Early Adolescent Executive Functioning, Intrauterine Exposures and Own Drug Use

Abstract

Individual differences in adolescents' executive functioning are often attributed either to intrauterine substance exposure or to adolescents' own substance use, but both predictors typically have not been evaluated simultaneously in the same study. This prospective study evaluated whether intrauterine drug exposures, the adolescents' own substance use, and/or their potential interactions are related to poorer executive functioning after controlling for important contextual variables. Analyses were based on data collected on a sample of 137 predominantly African-American/ African Caribbean adolescents from low-income urban backgrounds who were followed since their term birth. Intrauterine substance exposures (cocaine, marijuana, alcohol, cigarettes) and adolescents' substance use were documented using a combination of biological assays and maternal and adolescent self-report. At 12-14 years of age, examiners masked to intrauterine exposures and current substance use assessed the adolescents using the Delis-Kaplan Executive Function System (D-KEFS), an age-referenced instrument evaluating multiple dimensions of executive functioning (EF).

Results of covariate-controlled analyses in this study suggest that when intrauterine substance exposures and young adolescents' substance use variables were in the same analysis models, subtle differences in specific EF outcomes were identifiable in this non-referred sample. While further study with larger samples is indicated, these findings suggest that 1) research on adolescent substance use and intrauterine exposure research should evaluate both predictors simultaneously; 2) subtle neurocognitive effects associated with specific intrauterine drug exposures can be identified during early adolescence; and 3) intrauterine substance exposure effects may differ from those associated with adolescents' own drug use.

1. Introduction

Considerable concern was generated in the 1990's when popular media predicted that children with intrauterine cocaine exposure (IUCE) would experience severe and unusual developmental and educational impairments (Okie, 2009). However, in the intervening years, well-controlled, multivariate research beyond the neonatal period and into middle childhood has shown that the effects of IUCE are subtle, rather than global or devastating.

The ascertainment of potential specific neurocognitive effects of IUCE is complicated because IUCE consistently co-occurs with other potentially neurotoxic intrauterine exposures as well as in the context of multiple postnatal bio-psychosocial risk factors (Frank et al., 2002; Jacobson et al., 1996) whose effects may be misattributed to IUCE. In some multivariate studies of school-aged children, specific IUCE effects on neurocognitive skills were identified (Bandstra et al., 2001; Savage et al., 2005). These subtle effects have primarily been detected in specific developmental and behavioral domains, such as aspects of language (Bandstra et al., 2004; Beeghly et al., 2006), learning disabilities (Morrow et al., 2006), emotional arousal and regulation (Chaplin et al., 2009; Mayes, 2002) and developing executive cognitive functions (Rose-Jacobs et al., 2009; Warner et al., 2006). In other multivariate studies, variables such as children's age and IQ (Hurt et al., 2009) and intrauterine drug exposures to alcohol, marijuana, tobacco, and other drugs (Hurt et al., 2009; Jacobson and Jacobson, 2001; Sood et al., 2005) played an important role, whether or not IUCE effects were detected. Nearly all studies investigating potential IUCE effects into middle childhood have shown that multiple biopsychosocial variables, especially those associated with poverty, influence neurocognition (Hurt et al., 2009; Mayes, 2002; Sood et al., 2005). Further interpretive difficulties have emerged as the participants in longitudinal studies enter adolescence and may themselves begin using alcohol, marijuana, tobacco, and other drugs, (Fried et al., 2005; Tapert and Brown, 2000) in varying combinations, which may or may not be neurotoxic.

As with intrauterine drug exposures, demographic and bio-psychosocial risk factors (e.g. gender, factors associated with poverty, family history of drug use, mental health status) individually or together may be associated with adolescents' drug use (Fried et al., 2005; Tapert and Brown, 2000). The intergenerational nature of substance abuse is well established, but neuropsychological research has not extensively explored the corollary that adolescent substance users are more likely than non-users to have experienced intrauterine exposures (Frank et al., 2011). With the exception of Fried and colleagues (Smith et al., 2006), few investigators have attempted to ascertain whether neurocognitive effects ascribed to adolescent substance use are in fact predicted by intrauterine exposures or represent the combined effects of intrauterine exposures and adolescents' own drug use.

The present study addresses two questions related to this complex and understudied issue: 1) Whether the impact of IUCE or other intrauterine exposures on neurocognitive function can be detected in adolescence, particularly if the adolescent has initiated substance use; and 2) Whether the impact of adolescents' use of psychoactive substances on concurrent neurocognitive function can be identified if intrauterine exposures are controlled.

1.1 Executive Functions

Although definitions vary, executive functions (EF) generally refer to a set of skills necessary for independent, purposeful, goal-directed activity (Lezak, 1995; Stuss et al., 1986), including many higher-level neurocognitive functions (e.g., working memory, inhibitory control, organization, and planning). EF are critical to successful adaptation because they facilitate goal-directed behaviors and the capacity to manage multiple competing stimuli and/or performance demands (Alvarez and Emory, 2006; Fried, 2001; Noland et al., 2003). Although IQ and EF may share some variance, EF differ from “intelligence” as ascertained by global measures of IQ, which may reflect school experience and well-learned information (Fried, 2002b; Fried, 2001; Jacobson and Jacobson, 2001; Jurado and Rosselli, 2007). Once thought to be exclusively associated with the frontal lobes, increasing evidence has shown that EF require the participation and coordination of multiple anatomical and functional brain regions (Alvarez and Emory, 2006; Hazy et al., 2006). The negative effects of intrauterine exposures on EF may not emerge until middle childhood or adolescence (Rose-Jacobs et al., 2009), given the differential maturation rates of brain regions associated with various aspects of EF and the increasingly complex demands on children's cognitive and behavioral performance as they grow older (Blakemore and Choudhury, 2006; Fried, 2002a).

1.2 IUCE and EF during Middle Childhood and Adolescence

Among preschool and young school-aged children, findings vary as to whether IUCE correlates negatively with cognitive functioning (Bandstra et al., 2002; Behnke et al., 2006; Delaney-Black et al., 2000; Frank et al., 2001; Frank et al., 2005; Noland et al., 2005; Singer et al., 2004). However, during pre-adolescence and early adolescence, we (Rose-Jacobs et al., 2009) and others (Mayes et al., 2005; Salvage et al., 2005; Warner et al., 2006) have identified subtle negative associations of IUCE with EF measured by neuroimaging and/or functional assessments. Warner et al. (2006) evaluated 10-year-old children with and without IUCE using the Stroop, a measure of verbal inhibition (Stroop, 1935) and the Trail Making test (Moses, 2004), a measure of planning and set shifting, and evaluated whether IUCE was associated with their performance on these tests or to variations in neuroanatomical structure using MRI-based diffusion tensor imaging (DTI). Results showed that children with IUCE required significantly more time to complete a visual-motor set-shifting task, and exhibited poorer performance on the verbal inhibition task. IUCE also was associated with significantly higher average diffusion in the left frontal callosal and right frontal projection fibers. Test performance was correlated with fractional anisotropy of the frontal white matter. In further analyses controlling for gender and intelligence, Warner et al. found that intrauterine exposures to alcohol, marijuana, and the interaction between cocaine and marijuana were also associated with average diffusion in the left frontal callosal fibers.

In other research, Savage et al. (2005) assessed 10-year-old low-income children using the Gordon Diagnostic System, a visual continuous performance test measuring impulsivity and sustained attention, as well as Trail Making and Auditory Attention subtests of the Halstead-Reitan Neuropsychological Battery. While the performance of children in both IUCE and unexposed groups was below published norms, on average, children in the IUCE group made more errors of commission on the most difficult Gordon tasks, an indicator of mildly compromised attention and increased impulsivity. In a study of slightly younger children (7-9 year-olds) from another cohort, children with IUCE took longer to process stimuli than children without IUCE, and their effort was more likely associated with activity in more diverse cortical brain areas (Mayes et al., 2005). In our recent study of 9.5 and 11 year-old children with and without IUCE (Rose-Jacobs et al., 2009), heavier IUCE compared to lighter/no exposure was associated in covariate-controlled analyses with mild compromise on children's ability to inhibit prepotent verbal responses on the Stroop (Stroop, 1935), but not scores on the Rey-Osterrieth Complex Figure task (Rey and Osterrieth, 1993).

1.3 Intrauterine Drug Exposures (IUDE), Marijuana, Alcohol, Tobacco and EF

1.3.1 Marijuana

Several researchers following infants prospectively from birth have evaluated the longitudinal effects of intrauterine marijuana exposure on EF during middle childhood, adolescence, and young adulthood (Fried, 2001; Fried et al., 1998; Richardson et al., 2002; Smith et al., 2006) controlling for other prenatal substance exposures and contextual bio-psychosocial variables. However, specific findings vary. Richardson et al (2002) followed a cohort of low-income children whose mothers were recruited during pregnancy. At the 10-year follow-up visit, marijuana exposure during the second trimester of pregnancy was associated with lower test scores in learning and memory on the Wide Range Assessment of Memory and Learning (WRAML)(Adam and Sheslow, 1990), as well as with increased impulsivity (i.e., more commission errors on a continuous performance test). Among a predominantly middle-class European-Canadian cohort of children followed to young adulthood, Fried et al. (1998; 2001) found a somewhat different mix of neurocognitive deficits associated with intrauterine marijuana exposure. At 9 to 10 years of age, intrauterine marijuana exposure was negatively associated with performance on tasks requiring visual analysis, hypothesis testing, and attention, but not global IQ. Further analysis of visual-perceptual abilities in the same cohort between 9 and 12 years-old and again at 13 to 16 years indicated that intrauterine marijuana exposure was associated with greater difficulty in problem-solving, analysis, and synthesis associated with visual perception (Fried and Watkinson, 2000; Fried et al., 2003). In a later fMRI study of 31 of their participants between 18 and 22 years of age, Fried and colleagues reported that intrauterine marijuana exposure was related to altered neural activity during response inhibition (Smith et al., 2004) and visual-spatial working memory processing (Smith et al., 2006), after controlling for other prenatal exposures and the participants' own substance use.

1.3.2 Alcohol

Impairments in EF due to intrauterine exposure to high levels of alcohol are well documented through adulthood, particularly among individuals diagnosed with Fetal Alcohol Syndrome (FAS) (Chiriboga, 2003), now identified as the most severe manifestation of Fetal Alcohol Spectrum Disorder (FASD) (Riley and McGee, 2005). Commonly reported EF deficits associated with heavy intrauterine alcohol exposure include difficulties in planning, cognitive flexibility, selective inhibition, concept formation, reasoning, attention and working memory (Green et al., 2009; Mattson et al., 1999), but findings vary across studies (Connor et al., 2000; Cottencin et al., 2009; Green et al., 2009; Mattson et al., 1999; Richardson et al., 2002; Riley and McGee, 2005). This variation may reflect differences in the timing and dosage of exposure, variation in study design (e.g., cross-sectional/retrospective versus prospective), and/or inconsistency in the measurement and control of potential confounding variables, particularly other prenatal exposures (Connor et al., 2000; Cottencin et al., 2009; Green et al., 2009; Mattson et al., 1999; Richardson et al., 2002; Riley and McGee, 2005). Whether a lower level of intrauterine alcohol exposure below that associated with FAS or depressed birth weight exerts toxic effects on neurocognitive functioning in later childhood or adolescence (Riley and McGee, 2005) has not yet been established.

1.3.3 Tobacco Exposure

Several studies report that intrauterine tobacco exposure is associated with a variety of neurocognitive deficits in middle childhood and adolescence, including deficits in verbal and visual memory, impaired verbal learning, inattention, and impulsivity (Cornelius and Day, 2000; Milberger et al., 1996). In an MRI study of a subset of our cohort during middle childhood, intrauterine tobacco exposure was associated with significant reductions in cortical gray matter, total parenchyma volumes, and head circumference after covariate adjustment for demographics, a finding that retained marginal significance after additional covariate adjustment for other intrauterine drug exposures (Rivkin et al., 2008). Fried et al. (2002a) reported that intrauterine tobacco exposure was associated with poorer impulse control, impaired working memory, decrements in some components of visual-perceptual performance, and lower IQ scores during middle childhood and adolescence. In contrast, others have reported no significant association between intrauterine tobacco exposure and school-aged children's neurocognition or academic achievement (Gilman et al., 2008; Herrmann et al., 2008), or on expressive and overall language functioning (Beeghly et al., 2006), after controlling for biopsychosocial variables.

1.4 Adolescents' Substance Use and Neurocognitive Outcomes

Cross-sectional studies of adults have examined the relationship between substance use and neuropsychological functioning and have identified a range of neurocognitive deficits associated with varying levels of the use and abuse of alcohol and tobacco (Durazzo et al., 2007; Glass et al., 2009; Swan and Lessov-Schlagger, 2007), marijuana (McHale and Hunt, 2008; Nestor, 2008; Pope and Yurgelun-Todd, 1996; Ramaekers et al., 2006; Verdejo-Garcia et al., 2006), and cocaine (Browndyke et al. 2004; Hoff et al., 1996; Toomey et al., 2003; Verdejo-Garcia and Perez-Garcia, 2007). However, these results cannot readily be generalized to adolescents since their brains are still maturing and undergoing dynamic physiologic changes and synaptic reorganization (Blakemore and Choudhury, 2006; Conklin et al., 2007; Rutter, 2007). Scientists have speculated that, like early childhood, adolescence and emerging adulthood may be a “sensitive period” when the brain has enhanced vulnerability to environmental influences and toxic substance exposures (Blakemore and Choudhury, 2006; Lubman et al., 2007; Segalowitz and Davies, 2004). Adolescents and young adults who initiate drug use may expose the rapidly maturing brain to potentially neurotoxic substances and may exhibit longer-term impairments than older individuals who initiate drug use in adulthood (Jacobus et al., 2009; Lubman et al., 2007). Thus, adolescents may show different patterns of vulnerability or resilience following substance exposure than adults (Medina et al., 2007; Schweinsburg et al., 2008; Tapert and Schweinsburg, 2005).

1.5 Other Factors that Complicate Measurement of Outcomes in Studies of Intrauterine Substance Exposure and Substance Use

Methodological differences among studies of intrauterine substance exposure and adolescents' substance use may contribute to the inconsistent results reported in this literature. These differences include variations in study design, subject characteristics, demographics, enrollment criteria, and the specific covariates controlled for in the analyses; the unknown purity, route of administration, and timing of illegal drug use; and the specific neurocognitive outcomes tested (Frank et al., 2001). Additional contributing factors may include the chronicity and combinations of drug use, as well as variations across studies in the definition of abstinence from drug use, including the duration of non- use or amount that is classified as abstinence (Jacobus et al., 2009; Medina et al., 2009; Schweinsburg et al., 2008). When evaluating potential substance exposure and substance use effects, it is therefore critical to address (either by subject selection or statistical control) factors that are known to influence neurocognitive outcomes for all individuals (e.g., gender, child IQ, age, and maternal education) (Frank et al., 2001; Messinger et al., 2004; Morrow et al., 2006; Rose-Jacobs et al., 2009; Singer et al., 2008). Moreover, for studies evaluating low-income urban samples, it is imperative also to control for adverse environmental factors (e.g. exposure to violence, environmental exposure to lead) that may also contribute to neurocognitive deficits. Unless these factors are measured and controlled, there is danger that the effect of these factors may be misattributed to the effects of intrauterine substance exposure or adolescents' own substance use (Chiodo et al., 2004; Singer et al., 2008).

1.6 Purpose

Because few investigators have evaluated the effects of both intrauterine substance exposures and adolescents' own substance use on neuropsychological outcomes in the same study, it is possible that variations in children's neuropsychological functioning could be attributed to the effects of one factor while ignoring the effects of the other. Therefore, the purpose of this study is to evaluate whether the level of IUCE, other intrauterine drug exposures, adolescents' own substance use, and/or the interactions between intrauterine drug exposures and own substance use are related to poorer neurocognitive functioning as assessed using the Delis-Kaplan Executive Function System (D-KEFS) at early adolescence, after controlling for important contextual variables.

2. Methods

2.1 Data Sample

As described elsewhere (Beeghly et al., 2007; Frank et al., 2002; Frank et al., 1999; Tronick et al., 1996), the participants were part of a prospective longitudinal study evaluating the effects of level of IUCE on children's growth and development from birth to 14 years of age. All study children were born at Boston City Hospital (now Boston Medical Center) and were from low-income, urban backgrounds. The Human Studies Committees of Boston City Hospital/Boston Medical Center and Boston University School of Medicine approved the study at its inception and yearly thereafter. All birth mothers or other primary caregivers gave written informed consent. Beginning at the 8.5-year visit and continuing through the adolescent years, the children themselves also provided written assent. In addition, a Certificate of Confidentiality was obtained from the federal government to protect participants from having research data subpoenaed. After each study visit, the caregiver and child were given store vouchers and/or age appropriate gifts for completion of the interview and assessment. At the early adolescent protocol (12 to 14 years of age), the caregiver received $100 for completing the caregiver interview and for bringing their adolescent to the developmental assessment. In addition, the adolescent received $50 in store vouchers and two passes to the local cinema.

At the study's onset, infant-caregiver dyads were recruited on a daily basis from the postpartum unit of Boston City Hospital (October 1990 to March 1993) if they met the following inclusion criteria: Infant gestational age ≥ 36 weeks; no obvious major congenital malformations; no requirement for neonatal intensive (NICU) care; no diagnosis of Fetal Alcohol Syndrome in the neonatal record; no indication (either by maternal urine toxic screen, neonatal urine toxic screen, meconium assay, or medical records) of prenatal exposure to illegal opiates, methadone, amphetamines, phencyclidine, barbiturates, or hallucinogens; and no history of HIV seropositivity noted in the infant's or mother's medical record. In addition, mothers had to be at least 18 years old and fluent in English. These criteria excluded subjects with known major risk factors (e.g., premature birth) that might confound any specific effects, if any, of IUCE on child outcomes. English fluency was required because the neuropsychological measures planned for this cohort at older ages were not standardized for non-English speakers. Further details about recruitment procedures and sample characteristics are reported elsewhere (Frank et al., 1999; Tronick et al., 1996).

2.2 Intrauterine Drug Exposure Classification

Research staff interviewed study mothers at intake during the postpartum period about their pregnancy and lifetime use of cocaine, alcohol, marijuana, cigarettes, and other illicit drugs using an adaptation of the fifth edition of the Addiction Severity Index (ASI) (McLellan et al., 1992).

2.2.1 Intrauterine Cocaine Exposure Classification

IUCE was determined using a combination of biological markers and mothers' self-report. At least one biological marker (maternal or infant urine, or infant meconium) was obtained for each recruited dyad to confirm maternal self-reported infant exposure status. Urine samples were analyzed for benzoylecgonine, opiates, amphetamines, benzodiazepines and cannabinoids by radioimmunoassay using commercial kits (Abuscreen RIA, Roche Diagnostics Systems, Inc., Montclair, N.J.). We also sought to collect meconium specimens from all enrolled infants for analysis by radioimmunoassay for the presence of benzoylecgonine, opiates, amphetamines, benzodiazepines and cannabinoids, using a modification of Ostrea's method (Ostrea et al., 1992). Based on composite information derived from maternal self-report and/or the meconium assays, children with IUCE were further classified as having either heavier or lighter IUCE. We a priori defined “heavier” use as the top quartile of days of mother's self-reported use during the entire pregnancy and/or the top quartile of concentration for cocaine metabolites in the infant's meconium. The mean days of maternal self-reported cocaine use during pregnancy in this cohort was 20.6 days (range = 0 – 264); mothers reporting 61 or more days of cocaine use during pregnancy fell into the top quartile and were considered “heavier” users. The mean meconium concentration was 1143 ng of benzoylecognine per gram of meconium (range = 0 to 17,950 ng); infants with more than 3314 ng of benzoylecognine per gram of meconium were in the top quartile and were classified into the “heavier” group. All other IUCE was classified as “lighter”. A classification system based on both self-report and meconium assay was used because 18 of 141 (13%) of the infants in the present sample (14% in the source sample) had no meconium assay. Mothers in the top quartile for self-reported use were classified as heavier users, even if the benzoylecognine level in their infant's meconium was not in the top quartile or was missing. This procedure was used because women are more likely to underreport than over report illicit substance use during pregnancy and because not all infants with IUCE have positive meconium assays (Lester et al., 2002; Ostrea et al., 1989; van Gorp et al., 1999). This ordinal classification scheme is similar to that used by other investigators of prenatal substance exposure (Alessandri et al., 1998; Jacobson et al., 2001; Singer et al., 2004). Prior research in the present cohort indicated that level of cocaine exposure defined this way was significantly related in a dose-related manner to lower birth weight z-scores covariate adjusted for gestational age and gender (Frank et al., 1999), neonatal ultrasound findings (Frank et al., 1998), and less optimal patterns of newborn neurobehavior (Tronick et al., 1996).

2.2.2 Other Intrauterine Drug Exposure Classifications Determined During the Neonatal Period

Identification of prenatal marijuana exposure was based on results of positive urine assay, meconium assay, or maternal self-report. In previous reports, (Frank et al., 2002; Rose-Jacobs et al., 2009) we analyzed marijuana categorically as exposed or unexposed. We used this two-level variable for two reasons. First, ascertainment of intrauterine marijuana exposure by meconium concentration is not entirely valid due to storage of marijuana metabolites in mother's body fat (Ostrea et al., 1989) that do not get transferred to meconium. Second, self-reported use was denied by a third of the marijuana users in this cohort who were identified solely based on urine assay. In more recent reports, however, we utilized a 3-level index of intrauterine marijuana exposure: no marijuana exposure (i.e., no evidence for exposure based on meconium and urine assays and self-report); heavier use [positive urine assay at delivery or self-reported days of use in the top quartile of the sample, > 8 days during pregnancy); or lighter use (no positive urine assay at delivery and self-reported days of use below the top quartile). This classification index was associated (p<.0001) with level of mothers' alcohol, cigarettes, and cocaine use during pregnancy and with infants' mean birth weight [marijuana unexposed, M = 3210 ± 477 grams; lighter M = 3069 ± 464 grams; and heavier M =2943 ± 511 grams; p= 0.02].

At the time the study was initiated, there was no established biologic marker for gestational alcohol exposure, and cotinine assays were prohibitively expensive. Therefore, the ascertainment of alcohol and tobacco use in pregnancy by self-reports was state of the art at the time the current sample was recruited. We determined intrauterine alcohol exposure using mothers' self-reported average daily volume of alcohol in drinks per day during the 30 days prior to delivery. This variable was highly correlated (r = 0.81) with mothers' self-reported use through pregnancy. Consistent with the 3-level IUCE and intrauterine marijuana variables, we categorized prenatal alcohol use in this study as none, lighter (<0.5 drinks/day), and heavier (≥ 0.5 drinks/day). During the post partum interview, mothers also reported the average number of cigarettes per day that they consumed while pregnant. A 3-level intrauterine tobacco exposure variable was created for analytic purposes (none = no cigarettes during pregnancy, heavier > 10 cigarettes/day, and lighter (<10 cigarettes/day).

2.3 Design

Recruited caregiver/child dyads were evaluated in multiple follow-up assessments between children's birth and early adolescence (12-14 years of age) (Frank et al., 2002; Frank et al., 2005; Rose-Jacobs et al., 2009; Tronick et al., 1996). The dependent variables analyzed in this report are children's EF scores from the D-KEFS, which were collected during the early adolescent protocol. At each follow-up visit, primary caregivers brought their child to the Child and Adolescent Development Laboratory in the General Clinical Research Center at Boston Medical Center for a developmental and behavioral assessment administered by trained examiners masked to children's intrauterine exposures, background variables, caregivers' responses on the interviews, and scores on prior developmental assessments. During the child assessment, caregivers were interviewed in a separate room by a trained research interviewer regarding family demographics, their recent substance use, psychosocial adaptation, the caregiving environment, and child's current behavioral functioning.

2.3.1 Classification of Adolescents' Own Substance Use: Cocaine, Marijuana, Alcohol, Cigarettes, and Other Drugs

At the early adolescence follow-up visit, the adolescents reported on their own substance use during a computer-assisted self-interview (ACASI). Questions about cocaine, marijuana, alcohol, tobacco, and other substance use were taken from several validated components of the CDC's 2005 Youth Risk Behavior Surveillance System (YRBSS), the Wisconsin YRBS Middle School Questionnaire, the State and Local YRBS, and the Wisconsin YRBS High School Questionnaire (Eaton et al., 2006). Adolescents also provided a urine sample at the same visit, which was tested by the United States Drug Testing Laboratories, Inc. using the No-Excuse Urine Panel. This method has a limit of detection panel that screens using enzyme multiplied immunoassay technique (EMIT) at the lowest validated concentrations that can be achieved with the reagent set, rather than the higher levels required for judicial actions. The GC/MS confirmations used are either at 1/2 or 1/5 of the SAMHSA screening concentrations depending on the drug class for cannabinoids, opiates, amphetamines and cocaine metabolites and ELISA for cotinine. The No-Excuse Urine gives a detection window for most drugs up to approximately a week, longer for marijuana. If adolescents' self-report for substance use during the past 30 days or their urine assay was positive, adolescents were classified as having used a particular substance(s) during the past 30 days. If they reported substance use prior to the past 30 days, but not more recently, adolescents were classified as having used a particular substance(s) “ever”. To maintain adequate cell sizes for statistical analyses, adolescent cigarette use and cocaine use each were categorized as two-level variables, “never” and “ever”.

2.4 Measures

2.4.1 Executive Function

Following Goldberg (2005) and Anderson (2003), multiple dimensions of EF were evaluated at early adolescent follow-up using an age-referenced single instrument, the Delis-Kaplan Executive Function System (D-KEFS) (Delis et al., 2001). The D-KEFS confronts individuals with multiple tasks that are novel, complex, and require integration of information. To minimize participant burden, we selected five of the nine co-normed subtests comprising the D-KEFS: Color-Word Interference, Design Fluency, Trail Making, Word Context, and Tower. We chose these subtests based on theory, a review of literature, and experience with this cohort. Age-standardized scaled scores (population M = 10 ± 3) for each subtest (usually one measure associated with completion time, and one associated with accuracy/error rates) were used in the statistical analyses. Higher scores reflect more optimal performance. Some scales (e.g., those associated with number of errors) were reversed-scored as part of the standard scoring method.

2.4.1.1 Color-Word Interference

The D-KEFS Color-Word Interference test was adapted from the Stroop Color-Word test (Stroop, 1935) and evaluates an individual's cognitive flexibility and ability to inhibit a prepotent verbal response while attempting to generate the required conflicting response (Delis et al., 2001). In each condition, the individual is confronted with 50 stimuli randomly displayed in five rows of ten stimuli and asked to read them as quickly as possible without making mistakes. In Condition 1, the adolescent is asked to name the color of patches presented in one of three colors (red, blue, and green). In Condition 2, the adolescent is asked to read the color names printed in black ink. In Condition 3 (the subtest most similar to the most complex outcome of the Stroop Color-Word test), the color names are printed in a contradictory color ink and the adolescent is asked to ignore the word in order to name the ink color. In Condition 4, an even more complex interference task than the previous condition, the adolescent is asked to alternate between reading the color words and naming the discordant ink colors. In the present study, we used the age-referenced scaled scores for Completion Times in Condition 3 (Inhibition) and for Inhibition/Switching in Condition 4, and Total Errors for both conditions.

2.4.1.2 Design Fluency

The D-KEFS Design Fluency test assesses non-verbal fluency and cognitive flexibility (Delis et al., 2001). This test consists of three conditions in which the adolescent is asked to connect dots using four straight lines to complete as many different designs as possible in 60 seconds. The dots are arranged in response boxes presented in five rows of seven boxes. In Condition 1, Filled Dots, each box contains five filled dots. In Condition 2, Empty Dots Only, each box contains five filled dots and five unfilled dots, and the adolescent is asked to connect only the unfilled dots. In Condition 3, Switching, the most difficult task, requires the adolescent to alternate between connecting filled and unfilled dots. The scaled scores evaluated in this analysis included the Switching Total Correct score and the Percent Design Accuracy score.

2.4.1.3 Trail Making

The Trail Making Test is made up of several timed conditions and assesses flexibility of thinking on a visual-motor sequencing task (Delis et al., 2001). Condition 1, Visual Scanning, requires the adolescent to locate all occurrences of the number “3” among other numbers and letters randomly scattered across two pages. Condition 2, Number Sequencing, requires the adolescent to connect in serial order, numbers that are scattered across two pages. Condition 3, Letter Sequencing, is similar to Condition 2 but requires connecting letters in alphabetical order. Condition 4, Number-Letter Switching, requires the adolescent to connect alternately numbers and letters in order (e.g. 1-A-2-B-3-C) and serves as the primary measure of executive function for this test. During the test, errors made during the sequencing conditions were pointed out to the adolescents and they were asked to return to the last correct connection and continue from there. The scaled scores used in the analyses were the Number-Letter Switching Completion Time and Error Analysis Score from Condition 4.

2.4.5 Word Context

The D-KEFS Word Context test measures deductive reasoning, ability to integrate multiple pieces of information, hypothesis testing, and flexibility of thinking (Delis et al., 2001). The adolescent is presented with a fictional word and asked to decipher its meaning from sentences containing clues. Up to five sentences are shown for each word; the first provides an ambiguous clue and each subsequent sentence supplies increasingly more information. The goal of the task is to discover the meaning of the fictional word in as few clue sentences as possible and to report the correct response for all remaining sentences of that item. The scaled scores used in the current analyses were the Total Consecutively Correct and the Consistently Correct Ratio score.

2.4.1.5 Tower

The D-KEFS Tower Test is made up of nine test items increasing in difficulty. For each item, two to five disks are placed on one or more of three vertical pegs in a predetermined position and a picture of the tower to be built is presented. The goal is to move the disks across the pegs to build the target tower in as few moves as possible and within a set time limit. Only one disk may be moved at a time and a larger disk may not be placed on a smaller one. The scaled scores used in the current analyses were Total Achievement and the Rule-Violation-Per-Item Ratio score.

2.4.2 Control Variables

Potential candidate control variables including gender and current IQ were selected a priori based on theoretical considerations, previous literature, and earlier findings in this cohort. IQ was measured by the Wechsler Abbreviated Scale of Intelligence (WASI) (Weschsler, 1999). If time did not permit the adolescent to complete the WASI, the child's prorated IQ derived from the Wechsler Intelligence Scale for Children, 3^rd edition (WISC-III) (Weschsler, 1991) administered at the 8.5 or 11 year protocols was substituted. Because substance-using individuals often use multiple substances, multivariate analyses for IUCE and adolescent use were modeled to control for other psychoactive substances (e.g., alcohol, marijuana, tobacco) in the analyses. Other candidate covariates included: adolescents' age at the time of testing, birth weight z-score adjusted for gestational age and gender (National Center for Health Statistics Centers for Disease Control and Prevention, 2000); birth mother's education; birth mother's self-identified African American/ Caribbean ethnicity versus other; current category of caregiver (birth mother, kinship caregiver, or unrelated foster or adoptive parent), a partial proxy for home environment and a composite score reflecting adolescents' maximum exposure to violence between 8.5 years of age and early adolescence, as measured by the Violence Exposure Scale for Children-Revised (VEX-R)(Fox and Leavitt, 1995).

The VEX-R is a 21 item, 4-point Likert self-report scale using cartoon pictures that examines children's exposure to violence both as a witness and as a victim. Scores for violence exposure scores range in intensity from mild (yell, push, and spanking) to severe (threaten with a weapon, shoot, and stab). The standard VEX-R (with cartoons) was administered at the 8.5, 9.5, and 11-year visits. At the early adolescence research visit, we modified the VEX-R to make it more appropriate for adolescent subjects by removing the cartoons and administering only the text of the questions using ACASI. We also removed questions related to spanking, and made the questions more time-specific to determine if the event took place within the past year. The distribution of VEX-R scores was ranked at each time point and then subdivided into four groups of approximately equal size. The maximum of these quartile variables was entered as a covariate in the analyses.

Birth mothers' substance use during pregnancy was correlated with caregivers' postpartum and ongoing use of alcohol, cigarettes (by self-report), marijuana, and cocaine (by self-report or assay). Therefore, only variables representing pregnancy use were included in the analysis model (analyses available from author on request).

Children's maximum blood lead value (mg/dl) obtained as part of clinical care during the preschool years was evaluated as a potential covariate of EF at early adolescence. These values were not part of our original protocol and were abstracted from children's medical records. The blood lead levels recorded in the medical records were tested at the Massachusetts State Laboratory Institute, Department of Public Health (Centers for Disease Control and Prevention, 2001). Because only 106 of the 147 participants in the current analysis had blood lead levels, we evaluated this variable in separate analyses on the available sample.

2.6 Statistical Analyses

A multi-step analysis plan was used to evaluate the study's hypotheses. Descriptive statistics were first generated for each variable of interest, with means, standard deviations, and quantiles (for VEX-R and lead variables only) calculated for continuous variables and counts and percentages for categorical variables. Second, in bivariate analyses, the effect of level of IUCE on D-KEFS scaled scores and the effect of level of adolescents' drug use were tested using one-way analysis-of-variance (ANOVA). Third, multivariate analyses testing the effect of level of IUCE on continuous variables (sample characteristics, potential covariates, and D-KEFS scaled scores) were carried out using one-way analysis-of-variance (ANOVA) followed by Tukey post hoc tests, and on categorical variables using cross tabulations with chi-square tests of significance. For each D-KEFS outcome in separate models, interactions between the 3-level IUCE variable and each drug used during adolescence were tested as well as interactions between these variables and both the child's IQ and gender. In addition to this formal testing, we also performed our regression analyses in a stratified fashion separately for males and females. Spearman's correlations were used to determine whether intrauterine drug exposures were correlated substantially (and potentially collinear) with adolescent's own use of the same substance.

Following these preliminary analyses, the effects of each intrauterine substance exposure and adolescents' own substance use variables on D-KEFS outcomes were then evaluated using multiple linear regressions. Due to the sample size, we evaluated only main effects in all multivariate analyses. Although IUCE has always been the primary exposure of concern, we also always have evaluated the effects of other drug exposures to separate them from the potential specific effects of cocaine exposure. The interim model for all dependent variables included: level of IUCE, gender, IQ, level of intrauterine exposure to marijuana, alcohol, and cigarettes; adolescents' own use of cocaine, marijuana, alcohol, and cigarettes. We found no statistically significant interactions between each prenatal substance use variable and its counterpart among the adolescent use variables. Consistent with methods used in behavioral teratology research, other candidate covariates (described above) were evaluated one-by-one in the interim model for potential retention in the final model based on a 10% change-in-estimates criterion applied to the estimate of effect for any one of the substance exposure variables (Mickey and Greenland, 1989). VEX-R violence exposure did not meet these criteria for inclusion in the final multiple regression models. There were no significant interactions in these analyses.

To evaluate whether preschool blood lead value was a confounding variable, we repeated our final regression models using the sub sample (n = 106) of participants who had preschool blood lead values and compared them to the regression results for the whole sample. We also examined whether or not lead exposure was an independent statistically significant predictor of D-KEFS outcomes.

We used SAS version 9.1.2 for all of our analyses. Results were deemed statistically significant where two tailed p ≤ 0.05.

3. Results

3.1. Sample retention

Analyses were based on data from 137 adolescents who completed at least one D-KEFS subtest at the early adolescent follow-up visit. Potential retention bias was evaluated by comparing characteristics of the 137 who provided data for the D-KEFS at the early adolescent visit to characteristics of the 115 from the original birth cohort who did not provide data for this assessment. Participants and non-participants did not differ significantly on level of IUCE, intrauterine exposures to cigarettes, alcohol, or marijuana; infant birth weight, gestational age, gender; birth mothers' education, age, primiparity at delivery; public/private health insurance payment status; or African-American/Caribbean race/ethnicity (p ≥ 0.05).

Of the 137 participants in the D-KEFS sample, three adolescents did not complete all subtests due to scheduling difficulties and the need to shorten the assessment battery. Two did not complete Trail Making and one did not complete Word Context and Design Fluency.

In separate analyses, we also evaluated sample characteristics of participants with (n=108) and without (n=29) preschool lead values. There were no significant group differences (p<.05) on prenatal exposures to cigarettes, alcohol, or marijuana; infant birth weight, gender, gestational age; or maternal education, age, or primiparity at delivery; public/private health insurance payment status; or African-American/Caribbean race/ethnicity. However, those with a measured blood lead value had a higher intrauterine exposure level of cigarettes (mean = 1.24, s.d. = 1.26) compared to those without a lead value reported (mean = 0.73, s.d. = 1.07, p = .048).

3.2. Sample characteristics

Sample characteristics for the 137 adolescents with D-KEFS data are presented for each IUCE group in Table 1. Level of IUCE was not significantly associated with birth mother's race/ethnicity or education; adolescent's gender, age at the early adolescent assessment, IQ, preschool maximum blood lead values, or adolescents' use of marijuana, alcohol, or cigarettes. Level of IUCE was significantly related to intrauterine exposure to cigarettes, alcohol, marijuana, caregiver type, and birth weight z-score.

Table 1. Birth/Early Adolescent Characteristics by Three-Level Intrauterine Cocaine Exposure (n=137).

Participant characteristics Means (standard deviations) or %	Intrauterine cocaine exposure
	Unexposed (n = 62)	Lighter Exposed (n = 50)	Heavier Exposed (n = 25)	P-value
Birth mother characteristics
Ethnicity: African American/ African Caribbean	58 (94%)	42 (84%)	22 (88%)	0.24*
Maternal education at child's delivery (years) **	11.63 (1.31)	11.64 (1.32)	11.36 (1.19)	0.63†
Average daily cigarettes during pregnancy				<0.0001*
Non-smoker	48 (77%)	15 (30%)	3 (12%)
> 0, < ½ pack per day	5 (8)	17 (34)	11 (44)
≥ ½ pack per day	9 (15)	18 (36)	11 (44)
Average daily alcohol (last 30 days before delivery)				<0.0001*
None	60 (97%)	30 (60%)	7 (28%)
> 0, < 0.5 drinks per day	2 (3)	17 (34)	12 (48)
≥ 0.5 drinks per day	0 (0)	3 (6)	6 (24)
Marijuana use (dosage)				0.002*
Unexposed	56 (90%)	32 (64%)	16 (64%)
Lighter	5 (8)	9 (18)	4 (16)
Heavier	1 (2)	9 (18)	5 (20)
Caregiver category at adolescent protocol				< 0.0001*
Birth mother	56 (90%)	29 (58%)	11 (44%)
Kinship caregiver	5 (8)	15 (30)	10 (40)
Non-kinship caregiver	1 (2)	6 (12)	4 (16)
Adolescent characteristics
Birth weight (g) **	3350.18 (519.41)	3020.40 (307.31)	2854.52 (354.12)	<0.0001†
Gender – male	32 (52%)	25 (50%)	12 (48%)	0.95*
Age (years) **	14.12 (0.69)	14.19 (0.68)	14.36 (0.56)	0.30†
WASI IQ at EA **	92.15 (12.48)	91.58 (13.28)	95.32 (10.94)	0.45†
Maximum lead exposure (n=108)	8.85 (4.62)	8.88 (4.67)	11.60 (8.57)	0.15†
Age at Maximum Preschool Lead (years)	2.17 (1.36)	2.43 (1.37)	2.25 (1.28)	0.68†
Maximum sample quartile of VEX-R up to adolescence				0.02*
1st	0 (0%)	9 (18%)	1 (4%)
2nd	11 (18)	8 (16)	3 (12)
3rd	25 (40)	14 (28)	8 (32)
4th	26 (42)	19 (38)	13 (52)
Alcohol use				0.70
Never	45 (73%)	34 (68%)	15 (60%)
During lifetime but none during past 30 days	12 (19)	2 (24)	6 (24)
Within past 30 days	5 (8)	4 (8)	4 (16)
Marijuana use				0.28*
None	53 (85%)	36 (72%)	19 (76%)
During lifetime but none during past 30 days	3 (5)	8 (16)	2 (8)
Within past 30 days	6 (10)	6 (12)	4 (16)
Cigarette use				0.65*
Never	58 (94%)	46 (92%)	22 (88%)
Ever	4 (6)	4 (8)	3 (12)
Other drug use				<0.0001*
Never	62 (100%)	50 (100%)	21 (84%)
Any past 30 days	0 (0)	0 (0)	4 (16)

^*P-value via chi-square test or Fisher's exact test, where appropriate.

^**Means (S.D.) P-value via proc means.

^†P-value via one-way ANOVA (proc glm).

As we have described elsewhere, (Frank et al., 2011) use of other substances besides those listed above (e.g., inhalants, painkillers) during early adolescence was relatively rare and therefore could not be analyzed in multivariate analyses by those substances. The six adolescents who used substances other than those listed (inhalants, n = 4; painkillers, n = 2) also had used marijuana, alcohol, or cigarettes and were included in those multivariate analyses. No participants acknowledged cocaine use by self-report on the ACASI at this age. However, four adolescents had positive urine levels for cocaine metabolites using the No-Excuse panel. The latter values were all low in magnitude and could be explained by passive exposure. Therefore adolescent cocaine use was not listed in the tables but was considered to be adolescent exposure without making judgment whether the exposure were active or passive and was included as a covariate in all analyses that included adolescent exposures.

Table I describes the adolescent drug use by IUCE. Four of the 75 (5%) adolescents with IUCE versus none of the 62 cocaine-unexposed adolescents used some substance other than marijuana or alcohol; 9 of 33 (27%) intrauterine marijuana-exposed adolescents versus 20 of 104 (19%) intrauterine marijuana-unexposed adolescents used marijuana; and 15 of 40 (38%) adolescents who were intrauterine alcohol-exposed versus 28 of 97 (29%) intrauterine alcohol-unexposed adolescents used alcohol. Of the 29 identified as having used marijuana, 21 were indentified by self-reported alone, two were identified by urine alone, and six were identified by both urine and self-report. Of the two participants identified by urine alone, one had a carboxy-THC >100 ng/mL, above the confirmatory SAMHSA and the United States Drug Testing Laboratories, Inc No-Excuse panel. The other adolescent identified by urine alone had a confirmatory carboxy-THC of 9ng/mL that was below the SAMHSA level of15ng/mL but above the No-Excuse panel level of 2 ng/mL. Ultimately, all adolescent cigarette and alcohol use was by self-report.

3.3 D-KEFS Outcomes by Substance Exposure and Adolescent Substance Use

In Table 2, descriptive statistics are presented for the D-KEFS scaled scores for this sample. Although most scaled scores were within normal limits for age, the sample's average scores for most D-KEFS scaled scores fell below the standardization norm (population mean = 10) on most D-KEFS scaled scores, and often standard deviations were narrower than those reported in the standardization norming tables (population standard deviation = 3).

Table 2. D-KEFS scaled scores for 137 adolescents.

D-KEFS Outcome Variables	Mean	Std Dev
Color-Word Interference
Completion Time: Inhibition	8.84	2.77
Completion Time: Inhibition/Switching	7.96	2.67
Total Errors: Inhibition	8.26	3.22
Total Errors: Inhibition/Switching	7.28	3.28
Design Fluency
Total Correct: Switching	9.20	2.68
Percent Design Accuracy	7.51	3.20
Trail Making Number-Letter Switching
Completion Time	7.16	3.55
Error Analysis	9.59	2.69
Word Context
Total Consecutively Correct	7.76	3.57
Consistently Correct Ratio	8.18	3.94
Tower
Total Achievement	8.91	2.49
Rule-Violations-Per-Item Ratio	9.12	1.79

Tables 3 and 4 present outcomes where at least one outcome correlates with intrauterine exposure or adolescent substance use at p < 0.10. Therefore, results for Color-Word Completion Time: Inhibition/Switching, and Tower Rule-Violations-Per-Item Ratio were not included in the tables. Similarly, except for IUCE, only those intrauterine substance exposures and adolescent substance use variables that were associated (p < 0.10) with at least one of the outcomes are included in Tables 3 and 4. Therefore, intrauterine alcohol exposure < 0.5 drinks/day and adolescent alcohol use during the past 30 days were not included in the tables. All drug comparisons are to the absence of that intrauterine exposure or to non-use in adolescence. For each substance, results for intrauterine substance exposures as predictors of the outcome are presented first, followed by results for adolescents' own use as predictors.

Table 3a.

Unadjusted means, exposure by outcome variable for outcomes identified in Table 2. Sample size is 137 except where noted in the table. Prenatal alcohol exposure level <0.5 drinks/day (n=31) not shown because not associated with any of outcomes. Reference group is always the unexposed group in each category. *p < 0.05

Intrauterine Exposures
	Cocaine Heavier	Cocaine Lighter	Cocaine Unexposed	Marijuana Heavier	Marijuana Lighter	Marijuana Unexposed	Alcohol ≥0.5 drinks /day	Alcohol Unexposed	Cigarettes ≥ ½ pack/day	Cigarettes < ½ pack/day	Cigarettes Unexposed
N	25	50	62	15	18	104	9	97	38	33	66
Color-Word Interference
Completion Time: Inhibition	9.36	8.44	8.95	9.00	8.22	8.92	10.11	8.79	8.97	8.55	8.94
Total Errors: Inhibition	8.88	8.26	8.00	8.33	7.67	8.35	8.89	8.15	8.79	7.63	8.35
Total Errors: Inhibition/Switching	7.44	7.08	7.39	6.80	7.11	7.38	6.67	7.18	7.45	6.84	7.45
Design Fluency
Total Correct: Switching (n=136)	10.13	8.86	9.11	9.33	8.28	9.34	9.33	9.14	9.55	9.14	9.06
% Design Accuracy (n=135)	7.74	7.08	7.77	7.67	6.53	7.65	6.44	7.48	7.73	6.97	7.71
Trail Making Number-Letter Switching
Completion Time (n=135)	7.80	6.60	7.37	7.27	6.44	7.27	4.00*	7.39	7.24	7.13	7.14
Error Analysis (n=135)	10.04	9.76	9.26	10.00	10.06	9.44	8.25	9.60	10.21	9.55	9.28
Word Context
Total Consecutively Correct (n=136)	8.60	7.80	7.39	7.93	8.11	7.68	8.11	7.54	8.48	7.42	7.60
Consistently Correct Ratio (n=136)	8.16	8.40	8.00	8.00	7.44	8.33	8.00	7.99	9.33	7.52	7.97
Tower
Total Achievement	9.36	8.94	8.69	8.47	8.89	8.97	9.89	8.93	8.64	9.00	8.98

Table 3b: Unadjusted means, exposure by outcome variables identified in Table 2. Sample size is 137 except where noted in the table. Alcohol use only during the past 30 days (n= 13) not shown because not associated with any of the outcomes. Reference group is always the unexposed group in each category. *p < 0.05
Adolescent Use
	Marijuana within past 30 days	Marijuana during lifetime but none past 30 days	No Marijuana	Alcohol during lifetime but none past 30 days	No Alcohol	Ever Cigarettes	No Cigarette
N	16	13	108	30	94	11	126
Color-Word Interference
Completion Time: Inhibition	7.81	7.62	9.14	9.03	8.86	9.00	8.23
Total Errors: Inhibition	6.44*	7.31	8.64	8.20	8.46	9.00	8.19
Total Errors: Inhibition/Switching	6.44	7.92	7.33	6.33*	7.70	8.18	7.21
Design Fluency
Total Correct: Switching (n=136)	9.81	8.38	9.21	9.27	9.25	9.09	9.21
% Design Accuracy (n=135)	7.31	6.46	7.67	6.93	7.78	9.00	7.38
Trail Making Number-Letter Switching
Completion Time (n=135)	7.00	7.69	7.12	7.70	7.10	6.82	7.19
Error Analysis (n=135)	10.31	9.54	9.48	9.73	9.56	10.18	9.53
Word Context
Total Consecutively Correct (n=136)	7.50	8.31	7.74	8.03	7.85	9.27	7.63
Consistently Correct Ratio (n=136)	8.88	8.31	8.06	9.30	7.90	8.18	8.12
Tower
Total Achievement	8.19	9.08	8.99	8.83	8.95	7.81	9.00

Table 4a.

Adjusted means, exposure by outcome variable for outcomes identified in Table 2 (Adjusted for IQ; gender; prenatal cocaine, marijuana, alcohol and cigarettes; own use of marijuana, alcohol, cigarettes, and cocaine). Sample size is 137 except where noted in the table. Reference group is always the unexposed group in each category. *p < 0.05

Intrauterine Exposures
	Cocaine Heavier	Cocaine Lighter	Cocaine Unexposed	Marijuana Heavier	Marijuana Lighter	Marijuana Unexposed	Alcohol ≥0.5 drinks /day	Alcohol Unexposed	Cigarettes ≥½ pack/day	Cigarettes < ½ pack/day	Cigarettes Unexposed
N	25	50	62	15	18	104	9	97	38	33	66
Color-Word Interference
Completion Time: Inhibition	8.99	8.53	9.02	9.13	7.99	8.95	10.57	8.76	8.97	9.17	8.60
Total Errors: Inhibition	8.71	8.48	7.89	8.06	7.48	8.42	9.00	8.27	7.95	8.75	8.18
Total Errors: Inhibition/Switching	7.08	7.08	7.53	6.16	7.22	7.46	7.20	7.13	7.23	7.64	7.14
Design Fluency
Total Correct: Switching (n=136)	9.64	8.96	9.22	9.24	7.89*	9.42	9.18	9.28	9.40	9.88	8.75
% Design Accuracy (n=136)	7.03	7.19	7.94	7.42	6.64	7.66	7.05	7.43	7.53	8.07	7.20
Trail Making Number-Letter Switching
Completion Time (n=135)	7.91	6.52	7.36	7.58	6.30	7.24	3.67*	7.58	7.76	7.96	6.40
Error Analysis (n=135)	9.58	9.72	9.42	9.87	9.81	9.47	7.53*	9.86	9.78	10.51*	8.96
Word Context
Total Consecutively Correct (n=136)	7.56	7.87	7.76	7.09	7.74	7.87	8.61	7.68	8.33	7.94	7.38
Consistently Correct Ratio (n=136)	7.12	8.38	8.45	7.96	6.90	8.43	8.95	8.02	7.98	9.50	7.63
Tower
Total Achievement	9.05	9.24	8.57	7.90	8.49	9.12	10.11	8.96	9.28	8.56	8.86

Table 4b: Adjusted means, exposure by outcome variables identified in Table 2 (Adjusted for IQ; gender; prenatal cocaine, marijuana, alcohol and cigarettes; own marijuana, alcohol, cigarettes, and cocaine). Sample size is 137 except where noted in the table. Reference group is always the unexposed group in each category. *p < 0.05
Adolescent Use
	Marijuana within past 30 days	Marijuana during lifetime but none past 30 days	No Marijuana	Alcohol during lifetime but none past 30 days	No Alcohol	Ever Cigarettes	No Cigarette
N	16	13	108	30	94	11	126
Color-Word Interference
Completion Time: Inhibition	7.54*	7.18*	9.23	9.55	8.54	9.36	8.79
Total Errors: Inhibition	6.33*	6.99	8.69	8.46	8.22	9.94	8.11
Total Errors: Inhibition/Switching	6.78	8.35	7.23	5.95*	7.72	8.86	7.15
Design Fluency
Total Correct: Switching (n=136)	10.25	8.08	9.18	9.28	9.22	8.67	9.25
% Design Accuracy (n=136)	7.30	6.42	7.66	6.95	7.66	9.44*	7.33
Trail Making Number-Letter Switching
Completion Time (n=135)	7.36	7.29	7.11	7.64	7.02	6.69	7.20
Error Analysis (n=135)	10.59	8.98	9.48	9.52	9.59	10.00	9.52
Word Context
Total Consecutively Correct (n=136)	7.51	7.72	7.81	7.86	7.73	10.14*	7.56
Consistently Correct Ratio (n=136)	9.15	7.07	8.17	9.24	7.81	8.18	8.18
Tower
Total Achievement	8.31	8.95	8.99	8.88	8.81	8.50	8.94

Tables 3a and 3b present unadjusted means from the ANOVA analyses by intrauterine substance exposure or by adolescent use while Tables 4a and 4b present covariate adjusted means from the multivariable analyses by intrauterine substance exposure or by adolescent use.

3.3.1 Multivariate Analyses: Association of Intrauterine Cocaine Exposure with D-KEFS Scores

There were no significant main effects for the 3-level IUCE variable or interaction effects of level of IUCE and covariates for any D-KEFS outcome (Table 4a). To explore the possibility of a Type-II error based on our analysis sample size of 137, we have computed the smallest effect size detectable for DKEFS scores in contrasts between the lighter IUCE and unexposed groups as well as the heavier IUCE and unexposed groups. For lighter vs. unexposed, we could have detected a standardized effect size (difference in group means/common standard deviation) of 0.54 or greater with 80% power at a two-sided alpha of 0.05. For the Completion Time: Inhibition variable from the DKEFS that had a standard deviation in our sample of 2.77, an effect size of 0.54 translates to a difference in means of 1.50. For heavier vs. unexposed, we could have detected a standardized effect size (difference in group means/common standard deviation) of 0.68 or greater with 80% power at a two-sided alpha of 0.05. For the Completion Time: Inhibition variable, this effect size of 0.68 translates to a difference in means of 1.88. In terms of observed differences for this variable in our sample, we found a difference in mean Completion Time: Inhibition of 0.51 (effect size = 0.18) for lighter vs. unexposed and 0.41 (effect size = 0.15) for heavier vs. unexposed.

3.3.2 Multivariate Analyses: Marijuana

Lighter intrauterine marijuana exposure predicted less optimal Design Fluency Total Correct Switching condition scores compared with no intrauterine marijuana exposure (adjusted mean, lighter exposure = 7.89; adjusted mean unexposed = 9.42 difference=1.53, p < 0.05) (Table 4a).

Adolescents' own marijuana use was associated only with Color-Word Interference outcomes (Table 4b). Specifically, adolescents' marijuana use within the past 30 days predicted significantly less optimal Inhibition Condition Completion time scores (adjusted mean 30 day use = 7.54; adjusted mean no use = 9.23; difference =1.69, p < 0.05), and significantly poorer Total Error Scores (adjusted mean, 30 day use = 6.33; adjusted mean, no use = 8.69; difference = 2.36, p < 0.05]). Any adolescent marijuana use prior to the past 30 days also predicted less optimal Completion Time scores compared with no marijuana use (adjusted mean marijuana use prior to 30 days = 7.18, difference=2.05 p=0.05.

3.3.3 Multivariate Analyses: Alcohol

Intrauterine alcohol exposure ≥0.5 drink/day predicted significantly poorer Trail Making Test Number-Letter Switching Completion Time scores compared with no alcohol exposure (adjusted mean heavier exposed = 3.67; adjusted mean unexposed = 7.58; difference = 3.91, p = 001) as well as significantly poorer Errors scores (adjusted mean heavier exposed = 7.53; adjusted mean unexposed = 9.86; difference = 2.36, p < 0.05] (Table 4a).

Adolescents reporting alcohol use prior to the past 30 days versus those reporting no use had significantly poorer Color-Word Interference Total Errors scores for the Inhibition/Switching condition (adjusted mean alcohol use prior to 30 days = 5.95; adjusted. mean non-drinkers = 7.72; difference = 1.77, p < 0.05) (Table 4b).

3.3.4 Multivariate Analyses: Cigarettes

Lighter intrauterine cigarette exposure was paradoxically associated with significantly more optimal Trail Making Number-Letter Switching Error Analysis scores compared with no intrauterine cigarette exposure [adjusted. mean, lighter = 10.51, adjusted. mean, unexposed = 8.96, difference = 1.55, p < 0.05] (Table 4a).

Similarly, adolescents' own cigarette use was associated with more optimal Design Fluency Percent Design Accuracy Scores compared to non-smokers (adjusted mean smokers = 9.44; adjusted mean non-smokers = 7.33, difference = 2.11, p < 0.05). Cigarette use also was associated with more optimal Word Context Consecutively Correct scores (adjusted mean smokers = 10.14; adjusted mean non-smokers = 7.56, difference = 2.58, p < 0.05) (Table 4b).

3.4 Analyses on Participants with Lead Exposure Values

In the subsample of 106 participants with preschool blood lead values, no evidence for confounding effects of maximum lead level was found. Maximum preschool blood lead values were also not significantly associated with any D-KEFS outcomes (Results of these analyses are available upon request.).

4. Discussion

For most of the substance predictors and neurocognitive outcomes evaluated in our study, adjusted D-KEFS means were lower than (but within one standard deviation of) the population mean of 10. These somewhat depressed performances are consistent with the low-average scores described for individuals from relatively economically disadvantaged backgrounds (Brooks-Gunn and Duncan, 1997). Although the D-KEFS was normed on a United States nationally representative sample, the current sample differs from the standardization sample in that it is relatively homogeneous in terms of race/ethnicity and low socio-economic status.

Despite such relatively limited demographic variability, findings from covariate-controlled analyses indicate that even 12 to 14 years after birth, intrauterine marijuana and/or alcohol exposures, though not IUCE, are negatively associated with performance on specific EF tasks, after adolescent use was controlled. Also, results from covariate analyses indicate that adolescent use of marijuana and/or alcohol are negatively associated with performance on specific EF tasks, after intrauterine substance exposure was controlled. Notably, intrauterine exposures to marijuana and/or alcohol were associated with tasks that could be categorized as visual-motor-perceptual, while adolescent use of marijuana and/or alcohol were generally associated with tasks that had verbal components. Our results highlight the importance of evaluating intrauterine exposures to legal and illegal substances and adolescents' personal substance use in the same analyses even though our sample was recruited on the basis of maternal cocaine use during pregnancy.

In our sample, those with lighter intrauterine exposure to marijuana generated significantly fewer different designs by alternately connecting filled and empty dots within the allotted time of the Design Fluency test. This task requires, among other skills, sustained attention, visual-perceptual planning and working memory, and inhibition, all areas that Fried (2002a; 2003; 2006) identified as problematic following intrauterine marijuana exposure. Like Smith et al. (2004), we did not find that intrauterine marijuana exposure was associated with overall accuracy differences on an inhibition task. While Smith et al. (2004) did identify significantly more commission errors among adolescents with intrauterine marijuana exposure, the D-KEFS Error Analysis score does not separate errors of omission and commission and therefore we are unable to make that comparison in the present study.

Interestingly, although adolescents in our sample were not recruited for their own history of substance use, those who reported any history of personal marijuana use had significantly slower times and more errors on the Color-Word Interference test than those who reported no marijuana use. Other studies with older adolescents who presumably had used marijuana for longer periods of time than was likely among our young adolescents, reported a greater array of negative EF outcomes associated with adolescents' marijuana use, including: increased effort during an inhibition task (Tapert et al., 2007); more compromised spatial working memory strategy and increased errors (Harvey et al., 2007); poorer attention and processing speed (Jacobus et al., 2009); and increased response perseveration (Lane et al., 2007). Continued longitudinal study of our cohort is needed to older ages when it is assumed that there will be more drug use among participants. We also expect that some of the cohort will develop substance use disorders.

Although our sample was not selected for intrauterine alcohol exposure or adolescents' own use, participants with heavier intrauterine exposure to alcohol (but not FAS) took longer to complete the Trail Making Number-Letter test and made a greater number of errors on the task. Mattson et al (1999) tested a large age range of clinically referred children (8 to 15 years) exposed in utero to daily alcohol with intermittent binges. In analyses uncontrolled for other intrauterine exposures, they reported that children with FAS compared to children with lighter or no intrauterine alcohol exposure had significantly poorer scores (not specifically error scores) on a Number-Letter Switching task as well as on the Color-Word, Tower, and Word Context tests. Unlike many other alcohol exposure studies (Connor et al., 2000; Cottencin et al., 2009; Green et al., 2009; Mattson et al., 1999), our study is prospective, statistically controlled for multiple other substance exposures, and has a sample with lighter levels of alcohol exposure than FAS, all factors that may account for our detection of relatively circumscribed adverse effects of prenatal alcohol exposure.

Adolescents who reported lifetime alcohol use made significantly more errors on the most difficult Color-Word Interference test, Inhibition/Switching test (scaled scores greater than a standard deviation below the test mean and 1.77 scaled score points below the study comparison group) than those who did not report alcohol use. Groups did not differ on completion time. These findings are important, as little is known about the neurocognitive effects of alcohol use by young adolescents who are not from a clinically referred sample. Continued longitudinal neurocognitive testing of the current non-referred sample is needed to determine whether the adolescents who later heavily abuse alcohol will develop the significant neurocognitive deficits, such as poor attention, compromised verbal and nonverbal memory, and structural brain abnormalities that have been reported for older adolescents who are heavy drinkers (Brown and Tapert, 2004; Medina et al., 2008).

Intrauterine tobacco exposure was associated with seemingly paradoxical findings. Adolescents with intrauterine tobacco exposure had more optimal neurocognitive scores than those without intrauterine tobacco exposure. While the explanation for these associations are not known, these findings may reflect potential neuroprotective effects of tobacco exposure in the context of other substance exposures or the interaction of genetic and/or environmental factors that some (Gilman et al., 2008; Herrmann et al., 2008), but not others (Beeghly et al., 2007; Cornelius and Day, 2000; Milberger et al., 1996) have identified.

These paradoxical findings were also observed for those young adolescents reporting cigarette smoking. Some investigators have argued that adolescent and adult cigarette smoking could have diametrically distinct effects on neurocognitive performance, depending at least in part on the chronicity of such use (Glass et al., 2006). Our findings are consistent with those in other studies showing that acute nicotine use may enhance attention, vigilance, short-term memory, and cognitive inhibition (Dinn et al., 2004; Ernst et al., 2001; Lawrence et al., 2002; Potter and Newhouse, 2004). In contrast, chronic smoking by individuals has been found to be associated with poorer neurocognitive abilities including general cognitive function, rapid cognitive flexibility and processing, working and verbal memory (Ernst et al., 2001; Glass et al., 2006). The relatively few young adolescents reporting cigarette smoking versus marijuana use in this study is consistent to the population from which our sample was drawn.

Interestingly, IUCE was not significantly related to any EF outcomes in these early adolescent analyses. In our previous covariate-controlled analyses, we reported that between 8.5 and 11 years, children with heavier IUCE were more likely to have significantly lower Stroop scores than the lighter/unexposed group (Rose-Jacobs et al., 2009). The difference in our findings during middle childhood and adolescence may be influenced by the difference in the specific versions of the Color-Word test administered, the different methods of scoring each, or the specific ages tested. Alternatively, the children with heavier IUCE may have developed compensatory strategies in this area of EF by early adolescence.

Our analyses evaluated a non-clinical, early adolescent sample and many of the participants were in the early stages of drug experimentation and/or drug use. Therefore, our findings related to adolescent drug use are conservative as compared to what might be expected in later adolescence, when a greater proportion of our sample may develop more problematic patterns of substance use, which in turn may be associated with more systematic negative effects on EF.

4.1 Study Strengths and Limitations

Our covariate-controlled, longitudinal study has many strengths. First, it is one of the few to evaluate simultaneously the associations of a range of intrauterine drug exposures and adolescents' own drug use to several areas of EF in early adolescence (Fried et al., 2006; 2005). Second, our study evaluated multiple indicators of EF during early adolescence using process-oriented tests from a recently co-normed neuropsychological battery, which allow for clearer comparisons across the different domains of EF performance and across different categories of intrauterine exposures and adolescent substance use (Homack et al., 2005). Third, the youth who participated in this prospective study were recruited during the newborn period rather than during middle childhood or adolescence when their school performance, neurocognitive concerns, or own substance use might have prompted differential enrollment. Fourth, this study evaluated a host of possible covariates known to be associated with intrauterine substance exposures/or neurocognitive functioning including: other intrauterine substance exposures; number and type of caregivers; birth mothers' education and race/ethnicity; adolescents' gender; birth weight; age at assessment; IQ; preschool lead exposure; and exposure to violence between the ages of 8.5 years of age and early adolescence.

Our study also has several limitations. First, because of our recruitment criteria, the present findings may not be generalizable to adolescents who were born preterm or those living in more economically privileged or rural settings. Second, the D-KEFS, a standardized assessment developed for clinical as well as research settings, was administered on a one-to-one basis in a quiet laboratory setting. The assessment does not evaluate adolescents' neuropsychological functioning in an everyday setting such as school classrooms, which confront adolescents with multiple competing challenges and distractions. Third, although the D-KEFS measures multiple domains of executive functioning, the outcomes may not be representative of all aspects of EF central to academic and social success in the real world. Fourth, while from a neuropsychological perspective the process orientation and the multiple co-normed tests of the D-KEFS are study strengths, from a statistical perspective the multiple outcomes within and across tests on the same participants increases the possibilities of Type 1 error. In addition, our high-risk sample is also highly mobile. While we statistically compared our tested and non-tested sample and found few differences, it is always possible that attrition could play a non-obvious role in our findings. Although we did not identify specific associations of IUCE with D-KEFS scores, it is possible that there were Type II errors as a function of our sample size. However, we did have sufficient power to detect associations of other intrauterine exposures and adolescent substance use on EF outcomes. Also, although we controlled for gender in the analyses based on theoretical and preliminary analyses, it is possible that stratified analyses in a larger sample might have identified specific gender differences in the neuropsychological outcomes. In multivariate regression models gender was significantly associated with some DKEFS outcomes (males having higher Design Fluency scores and females having higher Word Context scores). However when the models were stratified by gender the relationships between IUSE or adolescent use were similar for both genders. Similarly, while in multivariate analyses IQ was a significant predictor of many of the DKEFS outcomes, IQ was not differentially distributed by IUSE or adolescent substance use, even in the four participants with IQs between 64 and 70.

Fifth, this study evaluated whether intrauterine exposures and adolescent substance use predicts EF during early adolescence, a developmental period of substance use experimentation. Heavier substance use and abuse may be more prevalent at later ages. Stronger associations (and larger cell sizes for different categories of substance use) may be found in later adolescence. Also, the complexity of controlling for intrauterine exposures and own adolescent substance abuse with our sample size precluded our analyzing possible additive effects of multiple combinations of substances across both time periods on the D-KEFS outcomes. In addition, although environmental lead exposure did not appear to be a confounding variable in this study, these values were not available for all our participants, making it difficult to form firm conclusions in regards to the role of lead exposure and intrauterine substance exposure and/or personal adolescent drug use. Similarly, as in any study, there is always the possibility of the influence of additional non-measured factors on the study outcome. Sixth, although EF was assessed at earlier ages in this cohort, different instruments were used at different ages. The D-KEFS was not available when the participants were younger. This makes it more difficult to evaluate trajectories of EF as a function of intrauterine exposures prior to and following initiation of substance use as measured by a single instrument. Because the current analyses measured EF at a single time point, the direction of causality between EF and adolescent's own substance use may not be clear. For instance, adolescents' own substance use may be a function of less adaptive EF. We also acknowledge that because the present study did not involve a neuroimaging component, any lack of findings on neuropsychological tests does not necessarily rule-out the possibility of differences in brain structure or function.

4.3 Conclusions

Results of these covariate-controlled analyses indicate that both intrauterine substance exposures and young adolescents' own substance use in a prospective non-clinical community sample are associated with different EF test outcomes, when each is controlled for the other. These results, although evaluated using a relatively small sample, suggest the need for evaluating both predictors simultaneously in future analyses. Our findings are notable in that the participants in the study were early adolescents between the ages of 12 and 14 years and heterogeneous in their drug use. Some of these adolescents may have been using drugs temporarily on an experimental basis, while others may be on a trajectory leading to heavier and more chronic substance use with progression to substance use disorders. Findings of this study do suggest that subtle neurocognitive effects associated with some intrauterine drug exposures can be identified as late as 12-14 years postnatal. Because EF related performance has been shown to continue to develop through young adulthood (Adleman et al., 2002), only longitudinal assessment of our sample into later adolescence and adulthood will allow us to evaluate the effect of the combination of maturation, intrauterine drug exposure, and participants' own chronic drug use or abstention, on EF.

Our study has several important public health implications. The effects of intrauterine alcohol and marijuana exposure on some neurocognitive functions even into early adolescence on EF were detected but IUCE effects were not. Therefore, prenatal counseling should emphasize abstinence from legal as well as illegal substances. Moreover, the use of drugs during early adolescence, even if below clinical levels of abuse, can have important negative effects on neurocognitive functioning, although the associations hold only for some substances and some aspects of EF. Notably, these associations differ from those observed for intrauterine exposures. Most of what is known about adolescent drug use and its effects on EF derives from studies of clinical populations and samples identified by heavy use, even though early experimental substance use is more prevalent than substance use disorders during early adolescence (O.o.A. Studies, 2007). Future longitudinal studies with larger sample sizes should continue to evaluate prospectively the neurocognitive effects of intrauterine drug exposures, early substance experimentation, as well as the longer-term effects of evolving substance use to abuse. Comprehensive assessments over a range of developmental epochs should include not only structured laboratory measures but also observational reports by parents or teachers of day-to-day function in areas associated with EF.