A total of 225 children between the ages of 8 and 18 years who were either at high or low risk for developing AD have been evaluated at approximately annual intervals (mean = 5.6 assessments) as part of our Longitudinal Cohort Initiative from our family study of relatives of alcoholics (Cognitive and Personality Factors Family Study). Of the 225 subjects tested with repeated assessments during childhood, 133 are being followed during young adulthood (ages 19 to 29 years); the others are maturing into this initiative. Presence or absence of a substance use disorder during young adulthood was determined for these 133 individuals based on repeated evaluations during the young adult period (an average of 2.3 biannual visits).
Demographic characteristics of the 225 youngsters from the Longitudinal Cohort Initiative and the 133 youngsters for whom young adult assessments were completed can be seen in . A subset of individuals (n = 82) had one or more follow-up visits with a full set of predictor variables collected between the ages of 8 and 13 years; another (n = 127) had one or more visits with predictor variables collected between 14 and 18 years. Some (n = 76) had visits in both developmental periods.
Demographic Characteristics of Samples Used in the Analyses
High-Risk Multiplex Families
The high-risk families were selected through a pair of alcohol-dependent brothers, an ascertainment scheme that results in multiplex AD families. Families were excluded if recurrent major depression, bipolar disorder, or schizophrenia disorders were present in the proband pair or their first-degree relatives. Additionally, AD must have been diagnosed as occurring at least 1 year before other drug dependence. All proband pairs and their living first-degree relatives (parents and siblings) were interviewed using the Diagnostic Interview Schedule (DIS) (41
). Using the DIS, diagnoses of AD and alcohol abuse by DSM-III and DSM-III-R criteria (42
) were made. In addition, presence or absence of alcoholism by Feighner Criteria (44
) was determined. Using the DIS information, a second clinician’s information, and family history reports of all other participating relatives, a best estimate diagnosis was determined. Because a multiplex sampling design was used, the offspring from the proband generation who are being followed as part of the longitudinal effort have an average of four first- and second-degree relatives with AD.
Inclusion and Exclusion Criteria for Low-Risk Families
Control families were selected on the basis of an adult family member volunteering to participate in the study and the family having the same structural characteristics of the high-risk families (two adult brothers). Multiple family members (proband siblings and their parents) were interviewed in person using the DIS to screen for the absence of alcohol or drug dependence, schizophrenia, recurrent major depressive disorder, and bipolar disorder in all first- and second-degree relatives of the adult index case. Low-risk offspring also had diagnostic data for their mothers and her first-degree relatives allowing for determination that offspring came from bilineal low risk for alcoholism pedigrees.
Longitudinal Childhood Evaluation
Because the study design used all available offspring between the ages of 8 and 18 years from the high- and low-risk pedigrees, children entered the childhood evaluation period at differing ages. At each annual evaluation, the children completed a battery of age-appropriate tests that were administered by trained master’s-level clinicians. This included the Child Manifest Anxiety Scale (45
), the Coopersmith Self-Esteem Inventory (46
), and the Wide Range Achievement Test-Revised (WRAT-R) (47
) or Wide Range Achievement Test-Third Edition (WRAT-III) (48
) math, reading, and spelling standard scores. The youth form of the Life Stressors and Social Resources Inventory (LISRES) (49
) (positive and negative events) was administered to all children who had reached age 13. Additionally, postural sway (Lipscomb and right monopedal stances) and P300 (visual and auditory) were collected at each evaluation and entered into the analysis. At the child’s first visit, the Junior Eysenck Personality Inventory (50
) was administered to provide measures of extraversion and neuroticism.
Assessment of Postural Sway
The children were asked to stand on a movement platform (Kistler-Model 9281 B, Kistler, Winterthur, Switzerland) while the output data of amplifiers at each corner recorded changes in pressure throughout the platform, which was digitized and stored at 18 Hz. A total of six trials (three with eyes open and three with eyes closed and blindfolded) in each of two procedures, a bipedal and a monopedal stance previously described (38
), were collected. In the monopedal stance, the child was asked to keep the right foot raised while he/she stood on their left foot. A 30-sec intertrial interval and a 1-min interval between tasks were provided in which the child was allowed to get off the platform. Scores obtained from the eyes closed condition were modeled in the present analysis.
Event-Related Potential Assessments
To provide exactly the same recording conditions throughout the longitudinal study, a PDP-11/23 was continuously maintained for stimulus presentation. Electrophysiological data were amplified by 20 k using a Grass Model 12 Neurodata system set to a bandpass of .01 Hz to 30 Hz. Each trial was sampled for 1200 msec at 8-msec intervals beginning with the 200-msec prestimulus baseline. For the visual task, the PDP-11/23 was slaved to coincide with an Atari 130 computer. Each child performed an auditory (choice reaction time) and a visual ERP task with electrodes placed at frontal, vertex, parietal, and occipital locations (Fz, Cz, Pz, Oz, P3, P4). All the active electrodes were referred to linked ears with a forehead ground. Eye blinks were tracked online using an oscilloscope and all trials affected by eye artifact (blinks or eye movements greater than 50 μV) were excluded online. Only the artifact-free trials were averaged according to condition offline.
Auditory ERPs were elicited with high (1500 Hz) and low pitched (800 Hz) tones, presented every 3 sec (70 dBA intensity; 40-msec duration) in a modified oddball paradigm as previously described (23
). The visual task consisted of presentation of a brief (33 msec) target (stick figure head with a nose and only one ear oriented with nose upward or nose downward) or a nontarget (circle) stimulus (modeled after Begleiter et al.
21 [ ]). The subject responded to the position of the ear (left or right) with a button press (left or right). Targets occurred on 80 of the 240 trials presented.
Young Adult Evaluation
Offspring who had completed the childhood portion of the study (had reached 19 years of age) were invited to participate in a young adult initiative. Each young adult was administered the Composite International Diagnostic Interview (CIDI) during each biannual evaluation (53
). Substance use disorder in this report is defined as any alcohol abuse/dependence or drug abuse/dependence diagnosed at any evaluation using DSM-IV criteria.
Statistical analyses were planned to investigate the relationship between a set of predictors measured in childhood during two developmental periods (8 to 13 years and 14 to 18 years) and their impact on young adulthood outcome. Outcome was defined by the presence or absence of a SUD diagnosed during any young adult assessment through use of survival curves.
Factor analysis was used to reduce the number of variables to be modeled. Five orthogonal factors were obtained by varimax rotation of the principal factors analysis. Factor scores were then created to represent the 13 neurobehavioral variables collected in childhood. To best represent the cross-sectional structure for each age, covariances were obtained for subjects having the same age and then pooled across the ages. Analyses were performed on the correlations that resulted.
PHREG (SAS version 9.1, SAS Inc., Cary, North Carolina) was used to perform regression analyses of the survival data. PHREG is based on a Cox proportional hazards model and uses the robust sandwich variance estimator to allow for adjusting for multiple siblings (each family contributed an average of 1.7 siblings). Young adult outcome was determined from the repeated CIDI evaluations performed after age 19 (mean age of 23.2 ± 2.7 years at last follow-up visit).
Main effects and all two-way interactions of each factor with risk status (high or low risk for AD) were tested within one model. Outcome was modeled using the 13 predictor variables (five factors) () collected in one of two developmental periods, 8 to 13 years and 14 to 18 years. The earliest record obtained during each period was used.
Results of Factor Analysis for the 13 Childhood Variables
Two variables were of special interest, P300 and postural sway. Developmental changes in P300 and postural sway during childhood and adolescence have been shown to differ by familial risk group status (38
). Separate analyses were planned to identify the critical age(s) that P300 would predict SUD outcome. Analyses were performed for P300 amplitude recorded at ages 9, 11, 14, 17, and 20 years and membership in either the SUD-positive or SUD-negative groups. Postural sway was evaluated at a single age (age 15) to control for age-related changes.