THE SENSE OF SMELL AND CARDIOVASCULAR HEALTH IN OLDER ADULTS By Keran Wang Chamberlin A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of Epidemiology – Doctor of Philosophy 2024 ABSTRACT Background and objectives: Poor olfaction is common but underrecognized in older adults. This sensory deficit has broader health implications beyond being a prodromal symptom of neurodegeneration. Although biologically plausible, its cardiovascular health implications are unclear. Therefore, we aimed to investigate the associations of poor olfaction with incident stroke, coronary heart disease (CHD), and heart failure (HF), as well as subclinical cardiac biomarkers, by using two well-established community-dwelling cohorts of older adults in the US. Methods: In the Health Aging, and Body Composition (Health ABC) Study, we analysed data of 2,537 participants (aged 75.6±2.8 years) who completed a 12-item Brief-Smell Identification Test in 1999-2000. We defined good olfaction as a test score of 11-12, moderate olfaction as 9-10, and poor olfaction as ≤8. We followed at-risk participants from baseline until the date of the first cardiovascular outcome of interest, death, last contact, or the end of the 12-year follow-up, whichever occured first. We used the cause-specific Cox regression to estimate the associations of olfaction with incident stroke, CHD, and HF, respectively. Further, we leveraged data from the Atherosclerosis Risk in Community (ARIC) Study which was designed for cardiovascular health research to independently investigate the associations of olfaction with risks of stroke, CHD, and HF. Olfaction was assessed using the 12-item Sniffin’ Sticks odor identification test in 2011-2013 and defined categorically the same as in the Health ABC Study. We followed at-risk participants to the date of the first cardiovascular event of interest, death, last contact, or December 31, 2020, whichever came first. We used the discrete-time sub-distribution hazard model to estimate the marginal absolute risk of each outcome of interest across olfactory statuses and adjusted risk ratios (aRRs), accounting for competing risk of death and covariates. The cross-sectional associations of olfaction with subclinical HF markers were estimated using the quantile regression for N-terminal pro-B-type natriuretic peptides (NT-proBNP) and high-sensitive cardiac troponin T (hs-cTnT) and using the logistic regression for electrocardiography-defined structural heart disease. Results: In the Health ABC Study, we identified 353 incident CHD, 258 strokes, and 477 HF events during up to 12 years of follow-up. Poor olfaction was significantly associated with HF, but not with CHD or stroke. In the ARIC Study, among 5,799 participants who were free of stroke at baseline, we identified 332 incident stroke events (256 ischemic) during up to 9.6 years of follow- up. Compared with good olfaction, poor olfaction was robustly associated with higher stroke risk throughout the follow-up, albeit the association was modestly attenuated after 6 years. Among 5,142 participants free of CHD at baseline, we identified 280 incident CHD events during up to 9.6 years of follow-up. Poor olfaction was associated with a higher CHD risk during the first 6 years of follow-up, but not beyond. Among 5,217 participants without clinical HF at baseline, we identified 622 incident HF hospitalizations during up to 9.6 years of follow-up, including 212 HF with reduced ejection fraction (HFrEF), 250 HF with preserved EF, and 160 with unknown left ventricular EF. Compared with good olfaction, poor olfaction was associated with a modestly higher risk of HF for 8 years. The association was largely limited to HFrEF. Participants with poor olfaction had higher median levels of NT-proBNP and hs-cTnT, and higher odds of structural heart disease than those with good olfaction. Conclusions: Among community-based older adults in the US, we found preliminary evidence that poor olfaction assessed by a single smell test is associated with the risk of major adverse cardiovascular outcomes. The data from both cohorts are consistent for HF, supported by subclinical HF biomarkers. However, associations of olfaction with stroke and CHD were observed only in the ARIC Study. We suggest future studies be conducted to confirm our findings and investigate the underlying mechanisms. To my family ─ especially my parents ─ for believing in my potential and supporting me unconditionally iv ACKNOWLEDGEMENTS I am immensely grateful to my committee chair and primary mentor, Dr. Honglei Chen, for his invaluable guidance, support, and trust throughout the course of my PhD journey. I have had the opportunity to work with a variety of cohorts and study designs, gaining valuable experience in using diverse research methodologies. Dr. Chen has always encouraged me to explore and provided me with resources and opportunities to learn and pursue my research interests, reinforcing my commitment to the academic path. I would like to express my sincere gratitude to my committee member, Dr. Zhehui Luo, who introduced me to the field of causal inference and offered invaluable methodological insights. I want to express my heartfelt thanks to my committee member, Dr. Anna Kucharska-Newton, who has always been supportive and responsive, helping me advance my dissertation and understand the ARIC data. I am very grateful to have Dr. Mathew Reeves as my committee member. Engaging in discussions and reviewing his constructive feedback has greatly enhanced my understanding of cardiovascular disease. I am also deeply appreciative of all the friends and colleagues from Dr. Chen’s group. I enjoyed our weekly or biweekly group meetings to freely discuss research questions and methodologies. This collaborative and inclusive team has inspired my learning and productivity. I want to thank all the Department faculty and staff for creating a supportive, welcoming, and intellectually enriching environment for us. Last, I would like to thank my husband Adam and my family for their patience, understanding, and encouragement through this journey. Their faith in me and their constant support have been a source of motivation and determination. This dissertation would not have been possible without the collective support of all my mentors, friends, and families. I am truly grateful for their contribution. v TABLE OF CONTENTS LIST OF ABBREVIATIONS………………………………………………………………. viii CHAPTER 1: INTRODUCTION………………………………………………………………1 1.1 Dissertation Overview………………………………………………………………………1 1.2 Dissertation Organization……………………………………………………………………2 CHAPTER 2: BACKGROUND………………………………………………………………...3 2.1 Olfaction…………………………………………………………………………………….3 2.2 Major Adverse Cardiovascular Outcomes………………………………………………….10 2.3 Biological Plausibility of Olfaction with Cardiovascular Health..….……………………...15 CHAPTER 3: METHODOLOGY…………………………………………………………….18 3.1 Study Populations…………………………………………………………………………..18 3.2 Smell Testing……………………………………………………………………………...19 3.3 Outcomes…………………………………………………………………………………...20 3.4 Covariates…………………………………………………………………………………..21 3.5 Statistical Considerations and Analyses……………………………………………………22 3.6 Institutional Review Board Approval……………………………………………………...25 CHAPTER 4: OLFACTORY STATUS IN RELATION TO MAJOR ADVERSE CARDIOVASCULAR OUTCOMES IN THE HEALTH ABC STUDY……………….……27 4.1 Introduction………………………………………………………………………………..27 4.2 Methods……………………………………………………………………………………28 4.3 Results……………………………………………………………………………………..34 4.4 Discussion…………………………………………………………………………………40 CHAPTER 5: OLFACTORY STATUS IN RELATION TO STROKE IN THE ARIC STUDY…………………………………………………………………………………………..46 5.1 Introduction………………………………………………………………………………...46 5.2 Methods…………………………………………………………………………………….47 5.3 Results……………………………………………………………………………………...52 5.4 Discussion………………………………………………………………………………….59 CHAPTER 6: OLFACTORY STATUS IN RELATION TO CORONARY HEART DISEASE IN THE ARIC STUDY……………………………...…………………….….……..64 6.1 Introduction………………………………………………………………………………...64 6.2 Methods…………………………………………………………………………………….65 6.3 Results……………………………………………………………………………………...69 6.4 Discussion………………………………………………………………………………….74 CHAPTER 7: OLFACTORY STATUS IN RELATION TO HEART FAILURE IN THE ARIC STUDY…………………………………………………………………………………...79 7.1 Introduction………………………………………………………………………………...79 7.2 Methods…………………………………………………………………………………….80 vi 7.3 Results……………………………………………………………………………………...85 7.4 Discussion………………………………………………………………………………….92 CHAPTER 8: DISCUSSION…………………………………………………………………..97 8.1 Summary of Findings………………………………………………………………………97 8.2 Summary of Limitations…………………………………………………………………...98 8.3 Future Directions…………………………………………………………………………..99 8.4 Conclusions……………………………………………………………………………….104 BIBLIOGRAPHY……………………………………………………………………………..105 APPENDIX 1: LITERATURE REVIEW ON PREVIOUS INVESTIGATIONS …………128 APPENDIX 2: LIST OF REGRESSION MODELS IN THE PRESENCE OF COMPETING EVENTS.……………………………..……….……………………...…………….………….135 APPENDIX 3: SUPPLEMENTAL MATERIAL FOR CHAPTER 4……...……………….137 APPENDIX 4: SUPPLEMENTAL MATERIAL FOR CHAPTER 5……….…………..….143 APPENDIX 5: SUPPLEMENTAL MATERIAL FOR CHAPTER 6………………………149 APPENDIX 6: SUPPLEMENTAL MATERIAL FOR CHAPTER 7…..………..…………152 vii LIST OF ABBREVIATIONS 3MS: Modified Mini-Mental State examination APOE: Apolipoprotein E ARIC Study: The Atherosclerosis Risk in Community Study BMI: body mass index B-SIT: Brief-Smell Identification Test CHD: coronary heart disease CHF: congestive heart failure CIF: cumulative incidence function CI: confidence interval CVD: cardiovascular disease DAG: directed acyclic graph ECG: electrocardiogram eGFR: estimated glomerular filtration rate Health ABC Study: The Health Aging, and Body Composition Study HDL-C: high-density lipoprotein-cholesterol HF: heart failure HFmrEF: heart failure with mildly reduced ejection fraction HFpEF: heart failure with preserved ejection fraction HR: hazard ratio HFrEF: heart failure with reduced ejection fraction hs-cTnT: high-sensitive cardiac troponin T ICD-CM: International Classification of Disease Clinical Modification viii IPW: inverse probability weighting LV: left ventricular LVEF: left ventricular ejection fraction MI: myocardial infraction NT-proBNP: N-terminal pro-B-type natriuretic peptides OR: odds ratio RD: risk difference RR: risk ratio SS: Sniffin’ Sticks ix CHAPTER 1: INTRODUCTION 1.1 Dissertation Overview Poor olfaction is common but often goes unnoticed in older adults. This sensory deficit is one of the most important prodromal symptoms of neurodegenerative diseases. Interestingly, emerging evidence has shown that poor olfaction robustly predicts all-cause mortality in older adults, but only a small portion of excess deaths related to poor olfaction can be attributed to dementia or Parkinson’s disease, suggesting that poor olfaction may have profound health implications beyond neurodegeneration. Cardiovascular disease, a group of heterogeneous adverse health conditions, represents a substantial public health burden and ranks as the leading cause of death. Given the potential structural and functional connections between olfaction and the cardiovascular system, poor olfaction may signify future adverse cardiovascular outcomes. On the one hand, poor olfaction in late life may be a sensitive marker of impaired cardiovascular health, like the “canary in the coal mine”. On the other hand, poor olfaction may contribute to the deterioration of cardiovascular health. Either as an early marker or a contributor, poor olfaction may signify future adverse cardiovascular events in the older population. However, empirical evidence on whether olfaction signifies cardiovascular health in older adults so far has been sparse. Leveraging two well-established community-based cohorts of older adults in the US, the overall objective of this project was to evaluate the associations of olfactory status with incident stroke, coronary heart disease (CHD), and heart failure (HF), as well as with established subclinical cardiac biomarkers. This project will provide empirical evidence on the under-investigated links between olfaction and incident major adverse cardiovascular outcomes in the context of aging, filling a critical knowledge gap. Further, it will potentially identify a novel and easy-to-assess biomarker 1 to monitor impaired cardiovascular health in older adults, potentially promoting early prevention and reducing cardiovascular-related morbidity and mortality. In addition, this work may inspire new research areas to study cardiovascular health through this sensory loss, eventually deepening our understanding and advancing geriatric care in cardiovascular health. Given that both poor olfaction and cardiovascular outcomes are common among older adults, the findings may potentially have significant public health implications. 1.2 Dissertation Organization This dissertation has been organized into eight chapters. Chapter 1 provides an overview of this dissertation project. Chapter 2 describes the background of poor olfaction, the three types of major adverse cardiovascular outcomes (including stroke, CHD, and HF), and their potential biological connections. Chapter 3 presents the overall methodology. Chapter 4 focuses on the first publication1, regarding the association of poor olfaction with incident stroke, CHD, and congestive heart failure (CHF) in the Health Aging, and Body Composition (Health ABC) Study. Chapter 5- 7 each represents a separate manuscript, focusing on poor olfaction in relation to risk of incident stroke, coronary heart disease, and heart failure, respectively, using data from the Atherosclerosis Risk in Community (ARIC) Study. Chapter 8 summarizes this project’s overall findings, limitations, future directions, and conclusions. 2 CHAPTER 2: BACKGROUND Poor olfaction affects over a quarter of older adults and its prevalence rapidly increases with age2,3. This sensory loss is best known as an early symptom of neurodegenerative diseases4. Accumulating empirical data have shown that poor olfaction is robustly associated with higher all- cause mortality in older adults5. Major cardiovascular disease is the leading cause of mortality, morbidity, and disability in older adults6. Despite wide speculations of the connections between olfaction and cardiovascular health, there is limited empirical evidence regarding cardiovascular health implications of poor olfaction in older adults. In this chapter, we will introduce each of these health phenotypes and discuss the biological plausibility of their connections and existing evidence regarding olfaction and cardiovascular health in older adults. 2.1 Olfaction Olfaction, also known as sense of smell, is an old sense in evolution. Human being has the comparable olfactory neuron number to rodents7 and can distinguish around one trillion different odor combinations8. The sense of smell may play a crucial role in the human well-being, supported by an increasing body of literature. In this section, we will detail our current understanding of olfaction by structuring this sub-chapter into the following 5 parts: Olfactory system; Olfactory dysfunction; Assessment techniques; Epidemiology; and Causes and health implications. 2.1.1 Olfactory System The olfactory system has sophisticated structures to support odor detection, signal processing, and smell-related cognitive functions. Peripheral olfactory structures start from the back of the nasal cavity with odorant-binding mucus covering the olfactory epithelium9. The olfactory epithelium consists of olfactory receptor cells, sustentacular (supporting) cells, basal cells (multipotent stem cells), and duct cells of the Bowman’s glands10. Bundles of olfactory receptor axons constitute 3 Cranial Nerve I, projecting to the olfactory bulb located on the cribriform plate. The interneurons in the olfactory bulb further project to the anterior olfactory nucleus connecting to ventral tenia tecta, anterior hippocampal continuation, and indusium griseum11. Neurons in the pathway further rapidly projects to olfactory tubercle, piriform cortices, amygdaloid nuclei, and entorhinal cortex. The olfactory bulb is also indirectly linked to orbitofrontal cortex and other cortices via the olfactory-related feedback from entorhinal cortex12,13. The hippocampus, amygdala, and orbitofrontal cortex controls one’s memory, emotion, and personality & behavior respectively14– 16. As a result, olfactory function is anatomically and functionally related to higher-order brain functions. Other supportive systems, such as the circulatory system, are also crucial for normal olfactory function. The epithelium of the nasal cavity has rich capillaries that warm and humidify the incoming air while providing protection against various pathogens9. The blood supply of the olfactory epithelium and the olfactory bulb comes from the olfactory artery and the accessory olfactory artery17. The olfactory artery, which may have up to three terminal branches, originates directly from the anterior cerebral artery, a branch of the internal carotid artery. The accessory olfactory artery is also called the posterior ethmoidal artery, which converges with the anterior ethmoidal artery on the cribriform plate. All these arteries are the end vessels and do not anastomose with other vascular territories, thus these arteries’ narrowing and occlusion may lead to abnormality of olfactory function. The anterior and middle cerebral arteries supply blood to the orbitofrontal cortex and hippocampus18, while the anterior choroidal artery, branching from internal carotid artery, supplies blood to the amygdala19. The impaired blood perfusion in any structure along the olfactory pathway may lead to a decline or loss in olfactory function. 4 2.1.2 Olfactory Dysfunction Olfactory dysfunction can be defined using different criteria20. Based on whether the olfactory abnormality involves the strength or the quality of the odor, it can be classified as quantitative or qualitative olfactory dysfunction. Olfactory dysfunction can also be categorized according to its pathological location, for example, whether the abnormal function is attributed to blockage of an airway, or to the impairment of neuroepithelium or central neural loci. Although the potential causes of olfactory dysfunction are various and largely unknown, it is not uncommon to classify olfactory dysfunction according to the underlying etiology. Table 2.1 lists the detailed categories of olfactory dysfunction following different classification criteria. Definitions Table 2.1 Types of olfactory dysfunction based on different criteria Terminology Dysfunction type Hyposmia Anosmia Hyperosmia Parosmia Phantosmia Pathological location Quantitatively declined smell ability Quantitatively complete loss of smell Quantitatively increased smell ability Distorted perception of the odor Perceiving an odor in the absence of a stimulus Conductive dysfunction Sensorineural dysfunction Central dysfunction Etiology Olfactory dysfunction due to sinonasal disease Post-infectious olfactory dysfunction Posttraumatic olfactory dysfunction Olfactory dysfunction due to neurodegeneration Olfactory dysfunction related to aging Others Blockage of the airway to inhibiting the transmission of the odors Damage of olfactory epithelium or olfactory nerve Damage of the key processing central nervous regions Some sinonasal diseases, like chronic rhinosinusitis, trigger one or more underlying pathogenesis21. Pathogens, especially viruses (e.g., common cold, influenza, COVID-19), result in transient or prolonged smell dysfunction. Traumatic head injury may cause instant or delayed smell loss22. This is a major cause of permanent smell loss. Smell loss due to neurodegenerative pathologies in the peripheral and/or central olfactory system.4 Largely unknown, may be related to age-related physiological or pathological changes olfactory dysfunction due to toxins or medications; congenital olfactory dysfunction; idiopathic olfactory dysfunction 5 2.1.3 Assessment Techniques Olfactory assessments can be divided into 4 categories20. The first category is subjective assessment. While self-reported sense of smell is an important measure in determining the impact of the smell impairment in one’s daily life, people often do not recognize this sensory deficit2,23,24. The second type of assessment is the psychophysical olfactory measurement, most frequently used in large population and clinical settings. Psychophysical smell tests primarily assess three olfactory modalities, separately or combined25. The first is the odor threshold which measures the lowest concentration of an odor that can be detected. This smell ability is usually affected by conductive dysfunction. Odor discrimination refers to one’s nonverbal ability to distinguish different odors. Last, odor identification indicates one’s ability to detect and match odors to verbal or visual clues that describe the smell. The latter two olfactory modalities also require the normal functioning of the central olfactory structures26,27. These psychophysical tests all have the weakness that they cannot determine the location of pathology, therefore more sophisticated examinations are required. Imaging techniques provide ways to pinpoint the underlying pathologies. For example, magnetic resonance imaging can measure olfactory bulb volume and olfactory sulcus depth. However, advanced imaging techniques are expensive and require special equipment and expertise, thus it is not widely used beyond lab research settings25. Electrophysiological techniques can test cellular ionic currents and the ion channels, thus recording the sequential processing at the neuron level28. However, this technique has been limited in its use due to the cost and logistic issues29. 2.1.4 Epidemiology The epidemiology of olfactory dysfunction is primarily from studies using objective 6 psychophysical smell tests, because compared to smell identification test results, self-reported olfactory function has relatively low sensitivity (~20-30%)24. A recent Meta analysis reported that the pooled prevalence of olfactory dysfunction among populations aged from 18 to 97 years was 22%30. It was estimated that nearly 32 million (27.5%) of American adults aged 50 years and older were affected by olfactory dysfunction. While the prevalence of olfactory dysfunction is affected by study populations and smell test types and cut-offs, it has been consistently found to increase with age. For example, Murphy et al. used an 8-item San Diego Odor Identification Test and reported 6% of olfactory impairment among adults in their 50s, increasing to over 60% when adults were older than 802. While a few studies focused on Eastern Asia, most studies were conducted in the US and Europe30. Multiple studies have identified racial and gender difference in olfactory function among the US adult population with olfactory dysfunction being more prevalent among Black individuals compared to White individuals, and more common in males than females23,31,32. While longitudinal investigations are limited, the existing empirical data have consistently shown that the rate of olfactory decline increases with age3,33–36. For example, among adults aged 57-85 years from the National Social Life, Health, and Aging Project, Pinto et al. found that the 5-year decline in odor identification score was 0.25 score higher with every 10-year age increase3. Other demographics’ associations with the rate of olfactory decline were not consistent across studies. 2.1.5 Causes and Health Implications Olfactory dysfunction can be caused by infection. As the olfactory system is directly exposed to various pathogens, upper respiratory tract infections which lead to nasal local inflammation and swelling will block the airflow, disturbing olfactory function. Luckily, such olfactory dysfunction is in general temporary and will recover once the inflammation is relieved. Influenza-like infection 7 may also cause smell abnormality without concurrent stuffy nose37. Interestingly, the smell loss without stuffy nose has also been found prevalent among patients with SARS-CoV-2 infection38.This type of olfactory dysfunction may be related to the downregulation of odor detection pathways39. Despite the existence of long-term smell loss in COVID-19 patients, over 85% recovered their sense of smell within 2 months40. Sinonasal diseases, including chronic and acute rhinosinusitis, are also a primary cause of olfactory dysfunction20. The mechanisms of smell loss with sinonasal disease can be complex. It may be caused by the mechanical obstruction of odor transmission due to edema with or without nasal polyps, the inflammation-mediated odorant binding dysfunction, or neuroepithelium/ olfactory bulb remodeling20,41,42. Depending on the mechanisms involved, olfactory dysfunction can be transient or chronic, usually paralleling the progress of sinonasal diseases. As olfactory modalities, especially those involving high-order functions, rely on both peripheral and central neural structures, any damage throughout the neural pathways may also affect olfactory function43. For example, traumatic head injury may distort nasal structure, shear the olfactory fila, or even lead to brain hemorrhage, causing olfactory impairment. Head-trauma related olfactory dysfunction mostly recovers quickly within months, while in some rare cases, the olfactory dysfunction may last over 5 years44. Olfactory dysfunction is also a common symptom of neurodegeneration45,46. Importantly, this sensory deficit often occurs in the early stages of neurodegenerative progression. Braak et al. first proposed the staging of Alzheimer’s disease and Parkinson’s disease based on the neuropathology in post-mortem samples47,48. Specifically, this theory posits that Alzheimer’s disease and Parkinson’s disease initiates in the olfactory structures years before the overt cardinal symptoms and signs of neurodegeneration. It sheds light on opportunities to pinpoint high-risk 8 populations in the early stage of neurodegeneration and prevent the disease from continuously progressing to clinical manifestations49. Notably, poor olfaction identified by a single smell test has been found associated with a 2- to 3- fold higher risk of dementia50–52 and a 4- to 5- fold higher risk of Parkinson’s disease during up to a decade of follow-up53. Despite the specific causes of olfactory dysfunction covered above, most cases with smell loss have unknown causes. Olfactory dysfunction may be the consequence of long-term exposure to environmental hazards, the manifestation of biological aging, or a mixture of the two. Olfactory epithelium is an interface of connecting interior and exterior body environments, and thus constantly exposed to diverse environmental insults. As a result of being located at such an unprotected position, olfactory system has a remarkable regenerating ability to recover from countless environmental insults54. However, neurogenesis in the olfactory system may slow down or become exhausted due to prolonged exposure to environmental hazards and the natural aging process55. As the first line of defense against external hazards, the olfactory system may exhibit early abnormalities before other symptoms become apparent. Olfactory dysfunction has been increasingly found to have broader health implications beyond its links to neurodegenerative diseases5,56. Emerging evidence has found that impaired olfaction is a strong predictor of all-cause mortality5, supporting that olfactory loss may provide insights into survival beyond chronological age and existing comorbidities in older adults. Interestingly, using data from the Health ABC Study, Liu et al. found that only 22% of excess mortality associated with poor olfaction could be explained by dementia and Parkinson’s disease in older adults57. This longitudinal mediation study provided empirical evidence regarding the potential health implications of poor olfaction in older adults beyond what is currently known. However, evidence on other health implications of olfactory dysfunction remains limited. 9 2.2 Major Adverse Cardiovascular Outcomes Cardiovascular disease (CVD) is a group of heterogeneous disorders related to the heart and circulatory system which represent a substantial disease burden. Globally, major CVDs are the leading cause of mortality, with a combined age-standardized death rate of 196.1 per 100,000 in 20216. In the US, CVDs account for a quarter of deaths and affect over 28.6 million (10% of) adults aged 20 years or older in 202058,59. Based on pooled data from 7 US cohorts, the lifetime risk of developing CVDs at age 55 ranged from 15.3% to 38.6% for females and from 21.5% to 47.8% for males, depending on diabetic status60. Therefore, primary prevention of CVDs remains critical in public health. CVDs share some underlying mechanisms, such as atherosclerosis and inflammation61, and have some common risk factors, for example, hypertension, diabetes, obesity, and hyperlipidemia62–64. Despite these similarities, each major CVD has its own distinct pathological features and progression trajectories. For example, the hallmark of CHD pathophysiology is the development of atherosclerotic plaque in the coronary artery65. While CHD is one of the most common causes of HF, clinical HF represents an advanced stage with unrecoverable functional and/or structural heart anomaly due to any cardiac pathologies, such as valvular disease and cardiomyopathy66. Like CHD, stroke occurs primarily due to obstructed blood arteries, but its pathology happens in the cerebral arteries with more complicated etiology, adding complexity to stroke prevention67. Given the distinctions across major CVDs, it is hereby crucial to investigate each individual CVD. 2.2.1 Stroke Stroke, a type of cerebrovascular disease, can be classified into two categories: over 80% in the US are ischemic, while the remaining cases are hemorrhagic68,69. Ischemic stroke occurs due to 10 artery blockage, while hemorrhagic stroke is caused by bleeding from a ruptured blood vessel67. Ischemic stroke can be classified into different subtypes based on clinical features, brain imaging, and other imaging or laboratory assessments, according to the TOAST classification70. Stroke due to large artery atherosclerosis requires either greater than 50% stenosis or occlusion of a major intracranial or extracranial artery on vascular imaging with clinical symptoms of cerebral cortical impairment, brains stem or cerebellar dysfunction. This type stroke accounts for ~13% of ischemic strokes68. Cardio-embolism is brain arterial occlusions due to an embolus happening in the heart, so the diagnosis of cardioembolic stroke requires at least one cardiac source identified for an embolus70. Its clinical features and brain imaging may resemble those of large artery atherosclerosis, making differential diagnosis between the two subtypes critical. Cardioembolic stroke explains ~27% of ischemic strokes68. The third subtype is lacunar stroke mainly due to small vessel occlusion in the brain’s deep structures. Unlike the first two subtypes, this type of stroke is characterized by typical lacunar syndromes rather than cerebral cortical syndromes, along with imaging evidence that supports subcortical lesions smaller than 1.5 cm and rules out large artery and cardioembolic strokes70. Lacunar stroke accounts for 23% of ischemic strokes68. Less than 3% of ischemic strokes are those with other determined etiology, such as hematologic disorders, nonatherosclerotic vasculopathies, and hypercoagulable states67. The last category of ischemic stroke is cryptogenic stroke, which accounts for around 35% of ischemic strokes69. This subtype is non-lacunar stroke confirmed by imaging but without an identified cause67. The incidence of stroke has declined significantly over the years. From 1990 to 2019, worldwide incidence of stroke decreased by 17%71. The age-standardized incidence of stroke was estimated as 151 per 100,000 people in 2019. In the US, the ARIC Study has found a reduction in stroke incidence over the last three decades in males and females as well as in White and Black 11 individuals72. Nevertheless, stroke has still been associated with substantial disease burden, especially as populations age. It is the second leading cause of death and the third leading cause of death and disability combined across the world71. In 2019 alone, it caused 6.55 million deaths worldwide, accounting for 11.6% of total deaths. In the US, stroke ranks the fifth leading cause of death, accounting for 4.8% of total deaths58. Further, stroke is also closely related to subsequent cognitive decline and dementia. One study found that stroke brought forward the onset of dementia by 4 years in people who have had minor strokes or by 25 years in those who have had severe strokes73. Given the great disease burden related to stroke, it is imperative to identify the at-risk population early and prevent stroke events. 2.2.2 Coronary Heart Disease CHD has also been referred to as coronary artery disease and ischemic heart disease. While CHD often first presents as an acute event, its genesis requires chronic buildup of pathologies. Cascades of inflammatory reactions triggered by various risk factors are linked to the accumulation of atherosclerosis in the endothelium of coronary arteries65. As the plaque progresses, the artery may calcify and become stenotic or even occluded. As arterial remodeling leads to decline in the blood supply to the heart, it may cause chest pain, and other chronic symptoms of angina pectoris74. Without proper intervention of the progression, the rupture of plaques potentially provokes acute coronary thrombosis, leading to acute myocardial infarction (MI)65. The acute MI is often fatal and among survivors result in reduced heart function, further affecting the normal functioning of the cardiovascular system as well as potentially compromising other organs and systems. CHD has been once one of the most common fatal conditions since the 1930s75. In the US, the mortality of CHD continued to increase until the 1960s76. This rise is probably attributed to the upward trend of smoking, changes in dietary choices, increased sedentary behaviors, and the 12 increasing identification of CHD with the assistance of electrocardiography75. In 1978, the expert panel in the famous 1978 Conference on the Decline in CHD in Bethesda, US, acknowledged the decline in CHD mortality since mid-1960s76. While the causes of the decline have been debated, the decline was likely to be owing to the improvements in different levels of CHD prevention, including the decline in CHD incidence due to public health initiatives and the improved survival among CHD patients due to advancements in medical care77,78. Despite the decline in CHD mortality since the late 1960’s, CHD still ranks as the top cause of death worldwide and in the US6,79. CHD affects more than 20 million adults in the US, with its prevalence increasing with age and being higher in men than in women across all age groups58. Notably, it is estimated that an individual in the US experiences an MI every 40 seconds. Therefore, CHD is still an important public health issue, requiring comprehensive systems of care designed to treat acute coronary events as well as continued public health efforts to control risk factors such as smoking, hypertension, and diabetes. 2.2.3 Heart Failure HF is a complex heart syndrome resulting from any functional or structural impairment of ventricular filling or ejection of blood66. Given that the progression to symptomatic HF is gradual and chronic, the American College of Cardiology and American Heart Association have developed a staging system for HF to highlight the importance of stage-specific preventive and prognostic interventions66. The most severe stage, stage D, is also called the advanced HF stage. In this stage, even with the use of medical therapy, HF signs and symptoms still interfere with daily life and often result in recurrent hospitalizations. Stage C is symptomatic HF which requires current or previous HF manifestations. In stage C and D, HF management seeks to control symptoms and increase overall survival. In contrast, patients in stage B which is also called the pre-HF stage do 13 not have HF symptoms but show the presence of structural or functional changes in the heart that portend clinical disease. Specifically, these changes can be identified by cardiac structural changes, increased filling pressure, or elevated levels of cardiac biomarkers indicating myocardial stretch or injury. N-terminal pro-B-type natriuretic peptides (NT-proBNP) and high-sensitive cardiac troponin T (hs-cTnT) are well-established HF biomarkers and widely used in clinical practices to assist the prevention, diagnosis, and prognosis of HF80. Individuals in stage A are those at risk of developing HF but without symptoms, structural heart disease or abnormal cardiac biomarkers. People classified as stage A include those with atherosclerotic CVDs, hypertension, diabetes, metabolic syndrome, obesity, genetic susceptibility of cardiomyopathy, or exposure to cardiotoxic agents. Among adults age 67-91 years in the ARIC Study, over half of them had Stage A HF, followed by 30% with Stage B HF, 13% with clinical HF, and only 5% without any HF-related risk factors and abnormalities81. Left ventricular ejection fraction (LVEF), a measurement of left ventricular systolic function, is defined as the fraction of the blood volume ejected in systole over the blood volume in the ventricle at the end of diastole82. This measure is related to disease severity and prognosis66. Based on LVEF, patients with HF events can be classified into three groups: HF with reduced ejection fraction (HFrEF) defined as LVEF ≤40%, HF with preserved ejection fraction (HFpEF) defined as LVEF ≥50%, and HF with mildly reduced ejection fraction (HFmrEF) defined as 40%< LVEF <50%. HFmrEF may be more similar to HFrEF than HFpEF, as the former two HF types are more likely attributed to CHD83,84. HF affected more than 64 million people worldwide in 201785. While the incidence of HF has been stable at the level of 1-20 per 1,000 person-years over the last two decades, the prevalence of HF keeps rising owing to the aging population, better survival from CHD, and the elongated 14 life expectancy of HF patients86. The US is seeing an increase in HF from a prevalence of 2.4% in 2012 to an estimated 3% in 203058. The incidence of HF rises with age, reaching 6.0-7.9 per 1000 person-years after age 45 and approximately 21 per 1000 person-years among those over 65 years87. Survival rates of HF have improved over time thanks to evidence-based treatments for HFrEF, including pharmacotherapies, coronary revascularization, cardioverter defibrillators, and cardiac resynchronization therapies88. However, the economic burden related to HF is substantial. In the US, it is expected to have over 8 million HF patients by 2030 with an annual cost of $30,000 per patient89. Given the significant disease burden associated with HF, it is crucial to assess HF risk early during preclinical stages, to prevent the progression to clinical HF events, and to protect Stage A HF from developing in the first place. 2.3 Biological Plausibility of Olfaction with Cardiovascular Health There are several reasons why olfaction could have a biologically plausible relationship with the development of major cardiovascular outcomes in older adults. First, olfactory dysfunction may be an early marker of the compromised cardiovascular health before clinical symptoms show up. Olfactory identification involves high-order cognitive functions and thus requires intact structures and functions of the peripheral and central olfactory systems. As sufficient blood perfusion is critical to the normal functioning of the olfactory system, olfactory function may be sensitive to compromised cardiovascular health. For example, empirical evidence found that some subclinical carotid atherosclerotic biomarkers, such as carotid intima media thickness and the number of sites in carotid artery with plaques, were associated with olfactory loss90,91. Interestingly, the main arteries of blood supply to the olfactory epithelium, the olfactory bulb, and certain central loci derive from the internal carotid artery17. In addition, patients with idiopathic intracranial hypertension, an established risk factor for stroke92, were also found to have olfactory 15 dysfunction93,94. Therefore, olfactory dysfunction may be sensitive to disturbed blood supply and serve as an early unspecific symptom of compromised cardiovascular health. Further, olfactory dysfunction may contribute to impaired cardiovascular health in older adults by jeopardizing one’s eating behaviors. This sense assists our decision making about food. Retro-nasal olfaction interacting with sense of taste contributes to our perception of food and drinks95. Smell perception may also entangle with the state of metabolism and food choices, affecting our dietary behaviors and nutritional status96,97. While the role of sense of smell in nutritional status can be complex, limited empirical evidence suggests that olfactory dysfunction may be adversely associated with one’s appetite, dietary intake, and diet quality98–101. Since dietary patterns and calorie intake are crucial for maintaining cardiovascular health, poor olfaction may elevate the risk of cardiovascular disease morbidity by affecting nutritional intake102,103. Finally, olfaction may signify future cardiovascular health as a general marker of accelerated aging. Although the direct evidence is limited, empirical data has consistently found that poor olfaction is associated with faster decline in cognitive and physical function104,105 and higher risk of developing depressive symptoms106. These cognitive, physical, and mental downturns emerge with advanced age and are closely related to mortality and morbidity in older adults107,108. In support, accumulating empirical evidence has found that cognitive impairment, depression, and reduced physical function are associated with higher risk of future CVD109–111. Therefore, poor olfaction may be associated with incident CVD as a marker indicating accelerated aging. Frailty is a geriatric syndrome featured by a multi-dimensional systematic decrease in physiological reserve108. This syndrome is prevalent and associated with substantial mortality and disability among older adults112. A growing body of literature has recently shown the connections 16 between frailty and incident cardiovascular outcomes among older adults113,114. Frailty may elevate one’s vulnerability to internal or external insults in late adulthood or share the similar pathologies with adverse cardiovascular outcomes and accelerated aging115,116. Given the growing empirical evidence connecting poor olfaction with frailty in older adults117, research on frailty and cardiovascular health should consider exploring the relationship among poor olfaction, frailty, and CVD. However, empirical evidence regarding olfaction and cardiovascular health is limited. In detail, a few studies have reported the cross-sectional connections between olfactory status and cardiovascular disease in older adults2,35,118–124. Such snapshot investigations cannot elucidate the temporal order and have limited empirical implications. Several longitudinal studies mainly focused on the metabolic and cardiovascular origin of olfactory dysfunction35,36,125,126, which is different from our study goals. To our knowledge, only one longitudinal study investigated the association of olfactory function with incident heart disease in the National Social Life, Health, and Aging Project127. Specifically, Siegel et al. reported that five-year olfactory decline was marginally associated with higher odds of incident heart diseases (odds ratio [OR]: 1.75, 95% CI: 0.93-3.31). However, their diagnosis of heart diseases was self-reported only once in their year-10 survey and heart diseases were analyzed only as the secondary outcome of interest. Appendix 1 lists the detailed information on all the related population studies. This current project will focus on associations of olfaction with future major cardiovascular outcomes and overcome previous limitations by leveraging two independent well-established longitudinal cohorts with long-term follow-up, and detailed outcome surveillance and adjudications. 17 CHAPTER 3: METHODOLOGY We used two well-established cohorts of older adults in the US to investigate our aims. The Health ABC Study served as the preliminary investigation to examine the association of olfactory status with incident stroke, CHD, and CHF in older adults (Chapter 4). In the ARIC Study, we conducted a more detailed investigation of each major cardiovascular outcome of interest, including stroke (Chapter 5), CHD (Chapter 6), and HF (Chapter 7). This chapter focuses on the overall methodology and related methodological considerations. 3.1 Study Populations The Health ABC Study was established in 1997-1998, aiming to study the interrelationships across behavioral factors, age-related conditions, and comorbidities in the context of aging128,129. In brief, the study recruited 3,075 well-functioning older adults aged 70 to 79 years (48.8% men and 41.6% Black participants) in Memphis, Tennessee, and Pittsburgh, Pennsylvania. White participants were randomly sampled from Medicare beneficiaries and Black participants of eligible age were identified in specified zip code areas. The eligibility criteria included no difficulty walking a quarter mile or climbing up ten steps, no active life-threatening cancer in the last 3 years, and no plan to move outside the study areas in the next 3 years. The study conducted clinic visits annually since enrollment (Year 1) through Year 6, then in Year 8, 10, 11, and 16. Participants were contacted through phone calls semiannually until Year 15, and then quarterly through Year 17. Year-3 clinic visit in 1999-2000, including a smell test, was considered the baseline of our analysis. The ARIC Study, established in 1987-1989, was designed to investigate atherosclerosis and its cardiovascular sequelae130,131. Briefly, the ARIC Study recruited 15,792 community- dwelling adults aged 45-64 years selected from four communities (Forsyth County, North Carolina, Jackson, Mississippi, suburbs of Minneapolis, Minnesota, and Washington County, 18 Maryland). Specifically, age-eligible participants from each community were selected by probability random sampling based on a predefined list of households or individuals. Since enrollment, participants underwent periodic in-person clinical examinations and annual phone interviews (semiannually since 2012) to update their health status. The fifth clinical examination (Visit 5) in 2011-2013 including a smell testing was considered as our study baseline. Overall, the two studies included both males and females, and white and Black participants. They had comparable average age, similar study designs, and data collection strategies. However, they were entirely independent and had differences in eligibility criteria of enrollment, calendar periods of the follow-up, and original cohort objectives. 3.2 Smell Testing Both studies used a 12-item brief smell identification test. The Health ABC Study used the “scratch and sniff” Brief-Smell Identification Test (B-SIT) at Year-3 clinic visit132. The ARIC Study used the “felt-tip pen” Sniffin’ Sticks (SS) test at Visit 5133. Both tests are reliable (test-retest reliability: 0.73-0.78) and have been widely used in large population and clinical settings134–139. Both tests required participants to smell 12 common odors, one at a time, and select the right odorant from 4 possible choices in a forced multiple-choice format. One correct answer was given one score, so the test score ranged from 0 to 12. As the two tests in the two cohorts had a very similar score distribution, we defined good olfaction as a score of 11-12, moderate olfaction as 9-10, and poor olfaction as 8 or lower, corresponding to the tertile of the score distribution among study participants from either cohort. Using these cut-offs, previous studies have identified the associations of olfactory status with risks of Parkinson’s disease, dementia, and all-cause mortality52,53,57. 19 3.3 Outcomes Both studies closely monitored the health and survival of study participants via clinic visits, telephone calls, and cohort-wide surveillance of hospitalizations and deaths140–142. Major cardiovascular adverse events and deaths were identified through cohort-wise surveillance or annual/semi-annual follow-ups. However, the specific identification and adjudication procedures varied between the two cohorts. In the Health ABC Study, local adjudicators extracted and reviewed inpatient/outpatient medical records according to a standardized study protocol and a central expert committee adjudicated the cause of death for fatal events. In the ARIC Study, possible CVD events were first identified through International Classification of Disease (ICD) codes and keywords in the discharge summary and related medical records were extracted. The possible events of CHD and stroke were first classified by the computer-based algorithm and confirmed by a physician in the ARIC Morbidity and Mortality Classification Committee. The HF hospitalizations were independently adjudicated by physicians in the ARIC Study. Table 3.1 presents the definition of each major cardiovascular outcome in the two studies. Table 3.1 The definition of major cardiovascular outcomes in the two cohorts Health ABC Study ARIC Study CHD MI: evolving/diagnostic ECG pattern + abnormal cardiac enzymes; ischemic symptoms + [either an evolving ST-T pattern or an obscure ECG pattern] 143,144 Angina pectoris Death with CHD as the underlying cause MI: evolving/diagnostic ECG pattern + abnormal cardiac enzymes; ischemic symptoms + [either an evolving ST-T pattern or an obscure ECG pattern] + abnormal cardiac enzymes144,145 Death with CHD as the underlying cause 20 Table 3.1 (cont’d) Stroke Stroke (probable or possible): with evidence of sudden or rapid onset of neurological symptoms lasting for over 24 hours or leading to death in the absence of evidence for a non- stroke cause143,146 Death with stroke as the underlying cause HF CHF: the first overnight hospitalization with CHF as the primary inpatient reason or a concurrent event 143 Stroke (definite or probable): stroke is categorized into thrombotic and cardioembolic brain infarction, subarachnoid and intracerebral hemorrhage147. The detailed definition of each subtype refers to the ARIC Stroke Cohort Surveillance Procedures, Manual of Operations146 Death with stroke as the underlying cause HF: the first overnight hospitalization with HF. HF is categorized into ADHF, chronic stable heart failure and heart failure unlikely or unclassifiable148,149. HF is further categorized into HFrEF(EF<50%) and HFpEF (EF≥50%)150 Abbreviations: CHD: coronary heart disease; MI: myocardial infarction; ECG: electrocardiogram; CHF: congestive heart failure; HF: heart failure; ADHF: acute decompensated heart failure; HFrEF: heart failure with reduced ejection fraction; HFpEF: heart failure with preserved ejection fraction; EF: ejection fraction. Notably, the ARIC Study was among the first of several large cohorts specifically designed to study CVD etiology and risk factors, substantially contributing to our knowledge about cardiovascular health over the past three decades. Accordingly, compared to the Health ABC Study, the ARIC Study presumably had more stringent event identification and adjudication protocols, along with more detailed information on CVD events. Further details on each study outcome are provided in the following chapters. 3.4 Covariates We consider a range of covariates mostly collected at each study baseline. Although the covariate list varied between studies and across outcomes of interest, we primarily considered three types of covariates in our analyses. The first type of covariates were basic demographics, including age, sex, race, study site, and education. The second type of covariates were established risk factors for adverse cardiovascular outcomes, such as smoking, body mass index 21 (BMI), hypertension, diabetes, blood cholesterol, and other prevalent major cardiovascular outcomes. The third type included potential predictors for adverse cardiovascular outcomes, for example, renal function and frailty. The lists and definitions of covariates are described in the Method of the following chapters (Chapter 4-7). 3.5 Statistical Considerations and Analyses In this project, our target population is older adults with an average baseline age of 75.5 years at risk of developing stroke, coronary heart disease, or heart failure in the US. Therefore, taking the issue of competing risk of death into statistical consideration is crucial. In the descriptive analysis, instead of using the Kaplan-Meier curve, we used the cumulative incidence function151,152, as the Kaplan-Meier survival curve 𝑆(𝑡) = Pr⁡(T > t) (where Pr(T > t) denotes the distribution of event times) assumes that the event of interest would occur for all subjects, which is impossible in the presence of the competing event of death. As a result, using the Kaplan-Meier curve will overestimate the incidence. In contrast, the cumulative incidence function (CIF), defined as ⁡𝐶𝐼𝐹𝑘(𝑡) = Pr(𝑇 ≤ 𝑡, 𝐷 = 𝑘) (where D denotes the type of event that occurred), will not necessarily approach unity as time becomes large, because this estimator considers that the occurrence of the competing event will preclude the occurrence of the event of interest. The non- parametric maximum likelihood estimator of the CIF of cause k is 𝐹𝑘(𝑡) = ∑ 𝑇𝑙≤𝑡 𝑑𝑘𝑙 𝑌𝑙 𝑆(𝑡𝑙−1) , Where 𝑘 ≥ 2 is the type of event, 𝑡1 < 𝑡2 < 𝑡3 … < 𝑡𝑙 are the distinct uncensored times, 𝑌𝑙 is the number of subjects at risk at 𝑡𝑙, 𝑑𝑘𝑙 is the number of events that occurred at 𝑡𝑙, 𝑆(𝑡𝑙) is the Kaplan- Meier estimator that would have been obtained by assuming that all failure causes are of the same type. In the presence of competing events of death, Figure 2.1 shows the conceptual relationship 22 among poor olfaction, major cardiovascular outcomes, and deaths before cardiovascular outcomes in the form of the directed acyclic graph (DAG). In the presence of competing risk of death, the association of olfaction with cardiovascular outcomes arises through two pathways: the first pathway is the direct association between olfaction and cardiovascular outcomes (Path 1 in Figure 2.1), and the other pathway is the indirect association through competing event of death (Path 2 in Figure 2.1). Notably, in our case, the existence of an indirect association pathway would attenuate the total association between olfaction and cardiovascular outcomes (i.e. Path 1 + Path 2), because poor olfaction is strongly associated with higher mortality and death is an absorbing status (i.e., nullifying the “risk” for the cardiovascular outcomes). Therefore, it is important to articulate which association is estimated in the presence of competing risk of death. Figure 2.1 Partial* directed acyclic graph of differential survival during the follow-up. * Other variables are omitted to avoid clutter In the survival analysis, hazard and risk are commonly used to quantify the association. Hazard is defined as the instantaneous rate of the event of interest among the at-risk population. Risk is the cumulative risk of the event among the at-risk population during a fixed equal exposure period. In other words, hazard is a velocity measure of event occurrence while risk is a cumulative measure over time. Hazard can be used to calculate the cumulative risk. In the presence of competing risk of death, there are two types of hazards151,153. One type is cause-specific hazard defined as 𝑐𝑠(𝑡) = lim 𝜆𝑘 ∆𝑡→0 𝑃𝑟𝑜𝑏(𝑡≤𝑇<𝑡+∆𝑡,𝐷=𝑘|𝑇≥𝑡) , ∆𝑡 23 where T is the time from baseline until the occurrence of the event of interest, D is the type of event of interest. It represents the instantaneous risk of having kth event among participants who do not yet have any types of events. The cause-specific hazard ratio (HR) can be directly estimated from the cause-specific Cox regression and estimates the direct association in the scale of HR. However, HR is less preferable than risk ratio (RR). First, HR is not an effect measure. Because hazard is conditional on individuals who have not had the outcome or competing events, and thus it is impossible to compare the hazard of the outcome of interest among the “same” individuals with different exposure levels154. Further, there are criticisms that this method considers death events the same as loss to follow-up, even though loss to follow-up is fundamentally different from death events. Censoring due to loss to follow-up is possible to be avoided in a study by implementing more flexible data collection strategies and improving participants’ awareness of the project; however, deaths are impossible to be eliminated, especially in an older population. As such, in the presence of competing risk of death, another way is to estimate absolute risk using the Fine-Gray sub-distribution hazard153. Sub-distribution hazard is defined as 𝑠𝑑(𝑡) = lim 𝜆𝑘 ∆𝑡→0 𝑃𝑟𝑜𝑏(𝑡≤𝑇<𝑡+∆𝑡,𝐷=𝑘|𝑇≥𝑡∪(𝑇<𝑡∩𝐾≠𝑘)) ∆𝑡 , where T is the time from baseline until the occurrence of the event of interest, D is the type of event of interest. In other words, it refers to the instantaneous risk from the kth event in participants not yet having the event of k. This hazard measure can be used to predict the cumulative risk of the kth event in the presence of competing risk of death. The RR and risk difference (RD) can be calculated subsequently. As the calculated absolute risk accounts for the competing risk of death, the risk- related association measures quantify the total association155. While this association measure is 24 affected by both direct and indirect pathways, it has a causal interpretation and is thus preferred in medical studies. Further, with more assumptions, it is also possible to estimate the direct association in the scale of risk-based association measures (i.e., RR and RD)155. Appendix 2 lists the commonly used regression models in the presence of competing risk of death. The Cox proportional hazards model has been widely used in survival analysis. However, the semiparametric nature of Cox regression requires the proportional hazard assumption during the whole follow-up, and it cannot correct the selection bias when loss to follow-up does not occur completely at random. To overcome these limitations, investigators from the Framingham Study first proposed to use the pooled logistic regression which showed decent performance when compared to the Cox proportional hazards model156. Because this modeling can easily incorporate time-varying covariates, time-varying coefficients, and inverse probability weighting (IPW), it has been increasingly used in the causal inference field157. Under the framework proposed by Young et al.155, we used pooled logistic regression to estimate sub-distribution hazard in the discrete-time scale, calculating marginalized absolute risk across olfactory statuses and estimating the total association in the scale of the RR. In the Health ABC Study, we used cause-specific Cox regression to estimate the direct associations of poor olfaction with incident major cardiovascular outcomes (Chapter 4). In the ARIC Study, we estimated both the total association in the RR scale and the direct association in the scale of the cause-specific HR to demonstrate the potential influence of competing risk of death in our results (Chapter 5-7). Methodological details are presented in the corresponding chapters. 3.6 Institutional Review Board Approval The work conducted for this dissertation was reviewed and approved by the Michigan State University Institutional Review Board (STUDY00009824). To obtain the data from the Health 25 ABC Study, investigators should submit an analytical proposal online at https://healthabc.nia.nih.gov/ancillary-biospecimen-proposals, which will be reviewed and approved by the Health ABC Study. To access the data from the ARIC Study, investigators should submit an analytical proposal which will be reviewed and approved by the ARIC Study. For both studies, Data Use Agreements need to be developed and signed. 26 CHAPTER 4: OLFACTORY STATUS IN RELATION TO MAJOR ADVERSE CARDIO- VASCULAR OUTCOMES IN THE HEALTH ABC STUDY This study has been already published in the Journal of the American Heart Association1. 4.1 Introduction The human sense of smell declines with age in older adults. Prevalence of poor olfaction, assessed by smell-identification screening, quickly increases from ~6% in age 50s to over 60% in 80s2. This age-dependent olfactory decline has been confirmed in multiple community-based studies3,23. Despite the high prevalence of poor olfaction in older adults, our understanding of its health implications has been largely limited to its role as a prodromal symptom of neurodegeneration and its robust association with mortality20. Interestingly, our recent findings57 indicated that only 22% of the excess mortality associated with poor olfaction could be explained by dementia and Parkinson’s disease, suggesting that poor olfaction may have more profound health implications than what is known to date. Further, this association with higher mortality was limited to participants who self-reported good-to-excellent health at baseline57, raising the possibility that poor olfaction may be a marker of deteriorating health that precedes the emergence of more traditionally recognized signs and symptoms of health decline. Beyond neurodegenerative diseases and mortality, the health implications of poor olfaction have been subject to wide speculation, with limited empirical evidence. Recent data suggest that poor olfaction is associated with carotid intima-media thickness and artery plaques90,91, suggesting that smell loss may be a marker of atherosclerosis – the underlying pathogenesis of cardiovascular disease. Further, poor olfaction may gradually degrade one’s food choices, adversely affecting dietary quality and nutrition98,100, which may contribute to cardiovascular disease over time. Therefore, as a nonspecific subclinical marker and/or a potential contributor, poor olfaction may 27 be related to cardiovascular risk. Because poor olfaction is prevalent among older adults and cardiovascular disease is the leading cause of death and disability, this potential association should be investigated further. To the best of our knowledge, only one study127 has prospectively examined the association between olfaction and heart diseases, and reported an elevated but statistically non-significant association with heart attack and/or heart disease. We hereby comprehensively examined olfactory status in relation to the risk of three major adverse cardiovascular conditions − CHD, stroke, and CHF among community-dwelling older adult participants in the Health ABC Study. 4.2 Methods 4.2.1 Study population The Health ABC Study aims to investigate the interrelationships among aging-related conditions, social and behavioral factors, and physiological and functional changes in older adults128. Briefly, in 1997 and 1998, this study recruited 3,075 well-functioning, community- dwelling older adults (51.5% women and 41.7% blacks) aged 70-79 years in the designated zip code areas surrounding Pittsburgh, Pennsylvania, and Memphis, Tennessee. Eligibility criteria included no reported difficulty in walking 1/3 mile or climbing up 10 steps, no active fatal cancers, and no plans to move in 3 years. Study participants were followed with annual clinic visits through Year 6, and then in Year 8, 10, 11, and 16. Phone interviews were conducted to update health status every 6 months until Year 15 and then quarterly through Year 17. In the current analysis, we used the Year-3 clinic visit (1999-2000) as the baseline which was when the olfaction test was conducted. The primary analysis was limited to 2,537 participants after excluding those who missed Year-3 clinic visit (n=154) and did not take the smell test (n=384). In the analysis of each cardiovascular outcome, we excluded prevalent cases of that outcome at baseline, respectively. As 28 case adjudications for major health outcomes (except death) were conducted through August 14, 2012, we followed at-risk participants from baseline until the first cardiovascular outcome, death, last contact (attrition rate < 2%), or the end of the 12-year follow-up, whichever came first. The Health ABC Study protocol was approved by all relevant institutional review boards, and all participants provided written informed consent at enrollment. 4.2.2 The Brief-Smell Identification Test Olfaction was tested at the Year-3 clinic visit, using the 12-item cross-cultural B-SIT. This test is a shortened version of the 40-item University of Pennsylvania Smell Identification Test and has been widely used in large populations132. The 12-item test is brief, convenient, and well-suited to field settings in large epidemiological studies and quick clinical screening139,158. Participants were instructed to smell each of the 12 odorants, one at a time, and then to identify the odorant from 4 possible answers in a forced multiple-choice format. One point was given for each correct answer with a total score ranging from 0 to 12. We defined poor olfaction as a B-SIT score ≤8, moderate as 9-10, and good as 11-12, approximately corresponding to the tertile distribution of the B-SIT testing score in the study population. Using these cut-points, we have reported strong associations of poor olfaction with Parkinson’s disease, dementia, total mortality, and pneumonia hospitalization in this cohort53,52,105,159. 4.2.3 Major adverse cardiovascular outcomes The Health ABC Study closely monitored the health and survival of study participants via study clinic visits, semiannual phone updates, and surveillance of hospitalization and death. As detailed previously 160–163, major adverse cardiovascular outcomes were first identified via the cohort’s routine follow-ups and health surveillance and then adjudicated according to a standard protocol. Briefly, at each clinic visit and semi-annual telephone interview, participants or their proxies were 29 asked directed questions about cardiovascular disease events diagnosed by a physician, overnight hospitalizations, and outpatient cardiovascular procedures such as angioplasty since the last interview. Once an event was reported, local medical event adjudicators collected and reviewed related medical records according to a standardized study protocol143. For each death event, study investigators had an exit interview with a knowledgeable proxy who provided detailed information on the death event and the participant's physical functioning before death. The immediate and underlying causes were adjudicated centrally by an expert committee after reviewing hospital records, death certificates, autopsy findings, and informant interviews. In this study, we defined incident CHD as the first event of myocardial infarction, angina pectoris, or death with CHD as the underlying cause. According to the protocol143, MI adjudication accounted for evolving diagnostic ECG pattern; diagnostic ECG pattern and abnormal cardiac enzymes; or ischemic symptoms and either an evolving ST-T pattern or an obscure ECG pattern. The adjudication of angina pectoris considered symptoms such as chest pain, chest tightness, shortness of breath, and a diagnosis from a physician, as well as medical treatment including nitroglycerin, beta-blocker, or calcium channel blocker. We defined stroke as the first event of stroke or death with cerebrovascular diseases as the underlying cause, considering evidence of a rapid onset of neurologic deficit attributed to obstruction or rupture of the arterial system and new CT/MRI lesion consistent with clinical presentation of stroke without evidence of alternative causes (e.g., tumor or infection). We defined CHF as the first admission of overnight hospitalization with CHF adjudicated as the primary inpatient reason or a concurrent event. The adjudication considered physician diagnosis, and medical treatments for CHF including both a diuretic and digitalis or a vasodilator, or the presence of cardiomegaly and pulmonary edema on chest X-ray, or evidence of a dilated ventricle and global/ segmental wall motion abnormalities 30 with deceased systolic function either by ECG or contrast ventriculography. 4.2.4 Covariates As few risk factors have been established for olfactory loss in older adults except for age, sex, and race, we mainly considered cardiovascular risk factors/predictors as covariates in the analyses. The adjustment of these covariates may help control for potential confounding and improve statistical efficiency164. With a few exceptions, we used covariate data from the Year-3 clinic visit when the smell testing was conducted. Age, sex, race, study site, education level, smoking status, minutes of brisk walking per week, and general health status were self-reported. BMI was calculated by dividing weight by height-squared (kg/m2) and systolic blood pressure by averaging two measures in the sitting position. The use of antihypertensive medication was assessed using the medication inventory method coded with the Iowa Drug Information System Drug Vocabulary and Thesaurus165. We defined comorbidities according to published protocols, in brief, 1) diabetes as self-reported diagnosis by a physician, the use of anti-diabetic drugs, a fasting blood glucose level of ≥126 mg/dL, or an oral glucose tolerance test of ≥200mg/dL166; 2) dementia as the score of the Modified Mini-Mental State examination (3MS) at the Year-1 clinic visit less than 80, a decline in 3MS score from Year 1 through Year 3 at least 1.5 race-stratified standard deviations, an adjudicated diagnosis of dementia based on hospitalization, or documented medication uses for dementia57; 3) Parkinson’s disease as adjudicated by two movement disorder specialists by consensus after review of self-reported diagnosis by a physician, medication uses, hospitalization records, and adjudicated cause of death53. Depressive symptoms were defined as a score of ≥10 on the Center for Epidemiologic Studies Depression Scale Short form167. When covariate data are not available for the Year-3 clinic visit, we used data from previous years. Resting heart rate was measured at Year 1. Left ventricular hypertrophy was diagnosed using Year-1 ECG according to 31 the Minnesota code criteria161. Abnormal lung function was defined as the forced expiratory volume in the 1st second measured at Year 1 below the lower limit of the age-, sex- and race- specific normalized reference values of the National Health and Nutrition Examination Survey Ⅲ equations168. Plasma total cholesterol and high-density lipoprotein-cholesterol (HDL-C) were measured using fasting EDTA plasma collected at Year 2 and Year 1, respectively169. Serum albumin was measured using samples collected at Year 1170, interleukin 6 using samples collected at Year 2171, and cystatin C and creatinine using samples collected at Year 3172. All these biomarkers have been widely analyzed in the Health ABC Study with details reported previously. We estimated eGFR mainly using the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) creatinine-cystatin C equation; for 9.6% of the sample with missing creatinine data, we estimated eGFR using the CKD-EPI cystatin C equation173. Among those with creatinine measures, these two eGFR estimates were highly correlated with a Spearman coefficient of 0.82. As the proportions of missingness in other covariates were <5%, we used simple imputation by the mode for discrete variables and the median for continuous variables. 4.2.5 Statistical analysis In descriptive analyses, we used linear regressions for continuous variables and logistic or multinomial regressions for categorical variables to estimate their age-adjusted marginal means/percentages in each olfaction group. We then calculated the CIF of each type of cardiovascular outcome and its corresponding competing risk of death, and tested the equality of CIF across baseline olfactory status using the Gray’s test174. In multivariable analyses, we used the Cox cause-specific hazard model with the robust sandwich standard error estimate175 to account for the competing risk of death and reported cause-specific HR and 95% CI for each type of cardiovascular event. This approach quantifies the direct association between olfactory status and 32 each outcome of interest, not affected by the association of olfaction with death, fitting our analytical goal153. In the analyses, we first controlled for age, sex, race, education, and study site (model 1), and then further adjusted for key lifestyle cardiovascular risk factors in model 263,110, including smoking status, brisk walking, BMI, self-reported general health status, antihypertensive medication use, diabetes, depressive symptoms, systolic blood pressure, total cholesterol, and HDL-C. As prevalent atherosclerotic diseases could be an important risk factor for CHF176, we also included prevalent CHD or stroke as a covariate in model 2 of the CHF analysis. Finally, we constructed model 3 for CHF by further adjusting for previously identified markers of CHF in the cohort161–163, including left ventricular hypertrophy, abnormal lung function, heart rate, serum albumin, interleukin 6, and eGFR. In all regression analyses, we applied the Supremum Test to check the proportional hazard assumption and, when applicable, stratified the covariates that did not satisfy the assumption in the regression model177. Finally, given the strong association of olfaction with dementia and Parkinson’s disease, we conducted a sensitivity analysis by excluding participants with prevalent dementia or Parkinson’s disease at baseline. For outcomes that showed a significant association with olfaction in the primary analysis, we conducted secondary subgroup analyses by age, sex, race, self-reported health status, and history of other major cardiovascular diseases at baseline. These analyses were pre-planned because the prevalence of poor olfaction is age-dependent and substantially higher in men than in women and in blacks than in whites, and our prior analysis showed that the association of poor olfaction with higher mortality was limited to people with self-reported good-to-excellent health at baseline57. Interestingly, for CHF, we found that the association was evident mostly among individuals who self-reported very-good-to-excellent health. We therefore conducted two post hoc exploratory analyses in this subgroup. First, we reestimated the full model in this subgroup. Next, 33 we further modeled the B-SIT score on a continuous scale, the non-linearity form of which was regressed by using the quadratic term. We used the SAS software (version 9.4; SAS Institute, Cary, NC) for all the analyses with a two-sided α of 0.05. 4.3 Results At baseline, participants were, on average, 75.6±2.8 years old, with 51.6% female and 38.5% Black. In the overall study sample, compared with participants with good olfaction, those with poor olfaction were more likely to be older, men, Black, smokers, and from Memphis (Table 4.1). They were also more likely to report a high-school education level or less and fair-to-poor general health status, and to have diabetes, abnormal or missing lung function, lower BMI, TC, HDL-C, and eGFR. As age is the most important risk factor for olfactory loss in older adults, we also presented age-adjusted covariates by olfaction in Table A3.1. Once age was adjusted, the imbalances of prevalent diabetes, lung function, BMI, and cholesterol level across olfaction groups disappeared. Variable Table 4.1 Population characteristics by baseline olfactory status (n=2,537) a Olfactory status Moderate (n = 867) 75.0 (5.0) 370 (42.7) 497 (57.3) 418 (48.2) 331 (38.2) Good (n = 845) 75.0 (4.0) 391 (46.3) 454 (53.7) 324 (38.3) 265 (31.4) Poor (n = 825) 76.0 (4.0) 279 (33.8) 546 (66.2) 485 (58.8) 380 (46.1) 428 (51.9) 397 (48.1) 510 (61.8) 315 (38.2) 313 (37.9) 338 (41.0) 174 (21.1) 432 (49.8) 435 (50.2) 490 (56.5) 377 (43.5) 270 (31.1) 363 (41.9) 234 (27.0) 377 (44.6) 468 (55.4) 414 (49.0) 431 (51.0) 273 (32.3) 366 (43.3) 206 (24.4) 34 Age in years, median (IQR) <75 years ≥75 years Male sex, n (%) Black race, n (%) Study site, n (%) Memphis Pittsburgh Education, n (%)b ≤high school >high school Body mass index, n (%)b <25 kg/m2 25-30 kg/m2 >30 kg/m2 Smoking status, n (%)b Table 4.1 (cont’d) Never Former & <30 pack-years Current or ≥30 pack-years Brisk walking, n (%) b <90 min/wk ≥90 min/wk General health status, n (%)b Very good to excellent Good Fair to poor Systolic blood pressure in mmHg, median (IQR) Antihypertensive drug use, n (%)b Diabetes, n (%) Depressive symptoms, n (%)b Heart rate in beats per minute, median (IQR) b LVH, n (%) Abnormal lung function, n (%) No Yes Missing Total cholesterol in mg/dL, median (IQR) b HDL-C in mg/dL, median (IQR) b Albumin in g/dL, median (IQR) b Interleukin 6 in pg/mL, median (IQR) b eGFR in mL/min/1.73m2, median (IQR) b Prevalent major cardiovascular diseases Prevalent CHD, n (%) Prevalent stroke, n (%) Prevalent CHF, n (%) 429 (50.8) 235 (27.8) 181 (21.4) 744 (88.0) 101 (12.0) 423 (50.1) 292 (34.6) 130 (15.4) 134 (26) 497 (58.8) 181 (21.4) 86 (10.2) 63 (14) 390 (45.0) 217 (25.0) 260 (30.0) 786 (90.7) 81 (9.3) 384 (44.3) 351 (40.5) 132 (15.2) 134 (24) 533 (61.5) 214 (24.7) 108 (12.5) 63 (14) 351 (42.5) 215 (26.1) 259 (31.4) 754 (91.4) 71 (8.6) 325 (39.4) 312 (37.8) 188 (22.8) 134 (28) 483 (58.5) 220 (26.7) 115 (13.9) 65 (16) 98 (11.6) 97 (11.2) 96 (11.6) 689 (81.5) 75 (8.9) 81 (9.6) 206.0 (51.0) 52.0 (19.0) 4.0 (0.4) 2.27 (2.15) 682 (78.7) 104 (12.0) 81 (9.3) 204.0 (49.0) 51.0 (21.0) 4.0 (0.4) 2.33 (2.49) 610 (73.9) 104 (12.6) 111 (13.5) 202.0 (51.0) 51.0 (20.0) 4.0 (0.5) 2.33 (2.24) 81.1 (24.3) 81.3 (24.5) 77.3 (26.6) 199 (23.6) 69 (8.2) 37 (4.4) 207 (23.9) 73 (8.4) 44 (5.1) 207 (25.1) 65 (7.9) 35 (4.2) 35 Table 4.1 (cont’d) Abbreviations: IQR: inter-quartile range; HDL-C: high-density lipoprotein-cholesterol; LVH: left ventricular hypertrophy; eGFR: estimated glomerular filtration rate; CHD: coronary heart diseases; CHF: congestive heart failure. a Prevalent case of outcomes of interest are included in this table. Please see Supplementary Materials for tables for each outcome of interest without corresponding prevalent cases. b Missing values (<5%) were singly imputed. Specifically, the numbers of missningness in covariates are as follows: eudaction: n=7 (0.28%), body mass index: n=2 (0.08%), smoking status: n=33 (1.3%), brisk walking: n=1 (0.04%), general health status: n=3 (0.12%), antihypertensive medication: n=1(0.04%), depressive symptoms: n=2 (0.08%), heart rate: n=1 (0.04%), total cholesterol: n=13 (0.51%), HDL-C: n=85 (3.35%), albumin: n=24 (0.95%), eGFR: n=71 (2.8%), and interleukin 6: n=104 (4.10%). After excluding prevalent cases at baseline, 1,924 participants were at risk for incident CHD, 2,330 for stroke, and 2,421 for CHF. During 12 years of follow-up, 353 individuals (18.3%) had an incident CHD event, 258 (11.1%) experienced an incident stroke, and 477 (19.7%) had an incident CHF hospitalization. In the descriptive analysis (Figure 4.1), baseline olfactory status was not statistically significantly associated with the cumulative incidence of CHD or stroke using the Gray’s test. However, this test showed a statistically significant difference for the cumulative incidence of CHF across olfaction groups. In all analyses, poor olfaction was associated with a higher competing risk of death, consistent with our previous findings on the association between olfaction and all-cause mortality using the same data source57. 36 Figure 4.1 Cumulative incidence function by baseline olfactory status (good, moderate, poor) of a) coronary heart diseases (CHD) and its competing event of death (n=1,924); b) stroke and its competing event of death (n=2,330); c) congestive heart failure (CHF) and its competing event of death (n=2,421) Multivariable models confirmed the unadjusted findings (Table 4.2). After adjusting for demographics, compared to participants with good olfaction, the cause-specific HR of CHF during a median 10.8 years of follow-up was 1.35 (95% CI: 1.08,1.70) for those with moderate olfaction and 1.39 (95% CI: 1.10, 1.75) for those with poor olfaction. The associations were barely changed with further adjustment for lifestyle risk factors and prevalent CHD/stroke, and were only modestly attenuated after further adjusting for ECG-based, spirometry-based, and blood-based biomarkers for CHF. In the fully adjusted model, the HR became 1.32 (95% CI: 1.05, 1.66) for moderate vs. good olfaction and 1.28 (95%CI: 1.01, 1.64) for poor vs. good olfaction. As in the descriptive analyses, neither CHD nor stroke outcome was statistically significantly associated with baseline olfactory status. For example, the cause-specific HR comparing poor with good olfaction were 0.97 (95% CI: 0.73, 1.28) for CHD and 1.12 (95 % CI: 0.82, 1.52) for stroke. After 37 removing prevalent cases of dementia or PD at baseline, the results were consistent with our primary findings (Table A3.2). Table 4.2 The association of baseline olfactory status with incident coronary heart diseases (CHD), stroke, and congestive heart failure (CHF) for up to 12 years of follow-up a Olfactory function No. of Event Person -years Incidence (per 1,000 person- year) Model 1 b Model 2 c Model 3 d HR (95% CI) P HR (95% CI) P HR (95% CI) P CHD (n=1,924) Good 119 127 6124 5939 19.4 21.4 Moderate Poor 107 5018 21.3 Stroke (n=2,330) Good 88 78 7665 7514 11.5 10.4 Moderate Poor 92 6428 14.3 CHF (n=2,421) Good 130 180 7792 7533 16.7 23.9 Moderate Poor 167 6561 25.5 Reference 1.06 (0.83,1.37) 1.01 (0.77,1.33) 0.636 0.920 Reference 1.01 (0.78,1.31) 0.97 (0.73,1.28) 0.937 0.815 Reference 0.86 (0.63,1.17) 1.13 (0.84,1.53) Reference e 0.334 0.429 0.85 (0.62,1.16) 1.12 (0.82,1.52) 0.298 0.476 Reference 1.35 (1.08,1.70) 1.39 (1.10,1.75) 0.009 0.006 Reference 1.31 (1.05,1.65) 1.37 (1.08,1.74) Reference f 0.019 0.010 1.32 (1.05,1.66) 1.28 (1.01,1.64) 0.017 0.043 Abbreviations: HR: hazard ratio; 95% CI: 95% confidence interval a Associations were estimated from Cox cause-specific models with the robust sandwich standard error estimate to account for the competing risk of death. b Model 1 included age, sex, race, education, and study site as covariates. c Model 2 further included smoking status, brisk walking, body mass index, self-reported general health status, systolic blood pressure, use of antihypertensive medications, diabetes, depressive symptoms, total cholesterol, and high-density lipoprotein-cholesterol as covariates. For CHF, Model 2 further included prevalent CHD/stroke in addition to the above covariates. d Model 3 (only for CHF) further included heart rate, left ventricular hypertrophy, abnormal lung function, albumin, interleukin 6, and estimated glomerular filtration rate. e Brisk walking and antihypertensive medication use were stratified in the Cox model. f Tertile of interleukin 6 was stratified in the Cox model. The associations of olfaction with CHF were robust across subgroups of age, sex, race, and baseline history of CHD and stroke (Figure 4.2). Although we did not observe a statistically significant interaction, the association between olfaction and CHF appears to be more evident and 38 monotonic among participants with very-good-to-excellent health at baseline. In contrast, the estimated associations were close to null among participants who self-reported fair-to-poor health. For example, compared with participants with good olfaction, the cause-specific HR of CHF for poor olfaction was 1.76 (95%CI: 1.20, 2.58) among participants with self-reported very good-to- excellent health, versus 0.92 (95%CI: 0.58, 1.47) among those with fair-to-poor health. Figure 4.2 Cause-specific hazard ratios (HRs) and 95% confidence intervals (CIs) of olfaction in relation to congestive heart failure with up to 12 years of follow-up in subgroup analyses (n=2,421). Each model was adjusted for the interaction between baseline olfactory status and the subgroup factor of interest, plus covariates of age, sex, race, education, study site, smoking status, brisk walking, body mass index, self-reported general health status (GHS), systolic blood pressure, use of antihypertensive medications, diabetes, depressive symptoms, total cholesterol, high-density lipoprotein-cholesterol, prevalent coronary heart diseases (CHD)/stroke, heart rate, left ventricular hypertrophy, abnormal lung function, albumin and estimated glomerular filtration rate, stratified by the tertile of interleukin 6 We, therefore, further explored details of this relationship among participants who self- reported a very-good-to-excellent health at baseline. When we analyzed the B-SIT score as a continuous variable using the perfect score of 12 as the reference, the cause-specific HR of CHF 39 ascended as the olfaction performance decreased until the B-SIT score of 4, after which the HRs were slightly attenuated (Figure 4.3). Figure 4.3 Cause-specific hazard ratios (HRs) and 95% confidence intervals (CIs) for congestive heart failure (CHF) by continuous olfaction score among participants who self-reported very- good-to-excellent health (n=1,100). Olfaction was measured by the Brief-Smell Identification test (B-SIT), the perfect score of which as 12 was used to be the reference. The model was adjusted for age, sex, race, education, study site, smoking status, brisk walking, body mass index, use of antihypertensive medications, diabetes, depressive symptoms, total cholesterol, high- density lipoprotein-cholesterol, prevalent coronary heart disease/stroke, heart rate, left ventricular hypertrophy, abnormal lung function, albumin, interleukin 6 and estimated glomerular filtration rate, stratified by groups of systolic blood pressure (140mmHg as the cut- off) 4.4 Discussion To our knowledge, this is the first study that aims to examine the association of olfaction with major cardiovascular diseases among older adults. Such an investigation is important because poor olfaction is prevalent in older adults, cardiovascular disease is the leading cause of death, and their connections are biologically plausible. In this large community-dwelling cohort, we found that a single test of olfactory status was associated with the risk of developing CHF for up to 12 years of 40 follow-up. This association was robust across subgroups of age, sex, race, and prevalent CHD/stroke, but appeared to be more evident among participants who reported very-good-to- excellent health at baseline. However, olfactory status was not statistically significantly associated with the risk of developing CHD or stroke. Taken together, this study provides interesting preliminary evidence that poor olfaction may be associated with long-term CHF risk in older adults, particularly among those who consider their general health as very good or excellent. Cardiovascular disease is the leading causes of death, and its incidence increases with age178. While cardiovascular risk factors are among the best characterized62, as people age, known associations with cardiovascular diseases may attenuate, possibly due to aging and resilience to existing risk factors among survivors179. There remains a critical need to identify novel factors associated with adverse cardiovascular outcomes in older adults to further inform risk prediction and intervention. In contrast to known cardiovascular risk factors, including hypertension, obesity, and smoking, even health-conscious individuals rarely pay attention to their sense of smell2,23,24. In older adults, poor olfaction is age-dependent2,3,23 and robustly predicts age-related neurodegenerative diseases and all-cause mortality180,181. Emerging evidence further suggests that poor olfaction is associated with a broad range of age-related adverse health conditions beyond neurodegenerative diseases, including cardiovascular diseases2,127,123,118,122,35, diabetes182,183, cirrhosis184, kidney dysfunction185, frailty186, pneumonia159, depression106, physical functioning decline105, and loss of body lean mass187. Some of the evidence comes from well-designed longitudinal cohorts105,106,127,159,185,187. These intriguing findings, together with well-documented associations of poor olfaction with neurodegenerative diseases and mortality, raise the possibility that poor olfaction may be a marker of accelerated biological aging across multiple systems. As olfaction identification tests are simple, fast, and non-invasive, future studies should investigate 41 this hypothesis and explore how it may help promote healthy aging by identifying potential health issues in older adults early. Although our findings are preliminary and need confirmation, we speculated that, in older adults, poor olfaction may be related to cardiovascular health either as a subclinical marker or a potential risk factor. As vascular remodeling develops and progresses, insufficient blood supply may gradually impair the health of nasal epithelium and structures in the olfactory signal pathway, limiting normal olfactory functioning9. Supporting this viewpoint, preliminary evidence suggests that carotid intima-media thickness and artery plaques, two subclinical markers of atherosclerosis, have been associated with the olfactory decline in older adults90,91. On the other hand, it is also biologically plausible that poor olfaction may adversely affect cardiovascular health. It has been speculated that impaired olfaction may alter ones’ diet and food choices, which could negatively impact their nutritional status and overall health over time98–100. Further, impaired olfaction may contribute to a depressed mood, social isolation, and physical inactivity188,189. All these may potentially increase one’s vulnerability to endogenous and exogenous stressors, contributing to or exacerbating cardiovascular disease risk. However, to date, the role of olfaction in cardiovascular health remains speculative with limited evidence. To the best of our knowledge, only one prospective study has explored the association of poor olfaction with the risk of cardiovascular diseases. In the National Social Life, Health, and Aging Project, Siegel et al reported that olfactory decline during the first 5 years was associated with marginally significantly greater odds of reporting a diagnosis of heart disease (odds ratio: 1.75, 95% CI: 0.93–3.31)127. Notably, in this study, the diagnosis of heart disease was self-reported only once at the year-10 follow-up survey and was analyzed as a secondary outcome. The current study is large, community-based, and specifically designed to examine olfaction in relation to the 42 risk of adjudicated incident CHD, stroke, and CHF. In the analyses, we carefully accounted for the competing risk of death and relevant covariates. We found that a single smell test was not statistically significantly associated with incident CHD or stroke events. However, compared to participants with good olfaction, those with moderate or poor olfaction had a robust, albeit modest, increase in CHF risk for up to 12 years of follow-up. In contrast to CHD and ischemic stroke where arterio- and/or atherosclerosis are major mechanisms190,191, CHF is etiologically more complex. The latter is a multiorgan syndrome with a net outcome of a failing heart, characterized by a reduced cardiac output and increased venous pressure192. Coronary arteriosclerosis contributes to CHF, but any sustained myocardial stress such as increased cardiac pressure and volume overload may lead to myocardial hypertrophic response and cardiac remodeling, eventually resulting in CHF176. Although speculative, the differential results of CHF from CHD/stroke support the possibility that poor olfaction may signal or elevate one’s vulnerability to myocardial stressors. Future studies are warranted to confirm our observations and to investigate potential mechanisms, which may lead to interventional opportunities with poor olfaction either as a red flag or a potential target of intervention, for example through olfactory training, dietary manipulation, and exercise193–195. Although the results were not statistically significant across subgroups of self-reported health status, the association of poor olfaction with incident CHF appeared to be more evident among participants who self-reported very-good-to-excellent health status at baseline, similar to our finding on the association of poor olfaction with mortality57. In contrast, the association was close to null among individuals who self-evaluated their overall health as fair or poor. Self-reported health is a subjective perception that one may integrate their biological, social, mental, and functional health perspectives with their personal and cultural beliefs and their attitudes towards 43 health196. While the report is subject to individual interpretation, it has been commonly used in health research to assess one’s general health status and it robustly predicts the risk of mortality in older adults197. We speculated individuals who rated their health as fair or poor might already have multiple comorbidities or risk factors that play a detrimental role in their myocardial health, outweighing that from poor olfaction. In contrast, among those who reported very-good-to- excellent health, poor olfaction may serve as an early signal for deteriorating myocardial health in the absence of other clinical signals for increased CHF risk. Notably, in this subgroup, the association estimate of poor olfaction with CHF was modestly higher than that of known leading causes of CHF such as prevalent cardiovascular disease and habitual smoking (Table A3.3). Taken together, we speculate poor olfaction is likely an early indicator for deteriorating myocardial health in apparently healthy older adults, awaiting independent confirmation and investigation of underlying mechanisms. Strengths of this study include the relatively large number of community-based participants, more than a decade of follow-up, meticulous health surveillance and outcome assessments, careful covariate identification, and statistical analyses. Our study also has several notable limitations. First, study participants were all older than 70 at enrollment but were relatively high-functioning. Therefore, study findings may not be readily generalizable to younger populations or populations with different demographics or health status. Second, olfaction was only assessed once. As olfactory function declines fast with age in older adults, participants' olfaction may continuously decline over follow-up, which was not captured in the current study. We therefore might have underestimated the role of olfaction in signifying or maintaining cardiovascular health in older adults. Future longitudinal studies with relatively younger participants and repeated assessments of olfaction may better characterize the role of olfaction in 44 cardiovascular health in the context of aging. Third, as the B-SIT was designed to screen for smell identification deficit in large populations, our study did not address the association of other olfactory modalities (e.g., detection and discrimination) with the risk of major cardiovascular outcomes. Fourth, while our study findings were robust as evidenced in multiple sensitivity analyses, as in any observational study, we could not exclude the possibilities of chance finding or residual confounding. Finally, while our study suggests that both poor and moderate olfaction are associated with the future risk of experiencing a CHF event, it provides little clue to the underlying mechanisms. In conclusion, in this well-established community-based study of older adults, we found that poor olfaction was statistically significantly associated with risk for CHF for up to twelve years, but not with risk for CHD or stroke. Future studies should confirm this observation and investigate underlying mechanisms. 45 CHAPTER 5: OLFACTORY STATUS IN RELATION TO STROKE IN THE ARIC STUDY 5.1 Introduction Stroke occurs in about 800,000 adults annually in the US and is the fifth leading cause of death58. While over 80% of stroke events can be categorized as ischemic, the pathologies and causes of stroke are heterogenous, even among ischemic strokes67. Less than half of ischemic strokes can be attributed to large artery atherothrombosis or cardiac-origin emboli68. Conventional risk factors may not sufficiently stratify stroke risk, especially for strokes with atypical or complex pathologies, such as those due to cerebral small vessel disease198. In addition, known risk factors may have attenuated associations with incident stroke in the elderly179. It is hereby imperative to identify novel markers to facilitate the risk stratification of stroke in older adults. Olfactory dysfunction is common in older adults, affecting over a quarter of those aged 65 years and older2. This sensory deficit can be simply tested with a non-invasive smell identification test199,200. Smell identification involves odor sensation and cognition, and thus requires intact functions of both the peripheral and central olfactory structures43. Cerebral hemodynamic abnormalities related to the olfactory system may compromise normal olfactory functioning. In support of this, major risk factors for stroke, such as subclinical intracranial atherosclerosis191 and disturbed intracranial fluid dynamics92, are found to be associated with olfactory dysfunction90,91,93,94. Moreover, olfactory dysfunction may adversely affect one’s diet and lifestyle99,100 which may in turn increase future risk of stroke201. Therefore, either as a marker or a contributor, poor olfaction may signify higher risk of stroke in older adults. Despite the intriguing biological plausibility of an association between olfaction and stroke, empirical evidence is scant. In a hospital-based magnetic resonance imaging study, stroke 46 patients were found to have smaller peripheral and central olfactory areas than control patients124. In contrast, we did not observe a statistically significant association between poor olfaction and risk of stroke in the Health ABC Study1. Therefore, we further investigated whether poor olfaction is related to future stroke risk in a large community-dwelling US cohort of older adults from the ARIC Study. 5.2 Methods 5.2.1 Study Population The ARIC Study is a community-based prospective cohort that aimed to investigate the etiology of atherosclerosis and its clinical sequelae in the US130,131. Briefly, between 1987-1989, this study recruited 15,792 participants aged 45-64 years from four U.S. communities (Forsyth County, North Carolina, Jackson, Mississippi, suburbs of Minneapolis, Minnesota, and Washington County, Maryland)130,131. Since enrollment, this study conducted periodic comprehensive in- person clinical examinations and annual or semi-annual (since 2012) phone calls to update participant’s health. Meanwhile, the study has closely monitored the health and survival of participants via the cohort-wide hospitalization and death surveillance. Over the past three decades, the ARIC study has provided invaluable information in the cardiovascular field and demonstrated the importance of population-based research in the understanding of cardiovascular health131. The ARIC’s fifth clinical examination (Visit 5) in 2011-2013 included a brief smell identification test and was thus considered as the baseline for the current analysis. Of 6,515 who attended Visit 5 in-person and provided their written informed consent, we excluded 18 participants with race other than Black or White, 24 Black participants from Minneapolis and Washinton County due to small numbers, 437 with missing olfactory score, and 237 with history 47 of stroke. We therefore followed 5,799 at-risk participants until the date of first stroke event, death, last contact, or December 31, 2020, whichever came first. The ARIC study protocol was approved by all participating institutions’ institutional review boards. This specific analysis was approved by the institutional review board of Michigan State University. 5.2.2 Olfactory Status Olfactory status was assessed using the 12-item SS odor identification test133. Briefly, participants were asked to smell 12 common odors in felt-tip pens, one at a time, and then to select the odor from four possible answers in a multiple-forced-choice format. This test is reliable and easy to administer133. It has been commonly used in clinical and epidemiological settings202–205. The test score ranges from 0 to 12, as each correct answer is given one point. We defined poor olfaction as a smell score ≤8, moderate as 9-10, and good as 11-12, which correspond to the tertile of the test score distribution in the study population. In the sensitivity analysis, we further categorized poor olfaction into anosmia (score≤6) and hyposmia (7-8), consistent with previous published studies22,206. 5.2.3 Stroke Events The fatal and non-fatal hospitalized strokes were identified by annual or semi-annual phone interviews and record review for hospitalizations146,147,207,208. Throughout the follow-up, hospitalizations with possible stroke-related discharge diagnosis codes were identified for ARIC participants. Specifically, the ARIC Study considered the ICD -9-Clinical Modification (CM) codes of 430-438 before 1997 as possible stroke-related hospitalizations, followed by ICD-9-CM codes 430-436 and ICD-10 codes G45, I60-I67 until 2019, and ICD-10 codes G45, I60-I64 afterward. The stroke-related deaths were identified through linkage to the vital statistics department for each death or else the National Death Index146,207. All the events were classified 48 independently by the computerized stroke algorithm and a physician reviewer from the Stroke- Mortality and Morbidity Classification Committee147. Disagreement between the two sources led to adjudication by another physician. In this study, we considered definite or probable incident fatal and non-fatal strokes as the primary outcome. As detailed before146,147, strokes were categorized into 1 of 4 primary types based on the standardized protocol, including subarachnoid hemorrhage, intracerebral hemorrhage, thrombotic brain infarction, and embolic brain infarction. Given that most strokes in the US are ischemic, in one sensitivity analysis, we restricted the outcome of interest to ischemic stroke which includes thrombotic and embolic brain infarction. 5.2.4 Covariates We considered a range of covariates at Visit 5 when olfaction was assessed. Date of birth, sex, race, and education level were self-reported at Visit 1 and smoking status at Visit 5. As Black participants were predominantly from Jackson, we further categorized race based on the study center as is commonly done in the analysis of ARIC data209. We defined education as less than high school, high school or equivalent, and at least some college. Apolipoprotein E (APOE) genotype was measured using TaqMan system and dichotomized to APOE4 carrier (≥1 ε4 alleles) and noncarrier (no ε4 alleles)210. BMI was calculated by dividing weight by height-squared (kg/m2) and the systolic blood pressure by averaging 2 measurements in the sitting position. The uses of lipid-lowering and antihypertensive medications were collected by asking participants to bring prescription and nonprescription drugs they had used in the last 4 weeks211,212. We defined comorbidities based on published protocols: 1) diabetes as a fasting glucose level ≥ 126 mg/dL, a non-fasting glucose level ≥200 mg/dL, HbA1C ≥6.5%, a self-reported physician diagnosis, or self- reported use of antidiabetic medications150; 2) CHD as a combination of self-reported CHD at Visit 1 and adjudicated events between Visit 1 and Visit 5213; 3) HF ascertained from adjudicated heart 49 failure hospitalization since 2005, self-reported HF at Visit 3-5, or hospitalization with ICD-9-CM code of 428 before 200581; 4) atrial fibrillation identified from the electrocardiogram or hospitalization discharge diagnosis214; 5) dementia adjudicated through in-person neuropsychological evaluations, or identified through telephone interview for cognitive status score, informant rating, or hospitalization215; 6) Parkinson’s disease adjudicated by experts’ review of self-reported diagnosis, medication uses, hospitalization discharges, or death certificate, along with additional diagnostic information from patients and their treating physicians216; 7) depressive symptoms defined as ≥ 9 out of 11 items on the Center for Epidemiologic Studies Depression questionnaire at Visit 5217. We assessed frailty using the Fried Frailty phenotype and combined prefrailty and frailty as having ≥1 of the five phenotypes, including weight loss, exhaustion, slow walking speed, low physical activity, and low grip strength218. Total cholesterol and HDL-C were measured in fasting plasma following standard procedures219. Plasma creatinine and cystatin C were used in the CKD-EPI creatinine-cystatin equation for eGFR173,220. 5.2.5 Statistical Analyses In descriptive analyses, we calculated the crude CIF of stroke along with its competing risk of death, and tested the equality of CIF across olfactory statuses at baseline using the Gray’s test174. In the multivariable analyses, we used the discrete-time Fine-Gray sub-distribution model to estimate the association of olfactory status with risk of stroke, accounting for covariates and competing events of death155,156. In brief, we used the pooled logistic regression with 1-month interval to estimate the sub-distribution hazard of developing stroke at each month. To investigate the potential time-varying association of olfactory status with the outcome of interest during the follow-up, we added interaction terms between olfactory statuses and follow-up time in the pooled logistic regression. For covariates, we used the Kolmogorov-Smirnov test and the Cramer von 50 Mises test to check the proportional sub-distribution hazard assumption221. If either test showed a statistically significant time-varying association of a covariate with the outcome, we added an interaction term between the covariate and time. We selected the cubic spline with the degree of freedom of 4 (one inner knot at 52 month) as the functional form of time, because it has the least prediction error222 at most of the follow-up years by using 100-fold cross-validation, compared with cubic splines with other degrees of freedom and the non-parametric step function. We calculated the counterfactual cumulative risk of stroke (Pr⁡[𝑌𝑡 𝑎]) at each follow-up month for all the study participants given their covariates under the hypothetical conditions of different olfactory statuses155. We used the following equation Pr[𝑌𝑡 𝑎] = ∑ 𝐸[𝑌𝑡 𝑎|𝑎, 𝑍 = 𝑧] Pr[𝑍 = 𝑧] 𝑧 𝑡 𝑗−1 𝑠𝑏(𝑎|𝑧) ∏[1 − ℎ𝑘 𝑘=0 = ∑ {∑ ℎ𝑗 𝑧 𝑗=1 𝑠𝑏(𝑎|𝑧)] } Pr⁡[𝑍 = 𝑧] at time 𝑡 (𝑡 = 1, … , 𝑇), where ℎ𝑡 𝑠𝑏(𝑎|𝑧) is the conditional sub-distribution hazard at time 𝑡 when the olfactory status is 𝑎 . Specifically, the sub-distribution hazard function is defined as 𝑠𝑢𝑏(𝑎|𝑧) = Pr⁡[𝑌𝑡 ℎ𝑡 𝑎 = 1|𝑌𝑡−1 𝑎 = 0, 𝑧] under the discrete time scale. Finally, we compared the absolute risk across olfactory statuses based on the baseline covariate distribution of the entire sample and then estimated the period-specific RD and RR with good olfaction as the reference. We presented three sets of models with sequentially increasing numbers of covariates adjusted. Model 1 was adjusted for basic demographics of age, sex, race‒center, and education. Model 2 aimed to assess the association of olfaction with risk of stroke independent of important vascular and cardioembolic risk factors223,63,62,67, including APOE4 carrier, smoking status, BMI, total cholesterol, HDL-C, lipid-lowering medication use, diabetes, systolic blood pressure, antihypertensive medication use, prevalent atrial fibrillation, CHD, heart failure, and eGFR. 51 Finally, as frailty is common in older adults and associated with both poor olfaction224–226 and incident stroke114,227, we further adjusted for frailty in Model 3. To further demonstrate the robustness of study findings to analytical approaches, we used the cause-specific Cox regression to estimate the period-specific cause-specific hazard ratio for olfactory status. Further, we conducted multiple subgroup analyses by age groups (<75 vs. ≥ 75 years), sex (male vs. female), race (White vs. Black), and history of major cardiovascular events including CHD and heart failure (no vs. yes). In addition, we conducted the following sensitivity analyses to check the robustness of study findings: 1) we classified poor olfaction into hyposmia and anosmia to further examine the dose-response pattern of the olfaction-stroke relationship; 2) to minimize the potential impact of dementia on olfactory testing, we redid the analysis by excluding participants with dementia at baseline; 3) we analyzed ischemic stroke as the outcome of interest; 4) we conducted multiple imputation for the 7.9% missing values of frailty and repeated the primary analysis, as detailed in Appendix 4. Finally, to demonstrate the strength of olfaction‒ stroke association in the context of other known major risk factors for stroke, we additionally presented the associations of for CHD, as a proxy of systemic atherosclerosis, and atrial fibrillation, as a risk factor for cardioembolic stroke, with stroke risk, using the same analytical approach as described above. We used SAS (version 9.4; SAS Institute Inc. Cary, NC, USA) for description and cause-specific hazard modeling, and R (version 4.1.3) for all the other analyses with a two-sided α of 0.05. 5.3 Results Eligible study participants included 3,423 women and 2,376 men, with an average age at baseline of 75.5±5.1 years old and 22.2% Black. Compared with participants with good olfaction, those with poor olfaction were more likely to be older, male, Black, APOE4 carriers, and current/former 52 smokers, and to report lower education level (Table 5.1). They were also more likely to use antihypertensive and lipid-lowering medications, and to have diabetes, atrial fibrillation, CHD, heart failure, prefrailty/frailty, dementia, Parkinson’s disease, depressive symptoms, and lower levels of total cholesterol, HDL-C, and eGFR. Table 5.1 Population characteristics by baseline olfactory status (n=5,799), the ARIC Study 2011-2013 Variables a Age in year Sex Male Race Black Center Forsyth county Jackson Minneapolis suburbs Washington County Race‒center White in Forsyth County White in Minneapolis suburbs White in Washington County Black in Forsyth County Black in Jackson Education Less than high school High school or equivalent At least some college APOE4 carrier Smoking status Never smoker Former smoker Current smoker Body mass index in kg/m2 <25.0 25.0 - <30 ≥30.0 Total cholesterol in mmol/L Good (n=2,121) 74 (71, 78) 721 (34) 231 (10.9) 532 (25.1) 208 (9.8) 739 (34.8) 642 (30.3) 509 (24) 739 (34.8) 642 (30.3) 23 (1.1) 208 (9.8) 176 (8.3) 894 (42.1) 1051 (49.6) 481 (22.7) 944 (44.5) 1077 (50.8) 100 (4.7) 550 (25.9) 851 (40.1) 720 (33.9) Olfactory status Moderate (n=1,924) 75 (71, 79) 797 (41.4) 420 (21.8) 423 (22) 382 (19.9) 567 (29.5) 552 (28.7) 385 (20) 567 (29.5) 552 (28.7) 38 (2) 382 (19.9) 259 (13.5) 823 (42.8) 842 (43.8) 501 (26) 785 (40.8) 1017 (52.9) 122 (6.3) 445 (23.1) 790 (41.1) 689 (35.8) 4.68 (4.03, 5.46) 4.63 (3.96, 5.40) HDL-C in mmol/L 1.34 (1.11, 1.58) 1.29 (1.09, 1.55) Use of lipid lowering agents 1143 (53.9) 1074 (55.8) 53 Poor (n=1,754) 77 (72, 81) 858 (48.9) 637 (36.3) 338 (19.3) 608 (34.7) 377 (21.5) 431 (24.6) 309 (17.6) 377 (21.5) 431 (24.6) 29 (1.7) 608 (21.8) 382 (21.8) 702 (40) 670 (38.2) 557 (31.8) 701 (40) 947 (54) 106 (6) 437 (24.9) 752 (42.9) 565 (32.2) 4.53 (3.83, 5.20) 1.28 (1.06, 1.50) 1016 (57.9) Table 5.1 (cont’d) Diabetes Systolic pressure in mmHg 1390 (79.2) 109 (5.1) 258 (12.2) 187 (8.8) 675 (38.5) 129 (118, 141.5) 601 (28.3) 128.5 (118, 140.5) 1503 (70.9) 641 (33.3) 128.5 (117.5, 140.5) 1459 (75.8) 67.7 (55.9, 79.9) 66.7 (54.2, 79.2) 64.2 (50.6, 76.3) 135 (7) 259 (13.5) 223 (11.6) Antihypertensive use eGFR in mL/min/1.73m2 Atrial fibrillation CHD Heart failure Fried Frailty phenotype 1077 (50.8) Robust 940 (44.3) Prefrail or frail 104 (4.9) Missing 8 (0.4) Dementia 3 (0.1) Parkinson’s disease Depressive symptoms 102 (4.8) Abbreviations: ARIC: Atherosclerosis Risk in Communities; HDL-C: high-density lipoprotein- cholesterol; eGFR: estimated glomerular filtration rate; CHD: coronary heart diseases. a Continuous variables are presented as median (25th, 75th percentile), and categorical variables as number (column %) 838 (43.6) 955 (49.6) 131 (6.8) 23 (1.2) 5 (0.3) 142 (7.4) 568 (32.4) 960 (54.7) 226 (12.9) 182 (10.4) 30 (1.7) 140 (8.0) 153 (8.7) 296 (16.9) 306 (17.4) During up to 9.6 years (median 8.3 years) of follow-up, 332 individuals had an incident stroke event, of which 256 were classified as ischemic stroke. The number of incident stroke events was 95 among participants with good olfaction, 100 among those with moderate olfaction, and 137 among those with poor olfaction. Figure 5.1 shows the crude association of olfactory status with incident stroke along with the competing event of death. For both outcomes, participants with poor olfaction had a higher cumulative incidence than those with better olfactory statuses during the follow-up. 54 Figure 5.1 Crude cumulative incidence function of stroke and its competing event of death by baseline olfactory status (n = 5,799) Multivariable analyses confirmed the association of poor olfaction with higher risk of stroke throughout the follow-up (Figure 5.2 and Table 5.2). At year 9.6, the marginal cumulative incidence of stroke was 5.3% (95% CI: 4.2-6.3%) for good olfaction group, 5.9% (95% CI: 4.8- 7.1%) for moderate olfaction, and 7.7% (95% CI: 6.5-9.1%) for poor olfaction. We however also found that the association appeared to be more evident during the first 6 years of follow-up and 55 was modestly attenuated afterwards. For example, the fully adjusted RR for poor vs. good olfaction was 2.14 (95% CI: 1.22, 3.94) at year 2, 1.98 (95% CI: 1.43, 3.02) at year 4, 1.91 (95% CI: 1.43, 2.77) at year 6, 1.49 (95% CI: 1.17, 2.00) at year 8, and 1.45 (95% CI: 1.16, 1.95) at year 9.6 (Table 5.2). Correspondingly, the RD between poor and good olfaction groups increased fast during the first 6 years, and then stabilized with extended follow-up (Figure 5.2). The findings were consistent in the alternative analysis using cause-specific hazard models (Table A4.1). In both analyses, we did not observe a statistically significant difference in risk or cause-specific hazard of stroke between moderate and good olfaction. Figure 5.2 Marginal adjusted cumulative incidence function of stroke by olfactory status and the risk difference comparing moderate and poor vs. good olfaction. The cumulative incidence was estimated by the discrete-time sub-distribution hazard model, adjusting for covariates in Model 3 Table 5.2 Marginal adjusted risk ratios of stroke comparing moderate/poor with good olfaction during the follow-up (n=5,799) Olfactory status Model 1b Risk ratio (95% confidence interval) a by follow-up years 2-Year 4-Year 6-Year 8-Year 9.6-Year 56 Table 5.2 (cont’d) Good Moderate Poor Model 2 c Good Moderate Poor Model 3 d Good Moderate Poor Reference 1.31 (0.69,2.49) 2.53 (1.46,4.67) Reference 1.22 (0.79,1.93) 2.22 (1.6,3.36) Reference 1.34 (0.94,1.98) 2.04 (1.57,2.96) Reference 1.2 (0.62,2.31) 2.17 (1.24,3.97) Reference 1.14 (0.72,1.81) 2 (1.44,3.09) Reference 1.18 (0.61,2.26) 2.14 (1.22,3.94) Reference 1.13 (0.71,1.8) 1.98 (1.43,3.02) Reference 1.29 (0.9,1.9) 1.94 (1.47,2.84) Reference 1.27 (0.89,1.86) 1.91 (1.43,2.77) Reference 1.1 (0.79,1.48) 1.55 (1.22,2.06) Reference 1.07 (0.78,1.46) 1.51 (1.2,2.02) Reference 1.07 (0.77,1.45) 1.49 (1.17,2) Reference 1.14 (0.89,1.56) 1.51 (1.2,2.02) Reference 1.11 (0.86,1.52) 1.47 (1.18,1.97) Reference 1.1 (0.85,1.53) 1.45 (1.16,1.95) a Marginal adjusted risk ratio was calculated through the multivariable discrete-time Fine-Gray model; 95% confidence interval was obtained through bootstrapping with 300 samples. b Model 1 includes age, sex, race-center, and education, plus interaction terms between time and olfactory status. c Model 2 further includes APOE4 carrier, smoking status, body mass index, diabetes, systolic blood pressure, antihypertensive medication, total cholesterol, HDL-cholesterol, lipid lowering medication, atrial fibrillation, coronary heart disease, heart failure, and estimated glomerular filtration rate, plus two-way interaction terms between time and education & HDL-cholesterol. d Model 3 further includes frailty. 57 Figure 5.3 Stratified marginal adjusted risk ratios (RRs) and 95% confidence intervals (95% CIs) of stroke in the subgroup analyses by baseline age groups (41.5 g/m2.7 in women and >45.0 g/m2.7 in men based on ARIC reference limits for LV mass indexed to height2.7, moderate or greater aortic stenosis as a peak transaortic velocity of >3.0 m/sec; moderate or greater mitral regurgitation based on a mitral regurgitation get area-to-left atrial area ratio of >0.20, and moderate or greater mitral stenosis based on a mean antegrade transoral gradient of ≥5mmHg. 7.2.5 Covariates Covariates were largely measured at Visit 5 except that date of birth, sex, race, and education was self-reported at Visit 1 and general health status was self-reported at the annual follow-up one year within the Visit 5 date. As White participants were primarily from Washington County, Minneapolis suburbs, and Forsyth County, race was categorized based on the study center. Education was classified as less than high school, high school or equivalent, and at least some college level. Age at Visit 5 was calculated as a continuous variable and BMI as weight divided by square of height (kg/m2). We defined self-reported smoking status as never, former, and current smokers. Use of lipid-lowering medication was assessed using medication inventory method211. We defined prevalent comorbidities based on published criteria: 1) hypertension as an average systolic blood pressure of ≥ 140 mmHg, or an average diastolic blood pressure of ≥ 90 mmHg or use of antihypertensive medications262; 2) diabetes as a fasting glucose level ≥ 126 mg/dL, a non- fasting glucose level ≥200 mg/dL, HbA1C ≥6.5%, self-reported physician diagnosis, or self- reported use of antidiabetic medications150; 3) CHD as self-reported CHD at Visit 1 or adjudicated events between Visit 1-5213; 4) atrial fibrillation as identified from the electrocardiogram or hospitalization214; 5) dementia as identified according to in-person neuropsychological evaluations, telephone cognitive assessment, informant rating, or hospitalization215. Prefrailty or frailty was defined as ≥1 symptoms of the Fried frailty phenotypes, including exhaustion, weight 83 loss, slow walking speed, low grip strength, and low physical activity218. Total cholesterol, HDL- C, creatinine, and cystatin were measured through standardized procedures219,220.The latter two biomarkers were used in the CKD-EPI creatinine-cystatin equation for eGFR173. The HF clinical stage at baseline was used as one of our stratification factors and defined following the published protocol in the ARIC Study81,150. Briefly, HF Stage A required having at least one of the following HF risk factors in the absence of structural heart disease or symptoms of HF, including prevalent atherosclerotic CVD, hypertension, diabetes mellitus, obesity, metabolic syndrome, and chronic kidney disease. Stage B was defined as having structural heart disease or elevated cardiac biomarkers, including NT-proBNP of ≥125 pg/mL or hs-cTnT of >14ng/L in women and >22ng/L in men. The rest of individuals who did not at Stage A or Stage B were considered at Stage 0. 7.2.6 Statistical analysis In the analysis of olfaction and risk of HF, we first used the Gray’s test to evaluate the crude association of olfactory statuses with the CIF of HF and its competing event of death174. In the multivariable analyses, we used discrete-time sub-distribution model to evaluate the association of olfactory status with risk of HF accounting for covariates and competing risk of death155,156. The details regarding the model building were presented in previous chapters. Using the estimated model, we calculated the absolute risk across olfactory statuses conditioning on the baseline covariate distribution across the entire sample and calculated RRs and RDs with good olfaction as the reference. These risk-based assessments indicate the total association which includes both the direct association between olfaction and HF and the indirect association through competing event of death155 (details in Chapter 3.5). We considered two sets of covariates with an increasing number of covariates added. 84 Model 1 adjusted for basic demographics, including age, sex, race-center, and education. Model 2 further adjusted for smoking status, self-reported general health status, BMI, diabetes, hypertension, total cholesterol, HDL-C, lipid lowering medication, CHD, atrial fibrillation, and eGFR to examine the independence of the association between olfactory status and HF risk. Given the close relationships between olfaction and frailty117 and between frailty and HF114,255, model 3 further adjusted for frailty. We then considered HFrEF and HFpEF as the outcomes of interest, respectively. Next, we conducted stratified analyses by age (0.05. Separating poor olfaction into anosmia and hyposmia, the positive association with HF risk appeared to be more evident for anosmia than hyposmia (Table A6.2). The results barely changed after using a more restrictive definition of HF hospitalizations or deleting prevalent dementia cases from the analysis (Table A6.3 and A6.4). In the cross-sectional analysis of olfaction and subclinical HF biomarkers, participants with poor olfaction had higher crude levels of NT-proBNP and hs-TnT, and a higher prevalence of structural heart disease. After adjusting for covariates, participants with poor olfaction had 13.3pg/mL higher median level of NT-proBNP and 0.8ng/L higher median level of hs-TnT, 91 compared to those with good olfaction (Table 7.4). Further, people with poor olfaction were also more likely to have structural heart disease with an OR of 1.24 (95% CI: 1.06, 1.46). Table 7.4 The cross-sectional association of olfactory status with NT-proBNP, hs-TnT, and structural heart disease Olfactory status Good Moderate Poor NT-proBNP, pg/mL (n=5,012) hs-TnT, ng/L (n=5,169) Structural heart disease (n=5,217) Crude value a 111.4 (146.6) 116.8 (172.1) 134.9 (203.4) Adjusted difference b Crude value a Adjusted difference b Crude value c Adjusted OR d Reference 9 (7) Reference 624 (31.7) Reference 7.0 (0,14.1) 10 (8) 0.1 (-0.2,0.5) 573 (32.9) 13.3 (4.6,22.1) 12 (10) 0.8 (0.3,1.3) 543 (36.1) 1.05 (0.90,1.21) 1.24 (1.06,1.46) Abbreviation: NT-proBNP: N-terminal B-type natriuretic peptides; hs-TnT: high-sensitive cardiac troponin T; OR: odds ratio a Median (IQR) is presented by olfactory statuses. b Adjusted differences in median across olfactory statuses (good olfaction as the reference) was estimated by the quantile regression, adjusting for age, sex, race-center, education, smoking status, BMI, self-reported general health status, diabetes, hypertension, total cholesterol, high-density lipoprotein-cholesterol, lipid-lowering medications, coronary heart disease, atrial fibrillation, renal function and frailty. c Number (row %) is presented by olfactory statuses. d OR was estimated by the logistic regression, adjusting for age, sex, race-center, education, smoking status, BMI, self-reported general health status, diabetes, hypertension, total cholesterol, high-density lipoprotein-cholesterol, lipid-lowering medications, coronary heart disease, atrial fibrillation, renal function, and frailty. 7.4 Discussion In a large community-dwelling cohort of older adults, we found that poor olfaction identified by a single smell test was modestly associated with higher risk of HF hospitalization for 8 years. Interestingly, poor olfaction appeared to have a evident association with HFrEF, but not with HFpEF. Further, we also identified that poor olfaction was associated with pre-HF markers indicating subclinical myocardial pathology and structural dysfunction. Despite the modest association we identified, study findings were robust in multiple subgroup and sensitivity analyses. Notably, this observation is consistent with our recent finding from another cohort of older US 92 adults, highlighting the potential relevance of this common sensory deficit to future HF risk. HF is a prevalent cardiac syndrome, especially among older adults251. While the clinical onset HF is usually acute, its underlying structural or functional cardiac dysfunction takes time to build263. The natural progression of subclinical HF consists of two distinct stages. Stage A is characterized by the presence of major HF risk factors, while Stage B involves the cardiac structural dysfunction264. The progressions can be driven by various reasons (such as atherosclerosis and cardiomyopathy) and further complicated by aging-related physiological changes, making the identification and prevention of HF challenging. Frailty is a common geriatric disorder characterized by a declined restoration of homeostasis after stress attacks254. While HF may lead to increased systematic vulnerability, frailty may in turn accelerate or signify the development of HF255. Accumulating evidence shows that frailty is associated with subclinical markers of structural and functional abnormalities in the vascular system and myocardium in older adults116,256–258. Notably, a recent proteomic study provided provocative mechanistic evidence linking frailty with HF115 by identifying multiple shared biological mechanisms, highlighting the extracardiac pathways of HF development in late life. Poor olfaction is common but often underrecognized among older adults2,3. This neglected sensory deficit, however, may have profound health implications5,181. In addition to its robust association with neurodegeneration and mortality, accumulating evidence shows the close association of poor olfaction with frailty117. Interestingly, some longitudinal studies have found that poor olfaction may occur prior to frailty and predict faster deterioration of its individual components105,187,226. For example, poor olfaction was associated with greater weight loss, including both fat and lean mass187, and with faster decline in physical functioning among community-based older adults105. Despite limited empirical data, poor olfaction in older adults 93 may contribute to the cascade of events leading to frailty by negatively affecting their nutritional behaviors, emotions, and lifestyles56,98,100. Given the increasingly recognized relationship between frailty and HF, it is biologically plausible that poor olfaction signifies future risk of both conditions among older adults. To the best of our knowledge, only one prospective study has investigated the association between olfaction and incident HF. In the Health ABC Study, we found that poor olfaction was associated with 28% higher cause-specific hazard of congestive HF during up to 12 years of follow-up1. Compared with the Health ABC Study, the ARIC Study was designed to study cardiovascular outcomes, was more inclusive with broader age range and functional status at baseline, conducted comprehensive HF adjudication protocols, and adjudicated HF sub-types. Further, in the analysis, we used risk-based association measures to account for the competing risk of death rather than simply treating these competing events as censoring151. Nevertheless, our finding supports that from the Health ABC Study. The current Study further suggests the association was limited to HFrEF, a novel observation that has not been reported. In support of these findings, we observed that poor olfaction was significantly associated with well-established subclinical HF biomarkers, including NT-proBNP, hs-cTnT, and structural cardiac abnormalities. While our findings are preliminary, they support a robust albeit modest association between poor olfaction and HF, which warrants further investigation. Our novel observation on HFrEF versus HFpEF deserves attention. HFrEF primarily involves the impaired contraction of the left ventricle, while HFpEF is characterized by diastolic dysfunction of the left ventricle252. Although they have shared risk factors and pathogenesis, HFrEF is more likely to be the consequence of cardiomyocyte loss owing to MI or myocarditis, while HFpEF is more relevant to aging-related inflammation and comorbidities (e.g., diabetes, 94 hypertension, and chronic obstructive pulmonary disease)253. Interestingly, we observed that poor olfaction was evidently associated with higher risk of HFrEF, showing a similar pattern to our independent investigations of the association between poor olfaction and CHD, a primary risk factor for HFrEF (Chapter 6). In contrast, we found little evidence for an association between poor olfaction and HFpEF. This is puzzling because frailty may be more relevant to HFpEF than HFrEF115, and poor olfaction is strongly linked to frailty117. Nevertheless, these observations are preliminary and should be further evaluated in future mechanistic studies. Strengths of this study included the broad representation of community-based US older adults, meticulously adjudicated HF hospitalizations, information on HF biomarkers, and comprehensive statistical analyses. However, this study also has several limitations. First, despite the large sample size and community representation, our findings may not be able to generalize to populations with other demographics, for example, younger populations, Asians, or Hispanics. Second, we only observed a modest association of poor olfaction with HF, which could potentially be explained by residual confounding. However, we have adjusted for a comprehensive list of potential confounders, and the findings appear to be robust within this study and consistent with the previous investigation in the Health ABC Study. Third, despite growing evidence on the relationships between frailty and HF and between olfaction and frailty, the connection and mechanisms between poor olfaction and HF remain largely unexplored. To address this gap, our study was the first to examine the association of olfaction with blood-based and ECG-based subclinical biomarkers for HF. However, since the association was cross-sectional, the longitudinal dynamics between olfaction and these biomarkers is unclear. In conclusion, among community-dwelling US older adults, we found that poor olfaction identified by a single smell test was associated with modestly higher risk of HF, especially HFrEF. 95 Future studies should confirm our findings and further investigate the underlying mechanisms. 96 CHAPTER 8: DISCUSSION 8.1 Summary of Findings In this project, we leveraged two large community-based cohorts of older adults in the US to comprehensively investigate the associations of poor olfaction with the risk of three major cardiovascular outcomes, including stroke, CHD, and HF. In the Health ABC Study, we found that poor olfaction measured by a single smell test was modestly associated with higher cause-specific hazard of CHF for up to 12 years of follow-up. This association was more evident among participants who reported very good to excellent health and was robust across subgroups of age, sex, race, and prevalent CHD/stroke. However, we did not observe a statistically significant association of poor olfaction with incident CHD or stroke. With these preliminary findings, we further investigated olfactory status in relation to each of these cardiovascular outcomes in detail in the ARIC Study. Notably, the ARIC Study was designed specifically to investigate risk factors for atherosclerosis and cardiovascular research with over 30-year continuous contributions to the field. In ARIC, we found an evident association of olfaction with stroke throughout the follow-up, albeit the strength of the association modestly attenuated after year 6. Notably, the magnitude of the association was comparable to established stroke risk factors, such as CHD and atrial fibrillation. A similar finding was observed for CHD, although the association lost its statistical significance after year 6. Finally, we found poor olfaction was associated with a modest risk for incident HF, a finding consistent with that from the Health ABC Study. Further analyses revealed that the association was largely limited to HFrEF. In support of this finding, we found poor olfaction was associated with higher median levels of HF biomarkers of NT-proBNP and hs-cTnT, and a higher odds of structural heart disease among older adults without clinical HF. 97 In summary, the ARIC Study confirmed our preliminary finding in the Health ABC Study about the association of poor olfaction with HF but showed different results on stroke and CHD. Potential explanations for these different findings are not clear. While these two studies had similar study designs, population demographics (i.e., sex, race, and mean age at the smell test), and data collection strategies, there are important differences between the two cohorts. First, the ARIC Study was originally designed to study cardiovascular health and has presumably more rigorous assessments of cardiovascular outcomes, biomarkers and covariates. In contrast, the Health ABC Study was designed to assess how body composition changes in the context of aging, with research focusing on body composition and functional outcomes. Second, there are minor yet important differences between these two study populations. The Health ABC Study recruited well- functioning older adults with a narrow age range (ages 70-79) at enrollment. In comparison, the ARIC Study had a much wider age range at baseline (ages 65-90) with no selection of health or functional status. Nevertheless, whether these differences could explain the differential findings on stroke and CHD is unclear, highlighting the importance of further investigations in other aging cohorts. 8.2. Summary of Limitations In this project, we leveraged extensive data from two large well-established cohorts of US older adults to investigate the associations of poor olfaction with major adverse cardiovascular outcomes among older adults. We conducted comprehensive statistical analyses and carefully accounted for a wide range of covariates and the competing risk of death. However, this project has several notable limitations. First, in both cohorts, the sense of smell was tested in participants with an average age of 75.5 years. Therefore, our findings may not be generalizable to younger populations. Further, our participants were exclusively White and Black individuals from the US, 98 limiting generalizability to populations of other races, ethnicities, and regions. Second, olfaction declines fast with advanced age, so it will be interesting to investigate how olfactory change may be related to future cardiovascular events. Although the ARIC Study tested participants olfaction again at Visit 6, we did not perform analyses due to a high attrition rate (over 40%) and a limited number of incident cases afterwards given the advanced age of our study population. We argue that this should be investigated in relatively young populations where olfactory loss begins to accelerate, informing whether olfactory decline could be an early marker of adverse cardiovascular events in older adults. In the next section (8.3.1), we will briefly present key methodological considerations for investigating time-varying olfactory function in the time-to-event analyses of olfaction and cardiovascular outcomes. Third, despite our comprehensive statistical analyses and relatively large sample size in both cohorts, our findings, even the consistent observation on olfaction and HF, were subject to chance and residual confounding. Fourth, the cross-sectional analyses on poor olfaction and pre-HF markers in the ARIC Study are preliminary, awaiting future longitudinal analysis to examine the temporality of this relationship. Finally, while we found statistical associations of poor olfaction with future risks of major CVDs in one or two cohorts, their clinical implication and underlying mechanism remains elusive, awaiting future investigation. 8.3 Future Directions 8.3.1 Mechanistic Investigations on the Associations The relationships between poor olfaction and cardiovascular health can be dynamic and complex in the life course. Most poor olfaction in older adults is idiopathic and emerges with advanced age2,20. This sensory loss itself, however, may be the consequence of lifelong physiological and pathological changes with age, including those due to metabolic disorders and other cardiovascular 99 risk factors. It may also be attributed to exposure to adverse environmental hazards, such as air pollutants and viral infections, as the peripheral olfactory system is directly exposed to the external environment265–267. Given that age, sex, and genetic variations together may only explain ~10-20% of smell perception variations268, it is essential to examine whether and how the exterior and interior pressures lead to olfactory loss in later life. Accumulating evidence, including that from the current project, implicates that olfactory dysfunction may have profound implications on the health of older adults. It therefore necessitates subsequent research on potential biological mechanisms. It is possible that poor olfaction may be a signal of accelerated aging across multiple physiological systems. In support, poor olfaction has been robustly linked to declines in physical, cognitive, and mental functioning52,105,106. As such, this sensory loss may also be a marker of the aging of the cardiovascular system in older adults. On the other hand, despite limited evidence, it is also plausible that poor olfaction may lead to poor dietary intake. This may further interact with mental and functional declines in the context of aging and accelerate a cascade of adverse health outcomes, including CVDs. While the investigation of the interplays of these possibilities will be challenging, thorough investigations of the role of poor olfaction in older adults will critically inform the maintenance of cardiovascular and overall health of older adults. 8.3.2 Olfactory Change with Incident Cardiovascular Disease As human olfaction starts to decline appreciably after age 50, it will be interesting to investigate if and how olfactory change is associated with the risk of cardiovascular disease among younger older adults. However, such investigations will require additional methodological considerations beyond what we did in the current project, for example, attrition during the follow-up and time- varying confounding. We will briefly describe the statistical issues and the potential solutions 100 using the ARIC Study as an example. In the ARIC Study, the first smell testing was conducted at Visit 5 (2010-2013, ages 75.6±5.2 years), and the second smell testing was conducted at Visit 6 (2016-2017, ages 79.5±4.7 years), with an average 4.9 years apart. Figure 8.1 shows the DAG incorporating the time-varying exposure, time-varying confounders, and missingness at Visit 6 (including both attrition at Visit 6 and missing measurements at Visit 6). The existence of U0 and U1 suggests that the censoring during the follow-up was not complete at random. The existence of the arrow from Y to Mis1 suggests that the missingness of olfactory status and other covariates at visit 6 was not complete at random. Therefore, we need to use statistical methodologies to mitigate the biases due to missingness and attrition. Of note, in this DAG, we did not include the competing risk of death for simplicity, as the total association (detailed in Chapter 3.5) does not require additional assumptions. Figure 8.1 The directed acyclic graph for repeated smell testing and incident stroke. Olf0 and Olf1 is the olfactory assessment at Visit 5 and Visit 6, respectively. C1 is censoring between Visit 5 and Visit 6; C2 is censoring after Visit 6. Y1 is incident stroke between Visit 5 and Visit 6; Y2 is incident stroke after Visit 6. U0 is unmeasured confounding at Visit 5 between censoring and stroke events; U1 stands for unmeasured confounding at Visit 6. L0 and L1 are known confounders at Visit 5 and at Visit 6. Mis1 stands for the missingness of olfaction or other covariates at Visit 6 101 Among 5,799 participants who were free of stroke (as an example) at Visit 5, we will have to delete those who died before/at Visit 6. Our final study population, therefore, will include 5,169 participants. This exclusion is critical because the subsequent multiple imputation for missingness in Visit 6 would be meaningless for those individuals who died before Visit 6. Of the 5,169 surviving participants, 2,712 (52.5%) had at least one missing value in covariates or olfactory status at Visit 6. In the multiple imputation, we will use the random forest method with 50 imputations and 100 iterations per imputation269, including three sets of covariates. The first set of variables are all the variables used in the primary analysis, including time-fixed and time-varying covariates, olfactory score and category at Visits 5 and 6, the indicators and the Nelson-Aalen estimates of the cumulative hazard to the survival time for incident stroke and the competing event of death during the follow-up270. The second set of covariates are potential variables related to non-response, including household income at Visit 1, and dementia status and depressive symptoms at Visit 5. The third set of covariates include those that could explain a considerable amount of variance in smell testing scores, such as the interval between Visit 5 and Visit 6 (i.e., the date of Visit 6 or the estimated median date of Visit 6), and cognitive function at Visit 5. While imputation relies on untestable assumptions, some graphs, e.g., the convergence plot and density plots of the variable distribution before and after the imputation, may assist in diagnosing the imputation. Similar analyses will be performed for all 50 imputed complete datasets. We will use the marginal structural model with IPW to address issues of treatment-confounder feedback and censoring at random. We will then estimate the marginal absolute risk across groups of olfactory changes and calculate the risk ratio with the reference level of constant good olfaction. The 95% CI was estimated by using bootstrapping. 102 Last, we will use Rubin’s Rule to pool the results from all the imputed datasets271. Point estimate is 𝑄̅ = 1 𝑚 𝑚 ∑ 𝑄̂𝑙 𝑙=1 , where 𝑄̂𝑙 is the estimated RR from each imputed dataset, m=50. The total variance 𝑇 comes from three sources: 𝑇 = 𝑈 + 𝐵 + 𝐵/𝑚 1) 𝑈: conventional statistical measure of variability, as we include a sample from a population; 2) 𝐵: extra variance because of missingness in the sample; 3) 𝐵/𝑚: the extra simulation variance as 𝑄 is estimated from a finite 𝑚. If the target association measure is RR. There are two ways to obtain the right pooled RR and its 95% CIs through bootstrapping. The first approach is to output log(𝑅1) − log⁡(𝑅2) after bootstrapping. Accordingly, pooled point estimate of log(𝑅1/𝑅2) and its 95% CI can be calculated based on the Rubin’s Rule and further transferred to RR. However, when the absolute risk is very low, the estimate of RR from the first approach can be inflated, thus the second approach may be preferable. Instead of directly outputting results for log(𝑅1) − log⁡(𝑅2), we can derive the point estimate and variance of log(𝑅1) − log⁡(𝑅2) using the delta method272 from the bootstrapping output for 𝑅1 and 𝑅2. 8.3.3 Incorporating Frailty into the Investigation Frailty is an increasingly appreciated geriatric clinical construct to characterize decreased physiological reserves and increasing vulnerability to adverse health consequences in the presence of stressors108. In older adults, frailty is highly predictive of morbidity, loss of independence, hospitalization, and mortality273,274. Interestingly, recent studies have robustly linked poor olfaction to frailty117 and frailty to cardiovascular health114–116. However, to our knowledge, no 103 study has explored the potential interplays between these two ageing phenotypes in the context of cardiovascular health. Such investigation may improve our understanding of both phenotypes and their relevance to cardiovascular health, identifying novel approaches to improve the health and quality of life of older adults. 8.4 Conclusions Using data from two well-established US cohorts of older adults, we found preliminary evidence that poor olfaction assessed by a single smell test is associated with the risks of major adverse cardiovascular outcomes. The data from both cohorts are mostly consistent for HF, supported by analysis involving subclinical cardiovascular biomarkers. The association of poor olfaction with stroke and CHD are only found in the ARIC Study but not the Health ABC Study. Nevertheless, the findings are provocative and deserve independent investigations. To our knowledge, this project is the first comprehensive investigation on olfaction and cardiovascular health in older adults. Poor olfaction is common in older adults but has long been overlooked by the medical community and the public. While the COVID-19 pandemic has suddenly brought this sensory deficit to people’s attention, we are far from understanding how it may affect human health, particularly in older adults. We expect my dissertation work, together with emerging findings on poor olfaction and a broad range of adverse outcomes, will fuel the research interest to unveil the potentially profound implications of olfaction on the health of older adults. 104 BIBLIOGRAPHY 1. Chamberlin KW, Yuan Y, Li C, et al. Olfactory Impairment and the Risk of Major Adverse Cardiovascular Outcomes in Older Adults. J Am Heart Assoc. Published online June 7, 2024:e033320. doi:10.1161/JAHA.123.033320 2. Murphy C, Schubert CR, Cruickshanks KJ, Klein BEK, Klein R, Nondahl DM. Prevalence of Olfactory Impairment in Older Adults. JAMA. 2002;288(18):2307-2312. doi:10.1001/jama.288.18.2307 3. Pinto JM, Wroblewski KE, Kern DW, Schumm LP, McClintock MK. The Rate of Age- Related Olfactory Decline Among the General Population of Older U.S. Adults. J Gerontol A Biol Sci Med Sci. 2015;70(11):1435-1441. doi:10.1093/gerona/glv072 4. Bhatia-Dey N, Heinbockel T. The Olfactory System as Marker of Neurodegeneration in Aging, Neurological and Neuropsychiatric Disorders. Int J Environ Res Public Health. 2021;18(13):6976. doi:10.3390/ijerph18136976 5. Choi JS, Jang SS, Kim J, Hur K, Ference E, Wrobel B. Association Between Olfactory Dysfunction and Mortality in US Adults. JAMA Otolaryngol Head Neck Surg. 2021;147(1):49-55. doi:10.1001/jamaoto.2020.3502 6. Naghavi M, Ong KL, Aali A, et al. Global Burden of 288 Causes of Death and Life Expectancy Decomposition in 204 Countries and Territories and 811 Subnational Locations, 1990–2021: a Systematic Analysis for the Global Burden of Disease Study 2021. The Lancet. 2024;0(0). doi:10.1016/S0140-6736(24)00367-2 7. McGann JP. Poor Human Olfaction is a 19th-century Myth. Science. 2017;356(6338):eaam7263. doi:10.1126/science.aam7263 8. Bushdid C, Magnasco MO, Vosshall LB, Keller A. Humans can Discriminate more than One Trillion Olfactory Stimuli. Science. 2014;343(6177):1370-1372. doi:10.1126/science.1249168 9. Doty RL, Kamath V. The Influences of Age on Olfaction: a Review. Front Psychol. 2014;5:20. doi:10.3389/fpsyg.2014.00020 10. Branigan B, Tadi P. Physiology, Olfactory. In: StatPearls. StatPearls Publishing; 2022. Accessed December 16, 2022. http://www.ncbi.nlm.nih.gov/books/NBK542239/ 11. Cleland TA, Linster C. Chapter 6 - Central olfactory structures. In: Doty RL, ed. Handbook of Clinical Neurology. Vol 164. Smell and Taste. Elsevier; 2019:79-96. doi:10.1016/B978- 0-444-63855-7.00006-X 12. Cenquizca LA, Swanson LW. Spatial Organization of Direct Hippocampal Field CA1 Axonal Projections to the Rest of the Cerebral Cortex. Brain Research Reviews. 2007;56(1):1-26. doi:10.1016/j.brainresrev.2007.05.002 105 13. Barbas H. Specialized Elements of Orbitofrontal Cortex in Primates. Annals of the New York Academy of Sciences. 2007;1121(1):10-32. doi:10.1196/annals.1401.015 14. Šimić G, Tkalčić M, Vukić V, et al. Understanding Emotions: Origins and Roles of the Amygdala. Biomolecules. 2021;11(6):823. doi:10.3390/biom11060823 15. Bird CM, Burgess N. The Hippocampus and Memory: Insights from Spatial Processing. Nat Rev Neurosci. 2008;9(3):182-194. doi:10.1038/nrn2335 16. Kringelbach ML. The Human Orbitofrontal Cortex: Linking Reward to Hedonic Experience. Nat Rev Neurosci. 2005;6(9):691-702. doi:10.1038/nrn1747 17. Hendrix P, Griessenauer CJ, Foreman P, Shoja MM, Loukas M, Tubbs RS. Arterial Supply of the Upper Cranial Nerves: A Comprehensive Review. Clinical Anatomy. 2014;27(8):1159-1166. doi:10.1002/ca.22415 18. Purves D, Augustine GJ, Fitzpatrick D, et al. The Blood Supply of the Brain and Spinal Cord. In: Neuroscience. 2nd Edition. Sinauer Associates; 2001. Accessed April 7, 2024. https://www.ncbi.nlm.nih.gov/books/NBK11042/ 19. AbuHasan Q, Reddy V, Siddiqui W. Neuroanatomy, Amygdala. In: StatPearls. StatPearls Publishing; 2024. Accessed April 7, 2024. http://www.ncbi.nlm.nih.gov/books/NBK537102/ 20. Hummel T, Whitcroft KL, Andrews P, et al. Position Paper on Olfactory Dysfunction. Rhinology. 2016;56(1):1-30. doi:10.4193/Rhino16.248 21. Gudis DA, Soler ZM. Chronic Rhinosinusitis-Related Smell Loss: Medical And Surgical Treatment Efficacy. Curr Otorhinolaryngol Rep. 2016;4(2):142-147. doi:10.1007/s40136- 016-0114-4 22. Schneider ALC, Gottesman RF, Mosley TH, et al. Associations of Prior Head Injury With Olfaction in Older Adults: Results From the Atherosclerosis Risk in Communities (ARIC) Study. JAMA Otolaryngology–Head & Neck Surgery. 2022;148(9):840-848. doi:10.1001/jamaoto.2022.1920 23. Dong J, Pinto JM, Guo X, et al. The Prevalence of Anosmia and Associated Factors Among U.S. Black and White Older Adults. J Gerontol A Biol Sci Med Sci. 2017;72(8):1080-1086. doi:10.1093/gerona/glx081 24. Cao Z, Yang A, D’Aloisio AA, et al. Assessment of Self-reported Sense of Smell, Objective Testing, and Associated Factors in Middle-aged and Older Women. JAMA Otolaryngol Head Neck Surg. 2022;148(5):408-417. doi:10.1001/jamaoto.2022.0069 25. Whitcroft KL, Hummel T. Clinical Diagnosis and Current Management Strategies for Olfactory Dysfunction: A Review. JAMA Otolaryngology–Head & Neck Surgery. 2019;145(9):846-853. doi:10.1001/jamaoto.2019.1728 106 26. Han P, Musch M, Abolmaali N, Hummel T. Improved Odor Identification Ability and Increased Regional Gray Matter Volume After Olfactory Training in Patients With Idiopathic Olfactory Loss. Iperception. 2021;12(2):20416695211005811. doi:10.1177/20416695211005811 27. Lu R, Aziz NA, Reuter M, Stöcker T, Breteler MMB. Evaluation of the Neuroanatomical Basis of Olfactory Dysfunction in the General Population. JAMA Otolaryngology–Head & Neck Surgery. 2021;147(10):855-863. doi:10.1001/jamaoto.2021.2026 28. Wickenden AD. Overview of Electrophysiological Techniques. Curr Protoc Pharmacol. 2014;64:11.1.1-17. doi:10.1002/0471141755.ph1101s64 29. Lötsch J, Hummel T. The Clinical Significance of Electrophysiological Measures of Olfactory Function. Behavioural Brain Research. 2006;170(1):78-83. doi:10.1016/j.bbr.2006.02.013 30. Desiato VM, Levy DA, Byun YJ, Nguyen SA, Soler ZM, Schlosser RJ. The Prevalence of Olfactory Dysfunction in the General Population: A Systematic Review and Meta-analysis. Am J Rhinol Allergy. 2021;35(2):195-205. doi:10.1177/1945892420946254 31. Liu G, Zong G, Doty RL, Sun Q. Prevalence and Risk Factors of Taste and Smell Impairment in a Nationwide Representative Sample of the US Population: a Cross-sectional Study. BMJ Open. 2016;6(11):e013246. doi:10.1136/bmjopen-2016-013246 32. Jones RE, Brown CC, Ship JA. Odor identification in young and elderly African‐Americans and Caucasians. Special Care in Dentistry. 1995;15(4):138-143. doi:10.1111/j.1754- 4505.1995.tb00501.x 33. Ship JA, Pearson JD, Cruise LJ, Brant LJ, Metter EJ. Longitudinal Changes in Smell Identification. The Journals of Gerontology: Series A. 1996;51A(2):M86-M91. doi:10.1093/gerona/51A.2.M86 34. Hedner M, Nilsson LG, Olofsson JK, et al. Age-Related Olfactory Decline is Associated with the BDNF Val66met Polymorphism: Evidence from a Population-Based Study. Front Aging Neurosci. 2010;2:24. doi:10.3389/fnagi.2010.00024 35. Shrestha S, Zhu X, Kamath V, et al. Factors Associated with Poor Olfaction and Olfactory Decline in Older Adults in the ARIC Neurocognitive Study. Nutrients. 2023;15(16):3641. doi:10.3390/nu15163641 36. Schubert CR, Cruickshanks KJ, Klein BEK, Klein R, Nondahl DM. Olfactory Impairment in Older Adults: Five-year Incidence and Risk Factors. The Laryngoscope. 2011;121(4):873-878. doi:10.1002/lary.21416 37. Henkin RI, Larson AL, Powell RD. Hypogeusia, Dysgeusia, Hyposmia, and Dysosmia Following Influenza-like Infection. Ann Otol Rhinol Laryngol. 1975;84(5 Pt 1):672-682. doi:10.1177/000348947508400519 107 38. Spinato G, Fabbris C, Polesel J, et al. Alterations in Smell or Taste in Mildly Symptomatic Outpatients With SARS-CoV-2 Infection. JAMA. 2020;323(20):2089-2090. doi:10.1001/jama.2020.6771 39. Zazhytska M, Kodra A, Hoagland DA, et al. Non-cell-autonomous Disruption of Nuclear Architecture as a Potential Cause of COVID-19-induced Anosmia. Cell. 2022;185(6):1052- 1064.e12. doi:10.1016/j.cell.2022.01.024 40. Tan BKJ, Han R, Zhao JJ, et al. Prognosis and Persistence of Smell and Taste Dysfunction in Patients with Covid-19: Meta-analysis with Parametric Cure Modelling of Recovery Curves. BMJ. 2022;378:e069503. doi:10.1136/bmj-2021-069503 41. Rombaux P, Potier H, Bertrand B, Duprez T, Hummel T. Olfactory Bulb Volume in Patients with Sinonasal Disease. Am J Rhinol. 2008;22(6):598-601. doi:10.2500/ajr.2008.22.3237 42. Lane AP, Turner J, May L, Reed R. A Genetic Model of Chronic Rhinosinusitis-associated Olfactory Inflammation Reveals Reversible Functional Impairment and Dramatic Neuroepithelial Reorganization. J Neurosci. 2010;30(6):2324-2329. doi:10.1523/JNEUROSCI.4507-09.2010 43. Doty RL. Clinical Studies of Olfaction. Chemical Senses. 2005;30(suppl_1):i207-i209. doi:10.1093/chemse/bjh187 44. Mueller CA, Hummel T. Recovery of olfactory function after nine years of post-traumatic anosmia: a case report. J Med Case Reports. 2009;3:9283. doi:10.4076/1752-1947-3-9283 45. Murphy C. Olfactory and Other Sensory Impairments in Alzheimer Disease. Nat Rev Neurol. 2019;15(1):11-24. doi:10.1038/s41582-018-0097-5 46. Postuma RB, Berg D. Advances in Markers of Prodromal Parkinson Disease. Nat Rev Neurol. 2016;12(11):622-634. doi:10.1038/nrneurol.2016.152 47. Braak H, Tredici KD, Rüb U, de Vos RAI, Jansen Steur ENH, Braak E. Staging of Brain Pathology Related to Sporadic Parkinson’s Disease. Neurobiology of Aging. 2003;24(2):197-211. doi:10.1016/S0197-4580(02)00065-9 48. Braak H, Alafuzoff I, Arzberger T, Kretzschmar H, Del Tredici K. Staging of Alzheimer Disease-associated Neurofibrillary Pathology Using Paraffin Sections and Immunocytochemistry. Acta Neuropathol. 2006;112(4):389-404. doi:10.1007/s00401-006- 0127-z 49. Chen H, Wang K, Scheperjans F, Killinger B. Environmental Triggers of Parkinson’s Disease – Implications of the Braak and Dual-hit Hypotheses. Neurobiology of Disease. 2022;163:105601. doi:10.1016/j.nbd.2021.105601 50. Roberts RO, Christianson TJH, Kremers WK, et al. Association Between Olfactory Dysfunction and Amnestic Mild Cognitive Impairment and Alzheimer Disease Dementia. JAMA Neurology. 2016;73(1):93-101. doi:10.1001/jamaneurol.2015.2952 108 51. Adams DR, Kern DW, Wroblewski KE, McClintock MK, Dale W, Pinto JM. Olfactory Dysfunction Predicts Subsequent Dementia in Older US Adults. J Am Geriatr Soc. 2018;66(1):140-144. doi:10.1111/jgs.15048 52. Yaffe K, Freimer D, Chen H, et al. Olfaction and Risk of Dementia in a Biracial Cohort of Older Adults. Neurology. 2017;88(5):456-462. doi:10.1212/WNL.0000000000003558 53. Chen H, Shrestha S, Huang X, et al. Olfaction and incident Parkinson disease in US white and black older adults. Neurology. 2017;89(14):1441-1447. doi:10.1212/WNL.0000000000004382 54. Schwob JE. Neural Regeneration and the Peripheral Olfactory System. The Anatomical Record. 2002;269(1):33-49. doi:10.1002/ar.10047 55. Kase Y, Shimazaki T, Okano H. Current Understanding of Adult Neurogenesis in the Mammalian Brain: How does Adult Neurogenesis Decrease with Age? Inflammation and Regeneration. 2020;40(1):10. doi:10.1186/s41232-020-00122-x 56. Croy I, Nordin S, Hummel T. Olfactory Disorders and Quality of Life--an Updated Review. Chem Senses. 2014;39(3):185-194. doi:10.1093/chemse/bjt072 57. Liu B, Luo Z, Pinto JM, et al. Relationship Between Poor Olfaction and Mortality Among Community-Dwelling Older Adults: A Cohort Study. Ann Intern Med. 2019;170(10):673- 681. doi:10.7326/M18-0775 58. Tsao CW, Aday AW, Almarzooq ZI, et al. Heart Disease and Stroke Statistics-2023 Update: A Report From the American Heart Association. Circulation. 2023;147(8):e93-e621. doi:10.1161/CIR.0000000000001123 59. Martin SS, Aday AW, Almarzooq ZI, et al. 2024 Heart Disease and Stroke Statistics: A Report of US and Global Data From the American Heart Association. Circulation. 2024;149(8):e347-e913. doi:10.1161/CIR.0000000000001209 60. Bancks MP, Ning H, Allen NB, et al. Long-term Absolute Risk for Cardiovascular Disease Stratified by Fasting Glucose Level. Diabetes Care. 2019;42(3):457-465. doi:10.2337/dc18-1773 61. Frąk W, Wojtasińska A, Lisińska W, Młynarska E, Franczyk B, Rysz J. Pathophysiology of Cardiovascular Diseases: New Insights into Molecular Mechanisms of Atherosclerosis, Arterial Hypertension, and Coronary Artery Disease. Biomedicines. 2022;10(8):1938. doi:10.3390/biomedicines10081938 62. Arnett DK, Blumenthal RS, Albert MA, et al. 2019 ACC/AHA Guideline on the Primary Prevention of Cardiovascular Disease: A Report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines. Circulation. 2019;140(11). doi:10.1161/CIR.0000000000000678 109 63. Goff DC, Lloyd-Jones DM, Bennett G, et al. 2013 ACC/AHA Guideline on the Assessment of Cardiovascular Risk. Circulation. 2014;129(25_suppl_2):S49-S73. doi:10.1161/01.cir.0000437741.48606.98 64. Smith SC, Benjamin EJ, Bonow RO, et al. AHA/ACCF Secondary Prevention and Risk Reduction Therapy for Patients With Coronary and Other Atherosclerotic Vascular Disease: 2011 Update: A Guideline From the American Heart Association and American College of Cardiology Foundation. Circulation. 2011;124(22):2458-2473. doi:10.1161/CIR.0b013e318235eb4d 65. Libby P, Theroux P. Pathophysiology of Coronary Artery Disease. Circulation. 2005;111(25):3481-3488. doi:10.1161/CIRCULATIONAHA.105.537878 66. Heidenreich PA, Bozkurt B, Aguilar D, et al. 2022 AHA/ACC/HFSA Guideline for the Management of Heart Failure. Journal of the American College of Cardiology. 2022;79(17):e263-e421. doi:10.1016/j.jacc.2021.12.012 67. Kleindorfer DO, Towfighi A, Chaturvedi S, et al. 2021 Guideline for the Prevention of Stroke in Patients With Stroke and Transient Ischemic Attack: A Guideline From the American Heart Association/American Stroke Association. Stroke. 2021;52(7):e364-e467. doi:10.1161/STR.0000000000000375 68. Kolominsky-Rabas PL, Weber M, Gefeller O, Neundoerfer B, Heuschmann PU. Epidemiology of Ischemic Stroke Subtypes According to TOAST Criteria: Incidence, Recurrence, and Long-term Survival in Ischemic Stroke Subtypes: a Population-based Study. Stroke. 2001;32(12):2735-2740. doi:10.1161/hs1201.100209 69. Gardener H, Sacco RL, Rundek T, Battistella V, Cheung YK, Elkind MSV. Race and Ethnic Disparities in Stroke Incidence in the Northern Manhattan Study. Stroke. 2020;51(4):1064- 1069. doi:10.1161/STROKEAHA.119.028806 70. Adams HP, Bendixen BH, Kappelle LJ, et al. Classification of Subtype of Acute Ischemic Stroke. Definitions for Use in a Multicenter Clinical Trial. TOAST. Trial of Org 10172 in Acute Stroke Treatment. Stroke. 1993;24(1):35-41. doi:10.1161/01.STR.24.1.35 71. GBD 2019 Stroke Collaborators. Global, Regional, and National Burden of Stroke and its Risk Factors, 1990-2019: a Systematic Analysis for the Global Burden of Disease Study 2019. Lancet Neurol. 2021;20(10):795-820. doi:10.1016/S1474-4422(21)00252-0 72. Koton S, Sang Y, Schneider ALC, Rosamond WD, Gottesman RF, Coresh J. Trends in Stroke Incidence Rates in Older US Adults: An Update From the Atherosclerosis Risk in Communities (ARIC) Cohort Study. JAMA Neurol. 2020;77(1):109-113. doi:10.1001/jamaneurol.2019.3258 73. Pendlebury ST, Rothwell PM. Incidence and Prevalence of Dementia Associated with Transient Ischaemic Attack and Stroke: Analysis of the Population-based Oxford Vascular Study. Lancet Neurol. 2019;18(3):248-258. doi:10.1016/S1474-4422(18)30442-3 110 74. Cassar A, Holmes DR, Rihal CS, Gersh BJ. Chronic Coronary Artery Disease: Diagnosis and Management. Mayo Clin Proc. 2009;84(12):1130-1146. doi:10.4065/mcp.2009.0391 75. Dalen JE, Alpert JS, Goldberg RJ, Weinstein RS. The Epidemic of the 20th Century: Coronary Heart Disease. The American Journal of Medicine. 2014;127(9):807-812. doi:10.1016/j.amjmed.2014.04.015 76. Proceedings of the Conference on the Decline in Coronary Heart Disease Mortality. Department of Health, Education, and Welfare, Public Health Service, National Instiutes of Health; 1979. 77. Rosamond WD, Chambless LE, Folsom AR, et al. Trends in the Incidence of Myocardial Infarction and in Mortality due to Coronary Heart Disease, 1987 to 1994. N Engl J Med. 1998;339(13):861-867. doi:10.1056/NEJM199809243391301 78. Tunstall-Pedoe H, Kuulasmaa K, Mähönen M, Tolonen H, Ruokokoski E. Contribution of Trends in Survival and Coronary-event Rates to Changes in Coronary Heart Disease Mortality: 10-year Results from 37 WHO MONICA Project Populations. The Lancet. 1999;353(9164):1547-1557. doi:10.1016/S0140-6736(99)04021-0 79. Centers for Disease Control and Prevention. Underlying Cause of Death, 2018-2022, Single Race Results Form. April 26, 2024. Accessed September 2, 2024. https://wonder.cdc.gov/controller/datarequest/D158;jsessionid=32224B8149E49A8DD4837 832CF91 80. Ibrahim NE, Januzzi JL. Established and Emerging Roles of Biomarkers in Heart Failure. Circ Res. 2018;123(5):614-629. doi:10.1161/CIRCRESAHA.118.312706 81. Shah AM, Claggett B, Loehr LR, et al. Heart Failure Stages Among Older Adults in the Community: The Atherosclerosis Risk in Communities Study. Circulation. 2017;135(3):224-240. doi:10.1161/CIRCULATIONAHA.116.023361 82. Kosaraju A, Goyal A, Grigorova Y, Makaryus AN. Left Ventricular Ejection Fraction. In: StatPearls. StatPearls Publishing; 2024. Accessed May 15, 2024. http://www.ncbi.nlm.nih.gov/books/NBK459131/ 83. Vedin O, Lam CSP, Koh AS, et al. Significance of Ischemic Heart Disease in Patients With Heart Failure and Preserved, Midrange, and Reduced Ejection Fraction: A Nationwide Cohort Study. Circ Heart Fail. 2017;10(6):e003875. doi:10.1161/CIRCHEARTFAILURE.117.003875 84. Koh AS, Tay WT, Teng THK, et al. A Comprehensive Population-based Characterization of Heart Failure with Mid-range Ejection Fraction. Eur J Heart Fail. 2017;19(12):1624-1634. doi:10.1002/ejhf.945 85. James SL, Abate D, Abate KH, et al. Global, Regional, and National Incidence, Prevalence, and Years Lived with Disability for 354 Diseases and Injuries for 195 Countries and 111 Territories, 1990–2017: a Systematic Analysis for the Global Burden of Disease Study 2017. The Lancet. 2018;392(10159):1789-1858. doi:10.1016/S0140-6736(18)32279-7 86. Savarese G, Becher PM, Lund LH, Seferovic P, Rosano GMC, Coats AJS. Global Burden of Heart Failure: a Comprehensive and Updated Review of Epidemiology. Cardiovasc Res. 2023;118(17):3272-3287. doi:10.1093/cvr/cvac013 87. Huffman MD, Berry JD, Ning H, et al. Lifetime Risk for Heart Failure among White and Black Americans: Cardiovascular Lifetime Risk Pooling Project. J Am Coll Cardiol. 2013;61(14):1510-1517. doi:10.1016/j.jacc.2013.01.022 88. Leon MB, Smith CR, Mack MJ, et al. Transcatheter or Surgical Aortic-Valve Replacement in Intermediate-Risk Patients. N Engl J Med. 2016;374(17):1609-1620. doi:10.1056/NEJMoa1514616 89. Heidenreich PA, Fonarow GC, Opsha Y, Sandhu AT, Sweitzer NK, Warraich HJ. Economic Issues in Heart Failure in the United States. J Card Fail. 2022;28(3):453-466. doi:10.1016/j.cardfail.2021.12.017 90. Schubert CR, Cruickshanks KJ, Fischer ME, et al. Carotid Intima Media Thickness, Atherosclerosis, and 5-Year Decline in Odor Identification: The Beaver Dam Offspring Study. J Gerontol A Biol Sci Med Sci. 2015;70(7):879-884. doi:10.1093/gerona/glu158 91. Schubert CR, Cruickshanks KJ, Fischer ME, Klein BEK, Klein R, Pinto AA. Inflammatory and Vascular Markers and Olfactory Impairment in Older Adults. Age Ageing. 2015;44(5):878-882. doi:10.1093/ageing/afv075 92. Adderley NJ, Subramanian A, Nirantharakumar K, et al. Association Between Idiopathic Intracranial Hypertension and Risk of Cardiovascular Diseases in Women in the United Kingdom. JAMA Neurology. 2019;76(9):1088-1098. doi:10.1001/jamaneurol.2019.1812 93. Kunte H, Schmidt F, Kronenberg G, et al. Olfactory Dysfunction in Patients with Idiopathic Intracranial Hypertension. Neurology. 2013;81(4):379-382. doi:10.1212/WNL.0b013e31829c5c9d 94. Schmidt C, Wiener E, Hoffmann J, et al. Structural Olfactory Nerve Changes in Patients Suffering from Idiopathic Intracranial Hypertension. PLOS ONE. 2012;7(4):e35221. doi:10.1371/journal.pone.0035221 95. Landis BN, Frasnelli J, Reden J, Lacroix JS, Hummel T. Differences Between Orthonasal and Retronasal Olfactory Functions in Patients With Loss of the Sense of Smell. Archives of Otolaryngology–Head & Neck Surgery. 2005;131(11):977-981. doi:10.1001/archotol.131.11.977 96. Riera CE, Tsaousidou E, Halloran J, et al. The Sense of Smell Impacts Metabolic Health and Obesity. Cell Metabolism. 2017;26(1):198-211.e5. doi:10.1016/j.cmet.2017.06.015 112 97. Zhao Y, Bhutani S, Kahnt T. Appetite-regulating Hormones Modulate Odor Perception and Odor-evoked Activity in Hypothalamus and Olfactory Cortices. Chemical Senses. 2023;48:bjad039. doi:10.1093/chemse/bjad039 98. Gopinath B, Russell J, Sue CM, Flood VM, Burlutsky G, Mitchell P. Olfactory Impairment in Older Adults is Associated with Poorer Diet Quality over 5 Years. Eur J Nutr. 2016;55(3):1081-1087. doi:10.1007/s00394-015-0921-2 99. Fluitman KS, Hesp AC, Kaihatu RF, et al. Poor Taste and Smell Are Associated with Poor Appetite, Macronutrient Intake, and Dietary Quality but Not with Undernutrition in Older Adults. J Nutr. 2021;151(3):605-614. doi:10.1093/jn/nxaa400 100. Roxbury CR, Bernstein IA, Lin SY, Rowan NR. Association Between Chemosensory Dysfunction and Diet Quality in United States Adults. Am J Rhinol Allergy. 2022;36(1):47- 56. doi:10.1177/19458924211016611 101. Kong IG, Kim SY, Kim MS, Park B, Kim JH, Choi HG. Olfactory Dysfunction Is Associated with the Intake of Macronutrients in Korean Adults. PLoS One. 2016;11(10):e0164495. doi:10.1371/journal.pone.0164495 102. Lichtenstein AH, Appel LJ, Vadiveloo M, et al. 2021 Dietary Guidance to Improve Cardiovascular Health: A Scientific Statement From the American Heart Association. Circulation. 2021;144(23):e472-e487. doi:10.1161/CIR.0000000000001031 103. Riccardi G, Giosuè A, Calabrese I, Vaccaro O. Dietary Recommendations for Prevention of Atherosclerosis. Cardiovascular Research. 2022;118(5):1188-1204. doi:10.1093/cvr/cvab173 104. Dintica CS, Marseglia A, Rizzuto D, et al. Impaired Olfaction is Associated with Cognitive Decline and Neurodegeneration in the Brain. Neurology. 2019;92(7):e700-e709. doi:10.1212/WNL.0000000000006919 105. Yuan Y, Li C, Luo Z, Simonsick EM, Shiroma EJ, Chen H. Olfaction and Physical Functioning in Older Adults: A Longitudinal Study. J Gerontol A Biol Sci Med Sci. 2022;77(8):1612-1619. doi:10.1093/gerona/glab233 106. Kamath V, Jiang K, Manning KJ, et al. Olfactory Dysfunction and Depression Trajectories in Community-Dwelling Older Adults. The Journals of Gerontology: Series A. 2024;79(1):glad139. doi:10.1093/gerona/glad139 107. Murman DL. The Impact of Age on Cognition. Semin Hear. 2015;36(3):111-121. doi:10.1055/s-0035-1555115 108. Cesari M, Calvani R, Marzetti E. Frailty in Older Persons. Clin Geriatr Med. 2017;33(3):293-303. doi:10.1016/j.cger.2017.02.002 113 109. O’Donnell M, Teo K, Gao P, et al. Cognitive Impairment and Risk of Cardiovascular Events and Mortality. European Heart Journal. 2012;33(14):1777-1786. doi:10.1093/eurheartj/ehs053 110. Harshfield EL, Pennells L, Schwartz JE, et al. Association Between Depressive Symptoms and Incident Cardiovascular Diseases. JAMA. 2020;324(23):2396-2405. doi:10.1001/jama.2020.23068 111. Hu X, Mok Y, Ding N, et al. Physical Function and Subsequent Risk of Cardiovascular Events in Older Adults: The Atherosclerosis Risk in Communities Study. J Am Heart Assoc. 2022;11(17):e025780. doi:10.1161/JAHA.121.025780 112. Xue QL. The Frailty Syndrome: Definition and Natural History. Clin Geriatr Med. 2011;27(1):1-15. doi:10.1016/j.cger.2010.08.009 113. Liu X, Tou NX, Gao Q, Gwee X, Wee SL, Ng TP. Frailty and Risk of Cardiovascular Disease and Mortality. PLoS One. 2022;17(9):e0272527. doi:10.1371/journal.pone.0272527 114. Damluji AA, Chung SE, Xue QL, et al. Frailty and Cardiovascular Outcomes in the National Health and Aging Trends Study. European Heart Journal. 2021;42(37):3856-3865. doi:10.1093/eurheartj/ehab468 115. Ramonfaur D, Buckley LF, Arthur V, et al. High Throughput Plasma Proteomics and Risk of Heart Failure and Frailty in Late Life. JAMA Cardiology. Published online May 29, 2024. doi:10.1001/jamacardio.2024.1178 116. Sesso J, Walston J, Bandeen-Roche K, et al. Association of Cardiovascular Fibrosis, Remodeling, and Dysfunction With Frailty, Prefrailty, and Functional Performance: The Multi-Ethnic Study of Atherosclerosis. J Gerontol A Biol Sci Med Sci. 2024;79(8):glae142. doi:10.1093/gerona/glae142 117. Yeo BSY, Chan JH, Tan BKJ, et al. Olfactory Impairment and Frailty: A Systematic Review and Meta-Analysis. JAMA Otolaryngology–Head & Neck Surgery. Published online July 11, 2024. doi:10.1001/jamaoto.2024.1854 118. Wehling E, Naess H, Wollschlaeger D, et al. Olfactory Dysfunction in Chronic Stroke Patients. BMC Neurol. 2015;15:199. doi:10.1186/s12883-015-0463-5 119. Seubert J, Laukka EJ, Rizzuto D, et al. Prevalence and Correlates of Olfactory Dysfunction in Old Age: A Population-Based Study. J Gerontol A Biol Sci Med Sci. 2017;72(8):1072- 1079. doi:10.1093/gerona/glx054 120. Akşıt E, Çıl ÖÇ. Olfactory Dysfunction in Patients with Ischemic Heart Failure. Acta Cardiol Sin. 2020;36(2):133-139. doi:10.6515/ACS.202003_36(2).20190812B 121. Okamoto K, Shiga H, Nakamura H, Matsui M, Miwa T. Relationship Between Olfactory Disturbance After Acute Ischemic Stroke and Latent Thalamic Hypoperfusion. Chem Senses. 2020;45(2):111-118. doi:10.1093/chemse/bjz077 114 122. Schlosser RJ, Desiato VM, Storck KA, et al. A Community-Based Study on the Prevalence of Olfactory Dysfunction. Am J Rhinol Allergy. 2020;34(5):661-670. doi:10.1177/1945892420922771 123. Roh D, Lee DH, Kim SW, et al. The Association between Olfactory Dysfunction and Cardiovascular Disease and its Risk Factors in Middle-aged and Older Adults. Sci Rep. 2021;11(1):1248. doi:10.1038/s41598-020-80943-5 124. Kültür T, Bayar Muluk N, Inal M, Kömürcü Erkmen SP, Rasulova G. Peripheral and central smell regions in patients with stroke: an MRI evaluation. Neurol Sci. 2022;43(7):4287- 4296. doi:10.1007/s10072-022-05960-w 125. Ekström I, Larsson M, Rizzuto D, Fastbom J, Bäckman L, Laukka EJ. Predictors of Olfactory Decline in Aging: A Longitudinal Population-Based Study. The Journals of Gerontology: Series A. 2020;75(12):2441-2449. doi:10.1093/gerona/glaa221 126. Palmquist E, Larsson M, Olofsson JK, Seubert J, Bäckman L, Laukka EJ. A Prospective Study on Risk Factors for Olfactory Dysfunction in Aging. The Journals of Gerontology: Series A. 2020;75(3):603-610. doi:10.1093/gerona/glz265 127. Siegel JK, Wroblewski KE, McClintock MK, Pinto JM. Olfactory Dysfunction Persists after Smoking Cessation and Signals Increased Cardiovascular Risk. Int Forum Allergy Rhinol. 2019;9(9):977-985. doi:10.1002/alr.22357 128. Health ABC. Accessed December 7, 2022. https://healthabc.nia.nih.gov/ 129. Simonsick EM, Newman AB, Nevitt MC, et al. Measuring Higher Level Physical Function in Well-functioning Older Adults: Expanding Familiar Approaches in the Health ABC Study. J Gerontol A Biol Sci Med Sci. 2001;56(10):M644-649. doi:10.1093/gerona/56.10.m644 130. The ARIC Investigators. The Atherosclerosis Risk in Community (ARIC) Study: Design and Objectives. American Journal of Epidemiology. 1989;129(4):687-702. doi:10.1093/oxfordjournals.aje.a115184 131. Wright JD, Folsom AR, Coresh J, et al. The ARIC (Atherosclerosis Risk In Communities) Study. Journal of the American College of Cardiology. 2021;77(23):2939-2959. doi:10.1016/j.jacc.2021.04.035 132. Doty RL, Marcus A, William Lee W. Development of the 12-Item Cross-Cultural Smell Identification Test(CC-SIT). The Laryngoscope. 1996;106(3):353-356. doi:10.1097/00005537-199603000-00021 133. Hummel T, Konnerth CG, Rosenheim K, Kobal G. Screening of Olfactory Function with a Four-minute Odor Identification Test: Reliability, Normative data, and Investigations in Patients with Olfactory Loss. Ann Otol Rhinol Laryngol. 2001;110(10):976-981. doi:10.1177/000348940111001015 115 134. Wilson RS, Schneider JA, Arnold SE, Tang Y, Boyle PA, Bennett DA. Olfactory Identification and Incidence of Mild Cognitive Impairment in Older Age. Arch Gen Psychiatry. 2007;64(7):802-808. doi:10.1001/archpsyc.64.7.802 135. Kjelvik G, Sando SB, Aasly J, Engedal KA, White LR. Use of the Brief Smell Identification Test for olfactory deficit in a Norwegian population with Alzheimer’s disease. Int J Geriatr Psychiatry. 2007;22(10):1020-1024. doi:10.1002/gps.1783 136. Menon C, Westervelt HJ, Jahn DR, Dressel JA, O’Bryant SE. Normative Performance on the Brief Smell Identification Test (BSIT) in a Multi-Ethnic Bilingual Cohort: a Project FRONTIER Study. Clin Neuropsychol. 2013;27(6):946-961. doi:10.1080/13854046.2013.796406 137. Cai Y, Schrack JA, Gross AL, et al. Sensory Impairment and Algorithmic Classification of Early Cognitive Impairment. Alzheimers Dement (Amst). 2023;15(2):e12400. doi:10.1002/dad2.12400 138. Knight JE, Yoneda T, Lewis NA, Muniz-Terrera G, Bennett DA, Piccinin AM. Transitions Between Mild Cognitive Impairment, Dementia, and Mortality: The Importance of Olfaction. J Gerontol A Biol Sci Med Sci. 2023;78(7):1284-1291. doi:10.1093/gerona/glad001 139. Palmqvist S, Rossi M, Hall S, et al. Cognitive Effects of Lewy Body Pathology in Clinically Unimpaired Individuals. Nat Med. Published online July 18, 2023:1-8. doi:10.1038/s41591-023-02450-0 140. National Institute on Aging. Overnight-Hospitalization adjudication protocol.pdf. 2005. Accessed December 7, 2022. https://healthabc.nia.nih.gov/sites/default/files/OverNghtHosp_adjudication_protocol.pdf 141. National Institute on Aging. Death adjudication protocol.pdf. 2005. Accessed December 8, 2022. https://healthabc.nia.nih.gov/sites/default/files/Death_adjudication_protocol.pdf 142. ARIC Study. Surveillance Manuals. Accessed July 25, 2023. https://sites.cscc.unc.edu/aric/surveillance-manuals 143. National Institute on Aging. CVD Adjudication Protocol.pdf. 2005. Accessed December 7, 2022. https://healthabc.nia.nih.gov/sites/default/files/CVD_adjudication_protocol.pdf 144. ARIC Study. Manual 3 Version 6.8.pdf. March 14, 2023. Accessed July 25, 2023. https://sites.cscc.unc.edu/aric/sites/default/files/public/manuals/Manual%203%20Version% 206.8.pdf 145. Soliman EZ, Lopez F, O’Neal WT, et al. Atrial Fibrillation and Risk of ST-Segment- Elevation Versus Non-ST-Segment-Elevation Myocardial Infarction: The Atherosclerosis Risk in Communities (ARIC) Study. Circulation. 2015;131(21):1843-1850. doi:10.1161/CIRCULATIONAHA.114.014145 116 146. The ARIC Study. Manual 3c Cohort Stroke Procedures Version 1.0.pdf. August 19, 2021. Accessed July 25, 2023. https://sites.cscc.unc.edu/aric/sites/default/files/public/manuals/Manual%203c%20Cohort% 20Stroke%20Procedures%20Version%201.0.pdf 147. Rosamond WD, Folsom AR, Chambless LE, et al. Stroke Incidence and Survival among Middle-aged Adults: 9-year Follow-up of the Atherosclerosis Risk in Communities (ARIC) Cohort. Stroke. 1999;30(4):736-743. doi:10.1161/01.str.30.4.736 148. Rosamond WD, Chang PP, Baggett C, et al. Classification of Heart Failure in the Atherosclerosis Risk in Communities (ARIC) Study. Circulation: Heart Failure. 2012;5(2):152-159. doi:10.1161/CIRCHEARTFAILURE.111.963199 149. ARIC Study. Manual 3a.1. Cohort HF Procedures Version 1.0_0.pdf. August 19, 2021. Accessed July 26, 2023. https://sites.cscc.unc.edu/aric/sites/default/files/public/manuals/Manual%203a.1.%20Cohor t%20HF%20Procedures%20Version%201.0_0.pdf 150. Echouffo-Tcheugui JB, Ndumele C, Zhang S, et al. Diabetes, Preclinical Heart Failure Stages, and Progression to Overt Heart Failure: The Atherosclerosis Risk in Communities (ARIC) Study. J Am Coll Cardiol. 2022;79(23):2285-2293. doi:10.1016/j.jacc.2022.03.378 151. Austin PC, Lee DS, Fine JP. Introduction to the Analysis of Survival Data in the Presence of Competing Risks. Circulation. 2016;133(6):601-609. doi:10.1161/CIRCULATIONAHA.115.017719 152. Austin PC, Fine JP. Accounting for Competing Risks in Randomized Controlled Trials: a Review and Recommendations for Improvement. Statistics in Medicine. 2017;36(8):1203- 1209. doi:10.1002/sim.7215 153. Li Y, Hwang WT, Maude SL, et al. Statistical Considerations for Analyses of Time-To- Event Endpoints in Oncology Clinical Trials: Illustrations with CAR-T Immunotherapy Studies. Clin Cancer Res. 2022;28(18):3940-3949. doi:10.1158/1078-0432.CCR-22-0560 154. Hernán MA. The Hazards of Hazard Ratios. Epidemiology. 2010;21(1):13-15. doi:10.1097/EDE.0b013e3181c1ea43 155. Young J, Stensrud M, Tchetgen E, Hernán M. A causal framework for classical statistical estimands in failure‐time settings with competing events. Statistics in Medicine. 2020;39. doi:10.1002/sim.8471 156. D’Agostino RB, Lee ML, Belanger AJ, Cupples LA, Anderson K, Kannel WB. Relation of Pooled Logistic Regression to Time Dependent Cox Regression Analysis: The Framingham Heart Study. Statistics in Medicine. 1990;9(12):1501-1515. doi:10.1002/sim.4780091214 157. Hernan MA, Robins JM. Causal Inference: What If. Chapman & Hall/CRC 117 158. Westervelt HJ, Bruce JM, Faust MA. Distinguishing Alzheimer’s Disease and Dementia with Lewy Bodies Using Cognitive and Olfactory Measures. Neuropsychology. 2016;30(3):304-311. doi:10.1037/neu0000230 159. Yuan Y, Luo Z, Li C, et al. Poor Olfaction and Pneumonia Hospitalisation among Community-dwelling Older Adults: a Cohort Study. The Lancet Healthy Longevity. 2021;2(5):e275-e282. doi:10.1016/S2666-7568(21)00083-0 160. Kalogeropoulos AP, Georgiopoulou VV, Murphy RA, et al. Dietary Sodium Content, Mortality, and Risk for Cardiovascular Events in Older Adults: The Health, Aging, and Body Composition (Health ABC) Study. JAMA Intern Med. 2015;175(3):410. doi:10.1001/jamainternmed.2014.6278 161. Butler J, Kalogeropoulos A, Georgiopoulou V, et al. Incident Heart Failure Prediction in the Elderly: The Health ABC Heart Failure Score. Circ: Heart Failure. 2008;1(2):125-133. doi:10.1161/CIRCHEARTFAILURE.108.768457 162. Eckhardt CM, Balte PP, Barr RG, et al. Lung Function Impairment and Risk of Incident Heart Failure: the NHLBI Pooled Cohorts Study. European Heart Journal. 2022;43(23):2196-2208. doi:10.1093/eurheartj/ehac205 163. Kalogeropoulos A, Georgiopoulou V, Psaty BM, et al. Inflammatory Markers and Incident Heart Failure Risk in Older Adults: The Health, Aging, and Body Composition Study. J Am Coll Cardiol. 2010;55(19):2129-2137. doi:10.1016/j.jacc.2009.12.045 164. Schisterman EF, Cole SR, Platt RW. Overadjustment Bias and Unnecessary Adjustment in Epidemiologic Studies. Epidemiology. 2009;20(4):488-495. doi:10.1097/EDE.0b013e3181a819a1 165. Marcum ZA, Perera S, Newman AB, et al. Antihypertensive Use and Recurrent Falls in Community-Dwelling Older Adults: Findings From the Health ABC Study. The Journals of Gerontology: Series A. 2015;70(12):1562-1568. doi:10.1093/gerona/glv095 166. Yaffe K, Falvey CM, Hamilton N, et al. Association Between Hypoglycemia and Dementia in a Biracial Cohort of Older Adults With Diabetes Mellitus. JAMA Intern Med. 2013;173(14):1300-1306. doi:10.1001/jamainternmed.2013.6176 167. Kaup AR, Byers AL, Falvey C, et al. Trajectories of Depressive Symptoms in Older Adults and Risk of Dementia. JAMA Psychiatry. 2016;73(5):525-531. doi:10.1001/jamapsychiatry.2016.0004 168. Hankinson JL, Odencrantz JR, Fedan KB. Spirometric Reference Values from a Sample of the General U.S. Population. Am J Respir Crit Care Med. 1999;159(1):179-187. doi:10.1164/ajrccm.159.1.9712108 169. Wang K, Luo Z, Li C, et al. Blood Cholesterol Decreases as Parkinson’s Disease Develops and Progresses. J Parkinsons Dis. 2021;11(3):1177-1186. doi:10.3233/JPD-212670 118 170. Gopal DM, Kalogeropoulos AP, Georgiopoulou VV, et al. Serum Albumin Concentration and Heart Failure Risk The Health, Aging, and Body Composition Study. Am Heart J. 2010;160(2):279-285. doi:10.1016/j.ahj.2010.05.022 171. Metti AL, Yaffe K, Boudreau RM, et al. Trajectories of Inflammatory Markers and Cognitive Decline over 10 Years. Neurobiol Aging. 2014;35(12):2785-2790. doi:10.1016/j.neurobiolaging.2014.05.030 172. Goldenstein L, Driver TH, Fried LF, et al. Serum Bicarbonate Concentrations and Kidney Disease Progression in Community-living Elders: the Health, Aging, and Body Composition (Health ABC) Study. Am J Kidney Dis. 2014;64(4):542-549. doi:10.1053/j.ajkd.2014.05.009 173. Inker LA, Schmid CH, Tighiouart H, et al. Estimating Glomerular Filtration Rate from Serum Creatinine and Cystatin C. N Engl J Med. 2012;367(1):20-29. doi:10.1056/NEJMoa1114248 174. Gray RJ. A Class of K-Sample Tests for Comparing the Cumulative Incidence of a Competing Risk. The Annals of Statistics. 1988;16(3):1141-1154. 175. Lin DY, Wei LJ. The Robust Inference for the Cox Proportional Hazards Model. Journal of the American Statistical Association. 1989;84(408):1074-1078. doi:10.2307/2290085 176. Givens RC, Schulze PC. Molecular Changes in Heart Failure. In: Eisen H, ed. Heart Failure: A Comprehensive Guide to Pathophysiology and Clinical Care. Springer; 2017:1- 26. doi:10.1007/978-1-4471-4219-5_1 177. Lin DY, Wei LJ, Ying Z. Checking the Cox Model with Cumulative Sums of Martingale- Based Residuals. Biometrika. 1993;80(3):557-572. doi:10.2307/2337177 178. Irawati S, Wasir R, Floriaan Schmidt A, et al. Long-term Incidence and Risk Factors of Cardiovascular Events in Asian Populations: Systematic Review and Meta-analysis of Population-based Cohort Studies. Curr Med Res Opin. 2019;35(2):291-299. doi:10.1080/03007995.2018.1491149 179. Howard G, Banach M, Kissela B, et al. Age-Related Differences in the Role of Risk Factors for Ischemic Stroke. Neurology. Published online January 18, 2023. doi:10.1212/WNL.0000000000206837 180. Pang NYL, Song HJJMD, Tan BKJ, et al. Association of Olfactory Impairment With All- Cause Mortality: A Systematic Review and Meta-analysis. JAMA Otolaryngology–Head & Neck Surgery. 2022;148(5):436-445. doi:10.1001/jamaoto.2022.0263 181. Doty RL. Olfactory Dysfunction in Neurodegenerative Diseases: is There a Common Pathological Substrate? The Lancet Neurology. 2017;16(6):478-488. doi:10.1016/S1474- 4422(17)30123-0 119 182. Le Floch JP, Le Lièvre G, Labroue M, Paul M, Peynegre R, Perlemuter L. Smell Dysfunction and Related Factors in Diabetic Patients. Diabetes Care. 1993;16(6):934-937. doi:10.2337/diacare.16.6.934 183. Catamo E, Tornese G, Concas MP, Gasparini P, Robino A. Differences in Taste and Smell Perception between Type 2 Diabetes Mellitus Patients and Healthy Controls. Nutr Metab Cardiovasc Dis. 2021;31(1):193-200. doi:10.1016/j.numecd.2020.08.025 184. Heiser C, Haller B, Sohn M, et al. Olfactory Function is Affected in Patients with Cirrhosis Depending on the Severity of Hepatic Encephalopathy. Annals of Hepatology. 2018;17(5):822-829. doi:10.5604/01.3001.0012.3143 185. Wang K, Luo Z, Li C, et al. Olfaction and Kidney Function in Community-dwelling Older Adults. PLoS One. 2022;17(2):e0264448. doi:10.1371/journal.pone.0264448 186. Nagururu NV, Bernstein IA, Voegtline K, Olson S, Agrawal Y, Rowan NR. The Association of Peripheral and Central Olfaction With Frailty in Older Adults. The Journals of Gerontology: Series A. 2023;78(7):1276-1283. doi:10.1093/gerona/glac237 187. Purdy F, Luo Z, Gardiner JC, et al. Olfaction and Changes in Body Composition in a Large Cohort of Older U.S. Adults. J Gerontol A Biol Sci Med Sci. 2020;75(12):2434-2440. doi:10.1093/gerona/glaa085 188. Joo YH, Hwang SH, Han K do, Seo JH, Kang JM. Relationship between Olfactory Dysfunction and Suicidal Ideation: The Korea National Health and Nutrition Examination Survey. Am J Rhinol Allergy. 2015;29(4):268-272. doi:10.2500/ajra.2015.29.4194 189. Leon M, Woo CC. Olfactory Loss is a Predisposing Factor for Depression, while Olfactory Enrichment is an Effective Treatment for Depression. Front Neurosci. 2022;16:1013363. doi:10.3389/fnins.2022.1013363 190. Vlodaver Z. Pathology of Chronic Obstructive Coronary Disease. In: Vlodaver Z, Wilson RF, Garry DJ, eds. Coronary Heart Disease: Clinical, Pathological, Imaging, and Molecular Profiles. Springer US; 2012:159-185. doi:10.1007/978-1-4614-1475-9_8 191. Banerjee C, Chimowitz MI. Stroke Caused by Atherosclerosis of the Major Intracranial Arteries. Circulation Research. 2017;120(3):502-513. doi:10.1161/CIRCRESAHA.116.308441 192. Ledley GS, Ahmed S, Jones H, Rough SJ, Kurnik P. Hemodynamics and Heart Failure. In: Eisen H, ed. Heart Failure: A Comprehensive Guide to Pathophysiology and Clinical Care. Springer; 2017:27-48. doi:10.1007/978-1-4471-4219-5_2 193. Birte-Antina W, Ilona C, Antje H, Thomas H. Olfactory Training with Older People. Int J Geriatr Psychiatry. 2018;33(1):212-220. doi:10.1002/gps.4725 194. Pinckard K, Baskin KK, Stanford KI. Effects of Exercise to Improve Cardiovascular Health. Front Cardiovasc Med. 2019;6:69. doi:10.3389/fcvm.2019.00069 120 195. Szczepańska E, Białek-Dratwa A, Janota B, Kowalski O. Dietary Therapy in Prevention of Cardiovascular Disease (CVD)—Tradition or Modernity? A Review of the Latest Approaches to Nutrition in CVD. Nutrients. 2022;14(13):2649. doi:10.3390/nu14132649 196. Wuorela M, Lavonius S, Salminen M, Vahlberg T, Viitanen M, Viikari L. Self-rated Health and Objective Health Status as Predictors of All-cause Mortality among Older People: a Prospective Study with a 5-, 10-, and 27-year Follow-up. BMC Geriatrics. 2020;20(1):120. doi:10.1186/s12877-020-01516-9 197. Lorem G, Cook S, Leon DA, Emaus N, Schirmer H. Self-reported Health as a Predictor of Mortality: A Cohort Study of its Relation to Other Health Measurements and Observation Time. Sci Rep. 2020;10(1):4886. doi:10.1038/s41598-020-61603-0 198. Koohi F, Harshfield EL, Markus HS. Contribution of Conventional Cardiovascular Risk Factors to Brain White Matter Hyperintensities. Journal of the American Heart Association. 2023;12(14):e030676. doi:10.1161/JAHA.123.030676 199. Doty RL, Marcus A, Lee WW. Development of the 12-Item Cross-Cultural Smell Identification Test(CC-SIT). The Laryngoscope. 1996;106(3):353-356. doi:10.1097/00005537-199603000-00021 200. Hummel T, Kobal G, Gudziol H, Mackay-Sim A. Normative Data for the “Sniffin’ Sticks” Including Tests of Odor Identification, Odor Discrimination, and Olfactory Thresholds: an Upgrade Based on a Group of More Than 3,000 Subjects. Eur Arch Otorhinolaryngol. 2007;264(3):237-243. doi:10.1007/s00405-006-0173-0 201. Guo N, Zhu Y, Tian D, et al. Role of Diet in Stroke Incidence: an Umbrella Review of Meta-analyses of Prospective Observational Studies. BMC Medicine. 2022;20(1):194. doi:10.1186/s12916-022-02381-6 202. Labeit B, Muhle P, von Itter J, et al. Clinical Determinants and Neural Correlates of Presbyphagia in Community-dwelling Older Adults. Front Aging Neurosci. 2022;14:912691. doi:10.3389/fnagi.2022.912691 203. Van Regemorter V, Dollase J, Coulie R, et al. Olfactory Dysfunction Predicts Frailty and Poor Postoperative Outcome in Older Patients Scheduled for Elective Non-Cardiac Surgery. J Nutr Health Aging. 2022;26(11):981-986. doi:10.1007/s12603-022-1851-3 204. Vandersteen C, Payne M, Dumas LÉ, et al. What about Using Sniffin’ Sticks 12 Items Test to Screen Post-COVID-19 Olfactory Disorders? Eur Arch Otorhinolaryngol. 2022;279(7):3477-3484. doi:10.1007/s00405-021-07148-y 205. Schmicker M, Frühling I, Menze I, et al. The Potential Role of Gustatory Function as an Early Diagnostic Marker for the Risk of Alzheimer’s Disease in Subjective Cognitive Decline. J Alzheimers Dis Rep. 2023;7(1):249-262. doi:10.3233/ADR220092 121 206. Palta P, Chen H, Deal JA, et al. Olfactory Function and Neurocognitive Outcomes in Old Age: the Atherosclerosis Risk in Communities Neurocognitive Study (ARIC-NCS). Alzheimers Dement. 2018;14(8):1015-1021. doi:10.1016/j.jalz.2018.02.019 207. Koton S, Schneider ALC, Windham BG, Mosley TH, Gottesman RF, Coresh J. Microvascular Brain Disease Progression and Risk of Stroke: The ARIC Study. Stroke. 2020;51(11):3264-3270. doi:10.1161/STROKEAHA.120.030063 208. Koton S, Schneider ALC, Rosamond WD, et al. Stroke Incidence and Mortality Trends in US Communities, 1987 to 2011. JAMA. 2014;312(3):259-268. doi:10.1001/jama.2014.7692 209. Johansen MC, Wang W, Zhang M, et al. Risk of Dementia Associated With Atrial Cardiopathy: The ARIC Study. J Am Heart Assoc. 2022;11(16):e025646. doi:10.1161/JAHA.121.025646 210. Powell DS, Kuo PL, Qureshi R, et al. The Relationship of APOE ε4, Race, and Sex on the Age of Onset and Risk of Dementia. Front Neurol. 2021;12:735036. doi:10.3389/fneur.2021.735036 211. Marrone MT, Mondul AM, Prizment AE, et al. Lipid-Lowering Drug Use and Cancer Incidence and Mortality in the ARIC Study. JNCI Cancer Spectrum. 2021;5(5). doi:10.1093/jncics/pkab080 212. Wijkman MO, Malachias MVB, Claggett BL, et al. Resistance to Antihypertensive Treatment and Long‐term Risk: The Atherosclerosis Risk in Communities study. J Clin Hypertens (Greenwich). 2021;23(10):1887-1896. doi:10.1111/jch.14269 213. Magnani JW, Norby FL, Agarwal SK, et al. Racial Differences in Atrial Fibrillation-Related Cardiovascular Disease and Mortality: The Atherosclerosis Risk in Communities (ARIC) Study. JAMA Cardiology. 2016;1(4):433-441. doi:10.1001/jamacardio.2016.1025 214. Mou L, Norby FL, Chen LY, et al. Lifetime Risk of Atrial Fibrillation by Race and Socioeconomic Status: The Atherosclerosis Risk in Communities (ARIC) Study. Circ Arrhythm Electrophysiol. 2018;11(7):e006350. doi:10.1161/CIRCEP.118.006350 215. Knopman DS, Gottesman RF, Sharrett AR, et al. Mild Cognitive Impairment and Dementia Prevalence: The Atherosclerosis Risk in Communities Neurocognitive Study (ARIC-NCS). Alzheimers Dement (Amst). 2016;2:1-11. doi:10.1016/j.dadm.2015.12.002 216. Huang X, Alonso A, Guo X, et al. Statins, plasma cholesterol, and risk of Parkinson’s disease: a prospective study. Mov Disord. 2015;30(4):552-559. doi:10.1002/mds.26152 217. Shukla A, Reed NS, Armstrong NM, Lin FR, Deal JA, Goman AM. Hearing Loss, Hearing Aid Use, and Depressive Symptoms in Older Adults-Findings from the Atherosclerosis Risk in Communities Neurocognitive Study (ARIC-NCS). J Gerontol B Psychol Sci Soc Sci. 2021;76(3):518-523. doi:10.1093/geronb/gbz128 122 218. Kucharska-Newton AM, Palta P, Burgard S, et al. Operationalizing Frailty in the Atherosclerosis Risk in Communities Study Cohort. J Gerontol A Biol Sci Med Sci. 2017;72(3):382-388. doi:10.1093/gerona/glw144 219. Quispe R, Elshazly MB, Zhao D, et al. Total cholesterol/HDL-cholesterol Ratio Discordance with LDL-cholesterol and Non-HDL-cholesterol and Incidence of Atherosclerotic Cardiovascular Disease in Primary Prevention: The ARIC study. European Journal of Preventive Cardiology. 2020;27(15):1597-1605. doi:10.1177/2047487319862401 220. Scheppach JB, Coresh J, Wu A, et al. Albuminuria and Estimated GFR as Risk Factors for Dementia in Midlife and Older Age: Findings From the ARIC Study. Am J Kidney Dis. 2020;76(6):775-783. doi:10.1053/j.ajkd.2020.03.015 221. Scheike TH, Zhang MJ. Flexible Competing Risks Regression Modeling and Goodness-of- fit. Lifetime Data Anal. 2008;14(4):464-483. doi:10.1007/s10985-008-9094-0 222. Heyard R, Timsit JF, Held L, COMBACTE-MAGNET consortium. Validation of Discrete Time-to-event Prediction Models in the Presence of Competing Risks. Biom J. 2020;62(3):643-657. doi:10.1002/bimj.201800293 223. DeCarli C, Reed T, Miller BL, Wolf PA, Swan GE, Carmelli D. Impact of Apolipoprotein E ε4 and Vascular Disease on Brain Morphology in Men From the NHLBI Twin Study. Stroke. 1999;30(8):1548-1553. doi:10.1161/01.STR.30.8.1548 224. Laudisio A, Navarini L, Margiotta DPE, et al. The Association of Olfactory Dysfunction, Frailty, and Mortality Is Mediated by Inflammation: Results from the InCHIANTI Study. J Immunol Res. 2019;2019:3128231. doi:10.1155/2019/3128231 225. Nagururu NV, Bernstein IA, Voegtline K, Olson S, Agrawal Y, Rowan NR. The Association of Peripheral and Central Olfaction With Frailty in Older Adults. The Journals of Gerontology: Series A. Published online December 11, 2022:glac237. doi:10.1093/gerona/glac237 226. Wang E, Wroblewski KE, McClintock MK, Pinto JM, Witt LJ. Olfactory Decline Develops in Parallel with Frailty in Older US Adults with Obstructive Lung Diseases. International Forum of Allergy & Rhinology. 2024;14(4):819-827. doi:10.1002/alr.23273 227. Renedo D, Acosta JN, Koo AB, et al. Higher Hospital Frailty Risk Score Is Associated With Increased Risk of Stroke: Observational and Genetic Analyses. Stroke. 2023;54(6):1538- 1547. doi:10.1161/STROKEAHA.122.041891 228. Kalaria RN, Akinyemi R, Ihara M. Stroke Injury, Cognitive Impairment and Vascular Dementia. Biochim Biophys Acta. 2016;1862(5):915-925. doi:10.1016/j.bbadis.2016.01.015 229. Wardlaw JM, Allerhand M, Doubal FN, et al. Vascular Risk Factors, Large-artery Atheroma, and Brain White Matter Hyperintensities. Neurology. 2014;82(15):1331-1338. doi:10.1212/WNL.0000000000000312 123 230. Tian F, Chen L, Qian Z (Min), et al. Ranking Age-specific Modifiable Risk Factors for Cardiovascular Disease and Mortality: Evidence from a Population-based Longitudinal Study. eClinicalMedicine. 2023;64. doi:10.1016/j.eclinm.2023.102230 231. Cecchini MP, Bojanowski V, Bodechtel U, Hummel T, Hähner A. Olfactory Function in Patients with Ischemic Stroke: a Pilot Study. Eur Arch Otorhinolaryngol. 2012;269(4):1149-1153. doi:10.1007/s00405-011-1819-0 232. Moo L, Wityk RJ. Olfactory and Taste Dysfunction after Bilateral Middle Cerebral Artery Stroke. J Stroke Cerebrovasc Dis. 1999;8(5):353-354. doi:10.1016/s1052-3057(99)80011-1 233. Rochet M, El-Hage W, Richa S, Kazour F, Atanasova B. Depression, Olfaction, and Quality of Life: A Mutual Relationship. Brain Sci. 2018;8(5):80. doi:10.3390/brainsci8050080 234. Boesveldt S, Yee JR, McClintock MK, Lundström JN. Olfactory Function and the Social Lives of Older Adults: a Matter of Sex. Sci Rep. 2017;7(1):45118. doi:10.1038/srep45118 235. Pinto JM, Wroblewski KE, Kern DW, Schumm LP, McClintock MK. Olfactory Dysfunction Predicts 5-Year Mortality in Older Adults. PLoS ONE. 2014;9(10). doi:10.1371/journal.pone.0107541 236. Jashari F, Ibrahimi P, Nicoll R, Bajraktari G, Wester P, Henein MY. Coronary and Carotid Atherosclerosis: Similarities and Differences. Atherosclerosis. 2013;227(2):193-200. doi:10.1016/j.atherosclerosis.2012.11.008 237. Polak Joseph F., Pencina Michael J., Pencina Karol M., O’Donnell Christopher J., Wolf Philip A., D’Agostino Ralph B. Carotid-Wall Intima–Media Thickness and Cardiovascular Events. New England Journal of Medicine. 2011;365(3):213-221. doi:10.1056/NEJMoa1012592 238. Nambi V, Chambless L, He M, et al. Common Carotid Artery Intima-media Thickness is as Good as Carotid Intima-Media Thickness of All Carotid Artery Segments in Improving Prediction of Coronary Heart Disease Risk in the Atherosclerosis Risk in Communities (ARIC) study. Eur Heart J. 2012;33(2):183-190. doi:10.1093/eurheartj/ehr192 239. Lichtenstein AH, Appel LJ, Vadiveloo M, et al. 2021 Dietary Guidance to Improve Cardiovascular Health: A Scientific Statement From the American Heart Association. Circulation. 2021;144(23):e472-e487. doi:10.1161/CIR.0000000000001031 240. Ding N, Sang Y, Chen J, et al. Cigarette Smoking, Smoking Cessation, and Long-Term Risk of 3 Major Atherosclerotic Diseases. Journal of the American College of Cardiology. 2019;74(4):498-507. doi:10.1016/j.jacc.2019.05.049 241. Haan MN, Mayeda ER. Apolipoprotein E Genotype and Cardiovascular Diseases in the Elderly. Curr Cardiovasc Risk Rep. 2010;4(5):361-368. doi:10.1007/s12170-010-0118-4 124 242. Dhindsa DS, Sandesara PB, Shapiro MD, Wong ND. The Evolving Understanding and Approach to Residual Cardiovascular Risk Management. Front Cardiovasc Med. 2020;7:88. doi:10.3389/fcvm.2020.00088 243. Nowbar AN, Gitto M, Howard JP, Francis DP, Al-Lamee R. Mortality From Ischemic Heart Disease. Circulation: Cardiovascular Quality and Outcomes. 2019;12(6):e005375. doi:10.1161/CIRCOUTCOMES.118.005375 244. Delgado J, Bowman K, Ble A, et al. Blood Pressure Trajectories in the 20 Years Before Death. JAMA Internal Medicine. 2018;178(1):93-99. doi:10.1001/jamainternmed.2017.7023 245. Wilkinson DJ, Piasecki M, Atherton PJ. The Age-related Loss of Skeletal Muscle Mass andFunction: Measurement and Physiology of Muscle Fibre Atrophy and Muscle Fibre Loss in Humans. Ageing Research Reviews. 2018;47:123-132. doi:10.1016/j.arr.2018.07.005 246. Duncan MS, Vasan RS, Xanthakis V. Trajectories of Blood Lipid Concentrations Over the Adult Life Course and Risk of Cardiovascular Disease and All-Cause Mortality: Observations From the Framingham Study Over 35 Years. J Am Heart Assoc. 2019;8(11):e011433. doi:10.1161/JAHA.118.011433 247. Sattar N, Preiss D. Reverse Causality in Cardiovascular Epidemiological Research: More Common Than Imagined? Circulation. 2017;135(24):2369-2372. doi:10.1161/CIRCULATIONAHA.117.028307 248. Nambi V, Chambless L, Folsom AR, et al. Carotid intima-media thickness and presence or absence of plaque improves prediction of coronary heart disease risk in the Atherosclerosis Risk in Communities (ARIC) study. J Am Coll Cardiol. 2010;55(15):1600-1607. doi:10.1016/j.jacc.2009.11.075 249. Lichtenstein AH, Appel LJ, Vadiveloo M, et al. 2021 Dietary Guidance to Improve Cardiovascular Health: A Scientific Statement From the American Heart Association. Circulation. 2021;144(23):e472-e487. doi:10.1161/CIR.0000000000001031 250. Lavie CJ, Ozemek C, Carbone S, Katzmarzyk PT, Blair SN. Sedentary Behavior, Exercise, and Cardiovascular Health. Circulation Research. 2019;124(5):799-815. doi:10.1161/CIRCRESAHA.118.312669 251. Roth GA, Dorsey H, Decleene N, et al. The Global Burden of Heart Failure: a Systematic Analysis for the Global Burden of Disease Study 2021. European Heart Journal. 2023;44(Supplement_2):ehad655.876. doi:10.1093/eurheartj/ehad655.876 252. Simmonds SJ, Cuijpers I, Heymans S, Jones EAV. Cellular and Molecular Differences between HFpEF and HFrEF: A Step Ahead in an Improved Pathological Understanding. Cells. 2020;9(1):242. doi:10.3390/cells9010242 125 253. Wj P, C T. A Novel Paradigm for Heart Failure with Preserved Ejection Fraction: Comorbidities Drive Myocardial Dysfunction and Remodeling through Coronary Microvascular Endothelial Inflammation. Journal of the American College of Cardiology. 2013;62(4). doi:10.1016/j.jacc.2013.02.092 254. Theou O, Rockwood K, Fülöp T. Frailty in Aging: Biological, Clinical and Social Implications. S. Karger AG; 2015. Accessed July 15, 2024. http://ebookcentral.proquest.com/lib/michstate-ebooks/detail.action?docID=3563953 255. Sergi G, Veronese N, Fontana L, et al. Pre-Frailty and Risk of Cardiovascular Disease in Elderly Men and Women: The Pro.V.A. Study. Journal of the American College of Cardiology. 2015;65(10):976-983. doi:10.1016/j.jacc.2014.12.040 256. Newman AB, Gottdiener JS, Mcburnie MA, et al. Associations of Subclinical Cardiovascular Disease with Frailty. J Gerontol A Biol Sci Med Sci. 2001;56(3):M158-166. doi:10.1093/gerona/56.3.m158 257. Nadruz W, Kitzman D, Windham BG, et al. Cardiovascular Dysfunction and Frailty Among Older Adults in the Community: The ARIC Study. J Gerontol A Biol Sci Med Sci. 2017;72(7):958-964. doi:10.1093/gerona/glw199 258. Gomez-Cabrero D, Walter S, Abugessaisa I, et al. A Robust Machine Learning Framework to Identify Signatures for Frailty: a Nested Case-control Study in Four Aging European Cohorts. Geroscience. 2021;43(3):1317-1329. doi:10.1007/s11357-021-00334-0 259. Mant J, Doust J, Roalfe A, et al. Systematic Review and Individual Patient Data Meta- analysis of Diagnosis of Heart Failure, with Modelling of Implications of Different Diagnostic Strategies in Primary Care. Health Technol Assess. 2009;13(32):1-207, iii. doi:10.3310/hta13320 260. Nambi V, Liu X, Chambless LE, et al. Troponin T and N-terminal pro-B-type natriuretic peptide: a biomarker approach to predict heart failure risk--the atherosclerosis risk in communities study. Clin Chem. 2013;59(12):1802-1810. doi:10.1373/clinchem.2013.203638 261. Saeed A, Nambi V, Sun W, et al. Short-Term Global Cardiovascular Disease Risk Prediction in Older Adults. Journal of the American College of Cardiology. 2018;71(22):2527-2536. doi:10.1016/j.jacc.2018.02.050 262. Schroeder EB, Liao D, Chambless LE, Prineas RJ, Evans GW, Heiss G. Hypertension, Blood Pressure, and Heart Rate Variability. Hypertension. 2003;42(6):1106-1111. doi:10.1161/01.HYP.0000100444.71069.73 263. Arrigo M, Jessup M, Mullens W, et al. Acute heart failure. Nat Rev Dis Primers. 2020;6(1):1-15. doi:10.1038/s41572-020-0151-7 264. Yancy CW, Jessup M, Bozkurt B, et al. 2013 ACCF/AHA Guideline for the Management of Heart Failure: a Report of the American College of Cardiology Foundation/American Heart 126 Association Task Force on Practice Guidelines. J Am Coll Cardiol. 2013;62(16):e147-239. doi:10.1016/j.jacc.2013.05.019 265. Zhang Z, Rowan NR, Pinto JM, et al. Exposure to Particulate Matter Air Pollution and Anosmia. JAMA Network Open. 2021;4(5):e2111606. doi:10.1001/jamanetworkopen.2021.11606 266. Andersson J, Oudin A, Nordin S, Forsberg B, Nordin M. PM2.5 Exposure and Olfactory Functions. International Journal of Environmental Health Research. 2022;32(11):2484- 2495. doi:10.1080/09603123.2021.1973969 267. Liu ZY, Vaira LA, Boscolo-Rizzo P, Walker A, Hopkins C. Post-viral Olfactory Loss and Parosmia. BMJ Medicine. 2023;2(1). doi:10.1136/bmjmed-2022-000382 268. Trimmer C, Keller A, Murphy NR, et al. Genetic Variation across the Human Olfactory Receptor Repertoire Alters Odor Perception. Proceedings of the National Academy of Sciences. 2019;116(19):9475-9480. doi:10.1073/pnas.1804106115 269. Shah AD, Bartlett JW, Carpenter J, Nicholas O, Hemingway H. Comparison of Random Forest and Parametric Imputation Models for Imputing Missing Data Using MICE: A CALIBER Study. Am J Epidemiol. 2014;179(6):764-774. doi:10.1093/aje/kwt312 270. White IR, Royston P. Imputing missing covariate values for the Cox model. Stat Med. 2009;28(15):1982-1998. doi:10.1002/sim.3618 271. Rubin DB. Multiple Imputation for Nonresponse in Surveys. John Wiley & Sons; 2004. 272. Oehlert GW. A Note on the Delta Method. Am Stat. 1992;46(1). 273. Cawthon PM, Fox KM, Gandra SR, et al. Do muscle mass, muscle density, strength, and physical function similarly influence risk of hospitalization in older adults? J Am Geriatr Soc. 2009;57(8):1411-1419. doi:10.1111/j.1532-5415.2009.02366.x 274. Hanlon P, Nicholl BI, Jani BD, Lee D, McQueenie R, Mair FS. Frailty and pre-frailty in middle-aged and older adults and its association with multimorbidity and mortality: a prospective analysis of 493 737 UK Biobank participants. Lancet Public Health. 2018;3(7):e323-e332. doi:10.1016/S2468-2667(18)30091-4 275. Scheike TH, Zhang MJ. Analyzing Competing Risk Data Using the R timereg Package. J Stat Soft. 2011;38(2). doi:10.18637/jss.v038.i02 127 APPENDIX 1: LITERATURE REVIEW ON PREVIOUS INVESTIGATIONS The empirical evidence of olfaction and major cardiovascular adverse outcome consists of two sections. Table A1.1 includes studies with olfaction as the outcome of interest and cardiovascular disease, cardiovascular subclinical markers, and/or cardiovascular risk factors as the exposures of interest. Table A1.2 includes olfaction as the exposure of interest and major adverse cardiovascular disease as the outcome of interest, which is in line with our study goal. Note: () under Exposure and Outcome means the approach of measurements. Table A1.1 Previous studies regarding cardiovascular disease, cardiovascular subclinical markers, and cardiovascular risk factors in relation to olfaction Study Murphy, 2002, JAMA2 Study design Cross- sectiona l Population Exposure outcome covariates effect Estimate The Epidemiology of Hearing Loss Study: n=2800 (≥ 55 y, White, WI) stroke (unknown); smoking status(unknown); diabetes Self-reported and objective measured olfaction impairment (SDOIT) ↑ ↑ - OR: [Yes vs. No] 1.99 (1.13-3.51); [Current vs. never] 2.15(1.49- 3.10); [Yes vs. No] 1.08 (0.79,1.47) Age, sex, occupation, sinus infection, nasal congestion, history of allergies, head injury, deviated septum, nasal polyps, chemotherapy , PD, epilepsy, use of medications 128 Table A1.1 (cont’d) Schubert, 2011, Laryngoscope3 6 Longitu dinal Study The Epidemiology of Hearing Loss Study: n=1556 (≥ 55 y, White, WI) (Self-reported) Statin use Longitu dinal cohort Schubert, 2014, J Gerontol A Biol Sci Med Sci90 The Beaver Dam Offspring Study (Epidemiology of Hearing Loss Study) (n=2302) (≥ 55 y, White, WI) Carotid IMT, Number of carotid plaque Schubert, 2015, Age and aging91 Longitu dinal cohort Carotid IMT, carotid plaque The Beaver Dam Offspring Study (Epidemiology of Hearing Loss Study) (n=1611 without olfactory impairment) 129 objective measured olfaction (SDOIT) decline between baseline and five years later objective measured olfaction (SDOIT) decline between baseline and five years later Incident objective measured olfaction (SDOIT) impairment between baseline and five years later Age, sex, history of nasal polyps and deviated septum, oral corticosteroid s used, history of heavy alcohol use, exercise Age, sex, hypertension, BMI, alcohol and smoking status Age, sex, smoking, exercise, nasal steroids, oral steroids, nasal polyps/ deviated septum ↓ OR: 0.68 (0.46, 0.99) - ↑ - - ↑ ↑ OR: [per 0.1mm] 1.13 (0.98, 1.31) [per site] 1.24 (1.01, 1.53) HR: {≥60 years} [T3 vs. T1] 1.03 (0.70-1.52) [per site] 1.00 (0.91-1.10) {<60 years} [T3 vs. T1] 4.35 (1.69-11.21) [per site] 1.56 (1.17-2.08) Table A1.1 (cont’d) Wehling, 2015, BMC Neurology118 Seubert, 2017, J Gerontol A Biol Sci Med Sci119 Cross- sectiona l study (Hospit al- based) Cross- sectiona l Hospital-based study: n=74 stroke patients vs. age and sex-matched controls (age: 67.2 years) Swedish National Study: N=2234 (60-90 y, no neurodegeneration) Stroke occurrence within one year History of coronary heart disease; Heart faulire; Afib; CBVD; Hypertension; TC; Okamoto, 2019, Chemical Senses121 Cross- sectiona l Hospital-based patients in Japan: n=19 acute ischemic stroke patients (69.8 y) No hypoperfusion vs. hypoperfusion in thalamus area Age and sex ↑ Linear correlation Age, education, APOE Ɛ4 carrier, BDNF, depression, Migraine, physical activity, BMI, occupation, appetite / - - - - - - ↑ Unknown of the exact value as only a forest plot is provided. Unknown of the exact value as only P value is provided Objective (SOIT) and self-reported olfactory function Olfactory dysfunction (16-item odor identification task) the T&T olfactometer (smell detection and recognition); olfactory identification using the Open Essence 130 Table A1.1 (cont’d) Longitu dinal cohort Ekstorm, 2020, J125 Gerontol A Biol Sci Med Sci the Swedish National Study on Aging and Care in Kungsholmen: n=1780 (70.5 y, 61.9% female, with ≥2 follow-ups) (Inpatient registries) cerebrovascular disease; cardiovascular disease burden (Afib, heart failure, coronary heart disease); diabetes; Average olfactory change per year (Sniffin’ Sticks battery) Longitu dinal cohort Palmquist, 2020, J Gerontol A Biol Sci Med Sci126 Swedish National Study on Aging and Care: n=1004 (60-90 y, without OD) (Inpatient registries) Smoking; atrial fibrillation; CBVD; Hypertension Incident olfactory impairment (Sniffin’ Sticks battery:≤10) - - ↑ ↑ ↑ ↑ ↓ Predictor * time -0.077 (-0.155, 0.002) -0.009 (-0.041, 0.023) -0.09 (-0.161, - 0.026)) OR: 1.92 (1.12-3.29) 2.07 (1.15, 3.75) 2.35 (1.02, 5.39) 0.66 (0.44, 1.00) Age, education, and test version, profession, vocabulary, number of medications, gait speed, APOE Ɛ4 carrier, BDNF Baseline odor identification , age, APOE Ɛ4 carrier, Episodic memory, Perceptual speed, MMSE, Physical inactivity, Head trauma, Complex leisure activity, social network index 131 Table A1.1 (cont’d) Schlosser, 2020, American J Rhinology and Allergy122 Cross- sectiona l A clinic at the Medical University of South Carolina (MUSC): N=176, (20-93 y) Heart problems; Roh, 2021, Scientific Report123 Cross- sectiona l Korean National Health and Nutrition Examination Survey: n=20016 (≥40 y) (self-reported) diabetes; hypertension; CAD; stroke; obesity; abdominal obesity hypertriglyceridem ia; low HDL Kultur, 2022, Neurological Sciences124 Cross- sectiona l Hospital-based population: n=82 (mean age: 54.3 y) Stroke 132 Age, MMSE, anxiety ↑ TDI score: -1.665, P=0.01 Age, sex, household income, educational level, smoking status, heavy drinking, sleep duration, lack of exercise, history of rhinosinusitis and rhinitis age - - ↑ - - ↑ - - ↑ OR: 1.08 (0.85, 1.38) 1.05 (0.88-1.27) 1.68 (1.15,2.47) 1.33(0.88, 2.00) 0.80 (0.64,1.01) 1.30 (1.03,1.63) 1.06 (0.88, 1.27) 1.05 (0.88, 1.26) Independent sample t test showed significant correlation between stroke and all olfactory MRI markers Threshold, discriminatio n, and identification , (TDI) score (Sniffin’ Sticks test) (Self- reported) history of olfactory dysfunction MRI imaging: Olfactory bulb volume, olfactory sulcus depth Insular gyrus area, corpus amygdala area Table A1.1 (cont’d) Shrestha, 2023, Nutrient35 Cross- sectiona l ARIC Study: m=6053 (mean age: 75.6 y) Olfaction- Sniffin’ Sticks Smoking; Obesity; Total cholesterol; Diabetes; Hypertension; MI history; CHD history; Stroke history Age, sex, education, race-site, alcohol, APOE Ɛ4, physical activity, CRP, vitamin B12, blood Hemoglobin ↑ - - ↑ ↓ - - - RR: 1.051 (1.000, 1.103); 1.127 (1.035, 1.227) 0.941 (0.881, 1.005); 0.920 (0.831, 1.020) 0.977 (0.952, 1.002) 1.075 (1.023, 1.129) 0.931 (0.881, 0.983) 0.982 (0.895, 1.077) 1.046 (0.970, 1.129) 1.037 (0.928, 1.160) Abbreviations: SDOIT: the San Diego Odor Identification Test; SOIT: Scandinavian odor identification test; PD: Parkinson’s disease; TC: total cholesterol; HDL-C: high-density lipoprotein-cholesterol; BMI: body mass index; IMT: intima media thickness; Afib: atrial fibrillation; CBVD: cerebrovascular disease; MMSE: Mini-Mental State examination; CAD: coronary artery disease. 133 Table A1.2 Previous study regarding olfaction in relation to incident cardiovascular disease Study Population Exposure outcome covariates effect Estimate - OR: 1.75 (0.93, 3.31) Study design Longitudinal Siegel, Int Forum Allergy Rhinol, 2019127 Olfactory decline (Sniffin’s Sticks) between baseline and year 5 Self-reported first heart attack or new heart disease at year 10 National Social Like Health and Aging Project, n=3528 Baseline age, gender, race/ethnicity, level of education, and cognition, baseline BMI and self- reported physical health 134 APPENDIX 2: LIST OF REGRESSION MODELS IN THE PRESENCE OF COMPETING EVENTS Table A2.1 The comparison of different regression models in the presence of competing events Parameter interpretation Measure of association a Meaning of association b Available package Conver gence Regressions Cause- specific hazard proportional model151,153 Fine-Gray proportional model (Based on Cox proportional model)151 Discrete-time Fine-Gray model155 proporti onal hazard assump tion Yes log⁡ 𝑐𝑠(𝑡, 𝑥 = 1) 𝜆𝑘 𝑐𝑠(𝑡, 𝑥 = 0) 𝜆𝑘 →derive cause-specific hazard ratio Yes log log⁡(1 − 𝐹𝑘(𝑡, 𝑥 = 1) log⁡(1 − 𝐹𝑘(𝑡, 𝑥 = 0) → The parameter does not have straightforward meanings No log⁡ 𝑠𝑑(𝑡, 𝑥 = 1) 𝜆𝑘 𝑠𝑑(𝑡, 𝑥 = 0) 𝜆𝑘 →derive sub-distribution hazard ratio Attriti on c Comp utation deman d Low No No Yes Using bootstr ap→ high Using bootstr ap→ high Cause-specific hazard measures the instantaneous rate ratio of the event[d] Direct association [Path 1] in the hazard ratio scale SAS, R “surv” package Good perform ance This model is used for prediction; but can obtain risk ratio/difference[i ] Risk difference/ ratio[i] Total association [Path 1+ Path 2] in RR/RD scale SAS, R “surv” package Good perform ance Total association [Path 1+ Path 2] + direct association [Path 1] in RR/ RD scale Good perform ance straightfor ward to implement by directly coding 135 Table A2.1 (cont’d) No With log link log⁡ 𝐹𝑘(𝑡, 𝑥 = 1) 𝐹𝑘(𝑡, 𝑥 = 0) →derive risk ratio Risk ratio[d] Total association [Path 1+ Path 2] in RR scale R “timereg” package Abs olute risk regre ssion 275 With logit link No log⁡ 𝐹𝑘(𝑡, 𝑥 = 1) 1 − 𝐹𝑘(𝑡, 𝑥 = 0) →derive risk ratio (when the risk of events is low, so OR≈RR) ≈Risk ratio[d] d Total association [Path 1+ Path 2] in RR scale R “timereg” package Yes Yes Too many covariat es may cause converg ence issue Some unident ified coding error Low (if not predict the margin al risk) Low (if not predict the margin al risk) Abbreviations: PH: proportional hazard assumption. a [d] Directly from parameter estimation; [i] from absolute risk prediction and then calculate the corresponding measure of associations b Path 1 and Path 2 refer to Figure 2.1. c Whether can correct the selection bias due to informative attrition. d When the absolute risk of events is low (e.g., <10%), odds ratio ≈ risk ratio. 136 APPENDIX 3: SUPPLEMENTAL MATERIAL FOR CHAPTER 4 Table A3.1 Age-adjusted population characteristics by baseline olfactory status (n=2,537) a Variable b Male sex Black race Study site of Pittsburgh c Education of >high school d Body mass index e 25-30 kg/m2 >30 kg/m2 Smoking status f Former & <30 pack-years Current or ≥30 pack-years Brisk walking of ≥90 min/wk General health status g Good Fair to poor Systolic blood pressure in mmHg Antihypertensive drug use Diabetes Depressive symptoms Heart rate in beats per minute LVH, n (%) Abnormal lung function Yes Missing Total cholesterol in mg/dL HDL-C in mg/dL Albumin in g/dL Interleukin 6 in pg/mL eGFR in mL/min/1.73m2 Prevalent major cardiovascular diseases Prevalent CHD Prevalent stroke Prevalent CHF Good (n = 845) 38.6 (35.4,41.9) 30.9 (27.9,34.1) 55.8 (52.4,59.1) 51.2 (47.8,54.6) Olfactory status Moderate (n = 867) 48.2 (44.9,51.6) 38.1 (34.9,41.4) 50.2 (46.9,53.5) 43.5 (40.2,46.8) Poor (n = 825) 58.5 (55.1,61.8) 46.6 (43.2,50) 47.7 (44.3,51.2) 38 (34.7,41.4) 43.5 (40.2,46.9) 23.7 (20.8,26.5) 42.1 (38.8,45.4) 26.8 (23.8,29.7) 41.2 (37.8,44.6) 21.5 (18.7,24.3) 27.8 (24.8,30.9) 21.1 (18.3,23.8) 11.9 (9.9,14.3) 25.1 (22.2,28) 29.9 (26.9,33) 9.3 (7.6,11.5) 26.1 (23.1,29.1) 31.8 (28.6,35) 8.6 (6.9,10.8) 34.4 (31.2,37.6) 15.6 (13.1,18) 136.1 (134.7,137.5) 58.9 (55.5,62.1) 21.5 (18.8,24.4) 10.3 (8.4,12.6) 64.3 (63.6,65.1) 11.6 (9.6,13.9) 8.6 (6.7,10.5) 9.7 (7.7,11.7) 208.3 (205.7,211) 54.8 (53.7,56) 4.00 (3.98,4.02) 3.3 (3.1,3.5) 79.6 (78.4,80.9) 40.5 (37.2,43.8) 15.2 (12.8,17.6) 135.4 (134,136.7) 61.5 (58.2,64.7) 24.7 (21.9,27.7) 12.4 (10.3,14.7) 64.8 (64.1,65.5) 11.2 (9.3,13.5) 11.9 (9.7,14) 9.3 (7.4,11.3) 204.2 (201.6,206.8) 53.3 (52.2,54.4) 3.98 (3.96,4.00) 3.3 (3.1,3.6) 80 (78.7,81.2) 38.1 (34.7,41.4) 22.4 (19.5,25.2) 134.9 (133.5,136.3) 58.5 (55.1,61.8) 26.6 (23.7,29.8) 13.5 (11.3,16) 65.8 (65.1,66.6) 11.7 (9.6,14.1) 12.8 (10.5,15.1) 13.3 (10.9,15.6) 203.7 (201,206.4) 53 (51.8,54.1) 3.98 (3.96,4.00) 3.5 (3.2,3.7) 75.9 (74.6,77.2) 23.9 (21.2,27) 8.3 (6.6,10.4) 4.4 (3.2,6) 23.7 (21,26.7) 8.3 (6.7,10.4) 5.1 (3.8,6.8) 24.4 (21.5,27.4) 7.5 (5.9,9.6) 4.2 (3.1,5.9) 137 Table A3.1 (cont’d) Abbreviations: IQR: inter-quartile range; HDL-C: high-density lipoprotein-cholesterol; LVH: left ventricular hypertrophy; eGFR: estimated glomerular filtration rate; CHD: coronary heart diseases; CHF: congestive heart failure; CI: confidence interval. a Linear regression for continuous variables, or logistic/ multinomial regression for categorical variables was used to calculate age-adjusted marginal means or percentage in each olfaction group, the average age of which was consistent with that of overall population as 75.6 years. b Continuous and categorical variables are presented as mean (95% CI) and % (95% CI), respectively. c Reference level of study site is Memphis. d Reference level of education is ≤ high school. e Reference level of BMI is <25 kg/m2. f Reference level of smoking status is never. g Reference level of general health status is very good to excellent. 138 Table A3.2 The association of baseline olfactory status with incident coronary heart diseases (CHD), stroke and congestive heart failure (CHF) after excluding prevalent cases of dementia or Parkinson’s disease a No. of Event Person- years Incidence (per 1,000 person-year) Model 1b Model 2 c Model 3d HR (95% CI) P HR (95% CI) P HR (95% CI) P Olfactory function CHD (n=1,718) Good Moderate 112 116 5906.00 5532.58 Poor 94 4078.08 Stroke (n=2,080) Good Moderate Poor CHF (n=2,160) Good Moderate Poor 83 70 73 123 165 137 7404.75 6927.00 5327.25 7496.92 6976.42 5403.42 19.0 21.0 23.1 11.2 10.1 13.7 16.4 23.7 25.4 Reference 1.06 (0.81,1.38) 1.11 (0.84,1.46) Reference 0.86 (0.63,1.19) 1.14 (0.82,1.58) Reference 1.37 (1.08,1.73) 1.41 (1.1,1.81) 0.667 0.477 0.374 0.431 0.009 0.006 Reference e 1.01 (0.77,1.31) 1.03 (0.78,1.38) Reference f 0.86 (0.62,1.19) 1.13 (0.81,1.58) Reference 1.33 (1.05,1.68) 1.37 (1.07,1.76) 0.963 0.817 0.354 0.459 0.019 0.014 Reference 1.32 (1.05,1.68) 1.29 (1.00,1.67) 0.020 0.051 139 Table A3.2 (cont’d) Abbreviations: HR: hazard ratio; 95% CI: 95% confidence interval a Associations were estimated from Cox cause-specific models with the robust sandwich standard error estimate to account for the competing risk of death. b Model 1 included age, sex, race, education and study site as covariates. c Model 2 further included smoking status, brisk walking, body mass index, self-reported general health status, systolic blood pressure, use of antihypertensive medications, diabetes, depressive symptoms, total cholesterol and high-density lipoprotein- cholesterol as covariates. For CHF, Model 2 further included prevalent CHD/stroke in addition to above covariates. d Model 3 (only for CHF) further included heart rate, left ventricular hypertrophy, abnormal lung function, albumin, interleukin 6 and estimated glomerular filtration rate. e Age category was stratified in the Cox model. f Brisk walking and antihypertensive medication use were stratified in the Cox model. 140 Table A3.3 Cause-specific hazard ratios and 95% confidence intervals a of each covariate in relation to congestive heart failure among all participants (n=2,421) and among participants who self-reported very-good-to-excellent health (n=1,100) Variable Categories Olfaction Age at baseline (year) Sex Race Study site Education Smoking status Brisk walking Body mass index Antihypertensive drug use Diabetes Depressive symptoms Total cholesterol (mg/dL) HDL-C (mg/dL) Prevalent coronary heart disease/stroke Heart rate (beat/minute) LVH Albumin (g/dL) eGFR (mL/min/1.73m2) Abnormal lung function General health status HR (95% CI) All participants b 1.32 (1.05,1.66) 1.28 (1.01,1.64) Those with very- good-to-excellent health c 1.40 (0.96,2.06) 1.70 (1.15,2.53) 1.04 (1.01,1.08) 1.04 (0.99,1.10) 1.03 (0.83,1.28) 1.10 (0.89,1.36) 0.81 (0.66,0.97) 0.8 (0.66,0.97) 0.99 (0.71,1.38) 0.93 (0.67,1.30) 0.72 (0.53,0.99) 0.80 (0.59,1.09) 1.09 (0.87,1.38) 1.12 (0.78,1.62) 1.32 (1.05,1.64) 1.53 (1.06,2.19) 0.73 (0.50,1.07) 0.74 (0.60,0.93) 0.9 (0.70,1.17) 0.73 (0.43,1.23) 0.77 (0.53,1.14) 0.93 (0.59,1.49) 1.44 (1.17,1.78) 1.45 (1.04,2.04) 1.24 (1.01,1.53) 1.22 (0.84,1.75) 0.95 (0.73,1.25) 0.95 (0.53,1.69) 1.00 (0.996,1.001) 1.00 (0.99,1.01) 1.00 (0.996,1.005) 0.99 (0.98,1.003) Moderate vs. good Poor vs. good − Male vs. female White vs. Black Memphis vs. Pittsburgh > high vs. ≤ high school Former & <30 pack-years vs. never Current or ≥30 pack-years vs. never ≥90 vs. <90 min/wk 25-30 kg/m2 vs. <25 kg/m2 >30 kg/m2 vs. <25 kg/m2 Yes vs. No Yes vs. No Yes vs. No − − Yes vs. No 1.65 (1.36,2.01) 1.63 (1.18,2.26) − Yes vs. No − − Yes vs. No Missing vs. No Good vs. very good to excellent 1.01 (1.00,1.02) 1.00 (0.99,1.02) 1.40 (1.08,1.83) 0.73 (0.54,0.99) 1.39 (0.87,2.20) 0.77 (0.46,1.29) 0.99 (0.98,0.99) 0.99 (0.98,1.00) 1.44 (1.09,1.90) 1.29 (0.81,2.06) 0.96 (0.71,1.30) 1.00 (0.60,1.66) 1.17 (0.95,1.45) / 141 Table A3.3 (cont’d) Fair to poor vs. very good to excellent 1.33 (1.02,1.74) / − − 1.01 (1.00,1.01) Systolic blood pressure (mmHg) Interleukin 6 (pg/mL) Abbreviations: CHF: congestive heart failure; HR: hazard ratio; 95% CI: 95% confidence interval; HDL-C: high-density lipoprotein-cholesterol; LVH: left ventricular hypertrophy; eGFR: estimated glomerular filtration rate. a The 95% confidence intervals were estimated using the robust sandwich standard error estimate. b Tertile of interleukin 6 was stratified in the Cox model. c The group of systolic pressure (cut-off as 140 mmHg) was stratified in the Cox model. Stratified variable Stratified variable 1.03 (0.99,1.07) 142 APPENDIX 4: SUPPLEMENTAL MATERIAL FOR CHAPTER 5 Methods We imputed missing frailty data and created 10 imputed datasets by using the random forest method with 100 iterations per imputation. In the imputation model, we included olfactory status, all the covariates in the primary analysis, the indicators of incident stroke and competing event of death and their corresponding Nelson-Aalen estimates of cumulative hazards, as well as additional variables that may be related to the missingness, including prevalent dementia, global cognitive function, and depressive symptoms. For each imputed dataset, we conducted the same analysis as the primary analysis and performed the statistical inference via bootstrapping with 300 samples. Finally, we used Rubins’ rule to obtain the pooled point estimates of risk ratios with good olfaction as the reference level and their pooled 95% confidence intervals at different time points. 143 4-Year b Reference 1.11 (0.69,1.79) 1.98 (1.26,3.16) Cause-specific hazard ratio (95% confidence interval) a by follow-up years Table A4.1 The period-specific associations of baseline olfactory status with incident stroke (n=5,799) Olfactory status Good Moderate Poor a Associations were estimated from the cause-specific Cox regression, adjusting for age, sex, race-center, education, APOE4 carrier, smoking status, body mass index, diabetes, systolic blood pressure, antihypertensive medication, total cholesterol, high-density lipoprotein-cholesterol (HDL-C), lipid lowering medication, atrial fibrillation, coronary heart disease, heart failure, estimated glomerular filtration rate, and frailty. b Quartiles of age were stratified in the model. c Quartile of HDL-C and frailty were stratified in the model. 8-Year c Reference 1.13 (0.84,1.54) 1.76 (1.31,2.38) 9.6-Year c Reference 1.09 (0.82,1.45) 1.61 (1.21,2.14) 6-Year c Reference 1.25 (0.88,1.8) 1.84 (1.3,2.62) Table A4.2 Marginal adjusted risk ratios of stroke comparing moderate/hyposmia/anosmia with good olfaction during the follow-up (n=5,799) Olfactory status Year 2 Reference 1.19 (0.63,2.27) 1.59 (0.85,3.15) 2.83 (1.63,5.23) Risk ratio (95% confidence interval) a by follow-up years Year 6 Reference 1.27 (0.95,1.82) 1.82 (1.24,2.69) 2.02 (1.43,2.97) Year 8 Reference 1.07 (0.82,1.48) 1.46 (1.06,1.96) 1.52 (1.08,2.07) Good Moderate Hyposmia Anosmia a Marginal adjusted risk ratio was calculated through the multivariable discrete-time Fine-Gray model; 95% confidence interval was obtained through bootstrapping with 300 samples. The model includes age, sex, race-center, education, APOE4 carrier, smoking status, body mass index, diabetes, systolic blood pressure, antihypertensive medication, total cholesterol, high density lipoprotein (HDL)-cholesterol, lipid lowering medication, atrial fibrillation, coronary heart disease, heart failure, estimated glomerular filtration rate, and frailty, plus two-way interaction terms between time and olfaction & education & HDL-cholesterol. Year 9.6 Reference 1.10 (0.80,1.53) 1.41 (1.02,1.96) 1.50 (1.06,2.08) Year 4 Reference 1.13 (0.73,1.74) 1.66 (1.06,2.59) 2.35 (1.52,3.91) 144 Figure A4.1 Marginal adjusted cumulative incidence function of stroke by 4-category olfactory status and the risk difference comparing moderate, hyposmia, anosmia with good olfaction. The cumulative incidence was estimated by the discrete-time sub- distribution hazard model, adjusting for covariates in Model 3 145 Table A4.3 Marginal adjusted risk ratios of stroke comparing moderate/poor with good olfaction among participants without dementia, Parkinson’s disease, and depressive symptoms (n=5,205) Olfactory status Year 2 Reference 1.32 (0.7,2.65) 2.21 (1.24,4.17) Risk ratio (95% confidence interval) a by follow-up years Year 4 Reference 1.19 (0.77,1.92) 1.92 (1.24,3.09) Year 8 Reference 1.12 (0.83,1.53) 1.52 (1.16,2.11) Year 6 Reference 1.31 (0.9,1.99) 1.84 (1.35,2.7) Good Moderate Poor a Marginal risk ratio was calculated through the multivariable discrete-time Fine-Gray model; 95% confidence interval was obtained through bootstrapping with 300 samples. The model includes age, sex, race-center, education, APOE4 carrier, smoking status, body mass index, diabetes, systolic blood pressure, antihypertensive medication, total cholesterol, high density lipoprotein (HDL)- cholesterol, lipid lowering medication, atrial fibrillation, coronary heart disease, heart failure, estimated glomerular filtration rate, and frailty, plus two-way interaction terms between time and olfaction & education & HDL-cholesterol. Year 9.6 Reference 1.17 (0.86,1.63) 1.52 (1.11,2.09) 146 Table A4.4 Marginal adjusted risk ratios of ischemic stroke comparing moderate/poor with good olfaction (n=5,799) Year 2 Reference 1.03 (0.52,2.15) 1.96 (1.13,4.12) Risk ratio (95% confidence interval) a by follow-up years Year 4 Reference 1.08 (0.67,1.7) 1.84 (1.22,2.95) Follow-up year Good Moderate Poor a Marginal risk ratio was calculated through the multivariable discrete-time Fine-Gray model; 95% confidence interval was obtained through bootstrapping with 300 samples. The model includes age, sex, race-center, education, APOE4 carrier, smoking status, body mass index, diabetes, systolic blood pressure, antihypertensive medication, total cholesterol, high density lipoprotein (HDL)- cholesterol, lipid lowering medication, atrial fibrillation, coronary heart disease, heart failure, estimated glomerular filtration rate, and frailty, plus two-way interaction terms between time and olfaction & education & HDL-cholesterol. Year 8 Reference 1.14 (0.83,1.58) 1.41 (1.09,1.93) Year 6 Reference 1.33 (0.91,1.92) 1.82 (1.32,2.66) Year 9.6 Reference 1.15 (0.83,1.56) 1.39 (1.06,1.95) Table A4.5 Marginal adjusted risk ratios of stroke comparing moderate/poor with good olfaction, after using multiple imputation (n=5,799) Olfactory status Year 2 Reference 1.18 (0.61,2.3) 2.13 (1.17,3.87) Risk ratio (95% confidence interval) a by follow-up years Year 4 Reference 1.13 (0.71,1.79) 1.97 (1.29,3) Year 9.6 Good Reference Moderate 1.1 (0.82,1.46) 1.44 (1.1,1.89) Poor a Marginal risk ratio was pooled from the results of 10 imputed datasets based on Rubin’s rule. For each imputed dataset, marginal absolute risks across olfactory statuses and risk ratios were calculated through the multivariable discrete-time Fine-Gray model; and their 95% confidence intervals were obtained through bootstrapping with 300 samples. The model includes age, sex, race-center, education, APOE4 carrier, smoking status, body mass index, diabetes, systolic blood pressure, antihypertensive medication, total cholesterol, high density lipoprotein (HDL)-cholesterol, lipid lowering medication, atrial fibrillation, coronary heart disease, heart failure, estimated glomerular filtration rate, and frailty, plus two-way interaction terms between time and olfaction & education & HDL-cholesterol. Year 8 Reference 1.06 (0.8,1.42) 1.48 (1.13,1.93) Year 6 Reference 1.27 (0.9,1.79) 1.9 (1.38,2.63) 147 Table A4.6 Adjusted marginal risk ratios a of stroke for common risk factors during the follow-up (n=5,799) Risk ratio (95% confidence interval) Year 4 1.98 (1.43,3.02) 1.83 (1.24,2.6) 1.99 (1.16,3.02) Year 2 2.14 (1.22,3.94) 1.84 (1.08,3.14) 2.33 (1.06,4.13) Follow-up year Poor vs. good olfaction CHD vs. no Atrial Fibrillation vs. no Abbreviation: CHD: coronary heart disease a Marginal risk ratio was calculated through multivariable discrete-time Fine-Gray model; 95% confidence interval was obtained through bootstrapping with 300 samples. To make the comparison comparable, the model included the interaction between the risk factor of interest and time. In addition, the model includes olfaction, age, sex, race-site, education, APOE4 carrier, smoking status, body mass index, coronary heart disease, heart failure, diabetes, systolic blood pressure, antihypertensive medication, total cholesterol, high density lipoprotein (HDL)-cholesterol, atrial fibrillation, lipid lowing medication, estimated glomerular filtration rate, and frailty, plus two-way interaction terms between time and education & HDL-cholesterol. Year 8 1.49 (1.17,2.00) 1.51 (1.15,2.02) 1.61 (1.10,2.21) Year 6 1.91 (1.43,2.77) 1.66 (1.21,2.29) 1.75 (1.14,2.45) Year 9.6 1.45 (1.16,1.95) 1.58 (1.14,2.19) 1.47 (1.01,1.97) 148 APPENDIX 5: SUPPLEMENTAL MATERIAL FOR CHAPTER 6 6-Year c Reference 1.37 (0.93,2.03) 1.65 (1.11,2.46) 4-Year b Reference 1.34 (0.82,2.22) 1.75 (1.07,2.91) Table A5.1 Olfactory status in relation to risk of coronary heart disease (n=5,142) using an alternative approach Cause-specific hazard ratio a (95% confidence interval) by years of follow-up Olfactory 9.6-Year e Status Reference Good 1.25 (0.93,1.68) Moderate 1.25 (0.91,1.72) Poor a Cause-specific hazard ratio was estimated from the cause-specific Cox proportional hazards regression, adjusting for age, sex, race-center, education, APOE4 carrier, smoking status, body mass index (BMI), diabetes, systolic blood pressure, antihypertensive medication, total cholesterol, high-density lipoprotein-cholesterol, lipid lowering medication, atrial fibrillation, stroke, heart failure, renal function, and frailty. b BMI is stratified in the model. c BMI and frailty are stratified in the model. d BMI and stroke are stratified in the model. Poor vs. good olfaction does not follow the proportional hazard assumption. e Race-center and stroke are stratified in the model. Poor vs. good olfaction does not follow the proportional hazard assumption. 8-Year d Reference 1.3 (0.96,1.77) 1.26 (0.91,1.76) Table A5.2 Four-category olfactory status in relation to risk of coronary heart disease (n=5,142) Olfactory status Year 2 Reference 1.52 (0.75,3.0) 1.71(0.64,3.84) 2.45 (0.9,5.84) Risk ratio a (95% confidence interval) by years of follow-up Year 6 Reference 1.34 (0.94,1.94) 1.61 (1.12,2.36) 1.56 (0.96,2.47) Year 8 Reference 1.32 (0.97,1.72) 1.25 (0.86,1.71) 1.20 (0.79,1.82) Good Moderate Hyposmia Anosmia a Marginal adjusted risk ratio was calculated through the multivariable discrete-time Fine-Gray model; 95% confidence interval was obtained through bootstrapping with 300 samples. The model includes age, sex, race-center, education, APOE4 carrier, smoking status, body mass index, diabetes, systolic blood pressure, antihypertensive medication, total cholesterol, high density lipoprotein- cholesterol, lipid lowering medication, atrial fibrillation, stroke, heart failure, renal function, and frailty, plus interaction terms between time and olfactory status. Year 9.6 Reference 1.15 (0.86,1.49) 1.06 (0.76,1.44) 1.10 (0.71,1.69) Year 4 Reference 1.49 (0.95,2.37) 2.14 (1.30,3.44) 1.93 (0.99,3.38) 149 Table A5.3 Olfactory status in relation to risk of coronary heart disease among participants without dementia (n=4,953) Risk ratio a (95% confidence interval) by years of follow-up Olfactory status Year 2 Reference 1.51 (0.7,3.44) 2.19 (1.10,4.72) Year 4 Reference 1.54 (0.96,2.57) 2.21 (1.43,3.43) Good Moderate Poor a Marginal adjusted risk ratio was calculated through the multivariable discrete-time Fine-Gray model; 95% confidence interval was obtained through bootstrapping with 300 samples. The model includes age, sex, race-center, education, APOE4 carrier, smoking status, body mass index, diabetes, systolic blood pressure, antihypertensive medication, total cholesterol, high density lipoprotein- cholesterol, lipid lowering medication, atrial fibrillation, stroke, heart failure, renal function, and frailty, plus interaction terms between time and olfactory status. Year 6 Reference 1.37 (0.94,2.15) 1.68 (1.12,2.5) Year 9.6 Reference 1.16 (0.85,1.59) 1.12 (0.81,1.57) Year 8 Reference 1.33 (1,1.83) 1.26 (0.9,1.74) 150 Figure A5.1 Stratified marginal adjusted risk ratios (aRRs) and 95% confidence intervals (CIs) of coronary heart disease (CHD) by a) age groups; b) sex; c) race; d) prevalent cardiovascular disease (CVD). * In subgroup of Black participants, due to the small number of incident events, the point estimate of year-2 RR was imprecise, so the data is not shown in the plot. The adjusted RR of CHD at year 4 was 2.91 (95% CI: 0.74, 3.1×108) for moderate olfaction and 3.4 (95% CI: 1.1, 2.6×108) for poor olfaction. ** In subgroup of participants with prevalent CVD, the adjusted RR of CHD at year 2 was 1.84 (95% CI: 0.23, 3.1×1012) for moderate olfaction and 1.79 (95% CI: 0.52, 3.9×1012) for poor olfaction 151 APPENDIX 6: SUPPLEMENTAL MATERIAL FOR CHAPTER 7 Figure A6.1 Marginal adjusted cumulative incidence of a) heart failure with reduced ejection fraction (HFrEF) and b) heart failure with preserved ejection fraction (HFpEF) by olfactory status. The cumulative incidence was estimated by discrete-time sub-distribution hazard model, adjusting for covariate in Model 3 152 Cause-specific hazard ratio (95% confidence interval) by follow-up years 4-Year b Reference 1.16 (0.83,1.62) 1.19 (0.85,1.68) Table A6.1 The period-specific associations of baseline olfactory status with incident heart failure (n=5,217) a Olfactory function Good Moderate Poor a Associations were estimated from Cox cause-specific models, adjusting for age, sex, race-center, education, self-reported general health status, smoking status, BMI, diabetes, hypertension, total cholesterol, high-density lipoprotein-cholesterol (HDL-C), lipid lowering medication, coronary heart disease, atrial fibrillation, renal function, and frailty. b Race-center was stratified in the model. c Race-center, BMI, quartile of HDL-C, and atrial fibrillation were stratified in the model. d Race-center, quartile of HDL-C, atrial fibrillation, and frailty were stratified in the model. e Quartile of HDL-C, atrial fibrillation, and frailty were stratified in the model. 8-Year d Reference 1.29 (1.04,1.61) 1.42 (1.13,1.78) 6-Year c Reference 1.21 (0.93,1.57) 1.28 (0.97,1.68) 9.6-Year e Reference 1.18 (0.97,1.45) 1.26 (1.02,1.56) Table A6.2 Adjusted marginal risk ratio a of heart failure for moderate/hyposmia/anosmia vs. good olfaction during the follow-up (n=5,217) Olfactory status Year 2 Reference 1.07 (0.65,1.65) 1.06 (0.59,1.98) 1.47 (0.89,2.58) Risk ratio (95% confidence interval) by follow-up years Year 6 Reference 1.22 (0.96,1.54) 1.09 (0.84,1.48) 1.26 (0.95,1.7) Year 8 Reference 1.22 (1.00,1.49) 1.15 (0.93,1.46) 1.28 (1.01,1.64) Good Moderate Hyposmia Anosmia a Marginal risk ratio was calculated through multivariable discrete-time Fine-Gray model; 95% confidence interval was obtained through bootstrapping with 300 samples. The model includes age, sex, race-center, education, self-reported general health status, smoking status, BMI, prevalent coronary heart disease, atrial fibrillation, diabetes, hypertension, total cholesterol, HDL-cholesterol, lipid lowering medication, and renal function, plus interaction terms between time and olfaction & body mass index & coronary heart disease & frailty. Year 9.6 Reference 1.07 (0.86,1.34) 1.01 (0.8,1.29) 1.16 (0.87,1.49) Year 4 Reference 1.16 (0.86,1.6) 0.97 (0.68,1.41) 1.36 (0.95,2.01) 153 No. of incident cases 141 160 167 Table A6.3 Adjusted marginal risk ratio a of acute decompensated heart failure for moderate/poor vs. good olfaction (n=5,217) Olfactory status Year 4 Reference Good 1.15 (0.81,1.64) Moderate Poor 1.27 (0.94,1.92) a Marginal risk ratio was calculated through multivariable discrete-time Fine-Gray model; 95% confidence interval was obtained through bootstrapping with 300 samples. The model includes age, sex, race-center, self-reported general health status, smoking status, BMI, prevalent coronary heart disease, stroke, diabetes, hypertension, total cholesterol, HDL-cholesterol, lipid lowering medication, and renal function, plus interaction terms between time and olfaction & body mass index & coronary heart disease & frailty. Risk ratio (95% confidence interval) by follow-up years Year 6 Reference 1.23 (0.94,1.61) 1.26 (0.98,1.65) Year 8 Reference 1.22 (0.97,1.49) 1.25 (1.01,1.56) Year 2 Reference 1.1 (0.64,1.94) 1.36 (0.89,2.37) Year 9.6 Reference 1.08 (0.86,1.33) 1.06 (0.8,1.35) Table A6.4 Adjusted marginal risk ratio a of heart failure for moderate/poor vs. good olfaction during the follow-up in participants without dementia (n=5,042) Olfactory status Year 2 Reference 1.08 (0.65,1.68) 1.23 (0.81,1.94) Risk ratio (95% confidence interval) by follow-up years Year 4 Reference 1.18 (0.88,1.55) 1.15 (0.84,1.52) Year 8 Reference 1.23 (1.04,1.52) 1.23 (0.99,1.5) Year 6 Reference 1.24 (0.99,1.6) 1.19 (0.95,1.52) Good Moderate Poor a Marginal risk ratio was calculated through multivariable discrete-time Fine-Gray model; 95% confidence interval was obtained through bootstrapping with 300 samples. The model includes age, sex, race-center, education, self-reported general health status, smoking status, BMI, prevalent coronary heart disease, stroke, diabetes, hypertension, total cholesterol, HDL-cholesterol, lipid lowering medication, and renal function, plus interaction terms between time and olfaction & body mass index & coronary heart disease & frailty. Year 9.6 Reference 1.07 (0.87,1.28) 1.09 (0.88,1.35) 154