EPIDEMIOLOGY OF COCAINE USE AND DEPENDENCE By German F. Alvarado A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Epidemiology 2012 ABSTRACT EPIDEMIOLOGY OF COCAINE USE AND DEPENDENCE By German F. Alvarado The following dissertation presents original research on selected topics in the epidemiology of cocaine use and dependence. The research consists of inter-related studies that deal with gaps in scientific knowledge about the topics pertinent to drug dependence. The dissertation is focused on four studies: In study one, I seek to estimate a suspected causal association between cocaine use and the occurrence of panic during adolescence and young adulthood, building upon prior study results based upon adult samples. The data are from an epidemiological sample of young adults, observed longitudinally from primary school to that time. The main finding is a modest but generally statistically robust association linking cocaine use and the occurrence of a panic attack-like experience. In study two, I attempt to estimate a suspected causal association linking earlier onset of cocaine use with later risk of newly incident panic attack in adults. The data are from an epidemiological sample of Americans aged 18-44 years at the time of crosssectional and retrospective assessment. The main findings of this study confirm prior evidence on the suspected causal association linking cocaine use with subsequent panic attack-like events. In this study, temporal sequencing of the association has been assured to the extent possible in retrospective age of onset data, with a conservative approach when the first cocaine and panic occurred in the same year. In study three, the aim is to estimate the risk of becoming cocaine dependent in the first months and years after onset of cocaine use, and to study suspected determinants of becoming dependent within 24 months after cocaine use starts, based on nationally representative samples of non-institutionalized citizens of the United States aged 12 years and older. The main findings of this study are as follows. First, an estimated 6-7% of recent-onset users develop cocaine dependence. Excess risk of cocaine dependence was found for females. Occurrence of cocaine dependence was greater among newly incident users who had started to smoke crack cocaine, as compared to those who had not smoked crack. No remarkable associations were found with the other variables under study (e.g., age, race-ethnicity). In study four, I seek to provide epidemiological evidence on predictors of who has started cocaine use in the last four years but has stopped using cocaine. The data for this research also are from nationally representative samples. The approach starts with exploratory analyses of data from calendar year 2003 (CY2003). Then, the model is reevaluated using data from an independent sample drawn in CY2008. The main findings are as follows. First, among those whose onset of cocaine use was within the four years prior to the date of the survey assessment, an estimated 13-16% had remained cocaine users within the prior 30 days; and an estimated 34% used cocaine within past 12 months but not in the past 30 days. An estimated 50-53% had not used cocaine in the previous 12 months before the assessment. It was found that people with age of cocaine onset at 20 years old or greater are twofold more likely to stop; this is the only finding from the CY2003 analysis that could be replicated in the CY2008 analysis. Cocaine cessation was not found to be reliably associated with sex, education, or using crack-cocaine. This dissertation is dedicated to God; to my wife, Yessika; to my parents, German & Ada; and my family and friends. iv ACKNOWLEDGEMENTS To my mentor, Professor James C. Anthony; to the committee members: Professor Naomi Breslau, Professor Chris E. Johanson, and Professor Hwan Chung; to NIH/NIDA/FIC; to the Faculty and Administrative Staff at the Department of Epidemiology at MSU; and to Professor Alejandro Llanos-Cuentas. v TABLE OF CONTENTS LIST OF TABLES ........................................................................................................... viii LIST OF FIGURES ............................................................................................................ x LIST OF ABBREVIATIONS ............................................................................................ xi CHAPTER 1: AIMS ........................................................................................................... 1 CHAPTER 2. BACKGROUND & SIGNIFICANCE ........................................................ 2  Section 2.1. Introduction and Overview to this Chapter ................................................. 2  Section 2.2. Introductory Material on Cocaine ............................................................... 2  Section 2.2.1 Introductory Material on Cocaine ............................................................. 2  Section 2.2.2 Epidemiology of Cocaine Use and Dependence. ...................................... 7  Section 2.3 Introductory Material on the Cocaine-Panic Association .......................... 20  Section 2.4. Introductory Material on Determinants of Becoming Dependent Soon After Onset of Cocaine Use. ......................................................................................... 25  Section 2.5. Introductory Material on Who Starts then Stops Cocaine. ....................... 27  Section 2.6. Summary and Anticipated Public Health Significance of the Dissertation Research ........................................................................................................................ 33 CHAPTER 3. RESEARCH MATERIALS AND METHODS......................................... 36  Section 3.1 Research Approach Details Specific to Results Chapter Section 4.1. Suspected Causal Association between Cocaine Use and Occurrence of Panic .......... 36  Section 3.2. Research Approach Details Specific to Results Chapter Section 4.2. New Evidence on the Suspected Causal Association Linking Earlier Cocaine Use with Later Occurrence of Panic .................................................................................... 41  Section 3.3 Research Approach Details Specific to Results Chapter Section 4.3. New Evidence on Risk of Becoming Cocaine Dependent: Epidemiological Estimates for the United States, 2002-2003 .................................................................................. 45  Section 3.4. Research Approach Details Specific to Results Chapter Section 4.4. Who Starts then Stops Cocaine? ................................................................................... 49 CHAPTER 4. RESULTS .................................................................................................. 54  Section 4.1. Suspected Causal Association between Cocaine Use and Occurrence of Panic .............................................................................................................................. 54  Section 4.2. New Evidence on the Suspected Causal Association Linking Earlier Cocaine Use with Later Occurrence of Panic ............................................................... 57  Section 4.3. New Evidence on Risk of Becoming Cocaine Dependent: Epidemiological Estimates for the United States, 2002-2003 ...................................... 58  Section 4.4. Who Starts then Stops Cocaine? ............................................................... 61 CHAPTER 5. DISCUSSION ............................................................................................ 66  Section 5.1. Discussion Specific to Results Chapter Section 4.1. Suspected Causal Association between Cocaine Use and Occurrence of Panic........................................ 66  vi Section 5.2. Discussion Specific to Results Chapter Section 4.2. New Evidence on the Suspected Causal Association Linking Earlier Cocaine Use with Later Occurrence of Panic ...................................................................................................... 71  Section 5.3. Discussion Specific to Results Chapter Section 4.3. New Evidence on Risk of Becoming Cocaine Dependent: Epidemiological Estimates for the United States, 2002-2003.......................................................................................................... 73  Section 5.4. Discussion Specific to Results Chapter Section 4.4. Who Starts then Stops Cocaine? .............................................................................................................. 79 CHAPTER 6. CONCLUSIONS ....................................................................................... 82  Section 6.1. Conclusions Specific to Results Chapter Section 4.1. Suspected Causal Association between Cocaine Use and Occurrence of Panic........................................ 82  Section 6.2. Conclusions Specific to Results Chapter Section 4.2. New Evidence on the Suspected Causal Association Linking Earlier Cocaine Use with Later Occurrence of Panic ...................................................................................................... 83  Section 6.3. Conclusions Specific to Results Chapter Section 4.3. New Evidence on Risk of Becoming Cocaine Dependent: Epidemiological Estimates for the United States. ............................................................................................................................ 83  Section 6.4. Conclusions Specific to Results Chapter Section 4.4. Who starts then stops cocaine? ............................................................................................................... 84 APPENDICES .................................................................................................................. 85  Appendix A: Tables to Accompany Chapter 4.1 .......................................................... 86  Appendix B: Tables to Accompany Chapter 4.2 .......................................................... 90  Appendix C: Tables to Accompany Chapter 4.3 .......................................................... 94  Appendix D: Tables to Accompany Chapter 4.4 ........................................................ 103 REFERENCES ............................................................................................................... 115  vii LIST OF TABLES Table 2.2.1. DSM-IV and ICD-10 criteria for diagnosis of cocaine dependence. ............. 7 Table 4.1.1. Characteristics of the baseline and follow-up samples in relation to occurrence of ‘panic’ in an urban public school system, 1985-2002. .............................. 87 Table 4.1.2. Estimated association for use of cocaine among ‘panic’ cases and controls in an urban public school system, 1985-2002 ................................................................... 88 Table 4.1.3. Subgroup variation in the association linking cocaine use with occurrence of ‘panic’ in an urban public school system 1985-2002 ................................................... 89 Table 4.2.1. The NESARC sample of adults, United States, 2000-2001. ....................... 91 Table 4.2.2. Estimated association linking earlier onset of cocaine use with later occurrence of panic attack-like experiences. Data from the NESARC adult sample, United States, 2001-2002. ................................................................................................. 92 Table 4.2.3. Subgroup variation in the association linking cocaine use with occurrence of panic.............................................................................................................................. 93 Table 4.3.1. Sociodemographic Characteristics of All Persons, All Recently Active Past-Onset Users, and the Subset of Recent-Onset Cocaine Users .................................. 95 Table 4.3.2. Selected Drug Use Characteristics of All Persons, All Recently Active Cocaine Users, and the Subset of Recent-Onset Cocaine Users ....................................... 97 Table 4.3.3. Relative Risk Estimates for Becoming Cocaine Dependent Among RecentOnset Cocaine Users, Without Statistical Adjustments .................................................... 98 Table 4.3.4. Relative Risk Estimates for Becoming Cocaine Dependent among RecentOnset Cocaine Users, with Statistical Adjustment for All Listed Covariates................. 101 Table 4.4.1A. Selected Sociodemographic Characteristics of All Persons, and the Subset of Cocaine Users ................................................................................................. 104 Table 4.4.1B. Selected Sociodemographic Characteristics of All Persons, and the Subset of Cocaine Users ................................................................................................. 105 viii Table 4.4.2A. Relative Risk Estimates for Stopping Cocaine Use Among Cocaine Users, Without Statistical Adjustments .......................................................................... 106 Table 4.4.2B. Relative Risk Estimates for Stopping Cocaine Use Among Cocaine Users, Without Statistical Adjustments .......................................................................... 107 Table 4.4.3. Relative Risk Estimates for Stopping Cocaine Use, with Statistical Adjustment. ..................................................................................................................... 108 Table 4.4.4A. Selected Sociodemographic Characteristics of All Persons, and the Subset of Cocaine Users ................................................................................................. 109 Table 4.4.4B. Selected Sociodemographic Characteristics of All Persons, and the Subset of Cocaine Users ................................................................................................. 110 Table 4.4.5A. Relative Risk Estimates for Stopping Cocaine Use Among Cocaine Users, Without Statistical Adjustments .......................................................................... 111 Table 4.4.5B. Relative Risk Estimates for Stopping Cocaine Use Among Cocaine Users, Without Statistical Adjustments .......................................................................... 112 Table 4.4.6. Relative Risk Estimates for Stopping Cocaine Use, with Statistical Adjustment. ..................................................................................................................... 113  ix LIST OF FIGURES Figure 2.2.2 The main rubrics of epidemiology (reproduced with permission of copyright holder James C. Anthony, 2003)......................................................................................... 9 Figure 2.2.3 Use of cocaine in 2008 ................................................................................. 11 Figure 2.2.4 Annual prevalence of cocaine use among the adult population in selected European countries............................................................................................................ 12  x LIST OF ABBREVIATIONS APA American Psychiatric Association CI Confidence Interval CY Calendar Year DSM Diagnostic and Statistical Manual of Mental Disorders HCl Cocaine Cocaine Hydrochloride ICD International Classification of Diseases IRB Institutional Review Board IV Intravenous MSA Metropolitan Statistical Area NESARC National Epidemiologic Survey on Alcohol and Related Conditions NIAAA National Institute on Alcohol Abuse and Alcoholism NIDA National Institute on Drug Abuse NIMH National Institute of Mental Health NSDUH National Survey on Drug Use and Health OR Odds Ratio PRC Prevention Research Center PSU Primary Sampling Unit RR Relative Risk SAMHSA Substance Abuse and Mental Health Services Administration xi SE Standard Error UCHRIS University Committee on Research Involving Human Subjects UNODC United States Office of National Drug Control Policy US United States WHO World Health Organization xii CHAPTER 1: AIMS The following dissertation presents original research on selected topics in the epidemiology of cocaine use and dependence. The research consists of inter-related studies that address gaps in scientific knowledge about the topics pertinent to drug dependence, and promises publication of three or more scientific articles in peerreviewed journals. The dissertation is focused on four aims. Aim 1: To estimate a suspected causal association between cocaine use and the occurrence of panic during adolescence and young adulthood, building upon prior study results based upon adult samples. The data are from an epidemiological sample observed longitudinally from primary school to adulthood. Aim 2: To estimate a suspected causal association linking earlier onset of cocaine use with later risk of newly incident panic attack in adults. The data are from an epidemiological sample of Americans aged 18-44 years at the time of cross-sectional and retrospective assessment. Aim 3: To estimate the risk of becoming cocaine dependent in the first months and years after onset of cocaine use, and to study suspected determinants of becoming dependent within 24 months after cocaine use starts, based on nationally representative samples of non-institutionalized citizens of the United States aged 12 years and older. Aim 4: To provide epidemiological evidence on predictors of who has started cocaine use in the last four years but has stopped using cocaine. The data for this research also are from nationally representative samples. 1 CHAPTER 2. BACKGROUND & SIGNIFICANCE Section 2.1. Introduction and Overview to this Chapter To begin, this chapter conveys background information on cocaine as well as on epidemiology of cocaine use and dependence. Before its end, the chapter will provide an introduction to the new research that is the motive of the dissertation and will fill some of the gaps in scientific literature on epidemiology of cocaine use and dependence. Section 2.2. Introductory Material on Cocaine [Note from the author: Some of this section of the dissertation report has been published in Alvarado GF, Anthony JC. [Epidemiology of cocaine use and dependence], Tratado Set de Trastornos Adictivos, Madrid. Editorial Médica Panamericana, 2006].  Section 2.2.1 Introductory Material on Cocaine A coca leaf (Erythroxylon coca) has a concentration of the cocaine hydrochloride (HCl) compound that is one percent when it is plucked from the coca bush. The coca leaf has been chewed for at least 3000 years by populations living in the Andean region of South America. The evidence from investigation of ancient sites is consistent with use of coca compounds in religious ceremonies and to avoid fatigue (Cartmell 1991, Indriati 2001, Rivera 2005). 2 In 1860, an important change in the use of coca compounds occurred when Dr. Albert Neimann of Gottingen University (Germany) extracted pure cocaine hydrochloride (HCl) from coca leaves. Rapidly, commercial products with cocaine were produced and their consumption expanded, mainly in Europe and North America. Dr. Sigmund Freud, in the book Über Coca published in Germany in 1883, was one of the first physicians to propose using cocaine for therapeutic reasons. Almost at the same time in United States, the famous surgeon Dr. William Halsted from Johns Hopkins University proposed using cocaine as an anesthetic (Olch 1975). Unfortunately, Halsted himself later became cocaine dependent (Olch 1975; Schneck 1988). Outside of the medical environment, several products were produced for mass consumption. Among these, Coca Cola in the United States and Vin Mariani in Europe are well known (Musto 1998). According to medical historian David Musto (1989), at the beginning of the twentieth century, North America experienced its first epidemic of cocaine use and cocaine dependence. Concern with respect to the negative consequences of cocaine use led to restrictions of this compound in many countries, including the United States. Coca Cola changed its formula and no longer contains active cocaine as an ingredient. The coca leaf used to make Coca Cola has been decocainized via chemical extraction (Petersen 1977). After the early 20th century regulations were issued, extra-medical cocaine use in the US was driven underground and became a forgotten drug for most Americans, although some medically prescribed use of cocaine persisted (e.g., in anesthesiology). A 3 well documented second cocaine epidemic took place the United States during the 1970s and the 1980s (Cornish 1996). Even though population prevalence of cocaine use declined markedly in the 1990s and is no longer at an especially elevated epidemic level, there is concern about a potential third cocaine epidemic in the 21st century in America, as well as concern about recent growth of cocaine use in countries of Europe (UNODC 2010). This situation encourages more research that will increase our understanding of the epidemiology of cocaine use and suspected or known hazards of its use, in anticipation of public health action to prevent new epidemics. With respect to cocaine pharmacology, when pure cocaine is extracted from coca leaves, after processing with different chemical products, an intermediate product –coca paste- is produced and there are two processed final chemical forms: cocaine hydrochloride (HCl) and cocaine base. When cocaine HCl is combined with an alkali (e.g. sodium bicarbonate or ammonia), it becomes a base; and after heating, pieces appear that are called ‘crack cocaine’, those pieces are suitable for smoking (Hatsukami, 1996). Use of cocaine HCl is typically via intranasal insufflations, or via intravenous (IV) or intramuscular injection route. Cocaine HCl is not suitable for smoking because it decomposes with the high temperatures needed for vaporization. Cocaine base is heated and the fumes are inhaled by mouth, as in ‘crack smoking’, a cheaper way of consuming cocaine, which appeared in the early 1980s (Cornish, 1996; Reinarman, 2004). The intranasal route can be inefficient, with a slow absorption rate, while inhalation of fumes and IV injection have the fastest rate. For illustration, peak venous concentration of cocaine is reached at 40 minutes after nasal insufflations whereas the peak after IV injection or after smoking crack maybe as fast as five minutes; also, subjective effects are 4 greater after smoking; if we add the effect of the price and convenience, then, the abuse liability is higher for crack cocaine (Hatsukami, 1996). Accordingly, the route of administration may be important in the prediction of dependence and associated problems as explained later in this chapter. In terms of neuropsychopharmacology, cocaine has a remarkable range of actions and effects, and serves reinforcing functions in a range of species, including humans and non-human primates (Johanson 1989, Carrera 2004). Most prominently at the neuropsychopharmacological level, cocaine blocks a dopamine transporter. Thereafter, dopamine accumulates in the synaptic cleft. Cocaine also can inhibit reuptake of 5hydroxytryptamine3 (5HT3). It has also effects in the binding sites of the sodium dependent transport area for dopamine and serotonin (as separate mechanisms from its reuptake). Among other neuropsychopharmacological actions, cocaine can block sodium channels interfering with the propagation of action potentials; its utility as a local anesthetic is related to this effect (Olch 1975, Wright 1997, Columb 2010). Cocaine is metabolized in the liver and excreted in the urine. Metabolites can be detected in urine (e.g., benzoylecgonine) and in hair (Barroso 2009). After cocaine administration, there are rapid effects in the brain (via the multiple pathways described above, including dopaminergic, serotoninergic and noradrenergic tracts). Use of cocaine is known to produce increases in heart rate, blood pressure, shortduration euphoria, increased energy, enhanced feelings of self confidence, and higher sensorial awareness. As far as adverse effects, paranoia has been described and cocaine may also cause hyperthermia, stroke, and heart attack. 5 There is good reason to investigate cocaine use as a cause of psychiatric disturbances. Its effects on mood have been recognized since the 19th century, when the term cocainomania was introduced, and described with clinical features resembling manic episodes (Tuke 1892). A more recently recognized form of psychiatric disturbance involves a cocaine dependence syndrome. For many years, it was thought that cocaine was an efficacious reinforcer of behavior, but with no ‘addiction’ or drug dependence consequences (Tatum & Seevers 1929, Campbell 2007). The perception of cocaine as a relatively safe drug (as compared to heroin) was due, in part, to an insistence that there should be a withdrawal syndrome of the type seen with abrupt discontinuation of opioid drug taking. This conceptualization changed in the 1950s and 1960s (Campbell 2007), but the American Psychiatric Association (APA) still had not included ‘cocaine dependence’ in the Diagnostic and Statistical Manual of Mental Disorders Third Edition (DSM-III 1981). This situation changed by 1987 when the DSM-III-R was published. Contemporary diagnostic criteria for cocaine dependence are presented in Table 2.2.1. 6 Table 2.2.1. DSM-IV and ICD-10 criteria for diagnosis of cocaine dependence. DSM-IV-T-R Substance Dependence Criteria Three (or more) of the following, occurring any time in the same 12-month period: 1. Tolerance 2. Withdrawal 3. The substance is often taken in larger amounts or over a longer period than was intended 4. Persistent desire or unsuccessful attempts to cut down or control substance 5. Time spent in activities necessary to obtain / use the substance, and recover from its effects 6. Important social, occupational, or recreational activities given up or reduced because of use of the substance 7. Use is continued despite the knowledge of having a persistent or recurrent physical or psychological problem that is likely to have been caused or exacerbated by the substance ICD-10 DIAGNOSIS OF DEPENDENCE should be made if three or more of the following have been experienced or exhibited at some time during the previous year: 1. A strong desire or sense of compulsion to take drugs. 2. Subjective awareness of an impaired capacity to control drug taking behavior in terms of its onset, termination, or levels of use 3. Narrowing of the personal repertoire of patterns of drug use 4. Progressive neglect of alternative pleasures of interests in favor of substance use 5. Persisting with drug use despite clear evidence of overtly harmful consequences 6. Evidence of tolerance such that increased doses of the substance are required in order to achieve effects originally produced by lower doses 7. A physiological withdrawal state 8. Substance use with the intention of relieving withdrawal symptoms and with awareness that this strategy is effective Section 2.2.2 Epidemiology of Cocaine Use and Dependence. As previously published in a Spanish language book chapter (Alvarado & Anthony 2006), epidemiology can be a support to investigators who wish to know the answers to central questions within the domain of five general rubrics: 7 (1) Quantity: In the community at large, how many are affected? Under this rubric we estimate incidence and prevalence (e.g., proportion of persons with currently active cocaine dependence, risk of becoming cocaine dependent). (2) Location: Where are the affected cases more likely to be found? This rubric, applied to cocaine, refers to the distribution of cocaine use and cocaine dependence in relation to characteristics of person, place, and time. (3) Causes: What accounts for some people in the community becoming cases whereas others do not? At this point, the epidemiologist must draw upon concepts, principles and methods of causal inference. (4) Mechanisms: What linkages of states and processes influence who becomes and remains a case? Under this rubric we research on natural history, clinical course, and consequences of cocaine use, case fatality rates, including co-morbidity and disability. (5) Prevention and Control: What can be done to prevent and intervene? This rubric encompasses prevention and intervention studies needed to reduce risk, delay onset, shorten duration, or remediate and rehabilitate (Anthony & Van Etten, 1998). In actual practice, epidemiologists can combine this rubric-orientation with the ecological concept of ‘scale.’ For example, as illustrated in Figure 2.2.2, scale can run from the microcosm (e.g., genetic polymorphisms) to the macrocosm (social groups, society in general), adopted from Anthony (2003), and applied here to cocaine epidemiology. An important task for epidemiologists who study cocaine use and cocaine dependence is to in fill this type of table, making research contributions within each cell of the table. 8 In this section I provide selected examples of research contributions, drawn from various cells of the table, as these cells are applied to the epidemiology of cocaine use and cocaine dependence. Scale from Microcosm to Macrocosm The Main Rubrics 1 Mechanisms 5 H Causes 4 G Location 3 Nations & Global Regions Quantity 2 Genes & Individual Social Simple Organisms Groups Gene Products A B C D E F Prevention & Control Figure 2.2.2 The main rubrics of epidemiology (reproduced with permission of  copyright holder James C. Anthony, 2003).    9 • In the community, how many people are cocaine users and cocaine dependent? The United Nations seeks global estimates for the number of users of each internationally regulated drug. With respect to cocaine, based on the data available to it (e.g., see Figure 2.2.3), the most recent UN estimate describes a range of values, with an expectation that the true number of active cocaine users in the world falls between 1,020,000 users and 2,670,000 users (UNODC, 2010). If cocaine dependence estimates from the United States are applied to these numbers, then the number of active cocaine dependence cases, globally, might be approach 500,000 (i.e., if one-sixth of users have become and remained cocaine dependent during active use of the drug, as described below). • Where are cocaine users and cocaine dependence cases more likely to be found? Coca leaf has been ingested in the western hemisphere for well over 3000 years, but the first reports of cocaine-related toxicity did not emerge until the late 19th century, within decades of cocaine hydrochloride (HCl) extraction from coca leaf. In recent years, cocaine use has been a behavior of late adolescence and early adulthood, with peak risk of starting to use cocaine observed between age 18 years and 22 years (Wagner & Anthony, 2002). The United States, to date, has been the country with greatest public concern about cocaine use and dependence. For this reason, most of the available epidemiological evidence about cocaine is based upon studies conducted in the United States. For example, in the US, an epidemic of newly incident cocaine use reached its peak in the early 1980s, with annual incidence rates declining to endemic plateau values in recent years. To illustrate, in recent years, the estimated number of newly incident cocaine users 10           Figure 2.2.3 Use of cocaine in 2008  For interpretation of the references to color in this and all other figures, the reader is referred to the electronic version of this dissertation. The image is for visual reference only but not meant to be readable. 11   Figure 2.2.4 Annual prevalence of cocaine use among the adult population in selected European countries.  in the United States has been roughly 600,000 (SAMHSA, 2009); the corresponding estimate for 2002 was 1 million users (SAMHSA, 2002). With respect to the occurrence of cocaine dependence, based upon recent national sample surveys in the United States, roughly one in six US residents age 12 years and older has tried cocaine at least one time, and among these newly incident users, approximately one in six has become cocaine dependent (Anthony & Helzer, 2002). Cumulative occurrence of cocaine dependence in the population has been estimated to be roughly three per cent. (Anthony et al, 1994; Anthony & Chen, 2003). Estimates from multiple sources indicate that for every treated case of cocaine dependence, there are at least three untreated cases with similar clinical features (e.g., see Anthony & Helzer, 2002). With respect to risk of developing the cocaine dependence syndrome, also estimated for the US and not yet from other countries, during the late 1980’s and the 1990’s, peak incidence rate estimates for cocaine dependence have been found at age 23-25. Once use began, cocaine dependence emerged early, with an estimated 5%-6% of cocaine users becoming cocaine dependent in the 12-24 months of use (Wagner & Anthony, 2002; O’Brien & Anthony, 2005). Most of the observed cases of cocaine dependence meet criteria for dependence within three years after initial cocaine use; almost 15-20% of cocaine users developed cocaine dependence within 10 years of first cocaine use (Wagner & Anthony 2002, Lopez-Quintero 2010). An increasing number of epidemiologic studies have started to produce prevalence estimates for other parts of the world (i.e, United Kingdom, Australia, Germany, Spain), as well as in Latin American countries (i.e, Brazil, Panama). To illustrate, cumulative occurrence of cocaine use (including coca paste use) was between 1%-4% in a survey of school attending teenagers in seven countries of Central America (Dormitzer et al., 2004). In a recent general community survey in Brazil, Galduroz and colleagues estimated that 2.3% of the community population had tried cocaine at least one time (Galduroz 2005). The global knowledge base about cocaine epidemiology may start changing with publication of results from the recent World Mental Health Surveys (Degenhardt et al., 2008; Demyttaenaere et al., 2004), and as cocaine spreads in Brazil (Galduroz 2004), and Europe (Haasen 2004, UNODC 2010), and in other countries such as South Africa or Nigeria (UNODC 2010). Degenhardt and colleagues (2008), among others (Michaud 2006, Schifano 2008) have documented recent increases in the occurrence of cocaine use in Europe, perhaps fueled by shifts in overseas trafficking of cocaine from source countries in South America, via intermediated transit countries of West Africa or directly to destinations with growing demand. The UN drug report depicts estimated cocaine use prevalence estimates for selected countries of Europe, as shown in Figure 2.2.4 Another facet of ‘location’ is neighborhood of residence. In an epidemiological sample, researchers found that youths living in the most disadvantaged neighborhoods (highest tertile of disadvantage) were 5-to-6 times more likely to have been offered a chance to try cocaine (Crum, Lillie-Blanton & Anthony, 1996). Other aspects of cocaine dependence also follow a socio-economic status gradient. For example, perceived risk of cocaine and experience with cocaine have been found to cluster within neighborhoods in the US (Petronis & Anthony, 2000). The same research team, using a model known as the ‘alternating logistic regressions’ (which estimates a pairwise cross product ratio, 14 PWCPR), also reported the first quantitative estimates of the magnitude of spatial cooccurrence of incident experience with cocaine at two levels of analysis: neighborhoods and cities (Petronis KR, Anthony JC, 2003). This research does not explain ‘why’ the cluster occurs, but it helps to quantify locations (e.g. impoverished neighborhoods) where cocaine involvement is more likely to be observed in clusters. • Why do some people in the community become cocaine users and cocaine dependent while others do not? Studies of monozygotic and dizygotic twins illuminate two facts about risk of becoming cocaine dependent. First, a good part of the variation in risk of becoming cocaine dependent can be traced to still-unknown genetic influences. Second, non-shared environmental circumstances, conditions, and processes surely play a role in determining why some people become cocaine dependent whilst others are spared (Kendler & Prescott, 1998). Whereas an increasing number of investigators have studied genetic sources of variation in the risk of transitioning from cocaine use to cocaine dependence, many observers appreciate that the influence of environmental conditions and processes might be strongest at the earliest stages of cocaine involvement—namely, in the stages that lead up to the first chance to try cocaine. In research on these earliest stages, Wagner & Anthony (2002) found that opportunity to try cocaine was associated with prior cannabis smoking, and a history of prior cannabis signaled excess risk in transitioning from opportunity to try cocaine into use of cocaine. 15 • What linkages of states and processes influence who becomes and who remains cocaine user and cocaine dependent? As for population level estimates of the chance to try cocaine, an estimated 20%-25% of the US population had had a chance to try this drug, based on the most recently available national survey data. Males have been more likely than females to have the chance to try cocaine, but males were not more likely than females to progress to actual use once the opportunity occurred (Van Etten & Anthony, 1999). With respect to genetic and individual-level traits such as personality that might be considered as suspected determinants of cocaine use and cocaine dependence, Cloninger has stressed a trait of ‘harm avoidance’ (Chakroun 2004). In contrast, other research has been designed in relation to the personality model of McCrae & Costa (1986). The McCrae & Costa model stresses ‘openness to experience’ and closely related facets such as ‘sensation seeking’ or ‘risk taking’ which might influence levels of drug involvement (Terracciano 2008). There are several conceptual models for the pathogenesis of drug dependence and its natural history, once drug use starts. One such model is the ‘Jellinek model,’ originally developed in relation to alcoholism, and according to which the clinical features occur in several distinct phases, one phase following another. Another model can be termed the ‘transitions model’ under which individuals make stage-transitions from ‘non-use’ to ‘first use,’ from ‘first use’ to ‘first problem’ and from ‘first problem’ to a formally diagnosed state of drug dependence. Alternative conceptualizations include the ‘transitions-progressions model’ that alternates between discrete stage-transitions of this 16 type and with intervening progressions along dimensions within each stage (Anthony & Helzer 2002). . Irrespective of the model chosen, drug dependence will not occur in the complete absence of a drug. We can think of each drug as an etiologic agent in relation to the occurrence of drug dependence. Analogous to an agent of infectious disease, with one case begetting others via person-to-person spread of an infective agent, there can be person-to-person spread in drug use and dependence, as a drug is passed from hand to hand (de Alarcon, 1969). This person-to-person spread might account for geographic clustering of cocaine experiences, as described above (Petronis & Anthony, 2003). Another aspect of ‘mechanisms’ of cocaine dependence has been illuminated by the work of Shaffer & Eber (2002). This research suggests that cocaine dependence might develop less rapidly among users whose early clinical features are characterized by psychosocial consequences, more rapidly among users with clinical features of neuroadaptation. Chen & Anthony focused their work on the rapid emergence of clinical features of cocaine dependence when crack-smoking was combined with intranasally used cocaine HCl powder, as compared to intranasal users of cocaine powder alone, once use of these compounds starts. As reported in 2004, they found that most clinical features of cocaine dependence occurred 2-3 times more often among cocaine HCl users who also were crack smokers as compared to individuals using cocaine hydrochloride powder only. In other study, O’Brien & Anthony (2005) reported that rapid-onset dependence was found for females, adolescents, and for non-Hispanic African Americans, as 17 compared to complementary reference groups among recent onset cocaine users. Use of crack cocaine and/or taking cocaine by injection also were associated with excess risk for becoming cocaine dependent soon after use. The study of ‘mechanisms’ also includes the possibility that cocaine use or dependence has tangible health consequences that appear in sequences of states or processes that follow first ues of the drug. For example, a greater risk of panic attack has been found among users of cocaine, as compared to non-users (Anthony, Tien & Petronis, 1989). Anthony & Petronis (1991) extended this line of research to include other psychiatric consequences of cocaine use. They found that cocaine was associated with occurrence of several psychiatric disturbances (i.e., a depression syndrome, a mania syndrome), and that these disturbances post-dated cocaine use in a sequence (i.e., had not occurred prior to onset of cocaine use). The public health significance of cocaine use with respect to psychiatric consequences, and hypotheses about cocaine-precipitated mental disorders have been reviewed by Chen (2009). With respect to consequences of cocaine use early in life, investigations of cocaineexposed infants sometimes show evidence of excess prenatal mortality, reduced gestational age and fetal growth, as well as neurobehavioral and neuropsychological sequelae (Bandstra et al. 2002; Morrow et al. 2003). Cocaine-exposed children have been found to have lower overall language skills than non-cocaine-exposed-children (Morrow et al., 2003). Also, there can be cocaine-associated deficits in mother-child interaction, with interactions most impaired when the mother continued to use cocaine while the children were three or more years old (Johnson et al., 2002). 18 According to pre-clinical studies, it is plausible that in utero drug exposure might be an early predisposing factor for drug dependence later in life. Nonetheless, in this field, no definitive evidence on this issue is available. • What can be done to prevent and intervene in cocaine use and cocaine dependence? At present, we have no definitive evidence about the impact of mass action interventions with respect to cocaine use and dependence. Despite expenditure of billions of dollars to intercept cocaine traffic, there is no clear evidence that these law enforcement actions have achieved lasting beneficial impact. This absence of evidence does not necessarily mean absence of effect, but rather it reflects lack of investment in proper research on the effect of law enforcement and other drug control approaches such as crop eradication and seizures of contraband before it enters the country (Manski et al., 2001). Fortunately, good progress is being made in clinical treatment research on cocaine dependence (Carroll 2004, Polling 2006). Research to produce comparable evidence for cocaine prevention programs is lagging (Behrens 2000, Caulkins 2004). Against this background on the history, pharmacology and public health research on cocaine use, it is now necessary to turn to the reader’s attention to the more specific aspects of cocaine epidemiology that motivate this doctoral dissertation research. The first hypotheses to be addressed will study the possibility that cocaine use might cause panic attacks mentioned under the rubric of ‘mechanisms’ in this section. 19 Section 2.3 Introductory Material on the Cocaine-Panic Association In the interest of full disclosure, I would like to draw the reader’s attention to the fact that some of the passages in this section already have been published in a journal article (Alvarado GF, Storr CL, Anthony JC. Suspected causal association between cocaine use and occurrence of panic. Subs Use Misuse, 2010;45:1019-32). The background for testing a cocaine-panic linkage began with small sample clinical case reports, substantiated by theory and evidence from neuroscience and psychopharmacology (e.g., Aronson & Craig 1986, Price & Gianini 1987, Geracioti & Post 1991). Many clinicians now believe that there is a cocaine-panic association, and there already is one report on how to treat cocaine-induced panic disorder (Louie et al. 1989). Anthony and colleagues (1989) nested a case-control study within a prospective study design and provided the first epidemiological estimates on this suspected causal association, finding cocaine users to have an excess risk of panic attack, consistent with the early clinical observations and case studies. The strength of association was quantified in a relative risk (RR) estimate of roughly 3.0 (i.e., three fold excess risk of panic attack among cocaine users as compared to never users). A more recent study on this topic has been conducted by O’Brien and colleagues, who used the ‘subject as own control’ case-crossover design, with participants drawn from a large general population sample. With this approach, the research group also found cocaine-associated excess risk of panic attack (RR=3.3; O’Brien et al. 2005). 20 Quite apart from the prior evidence and theories based upon neurobiology of panic and psychopharmacology (e.g., see Anthony et al. 1989; Geracioti & Post, 1991; Breslau & Klein, 1999; Almodovar-Fabregas et al. 2002; Millan, 2003), a series of methodological issues can be raised to set the stage for the present study. First, it is possible that large-sample epidemiological research methods are required to probe the suspected cocaine-panic linkage with any degree of thoroughness. A formalized experimental trial is not ethical; we cannot administer cocaine under laboratory conditions to cocaine-naive human subjects (i.e, ‘never’ users) to the point of their developing cocaine dependence, although acute effect studies have been completed for cocaine and for other drugs (e.g., see Griffiths, 2006; Foltin & Fishman, 1997). A trial restricted to experienced cocaine users might necessarily exclude participants who are most vulnerable to cocaine–induced panic during the first 1-100 occasions of use, by virtue of the sample restriction. A clinical study of patients seeking treatment for panic can be biased to the extent that the combination of panic and cocaine use might be overrepresented in the clinical population as compared to the community population. The need for a large community sample is a function of parameters such as the cumulative occurrence of cocaine use, estimated as roughly one in six for 15-54 year olds (Anthony et al. 1994) and as 19%-20% for young adults in recent national sample surveys (United States, 2003), as well as estimates of relative risk of panic attack among cocaine users versus never users (e.g., the roughly three fold cocaine-associated excess risk, just noted). If the relative risk parameter truly is 3.0, then the minimum sample size for a statistically powerful epidemiological study with a 1:1 case-control ratio is more 21 than 110 panic attack cases (and more than 110 non-cases, based upon alpha=0.05 and beta= 0.20). Notwithstanding the statistical robustness and presumed definitiveness of large sample epidemiological research on a cocaine-panic association, other limitations are confronted in this type of research. First and foremost is a limitation in relation to assessment. Panic attacks typically can be assessed via standardized field survey questions. As gauged against standardized diagnostic assessments completed by experienced psychiatrists, diagnostic field survey assessments of panic attacks based upon self or lay-administered interviews are suspect, even though the measurement properties (reliability, validity) of standardized items to assess panic attack (including ‘limited symptom’ panic attack) seem to be better than those observed for panic disorder per se (e.g., see Helzer et al. 1985; Anthony et al. 1985). Nonetheless, to be clear, a formal ‘panic attack’ as diagnosed by an expert clinician can be distinguished from the panic-like or panic attack-like experience elicited via standard field survey questions. A note about ‘limited symptom’ panic attack may be in order. This concept appears in the revised Third Edition of the American Psychiatric Association Diagnostic and Statistical Manual (DSM-III-R; APA, 1987). As defined there, a ‘limited symptom’ panic attack is one that involves “an unprovoked attack of sudden fear that includes fewer than four symptoms.” Research on the validity of diagnostic assessment of panic (e.g. Anthony et al.,1985) indicates that experienced clinicians can differentiate fully fledged panic attack from ‘limited symptom’ panic attack (e.g., when the clinician pays close attention to principles of psychopathology, pharmacology, and etiology at the symptom and patient level). It is less clear that this differential diagnosis is made successfully with 22 large sample field survey methods in the absence of a clinical reappraisal. In the proposed research, a panic attack assessment protocol for lay (non-clinician) interviews akin to the one used in the NIMH Epidemiologic Catchment Area (ECA) Project was used (Anthony et al. 1989). Specifically, the assessment requires an affirmative response to the following question adapted from the NIMH Diagnostic Interview Schedule: “Have you ever in your life had an attack of fear or panic when all of a sudden you felt very frightened, anxious, or uneasy?” Experts on panic disorder will appreciate that this question should elicit most (but not all) forms of genuine panic attack, but it is somewhat non-specific and liable to encompass a ‘limited symptom’ panic attack, and possibly other panic-like anxiety states. As such, I regard the object of study in the present research to be a panic attack-like experience, rather than panic attack per se. A note about large-sample epidemiological survey research response rates also may be in order. Response rates from recent large-sample cross-sectional surveys for mental disorders and drug use generally have been in a range from 70-80% (e.g., see United States, 2002, 2003; Grant et al. 2004; World Health Organization World Mental Health Consortium, 2004). In most recent long-span prospective research with large epidemiological samples studied from birth or primary school, interview participation at follow-up in young adulthood is reflected in ‘response rate’ values on the order of 60%70% (e.g., see Fergusson et al. 2001). Hence, there is some greater potential for attrition bias in long-span prospective studies (e.g., if participants with early mental health problems are more likely to drop out of the prospective study). Against this background and with only two prior epidemiological community sample studies on the topic of the cocaine-panic association, an additional contribution of 23 epidemiological evidence from independent study samples was judged to be valuable. To add this new evidence, in a first analysis, the dissertation has involved a re-framing of epidemiologic data gathered for another purpose as part of a prospective and longitudinal study of the mental health and behavior of youths growing up and going to school in an urban American community. Whereas that overall research program was prospective and longitudinal in relation to independent research aims, here the data were gathered on cocaine use and separately on panic attack-like experiences, assessed via standardized follow-up assessment protocols administered during young adult years of participants originally recruited as children. Due consideration will be given to a history of panic experiences that had been measured when the young adults were in adolescence. As such, the resulting study design is that of a retrospective case-control study nested within the prospective study. The ‘cases’ are young adults who have had experiences resembling a panic attack; the ‘controls’ are all others. The exposure of interest pertains to whether these cases and controls had tried cocaine at least once. Here, as in prior research (e.g., Anthony et al. 1989; O’Brien et al. 2005), the expectation is an estimate of about three fold risk of panic attack among those who had used cocaine versus those who had not. In a second analysis, the open research question about prior cocaine use and subsequent onset of panic motivated an effort to synthesize a longitudinal research approach by drawing upon retrospective age of onset date from a cross-sectional sample survey. In this second analysis, the main aim is to estimate a suspected causal association linking earlier onset of cocaine use with later risk of newly incident panic attack in adults. The data will be from an epidemiological sample of Americans aged 18-44 years at the time of cross-sectional and retrospective assessment. 24 This inquiry involves cross-sectional assessment of the age of onset of cocaine use and of panic attack, as well as an application of discrete-time survival analysis methods for estimation of the strength of suspected causal association linking earlier cocaine use with later risk of panic attack. For focus, the sample has been restricted to 1844 year old participants in a community population survey (NESARC, National Epidemiologic Survey on Alcohol and Related Conditions); onset of panic disorder (and panic attack) after age 45 is quite rare (see Eaton et al; 1989). Section 2.4. Introductory Material on Determinants of Becoming Dependent Soon After Onset of Cocaine Use. The topic of the cocaine dependence syndrome already has been addressed in section 2.2.2, and Table 2.1 provides a list of the ICD and DSM-IV criteria for this specific syndrome. In epidemiological field research the syndrome can be assessed via standardized field survey questions. In this part of the dissertation research, the main aim has been to estimate the risk of becoming cocaine dependent in the first months and years after onset of cocaine use, and to study a potential male-female difference in this risk, with the context of a more general conceptual model that includes other suspected determinants of becoming cocaine dependent. The study estimates are based on nationally representative samples of non-institutionalized citizens of the United States aged 12 years and older. As such, drawing upon a public use dataset from epidemiological data for the United States (from 25 the NSDUH, National Survey on Drug Use and Health), the goal is to challenge or to confirm a previously a published epidemiological estimate for the risk of developing a dependence syndrome within a span of 0-24 months after first use of cocaine (O’Brien, 2005) and to probe into a previously reported male-female variation in this risk, as described in later paragraphs of this introduction. Tapping NSDUH data gathered during the 2002–2003 calendar years, the probability of rapid transition from cocaine use to dependence is to be re-estimated, and possible malefemale difference in risk of becoming cocaine dependent soon after onset of cocaine use will be studied. There also is hypothesis testing whether males and females are equally likely to experience clinical features associated with cocaine dependence in the first months and years after cocaine use starts. This work builds from the prior research findings of an affiliated research team. In particular, based upon data gathered in 2000-2001, O’Brien & Anthony (2005) reported that an estimated 5%-6% of cocaine users in the US develop a cocaine dependence syndrome within a span of roughly 24 months after first cocaine use, with a median interval of roughly 12-13 months. In addition, that research team found excess risk associated with being female, even with other covariates held constant (e.g., cracksmoking, being of African heritage). Of course, as in the dissertation research based upon the NESARC data, the NSDUH data are from a cross-sectional survey, and do not have the character of a prospective study. 26 Nonetheless, via a focus on newly incident and recent-onset cocaine users, the goal is to simulate relative risk estimates as might be obtained through a prospective design. To the extent that advancing cocaine dependence might induce selective attrition from a prospective study sample, then the cross-sectionally derived estimates may actually be superior to prospectively-derived estimates, as explained by Chen and Anthony (2005). Section 2.5. Introductory Material on Who Starts then Stops Cocaine. Some members of the dissertation guidance committee recommended an investigation of who starts using cocaine and then stops. This suggestion has prompted an inquiry that was not thought out in advance of the dissertation research project. The approach involved analyses of cross-sectional data from 2003, in order to generate a set of testable hypotheses. Then, the approach was repeated in 2009, using analysis of crosssectional data from 2008 (five years later), with formal tests of the previously generated hypotheses. In this study, the main aim is to provide epidemiological evidence about predictors or correlates of who starts using cocaine and then stops within a four year span of time. The study data are from nationally representative cross-sectional survey samples of non-institutionalized residents of the United States (US). The approach starts with estimation of an initial set of estimates derived from initial exploratory and developmental analyses of data from calendar year 2003 (CY2003). Then, in crossvalidation mode, the model based on CY2003 data is re-evaluated using data from an independent sample drawn in CY2008. 27 A search for citations and a subsequent literature review with epidemiological estimates on starting and stopping cocaine use disclose that there are roughly three times as many published articles on the topic of starting, as compared to the topic of stopping, with very few identified predictors of cessation of cocaine use in contrast to a rich array of suspected predictors and explanatory variables with respect to onset of cocaine use. One explanation for this imbalance may relate to the relative ease of recruiting community dwelling individuals who never used cocaine and who are at risk for starting to use, combined with the greater difficulty of identifying in community samples cocaine users who started and then stopped. In addition, until recently, community surveys have not generally gathered the fine-grained month to month data on intervals between starting and stopping cocaine use as might be required to investigate the issue of cessation as thoroughly as is required. The addition of these month-specific assessments in national surveys completed between CY2003 and CY2008 has created new opportunities for the type of research completed here. In the background of this dissertation research project, there actually are several longitudinal studies of pertinence, mainly reflecting the experience of young people who first started to use cocaine before 1985. For example, Kandel and Raveis (1989) as well as White and Bates (1995) conducted longitudinal research on cessation of cocaine use, with each set of estimates based upon non-clinical samples of young people who had started cocaine use in the 1970s and early 1980s. The focus of the Kandel-Raveis research was on a cohort of adolescents followed until young adulthood from the early 1970s forward. In the Kandel-Raveis study, the most noteworthy relationships involved 28 parenthood (i.e., becoming a parent, which was associated with increased likelihood of stopping cocaine use), as well as peer relationships (i.e., having friends who use cocaine, which was associated with persistence of cocaine use, and inversely with stopping cocaine use). The work by White and Bates focused on a different cohort of adolescents recruited in late 1970s and early 1980s and followed until young adulthood. They found that stopping cocaine use had a positive association with being female, being older, and an inverse association with the drug use of friends. An unpublished internet document posted by the United States Office of National Drug Control Policy (2004) describes a study of cocaine use by young people who were assessed longitudinally between 1979 and 1998. Whereas the main findings were about predictors of heavy cocaine use, the results portray movement of adolescents into and out of cocaine use. For example, an estimated 7% of non-cocaine users observed in 1984 had become users in 1988. Moreover, an estimated 40% of cocaine users observed in 1984 were found to be using cocaine in 1988 as well; the other 60% had suspended or stopped cocaine use and did not use in 1988. Regrettably, this report is of a descriptive character, and does not provide estimates on the predictors of cocaine cessation. Issues pertinent to cocaine cessation were introduced in a cross-sectional study of cocaine dependence among newly incident cocaine users, completed by Chen & Anthony and published in Psychopharmacology in 2004. Specifically, this work suggested that users of crack-cocaine might have a more aggravated clinical course, a possibility that previously had been raised in the clinical literature on crack-cocaine (e.g., see Verebey & Gold et al., 1988). 29 One of the more important issues in evaluation of these studies, with the exception of the work by Chen and Anthony, is that they all are based on samples that were recruited during or shortly after the United States’ second cocaine epidemic, which started in the 1970s and continued into the 1990s. There is some evidence that individuals starting to use drugs (including cocaine) during the post-epidemic interval (after the early 1990s) have different characteristics than individuals who have started to use drugs more recently. For example, Golub and Johnson (1994) have described ‘cohort changes’ in patterns of drug use onset, and characterize those who started to use cocaine during the cocaine epidemic years as individuals born between 1955-69, with the more recent birth cohorts much less likely to use cocaine and more likely to smoke cannabis or cannabistobacco combinations known as ‘blunts’. It is these recent gaps in epidemiological evidence about cessation of cocaine use outside of clinical treatment settings that have prompted the present investigation. All of the cocaine users under study for this research were sampled, recruited, and assessed by self-report methods during the National Survey on Drug Use and Health in CY2003 and in CY2008. The NSDUH research approach raises some important issues that must be considered in anticipation of framing specific research questions. Specifically, there is the self-report character of the NSDUH assessments without bioassay confirmation. Some recent bioassay comparisons under confidential field survey conditions suggest that some confirmed cocaine users in these samples may be unwilling to disclose recent cocaine use even when they are willing to provide information about the lifetime history of past cocaine use (Harrison 2007). Second, there is the NSDUH’s cross-sectional sample 30 design, without longitudinal follow-up to the time of cocaine cessation (or persistent use). Third, there are constraints on the explanatory variables that can be considered in relation to major hypotheses, with interpretation of cause-effect estimates constrained by incompletely specified models and often by NSDUH failure to measure important confounding variables that might account for the observed associations or that might involve violations of assumptions about endogeneities. Finally, there is potential differential attrition or non-participation of the most seriously affected cocaine users (e.g., those with severe forms of cocaine dependence), such that the cocaine users represented in the NSDUH samples might represent the ones most likely to quit using cocaine during a short span of time. Other cocaine users in the same originating source population and birth cohorts under study might not be observed due to cocaine-related mortality (e.g., acute overdose) or disengagement of social conventional bonds that might otherwise, if strong, tend to promote participation in survey research sponsored by the US federal government. These issues, among others, will be covered in more detail within the discussion section of this paper, after presentation of methods and results. Nonetheless, a few notes about each limitation may be in order here. First, the existing research on self-report methods to assess cocaine use in large samples is based largely upon comparisons of potentially fallible toxicological assays with self-reports derived from face to face interviews in past years. In more recent years, including the years under study, the NSDUH has involved audio-enhanced computer-assisted personal interviews so that the respondents do not have to vocalize or describe their drug use to the field staff member. Instead, answers are 31 keyed into a laptop computer directly by the respondent. Second, longitudinal follow-up studies of cocaine users, especially those with severe cocaine dependence, may be more subject to attrition-related biases as compared with the cross-sectional survey estimates derived in research of this type. Third, the aims of the present research are predictive or correlational in nature rather than explanatory in a cause-effect sense. The resulting estimates pertain to the strength of association linking antecedent characteristics with cessation of use among cocaine users, but cannot be interpreted as evidence of causeeffect relationships. Finally, the inferences from studies of this type necessarily are constrained in relation to external generalizability. What is observed should hold for other community survey populations with methods like those used in the NSDUH if all assumptions of generalizability can be made. Nonetheless, this question can be checked only via systematic replication studies conducted in other populations, times, and places. In this context, the following research questions are posed: Are females more likely to stop? Is age of onset associated with cocaine cessation? Are crack users less likely to stop? These three research questions reflect plausible testable hypotheses based upon review of the background literature cited above. The initial tests will be made with NSDUH CY2003 data; the confirmatory tests will be made with NSDUH CY2008 data. 32 Section 2.6. Summary and Anticipated Public Health Significance of the Dissertation Research There is a growing body of evidence about the epidemiology of cocaine use and suspected hazards. Most of the published estimates have a descriptive character, with no hypothesis-testing. One exception to this general rule involves cocaine toxicity, and in particular the dissertation research is focused on a suspected causal linkage between cocaine use and panic attacks, which started when clinicians observed cases that seemed to be cocaine-precipitated panic attack and disorder. The subsequent epidemiological probing of this suspected causal hypothesis was generally supportive, but does not yet have a definitive character. For this reason, the first two studies completed in this dissertation research involve additional probing of that same cocaine-panic association, with a goal of increasing the definitiveness of the evidence base as might support a causal inference, if one is warranted. In the two dissertation studies on the cocaine-panic association, the first investigation has been nested within a longitudinal and prospective study of children growing up and going to school in a single metropolitan area, which has the advantage of holding constant local area characteristics that might otherwise function to confound and possibly bias the estimation of cocaine-associated excess risk of panic attacks. The second investigation has involved an application of advanced survival analysis models for the study of cocaine-associated excess hazard of panic attack, in an extension of the nested casecontrol research originally completed by Anthony and colleagues (1989). These new investigations were intended to build up the body of epidemiological evidence on the 33 cocaine-panic association, to subject it to new scrutiny, and to extend the modeling of suspected confounding variables in order to help confirm or disconfirm what previously had been published. In the dissertation study about occurrence of a rapid-onset syndrome of cocaine dependence soon after cocaine use, a completely different type of cocaine toxicity is under investigation. Here, the toxic response of interest is a syndrome of cocaine dependence as defined in the APA DSM-IV, and the research question involves whether and to what extent this syndrome emerges quite rapidly in the first 24 months after onset of cocaine use. This study entailed a replication of a prior epidemiological research conducted by Wagner & Anthony (2002), Chen & Anthony (2004), and O’Brien and Anthony (2005). As such, this investigation also has been an effort to confirm or to disconfirm the estimates from that prior research, in which it was found that an estimated 5-6% of cocaine users developed a rapid-onset form of the cocaine dependence syndrome, with some interesting population subgroup variation in risk estimates, as well as crack-associated excess risk of becoming cocaine dependent. In the fourth and final investigation, the dissertation covers new ground in an epidemiological study of starting and stopping cocaine use. That is, essentially all prior epidemiological studies have been focused upon cocaine-associated toxicity or upon the initiation of cocaine use. In this investigation, the goal was to look into cessation of cocaine use once it has started, and to evaluate sub-group variation in the odds of stopping cocaine use once it starts. Because this investigation was the first of its type, no 34 advance hypotheses or complex models were specified. Instead, the work has somewhat of an exploratory character in the analysis of 2003 data, with a more confirmatory approach in analysis of the 2008 data. In sum, this dissertation was designed to produce incremental steps toward the development of a more complete understanding of the epidemiology of cocaine use, with focused attention on two forms of toxic responses (panic attack; cocaine dependence), as well as a focused look at cessation of cocaine use in an epidemiologically defined population sample. The methods of investigation are described in the next chapter, and thereafter the results chapter is organized in the form of four submitted manuscripts for publication. The dissertation report then presents a general discussion of the main findings of the dissertation research, a consideration of limitations in these investigations, and an outline of directions for new research. The conclusions chapter provides a brief overview of what has been found and what might come next in these lines of research. 35 CHAPTER 3. RESEARCH MATERIALS AND METHODS As stated in Chapter 1, this dissertation research has been oriented in relation to a series of research projects. This chapter provides information on research materials and methods pertinent to each study (i.e, populations under study, sampling approaches, recruitment and informed consent processes, assessment approaches, quality control, analysis approaches). Section 3.1 Research Approach Details Specific to Results Chapter Section 4.1. Suspected Causal Association between Cocaine Use and Occurrence of Panic In the interest of full disclosure, I would like to draw the reader’s attention to the fact that some of the passages in this section already have been published in a journal article (Alvarado GF, Storr CL, Anthony JC. Suspected causal association between cocaine use and occurrence of panic. Subs Use Misuse, 2010;45:1019-32). The present inquiry builds from a program of epidemiology and prevention research initiated by Professors Sheppard Kellam, James C. Anthony, and their colleagues at the Prevention Research Center of Johns Hopkins University School of Hygiene and Public Health. Detailed reports on the research design and methods have been described by Kellam & Anthony (1998), Kellam et al. (1991), and later collaborators who joined the research team (e.g., Chilcoat et al. 1995; Ialongo et al. 1999; Storr et al. 2004; Wilcox & Anthony, 2004). In this program of research, the basic design 36 was that of a prospective and longitudinal study, with multiple waves of follow-up assessment after initial recruitment of an epidemiologically credible sample of children as they entered primary school in a single metropolitan area. As described in the just-cited detailed reports (e.g., Kellam and Anthony, 1998), the study population was designated to include all first-graders entering 19 public elementary schools of a single school system during two successive school years (“Cohort 1” entering in 1985 and “Cohort 2” entering in 1986). All of these first-graders were residents of urban neighborhoods within the catchment area of this school system, which is located in the mid-Atlantic United States. There was no sub-sampling: efforts were made to recruit all entering first graders, and more than 90% were recruited (n=2311). After annual assessments between 1985 and 1994, a young adult re-assessment was completed in 2000/2001, 15 years after initial recruitment, when respondents had reached a mean age of 21 years old. These young adult participants completed 90 minute face-to-face interviews. Some of the designated participants did not consent to participate as young adults (n=142); others were traced but were unable to be interviewed (n=133, e.g., in military); there were some verified deaths (n=32), and 317 could not be located. Sample attrition analyses disclosed greater follow-up participation for African-Americans in the sample (66% of the baseline sample; 71% of the follow-up sample), and for females (50% at baseline; 53% at follow-up). The protocol for the research was reviewed and approved by the cognizant institutional review board for protection of human subjects in research at Johns Hopkins Bloomberg School of Public Health. The analysis protocol was reviewed and approved at Michigan State University. 37 The key response variable in this study is the occurrence of a ‘panic attack-like experience’ including ‘limited symptom’ panic attacks, assessed by means of the standardized survey question mentioned in this paper’s introduction, as it was administered at the time of young adult assessment: “Have you ever in your life had an attack of fear or panic when all of a sudden you felt very frightened, anxious, or uneasy?” This question was asked early in the young adult interview as part of a ‘warm-up’ review of headaches, general medical conditions, depressed mood, anxiety states, and fears. It was separated from questions about drug use, which appeared about 30 minutes into the lengthy follow-up interview. Young adults who answered ‘yes’ to the standardized question on panic-like experiences are the ‘cases’ in this case-control study; all other young adult participants are non-case ‘controls’. There was also a question about panic attack-like experiences at the time of assessment in early adolescence:”Have you ever in your life had a spell or attack when all of a sudden you felt frightened, anxious or very uneasy in situations when most people would not be afraid or anxious?” The responses to this question are not used to designate panic attack cases. Rather, in a set of exploratory analyses, I have used this self-report history to exclude individuals with an early panic-like experience, before the years when cocaine self-administration begins. In this way, I achieve greater focus on panic-like experiences arising during the years of late adolescence and young adulthood, contemporaneous with onset of cocaine use. The suspected determinant of central interest is cocaine use on one or more occasions. At follow-up assessment, participants were asked the following question ‘Have you ever used cocaine in any form, including powder, crack, free base, coca leaves 38 or paste? ” This covariate is coded 1 for cocaine users; for non-users, the covariate value is 0. Neither the interviewers nor subjects knew that the cocaine-panic association would be a topic of inquiry for this project, which was conceptualized after all data had been gathered. The relationships linking cocaine use with occurrence of panic attack-like experiences were estimated within the context of a general conceptual model for panic attack, based on a summary of published literature and previous theory. Key covariates within this conceptual model encompassed basic demographic characteristics (age, sex, race-ethnicity), with measures derived from school administrative records abstracted early in the study. Other covariates included suspected determinants of panic attack such a history of tobacco smoking, drinking problems, history of a depression syndrome, and marijuana use (e.g., see Anthony et al. 1989; Breslau and Klein, 1999): all of these variables were assessed via standardized questions during young adult assessment. Lastly, I also have taken into account childhood misbehavior, as assessed via a standardized teacher rating in first grade (e.g., see Kellam and Anthony, 1998), which might be linked directly, or indirectly, with anxiety disturbances in general or panic attack specifically, as well as with cocaine use. The key response variable of interest—namely the occurrence of panic attack-like experiences—has been expressed as a function of whether the young adult ever had tried cocaine, with statistical adjustment for covariates listed above, and with a general plan for data analysis organized in relation to our research group's standard "explore, analyze, explore" cycles. The first exploratory steps involve data analyses to shed light on the underlying distributions of each response variable and covariate of interest. 39 In the initial analysis step, the task was to gauge the strength of the association thought to link cocaine use with occurrence of panic, estimated in the context of a regression model. In this research, as in a large proportion of contemporary epidemiological studies of relatively rare disease events (i.e., cumulative incidence proportions, P, below 10%), the parameter of interest is the relative risk estimate (RR). Here, the statistical models for estimation actually yield RR estimates by way of intermediate differencing of natural log odds, where the odds equals the cumulative incidence proportion, P, divided by its complement (1-P). For example, under the logistic regression model, the slope estimate conveys the dependence of a change in one unit of the y variable upon a change in one unit of the x variable on the natural log scale. Via exponentiation, this slope estimate serves as an odds ratio (OR) estimate, and in turn, the OR estimate serves in this context as an estimate of the desired population parameter of relative risk. In subsequent analysis steps, the statistical approach involved refining and checking the model estimates for robustness by fitting terms for additional covariate adjustment. Our final exploratory analysis steps involved a search for possible subgroup variation in the strength of the observed association, regression diagnostics, re-fitting of models to identify overly influential observations and outliers, as well as sensitivity analyses used to probe for model mis-specifications and invalid assumptions. In this work, I present and interpret the width of the 95% confidence intervals and the actual p-values as gauges for statistical uncertainty of the study evidence. However, readers with a frequentist orientation may wish to pay more attention when p values are below 0.05 and less attention when p values are 0.05 or greater. 40 Section 3.2. Research Approach Details Specific to Results Chapter Section 4.2. New Evidence on the Suspected Causal Association Linking Earlier Cocaine Use with Later Occurrence of Panic The study data are from the National Epidemiologic Survey on Alcohol and Related Conditions (NESARC) which has been described as a “nationally representative face-to-face survey of 43,093 respondents, aged 18 years and older, conducted by the NIAAA (National Institute on Alcohol Abuse and Alcoholism) in 2001-2002. The NESARC target population of the Wave 1 was the civilian, non-institutionalized population residing in the United States and the District of Columbia, including Alaska and Hawaii.” A detailed methodology overview has been published by Grant and colleagues (2004). In brief, the NESARC sampling approach involved multistage probability sampling methods, with PSU (primary sampling unit) selection in the first stage, withinPSU selection in the second stage, and one sampled person being selected from each household in the third stage. Within the resulting sample of 43,093 respondents, there were 22,048 18-44 year olds. The household response rate was 89% and the person response rate was 93%. After recruitment, all participants were assessed via standardized field survey interview methods, described below, according to a study protocol reviewed and 41 approved by a cognizant Institutional Review Board (IRB); the analysis protocol also was approved by the Michigan State University IRB. The key response variable in this ‘time to event’ analysis is the time to the occurrence of a ‘panic attack-like experience’ as measured in years since birth. The NESARC research team assessed several forms of panic attack, including ‘limited symptom’ panic attacks, as well as DSM-IV panic attack and panic disorder. For example, limited symptom panic attack was measured using a standardized survey question: ‘Have you ever had a panic attack, when all of a sudden you felt frightened, overwhelmed or nervous, almost as if you were in great danger, but really weren’t?’, followed by a question on age at first panic event of this type (hereinafter, “panic” or “panic event”). The suspected determinant of central interest is time-specific occurrence of cocaine use on one or more occasions, including crack-cocaine smoking, with a focus upon cocaine use that occurred prior to onset of panic attack (or until the date of NESARC assessment, if the person never had a panic attack). For this construct, the standardized interview questions were: ‘Have you ever used cocaine or crack?’ and ’How old were you when you first used cocaine or crack?’ The guiding conceptual model was one based on a summary of published literature and previous theory. Key covariates within this conceptual model encompassed basic demographic characteristics (sex, cohort, race-ethnicity), with age taken into account in the survival analysis model. Other covariates included suspected determinants of panic attack such a history of tobacco and/or cannabis smoking and a history of a 42 depression syndrome. All such variables were assessed via standardized questions within the NESARC interview. In our analyses, the key response variable of interest—namely the time to occurrence of the panic event—has been expressed as a function of whether the adult had tried cocaine prior to the panic event, with statistical adjustment for covariates listed above, and with a general plan for data analysis organized in relation to our research group's standard "explore, analyze, estimate, explore" cycles. In these cycles, the first exploratory steps involve data analyses to shed light on the underlying distributions of each response variable and covariate of interest. In the initial analysis step, the task was to estimate the strength of the association thought to link cocaine use with occurrence of panic, estimated in the context of a discrete time survival analysis in regression modeling. Singer and Willett (1993) attribute to Cox the initial idea that the hazard can be viewed as estimated probabilities, with re-expression of the parameters to have a logistic relationship with covariates in the model and with the discrete time intervals during which the events can occur. They express the ‘population discrete-time hazard model’ in the following excerpt from their paper (page 166): 43 Under this model, “the baseline level of hazard in each time period, and the slope parameters [β1, β2,....., βn] describe the effects of the predictors on the baseline hazard function, albeit on a logistic scale.” Expressed in terms of logistic transformations of the equation, the result yields conditional log-odds estimates for the relative occurrence of the event under study, during each discrete time interval (Singer & Willett 1993) – i.e,. via the following equation: Once the panic attack occurs, the participant is ‘censored’ and no longer contributes with information. This censoring ensures that observed cocaine use has not occurred after panic attack in the estimation of relative risk. In addition, to be 44 conservative, our estimates of relative risk are not based upon information from panic cases whose age at first panic was the same as the age at first cocaine use, for whom the cocaine and panic temporal sequencing is uncertain. I return to the ‘conservative’ nature of the observed estimates in this paper’s discussion section. In subsequent analysis steps, the statistical approach involved refining and fitting terms for additional covariate adjustment. The post-estimation ‘explore’ step structured with a search for possible subgroup variation in the strength of the observed association. Regression diagnostics, re-fitting of models to identify overly influential observations and outliers, and sensitivity analyses were used to probe for model mis-specifications and invalid assumptions. In this work, I present and interpret the width of the 95% confidence intervals (CI) and the actual p-values as gauges for statistical uncertainty of the study evidence; using an appropriate Taylor series approach. However, readers with a frequentist orientation may wish to pay more attention when p values are below 0.05 and less attention when p values are 0.05 or greater. Section 3.3 Research Approach Details Specific to Results Chapter Section 4.3. New Evidence on Risk of Becoming Cocaine Dependent: Epidemiological Estimates for the United States, 2002-2003 As noted, in this part of the dissertation research, the research design is cross-sectional. The study population was designated to include US citizens aged 12 years and over. The present analysis has a focus on the newly incident recent onset cocaine users in order to 45 constrain sample attrition that might occur as cocaine use progresses toward dependence syndromes and related clinical features (e.g., cocaine-associated disengagement from conventional social bonds and roles). The sample approach involved multistage area sampling procedures, as described elsewhere, with large Primary Sampling Units (PSUs), smaller strata within these PSU (e.g., block groups), and then sampling of dwelling units within strata before rostering and sampling of designated respondents within each dwelling unit. The resulting sample, as presented in the NSDUH public use datafiles, consists of 54,079 respondents in CY 2002 and 55,230 respondents in CY 2003. Some of the designated participants did not consent to participate, other had missing or invalid responses to key study variables. For this reason, the effective sample size for the present investigation is 109,309. The response rate was 85-91% for household screening and 77-79% for completed interviews. Recruitment and consent procedures were based upon a study protocol that had been reviewed and approved by the cognizant institutional review board for protection of human subjects in research. The key response in this study is occurrence of the DSM IV cocaine dependence syndrome within 24 months after onset of cocaine use. It was measured via a series of standardized survey questions that measure seven clinical features of cocaine dependence specified in the Diagnostic and Statistical Manual, Fourth Edition (APA, 1994); the questions are framed in relation to the prior 12 months (i.e., 12 months prior to the date 46 of assessment). The NSDUH research team created a computerized diagnostic algorithm to identify respondents whose profile of clinical features included three or more of the seven cocaine-dependence criteria. The public use dataset includes the results of applying this algorithm, with coding as follows: cocaine users qualifying for DSM-IV dependence=1; users who did not meet criteria for dependence and all others (including those who never used cocaine) were coded ‘0’. This approach is consistent with the approach used by O’Brien & Anthony (2005) and their resulting estimates. Hence, for replication purposes, this aspect of method is held constant here. Nonetheless, there is a point for discussion, involving sensitivity and specificity of the survey assessment, which will be addressed in the discussion section of this dissertation. In addition to a goal of estimating male-female variations in transition from cocaine use to cocaine dependence soon after onset of cocaine use, the suspected causal determinants or covariates of central interest are age, race-ethnicity, route of administration, family income, education, size of the metropolitan statistical area (MSA), and number of drugs used before cocaine. These covariates were measured via a fixed sequence of standardized survey questions designed to obtain information about drug experiences and other personal characteristics under study. The guiding conceptual model was one in which the key response variable of interest— namely, the occurrence of DSM-IV cocaine dependence—was expressed as a function of sex, as well as age, race-ethnicity, marital status, education, family income, size of the MSA, and number of drugs used before age 11 years (i.e., before onsets of first cocaine 47 use). As noted, race-ethnicity has been found to be associated with risk of becoming cocaine dependent, as has crack-smoking. Marital status, education, and family income are correlates of cocaine use that have been prominent in the prior literature (e.g., see Anthony 1994; O’Brien and Anthony, 2005; Miech and Chilcoat, 2005). Metropolitan area serves as a crude measure of street level availability of cocaine in the vicinity of the sampled dwelling units. Number of drugs used in childhood might function as a prognostic indicator of the reinforcing functions served by psychoactive drugs such as cocaine, in addition to its service as an indicator of general drug seeking behavior. The plan for data analysis was organized in relation to our research group’s standard ‘explore, analyze, explore’ cycles. The first exploratory steps in this cycle involve Tukeystyle box-and-whisker plots and other exploratory data analyses to shed light on the underlying distributions of each response variable and covariates of interest. Thereafter, in an initial analysis step, the tasks were crosstabular estimation of the frequency of occurrence of cocaine dependence as observed within 24 months after onset of cocaine use, and estimation of the strength of the association thought to link being male or female with the occurrence of cocaine dependence, estimated in the context of a multiple logistic regression model. Subsequently, covariate adjustments within the regression framework probe for the independence of the resulting associations (i.e., via covariate-adjusted estimates of the male-female relationship). Via these analysis steps, the statistical approach involves refining and checking the model estimates for robustness (e.g., via successive covariate control). For both weighted and unweighted analyses, STATA survey commands were used to conduct logistic regression 48 for complex survey data, with Taylor series linearization to address variance estimation complexities due to nested structures within the multistage sampling plan. The final exploratory steps involved a search for possible subgroup variation in the strength of the observed association, regression diagnostics, re-fitting of models to identify overly influential observations and outliers. A bias analysis approach was used to probe for selected model mis-specifications and potentially invalid assumptions. In this work, the width of the 95% confidence intervals and the nominal p-values are used as gauges for statistical precision and uncertainty in the study evidence. Readers with a frequentist orientation may wish to pay more attention when p-values are below 0.05 and less attention when p-values are 0.05 or greater. Section 3.4. Research Approach Details Specific to Results Chapter Section 4.4. Who Starts then Stops Cocaine? In this study, the research design is cross-sectional. As explained before, the approach generally has involved analyses of cross-sectional data from CY2003, in order to generate a set of testable hypotheses. Then, the approach has been repeated in crossvalidation mode, using analysis of cross-sectional data from CY2008 (five years later), with formal tests of the previously generated hypotheses. 49 For the CY2003 sample, the source population consisted of non-institutionalized community residents of the United States age 12 years and older. The study sample included 2,376 self-reported cocaine users in the CY2003 sample, all with reported onset of use within four years of the date of the survey assessment – i.e., within the interval of four years prior to assessment, they were the most recent onset cocaine users with no prior history of cocaine use predated that four year interval. These cocaine users were recruited during nationally representative sampling for the 2003 National Survey on Drug Use and Health (n=55,230 designated respondents after informed consent and creation of public use datafiles). The sampling approach involved multistage sampling procedures. Some of the designated participants did not consent to participate; others had missing or invalid responses to key study variables. For this reason, the effective sample size for the present CY2003 investigation is 2,376 cocaine users. For the 2008 sample, the source population again consisted of noninstitutionalized community residents of the United States age 12 years and older. The study sample included 2,227 cocaine users within the CY2008 sample, all with reported onset of use within four years of the date of the survey assessment. As in CY2003, these cocaine users were recruited during nationally representative sampling for the CY2008 National Survey on Drug Use and Health (n=55,739 designated respondents after informed consent and creation of public use datafiles). As in CY2003, the sampling approach involved multistage sampling procedures, and some of the designated participants did not consent to participate; others had missing or invalid responses to key 50 study variables. For this reason, the effective sample size for the present CY2008 confirmatory cross-validation analysis is 2,227. For the 2003 data, the response rate was 91% at the level of the sampled dwelling unit, and 77% for completed interviews. For the 2008 data, the corresponding values were 89% and 75%. The degree to which cocaine users are over-represented among the non-participants is unknown. The study protocols for the CY2003 and CY2008 surveys were reviewed and approved by the cognizant institutional review board for protection of human subjects in research. The protocol for this analysis of the NSDUH public use data was reviewed by the Michigan State University UCRIHS. The key response variable in this research is based upon the time of last use of cocaine, as measured by self-report. This variable has three mutually exclusive categories: (1) used cocaine within the past 30 days (~16% of users for 2003 data, ~13% for 2008); (2) used within past 12 months but not in the past 30 days (~34% for 2003 and 2008 data), (3) used before that (all others). This response variable was measured via a series of standardized survey questions. The primary analyses were framed in relation to the research questions and testable hypotheses identified in this study’s introduction. As indicated in prior research (e.g., Chen & Anthony, 2004; White and Bates, 2006), there was reason to ask whether cocaine cessation might be associated with older age of cocaine onset, whether female 51 cocaine users were more likely to stop using cocaine, and whether crack-cocaine users might be less likely to stop. The models to evaluate the three testable hypotheses were extended to include other covariates of less central interest – namely, race-ethnicity, family income, education, size of the area of the dwelling unit, and number of prior drugs used. All of these variables were measured within NSDUH modules of standardized questions presented to the respondent in a fixed sequence. The guiding conceptual model was one in which the key response variable of interest—namely, stopping cocaine use—was expressed as a function of the primary and secondary covariates, initially in relation to regression models with a single covariate, and then with covariate-adjusted models. The plan for data analysis was organized in relation to conventional standard ‘explore, analyze, explore’ cycles, in which the first exploratory steps involve Tukeystyle box-and-whisker plots and other exploratory data analyses to shed light on the underlying distributions of each response variable and covariates of interest. In the initial analysis step, the task was to estimate the strength of the association thought to link the individual covariates and stopping cocaine use, estimated in the context of a regression model. In subsequent analysis steps, the statistical approach involved refining and checking the model estimates for robustness by fitting terms for covariate adjustment. For both weighted and unweighted analyses, Stata survey commands were used to conduct 52 multinomial logistic regression for complex survey data, with Taylor series linearization to address variance estimation complexities due to nested structures within the multistage sampling plan. The tables of this paper will display the estimated risk for cessation comparing stoppers (i.e., those who have not used cocaine in the 12 months prior to the assessment) vs. current users (i.e., those who used cocaine within the 30 days prior to the assessment) as the reference category in the outcome. Final exploratory steps involved a search for possible subgroup variation in the strength of the observed association, regression diagnostics, re-fitting of models to identify overly influential observations and outliers, as well as sensitivity analysis used to probe for model mis-specifications and invalid assumptions. In this work, the width of the 95% confidence intervals and the nominal p-values are used to gauge statistical uncertainty of the study evidence. However, readers with a frequentist orientation may wish to pay more attention when p-values are below 0.05 and less attention when p-values are 0.05 or greater. 53 CHAPTER 4. RESULTS This chapter has four sections, one for each of the four studies completed as dissertation research projects. Section 4.1. Suspected Causal Association between Cocaine Use and Occurrence of Panic In the interest of full disclosure, I would like to draw the reader’s attention to the fact that some of the passages in this section already have been published in a journal article (Alvarado GF, Storr CL, Anthony JC. Suspected causal association between cocaine use and occurrence of panic. Subs Use Misuse, 2010;45:1019-32). Table 4.1.1 offers a description of the epidemiologic study sample, for this section of dissertation, showing males and females in quite balanced proportions. With respect to race-ethnicity, school records characterized two-thirds as African-American (Table 4.1.1). In the young adult follow-up sample, 641 reported at least one panic attack-like experience. A total of 46 (7.2%) of these cases had tried cocaine at least once. By comparison, among 1043 non-cases, only 40 (3.8%) had tried cocaine (Table 4.1.2). The ratio of 641 cases to 1043 non-cases is advantageous from the standpoint of statistical precision and power, but the ratio also suggests that this particular standardized survey question might have been over-inclusive, possibly not specific to panic attack per se, as discussed in our introduction. The estimated odds ratio (OR) for the association linking cocaine use with panic attack-like experiences was 1.9 (p=0.003), as shown in Table 4.1.2. The other results presented in Table 4.1.2 depict variation in the summary estimate with increasing 54 elaboration of the regression model to include multiple covariates. After addition of terms for covariates, the association remains positive (OR >1) and statistically robust (p< 0.05). Several covariates are of special interest. For example, an empirical report on a possible tobacco-panic association motivated a special look at the possibility that the cocaine-panic association might depend upon a prior history of tobacco smoking (Breslau & Klein, 1999). In our sample, virtually all cocaine users also had smoked tobacco. However, in an analysis focused exclusively upon young adults who had smoked tobacco, I observed a cocaine-panic association at the margin of statistical significance as defined conventionally (OR=1.7; 95% Confidence Interval CI: 1.0, 3.0; p=0.057). Once I excluded participants whose first panic-like experience was before mid-adolescence, the result was a slightly larger OR estimate that was statistically robust (OR=1.9; p=0.030). There were too few depression cases for estimation of the cocaine-panic association in that subgroup. Nonetheless, in an analysis focused exclusively upon young adults who did not have a history of a depression syndrome, I observed an appreciable cocaine-panic association (OR=2.4; 95% CI: 1.2, 4.8; p= 0.016); the exclusion of cases whose first panic-like experience was before mid-adolescence led to the same estimate (OR=2.4; p=0.018). In an analysis focused exclusively upon young adults who did not have alcohol problems, the cocaine-panic association remained (OR=2.5; 95% CI: 1.2, 5.3; p=0.013), even with exclusion of the early panic cases (OR=3.2; p=0.004). In our sample virtually all cocaine users had smoked marijuana, but in an analysis focused exclusively upon young adults who had smoked marijuana, I observed a cocaine-panic association at the margin of statistical significance as defined conventionally (OR=1.6; 95% CI: 0.9, 2.7; p=0.113). This estimate increased slightly and became statistically more robust with 55 exclusion of young people who had had early panic-like experience (OR=1.8; p=0.060). (Table 4.1.3, final row). Limits on sample size and the number of cocaine users placed constraints on the statistical precision and robustness of estimates for all subgroups of interest. However, in an exploratory search for subgroup variation, it was possible to check whether the regression models were improved with the addition of product-terms. For example, there was no appreciable improvement of fit in the model with inclusion of a product term for cocaine*tobacco (p=0.78 for model fit improvement); hence, a single slope appears to serve well as a summary estimate for the cocaine-panic association for young adults with and without a history of tobacco smoking. When I added a product term for cocaine*alcohol there was an improvement of the model fit (p=0.17); hence, with alpha set up at 0,20, more than one slope is required to model this association. Last, when I added a product term for cocaine*depression there was not an improvement of the model fit (p=0.36). Under the model with cocaine*alcohol product term, those with ‘cocaine use but no alcohol problems’ (n=41) were more likely to have had a panic attack-like experience as compared to the 1394 young people with ‘no history of cocaine use, nor alcohol problems’ (OR=1.8; p=0.056). For the 204 young adults with ‘alcohol problems but no cocaine use’ the corresponding odds ratio estimate is 2.7 (p<0.001). For the 45 young adults with a history of ‘cocaine use and also alcohol problems’, the corresponding odds ratio is 2.6 (p=0.002). 56 Section 4.2. New Evidence on the Suspected Causal Association Linking Earlier Cocaine Use with Later Occurrence of Panic Table 4.2.1 offers a description of the unweighted epidemiologic study sample for this section of the dissertation research. For example, the mean age of the study sample was 31.7 years, 56% were female, and 50% were Non-Hispanic White. In the study sample, 2,688 people reported at least one panic attack experience. Among these panic cases, 425 (15.8%) people had tried cocaine at least once. By comparison, among the 18,667 non-cases, only 1,291 (6.9%) had tried cocaine. The main estimates of the study are presented in Table 4.2.2, based on the statistical approach described above, with and without application of the survey weights. As shown in Table 4.2.2, the corresponding overall crude relative risk estimate was 2.2 (95% CI= 1.9, 2.5; p<0.001). The other estimate presented in Table 4.2.2 is based on an elaboration of the regression model to include terms for sex and race-ethnicity. With this covariate adjustment, the association remains positive (RR >1) and statistically robust (p< 0.05). Exploratory analyses probed into sub-group variation in the magnitude of the cocaine-panic association, with focus upon covariates that might be causal determinants of panic attack (Table 4.2.3). For example- as most of the cocaine cases had also smoked tobacco- in an analysis focused exclusively upon adults who had smoked tobacco, I observed a cocaine-panic association of RR=2.0; 95% (CI: 1.7, 2.4; p<0.001). Furthermore, in an analysis focused exclusively upon adults who did not have a history of a depression syndrome, the relative risk estimate was somewhat larger (RR=2.5; 95% CI: 2.0, 3.1; p< 0.001). In an analysis focused exclusively upon adults who had smoked 57 cannabis, I also observed a cocaine-panic association (RR=1.5; 95% CI: 1.3, 1.8; p<0.001). Post-estimation regression diagnostics and sensitivity analyses disclosed that the observed main study estimates were robust. For example, there were no overly influential observations. Section 4.3. New Evidence on Risk of Becoming Cocaine Dependent: Epidemiological Estimates for the United States, 2002-2003 Table 4.3.1 offers a description of the epidemiologic study sample (with and without weights) for this section of the dissertation research. It shows selected sociodemographic characteristics of all persons, all recently active past onset users, and the subset of newly incident recent onset cocaine users as defined in Chapter 3. For example, in the weighted estimates, I found that 52% of all persons in the source population were female; 70% were Non-Hispanic White. With respect to socioeconomic status, 20% of the sample may be characterized as lower income. A total of 1,638 respondents (roughly 1.5% of the total sample of 109,309 individuals) were found to have started using cocaine for the first time within 24 months of interview assessment. Among these, 55% were female, 75% were Non-Hispanic White, and 30% were considered as lower income. Table 4.3.2 offers a description of the epidemiologic sample for selected drug use characteristics. For example, based on weighted data, with respect to the general sample of all persons and resulting projections to the source population, an estimated four in one thousand persons (0.4%) qualified as cocaine dependent; 3.5% had engaged in crack58 smoking. Based upon the sample of newly incident recent onset cocaine users, an estimated 6% had become active cases of the cocaine dependence syndrome; 12% had used crack. The main estimates of the study are presented in Table 4.3.3, based on the statistical approach described in the methods section of the paper. For example, in the third column, we see that an estimated 4.3% of male newly incident recent onset users rapidly develop cocaine dependence. In contrast, the estimate for female newly incident recent-onset users is 9.5%. As shown in Table 4.3.3, the overall crude estimate (statistically unadjusted) for the association between female sex and rapid onset of cocaine dependence among newly incident users is 2.1, as compared to the experience of newly incident cocaine using males, and the association is statistically robust (p=0.01). Age of the user was not associated with rapid onset cocaine dependence (p>0.05). With respect to race-ethnicity, 7% of Non-Hispanic Whites developed cocaine dependence soon after onset of cocaine use, whereas 6% of Non-Hispanic-African Americans did – indicative of an unremarkable association that was not statistically robust by conventional standards (p>0.05). In these bivariate analyses, education was not associated with rapid onset of cocaine dependence among newly incident cocaine users (p>0.05), nor was family income. 59 Concerning crack smoking, newly incident cocaine users with concurrent crack-smoking (also newly incident) were found to have threefold excess risk of becoming cocaine dependent soon after onset of cocaine use, as compared with cocaine users with no concurrent crack-smoking (16% vs. 5%; RR=3.2, p<0.001). Statistical associations between the variables ‘ever used needle’ and ‘number of drugs used by age 11’ and the cocaine dependence outcome were unremarkable (p>0.05). The results presented in Table 4.3.4 depict robustness in the estimate of the male-female association with increasing elaboration of the regression model to include multiple covariates. With the addition of terms for several covariates, the association linking female sex with rapid transition from cocaine use to cocaine dependence remains robust, with p-values below the conventional 0.05 level of statistical significance. Even in the most elaborated regression model (Row 1 of Table 4.3.4), it can be seen that newly incident female cocaine users are more likely than male counterparts to experience rapid onset of cocaine dependence (estimated relative risk=2.2; p=0.005). In the model with multiple covariates, some variables thought to be related to rapid onset cocaine dependence were not found to have robust associations. For example, in this model, race-ethnicity was not associated with rapid onset of cocaine dependence soon after onset of cocaine use, (RR=2.2, p>0.05). 60 The final set of exploratory analyses (e.g., regression diagnostics, sensitivity analyses) disclosed that the observed main study estimates were robust. For example there were no overly influential observations. Section 4.4. Who Starts then Stops Cocaine? Based upon CY2003 data, Table 4.4.1 offers a description of the epidemiologic study sample for this section of the dissertation research (with and without weights). As noted, 2,376 respondents (4% of the total sample of 55,230 individuals) were found to have started using cocaine for the first time within four years prior to the date of survey assessment. The table depicts selected socio-demographic characteristics of all persons, as well as the subset of recent-onset cocaine users. For example, in the weighted data, 52% of the recent-onset cocaine users in CY2003 were found to be female; 70% were Non-Hispanic White. With respect to socioeconomic status, 20% of the sample may be characterized as lower income. For the subset of cocaine users, 43% were female, 71% were Non-Hispanic White, and 31% were considered as lower income. In the same subset of cocaine users, 44% started to use cocaine at 20 years old or older. The main estimates from the CY2003 analyses are presented in Table 4.4.2, based on the statistical approach described in the methods section of the paper. For instance, 50% in the sample of users have not used cocaine within the past 12 months previous to the assessment (i.e., have stopped using cocaine). Males are as likely as females (~50%) to stop cocaine use (RR=1.0; p>0.05). With respect to the issue of age at onset of cocaine use, I see that older recent-onset users (>=20 years old) are more likely to have 61 stopped using cocaine by the date of assessment (55%), the corresponding estimate for the younger (<=15 years old) recent-onset cocaine users is 44%. As shown in Table 4.4.2, the overall crude estimate (statistically unadjusted) for the association linking age of cocaine use onset with stopping cocaine use is 1.9 (p=0.01). With respect to raceethnicity, those with African American heritage are more likely to stop than Whites (RR=3.1; p=0.01), once cocaine use has started. In this bivariate analysis, there was not statistically robust association between education and stopping. With respect to history of number of drugs used, those who have not used any drug prior to cocaine are apparently more likely to stop than those who have used three drugs or above (59% vs. 48%) but the p-value was larger than the conventional threshold of significance. (RR=3.5, p=0.08). If cocaine users who engaged in crack-smoking are less likely to stop as compared to noncrack users (47% vs. 50%), the observed nominal p-value is larger than the conventional threshold of statistical significance (RR=0.7; p=0.08). The results presented in Table 4.4.3 depict robustness in the previous estimate with increasing elaboration of the regression model to include multiple covariates. With the addition of terms for several covariates, I see that there is not an association between being female and stopping cocaine use. Nonetheless, under the multinomial logistic regression model for multiple covariates, the estimated association between age of cocaine onset and stopping remained statistically robust, with a nominal p-value below the conventional 0.05 level of statistical significance. In specific, people who start using cocaine at 20 years and older are found to be at increased risk of stopping cocaine use (adjusted RR=2.2; p=0.003), as compared to other age of onset categories. With respect 62 to race-ethnicity, African Americans were found to be more likely to stop cocaine use than Whites (adjusted RR=2.6; p=0.02). Those who did not engage in extramedical drug use before cocaine use were more likely to stop than those who had used three drugs or above (adjusted RR=5.5, p=0.03). Under this model, if crack cocaine users are less likely to stop than non-crack users, the associated p-value was observed to be above the conventional threshold (adjusted RR=0.7, p=0.12). Lastly, in this model, I tested an interaction between “age of onset” and recency of cocaine onset (<=24 months vs. >24 months); a single slope was a better fit (p>0.05), and the interaction was not necessary in the final model. For CY 2008 data, Table 4.4.4 offers a description of the study sample (with and without weights). A total of 2,227 respondents (4% of the total sample of 55,734 individuals) were found to have started using cocaine for the first time within four years of interview assessment. The table depicts selected socio-demographic characteristics of all persons, and the subset of cocaine users. For example (using weighted estimates) it was found that of the recent-onset cocaine users in CY 2008, 52% were female; 68% were Non-Hispanic White. With respect to socioeconomic status, 17% of the sample may be characterized as lower income. For the subset of cocaine users, 47% were female, 76% were Non-Hispanic White, and 30% were considered as lower income. In the same subset of cocaine users, 49% started to use cocaine at 20 years old or older. These study estimates from CY2008 are not appreciably distant from those observed in CY 2003. The main estimates from the 2008 CY analyses are presented in Table 4.4.5. For example, 53% in the sample of users had not used cocaine within the 12 months previous 63 to the assessment (i.e., have stopped to use cocaine). Males were as likely as females (51% vs. 57%) to have stopped cocaine use (RR=0.9, p>0.05). With respect to age at onset of cocaine use, the older recent-onset users (>= 20 years old) are found to be more likely to have stopped using cocaine by the date of assessment (58%); the corresponding estimate for the younger (<= 15 years old) recent-onset cocaine users is 49%. As shown in Table 4.4.5, the overall crude estimate (statistically unadjusted) for the association linking age of onset (>=20 years old) with stopping cocaine use is 1.8 (p=0.07), as gauged in relation to a comparison group of cocaine users who had started to use cocaine at age 15 years of younger. With respect to race-ethnicity, perhaps the African American cocaine users among the recent onset users are apparently less likely to stop than White cocaine users, but the p-value did not reach conventional statistical significance (RR=0.4, p=0.1). No association was found between education and stopping. With respect to number of drugs used early in life, I did not find an association with stopping. Moreover, crack use was not associated with stopping cocaine use (RR=0.6, p>0.05). The results presented in Table 4.4.6 depict robustness in the previous estimate with increasing elaboration of the regression model to include multiple covariates. With the addition of terms for several covariates, it can be seen that there is not an association between sex and stopping cocaine use. Nevertheless, with respect to the age of cocaine onset and stopping, I see that an association with later onset use remains statistically robust, with p-value below the conventional 0.05 level of statistical significance. With respect to race-ethnicity, crack-cocaine use, and prior drug use, no statistically robust associations were found in this context (p>0.05). Lastly, I tested an interaction between 64 “age of onset” and recency of cocaine onset (<=24 months vs. >24 months); in this model (not shown), a single slope was found and the interaction was not necessary (p>0.05). The final set of exploratory analyses (e.g., regression diagnostics, sensitivity analyses) disclosed that the observed main study estimates were robust. For example, there were no overly influential observations. 65 CHAPTER 5. DISCUSSION This chapter of the dissertation will cover a synopsis of the main research findings from each project, the major limitations of the research, an overview of the meaning of the findings in light of these limitations, with some notes on possible implications for theory or additional subject matter research in this area. Section 5.1. Discussion Specific to Results Chapter Section 4.1. Suspected Causal Association between Cocaine Use and Occurrence of Panic In the interest of full disclosure, I would like to draw the reader’s attention to the fact that some of the passages in this section already have been published in a journal article (Alvarado GF, Storr CL, Anthony JC. Suspected causal association between cocaine use and occurrence of panic. Subs Use Misuse, 2010;45:1019-32). The main finding of this retrospective case-control study of young adults is a modest but generally statistically robust association linking cocaine use and the occurrence of a panic attack-like experience, somewhat more modest than the published three fold relative risk estimates for this suspected causal association, which were based upon more broadly representative and older community samples. Whereas slight measurement differences might account for some of the variation in the relative risk estimate, and the more modest association observed in this study, I also note prior epidemiological evidence showing that incidence (risk) of panic attack and panic disorder reaches peak values between age 25 years and age 34 years (e.g., see Eaton et al., 1989; Eaton et al., 1998), and that the period of risk for first-time use of cocaine extends across 66 an age span of 20-29 years, with peak risk values for African-American community residents occurring somewhat later than the mean (e.g., see Wagner et al., personal communication; unpublished analyses of national survey data). None of our follow-up sample's participants (71% African-American) had reached age 24 years by the time their cocaine and 'panic' experiences were assessed. As such, the more modest association observed in this study could signify that cocaine-precipitated panic occurs more often during the age span from 25 years through 44 years, with a more modest association in young adulthood. However, before more detailed discussion of the results and speculations of this type, several of the more important study limitations merit attention. First, I have been careful to note a lack of specificity in this study’s standardized assessment of panic attack-like experiences, which is a topic that did not surface in prior research on this specific topic. In fact, the panic assessment question in the current study is almost the same as the panic assessment questions used in all epidemiological studies on panic disorder based on the Diagnostic Interview Schedule. Whether cocaine precipitates a ‘panic-like experience’ or a ‘DSM panic attack’ is an open question for future research. Whereas this distinction is a crucial one for psychopathology in general, it may be of less central importance in our attempts to understand cocaine pharmacology and neurobiology of panic-like states (e.g., see Almodovar-Fabregas et al., 2002; Millan, 2003). Second, in this study, I have no data on temporal sequence of cocaine use and panic, but the associations tended to be larger when I excluded young people whose early assessments showed a panic-like experience before mid-adolescence (i.e., the assessments made between 1985-1994). I have been unable to locate any published 67 literature or speculation that panic attack precipitates cocaine use, or that individuals with panic attack self-medicate by taking cocaine. Self medication might help explain any observed alcohol-panic or anxiolytic-panic association, but I speculate that selfmedication with cocaine does not occur with panic attack. In its lack of detail about temporal sequencing, this report is somewhat like the currently accumulating case-control studies on risk of Alzheimer’s disease and mood disturbances in that there still remains some degree of possibility that the mood disturbances follow onset of Alzheimer’s disease rather than precede it. The same is true in this case-control study’s evidence about cocaine use and panic attack; I speculate that the cocaine use preceded the panic attack, but I do not have the evidence to test this speculation at present. Within our own study sample, I see no evidence that risk of starting cocaine use depends upon early panic experiences. In the overall sample, roughly 5% had tried cocaine by the time of the young adult assessment; among youths who had experienced an early panic-like experience by age 14 (calendar year 1994), only 3.2% had tried cocaine. Prompted by a reviewer's question, I note that the study participants were not promised a health benefit from participation, although some may derive a sense of personal satisfaction and well-being from their participation, and all participants receive a tangible monetary incentive large enough to overcome an increasing tendency toward refusal in the United States (e.g., see Atrostic et al., 2001). In some respects, the absence of direct health benefit may help to promote validity of the study estimates. For example, there is some concern about secondary gains that become tangible benefits in the form of health system responses only when the participant describes serious mental health or behavioral problems (e.g., see Craig et al., 1994). Under these circumstances, participants 68 might come to believe that there will be a health system response only when a certain threshold of problems are reported (e.g., only when the drug taking is accompanied by a psychiatric disturbance such as a panic attack). To thwart this type of participant belief, our introduction to the study makes clear that participation will not affect delivery of health services one way or the other. Since they were 8-10 years old, our participants have heard our interviewers introduce the study with language of the following form: "Whatever you tell me is going to be between you and me. You can tell me what you think and what you have done, and I am not going to tell anyone else what you say, unless something really scary or dangerous comes up. I will not tell your teacher or principal or parents what you say. Even though I will keep all of this a secret, you don't have to keep it a secret. You can tell anyone you want about this interview and the questions I am asking. But remember, I am not going to tell other people what you say to me. Is this okay with you?" This language helps to establish the private and confidential nature of the assessment, and it also has allowed the study team to take necessary actions when something truly scary or dangerous is reported, without violating the terms described above (e.g., active suicidal ideation and planning of a very depressed young person; a report of child abuse). As such, the data reported in this article should not be regarded as having the same qualities as data gathered in the context of a health visit. In some respects, the circumstances of this study's epidemiological field survey interviews represent limitations (e.g., the use of a lay interview method to assess panic as opposed to a diagnostic assessment by an experienced clinician-psychopathologist), but in other respects these same circumstances may promote validity of the estimated associations (e.g., avoidance of a clinician's diagnostic suspicion bias, or falsely positive reporting that 69 might occur in the health context, when the participant believes that there might be secondary gain). Notwithstanding limitations and circumstances such as these, the present study also possesses a number of counterbalanced strengths. As noted in the introduction, there are few epidemiological studies to investigate whether and by how much cocaine use might influence occurrence of panic attack or panic-like experiences. In addition, this study’s epidemiologically credible sample helps to constrain sources of bias and error that otherwise can complicate cross-sectional, case-control or retrospective research on highly selected samples. I also note that an association between cocaine use and panic is of theoretical interest in the study of anxiety disorders — even if the panic is a ‘panic like experience’ as opposed to the more specific concept of a ‘panic attack’ as laid out in the DSM-IV. Readers interested in the underlying neurobiological and pharmacological mechanisms that might subserve a cocaine-panic association will appreciate the early contribution of Geraciotti and Post (1991), as well as more recent contributions on cocaine use and psychostimulants generally (e.g., see Almodovar-Fabregas et al., 2002; Millan, 2003). In closing, I acknowledge that the size of the observed cocaine-panic association may be too modest to be of substantial public health importance, even if the relative risk value truly is as large as 3.0. Nonetheless, the suspected causal link from cocaine use to excess risk of panic-like experiences (however modest) remains of interest in the history of research on suspected hazards of using cocaine. In this study’s evidence, I cannot rule out the possibility that panic precipitates cocaine use ― although I assign low plausibility to this idea and cannot find it referenced in the published literature. However, as a study 70 strength, I have constrained other potential sources of confounding, via regression-based covariate adjustment and via use of subgroup analyses to eliminate the possibly confounding influences of tobacco, alcohol problems, and a history of depression syndromes. It is my hope that new epidemiological and clinical investigations can be completed with a more careful and systematic approach to assessment of panic attack, and with due attention to this type of methodological refinement. Section 5.2. Discussion Specific to Results Chapter Section 4.2. New Evidence on the Suspected Causal Association Linking Earlier Cocaine Use with Later Occurrence of Panic The main findings of this study confirm prior evidence on the suspected causal association linking cocaine use with subsequent panic attack-like events. In this study, temporal sequencing of the association has been assured to the extent possible in retrospective age of onset data, with a conservative approach when the first cocaine and panic occurred in the same year. (I have dropped from the analysis the panic events occurring in the same year). Before detailed discussion of these results, several of the more important study limitations merit attention, and some of the limitations mentioned in the prior section pertain to this dissertation sub-project as well (e.g., with regard to reporting of cocaine and other drug use; with respect to the concept of panic attack experience). With respect to the research design it is cross-sectional analyzed as longitudinal. More convincing 71 estimates might come from a truly longitudinal study with frequent follow-up assessments (e.g. with associated reductions in the likelihood of recall bias). With respect to the data analysis plan, it would have been preferable to have finegrained data about the timing of cocaine and panic attack. For example, I cannot look in this dataset if the panic occurred within days or weeks or months after onset of cocaine use, as would be required if first cocaine use precipitates the panic event (e.g., unexpected cocaine-induced tachycardia prompting acute anxiety about a possible heart attack). Notwithstanding limitations such as these, there are few epidemiological studies investigating whether and first how much cocaine use might influence later occurrence of panic attack in years after onset of cocaine use. In this study, I have been able to specify the temporal relationship of cocaine exposure and panic attack by using age of onset data in a new independent epidemiological sample. Also, I have constrained potential sources of confounding, via regression-based covariate adjustment and via use of subgroup analyses to constrain the suspected causal influences of tobacco, cannabis and a history of depression syndromes. The accumulating evidence on cocaine-associated panic may have important implications as I seek a more complete picture of cocaine-associated hazards. For example, panic attack has been found to aggravate the course of depression and other psychiatric disturbances (Goodwin 2002). In future research, it should be possible to investigate in epidemiological samples whether cocaine-associated panic attack carries this negative prognosis. Also, it may be possible to do a follow-up with fine-grained data to seek for specific time risk sets for panic experiences shortly after cocaine use (see 72 O’Brien & Anthony, 2005) and to look for a dose-response curve. It will be also interesting to look if this cocaine-panic association motivates cocaine users to change their behavior (i.e., to stop). Moreover, it may be possible to look for candidate genes that might explain why some cocaine users develop panic experiences while others do not as has been done for cocaine induced paranoia (Gelernter 2005). Section 5.3. Discussion Specific to Results Chapter Section 4.3. New Evidence on Risk of Becoming Cocaine Dependent: Epidemiological Estimates for the United States, 2002-2003 The main findings of this study may be summarized succinctly. First, based on these study estimates, within 24 months after starting cocaine use, an estimated 6-7% of these recent-onset cocaine users develop cocaine dependence. This estimate is not appreciably different from the estimate of 5%-6% reported by O’Brien and Anthony (2005). Also consistent with speculations offered by O’Brien and Anthony (2005), excess risk of cocaine dependence was found for females who recently started to use cocaine. Occurrence of cocaine dependence was greater among newly incident users who had started to smoke crack cocaine, as compared to those who had not smoked crack. No remarkable associations were found with the other variables under study (e.g., age, raceethnicity). Before detailed discussion of these results, several of the more important study limitations merit attention. Of central concern are the study’s self-report data, its 73 specifications for the interval of interest across a span of 0-23 months after onset of cocaine use, and the cocaine dependence assessment, which has a focus on the 12 months prior to NSDUH assessment. I note that not all active cocaine users are willing to provide self-reports about this illegal behavior (e.g., Harrison 2007). In addition, as noted in the methods section, at first glance, the 12-month recall interval for the NSDUH assessment of cocaine dependence differs from the 24-month interval for newly incident cocaine use; a concern may be lack of sensitivity of the NSDUH assessment with respect to cocaine dependence that starts and then stops before 12 months have passed from the first occasion of cocaine use. Drawing upon clinical observations about cases of cocaine dependence, it would seem that this rapid-onset-to-rapid-offset of cocaine dependence is not a very frequent response to cocaine use; cocaine dependence is described as a chronic condition, often with insidious onset and a prognosis of relapse in the absence of effective treatment (e.g., see Silva de Lima 2002; Kleber 2003; Rounsaville 2004; Kleber 2007). Nonetheless, it must be acknowledged that the NSDUH assessment method actually might yield an under-estimate of the probability of developing cocaine dependence soon after onset due to this incomplete sensitivity. Nonetheless, as noted by Wagner & Anthony (2002), Chen & Anthony (2005), O’Brien and Anthony (2005), among others, it is likely that there is an exceptionally small number of cases of cocaine dependence with very rapid onset of its clinical features, followed by very rapid offset of these clinical features, within a matter of a few months after onset of use. With respect to the research design, the NSDUH study is cross-sectional, not prospective. Nonetheless, as noted in the introduction, a prospective study of this topic necessarily 74 faces problems of sample attrition and losses to followup that are more pronounced than the non-participation issues faced in cross-sectional research. Chen and Anthony (2005) argue that the cross-sectional approach in this context actually may be superior to the prospective study – given the likelihood of selective losses to followup when cocaine dependence shifts into advanced stages, with disengagement from conventional social bonds and associated social maladaptation. With respect to the population under study, it can be said that findings based on the United States population may not have external validity and generalizability or be representative of human responses to cocaine use in all places and in all times. Replication elsewhere is the only solution to this potential study limitation. With respect of the sampling approach, study sample, and sample size, there is a concern that the NSDUH approach may neglect the most seriously affected cocaine users, who might well be underrepresented. It should be noted that this defect also is commonly present in prospective studies, which generally start with cross-sectional sampling and assessment. With respect to participation levels (and sample attrition), perhaps cocaine users are overrepresented among non-participants. The only solution to this limitation is to work at ways to secure essentially perfect participation or to bring non-participation under experimental control so as to derive better sensitivity analyses to this potential source of non-randomly missing data. 75 With respect to the data analysis plan, the use of the logistic regression approach can be justified. Nonetheless, in future research on this topic, it might be useful to reformulate the cocaine dependence outcome in an alternative to the DSM-IV diagnostic approach; for example, Chen and Anthony (2004) studied individual clinical features as interdependent response variables and used a GEE approach. McBride et al. (under review) have used an alternative “summarize, then analyze” approach in the form of latent structure models. As noted below, these alternative data analyses approaches represent new directions for this line of research. Notwithstanding limitations such as these, the present study also possesses a number of counterbalanced strengths. There is only one recent study in this topic; it used older data from 2000-2001, and required replication as completed here. As noted in the introduction, the cross-sectional approach is one that focuses attention on the newly incident recent onset cases. This is a strength in that recall problems over longer spans of time are constrained. In addition, as noted above, the approach helps to constrain problems of differential attrition that can occur in cross-sectional samples of the population and even in prospective studies. In light of these counterbalanced limitations and strengths, it is noteworthy that the original findings of O’Brien and Anthony (2005) with respect to estimated risk of cocaine dependence and male-female variation were substantiated in these new estimates. It is useful that these findings are observed to be replicated in a different epidemiological 76 sample from national surveys completed in CY 2002 and CY 2003, with different samples but with essentially the same approach otherwise. As compared to the 5%-6% estimate reported by O’Brien and Anthony, in this research with more recent samples, an estimated 6%-7% developed cocaine dependence soon after onset of cocaine use. The OBrien-Anthony observation of a female-associated excess risk of cocaine dependence soon after onset of cocaine use also has been replicated, even with statistical adjustment for other covariates. Although not formally tested here, the association of cocaine dependence with crack-smoking is noteworthy. In addition to its agreement with the work of O’Brien and Anthony (2005), there also is agreement with research conducted by Chen & Anthony (2004), who worked in an epidemiological sample of cocaine users from the 1990s. What might be the implications of study results of this type? In addition to quantifying the public health burdens associated with cocaine dependence and other potential harmful consequences of extra-medical drug use, the study estimates for cocaine can be placed into context with related estimates for other psychoactive drug compounds. For example, using different methods, Wagner and Anthony found that only 1-2% of newly incident drinkers of alcoholic beverages developed alcohol dependence during the first 1-2 years after onset of drinking, and that only 2-3% of newly incident cannabis smokers developed cannabis dependence during the 1-2 years after onset of cannabis smoking (Wagner & Anthony, 2002). By comparison, with respect to rapid transition from cocaine use to cocaine dependence, the estimated 5%-6% values observed by O’Brien & Anthony (2005) and the 6%-7% values observed in this study are congruent with pre-clinical 77 laboratory research on the reinforcing efficacy of cocaine self-administration, and with clinical research as well (Williams 2008; Hemby 2010). To be sure, the epidemiological estimates from these studies are not straightforward in their interpretation as indicators of the reinforcing functions served by drug self-administration; other variables such as relative drug availability are at play. Nonetheless, from a public policy perspective, it is useful to know that the likelihood of developing cocaine dependence soon after onset of use seems to be higher than is the case for other drugs of public concern. With respect to the observed male-female differences, it is necessary to think back toward prior discussions of sexual dimorphism and other sources of male-female variations in the occurrence of neuropsychiatric and behavioral disturbances. Weissman and Klerman (1977), Klerman and Weissman (1988) provided an exceptionally useful overview of male-female differences in the context of occurrence of major depression and related neuropsychiatric conditions. According to their review, there are several issues that should be considered in this context. First, as females were more likely than males to present depression, some people thought –at first-that this excess risk was an artifact due to the fact that women were more free than men to express their clinical features of depression because of social approval, but it was shown that the excess risk for females was real and not an artifact of response bias. I do not have indication in the present research that the observed male-female difference could be an artifact of response bias given the aim-blinded nature of the survey with respect to clinical features of cocaine dependence. Second, given that the excess rate for depression among women is real, I 78 should look for other explanations: a genetic theory was proposed, as well as an explanation with female endocrine physiology. More research on these topics is needed. In future research that builds from findings such as these, as this excess risk was seen also in preclinical studies (Hu 2004), it may be possible to look for replication in other epidemiological samples. If the sex difference finding is replicated, it may be possible to look for associated genes (Gelernter 2005) and interactions between genes and sex as observed for tobacco (Bierut, Madden, Breslau et al. 2007). It will be interesting to dig into other epidemiological facets as interactions with other variables (e.g., stress and stressful events). Section 5.4. Discussion Specific to Results Chapter Section 4.4. Who Starts then Stops Cocaine? The main findings of this study may be summarized succinctly. First, based on these study estimates, among those whose onset of cocaine use was within the four years prior to the date of the survey assessment, an estimated 13-16% had remained cocaine users within the prior 30 days; and an estimated 34% used cocaine within past 12 months but not in the past 30 days. An estimated 50-53% had not used cocaine in the previous 12 months before the assessment. It was found that people with age of cocaine onset 20 year old or more are two fold more likely to stop as the only finding from the 2003 CY 79 analysis that could be replicated in the 2008 analysis. I have not been able to find an association between sex, education, and crack use with cocaine cessation. Before detailed discussion of these results, several of the more important study limitations merit attention, in addition to those mentioned in prior sections of this chapter. Of central concern is left truncation of the cross-sectional sample. As noted in relation to the possibility for sample frame error, there is a potential non participation bias (e.g., systematic neglect of the most severely affected cocaine users). In addition, of special concern in relation to cocaine cessation research is left-truncation or censoring due to out-migration by death, which Rosenberg (2001) termed the “Len Bias” bias (e.g., an individual within the sample frame who started to use cocaine by January of CY of the assessment but died because of cocaine use before the actual survey, will not appear in the final sample). With respect to the assessment of the key response variables and the key covariates of interest, NSDUH has relied on self-reporting, as pointed out in the introduction. Despite of the fact that reliability of self-report measures is good, it may be possible-in the future- to add bioassays in a sub-sample, as was done in Harrison (2007). With respect to the cross-sectional design, it has been discussed that prospective designs are prone to bias once non-participation is taken into account (Chen 2005, O’Brien 2005), especially for severe cocaine users. I acknowledge that this potential selection bias is an important limitation of all epidemiological surveys in this topic. With respect to the conceptual model, it would have been better if the study had measured friend’s cocaine use as Kandel and Bates had done. Notwithstanding 80 limitations such as these, the present study also possesses counterbalanced strengths. There is one of the few studies on this subject with an epidemiological sample. The association between later onset of use (>=20 years old) and the likelihood of stopping has been found in the 2003 NSDUH sample and has been replicated in the 2008 sample. It is possible that early cocaine onset in adolescents (<20 years old) with an immature nervous system may be related to their decisions not to use or to stop use (Chambers 2003). The results from this study may have important implications as I seek to account for public health targets when dealing with cocaine stoppers. In future research that builds from findings such as these, it may be possible to do biological assessments and genetic testing in a subsample. Moreover, it will be possible to shorten the analysis time-periods, in order to diminish the likelihood of loss to follow-up, as well as to study in an epidemiological sample, the rapid onset and offset of cocaine use. 81 CHAPTER 6. CONCLUSIONS This chapter of the dissertation will cover a synopsis of the conclusions from each project. Section 6.1. Conclusions Specific to Results Chapter Section 4.1. Suspected Causal Association between Cocaine Use and Occurrence of Panic In the interest of full disclosure, I would like to draw the reader’s attention to the fact that the following paragraph of this section has been published in a journal article (Alvarado GF, Storr CL, Anthony JC. Suspected causal association between cocaine use and occurrence of panic. Subs Use Misuse, 2010;45:1019-32). The main finding of this retrospective case-control study of young adults is a modest but generally statistically robust association linking cocaine use and the occurrence of a panic attack-like experience, somewhat more modest than the published three fold relative risk estimates for this suspected causal association, which were based upon more broadly representative and older community samples. 82 Section 6.2. Conclusions Specific to Results Chapter Section 4.2. New Evidence on the Suspected Causal Association Linking Earlier Cocaine Use with Later Occurrence of Panic The main findings of this study confirm prior evidence on the suspected causal association linking cocaine use with subsequent panic attack-like events. In this study, temporal sequencing of the association has been assured to the extent possible in retrospective age of onset data, with a conservative approach when the first cocaine and panic occurred in the same year. Section 6.3. Conclusions Specific to Results Chapter Section 4.3. New Evidence on Risk of Becoming Cocaine Dependent: Epidemiological Estimates for the United States. The main findings of this study may be summarized succinctly. First, based on these study estimates, within 24 months after starting cocaine use, an estimated 6-7% of these recent-onset cocaine users develop cocaine dependence. Excess risk of cocaine dependence was found for females who recently started to use cocaine. Occurrence of cocaine dependence was greater among newly incident users who had started to smoke crack cocaine, as compared to those who had not smoked crack. No remarkable associations were found with the other variables under study (e.g., age, race-ethnicity). 83 Section 6.4. Conclusions Specific to Results Chapter Section 4.4. Who starts then stops cocaine? The main findings of this study may be summarized succinctly. First, based on these study estimates, among those whose onset of cocaine use was within the four years prior to the date of the survey assessment, an estimated 13-16% had remained cocaine users within the prior 30 days; and an estimated 34% used cocaine within past 12 months but not in the past 30 days. An estimated 50-53% had not used cocaine in the previous 12 months before the assessment. It was found that people with age of cocaine onset 20 year old or more are two fold more likely to stop as the only finding from the 2003 CY analysis that could be replicated in the 2008 analysis. I have not been able to find an association between sex, education, and crack use with cocaine cessation. 84 APPENDICES 85                     Appendix A: Tables to Accompany Chapter 4.1  86 Table 4.1.1. Characteristics of the baseline and follow-up samples in relation to occurrence of ‘panic’ in an urban public school system, 1985-2002.* Baseline (first grade) (1985- Young adulthood (2000-2002) Subset excluding youths with 1986) early panic-like experiences Number % Number % Number % 2,311 100.0 1,692 100.0 1,400 100.0 Female 1,160 50.2 902 53.3 734 52.4 Male 1,151 49.8 790 46.7 666 47.6 1,550 67.1 1,218 72.0 965 68.9 761 32.9 474 28.0 435 31.1 1985 1,196 51.8 856 50.6 724 51.7 1986 1,115 48.2 836 48.9 676 48.3 Total Sex Race/ ethnicity Minority Nonminority Year of first grade entry * Data were obtained from 1,692 participants at the time of the young adult follow-up (2000-2002), i.e., the successfully traced and re-recruited segment of the original first grade sample of 2311 children. 87 Table 4.1.2. Estimated association for use of cocaine among ‘panic’ cases and controls in an urban public school system, 1985-2002* Number of cases (n=641) Number of non-cases (n=1043) Cocaine use Yes 46 40 No 595 1003 p-value Crude odds ratio 0.003 Adjusted odds ratio ‡ 1.7 (1.1, 2.7) 0.020 Adjusted odds ratio § * 1.9 (1.3, 3.0) † 1.9 (1.1, 3.3) 0.014 Data were obtained from 1,692 participants at the time of the young adult follow-up (2000-2002), i.e., the successfully traced and re-recruited segment of the original first grade sample of 2311 children. † Numbers in parentheses, 95% confidence interval. ‡ Adjusted for being male, age and race-ethnicity. § Adjusted for being male, age and race-ethnicity plus covariate term for first grade teacher ratings of aggression and misbehavior at school. 88 Table 4.1.3. Subgroup variation in the association linking cocaine use with occurrence of ‘panic’ in an urban public school system 19852002,* with estimates based on a sample restricted to tobacco smokers, restricted to marijuana smokers, restricted to young adults without history of major depression syndrome and restricted to young people without a history of alcohol problems. All persons † All persons for whom there was no prior history of ‘panic’ ‡ OR§ 95% CI§ p OR 95% CI p No restrictions 1.9 1.1, 3.3 0.014 2.3 1,3, 4.0 0.005 Restricted to tobacco smokers 1.7 1.0, 3.0 0.057 1.9 1.1, 3.6 0.030 Restricted to marijuana smokers 1.6 0.9, 2.7 0.113 1.8 1.0, 3.2 0.060 Restricted to persons without history 2.4 1.2, 4.8 0.016 2.4 1.2, 4.8 0.018 1.2, 5.3 0.013 3.2 1.5, 7.1 0.004 of a depression syndrome Restricted to persons without history 2.5 of alcohol problems * Data were obtained from 1,692 participants at the time of the young adult follow-up (2000-2002), i.e., the successfully traced and re-recruited segment of the original first grade sample of 2311 children. † Irrespective of a history of early ‘panic’ and adjusted by age, sex, race-ethnicity and early aggressive behavior. ‡ Adjusted by age, sex, race-ethnicity and early aggressive behavior. § OR, odds ratio; CI, confidence interval. 89                     Appendix B: Tables to Accompany Chapter 4.2  90 Table 4.2.1. The NESARC sample of adults, United States, 2000-2001. All adults 18-44 year old (2001-2002) Unweighted Weighted % % 22,048 100.0 100.0 n Total Sex Female Male 12,374 9,674 Race/ ethnicity Non-Hispanic White Non-Hispanic Black/African American Hispanic Other 11,118 4,237 5,478 1,215 † Mean (SE). 91 50.6 49.4 31.7(0.05) † Age 56.1 43.9 31.5 (0.09) † 50.4 19.2 24.9 5.5 64.9 12.4 15.4 7.3 Table 4.2.2. Estimated association linking earlier onset of cocaine use with later occurrence of panic attack-like experiences. Data from the NESARC adult sample, United States, 2001-2002. Number of panic cases (n=2,688) Cocaine use-onset before year of first panic event Yes (n=1,716) No (n=19,629) 425 2,263 Unweighted Estimated Relative Risk & 95% Confidence Interval. Bivariate association, no covariates. Covariate-adjusted association ‡ 2.2 (1.9, 2.5) 2.4 (2.1, 2.7) 1,291 17,376 p-value <0.001 <0.001 ‡ Adjusted for being male, cohort and race-ethnicity. 92 Number of non-cases (n=18,667) Weighted Estimated Relative Risk & 95% Confidence Interval. 2.1 (1.8, 2.5) 2.6 (2.1, 3.1) p-value <0.001 <0.001 Table 4.2.3. Subgroup variation in the association linking cocaine use with occurrence of panic with estimates based on a sample restricted to tobacco smokers, restricted to cannabis smokers and restricted to adults without history of major depression syndrome. Data from NESARC, United States, 2001-2002. All adults 18-44 year old† Unweighted 95% CI§ p RR † Weighted 95% CI§ p RR † No restrictions 2.4 2.1, 2.7 <0.001 2.6 2.1, 3.1 <0.001 Restricted to tobacco smokers* 2.0 1.7, 2.4 <0.001 2.2 1.8, 2.7 <0.001 Restricted to cannabis smokers 1.5 1.3, 1.8 <0.001 1.8 1.4, 2.3 <0.001 Restricted to persons without history of a depression 2.5 2.0, 3.1 <0.001 3.0 2.2, 4.1 <0.001 syndrome † Adjusted by sex, cohort and race-ethnicity. § RR, Relative Risk; CI, confidence interval. * It was not possible to produce estimates for the non-smokers; most cocaine users had smoked both tobacco and cannabis. These restrictions constrain the possibility that the observed association is the entirely to tobacco or to cannabis. 93                    Appendix C: Tables to Accompany Chapter 4.3  94 Table 4.3.1. Sociodemographic Characteristics of All Persons, All Recently Active Past-Onset Users, and the Subset of RecentOnset Cocaine Users All persons Recently active past-onset Recent onset users of users of cocaine† cocaine‡ n unwtd% wtd% n unwtd% wtd% n unwtd% wtd% All persons 109 309 100.0 100.0 2587 100.0 100.0 1638 100.0 100.0 Sex Male 52 723 48.2 48.3 1635 63.2 69.3 882 53.9 44.6 Female 56 586 51.8 51.7 952 36.8 30.7 756 46.2 55.4 Age at interview (in years) 12-13 12 291 11.2 3.6 5 0.2 0.0 28 1.7 1.1 14-15 12 046 11.0 3.6 44 1.7 0.6 156 9.5 7.3 16-17 11 576 10.6 3.4 146 5.6 2.2 387 23.6 18.5 18-20 14 115 12.9 5.2 462 17.9 8.8 595 36.3 34.7 21-25 21 996 20.1 8.0 1127 43.6 20.4 415 25.3 25.0 26-34 11 092 10.2 14.8 350 13.5 25.7 52 3.2 12.8 35 and older 26 193 24.0 61.4 453 17.5 42.2 5 0.3 0.7 Race/ethnicity Non-Hispanic White 73 855 67.6 70.1 1898 73.4 68.9 1258 76.8 75.0 Non-Hispanic Black/African Am 13 528 12.4 11.5 178 6.9 14.4 51 3.1 4.1 Hispanic 14 542 13.3 12.5 370 14.3 13.2 227 13.9 15.9 Other 7 384 6.8 5.9 141 5.5 3.6 102 6.2 5.0 Education 11.6 6.5 107 16.2 10.4 270 22.7 13.4 14 629 College senior or graduate 24.0 21.6 353 28.8 28.1 728 22.3 18.8 20 552 Some college 23.5 23.6 387 28.9 30.5 790 28.7 23.0 25 146 High school graduate 41.0 48.3 791 26.1 30.9 799 26.3 44.1 Less than high school graduate 48 982 95 Table 4.3.1 (cont’d) All persons Recently active past-onset users of cocaine† n unwtd% wtd% 2587 100.0 100.0 n unwtd% wtd% All persons 109 309 100.0 100.0 Family income 0 to $19 999 25 354 23.2 19.5 824 31.9 $20 000-$49 000 41 935 38.4 37.9 1034 40.0 $50 000-$74 999 19 446 17.8 18.5 341 13.2 $75 000+ 22 574 20.7 24.2 388 15.0 Size of metropolitan statistical area MSA of 1 million+ 39 006 35.7 44.9 947 36.6 MSA < 1 million 40 775 37.3 33.2 1066 41.2 Segment not in MSA 29 528 27.0 21.9 574 22.2 † Recently active use in the past 12 months, but with onset 2+ years before survey assessment date. ‡ Within 24 months of the survey assessment date. Data from 2002-2003 National Survey on Drug Use and Health. 96 Recent onset users of cocaine‡ n unwtd% wtd% 1638 100.0 100.0 29.3 39.4 14.2 17.2 503 603 247 285 30.7 36.8 15.1 17.4 29.6 37.5 14.6 18.4 49.3 34.0 16.8 511 669 458 31.2 40.8 30.0 42.2 37.0 20.8 Table 4.3.2. Selected Drug Use Characteristics of All Persons, All Recently Active Cocaine Users, and the Subset of RecentOnset Cocaine Users All persons Recently active past-onset Recent onset users of users of cocaine† cocaine‡ n unwtd% wtd% n unwtd% wtd% n unwtd% wtd% All persons 109 309 100.0 100.0 2587 100.0 100.0 1638 100.0 100.0 Occurrence of DSM-IV cocainedependence syndrome Yes, 3+ clinical features 545 0.5 0.4 431 16.7 18.3 111 No 108 764 99.5 99.6 2156 83.3 81.7 1527 Occasions of cocaine use (all forms) in past 12 months 1-2 days 1 042 1.0 0.6 485 18.8 17.5 557 3-11 days 1 226 1.1 0.8 792 30.6 29.2 434 12-100 days 1 249 1.1 0.9 953 36.8 37.3 296 101 or more days 415 0.4 0.3 357 13.8 15.9 58 Never/not in past year/dk/ref 105 377 96.4 97.5 0 0.0 0.0 293 Crack use in lifetime Yes 3 501 3.2 3.5 1170 45.2 50.6 234 No 105 808 96.8 96.5 1417 54.8 49.4 1404 Ever used needle to inject cocaine 19 12.1 219 8.5 1.0 0.7 760 Yes 1619 91.5 87.9 2368 99.3 99.0 108 549 No Number of drugs used by age 11 years 88.7 87.5 0 95 666 1873 72.4 75.5 1250 9.4 9.8 10 729 1 441 17.1 15.0 274 1.5 2.1 2 2 264 194 7.5 6.9 73 0.3 0.6 650 3 79 3.1 2.6 41 † Recently active use in the past 12 months, but with onset 2+ years before survey assessment date. ‡ Within 24 months of the survey assessment date. Data from 2002-2003 National Survey on Drug Use and Health. 97 6.8 93.2 6.1 93.9 34.0 26.5 18.1 3.5 17.9 33.8 26.2 16.6 3.6 19.8 14.3 85.7 12.2 87.8 1.2 98.8 0.8 99.2 76.3 16.7 4.5 2.5 77.3 15.9 4.5 2.3 Table 4.3.3. Relative Risk Estimates for Becoming Cocaine Dependent Among Recent-Onset Cocaine Users, Without Statistical Adjustments Unweighted† Weighted‡ Estimated risk of becoming Estimated risk of dependent becoming dependent Number Number of % RR 95% p-value RR 95% p-value of cocaineCI CI recent dependence onset cases cocaine users All persons 1638 111 6.8 Sex Male 882 39 4.3 1.0 1.0 Female 756 72 9.5 2.3 1.5-3.4 <0.001 2.1 1.2-3.7 0.01 Age at interview (in years) 26+ 57 3 5.2 0.8 0.2-2.5 0.64 0.8 0.2-3.0 0.75 21-25 415 18 4.3 0.6 0.3-1.1 0.09 0.5 0.3-1.1 0.08 18-20 (ref) 595 41 6.9 1.0 1.0 16-17 387 32 8.3 1.2 0.8-2.0 0.42 1.1 0.6-2.1 0.72 14-15 156 15 9.6 1.4 0.8-2.7 0.25 1.6 0.8-3.3 0.19 12-13 28 2 7.1 1.0 0.2-4.5 0.95 0.4 0.1-2.0 0.27 Race/ethnicity Non-Hispanic White (ref) 1258 83 6.6 1.0 1.0 Non-Hispanic Black 51 3 5.9 0.9 0.3-2.9 0.84 1.4 0.3-6.5 0.63 Hispanic 227 17 7.5 1.1 0.7-2.0 0.62 1.2 0.5-2.8 0.67 Other 102 8 7.8 1.2 0.6-2.6 0.63 0.6 0.2-1.9 0.43 98 Table 4.3.3 (cont’d) All persons Education College Senior or graduate Some college High school graduate Less than high school graduate (ref) Family income 0 to $19 999 $20 000-$49 000 (ref) $50 000-$74 999 $75000+ Population density MSA of 1 million+ MSA < 1 million (ref) Segment not in MSA Crack use in lifetime No (ref) Yes Number of recent onset cocaine users 1638 Number of cocainedependence cases Unweighted† Weighted‡ Estimated risk of becoming Estimated risk of dependent becoming dependent % RR 95% p-value RR 95% p-value CI CI 111 6.8 107 353 387 791 3 21 23 64 2.8 5.9 5.9 8.1 0.3 0.7 0.7 1.0 0.1-1.1 0.4-1.2 0.4-1.2 0.06 0.20 0.19 0.5 0.8 1.1 1.0 0.1-2.1 0.4-1.6 0.5-2.1 0.34 0.60 0.87 503 603 247 285 27 34 25 25 5.4 5.6 10.1 8.8 0.9 1.0 1.9 1.6 0.6-1.6 0.84 0.4-1.4 0.38 1.1-3.2 0.9-2.8 0.02 0.08 0.8 1.0 2.3 1.7 1.1-5.0 0.9-3.4 0.03 0.12 511 669 458 33 45 33 6.5 6.7 7.2 1.0 1.0 1.1 0.6-1.6 0.85 0.4-1.5 0.46 0.7-1.7 0.76 0.8 1.0 0.9 0.5-1.6 0.68 1404 234 73 38 5.2 16.2 1.0 3.5 2.3-5.4 <0.001 1.0 3.2 1.8-5.8 <0.001 99 Table 4.3.3 (cont’d) Number of recent onset cocaine users 1638 Number of cocainedependence cases Unweighted† Weighted‡ Estimated risk of becoming Estimated risk of dependent becoming dependent % RR 95% p-value RR 95% p-value CI CI All persons 111 6.8 Ever used needle to inject cocaine Yes 19 2 10.5 No (ref) 1619 109 6.7 Number of drugs used by age 11 years 0 (ref) 1250 84 6.7 1 274 18 6.6 2 73 5 6.8 3 41 4 9.8 † Estimated based on logistic regression. ‡ Data based on variance estimates via Taylor Series linearization. Data from 2002-2003 National Survey on Drug Use and Health. 100 1.6 1.0 1.0 1.0 1.0 1.5 0.4-7.1 0.6-1.7 0.4-2.6 0.5-4.3 0.51 0.93 0.97 0.45 2.4 1.0 0.414.7 0.34 1.0 0.9 0.8 0.3 0.5-1.9 0.3-2.5 0.1-1.1 0.85 0.71 0.08 Table 4.3.4. Relative Risk Estimates for Becoming Cocaine Dependent among Recent-Onset Cocaine Users, with Statistical Adjustment for All Listed Covariates Unweighted† Weighted‡ RR 95%CI p-value RR 95%CI p-value Recent onset users n=1638 Sex Male (ref) 39 1.0 1.0 Female 72 2.4 1.6-3.6 <0.001 2.2 1.3-3.7 0.005 Age group (in years) 26+ 3 0.7 0.2-2.5 0.64 0.8 0.2-2.7 0.68 21-25 18 0.6 0.4-1.2 0.14 0.5 0.3-1.1 0.10 18-20 (ref) 41 1.0 1.0 16-17 32 1.1 0.7-1.8 0.76 1.0 0.5-1.9 0.93 14-15 15 1.1 0.6-2.2 0.68 1.3 0.6-2.8 0.57 12-13 2 0.7 0.1-3.2 0.62 0.3 0.0-1.7 0.16 Race/ethnicity Non-Hispanic White (ref) 83 1.0 1.0 Non-Hispanic Black 3 1.1 0.3-3.7 0.88 2.2 0.5-9.8 0.28 Hispanic 17 1.3 0.7-2.2 0.35 1.6 0.7-3.9 0.26 Other 8 1.3 0.6-2.8 0.51 0.8 0.2-2.3 0.62 Family income 0.31 0.4-1.4 0.7 0.84 0.6-1.6 0.9 0-$19 999 27 1.0 1.0 34 $20 000-$49 000 (ref) 0.06 1.0-5.5 2.3 0.02 1.1-3.3 1.9 25 $50 000-$74 999 0.08 0.9-4.2 0.06 1.9 0.6-5.3 1.7 $75 000+ 25 101 Table 4.3.4 (cont’d) RR Unweighted† 95%CI p-value RR Recent onset users n=1638 Number of drug used by age 11 years 0 (ref) 84 1.0 1.0 1 18 1.0 0.5-1.6 0.86 0.9 2 5 1.2 0.4-3.0 0.77 0.9 3 4 1.7 0.6-5.3 0.34 0.4 † Estimated based on logistic regression, with statistical adjustment for covariates. ‡ Data based on variance estimates via Taylor Series linearization with statistical adjustment for covariates. Data from 2002-2003 National Survey on Drug Use and Health. 102 Weighted‡ 95%CI p-value 0.5-1.9 0.3-2.7 0.1-1.6 0.84 0.81 0.22                       Appendix D: Tables to Accompany Chapter 4.4  103 Table 4.4.1A. Selected Sociodemographic Characteristics of All Persons, and the Subset of Cocaine Users All persons Users of cocaine‡ n unwt wt % n unwt wt % All persons 55 230 100.0 100.0 2 376 100.0 100.0 Sex Male 26 766 48.5 48.4 1 316 55.4 56.8 Female 28 464 51.5 51.6 1060 44.6 43.2 Age at interview (in years) 12-13 6 095 11.0 3.5 18 0.8 0.6 14-15 6 083 11.0 3.5 116 4.9 3.6 16-17 6 026 10.9 3.5 326 13.7 9.2 18-20 7 212 13.1 5.3 798 33.6 29.9 21-25 11 171 20.2 8.1 963 40.5 36.3 26-34 5 562 10.1 14.7 123 5.2 14.9 35 and older 13 081 23.7 61.4 32 1.4 5.4 Age of cocaine onset 11.9 376 15.8 1.0 895 1.6 15 or younger 24.8 20.5 590 2.2 2.7 16-17 1 466 25.8 23.2 2.9 613 2.9 1 592 18-19 797 33.5 44.4 8.5 2 929 5.3 20 and older 0 87.5 87.3 48 348 Never used Race/ethnicity 71.3 73.3 1 742 70.0 66.1 36 502 Non-Hispanic White 3.7 3.0 70 11.6 12.5 6 901 Non-Hispanic Black 19.7 17.3 410 12.6 14.0 7 753 Hispanic 5.2 6.5 154 5.9 7.4 Other 4 074 ‡ Within four years of the survey assessment date. Data from 2003 National Survey on Drug Use and Health. 104 Table 4.4.1B. Selected Sociodemographic Characteristics of All Persons, and the Subset of Cocaine Users All persons n 55 230 All persons Education College senior or graduate 7 178 Some college 10 232 High school graduate 12 536 Less than high school graduate 7070 12-17 yo 18204 Family income 0 to $19 999 13 196 $20 000-$49 000 21 016 $50 000-$74 999 9 670 $75 000+ 11 348 Size of metropolitan statistical area MSA of 1 million+ 19 467 MSA < 1 million 20 678 Segment not in MSA 15 085 Number of drugs used prior to cocaine 48 459 None 228 1 601 2 5 942 3 or above Crack use in lifetime 1798 Yes 53408 No ‡ Within four years of the survey assessment date. Data from 2003 National Survey on Drug Use and Health. unwt 100.0 wt % 100.0 Users of cocaine‡ n 2 376 13.0 18.5 22.7 12.8 33.0 23.0 22.3 28.2 16.0 10.5 176 563 653 524 460 7.4 23.7 27.5 22.1 19.4 9.6 25.7 28.4 22.9 13.4 23.9 38.1 17.5 20.6 19.5 37.5 18.7 24.2 758 945 340 333 31.9 39.8 14.3 14.0 30.6 40.1 14.6 14.7 35.3 37.4 27.3 45.0 33.2 21.8 744 975 657 31.3 41.0 27.7 41.2 37.3 21.5 87.7 0.4 1.1 10.8 85.5 0.4 1.4 12.7 30 55 171 2120 1.3 2.3 7.2 89.2 1.3 2.3 8.7 87.7 3.3 96.7 3.4 96.6 501 1875 21.1 78.9 19.0 81.0 105 unwt 100.0 wt % 100.0 Table 4.4.2A. Relative Risk Estimates for Stopping Cocaine Use Among Cocaine Users, Without Statistical Adjustments Estimated Number of Number of stoppers (for risk for cocaine at least one stopping users ‡ year) wt % RR 95% CI pvalue All persons 2 376 1 098 49.7 Sex Female (ref) 1060 517 49.8 1.0 Male 1 316 581 49.7 1.0 0.7, 1.4 0.96 Age of cocaine onset 15 or younger (ref) 376 159 44.1 1.0 16-17 590 249 45.2 1.4 0.9, 2.2 0.18 18-19 613 272 45.9 1.2 0.8, 1.8 0.42 20 and older 797 418 55.3 1.9 1.2, 3.1 0.01 Race/ethnicity Non-Hispanic White (ref) 1 742 767 47.2 1.0 Non-Hispanic Black 70 38 73.0 3.1 1.4, 6.9 0.01 Hispanic 410 225 55.7 1.1 0.7, 1.8 0.57 Other 154 68 45.5 0.5 0.3, 1.0 0.06 Education College senior or graduate 176 91 48.6 0.6 0.3, 1.1 0.13 Some college 563 250 47.6 0.8 0.5, 1.4 0.47 High school graduate 653 341 57.0 1.5 0.9, 2.5 0.10 Less than high school graduate 524 281 54.6 1.0 (ref) 460 135 30.7 0.4 0.2, 0.6 12-17 yo <0.001 ‡ Within four years of the survey assessment date. Data from 2003 National Survey on Drug Use and Health. 106 Table 4.4.2B. Relative Risk Estimates for Stopping Cocaine Use Among Cocaine Users, Without Statistical Adjustments Estimated Number of Number of stoppers (for risk for cocaine stopping at least one users‡ year) wt % RR 95% CI pvalue All persons 2 376 1 098 49.7 Size of metropolitan statistical area 1.0 349 48.0 744 MSA of 1 million+ (ref) 0.29 0.8, 1.8 50.8 1.2 457 MSA < 1 million 975 0.18 0.9, 2.1 1.3 51.1 657 292 Segment not in MSA Number of drugs used prior to cocaine 0.08 0.9,14.3 3.5 59.2 15 30 None 0.7, 4.9 0.21 1.9 60.6 33 55 1 0.01 1.3, 5.2 2.6 66.5 108 171 2 1.0 47.6 942 2 120 3 or above (ref) Crack use in lifetime No (ref) 1 875 897 50.4 1.0 Yes 501 201 47.0 0.7 0.5, 1.0 0.08 ‡ Within four years of the survey assessment date. Data from 2003 National Survey on Drug Use and Health. 107 Table 4.4.3. Relative Risk Estimates for Stopping Cocaine Use, with Statistical Adjustment. RR 95%CI p-value Users (n=2 376) Sex Male 1.1 0.7, 1.5 Female (ref) 1.0 Age of cocaine onset 15 or younger (ref) 1.0 16-17 1.6 1.0, 2.6 18-19 1.4 0.9, 2.3 20 and older 2.2 1.3, 3.7 Race/ethnicity Non-Hispanic White (ref) 1.0 Non-Hispanic Black 2.6 1.2, 5.9 Hispanic 1.0 0.6, 1.6 Other 0.5 0.3, 0.9 Number of drugs used prior to cocaine 5.5 1.2,24.5 None 0.9, 6.6 2.4 1 1.4, 5.6 2 2.8 3 or above (ref) 1.0 Crack use in lifetime 0.5, 1.1 No (ref) 1.0 Yes 0.7 Adjusted by listed covariates and family income, population density Data from 2003 National Survey on Drug Use and Health. 108 0.77 0.08 0.12 0.003 0.02 0.96 0.04 0.03 0.08 0.005 0.12 Table 4.4.4A. Selected Sociodemographic Characteristics of All Persons, and the Subset of Cocaine Users All persons Users of cocaine‡ n unwt wt % n unwt wt % All persons 55 739 100.0 100.0 2 227 100.0 100.0 Sex Male 26 743 48.0 48.5 1 232 55.3 55.4 Female 28 996 52.0 51.5 995 44.7 44.6 Age at interview (in years) 12-13 5 477 9.8 3.1 15 0.7 0.4 14-15 6 028 10.8 3.4 66 3.0 1.9 16-17 6 337 11.4 3.5 253 11.4 7.3 18-20 7 682 13.8 5.3 754 33.9 28.3 21-25 11 459 20.6 7.9 979 44.0 37.2 26-34 5 571 10.0 14.3 128 5.8 18.1 35 and older 13 185 23.7 62.6 32 1.4 6.8 Age of cocaine onset 8.7 283 12.7 1.0 793 1.4 15 or younger (ref) 24.9 20.8 554 2.5 2.7 16-17 1 506 28.4 21.7 3.1 633 3.0 1 666 18-19 757 34.0 48.8 8.2 2 542 4.7 20 and older 0 88.2 85.2 49 132 Never used Race/ethnicity 75.9 71.4 1 589 67.8 62.0 34 549 Non-Hispanic White 4.5 3.2 72 11.8 13.0 7 725 Non-Hispanic Black 16.2 17.2 383 14.0 16.4 9 122 Hispanic 3.4 8.2 183 6.3 8.7 Other 4 843 ‡ Within four years of the survey assessment date. Data from 2008 National Survey on Drug Use and Health. 109 Table 4.4.4B. Selected Sociodemographic Characteristics of All Persons, and the Subset of Cocaine Users All persons n 55 739 Users of cocaine‡ n 2 227 unwt wt % unwt All persons 100.0 100.0 100.0 Education College senior or graduate 7 555 13.6 24.9 194 8.7 Some college 10 909 19.6 23.0 598 26.7 High school graduate 12 714 22.8 28.3 690 31.0 Less than high school graduate 6 719 12.1 13.9 411 18.5 12-17 year olds 17 842 32.0 10.0 334 15.0 Family income 0 to $19 999 12 024 21.6 16.8 659 29.6 $20 000-$49 000 19 051 34.2 32.4 791 35.5 $50 000-$74 999 9 974 17.9 18.6 312 14.0 $75 000+ 14 690 26.4 32.3 465 20.9 Size of metropolitan statistical area MSA of 1 million+ 23 303 41.8 51.8 875 39.3 MSA < 1 million 27 514 49.4 41.9 1190 53.4 Segment not in MSA 4 922 8.8 6.3 162 7.3 Number of drugs used prior to cocaine None 49 265 88.4 85.4 41 1.8 1 277 0.5 0.7 64 2.9 2 643 1.2 1.5 180 8.1 3 or above 5 554 10.0 12.4 1942 87.2 Crack use in lifetime 16.3 363 3.3 2.9 Yes 1624 83.7 96.7 1861 97.1 54086 No ‡ Within four years of the survey assessment date. Data from 2008 National Survey on Drug Use and Health. 110 wt % 100.0 14.3 30.3 29.1 16.8 9.6 26.3 36.3 14.0 23.5 50.3 45.6 4.1 1.7 2.6 7.1 88.6 16.1 83.9 Table 4.4.5A. Relative Risk Estimates for Stopping Cocaine Use Among Cocaine Users, Without Statistical Adjustments Estimated Number of Number of stoppers (for risk for cocaine stopping at least one users‡ year) wt % RR 95% CI p-value All persons 2 227 1 155 53.3 Sex Female (ref) 995 527 56.6 1.0 Male 1 232 628 50.7 0.9 0.6, 1.3 0.49 Age of cocaine onset 15 or younger (ref) 283 137 49.1 1.0 16-17 554 263 47.6 1.1 0.6, 1.8 0.76 18-19 633 327 50.7 1.4 0.8, 2.3 0.21 20 and older 757 428 57.7 1.8 1.0, 3.3 0.07 Race/ethnicity Non-Hispanic White (ref) 1 589 811 53.5 1.0 Non-Hispanic Black 72 35 46.0 0.4 0.1, 1.2 0.10 Hispanic 383 212 56.9 0.7 0.4, 1.4 0.33 Other 183 97 42.4 0.5 0.1, 1.5 0.21 Education College senior or graduate 194 105 53.4 1.5 0.6, 4.0 0.43 Some college 598 299 53.1 0.9 0.5, 1.8 0.76 High school graduate 690 381 54.9 1.1 0.7, 1.8 0.61 Less than high school graduate 411 255 61.8 1.0 (ref) 334 115 34.1 0.4 0.2, 0.7 0.003 12-17 year olds ‡ Within four years of the survey assessment date. Data from 2008 National Survey on Drug Use and Health. 111 Table 4.4.5B. Relative Risk Estimates for Stopping Cocaine Use Among Cocaine Users, Without Statistical Adjustments Estimated Number of Number of stoppers (for risk for cocaine stopping at least one users ‡ year) wt % RR 95% CI pvalue All persons 2 227 1 155 53.3 Size of metropolitan statistical area 1.0 414 50.5 875 MSA of 1 million+ (ref) 0.02 1.1, 2.5 55.8 1.6 652 MSA < 1 million 1190 0.57 0.4, 4.3 1.4 60.3 162 89 Segment not in MSA Number of drugs used prior to cocaine 0.79 0.3, 2.9 0.9 54.2 22 41 None 0.6, 5.9 0.32 1.8 62.8 41 64 1 0.43 0.6, 2.7 1.3 67.2 117 180 2 1.0 51.9 975 1 942 3 or above (ref) Crack use in lifetime No (ref) 1 861 987 54.5 1.0 Yes 363 167 47.5 0.6 0.3, 1.3 0.19 ‡ Within four years of the survey assessment date. Data from 2008 National Survey on Drug Use and Health. 112 Table 4.4.6. Relative Risk Estimates for Stopping Cocaine Use, with Statistical Adjustment. RR 95%CI p-value Users (n=2 227) Sex Male 0.9 0.6, 1.3 Female (ref) 1.0 Age of cocaine onset 15 or younger (ref) 1.0 16-17 1.0 0.6, 1.8 18-19 1.4 0.8, 2.5 20 and older 2.1 1.1, 4.0 Race/ethnicity Non-Hispanic White (ref) 1.0 Non-Hispanic Black 0.4 0.1, 1.0 Hispanic 0.7 0.4, 1.4 Other 0.5 0.1, 1.7 Number of drugs used prior to cocaine 1.5 0.4, 5.4 None 0.7, 8.6 2.4 1 0.7, 3.7 2 1.6 3 or above (ref) 1.0 Crack use in lifetime 0.3, 1.3 No (ref) 1.0 Yes 0.6 Adjusted by listed covariates and family income, population density Data from 2008 National Survey on Drug Use and Health. 113 0.51 0.85 0.22 0.03 0.05 0.33 0.25 0.53 0.16 0.25 0.17 REFERENCES 114 REFERENCES 1. Almodovar-Fabregas LJ, Segarra O, Colon N et al. Effects of cocaine administration on VTA cell activity in response to prefrontal cortex stimulation. Ann N Y Acad Sci, 2002, 965: 157-171. 2. American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders, 3rd Ed. Washington, DC, American Psychiatric Association; 1981. 3. American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders, 3rd Ed Rev. Washington, DC, American Psychiatric Association; 1987. 4. American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders, 4rd Ed. Washington, DC, American Psychiatric Association; 1994. 5. American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders, 4rd Ed. Text Rev. Washington, DC, American Psychiatric Association; 2000. 6. Alvarado GF, Anthony JC. [Epidemiology of cocaine use and dependence], Tratado Set de Trastornos Adictivos, Madrid. Editorial Medica Panamericana, 2006. 7. Alvarado GF, Storr CL, Anthony JC. Suspected causal association between cocaine use and occurrence of panic. Subs Use Misuse, 2010;45:1019-32. 8. Anthony JC, Chen CY. Epidemiology of Drug Dependence. In: Textbook of Substance Abuse Treatment, Washington: American Psychiatric Publishing, Inc; 2003. p. 55-71. 9. Anthony JC, Forman V. Toward a Drugs and Crime Research Agenda for the 21st Century. In: At the intersection of public health and criminal justice research on drugs and crime, Washington D.C. National Institute of Justice; 2003. p. 283–342. 10. Anthony JC, Helzer JE. Epidemiology of Drug Dependence. In: Psychiatric Epidemiology, New York: John Wiley and Sons, Inc; 2002. p. 479-561. 11. Anthony JC, Petronis KR. Epidemiologic evidence on suspected associations between cocaine use and psychiatric disturbances. NIDA Res Monogr, 1991, 110: 71-94. 115 12. Anthony JC, Petronis KR. Suspected risk factors for depression among adults 1844 years old. Epidemiology, 1991, 2: 123-132. 13. Anthony JC, Van Etten ML. Epidemiology and its Rubrics. In: Comprehensive Clinical Psychology. Oxford, UK: Elsevier Science Publication; 1998. p. 355-89. 14. Anthony JC, Tien AY, Petronis KR. Epidemiologic evidence on cocaine use and panic attacks. Am J Epidemiol, 1989, 129: 543-549. 15. Anthony JC, Folstein M, Romanoski AJ. Comparison of the lay Diagnostic Interview Schedule and a standardized psychiatric diagnosis. Experience in eastern Baltimore. Arch Gen Psychiatry, 1985, 42: 667-675. 16. Anthony JC, Warner LA, Kessler RC. Comparative epidemiology on dependence on tobacco, alcohol, controlled substances, and inhalants: basic findings from the National Comorbidity Survey. Exp Clin Psychopharmacol. 1994; 2: 244-68. 17. Aronson TA, Craig TJ. Cocaine precipitation of panic disorder. Am J Psychiatry, 1986, 143: 643-645. 18. Atrostic BK, Bates N, Burt G, Silberstein, A. Non response in US government household surveys: consistent measures, recent trends, and new insights. Journal of Official Statistics, 2001, 17: 209-226. 19. Bandstra ES, Morrow CE, Vogel AL et al. Longitudinal influence of prenatal cocaine exposure on child language functioning. Neurotoxicology and Teratology. 2002; 24:297-308. 20. Barroso M, Gallardo E, Queiroz JA. Bioanalytical methods for the determination of cocaine and metabolites in human biological samples. Bioanalysis, 2009, 1(5): 977-1000. 21. Behrens DA, Caulkins JP, Tragler G, Feichtinger G. Optimal control of drug epidemics: prevent and treat-but not at the same time? Management Science, 2000, 46(3):333-47. 22. Bierut LJ, Madden PA, Breslau N et al. Novel genes identified in a high-density genome wide association study for nicotine dependence. Hum Mol Genet. 2007;16(1):24-35. 23. Breslau N, Klein DF. Smoking and panic attacks. An epidemiologic investigation. Arch Gen Psychiatry, 1999, 56: 1141-1147. 24. Campbell ND. Discovering addiction. The science and politics of substance abuse research. Ann Arbor, the University of Michigan Press, 2007. 116 25. Carrera MR, Meijler MM, Janda KD. Cocaine pharmacology and current pharmacotherapies for its abuse. Bioorg Med Chem, 2004, 12(19): 5019-30. 26. Carroll KM, Felton LR, Ball SA, Nich C, Frankforter TL, Shi J, Rounsaville BJ. Efficacy of disulfiram and cognitive behavior therapy in cocaine-dependent outpatients: a randomized placebo-controlled trial. Arch Gen Psychiatry, 2004, 61(3): 264-72. 27. Cartmell LW, Aufderhide A, Weems C. Cocaine metabolites in pre-Columbian mummy hair. J Okla State Med Assoc, 1991, 84(1): 11-2. 28. Caulkins JP, Pacula RL, Paddock S, Chiesa J. What we can-and cannot-expect from school-based drug prevention. Drug Alcohol Rev, 2004, 23(1): 79-87. 29. Chakroun N, Doron J, Swendsen J. [Substance use, affective problems and personality traits: test of two association models]. Encephale, 2004, 30(6): 564-9. 30. Chambers RA, Taylor JR, Potenza MN. Developmental neurocircuitry of motivation in adolescence: a critical period of addiction vulnerability. Am J Psychiatry, 2003, 160(6): 1041-52. 31. Chen CY, Anthony JC. Epidemiological estimates of risk in the process of becoming dependent upon cocaine: cocaine hydrochloride powder versus crack cocaine. Psychopharmacology. 2004;172:78-86. 32. Chen CY, O’Brien MS, Anthony JC. Who becomes cannabis dependent soon after onset of use? Epidemiological evidence from the United States: 2000-2001. Drug Alcohol Depend, 2005, 79(1):11-22. 33. Chen CY, Storr CL, Anthony JC. Early-onset drug use and risk for drug dependence problems. Addict Behav. 2009;34(3):319-22. 34. Chilcoat HD, Dishion TJ, Anthony JC. Parent monitoring and the incidence of drug sampling in urban elementary school children. Am J Epidemiol, 1995, 141: 25-31. 35. Columb MO, Ramsaran R. Local Anaesthetic Agents. Anaesth Intens Care, 2010,11(3):113-117. 36. Cornish JW, O’Brien CP. Crack cocaine abuse: an epidemic with many public health consequences. Ann Rev Public Health, 1996, 17:259-73. 37. Craig TK, Drake H, Mills K, Boardman AP. The South London Somatisation Study. II. Influence of stressful life events, and secondary gain. Br J Psychiatry, 1994, 165(2):248-258. 117 38. Crum RM, Lillie-Blanton M, Anthony JC. Neighborhood environment and opportunity to use cocaine and other drugs in late childhood and early adolescence. Drug Alcohol Depend. 1996;43:155-61. 39. De Alarcon R. The spread of heroin abuse in a community. Community Health. 1969;1(3):155-61. 40. Degenhardt L, Chiu WT, Sampson N, et al. Toward a global view of alcohol, tobacco, cannabis, and cocaine use: findings from the WHO World Mental Health Surveys. PLoS Med 2008;5(7):1053-1067. 41. Demyttaenaere K, Bruffaets R, Posada-Villa J et al. Prevalence, severity, and unmet need for treatment of mental disorders in the World Health Organization World Mental Health Surveys. JAMA. 2004;291(21):2581-90. 42. Dormitzer CM, Gonzalez GB, Penna M et al. The PACARDO research project: youthful drug involvement in Central America and the Dominican Republic. Rev Panam Salud Publica. 2004;15(6):400-16. 43. Eaton WW, Anthony JC, Romanowsky A et al. Onset and recovery from panic disorder in the Baltimore Epidemiologic Catchment Area follow-up. Br J Psychiatry, 1998, 173: 501-507. 44. Eaton WW, Kramer M, Anthony JC, Dryman A, Shapiro S, Locke BZ. The incidence of specific DIS/DSM-III mental disorders: data from the NIMH Epidemiologic Catchment Area Program. Acta Psychiatr Scand, 1989, 79: 163178. 45. Fergusson DM, Horwood LJ. The Christchurch Health and Development Study: review of findings on child and adolescent mental health. Aust N Z J Psychiatry, 2001, 35: 287-296. 46. Foltin RW, Fishman MW. Residual effects of repeated cocaine smoking in humans. Drug Alcohol Depend. 1997, 47(2): 117-24. 47. Galduroz JC, Noto AR, Nappo SA, Carlini EA. Household survey on drug abuse in Brazil: Study involving the 107 major cities of the country-2001. Addict Behav. 2005;30(3):545-56. 48. Galduroz JC, Noto AR, Nappo SA, Carlini EA. Trends in drug use among students in Brazil: analysis of four surveys in 1987, 1989, 1993 and 1997. Braz J Med Biol Res 2004, 37(4): 523-31. 118 49. Gelernter J, Panhuysen C, Weiss R et al. Genomewide linkage scan for cocaine dependence and related traits: significant linkages for a cocaine-related trait and cocaine-induced paranoia. Am J Med Genet B Neuropsychiatr Genet. 2005;136B(1):45-52. 50. Geracioti TD, Post RM. Onset of panic disorder associated with rare use of cocaine. Biol Psychiatry, 1991, 29: 403-406. 51. Golub A, Johnson B. A recent decline in cocaine use among youthful arrestees in Manhattan, 1987 through 1993. AJPH, 1994, 84(8): 1250-54. 52. Goodwin RD. Anxiety disorders and the onset of depression among adults in the community. Psychol Med. 2002;32(6):1121-4. 53. Grant BF, Dawson DA, Stinson FS, Chou SP, Dufour MC, Pickering RP. The 12month prevalence and trends in DSM-IV alcohol abuse and dependence: United States, 1991-1992 and 2001-2002. Drug Alcohol Depend, 2004, 74: 223-34. 54. Griffiths RR, Richards WA, McCann U, Jesse R. Psilocybin can occasion mystical-type experiences having substantial and sustained personal meaning and spiritual significance. Psychopharmacology 2006; 187(3): 268-83. 55. Haasen C, Prizleve M, Zurhold H et al. Cocaine use in Europe. A multicenter study. Eur Addict Res 2004,10(4): 147-55. 56. Harrison L, Martin SS, Enev T, Harrington D. Comparing drug testing and selfreport of drug use among youths and young adults in the general population. (Office of Applied Studies, DHHS Publication No. SMA 07-4249, Methodology Series M-7). Rockville, MD: Substance Abuse and Mental Health Services Administration, 2007. 57. Hatsukami DK, Fishman MW. Crack cocaine and cocaine hydrochloride. Are the differences myth or reality? JAMA, 1996; 276(19):1580-8. 58. Helzer JE, Robins LN, McEvoy LT et al. A comparison of clinical and diagnostic interview schedule diagnoses. Physician reexamination of lay-interviewed cases in the general population. Arch Gen Psychiatry, 1985, 42: 657-666. 59. Hemby SE. Cocainomics: new insights into the molecular basis of cocaine addiction. J Neuroimmune Pharmacol, 2010;5(1):70-82. 60. Hu M, Crombag HS, Robinson TE, Becker JB. Biological basis of sex differences in the propensity to self-administer cocaine. Neuropsychopharmacology. 2004;29(1):81-5. 119 61. Ialongo NS, Werthamer L, Kellam SG et al. Proximal impact of two first- grade preventive interventions on the early risk behaviors for later substance abuse, depression and antisocial behavior. Am J Community Psycol, 1999, 27: 599-641. 62. Indriati E, Buikstra JE. Coca chewing in prehistoric coastal Peru: dental evidence. Am J Phys Anthropol 2001, 114 (3): 242-57. 63. Johanson CE, Fishman MW. The pharmacology of cocaine related to its abuse. Pharmacol Rev, 1989, 41(1): 3-52. 64. Johnson AL, Morrow CE, Accornero VH, Xue L, Anthony JC, Bandstra ES. Maternal Cocaine Use: Estimated Effects on Mother-Child Play Interactions in the Preschool Period. Developmental and Behavioral Pediatrics. 2002; 23:4: 191202. 65. Kandel DB, Raveis VH. Cessation of illicit drug use in young adulthood.Arch Gen Psychiatry, 1989, 46(2): 109-116. 66. Katerndahl DA, Realini JP. Relationship between substance abuse and panic attacks. Addict Behav, 1999, 24: 731-736. 67. Kellam SG, Anthony JC. Targeting early antecedents to prevent tobacco smoking: findings from an epidemiologically based randomized field trial. Am J Public Health, 1998, 88: 1490-1495. 68. Kellam SG, Werthamer-Larsson L, Dolan LJ et al. Developmental epidemiologically based preventive trials: baseline modeling of early target behaviors and depressive symptoms. Am J Community Psychol, 1991, 19: 56384. 69. Kendler KS, Prescott CA. Cocaine use, abuse and dependence in a populationbased sample of female twins. Br J Psychiatry. 1998; 173:345-50. 70. Kleber HD. Pharmacologic treatments for heroin and cocaine dependence. Am J Addict. 2003;12 Suppl 2:S5-S18. 71. Kleber HD, Weiss RD, Anton RF Jr. et al. Treatment of patients with substance use disorders. Am J Psychiatry. 2007 Apr;164(4 Suppl):5-123. 72. Klerman GL, Weissman MM. The changing epidemiology of depression. Clin Chem. 1988;34(5):807-12. 73. Louie AK, Lannon RA, Ketter TA. Treatment of cocaine induced panic disorder. Am J Psychiatry, 1989, 146: 40-44. 120 74. Louie AK, Lannon RA, Rutzick EA, Browne D, Lewis TB, Jones R. Clinical features of cocaine induced panic. Biol Psychiatry, 1996, 40: 938-940. 75. Manski C, Pepper J, Petrie C. Informing America's Policy on Illegal Drugs: What We Don't Know Keeps Hurting Us. Washington DC: National Academy Press; 2001. 76. McCrae RR, Costa PT. Validation of the five-factor model of personality across instruments and observers. J Pers Soc Psychol, 1987, 52(1):81-90. 77. Michaud PA, Berchtold A, Jeannin A, Chossis I, Suris JC. Secular trends in legal and illegal substance use among 16 to 20 year old adolescents in Switzerland. Swiss Med Wkly. 2006;136(19-20):318-26. 78. Miech RA, Chilcoat H, Harder VS. The increase in the association of education and cocaine use over the 1980s and 1990s: evidence for a 'historical period' effect. Drug Alcohol Depend. 2005;79(3):311-20. 79. Millan MJ. The neurobiology and control of anxious states. Prog Neurobiol, 2003, 70: 83-244. 80. Morrow CE, Bandstra ES, Anthony JC, Ofir AY, Xue L, Reyes MB. Influence of Prenatal Cocaine Exposure on Early Language Development: Longitudinal Findings from Four Months to Three Years of Age. Developmental and Behavioral Pediatrics. 2003; 24:39-50. 81. Musto DF. America’s First Cocaine Epidemic. Wilson Q, 1989, 13(3): 53-64. 82. Musto DF. International Traffic in Coca through the early 20th century. Drug Alcohol Depend, 1998, 49(2): 145-56. 83. Office of National Drug Control Policy. Predicting Heavy Drug Use (Publication No NCJ 208382). Washington, DC, Executive Office of the President, 2004. 84. Olch PD. William S. Halsted and Local Anesthesia. Contributions and Complications. Anesthesiology, 1975, 42(4): 479-86. 85. O’Brien MS, Anthony JC. Risk of Becoming Epidemiological Estimates for the United Neuropsychopharmacology. 2005, 30(5): 1006-18. Cocaine States, Dependent: 2000-2001. 86. O’Brien MS, Wu LT, Anthony JC. Cocaine use and the ocurrence of panic attacks in the community. A case-crossover approach. Subst Use Misuse, 2005;40(3):285-97. 87. Petersen RC. History of Cocaine. NIDA Res Monogr. 1977, 13:17-34. 121 88. Petronis KR, Anthony JC. Perceived risk of cocaine use and experience with cocaine: do they cluster within US neighborhoods and cities? Drug Alcohol Depend. 2000;57:183-92. 89. Petronis KR; Anthony JC. A different kind of contextual effect: geographical clustering of cocaine incidence in the USA. J Epidemiol Community Health. 2003; 57:893-900. 90. Polling J, Oliveto A, Petry N, et al. Six-month trial of bupropion with contingency management for cocaine dependence in methadone-maintained population. Arch Gen Psychiatry, 2006, 63(2): 219-28. 91. Price WA, Giannini AJ. Phencyclidine and "crack"-precipitated panic disorder. Am J Psychiatry, 1987, 144: 686-687. 92. Renarman C, Levine HG. Crack in the rearview mirror. Deconstructing drug war mythology. Social Justice, 2004;31(1-2):182-99. 93. Rivera MA, Aufderheide AC, Cartmell LW, Torres CM, Langsjoen O. Antiquity of coca-leaf chewing in the south central Andes: a 3000 year archaeological record of coca-leaf chewing from northern Chile. J Psychoactive Drugs, 2005;37(4): 455-8. 94. Rosemberg MF. Pharmacothanatology: an epidemiological investigation of drug related deaths. A dissertation submitted to JHU in conformity with the requirements for the degree of Doctor of Philosophy. Baltimore, MD, 2001. 95. Rounsaville BJ. Treatment of cocaine dependence and depression. Biol Psychiatry. 2004;56(10):803-9. 96. Schifano F, Corkery J. Cocaine/crack cocaine consumption, treatment demand, seizures, related offences, prices, average purity levels and deaths in the UK (1990 - 2004). J Psychopharmacol. 2008;22(1):71-9. 97. Schneck JM. Cocaine addiction and Dr. William S. Halsted. J Clin Psychiatry, 1988, 49(12): 503-4. 98. Shaffer HJ, Eber GB. Temporal progression of cocaine dependence symptoms in the US National Comorbidity Survey. Addiction. 2002; 97:543-54. 99. Silva de Lima MS, de Oliveira Soares BG, Reisser AA, Farrell M. Pharmacological treatment of cocaine dependence: a systematic review. Addiction. 2002;97(8):931-49. 122 100. Singer JD, Willett JB. It’s about time. Using discrete-time survival analysis to study duration and the timing of events. J Educ Stat, 1993;18(2):155-95. 101. StataCorp. Stata Statistical Software: Release 9.0, College Station, Tx, 2005, Stata Corporation. 102. Storr CL, Reboussin BA, Anthony JC. Early Childhood Misbehavior and the Estimated Risk of Becoming Tobacco-dependent. Am J Epidemiol 2004, 160: 126130. 103. Tatum AL, Seevers MH. Experimental cocaine addiction. J Pharmacol Exp Ther, 1929, 36: 401-10. 104. Terracciano A, Lockenhoff CE, Crum RM, Bienvenu J, Costa PT. Five-factor model personality profiles of drug users. BMC Psychiatry, 2008, 8:22. 105. Tuke DH. A dictionary of psychological medicine. London. J & A Churchil, 1892. 106. United Nations Office on Drugs and Crime. World Drug Report 2010 (United Nations Publication, Sales No E.10.XI.13). UNODC. New York, 2010. 107. United States Substance Abuse and Mental Health Services Administration. Overview of Findings from the 2002 National Survey on Drug Use and Health. (Office of Applied Studies, NHSDA Series H-21, DHHS Publication No. SMA 03– 3774). Rockville, MD, SAMHSA, 2003. 108. United States. Substance Abuse and Mental Health Services Administration. Overview of Findings from the 2003 National Survey on Drug Use and Health. (Office of Applied Studies, NHSDA Series H-24, DHHS Publication No. SMA 04– 3963). Rockville, MD, SAMHSA, 2004. 109. United States. Substance Abuse and Mental Health Services Administration. Overview of Findings from the 2008 National Survey on Drug Use and Health. (Office of Applied Studies, NHSDA Series H-36, DHHS Publication No. SMA 09– 4434). Rockville, MD, SAMHSA, 2009. 110. United States. Substance Abuse and Mental Health Services Administration. Overview of Findings from the 2009 National Survey on Drug Use and Health. (Office of Applied Studies, NHSDA Series H-38-A, DHHS Publication No. SMA 10–4856). Rockville, MD, SAMHSA, 2010. 111. Van Etten ML, Anthony JC. Comparative epidemiology of initial drug opportunities and transitions to first use: marijuana, cocaine, hallucinogens and heroin. Drug Alcohol Depend. 1999; 54:117-25. 123 112. Verebey K, Gold MS. From coca leaves to crack: the effects of dose and route of administration in abuse liability. Psychiatr Ann, 1988, 18(9): 513-20. 113. Wagner FA, Anthony JC. Into the World of Illegal Drug Use: Exposure Opportunity and Other Mechanism Linking the Use of Alcohol, Tobacco, Marijuana and Cocaine. Am J Epidemiol. 2002; 155:918-25. 114. Wagner FA, Anthony JC. From first drug use to drug dependence; developmental periods of risk for dependence upon marijuana, cocaine, and alcohol. Neuropsychopharmacology, 2002, 26(4): 479-88. 115. Weissman MM, Klerman GL. Sex differences and the epidemiology of depression. Arch Gen Psychiatry. 1977;34(1):98-111. 116. White HR, Bates ME. Cessation from cocaine use. Addiction, 1995, 90(7): 94757. 117. Wilcox HC, Anthony JC. The development of suicide ideation and attempts: an epidemiologic study of first graders followed into young adulthood. Drug Alcohol Depend. 2004, 7;76 Suppl:S53-67. 118. Williams MJ, Adinoff B. The role of acetylcholine in cocaine addiction. Neuropsychopharmacology. 2008 Jul;33(8):1779-97. 119. Wright SN, Wang SY, Kallen RG, Wang GK. Differences in steady-state inactivation between Na channel isoforms affect local anesthetic binding affinity. Biophys J, 1997, 73(2): 779-88. 120. World Health Organization (WHO) World Mental Health Consortium. Prevalence, severity, and unmet need for treatment of mental disorders in the World Health Organization World Mental Health Surveys. JAMA, 2004, 291: 2581-2590. 121. World Health Organization (WHO). International Statistical Classification of Diseases and Health Related Problems (ICD-10). 10th Revision. Second Edition, Geneva, WHO, 2004. 124