GSMS Research Summary

To view the last GSMS Progress Report, Click Here.


The increase in rates of drug use and substance use disorders (SUD) across the teens and early 20’s has been extensively studied at the level of epidemiology, theory, and intervention. In contrast, the decrease in drug use and SUD that begins in the later 20’s (Figure A1) has received little scientific attention. Yet there are both scientific and public health reasons to pay close attention to this phenomenon. Scientifically, recovery from SUD is hard to explain within a simplistic model of gene-driven brain structure or function, or within current risk models. A more multifaceted, developmental understanding is needed. The public health implications arise from the possibility of developing ways to support natural recovery. Barriers to progress include the following: (1) Most research on natural (untreated) recovery uses samples of convenience; (2) representative population studies are largely cross-sectional, and so (a) rely on retrospective recall and (b) lack the ability to address mechanisms.

The goals of the present application are (1) to distinguish stable recovery (recovery at age 26 and 30) from unstable patterns of SUD (e.g., relapse at age 30, new onsets at 26 or 30); and (2) to identify predictors and potential mechanisms that lead to stable recovery. The data will come from the Great Smoky Mountains Study (GSMS), a longitudinal, population-based study of 1,420 youth (70% Anglo, 25% American Indian, 5% African American, 50% female) recruited in 1993 at age 9-13 and assessed on average 8 times since then. Like the cross-sectional studies in Figure A1, the longitudinal GSMS data also shows SUD peaking in the early 20s and declining at the last assessment at age 26 (Figure B1). Data collection will include (as at every preceding wave) a full interview-based assessment of 18 DSM-IV substance use disorders and most psychiatric disorders, treatments and medications, level of functioning, health, family structure and functioning, education, employment, income, recent life events and quality of life. In addition we shall complete on all participants the

CANTAB neurocognitive assessment battery that we are currently using with a subgroup of 160 who are participating in a fMRI study of adolescent-limited and persistent SUD.

We will address these hypotheses:

H1. Stable recovery at age 30 will be predicted by fewer comorbid psychiatric disorders, a higher level of functioning, better health and fewer traumatic life events by age 26, compared with relapse or persistence.

H2. High levels of stress biomarkers (EBV, DHEAS, CRP, cortisol reactivity) in response to stressors at age 26 will predict relapse at age 30.

H3. Stable recovery vs. relapse will be associated with poor delay of gratification and high sensation seeking.

H4. Despite higher levels of SUD in Indians and males, neurocognitive mechanisms of risk will not differ.


This study builds on the Great Smoky Mountains Study (GSMS), a longitudinal epidemiologic study now in its thirteenth year of examining the development of psychiatric and substance abuse disorders. A key question in the previous funding period for this study was whether early-onset drug–users were all at risk of later problem drug use or only those with comorbid psychiatric problems. We showed that while having a conduct disorder (CD) quadrupled risk of SUD in adolescents, 80% of adolescents with SUD did not have a history of CD; and the effects of CD diminished after age 14.  A range of other risk factors showing a reduction in drug use and abuse after early 20s are used as the basis for our hypotheses about SUD in early adulthood.



This is a continuation of a study of normal and abnormal development that began in 1993 with three cohorts of girls and boys aged 9, 11, and 13 selected from a population of 20,000 children.  These participants have been re-interviewed every year for the last 18 years.  The study participants are ages 26 to 30. This year we will start re-interviewing participants in all cohorts at age 30. This interviewing will continue through 03/31/2016  We anticipate a total of 1130 participants.


The Young Adult Psychiatric Assessment (YAPA) and other measures will be administered in the home or our office located in Sylva, NC, based on participant preference. This interview will take approximately 2 hours to complete

The YAPA is an interview that elicits information contributing to a wide range of substance use and psychiatric symptoms and disorders, and a range of risk factors, with particular attention to psychiatric comorbidity.

After the YAPA interview, we will schedule another time to meet with participant at our office in Sylva, NC to administer the following well-known and validated instruments to measure the following: physical activity, personality, interpersonal relationships and IQ. We estimate it will take approximately 2 hours to complete all questionnaires.

  1. International Physical Activity Questionnaire- measures level of physical activity (which may interact with the stress markers)
  2. Multidimensional Personality Questionnaire and the Adult Personality Functional Assessment- assess personality (which may interact with measures of psychopathology and stress)
  3. Conflict Tactic Scale- measure interpersonal relations with spouse or partner (not measured before because subjects were too young).
  4. Intellectual capacity (IQ) (which may be one of the protective factors predicting desistence).
  5. The Life Trajectory Interview- identifies how individuals position themselves with respect to models of economic and social success to test how this relates to individuals mental health trajectories across the lifespan as well as characterizing participants’ understanding of how individual behavior and extrinsic event may act to “derail” life course goals.
  6. Eating Disorder Examination Questionnaire (EDE-Q)- It is important to distinguish food addiction from  other eating disorders. This is a self-report measure for the diagnosis of anorexia nervosa, bulimia nervosa, and binge eating disorder.
  7.  Addiction Severity Index (McLellan et al., 2006….) In our longitudinal study we have until now used measures developed primarily for children. We now need to include measures designed specifically for adults. The ASI will be used as a bridge between the youth-upward YAPA and standard adult measures of addiction. The ASI was developed more than 30 years ago to be administered by a technician or clinician, and has a strong emphasis on functioning and the life events that accompany addiction. Like the YAPA, it provides information on the frequency, duration, and intensity of alcohol and drug use, and psychosocial functioning. It is widely used clinically and has adequate reliability.(Makela, 2004)

Continued Data Collection

We will continue, as we have since the beginning of the study, to perform finger pricks. We collect 5 drops of blood twice, at a 20-minute interval, to study cortical reactivity and other potential stress markers (DHEAS, C-reactive protein, Epstein-Barr virus antibodies) In addition to studying hormones, we will perform DNA analysis on the stored and future blood samples to study gene-environment interactions in the development of psychiatric disorders.


There are no expenses associated with this study. Costs to participants are their time and effort. The risks of the study are those associated with discussing personal issues with a trained interviewer: possible embarrassment and distress. Based on preliminary results, the benefits of the study are the chance to talk through problems with a sympathetic interviewer and to contribute to knowledge about human development.


Participants have already been recruited and previously interviewed, six times each, on average.  Each participant will be paid $45 to complete the psychiatric interview and another $45 to complete additional measures.


Only participants who have previously given consent and have been interviewed will be consented for this interview.


Study participants will incur no costs at as result of participating in this study.


Data entry programs have been created.  Over the past 18 years we have built an extremely efficient data management team, capable of generating data for analysis very quickly and accurately


We use newly-developed technology to turn the Young Adult Psychiatric Assessment, into a computer-based interview, using tablet PCs. This speeds the administration of the interview, downloads the data directly onto the study server, thus eliminating data entry errors and the risk of lost interview paperwork, and increases data security.   Participant information on the computer files is identified only by a study-specific identification number.  This study is being carried out under a Certificate of Confidentiality given by NIH in 2001.

Amendment for Methylation Study: 2/24/2014

We intend to examine how adverse experiences are biologically embedded and may persist over time we will use PREVIOUSLY-COLLECTED bloodspots from the Great Smoky Mountains Study from children who experienced multiple DSM IV “extreme stressors” and that were collected at 3 time points: prior to the first report of an extreme stressor event, at the time the last event was reported in childhood, and prior to any event in adulthood. To identify event-induced changes, we will compare methylomes before any event and after the last events in childhood (9-16). To identify marks that persist over time, we will compare the adult methylome (> 19) assayed prior to any event in adulthood with the methylome assayed at the last event in childhood. Pre-existing differences between adversity exposed and non-exposed groups (e.g. caused genetic or environmental factors such as living in urban environment or access to health care) associated with events could lead to false positive findings. Rather than simply comparing exposed and non-exposed groups, we will consider within-subject changes where comparisons involve methylation data from the same subject before and after events.

No recruitment of subjects will take place under the Developmental Methylomics Grant. All the data have been collected under proposals approved by the IRB of Duke University Medical Center. The proposal has also been approved by Emory University, and by the Tribal Council of the Eastern Band of Cherokee Indians. Only the previously collected bloodspots of those subjects who have consented to genetic testing will be used.

Specific aims are:

1 Characterize the biological embedding of adverse childhood experiences
a Select 600 GSMS blood DNA samples from 250 children exposed to multiple DSM-IV extreme stressors. For each child we will select samples from 3 time points: prior to the first report of an event, at the time that the last event was reported in childhood (9-16), and prior to the report of any event in adulthood (> 19).
b Shear DNA into ~150 bp fragments through ultrasonication and use a methyl-CpG binding domain protein to capture the methylated fraction. Sequence 60 million fragments for each sample at 50bp read length using our SOLiD 5500xl W. Use an existing pipeline to align and QC the reads, eliminate sites showing alignment problems, obtain methylation measures for all >28 million CpGs, and perform data reduction by collapsing correlated sites using a 2-stage adaptive procedure.
c Compare methylomes before and after the last event to identify adversity induced changes. Compare adult versus last-event methylomes to identify marks that persist over time. All comparisons involve within-subject changes to improve causal inferences. Use bioinformatics and pathways analyses to interpret the findings.
2 Identify processes that are related to short and long term health risks
a Select an additional 800 DNA samples from 2 time points in 2 groups of 200 children 1) exposed to family maltreatment (physical, sexual abuse, neglect) and 2) controls never exposed to an extreme stressor. Age match to the last event and adulthood assessments in the extreme stressor group, for each subject. Perform the whole methylome assays (see 1b). For the extreme stressor group use the already generated data in 1b.
b Extract health risk data including emotional distress, social functioning, physical health, risky/illegal behaviors, socio-economic status, and stress biomarkers. To examine how methylation mediates the effects of events on health risks, for short term outcomes (time of the last event) associate sites that changed as a result of the events and for long term effects (adulthood) consider sites that showed persistent changes.
c Parse the 175 most meaningful associations from 2b using bioinformatics and after integration expression plus methylation data from human post-mortem brain studies. Replicate these findings using pyrosequencing of bisulfite converted DNA in 392 independent GSMS subjects using a parallel three group design.
3 SNP analyses and the development of prediction algorithms
a Combine the replication data (2c) with already available GWAS data to explore the role of sequence variation.
b Use all available clinical, GWAS, and methylation data, to start developing algorithms that predict the long term response to adversities and that may eventually be used to improve prevention, prognosis, and treatment.

Successful completion of this project implies that we gained insight into how childhood adversities alters the methylome and what changes persist over time. We will also have identified processes associated with health risks in childhood/adulthood and found replicable methylation biomarkers associated with these risks. Methylation markers are stable and can be measured cost-effectively in blood, which is relatively easy to collect. Our findings therefore also have considerable translational potential as, for example, diagnostic “biomarkers of health risk” that could guide intervention strategies.

Amendment to study:

As part of the dissemination and outreach plan described in the grant, we propose to write a book on the study, with a focus on adult outcomes of early experiences. As part of the book, we would like to re-interview a small group of participants (about 20 subjects) to check whether our conclusions, based on the structured interviews we have conducted, are consistent with their own recollections of past events as recorded in our data. For this data-validation exercise we propose to use the “Life History Calendar” developed by Avshalom Caspi and colleagues for their longitudinal study in New Zealand, and now widely used around the world. The LHC is completed by an interviewer at a one-on-one meeting with the study participant.

Study interviewers will be two of the interview team who have worked on the study for the past 20 years. They are extremely familiar with the participants and have very extensive experience of psychosocial interviewing. They are accustomed to visiting participants in their homes.

The 20 participants will be selected to represent both sexes and race/ethnic groups, and the full range of socioeconomic groups, based on and analysis of the past 20 years of data.  Interviews, which will take approximately 2 hours, will be recorded and transcribed. Participants will be paid $100 for a completed interview.  Participants are very familiar with the process of being interviewed, and we anticipate that they will find this data validation interview similar to those that they have completed over the past 20 years.

We have followed these participants for more than 20 years using the same interview (or nearly the same interview). They are aware of the length of interview and we consistently have 85% or better compliance rate year after year. We will call them on the phone and those that we do not have a current phone number, we will go out to their homes. They have agreed to be contacted in the consent that they signed in their last follow-up. Additionally, they have provided multiple contact sources such as email addresses and relatives.

As part of the study, we would like to take a photograph of the subject.  Some portion of a participant’s “Life History” may appear in the book. We will not use names to tell any part of anybody’s “Life History”. We may use the subjects photograph as part of their Life History. We will ask the subject for their approval or refusal for any parts of their Life History that we would like to use in the book. If we want to use their picture, we will ask their permission before doing so.

We will not collect blood spots or take height/weight measurements as we have done on all previous visits.

Amendment to study:

Early substance use (E-SU) is known to increase the risk of abuse, dependence, addiction, polydrug use, and other problematic outcomes in adulthood. Therefore, the U.S. Surgeon General’s office has issued a call-to-action to stop underage use, especially in adolescents less than 15 years of age. This goal is especially compelling for American Indians, for some of whom lifetime burden from substance use is particularly high. In order to prevent E-SU effectively, however, we need firm evidence regarding the childhood environment that increases (or decreases) children’s risk for E-SU. Yet very little is presently known about early environmental risks for E-SU specifically and in diverse samples, which constitutes a major gap in knowledge with substantial implications for prevention. The proposed research fills in this gap by applying a developmental life course model to longitudinal data spanning childhood to age 30. Specifically, we propose to examine developmental features of environmental risk, including distinctions of timing, duration, and co-occurrence with other risks; the earliest ages of onset; protective factors that could neutralize risk; and vulnerability factors that could exacerbate the contributions of the risk environment to E-SU. Many studies are also not informative about sex differences in associations between risk and E-SU or co-occurring adolescent psychological, social, and biological transitions such as puberty. The proposed program of secondary data analyses tests a life-course developmental model of E-SU that examines whether youth with E-SU are characterized by distinct profiles of risk and protective factors compared to their peers. Because E-SU occurs during periods of major psychosocial and brain developments, our life-course model also tests whether E-SU itself contributes to poor adult outcomes, even when holding constant prior environmental risk and child behaviors. Data come from a large ongoing, community-representative, prospectively longitudinal, study, the Great Smoky Mountains Study (GSMS). The GSMS has assessed a wide array of risk factors, use of many different substances, additional psychiatric symptoms and disorders, and the developmental transitions from childhood to young adulthood on up to eleven occasions spanning ages 9 to 30, providing an unparalleled opportunity to study pathways to and from E-SU. The dataset also includes a sizable American Indian subsample, allowing us to test our life-course model of E-SU in this understudied population. The proposed analyses offer an unprecedented opportunity to generate new insights into risk, resilience, and outcomes for E-SU in rural American Indian and Anglo females and males. Results from the proposed research will have implications for both prevention and intervention research by identifying who is most at risk for early substance use, when, and why. Results will also have implications for basic research investigating associations between E-SU and long-term outcomes, because such associations can only be understood when the early risk environment of E-SU is well-characterized and taken into account.

Amendment to study: DoD Grant

After decades of improvement, premature mortality is uniquely on the rise in the U.S. among White non-Hispanic adults with low education. Suicide, drug poisoning (particularly from opiates), and alcoholic liver disease appear to be the culprits and have been coined “deaths of despair.” Suicidal thoughts and behaviors, illicit drug use, and alcohol problems (or “diseases of despair,” DoD)—the conditions that likely precede these deaths—are the focus of this application, as are the pathways to these DoD. Despite many years of research and in the face of rising suicides and a nationwide opiates public health emergency, we lack accurate and appreciable predictions of who will succumb to DoD and who will be shielded from them. Speedy new insights on the development of DoD are needed to inform efforts to reverse the rising tide of DoD. These can only be generated with decades’ worth of prospective-longitudinal data with rich coverage of multiple levels of risk and protective factors—from community contexts to molecular mechanisms—with clinically-relevant characterizations of DoD. Realistically, no single extant dataset can fulfill these requirements. Secondary data analysis of multiple long-term longitudinal studies of recent cohorts, with recent young adult assessments, can provide an unprecedented opportunity here. We capitalize on three complementary, long-standing, prospective longitudinal data sets spanning childhood, adolescence, and young adulthood, with recent assessments in young adulthood. (1) The nationally representative National Longitudinal Study of Adolescent to Adult Health (Add Health) allows for fine-grained socio-structural characterizations of individuals and communities affected by DoD. (2) The community-representative psychiatric-diagnostic Great Smoky Mountains Study (GSMS) was collected in mostly impoverished rural communities in Appalachia—one of the epicenters of the DoD epidemic. It features a quasi-experiment that allows testing for whether cash transfers received by a subgroup of participants for over 15 years are protective against DoD. (3) Fast Track is a comprehensive 10 year randomized clinical intervention trial with 15 years of follow-up data that targeted mechanisms that are key in recent models of pathways to DoD.

Aim 1 is to characterize the developmental epidemiology of DoD across the early lifespan in the nationally-representative Add Health and in the rural-Appalachian economically-challenged GSMS with a large subsample of American Indians. We propose a comprehensive, longitudinal analysis of DoDs across the first 3 decades of life.

Aim 2 is to conduct hypothesis-based testing and also to use discovery-focused machine learning techniques to examine developmental pathways from childhood/adolescent strain to adult DoD in Add Health and GSMS. Both datasets feature rich longitudinal characterizations of participants’ educational, economic, work, and social trajectories from childhood to young adulthood.

Aim 3 assesses the impact of childhood interventions in GSMS and Fast Track on pathways to DoD. Two of the datasets featured human capital interventions.