OLFACTION IMPAIRMENT AND ITS ASSOCIATION WITH WEIGHT AND PROGRESSIVE WEIGHT LOSS IN OLDER ADULTS By Frank Daniel Purdy A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of Biostatistics - Master of Science 2018 ABSTRACT OLFACTION IMPAIRMENT AND ITS ASSOCIATION WITH WEIGHT AND PROGRESSIVE WEIGHT LOSS IN OLDER ADULTS By Frank Daniel Purdy Olfaction impairment (OI) is a very common, but underreported, sensory deficiency among older adults that can lead to a host of adverse health conditions, quality of life issues, and is a predictor of 5 - year mortality. Similarly, body composition and weight are driving factors of poor health among older populations, with malign effects including decreased mobility and sarcopenia. We theorize that older adults with OI both weigh less, and lose weight faster than those with without any impairment. Our goal is to examine how OI associated with weight and weight loss two ways: 1) cross - sect ional, where we assess the association during the year sense of smell was tested; and 2) longitudinally, where we inspect whether OI indicates more rapid weight loss over a seven - year period. The outcome variables of interest were total mass, total lean ma ss, total bone - free lean mass and total fat mass. Analyses were also performed in race - specific populations and in sex - specific populations. Data were obtained from the Health, Aging and Body Composition Study. The study found in the cross - sectional analys is that those with poor olfaction has significantly less total mass ( - 1.48 kg) and fat mass ( - 1.05 kg) than those with good olfaction. Longitudinally, those with poor olfaction showed a significantly faster decrease in total mass ( - 245.7 g/year), lean mass ( - 171.4 g/year) and bone free mass ( - 164.7 g/year) but not fat mass. Differing results were found in race - specific analyses but not in sex - specific analyses, with the decrease in weight appearing in whites but not blacks. Copyright by FRANK DANIEL PURDY 2018 iv Dedicated to Heidi, Story and Collins. For your support, love and understanding. v TABLE OF CONTENTS LIST OF TABLES ................................ ................................ ................................ ................................ vi LIST OF FIGURES ................................ ................................ ................................ ............................. vii INTRODUCTION ................................ ................................ ................................ ............................... 1 METHODS AND MATERIALS ................................ ................................ ................................ ............ 3 Study Population ................................ ................................ ................................ ................. 3 Anthropometr ic Outcomes ................................ ................................ ................................ . 4 Brief Smell Identification Test ................................ ................................ ............................. 4 Cova riates and Possible Confounders ................................ ................................ ................ 5 STATISTICAL ANALYSIS ................................ ................................ ................................ .................... 7 Summary Statistics ................................ ................................ ................................ .............. 7 Cross - sectional Analysis ................................ ................................ ................................ ...... 7 Longitudinal Analysis ................................ ................................ ................................ ........ 10 RESULTS ................................ ................................ ................................ ................................ ......... 13 Summary Statistics ................................ ................................ ................................ ............ 13 Cross - sectional Analysis ................................ ................................ ................................ .... 13 Longitudina l Analysis ................................ ................................ ................................ ........ 15 DISCUSSION ................................ ................................ ................................ ................................ ... 17 Outcome Evaluation ................................ ................................ ................................ ......... 17 Study Strengths ................................ ................................ ................................ ................. 19 Limitations ................................ ................................ ................................ ......................... 19 Future Analyses/Studies ................................ ................................ ................................ ... 20 CONCLUSION ................................ ................................ ................................ ................................ . 22 REFERENCES ................................ ................................ ................................ ................................ .. 23 vi LIST OF TABLES Table 1 Participant characteristics by olfaction score on B - SIT ................................ ...................... 8 Table 2 Whole coh ort summary statistics for outcome measures in Year 3 and Year 10 in kg ................................ ................................ ................................ ................................ .. 11 Table 3 Mixed model comparisons using likelihood ratio tests for Total Mass ........................... 12 Table 4 Multivariable linear regression results for mean response by olfaction group in for the whole cohort, race - specific analyses and sex specific analyses .......................... 14 Table 5 Mixe d model longitudinal regression results for linear trend of weight over time by olfaction group in for the whole cohort, race - specific analyses and sex specific analyses ................................ ................................ ................................ ............................ 16 vii LIST OF FIGURES Figure 1 Least square means by olfaction group for total mass ................................ ................... 10 Figure 2 Fitted total mass means per olfaction group ................................ ................................ .. 15 1 INTRODUCTION In older adults, olfaction impairment (OI) is a very common ailmen t that represents a crucial factor in many health issues , with studies indicating that the prevalence of OI increases with age . The prevalence of OI has been shown to be approximately 25% 1 in those aged 50 and older, but the prevalence increases to a rang e of 60% - 75% when looking at adults 80 and older. 1, 2 D isparities are present when examining OI by racial group, with the rate of OI in blacks approximately twice that of OI in similar white populations, 1, 3 ,4 even after controlling for demographics, socio - e conomic considerations and health statuses. Poor olfaction can lead to a host of health problems that adversely affect older populations, including quality of life decline , safety concerns 5 and dietary changes. 6 Perhaps even more importantly, s tudies have also revealed that olfaction has a significant impact on cognitive health, with poor olfaction regarded as a potential prodromal symptom for neuro degenerative diseases like an Alzheimer Disease , 7,8 and a risk factor for schizophrenia as well . 9 Indications also show that poor olfaction is a predictor of 5 - year mortality in older adults . 10 This makes OI one of the leading factors of ill health in older adults. However, there is still much regarding the relationship between olfaction impai rment and general health status in older populations that is unknown and need s to be explored further. This study will focus one area of potential concern, poor olfaction and its association with weight and weight loss. Weight , body composition, and weight loss have been found to be significant factors related to the general health status and quality of life within this population. Losses in body weight can have significant negative effects on mobility , 1 1 , 1 2 lean muscle mass 13 , 1 4 and bone mineral 2 density, 15 and is associated with many later - in - life conditions, including depression 16 and cognitive impairment . 17 It has also been theorized to be a potential non - cognitive indicator for 18 Additionally, weight loss (change >5%) can be an indic ator for an increased risk of mortality . 19,2 0 Given the combination of these effects and associations, a concerns when considering elderly health. Research exa mining the association between olfaction and weight changes in older population s have been limited, especially with regards to whether different race s and sex es are affected the same way . If poor olfaction can be shown to be an indicator for lower weight a nd weight loss, it would be an important step in understanding the connection between olfaction impairment and poor health in older adults. Therefore, this study aims to examine this relationship , using a biracial community - based cohort of older adults in the United States, in two ways : 1) a cross - sectional approach to explore the how poor olfaction is associated with weight and body composition and 2) a longitudinal approach that further scrutinizes whether poor olfaction can be predictive of weight loss. Both analyses will examine the overall population, as well as race - and sex - specific populations. 3 METHODS AND MATERIALS Study Population The Health, Aging and Body Composition Study (HABC) was created as a way to delve into the multitude of factors that lead to the decline of functions in older adults, with its focus on the change in body composition in relation to age, behavioral and physiological conditions. The study began in 1997 - 1998, enrolling a total of 3075 community - dwelling, healthy older a dults, aged 70 - 79 . The eligibility criteria for entry included a self - reported ability to climb 10 stairs and walk ΒΌ mile with no difficulty, no mobility - related problems in performing every - day tasks, cancer - free for the previous three years and no intent ion of changing residence for at least three years. The participants were recruited from two separate sites, Memphis, Tennessee and Pittsburgh, Pennsylvania and were 45% women and 33% African - American. Contact was made with participants on an annual or bi ennial basis through clinical and home visits through year 10 and then again at year 16. Further contact made by semiannual phone calls to update contact information and health status through year 15, and then quarterly until year 17 . Olfaction was examine d during the year 3 clinical visit using the 12 - item Brief Smell Identification Test (BSIT) . Body composition and weight measurements were assessed during annual clinical visits beginning in year 1, and then were subsequently switched to biennial clinic vi sits starting after year 6. Institutional Review Boards at the University of Tennessee, the University of Pittsburgh and the University of California - San Francisco approved all study protocols and informed written consents were obtained for all participant s. 21 4 Anthropometric Outcomes Four body composition measures of interest were identified that have shown to affect the health status of older adults: 1) t otal mass 12 (or whole body mass) , 2) total lean mass (also called fat - free mass), 3) total bone - free lean mass 22 and 4) total fat mass . 14 These a nthropometrics were collected annually for years 1 - 6, and then biennially for years 8 and 10, using whole body dual - energy X - ray absorptiometry (DXA). DXA is a non - invasive method that has the ability to measure both total body composition and regional subsections as well. Two separate x - rays are generated using different energy levels and then, through the comparison of the two results, accurate totals can be assessed for various body composition measures. 23 Tota l mass, lean mass and fat mass are obtained directly from the DXA while bone - free mass is calculated from the difference of lean mass and bone mineral content ( also obtained directly through DXA). Validation for DXA in gathering lean mass measurements 24 an d fat mass measurements 25 has been published previously. Brief Smell Identification Test The BSIT , an abbreviated version of the 40 - item Pennsylvania Smell Identification Test, is a widely - used 10,12,26 exam for olfaction in epidemiological studies and was performed during the year 3 clinical visit. It is a cross - cultural test that asks respondents to identify 12 common odors, via a scratch - and - sniff form, and choose the correct answer from a selection of four options. Each correct answer is given a point a nd the final score is the total of all correct answers, ranging from 0 - 12 , with a higher score indicating a better sense of smell . 27 Validation for this test of olfaction has been show in multiple studies. 27 , 28 5 For the purposes of this study, a score >10 will be considered within the good olfactio n group , a within the m oderate olfaction group , and any score <9 will be considered a case within the poor olfaction group . The BSIT was successfully performed on 2537 particip ants, and all the data collected from year 3 will be considered their baseline data. Covariates and Possible Confounders Other covariates to be considered in this study will include demographic information (age, sex, race, clinic site, height, education and family income), lifestyle choices (smoking status and physical activity ), and self - reported general health status and general appetite status. The demographic data was collected upon entry into the study in year 1 and remained unchanged when the olfact ion test was given in year 3. Education was categorized by three levels (less than high school, high school graduate, and postsecondary) and annual family income was categorized by four levels (less than 10k, between 10k and 25k, between 25k and 50k, and g reater than 50k). Lifestyle choices and the self - reported statuses were gathered in year 3. Smoking status has three levels (never smoker, former smoker, and current smoker), physical activity is binary based on whether the participant walks more than 90 m inutes per week at a brisk pace, health status has three levels (excellent/very good, good, and fair/poor) and appetite status has three levels (very good, good, and moderate to poor). We also controlled for diseases and conditions that are potential confo unders within our analysis , following definitions set by already published criteria within the HABC cohort. Disease status at baseline was considered a binary variable for all conditions. Cancer 2 9 (excluding melanoma) , cardiovascular disease 30 ( coronary he art disease, congestive heart failure, 6 cerebrovascular disease, and peripheral vascular disease) and cognitive conditions Disease and dementia) 31 were assessed and adjudicated by experienced physicians using hospital records, current medicatio n information, annual surveillance data and self - reported cases of doctor - diagnosed conditions. Specifically, through consensus agreement between two mobility disorder specialist s after examination of the relevant rec ords. Additionally, for cognitive conditions, the Modified Mini - Mental Status Examination (3MS) 32 was assessed in years 1, 3, 5, 8, and 10. Using this test (with scores ranging from 0 - 100), cases of cognitive dementia wer e assessed if 1) the 3MS score in y ear 1 was <80 or 2) the race - We also defined depressive symptoms 33 - question Center for Epidemiological Studies Depression (CES - D) scale. Diabetes 34 was assess ed through self - or through self - reported medication use. Finally, hypertension 35 was defined through self - reported diagnoses and medi cation use, or if systolic blood pressure 7 STATISTICAL ANALYSIS Summary Statistics Using the above anthropometric outcomes variables, covariates and confounders, we reduced the 2537 participant s who had taken the BSIT to a final analysis sample. We removed 136 participants due to either missing outcome variables or DXA readings that were invalidated due to artifacts found during screenings that prevented accurate measurements. Participants missi ng other covariates were also dropped from the analysis (each covariate had, at most, 7 missing data and some missing data overlapped). We believe this exclusion was justified due to the small percentage of missing values and does not affect the robustness or power of the analysis. A single exception was made regarding the variable for family income. This variable had 304 missing values, and we created a missing category in addition to the original 4 levels in order to ensure no further data was lost due th eir exclusion. The final sample size used in the statistical analysis was 2311 participants , with 1128 males and 1183 females, and 1438 white participants and 873 black participants. Analysis of variance was conducted to analyze the means of continuous va riables across the olfactory groups at baseline , while chi - square test s w ere performed to assess the proportion of subjects within categorical variables. (Table 1) Cross - sectional Analysis It is important to understand the relationship between OI and weigh t during the year that the give us the clearest picture regarding the expected body composition by olfaction group, it will 8 Table 1 Participant characteristi cs by o lfaction s core on B - SIT Poor: Score <9 N=753 Moderate: Score 9 and 10 N=785 Good: Score >10 N=773 p - value* Continuous Variables, Mean (SD) Age (years) 76.1(2.93) 75.6(2.91) 75.1(2.61) <.0001 Standing Height (m) 1.68(0.09) 1.66(0.10) 1.6 5(0.09) 0.0278 Daily Caloric Intake (Kcal) 1982.6(852.8) 1872.1(698.9) 1775.5(698.8) 0.0020 Categorical Variables, N (%) Sex Female Male 303(25.6) 450(39.9) 400(33.8) 385(34.1) 480(40.6) 293(26.0) <.0001 Race White Black 411(28.6) 342(39.2) 494(34.4) 291(33.3) 533(37.1) 240(27.5) <.0001 Clinic Site Pittsburgh Memphis 362(30.6) 391(34.6) 395(33.4) 390(34.5) 425(36.0) 348(30.8) 0.0023 Physical Activity < 90 mins/week 71(29.1) 682(33.0) 78(32.0) 707(34.2) 95(38.9) 678(32.8) .1501 Smoking Status Never Smoker Former Smoker Current Smoker 307(29.6) 79(44.9) 367(33.4) 349(33.7) 59(33.5) 377(34. 3) 380(36.7) 38(21.6) 355(32.3) 0.0002 Education Less than High School High School Graduate Postsecondary 242(45.9) 212(28.5) 299(28.8) 166(31.5) 269(36.2) 350(33.7) 119(22.6) 263(35.4) 391(37.6) <.0001 Family Income Less than 10k 10k to 25k 25k to 50k Greater than 50k Missing 98(42.1) 270(35.5) 203(29.9) 96(26.5) 86(31.4) 76(32.6) 242(31.8) 247(36.3) 117(32.2) 103(37.6) 59(25.3) 249(32.7) 230(33.8) 150(41.3) 85(31.0) 0.0002 Health Status Excellent/Very Good Good Fair/Poor 296(28.3) 287(33.1) 170(42.8) 354(33.8) 315(36.3) 116(29.2) 396(37.9) 266(30.7) 111(28.0) <.0001 Appetite Status Very Good Good Moderate to Poor 352(29.4) 288(35.0) 113(39.1) 422(35.2) 262(31.8) 101(35.0) 425(35.5) 273(33.2) 75(26.1) 0.0019 Diabetes at Baseline Yes No 189(35.4) 564(31.7) 187(35.1) 598(33.6) 157(29.5) 616(34.7) 0.0701 Hypertension at Baseline Yes No 564(32.8) 189(31.9) 597(34.7) 188(31.8) 558(32.5) 215(36.3) 0.2009 Cancer at Baseline Yes No 146(32.1) 607(32.7) 158(34.7) 627(33.8) 151(33.2) 622(33.5) 0.9278 CVD at Baseline Yes No 220(33.0) 533(32.4) 225(33.8) 560(34.0) 221(33.2) 552(33.6) 0.9574 9 Table 1 Participant characteristics by o lfaction s core on B - SIT Cont. Poor: Score <9 N=753 Moderate: Score 9 and 10 N=785 Good: Score >10 N=773 p - value * Dementia at Baseline Yes No 196(56.7) 557(28.4) 93(26.9) 692(35.2) 57(16.4) 716(36.4) <.0001 Depression at Baseline Yes No 126(38.2) 627(31.7) 116(35.2) 669(33.8) 88(26.6) 685(34.5) 0.0103 *P - values for continuous variables were obtained through analysis of variance. P - values for categorical variables were obtained through chi - square tests. be informative for comparing the differences between olfaction groups, a nd it will allow us to compare the race - and sex - specific associations. The analysis was completed by means of multiple linear regression, using data from year 3, for four outcomes: 1) total mass, 2) total lean mass, 3) total bone - free lean mass, and 4) to tal fat mass . Within the exposure variable, the group with the best sense of smell (olfaction score >10 ) was used as the reference group and all outcomes were adjusted to control for demographics, lifestyle choices, and health, appetite and disease status. The model , in an abbreviated form, can be written as: Dummy variables were created for categorical variables and where multiple The following reference levels were used race: black, sex: female, site: Pittsburgh, physical activity: <90mins /week , smoke: never smoker, education: less than high school, income: less than 10k, health status: fair - poor, appetite status: moderate - poor, all disease statuses: not present. 10 Least square m eans (or adjusted means) were estimated from the model and the difference between them were calculated as well. (Figure 1) Longitudinal Analysis As previously mentioned, weight loss in older populations is often considered a critical indicator of poor health. A longitudinal analysis will allow us to examine how weight changes over time for all olfaction groups and inspect how the rate of weight loss is different when assessed between separate groups . Again, race - and sex - specific analyses will be explored as well. For the analysis, year 3 was used as the baseline and year 10 as the final measurement , with the interest being the comparison of linear weight change over time between olfaction groups. To perform the analysis, a mixed model regression with random effects was constructed with estimations obtained through maximum likelihood, rather than the restricted maximum likelihood. This was done for two reasons: 1) It allowed for performances of the l ikelihood r atio t est (LRT) to examine whether par simonious models existed and 2) even though maximum likelihood produces biased estimates for the standard errors, this bias is reduced when sample Figure 1 Least square means by olfaction group for total mass 11 sizes are large enough. Our sample size (n=2311) should be sufficient. Fixed effects were defined as the same c ovariates which were used in the cross - sectional analysis (disease status was updated to include incident cases that occurred during the seven years of interest) and the outcomes examined remained the same as well. Additionally, year was added to fixed e ffects, as was the interaction between year and olfaction groups. Summary statistics for each outcome in Year 3 and Year 10 can be seen in Table 2. Table 2 Whole cohort s ummary statistics for outcome measures in Year 3 and Year 10 in kg Mean Standard Dev iation Range Year 3 Total mass 75.2 14.5 41.3 - 133.2 Total lean mass 48.6 10.2 25.8 - 84.5 Total bone - free lean mass 46.4 9.7 24.3 - 81.2 Total fat mass 26.7 8.5 7.0 - 67.3 Year 10 Total mass 72.4 14.4 35.3 - 126.3 Total lean mass 46.9 9.7 24.6 - 77.2 Total bone - free lean mass 44.9 9.4 23.5 - 74.8 Total fat mass 25.5 8.2 6.1 - 64.9 To develop a mixed model that was as reduced as possible, we began with the most complicated form : a full mean function, random effects for both intercept and slope and an unstructu red variance , and an unstructured covariance matrix that is not common to each olfactory group. Starting with this most complex model, LRTs were used to reduce the model beginning with the mean function, then turning to the random effects and covariance st ructures. A summary of th ese tests can be seen in Table 3 , using total mass as the anthropometric example . Similar assessments were made with similar results for the other outcomes. 12 Tabl e 3 Mixed model comparisons using likelihood ratio tests for Total M ass Chi - Square Statistic* DF P - value Mean Stucture Unstructured vs. Different linear average trends 18.2 12 0.6684 Different linear average trends vs Same linear average trends 5.5 1 0.0190 Random Effects Both intercept and slope vs Just inter cept No difference 1 NA Both intercept and slope vs Just slope 823.1 1 <.0001 Just intercept vs No random effects 3682.2 1 <.0001 Covariance Structure Common covariance vs. Separate covariance 5791.9 42 <.0001 Unstructured variance vs Toeplitz 331 .8 45 <.0001 Calculated from the difference of the - 2 log likelihood for the nested model and the full model Based on the table above, the most parsimonious mixed model will have a mean structure with different linear average trends, will include a ran dom intercept and have an unstructured variance structure for each of the olfaction groups. Written out, an abbreviated form would be represented by: , w here ij represents the ith subject at time j=3,4,5,6,8,10, denotes the model errors, represents similar co variates shared with the cross - sectional model and represents the random intercept. SAS 9.4 was used for all analyses. 13 R ESULTS Summary Statistics In our analysis, (Table 1) olfaction was associated with age, daily caloric intake , sex, race, se lf - reported health status , self - reported appetite status, and education level. Specifically, poor olfaction was an indicator of older age, a higher daily caloric intake, male sex, black race, a self - reported health status of fair/poor, a self - reported appe tite status of moderate to poor and lower education level. For the disease statuses included, associations with olfaction could be obse rved for dementia and depressive states , but it was no t associated for cardiovascular disease, hypertension , diabetes and cancer. Cross - sectional Analysis When examining the whole cohort, poor olfaction was significantly associated with lower total mass and lower fat mass when compared to the reference group (good olfaction) . (Table 4 ) Specifically, older adults in the poor olfaction group would have a total mass that, on average, was 1.48kg less than those in the good olfaction group, and a total fat mass that was 1.05kg less. No significance was determined when contrasting good and moderate olfaction scores for total mass or total fat mass, nor was there any significance when comparing any olfaction group for lean mass or bone - free lean mass. Interestingly, this relationship changes when we separate the analysis by race. When looking only at the white population of older ad ults, a poor sense of smell is significantly associated with lower expected mass for all four measures of body composition, when compared to those 14 with a good sense of smell. Conversely, the opposite is true when turn to a solely black population. No olf action group was found to have a significantly different expected mass for any of the four outcomes of interest. Even though the results are not signif icant, there are indications that total mass, lean mass, bone - free mass and fat mass might actually incre ase for those with poor olfaction , instead of decrease. The sex - specific analysis showed little variation when comparing olfactory groups. In fact, no significance was seen in any of the expected body composition masses , for either males or females , when comparing p oor olfaction to good olfaction. Table 4 Multivariable linear regression results for mean response by olfac tion group in for the whole cohort, race - specific analyses and sex specific analyses Whole Cohort Race Sex White N=1438 Black N=873 Male N=1128 Female N=1183 Estimate(SE) P - value Estimate(SE) P - value Estimate(SE) P - value Estimate(SE) P - value Estimat e(SE) P - value Outcome Measure Total mass (kg) B - SIT score <9 - 1.48(0.64) 0.0205 - 2.55(0.73) 0.0005 0.54(1.17) 0.6441 - 1.07(0.88) 0.2245 - 1.50(0.91) 0.0995 B - SIT score 9 and 10 0.06(0.60) 0.9191 - 0.52(0.67) 0.4393 0.95(1.15) 0.4709 0.04(0.88) 0.9624 0.08(0.82) 0.9263 B - SIT score >10 Ref Ref Ref Ref Ref Total lean mass (kg) B - SIT score <9 - 0.43(0.28) 0.1291 - 0.78(0.33) 0.0173 0.22(0.51) 0.6600 - 0.32(0.44) 0.4642 - 0.36(0.35) 0.3042 B - SI T score 9 and 10 0.001(0.27) 0.9966 - 0.11(0.30) 0.7210 0.11(0.50) 0.8272 - 0.17(0.44) 0.7078 0.09(0.31) 0.7657 B - SIT score >10 Ref Ref Ref Ref Ref Total bone - free lean mass (kg) B - SIT score <9 - 0.41(0.27) 0.1347 - 0.79(0.32) 0.013 8 0.28(0.50) 0.5776 - 0.32(0.43) 0.4602 - 0.33(0.34) 0.3286 B - SIT score 9 and 10 0.002(0.26) 0.9928 - 0.10(0.30) 0.7261 0.10(0.49) 0.8335 - 0.15(0.43) 0.7234 0.09(0.31) 0.7741 B - SIT score >10 Ref Ref Ref Ref Ref Total fat mass (kg) B - SIT score <9 - 1.05(0.42) 0.0126 - 1.77(0.48) 0.0003 0.31(0.77) 0.6813 - 0.75(0.54) 0.1695 - 1.14(0.63) 0.0708 B - SIT score 9 and 10 0.06(0.40) 0.8800 - 0.41(0.44) 0.3557 0.84(0.76) 0.2658 0.21(0.54) 0.7022 - 0.02 0.9751 B - S IT score >10 Ref Ref Ref Ref Ref Model fully adjusted for demographics (age, sex, race, height, education, clinic site), lifestyle choice (smoking status, phy sical activity, average daily caloric intake) and health factors (health status, appetite st atus, disease status) 15 Longitudinal Analysis When comparing the rates of change in total mass between the poor olfaction group and the good olfaction group, we f oun d that the poor olfaction group decline d more rapidly, by approximately 245g per year. This indicates that at the end of the seven years used in this study, a person with a poor sense of smell would have lost about 1.715kg more total mass than a good sense of smell. (Figure 2 ) Similarly, older adults in the mode rate olfaction group also showed a faster decrease (180.3g per year) in total mass than those in the good olfaction group . Poor olfaction was also found to be a predictor of faster weight loss for lean mass and bone - free lean mass, comparatively to good ol faction, but not for fat mass. (Table 5 ) Within our race - specific analysis, we found results that mirrored our cross - sectional analysis. W hite participants followed the same pattern as the overall cohort, with both poor and moderate olfaction groups show ing a significantly faster decline in their total mass when compared to the good olfaction group. Additionally, the same was true solely for poor olfaction when examining lean mass and bone - free lean mass. No associations were found regarding Figure 2 16 olfaction gro ups and the decrease of fat mass. For black participants, poo r olfaction had no significant e ffect on any body composition m easure of weight loss over time. Olfaction was not also associated with change in weight over time for any outcome when considering the male population. Women showed significance between poor olfaction and good olfaction for lean mass and bone - free mass, and between moderate olfaction and good olfaction for total mass and fat mass. Table 5 Mixe d model longitudinal regression results for linear trend of weight over time by olfaction group in for the whole cohort, race - specific analyses and sex specific analyses Whole Cohort Race Sex White N=143 8 Black N=873 Male N=1128 Female N=1183 Estimate(SE) P - value Estimate(SE) P - value Estimate(SE) P - value Estimate(SE) P - value Estimate(SE) P - value Outcome Measure Total mass (g) B - SIT score <9*Year - 245.7(93.6) 0.0087 - 189.8(83.9) 0.02 38 - 226.1(352.5) 0.5213 - 252.7(250.8) 0.3138 - 245.2(139.6) 0.0791 B - SIT score 9 and 10*Year - 180.3(87.2) 0.0387 - 176.2(73.2) 0.0161 - 201.3(280.3) 0.4726 - 128.6(248.7) 0.6052 - 274.1(105.4) 0.0094 B - SIT score >10*Year Ref Ref Ref Ref Ref Total lean mass (g) B - SIT score <9*Year - 171.4(44.3) 0.0001 - 1 32.6(49.4) 0.0073 - 159.8(140.7) 0.2561 - 103.9(124.4) 0.4035 - 92.6(42.8) 0.0307 B - SIT score 9 and 10*Year - 74.2(41.8) 0.0759 - 89.8(49.4) 0.0717 - 43.3(135.6) 0.7495 - 37.2(124.10 0.7651 - 44.8(34.6) 0.1956 B - SIT score >10*Year Ref Ref Ref Ref Ref Total bone - free lean mass (g) B - SIT score <9*Year - 164.7(42.3) <.0001 - 129.3(34.6) 0.0002 - 152.5(139) 0.2725 - 128.0(82.5) 0.1206 - 88.0(42.1) 0.0369 B - SIT score 9 and 10*Year - 78.1(40.8) 0.0555 - 79.4(41.0) 0.0530 - 38.7(124.8) 0.7565 - 67.5(83.2) 0.4172 - 43.8(35.0) 0.2104 B - SIT score >10*Year Ref Ref Ref Ref Ref Total fat mass (g) B - SIT score <9*Year - 92.1(77.6) 0.2355 - 97.9(78.9) 0.2147 - 82.5(184.9) 0.6553 - 125.9(150.5) 0.4029 - 205.6 (117.4) 0.0799 B - SIT score 9 and 10*Year - 107.6(74.6) 0.1493 - 74.2(58.1) 0.2012 - 140.4(181.2) 0.4384 - 100.1(117.9) 0.3959 - 196.6(94.3) 0.0373 B - SIT score >10*Year Ref Ref Ref Ref Ref Model fully adjusted for fixed effects demographics (age, sex, race, height, educat ion, clinic site, previous year weight measures), lifestyle choice (smoking status, physical activity, average daily caloric intake) and health factors (health status, appetite status, disease status). Random effects included intercept. Covariance structur e is unstructured across olfaction groups. 17 D ISCUSSION Outcome Evaluation In this biracial co hort of healthy, community - dwelling older adults, we examined the relationship between olfaction and baseline body weight, as well as weight change over a period of seven years. Our initial analysis found similar OI distribution differences between males a nd females and blacks and whites as in previously published studies , 1,3 with prevalence of OI among blacks higher than among whites and prevalence of OI among men higher than among women. Overall, we found that poor olfaction in older adults was associated with a lower total mass and lower fat mass, but not for lean mass or bone - free mass. Prospectively, poor olfaction is also an indicator of a quicker change in total mass, lean mass and bone - free lean mass when compared to good olfaction. This result is i mportant as loss of lean mass , especially relative to the retention of fat mass, is considered to be one of the greatest health risks to older adults, as other studies have shown that the deterioration of muscle can lead to a multitude of public health and quality of life issues. 14,22 By showing that this relationship between olfaction and the loss of lean mass exists, it brings us one step closer to fully understanding the implications poor sense of smell has on health in an older population. Stratum - speci fic examinations by race and sex allowed use to examine how the relationship between poor sense of smell and our body composition measures for each of the subgroups. For example, whites in both the cross - sectional and longitudinal analyses exhibited signif ican ce for almost every outcome (with the exception of change in fat mass over time in the 18 longitudinal analysis) when comparing poor olfaction and good olfaction. However, blacks lacked significance in all race - specific analyses. Other researchers have st udied the racial disparities of OI 1,3,4 but none have examined how those distinctions translate to body composition. by the loss or reduction of the sense of smell . If we examine the racial breakdown in our baseline population by family income, we see a disproportionate number of blacks who fall into the lowest income bracket of less than $10000 per year (~25%) while the number of whites in the same category is much lower (~4%). Income can be tied to the quantities and varieties of food which can be purchased, leading to a potentially different diet based on said income. Although caloric intake was accounted for in this analysis, different food groups and cooking met hods could lead to similar calorie counts while at the same time contributing to different However, diet may not be the only concern. In an animal trials using mice, researchers showed that a poor sense of smell result ed in a significant decrease in weight, even though the caloric intake was held constant among all mice. 3 6 be altered due to various metabolic disorders, including obesity, anorexia and diabetes. 36,37,38 This could indicate that olfaction is directly related to metabolic processes, although the physiological process of why this may occur is still not well known. 19 Study Strengths There are a few strengths to this study. First, t he Health ABC study is an extremely well - designed prospective cohort, with frequent points of data collection and contact with the study participants on a regular basis . The data addressed a wide range of topics and areas of interest, allowing us to examine a variety of covariates and control for potential confounders. A djudications by experienced professionals regarding disease status added to the reliability of the data used in this study. Potential bias due to inaccurate or missing information should be considered minimal. Additi onally, the large sample size (N=2311) allowed for any concerns regarding power to be mitigated. Even when splitting the analysis into race - and sex - based stratum, the number of study participants in each group remains large enough to ensure that the resul ts are meaningful and accurate. Limitations Although we believe our results to be valid, there are a few limitations to our current study. First, the olfaction test used in the HABC Study was given to participants whose mean age was ~75 at baseline. Even though OI increases with age, it would be appropriate to also examine younger individuals and track their olfaction decline to assess whether even earlier cases of OI could be predictive of weight and body composition problems as they age. Otherwise, the g eneralizability to a younger population would be limited and the results are only indicative of the association of olfaction and weight loss in an older population. 20 Secondly , sense of smell can change over time and the BSIT in this study was only administe red in a single year. Without additional tests, any mistakes or errors that occurred when testing could not be assessed or potentially even know. Further, w ithout repeated measurements for olfaction, we cannot examine how the temporal trend for sense of sm ell might affect weight and weight loss . Thirdly, the BSIT only tests whether or not a person can identify a particular smell. It does not for olfaction i mpairment and caution should be used in clearly defining different levels of impairment. Finally, the loss of sense of smell can be due to reasons other than aging, which the BSIT does not adequately account for. This study did not seek to determine whethe r these other causes influenced the measures of body composition. Future Analyses/Studies Even though the Health ABC study enrolled participants who were healthy at enrollment, the target population of those aged >70 are inherently at risk for negative he alth status changes. So even by year 3, when the baseline began for this analysis, many of the participants had had adverse health events. To counter the effects of these events, preliminary work on sensitivity a poor health status or who are suffering from a disease known to have a direct impact on both sense of smell and weight loss (or from its treatment, like cancer). This will allow for 21 conclusions to be drawn indicating that the association between poor olfaction and weight loss is not due to the poor health status of a participant at baseline. Another issue that arises with a population of older adults, especially in prospective studies, is how to account for mortality. An analysis covering the entire course of the study shows that by the end of the 17 th year, approximately 65% of the original participants had passed away, and - up. A joint survival/longitudinal analysis has been proposed to examine this issue and create a model that can express the relationship between the censoring caused by mortality while assessing olfaction and the change in body composition over time. As mentioned previously, animal studies have sho wn that olfaction may have a closer relationship to metabolic processes than previously supposed. The Health ABC study conducted a sub - study in years 2 and 3 which looks specifically at energy expenditure, in - depth food frequency questions and metabolic ra tes. Although the sample size is smaller than the original study (N=322), the subset of data could be very informative to explore the olfaction - metabolism relationship and assess whether results can be achieved that mirror the animal studies. Additionally, with the dietary information provided, an opportunity exists to examine whether difference in diets can adequately explain the differences found in this study among blacks and whites in regard to olfaction and weight. 22 CONCLUSION Poor olfaction in olde r adults has been shown to be associated with lower weight (for total mass and fat mass) and a more rapid decline of weight loss (for total mass, lean mass and bone - free mass) when compared to good olfaction groups . As these changes in body composition can lead to many health hazards, it i s important to understand the processes that lead to them and potentially find ways to identify those at greatest risk. Differences between whites and blacks offer new and interesting areas of research that could further e xplore the intricacies of how olfaction, and potentially diet and metabolism, can affect the weight and body composition of older adults. 23 REFERENCES 24 REFERENCES 1. Murchy C, Schubert CR, Cruickshan ks KJ, Klein BEK, Klein R. Nondahl DM. Prevalence of Olfactory Impairment in Older Adults. JAMA. 2002;288:2307 - 2312. 2. Doty RL, Shaman P, Applebaum SL, Giberson R, Siksorski L, Rosenberg L. Smell Identification Ability: Changes with Age. Science. 1984;22 6:1441 - 1443. 3. Dong J, Pinto J, Guo X, et al. The Prevalence of Anosmia and Associated Factors Among US Black and White Older Adults. Journals of Gerontology Series A - Biological Sciences and Medical Sciences . 2017;72:1080 - 1086. 4. Pinto J, Schumm L, Wro blewski K, Kern D, McClintock M. Racial Disparities in Olfactory Loss Among Older Adults in the United States. Journals of Gerontology Series A - Biological Sciences and Medical Sciences. 2014;69:323 - 329. 5. Croy I, Nordin S, Hummel T, et al. Olfactory diso rders and quality of life - an updated review. Chemical Senses. 2014;39:185. 6. Murphy C. The Chemical Senses and Nutrition in Older Adults. Journal of Nutrition For the Elderly. 2008;27:247 - 265. orders. Neurobiology of Disease. 2011;2012;46:527 - 552. 8. Fischer ME, Cruickshanks KJ, Schubert CR, et al. Age - Related Sensory Impairments and Risk of Cognitive Impairment. Journal of the American Geriatrics Society. 2016;64:1981 - 1987. 9. Brewer WJ, Wood SJ, McGorry PD, et al. Impairment of Olfactory Identification Ability in Individuals at Ultra - High Risk for Psychosis Who Later Develop Schizophrenia. American Journal of Psychiatry. 2003;160:1790 - 1794. 10. Pinto J, Kern D, Schumm L, McClintock M. Olfact ory Dysfunction Predicts 5 - Year Mortality in Older Adults. PLOS ONE. 2014;9:e107541. 11. Murphy RA, Patel KV, Kritchevsky SB, et al. Weight Change, Body Composition, and Risk of Mobility Disability and Mortality in Older Adults: a Population - Based Cohort Study. Journal of the American Geriatrics Society. 2014;62:1476 - 1483. 12. Lee JS, Kritchevsky S, Tylavsky F, et al. Weight change, weight change intention, and the incidence of mobility limitation in well - functioning community - dwelling older adults. Jour nals of Gerontology Series A - Biological Sciences and Medical Sciences. 2005;60:1007 - 1012. 25 13. Beavers KM, Beavers DP, Houston DK, et al. Associations between body composition and gait - speed decline: results from the Health, Aging, and Body Composition stud y. American Journal of Clinical Nutrition. 2013;97:552 - 560. 14. Koster A, Ding J, Stenholm S, et al. Does the Amount of Fat Mass Predict Age - Related Loss of Lean Mass, Muscle Strength, and Muscle Quality in Older Adults? Journals of Gerontology Series A - B iological Sciences and Medical Sciences. 2011;66:888 - 895. 15. Lloyd JT, Alley DE, Hochberg MC, et al. Changes in bone mineral density over time by body mass index in the health ABC study. Osteoporosis International. 2016;27:2109 - 2116. 16. Callen BL. Nutr itional screening in community dwelling older adults. International Journal of Older People Nursing. 2011;6:272 - 281. 17. Kitamura K, Watanabe Y, Nakamura K, et al. Weight loss from 20 years of age is associated with cognitive impairment in middle - aged and elderly individuals. PLOS ONE. 2017;12:e0185960. 18. Jimenez A, Pegueroles J, Carmona - Iragui M, et al. Weight loss in the healthy elderly might be a non - ONCOTARGET. 2017;8:104706 - 104716. 19. Newman AB, Yanez D, Harris T, et al. Weight Change in Old Age and its Association with Mortality. Journal of the American Geriatrics Society. 2001;49:1309 - 1318. 20. Corrada M, Kawas C, Mozaffar F, Paganini - Hill A. Association of body mass index and weight change wit h all - cause mortality in the elderly. American Journal of Epidemiology. 2006;163:938 - 949. 21. National Institute on Aging. Introducing the Health ABC Study: The Dynamics of Health, Aging, and Body Composition. https://healthabc.nia.nih.gov/ . Accessed Apri l 25, 2018. 22. Lee JS, Visser M, Tylavsky FA, et al. Weight loss and regain and effects on body composition: the Health, Aging and Body Composition Study. Journals of Gerontology. Series A: Biological Sciences & Medical Sciences. 2010;65A:78 - 83. 23. Las key MA. Dual - energy X - ray absorptiometry and body composition. Nutrition. 1996;12:45 - 51. 24. Visser M, Fuerst T, Lang T, Salamone L, Harris TB. Validity of fan - beam dual - energy X - ray absorptiometry for measuring fat - free mass and leg muscle mass. Health, Aging, and Body Composition Study - Dual - energy x - ray absorptiometry and body composition working group. Journal of Applied Physiology. 1999;87:1513 - 1520. 26 25. Salamone LM, Fuerst T, Visser M, Kern M, Lang T, et al. Measurement of fat mass using DEXA: a vali dation study in elderly adults. Journal of Applied Physiology. 2000;89:345 - 352. 26. Chen H, Shrestha S, Huang X, et al. Olfaction and incident Parkinson disease in US white and black older adults. Neurology. 2017;89:1441 - 1447. 27. Doty RL, Marcus A, Lee WW. Development of the 12 - item Cross - Cultural Smell Identification Test (CC - SIT). The Laryngoscope. 1996;106 :353 - 356 . 28. Doty RL. Olfactory dysfunction and its measurement in the clinic. World Journal of Otorhinolaryngology - Head and Neck Surgery. 2015;1 : 28 - 33 . 29. Izano M, Wei EK, Tai C, et al. Chronic inflammation and risk of colorectal and other obesity - related cancers: The health, aging and body composition study. Int J Cancer. 2016;138 :1118 - 1128. 30. Kalogeropoulos AP, Georgiopoulou VV, Murphy RA, e t al. Dietary sodium content, mortality, and risk for cardiovascular events in older adults: the Health, Aging, and Body Composition (Health ABC) Study. JAMA Internal M edicine. 2015;175 :410 - 419 31. Yaffe K, Freimer D, Chen H, et al. Olfaction and risk of dementia in a biracial cohort of older adults. Neurology. 2017;88 :456 - 462. 32. Teng EL, Chui HC. The Modified Mini - Mental State (3MS) examination. The Journal of Clinical P sychiatry. 1987;48 :314 - 318. 33. Andresen EM, Malmgren JA, Carter WB, Patrick DL. S creening for depression in well older adults: evaluation of a short form of the CES - D (Center for Epidemiologic Studies Depression Scale). American Journal of Preventive M edicine. 1994;10 :77 - 84. 34. Yaffe K, Falvey CM, Hamilton N, et al. Association Betwe en Hypoglycemia and Dementia in a Biracial Cohort of Older Adults With Diabetes Mellitus. JAMA internal medicine. 2013;173 :1300 - 1306. 35. Shea MK, Booth SL, Weiner DE, et al. Circulating Vitamin K Is Inversely Associated with Incident Cardiovascular Disea se Risk among Those Treated for Hypertension in the Health, Aging, and Body Composition Study (Health ABC). The Journal of N utrition. 2017;147 :888 - 895. 36. Riera CE, Tsaousidou E, Halloran J, et al. The Sense of Smell Impacts Metabolic Health and Obesity. Cell Metabolism. 2017;26:298 - 211.e5. 37. Islam M, Fagundo A, Arcelus J, et al. Olfaction in eating disorders and abnormal eating behavior: a systematic review. Frontiers in Psychology. 2015;6:1431. 27 38. Palouzier - Paulignan B, Lacroix M, Aime P, et al. Ol faction Under Metabolic influences. Chemical Senses. 2012;37:769 - 797.