Cognitive Functioning of Adolescent and Young Adult Cannabis Users in the Philadelphia Neurodevelopmental Cohort

Abstract

Cannabis use in youth is rising and has been linked to deficits in cognitive functioning. However, cognitive findings have primarily been based on small samples of users seeking treatment, and few studies have evaluated cognition in occasional cannabis users. Here, we examined 4,568 adolescents and young adults (ages 14–21) drawn from the Philadelphia Neurodevelopmental Cohort, a prospective, population-based study. Participants were classified as cannabis Non-Users (n=3,401), Occasional Users (twice per week or less; n=940), or Frequent Users (>3 times per week; n=227). Mixed-model analyses examined main effects of cannabis use and interactions between age and cannabis use on cognitive functioning. There was a significant interaction between cannabis group and age, such that adolescent (but not young adult) Frequent Users performed worse than Non-Users on measures of executive control (p=0.002). Earlier age of cannabis use was associated with worse performance in executive control in Occasional Users (p=0.04). Unexpectedly, Occasional Users exhibited better executive control, memory, and social cognition than Non-Users (ps<.05). Although mild executive control deficits in adolescent frequent users and a relation between early cannabis initiation and cognitive performance are partially consistent with prior research, cognitive deficits were not found in other hypothesized domains in this community-based sample. Moreover, occasional cannabis users displayed equivalent or even slightly better executive control, social cognitive, and memory abilities compared to non-users, suggesting complex relationships between cannabis use and cognition in youth. Longitudinal studies with community samples are needed to identify variables affecting risk and resilience to cognitive deficits associated with cannabis.

Introduction

Cannabis is the most used illicit substance among adolescents and young adults in the United States (US). The number of frequent users has increased over the last decade in the US, with 8.1 million persons over age 12 reporting daily or almost daily use (Substance Abuse and Mental Health Services Administration, 2014). Societal acceptance of cannabis has increased concurrent with decreases in perceived harm by adolescents (Substance Abuse and Mental Health Services Administration, 2014). These trends emphasize the importance of understanding antecedents and consequences of cannabis use in youth.

First use of cannabis typically occurs in adolescence, a period marked by substantial increases in cognitive capacities, particularly working memory, cognitive control, and other executive functions (Gur et al., 2012). These improvements are related to maturation of prefrontal regions and associated neural circuitry, which continues into early adulthood (Blakemore & Choudhury, 2006; Satterthwaite et al., 2013). There have therefore been increasing concerns about whether use of cannabis may disrupt healthy trajectories of brain, emotional, and cognitive development (Hadland, Knight, & Harris, 2015; Jacobus & Tapert, 2014), which may involve significant restructuring of the endocannabinoid system during adolescence (Ellgren et al., 2008). This concern is heightened for individuals who begin using substances regularly before completing “critical periods” in brain development (Schneider, 2008).

While there is broad consensus that cannabis use acutely causes deficits in attention, executive functions, and memory, its residual cognitive effects are still debated. Studies have linked frequent cannabis use in adolescence and young adulthood to dysfunction across multiple neuropsychological domains, including attention, executive control, processing speed, and episodic memory (Dougherty et al., 2013; Fontes et al., 2011; Gruber, Sagar, Dahlgren, Racine, & Lukas, 2012; Harvey, Sellman, Porter, & Frampton, 2007; Medina et al., 2007; Solowij et al., 2011). Dose-dependent cognitive effects have been shown in some (Gruber et al., 2012) but not all (Tait, Mackinnon, & Christensen, 2011) studies. Consistent with heightened vulnerability during adolescence, relationships between earlier initiation of cannabis use and cognitive deficits have also been reported (Fontes et al., 2011; Gruber et al., 2012; Solowij et al., 2011), although a number of studies have failed to find such associations (Crane, Schuster, Fusar-Poli, & Gonzalez, 2013). Other longitudinal studies have linked continued cannabis use to cognitive decline in IQ, executive functions, and episodic memory (Hanson et al., 2010; Tait et al., 2011), especially when use begins before age 15–17 and continues throughout young adulthood (Meier et al., 2012; Pope et al., 2003). Yet with sustained abstinence of a few weeks or more, many deficits appear minimal or subtle (Fried, Watkinson, & Gray, 2005; Hooper, Woolley, & De Bellis, 2014; Medina et al., 2007; Pardini et al., 2015; Tait et al., 2011), especially in adults (Grant, Gonzalez, Carey, Natarajan, & Wolfson, 2003; Schreiner & Dunn, 2012). In addition, two recent longitudinal studies reported that cannabis use in adolescence was not associated with IQ after adjusting for confounders and familial factors (Jackson et al., 2016; Mokrysz et al., 2016).

Despite established links between substance use and changes in affective processing during adolescence (Baker, Piper, McCarthy, Majeskie, & Fiore, 2004), compared with measures of “cold” cognitive functioning described above, measures of social cognitive functioning have been infrequently examined in cannabis using adolescents and emerging adults. A growing literature in adults suggests cannabis-associated deficits in social cognition, including poorer performance in identifying facial emotions (Hindocha et al., 2014; Platt, Kamboj, Morgan, & Curran, 2010), possibly related to cannabis-associated psychopathology (Huijbregts, Griffith-Lendering, Vollebergh, & Swaab, 2014). Affective processing is regulated by a complex system including prefrontal and temporal cortices, and limbic structures such as the amygdala (Nelson, Leibenluft, McClure, & Pine, 2005). Cannabinoid (e.g., CB₁) receptors are abundant in the limbic system, where they modulate mood and anxiety states (Witkin, Tzavara, & Nomikos, 2005). To this end, heavy cannabis-using adults show decreased activation in the anterior cingulate and amygdala during facial emotion viewing, suggesting altered brain mechanisms of affective processing (Gruber, Rogowska, & Yurgelun-Todd, 2009). Yet behavioral measures of affective processing have not, to date, been examined in adolescent cannabis users.

In sum, the presence of cannabis-induced cognitive and social cognitive deficits in adolescents and young adults remains controversial (Hooper et al., 2014; Jackson et al., 2016; Pope, 2002). It is difficult to draw firm conclusions from the extant literature on cannabis and cognition because of heterogeneity in methods, measurement, and sample characteristics. Most cognitive effects associated with cannabis use are small to medium in magnitude (Grant et al., 2003; Hanson et al., 2010; Medina et al., 2007; Schreiner & Dunn, 2012), and many studies were small and likely underpowered, which may result in both Type I and Type II errors (Ioannidis, 2005). Moreover, with few exceptions (Meier et al., 2012; Mokrysz et al., 2016; Tait et al., 2011) most studies have examined samples drawn from drug treatment facilities or universities, or have recruited individuals through advertising. These sampling methods may create selection biases and raise questions about generalizability. Here, we extend prior literature by examining cognitive functioning in cannabis users drawn from a large, prospectively accrued, community-based sample of adolescents and young adults from the Philadelphia Neurodevelopmental Cohort (PNC). The PNC was not enriched for substance use disorders or psychopathology, avoiding potential sampling biases. In addition, few studies have separately examined the cognitive functioning of less frequent cannabis users, despite the fact that these individuals constitute an overwhelming majority of cannabis-using youth. In addition to examining frequent cannabis users, similar to much of the extant literature, here we also examine cognitive functioning in occasional cannabis users from the PNC. Many prior studies also do not account for mental health comorbidity in cannabis users or recruit samples who are psychiatrically healthy, raising questions about generalizability. The size of the PNC offered the opportunity to examine associations with cannabis use along with psychopathology. Finally, we examined social cognition, which is rarely investigated in the context of cannabis but important for understanding antecedents and effects of drug use.

Given prior research, we hypothesized that cannabis users would show deficits in executive functioning, memory, and social cognition, with occasional cannabis users showing lesser deficits than frequent users. We also hypothesized that cognitive deficits would be more pronounced during adolescence and in those who initiated use at earlier ages, and that deficits would be diminished (but not eliminated) after controlling for psychopathology.

Method

Participants

The PNC is a large-scale, single-site, community-based study of 9,498 youths aged 8–21. The first wave of data, reported here, was collected from November 2009 to December 2011. Please see Calkins et al. (2015) for more extensive information on PNC recruitment, enrollment, and procedures. Participants were drawn from a pool of 50,293 youths who were recruited and genotyped by the Center for Applied Genomics at the Children's Hospital of Philadelphia (CHOP). Participants were selected at random after stratification by sex, age, and race/ethnicity. Inclusion criteria were intentionally liberal to extend generalizability (Calkins et al., 2015). See Supplementary Methods for inclusion/exclusion criteria and Figure S1 for a schematic of recruitment and formation of the analyzed sample. Participants and their guardians (for participants <18 yrs) provided written informed consent or assent. The Institutional Review Boards at Penn and CHOP approved the protocol.

Substance Use Assessments

PNC participants were assessed for substance use with either the Minnesota Twin and Family Study (MTFS) computerized substance use (CSU) assessment (Han, McGue, & Iacono, 1999) or Kiddie-Schedule for Affective Disorders and Schizophrenia (K-SADS) substance screener, depending on their study entry date. The majority of participants (n=6,298) received the CSU, which was privately self-administered on a laptop computer in the context of the Penn Computerized Neurocognitive Battery (CNB; described below). Probands were assessed independent of parents/collaterals and were informed that information reported would be kept confidential except as legally required. The CSU measure was designed for large-scale studies of adolescent substance use and has been applied to evaluate genetic and environmental contexts, influences, and predictors of substance use (Derringer, Krueger, McGue, & Iacono, 2008; Isen, Sparks, & Iacono, 2014; Samek, Keyes, Iacono, & McGue, 2013). It has been shown to demonstrate good reliability compared to more thorough diagnostic interviews, as well as greater disclosure of substance use compared to face-to-face interviews (Jackson et al., 2016). The CSU assesses whether individuals have ever used cannabis, alcohol, tobacco, downers, inhalants, LSD/hallucinogens, stimulants, cocaine, opiates, steroids, prescription drugs, and over-the-counter drugs of abuse. For cannabis, alcohol, and tobacco, additional questions gathered details about age at first use, age at first daily use (if applicable), frequency of use over the past year, methods of access, and consequences of use. Included in the assessment are probes to identify invalid assessments, including the use of fictitious street drugs (Derringer et al., 2008); endorsers of these drugs were excluded from analysis (n=32).

Prior to implementing the CSU measure, 2,964 participants recruited in the cohort received an assessor-administered version of the K-SADS substance screener questions, which were embedded within the clinical interview further described below (GOASSESS). The switch to the self-report measure was made to reduce interview length to accommodate a high participant volume (as many as 150 participants assessed per week), and to facilitate disclosure by adolescents (e.g., to avoid social desirability biases that might lead to underreporting of substance use). Since questions were not asked in GOASSESS about frequency of cannabis use, here we include only participants who denied use of cannabis during the GOASSESS screener. This resulted in excluding 238 participants. Participants who received the GOASSESS substance screener were slightly older than those who received the self-report assessment (18.7 vs. 16.5, respectively; p < 0.001).

As shown in Table 1, no collaterals for the 8–10 year old probands and few 11–13 year old probands endorsed cannabis use. Thus, analyses were only conducted in participants 14 years and older (n=4,568).

Table 1.

Cannabis Use by Age in the Philadelphia Neurodevelopmental Cohort (PNC)

	Age
Cannabis Group	8–11	12	13	14	15	16	17	18	19	20	21	Total
Non-User	2,975 (99.9%)	689 (99.3%)	689 (97.2%)	687 (92.8%)	638 (84.4%)	595 (75.5%)	483 (67.3%)	469 (64.5%)	328 (68.5%)	136 (56.4%)	64 (53.8%)	7,753 (86.7%)
Occasional User	2 (0.1%)	4 (0.6%)	16 (2.3%)	46 (6.2%)	98 (13.0%)	159 (20.2%)	191 (26.7%)	202 (27.8%)	119 (24.8%)	81 (33.6%)	44 (37.0%)	962 (10.8%)
Frequent User	0 (0%)	1 (0.1%)	4 (0.6%)	7 (1.0%)	20 (2.7%)	34 (4.3%)	43 (6.0%)	56 (7.7%)	32 (6.7%)	24 (10.0%)	11 (9.2%)	232 (2.6%)
Total	2,977	694	709	740	756	787	715	727	479	241	119	8,947

Classification of Cannabis Use

After considering a number of possibilities for cannabis use classifications, we classified cannabis users as Frequent Users (endorsed cannabis use over the past year either “Daily or Almost Daily” or “3–4 Times Per Week”) and Occasional Users (endorsed past use of cannabis or use “1–2 Times Per Week” or less over the past year). The cannabis literature is inconsistent regarding clinically meaningful levels or markers of cannabis use for cognitive and neurobiological outcomes. We constructed our classification of Frequent User to be consistent with a number of studies examining frequent or heavy use (e.g., Mathias et al., 2011; Meier et al., 2012), and followed these analyses by examining early initiation of cannabis given its prominence in the literature (Lisdahl, Gilbart, Wright, & Shollenbarger, 2013). Our Occasional User group was constructed to provide a group with less frequent cannabis use than is typically studied (and is often not considered detrimental to cognitive functioning), and criteria are consistent with the few studies that have examined occasional users (Desrosiers, Ramaekers, Chauchard, Gorelick, & Huestis, 2015; Harvey et al., 2007). While prior research has rarely utilized continuous measures of cannabis use as predictors, we considered employing a continuous cannabis use measure, but ultimately decided against it given the questionable precision of this measurement with retrospective recall at this timescale, and the inability of a continuous measure to capture a threshold of “problematic use,” especially if there is a nonlinear relationship between use and cognitive functioning. Moreover, our measurement of cannabis use was also not collected as a continuous variable but instead queried in discrete categories that may not be equally spaced; thus, transformation into a continuous measure did not appear justified.

Clinical Assessment

Psychopathology assessment was conducted through GOASSESS, an abbreviated, computerized version of the K-SADS adapted to collect information on symptoms, duration, distress, and impairment for lifetime mood, anxiety, behavioral (externalizing), psychosis spectrum symptoms, suicidal thinking and behavior, and treatment history. GOASSESS has been described in detail previously (Calkins et al., 2015).

Neurocognitive Assessment

Cognitive functioning was assessed with the 1-hour Penn Computerized Neurocognitive Battery (CNB; Gur et al., 2012; Moore, Reise, Gur, Hakonarson, & Gur, 2015), a collection of twelve neurocognitive tests assessing a range of cognitive functions (see Table 2). Numerous studies support the validity and reliability of the CNB across a wide age range, in both healthy and psychiatric samples (e.g., Almasy et al., 2008; Moore et al., 2015). The sensitivity of the CNB has been previously established by showing age and sex effects on performance in a subsample of 3,500 PNC participants (Gur et al., 2012). Moreover, CNB tests reflect the recruitment of specific brain systems (Roalf et al., 2014; Satterthwaite et al., 2013), making them potential biomarkers of brain dysfunction. Validity indicators embedded within the CNB, such as impossible reaction times, outliers, pattern making in responses, and lack of responses/disengagement from task, were inspected. Non-valid data for individual tests were removed from analyses. Raw accuracy and speed scores were normalized by age (i.e., z-transformed) within the entire cohort of the PNC study, based on the mean and standard deviation of participants within a 2-year age bin, as previously detailed (Gur et al., 2012). For data reduction and Type I error control, we analyzed neurocognitive domain subscale scores, which were previously created using a factor analysis of the PNC CNB data (Moore et al., 2015). These scores are latent variables that reflect neurocognitive efficiency, which is the sum of an individual’s standardized accuracy and speed scores. The subscale scores were developed using correlated factors models, which suggested four neurocognitive factors corresponding to “Executive Control,” “Episodic Memory,” “Complex Cognition,” and “Social Cognition” (see Moore et al., 2015). Each of these factors integrates data from two to four tests (Table 2). See Supplementary Methods for further details on the CNB.

Table 2.

Tests and domains in the Penn Computerized Neurocognitive Battery (CNB)

Neurocognitive Domain	Cognitive Functions Assessed	Test
Executive Control	Sustained Attention	Penn Continuous Performance Test (PCPT)
	Working Memory	Letter N-Back Test (LNB)
Episodic Memory	Verbal Episodic Memory	Penn Word Memory Test (PWMT)
	Face Memory	Penn Face Memory Test (PFMT)
	Spatial Episodic Memory	Visual Object Learning Test (VOLT)
Complex Cognition	Abstraction/Mental Flexibility	Penn Conditional Exclusion Test (PCET)
	Language Reasoning	Penn Verbal Reasoning Test (PVRT)
	Nonverbal Reasoning	Penn Matrix Reasoning Test (PMAT)
	Visuospatial Ability	Penn Line Orientation Test (PLOT)
Social Cognition	Emotion Identification	Penn Emotion Identification Test (PEID)
	Emotion Differentiation	Penn Emotion Differentiation Test (PEDT)
	Age Differentiation	Penn Age Differentiation Test (PADT)

Statistical Analyses

Groups were compared on demographic and substance use variables using ANOVAs and chi-square tests. Neurocognitive efficiency scores were entered as dependent measures in mixed-model repeated measures analysis in Stata 13, with cannabis group and age (continuous) as fixed factors, and neurocognitive domain as a within-group factor. Previous results showed greater cannabis-related cognitive deficits in adolescents compared to adults (Jacobus & Tapert, 2014); thus, we examined three-way interactions among group, cognitive domain, and age. Planned follow-up analyses examined interactions and the directionality of main effects using family-wise error rate adjustments to control for multiple tests. All analyses were adjusted for sex because of reported sex differences in cognitive performance (Gur et al., 2012). Interactions between sex and cannabis group were also examined given prior results (Crane et al., 2013).

We examined the relationship of neurocognitive performance to age at first cannabis use with multiple regressions controlling for age and sex. To evaluate the specificity of effects, we conducted a series of supplementary analyses that included environmental factors, comorbid substance use, and psychopathology. Specifically, these analyses included a factor score representing participant neighborhood socioeconomic status (SES) (derived previously by linking census data to our sample; see Supplemental Methods and Moore et al., 2016); alcohol and tobacco use in the past year; and empirically-derived psychopathology factor scores from a bi-factor confirmatory model constructed from GOASSESS item-level data. This model parsed psychopathology into a general factor, representing the overall burden of psychopathology, and four orthogonal symptom dimensions representing anxious-misery (mood & anxiety), fear (phobias), behavioral/externalizing, and psychosis-spectrum symptoms (see Supplementary Methods and Shanmugan et al., 2016).

Results

Demographic, Clinical, and Substance Use

As shown in Table 2, both cannabis groups were significantly older and included more males than Non-Users, and Frequent Users included more males than Occasional Users. Groups were relatively similar in racial composition. Both cannabis groups evidenced slightly lower estimated IQs than Non-Users, although the magnitude of this difference was small, and all groups had IQs within the average range. Frequent Users started using cannabis at younger ages than Occasional Users. Users reported more frequent alcohol use and higher levels of overall psychopathology. Moreover, both cannabis groups had higher levels of externalizing/behavior symptoms and lower levels of fear/phobia symptoms than Non-Users. Occasional Users had higher levels of mood-anxiety symptoms but lower levels of psychosis symptoms than Non-Users.

Cannabis and Cognitive Functioning

The mixed model analysis on neurocognitive efficiency scores showed significant main effects for cognitive domain (χ²(3)=31.74; p<0.0001) and sex (χ²(1)=4.32; p=0.04), but not age (χ²(1)=0.15; p=0.70) or cannabis group (χ²(2)=4.73; p=0.09). Significant two-way interactions were found for cognitive domain x age (χ²(3)=32.21; p<0.0001) and cannabis group x cognitive domain (χ²(6)=35.52; p<0.0001), but not age x cannabis group (χ²(2)=4.45; p=0.11) or cannabis group x sex (χ²(2)=1.20; p=0.55). Notably, the age x cognitive domain x cannabis group interaction was significant (χ²(6)=36.43; p<0.0001). Figure 1 presents the cognitive profile for cannabis groups across all ages (values shown in Table S2). Figure 2 shows interactions between age and cannabis group within each cognitive domain.

Figure 1.

Cognitive performance of Philadelphia Neurodevelopmental Cohort Youth by Cannabis Use Status

Note: Data shown are mean factor scores for neurocognitive efficiency (z-transformed and age- normed) and ±95% confidence interval.

Figure 2.

Cognitive Performance in the Philadelphia Neurodevelopmental Cohort Youth by Cannabis Use Status and Age.

Note. Data shown are estimated marginal means re-centered to the means of the Non-Users and adjusted for sex.

Decomposing the significant three-way interaction for neurocognitive efficiency showed a significant two-way interaction of age x cannabis group in executive control (F[2, 4560]=6.14; p=0.002), and a main effect of cannabis group in executive control (F[2, 4560]=5.91; p=0.003). As shown in Figure 2, Frequent Users performed worse in executive control than Non-Users at younger but not older ages (t=3.25; p=0.001), roughly corresponding to adolescence (i.e., ages 14–17). Occasional Users performed better than Non-Users in executive control (t=3.11; p=0.002). Both of these effects were relatively small in magnitude (Figure 2).

With family-wise error correction, there were no interactions between age and cannabis group in episodic memory (F[2, 4560]=0.09; p=0.92), complex cognition (F[2, 4560]=3.15; p=0.04), or social cognition (F[2, 4560]=2.70; p=0.07). However, there were main effects of cannabis group on episodic memory (F[2, 4560]=9.00; p=0.001) and social cognition (F[2, 4560]=4.28; p=0.01), but not complex cognition (F[2, 4560]=3.09; p=0.05). Follow-up analyses indicated that Occasional Users performed better than Non-Users in memory (t=4.22; p<.001) and social cognition (t=2.65; p=0.008), but Frequent Users were not significantly different from other groups in these domains (ps>.10).

Supplementary analyses evaluated whether heterogeneity in the Occasional User group (e.g., including past users of cannabis in the group) may have influenced observed results. We re-ran analyses including only those who endorsed cannabis use in the past year in the Occasional Users (n=597). Convergent results were obtained when analyzing this subgroup (see Supplemental Results). Further supplementary analyses revealed that there was not substantial variability across use levels of the Occasional User groups in terms of cognitive performance in any domain (see Supplemental Results).

Associations of Cognitive Functioning with Age at First Cannabis Use

In Occasional Users, earlier age of cannabis use was associated with poorer executive control (β=.08; p=0.04), with trend-level findings for complex cognition (β=.07; p=.07) and social cognition (β=.07; p=.07). In Frequent Users, earlier age of cannabis use was associated with trend-level findings for executive control (β=.14; p=.05) and complex cognition (β=.13; p=.07). Additional post hoc analyses were conducted to examine whether those who initiated cannabis at earlier ages (i.e., before age 16) showed reduced neurocognitive performance compared to those who initiated cannabis at later ages. Results showed no significant influence of age at cannabis initiation (see Supplement).

Neighborhood SES, Psychopathology, and Other Substance Use

To evaluate the specificity of findings and examine whether psychopathology and alcohol and tobacco use influenced observed results, we also performed a series of post-hoc supplementary analyses. These follow-up analyses included neighborhood SES, the four orthogonal dimensions of psychopathology (from the factor analysis described above), alcohol use, and tobacco use as covariates in each model. As detailed below, findings convergent with primary analyses were predominantly obtained when these were included in analyses.

As shown in Table S2, each psychopathology factor and alcohol use were significantly associated with executive control performance (though mood/anxiety symptoms and alcohol use were positively associated with cognitive performance), and the cannabis group x age interaction (F[2, 4074]=5.95; p=0.003) remained significant, with a similar pattern of results as in primary analyses. In addition, although alcohol use, psychosis spectrum, mood/anxiety, and fear/phobia symptoms were associated with memory performance, the relationship between cannabis use and memory was relatively unchanged, with a significant main effect of cannabis group (F[2, 4076]=4.91; p=0.007) and Occasional Users performing better than Non-Users (Table S3). As shown in Table S4, each psychopathology factor was significantly associated with complex cognition performance, but there was not a significant main effect of cannabis group (F[2, 4076]=0.52; p=0.595) once psychopathology, alcohol use, and SES were included in analyses. In social cognition, although alcohol and tobacco use, externalizing/behavior, psychosis spectrum, mood/anxiety, and fear/phobia symptoms were associated with performance, the main effect of cannabis group on social cognition was reduced to a trend (F[2, 4076]=3.39; p=0.034) in these analyses (Table S5).

Discussion

This study represents the largest to date examining associations between cannabis use and cognitive functioning in adolescents and young adults. Consistent with our hypotheses, we found that adolescent frequent cannabis users showed small performance deficits on measures of executive control compared to non-users. We also found that earlier age of initiated use was associated with slightly worse cognitive functioning in executive control. However, occasional users of cannabis performed similarly to non-users on measures of complex cognition, and even showed higher performance on measures of executive control, social cognition, and memory compared to non-users. Importantly, convergent results were predominantly obtained after accounting for the influence of neighborhood SES, alcohol and tobacco use, and psychopathology on cognitive functioning.

Frequent Cannabis Use and Cognitive Functioning

Our findings of mild executive control deficits in adolescent frequent users are consistent with prior research showing that adolescent cannabis abusers have difficulty efficiently sustaining attention (Dougherty et al., 2013; Fontes et al., 2011). In addition, working memory deficits in adolescent cannabis users have been shown to persist after abstinence and predict subsequent increases in use (Meier et al., 2012; Winward, Hanson, Tapert, & Brown, 2014). Thus, adolescent frequent cannabis users may have difficulty efficiently sustaining attention and mentally manipulating information, potentially due to dysfunction in prefrontal executive networks (Jacobus & Tapert, 2014). However, in contrast to adolescents, minimal evidence for executive deficits was found in young adult frequent users, which was particularly surprising since 92.8% of these individuals reported initiating cannabis use before age 17.

The absence of deficits in memory, complex cognition, and social cognition in frequent cannabis users was unexpected. These findings differ from much prior research in adolescents and young adults, which shows that frequent cannabis use is associated with cognitive deficits within these domains, even after sustained abstinence (Hanson et al., 2010; Medina et al., 2007; Meier et al., 2012; cf. Schreiner & Dunn, 2012). Limited statistical power cannot explain this discrepancy, as our sample size provided ample power, and the CNB has shown sensitivity to detect small-to-medium sized effects in prior studies (Gur et al., 2012, 2014). Moreover, our user groups had greater levels of overall and externalizing psychopathology, greater levels of alcohol use, and lower estimated IQs compared to non-users, which makes the relatively modest cannabis-associated effects even more notable. The most likely explanation for our divergent findings is differences in sample characteristics. To extend generalizability, the PNC recruited individuals representative of the greater Philadelphia population and was not enriched for substance use or psychopathology. Therefore, although frequent users reported daily or almost daily use over the past year, similar to criteria used in prior studies (e.g., Solowij et al., 2011), our sample may have had less frequent or problematic cannabis use compared to prior help-seeking samples. Moreover, a longer duration of drug exposure or greater psychopathology may be necessary to produce cognitive effects observed in prior work.

Occasional Cannabis Users and Cognitive Functioning

Adolescents and young adults who endorsed occasional use of cannabis (i.e., twice weekly or less) displayed equivalent or slightly better cognitive performance than youth who denied ever using cannabis. Together with results from frequent users, these findings suggest a complex, potentially non-linear relationship between cannabis use and cognitive functioning, such that occasional users generally show slightly better cognitive performance, while frequent users generally show slightly worse performance in certain cognitive domains, compared to non-users. Given this cross-sectional study spanning childhood to young adulthood, the better performance of occasional users compared to non-users may suggest that individuals with higher social cognitive, executive, and memory abilities are more vulnerable to cannabis use. It should be noted that the better performance in participants who used cannabis occasionally was substantially attenuated by adjustment for neighborhood SES, alcohol and tobacco use, and psychopathology, especially in complex cognition and executive control. However, the lack of deficits in this group support recent work reporting that cognitive deficits in occasional adolescent cannabis users were minimal once pre-use psychopathology and cognitive functioning were accounted for (Pardini et al., 2015), although a recent study with a long follow-up period found modest effects of occasional cannabis use on episodic memory functioning (Auer et al., 2016). Results are also informative when considered alongside studies showing minimal effects of occasional cannabis use on academic functioning, especially in individuals who ceased use (Maggs et al., 2015; Pardini et al., 2015).

Early Initiation of Cannabis Use

Consistent with prior research (Fontes et al., 2011; Gruber et al., 2012), earlier age of cannabis initiation was associated with reduced efficiency in executive control, although these effects were small in magnitude (i.e., of unknown clinical significance) and were not present when examining “early users” as a group. Thus, there may be small negative effects of earlier initiation or longer duration of use; alternatively, pre-existing deficits could increase risk for earlier use. Longitudinal research is needed to infer causality and resolve inconsistency in findings (Pope et al., 2003).

Cannabis Use and Social Cognition

To our knowledge, this is the first study to examine behavioral measures of social cognition in adolescent cannabis users. We found that occasional cannabis users evidenced better performance than non-users in the social processing of faces, while frequent cannabis users displayed similar performance to non-users. Notably, females typically outperform males in emotion identification and differentiation (Gur et al., 2012; Hindocha et al., 2014), making it even more striking that occasional users performed better than non-users given the higher percentage of males in this group. While this performance advantage could theoretically arise as a result of cannabis use, it more likely precedes cannabis initiation and may be linked with proneness to use. Enhanced social cognition associated with occasional cannabis use may relate to the social context of use, which is often driven by peer interactions and influences (Ali, Amialchuk, & Dwyer, 2011). In other words, obtaining cannabis as an adolescent may involve enhanced social functioning. Prior work has shown potential alterations in brain mechanisms underlying social cognition in young adult cannabis users (Gilman, Lee, et al., 2016; Gilman, Schuster, et al., 2016). Interestingly, previous studies assessing emotion identification during acute tetrahydrocannabinol administration or in adult cannabis abusers reported deficits in facial emotion processing (Bossong et al., 2013; Hindocha et al., 2014; Platt et al., 2010). The absence of emotion processing deficits in our frequent users may again relate to our community-ascertainment approach. Prior studies based on convenience or treatment-seeking samples may have had greater psychiatric comorbidity and consequently poorer emotion perception. In our sample, better social cognition performance was reduced to a trend in occasional users when psychopathology was included as a covariate, lending support to this interpretation.

Interpretation Considerations

Our cross-sectional data do not allow the conclusion that cannabis is not detrimental to cognitive functioning. Instead, our results suggest that large magnitude cognitive effects associated with cannabis use may not be evident at a population level in adolescents and young adults. Yet there may be critical variables that increase one’s risk for cognitive deficits associated with cannabis use. For example, most group differences in measures of brain structure and function or in cognitive functioning were found in youth who either used cannabis heavily for an extensive period or before particular ages. Most participants were also seeking treatment, suggesting a greater level of mental health concerns, which have been related to risk for cognitive impairment (Weiser et al., 2004). As such, psychopathology and alcohol use were broadly associated with cognitive functioning in our sample and should be accounted for in future research. Moreover, results have been heavily dependent on sample characteristics and research design. For example, although memory deficits have been noted as a robust, consistent finding in studies of heavy, chronic cannabis use in youth (Crane et al., 2013; Jacobus & Tapert, 2014), some studies have not replicated prior findings (Gruber et al., 2012; Mahmood, Jacobus, Bava, Scarlett, & Tapert, 2010), have reported that associations are accounted for by other correlated factors (Pope et al., 2003) or comorbidities (Lisdahl & Price, 2012), or found that effects were minimized after abstinence of at least three weeks (Fried et al., 2005; Hanson et al., 2010; Hooper et al., 2014; Medina et al., 2007). Thus, specific groups of cannabis users—such as those in treatment, with greater psychiatric comorbidity, or with extended heavy use—might show larger associations between cannabis use and cognitive functioning. Small effects could also be clinically meaningful for a given individual, especially considering other factors within an individual that may influence performance, such as intellectual endowment, socioeconomic status, psychiatric comorbidity, or other substance use. Studies that examine interactions of such factors in predicting cognitive outcomes will be especially informative to help identify brain-behavior features of individuals at high risk of substance initiation or problems.

As with all cross-sectional studies, we are unable to determine whether cognitive performance differences observed existed prior to cannabis use (i.e., may predict initiation (Giancola & Tarter, 1999)), are a consequence of use, or reflect interactions with pre-existing cognitive vulnerabilities. Moreover, deficits may reflect shared causal pathways between cannabis use and correlated sociobehavioral factors (Daly, 2013; Hooper et al., 2014; Rogeberg, 2013). Thus, prospective, longitudinal studies of youth before and after cannabis initiation are essential to examine causality and the complex interplay of factors affecting cognitive functioning in cannabis users (Macleod et al., 2004). We are following a subset of these individuals longitudinally with extensive substance use assessments, which will hopefully contribute to enhanced understanding of such crucial questions.

Study Strengths and Limitations

The notable strengths of this study include its large sample; deep phenotyping of a cohort not enriched for substance use or psychopathology, nor recruited through advertisements; administration of a comprehensive neurocognitive battery; and exploration of social cognition. Study limitations include the brevity of our substance use assessment, potential unreliability of retrospective recall, absence of objective use indicators, lack of information regarding recency of substance use, cannabis use disorder symptoms, and limited information on functional outcomes. Future studies would benefit from more comprehensive assessments of use patterns and detailed information about other drug use and cannabis use disorders, which was infeasible with the large-scale data collection required for this study. Moreover, future community-based studies of cannabis users should examine additional areas of cognitive functioning such as planning and decision-making. It should also be noted that our findings do not speak to associations between cannabis use and other significant physical and mental health outcomes, such as lung functioning, motivation, anxiety/depression, or risk for psychosis (Arendt, Rosenberg, Foldager, Perto, & Munk-Jørgensen, 2005).

Conclusions

Consistent with prior research, we found small deficits in executive functioning in adolescents endorsing frequent cannabis use. However, results also highlighted relatively modest cognitive deficits associated with cannabis use in adolescents and young adults at a population level when participants were recruited with a community ascertainment approach. Moreover, we identified relative strengths in executive control, memory, and social cognition in occasional cannabis users, which may inform future studies to identify factors associated with cannabis use initiation.

Table 3.

Demographic and substance use characteristics of Philadelphia Neurodevelopmental Cohort (PNC) participants 14 and older, separated by cannabis use

	Non-User n = 3,401	Occasional Cannabis User n = 940	Frequent Cannabis User n = 227	Occasional vs. Non-User p	Frequent vs. Non-User p	Occasional vs. Frequent p	p
Age	16.4 (1.9)	17.4 (1.8)	17.6 (1.7)	<0.001	<0.001	0.134	<0.001
Sex M/F (%Male)	1,419/1,982 (41.7%)	437/503 (46.5%)	156/71 (68.7%)	0.009	<0.001	<0.001	<0.001
Race				---	---	---	0.434
Caucasian	1,889 (55.5%)	549 (58.4%)	119 (52.4%)	---	---	---
African-American	1,147 (33.7%)	300 (31.9%)	82 (36.1%)	---	---	---
Asian/Native American/Other	365 (10.7%)	92 (9.7%)	26 (11.5%)	---	---	---
WRAT-4 Word Reading	101.7 (17.5)	100.6 (15.8)	98.5 (14.7)	0.059	0.005	0.100	0.006
Age at First Cannabis Use	---	14.6 (1.7)	13.7 (1.6)	---	---	<0.001
Age at Daily Cannabis Use	---	---	15.4 (1.8)	---	---	---
Cannabis Use During Past Year				---	---	---
No use in past year	3,401 (100.0%)	343 (36.5%)	0 (0.0%)	---	---	---
< 1 time a month	0 (0.0%)	272 (28.9%)	0 (0.0%)	---	---	---
Once a month	0 (0.0%)	94 (10.0%)	0 (0.0%)	---	---	---
2–3 times a month	0 (0.0%)	137 (14.6%)	0 (0.0%)	---	---	---
1–2 times a week	0 (0.0%)	94 (10.0%)	0 (0.0%)	---	---	---
3–4 times a week	0 (0.0%)	0 (0.0%)	77 (33.9%)	---	---	---
Everyday or nearly everyday	0 (0.0%)	0 (0.0%)	150 (66.1%)	---	---	---
Alcohol Use During Past Yeara				<0.001	<0.001	<0.001	<0.001
No use	2,184 (78.9%)	131 (16.3%)	24 (11.3%)	---	---	---
< 1 time a month	342 (12.4%)	231 (28.7%)	41 (19.4%)	---	---	---
Once a month	95 (3.4%)	120 (14.9%)	32 (15.1%)	---	---	---
2–3 times a month	103 (3.7%)	167 (20.7%)	61 (28.8%)	---	---	---
1–2 times a week	40 (1.5%)	116 (14.4%)	42 (19.8%)	---	---	---
3–4 times a week	4 (0.1%)	36 (4.5%)	11 (5.2%)	---	---	---
Everyday or nearly everyday	0 (0.0%)	5 (0.6%)	1 (0.5%)	---	---	---
Psychopathology Factors b
Overall Psychopathology	0.07 (0.5)	0.25 (0.5)	0.42 (0.5)	<0.001	<0.001	<0.001	<0.001
Mood-Anxiety	0.01 (0.4)	0.04 (0.4)	0.03 (0.4)	0.006	0.391	0.566	0.020
Fear/Phobia	0.04 (0.6)	−0.09 (0.6)	−0.20 (0.6)	<0.001	0.016	<0.001	<0.001
Externalizing/Behavior	−0.19 (1.1)	0.06 (1.2)	0.49 (1.2)	<0.001	<0.001	<0.001	<0.001
Psychosis Spectrum	0.04 (0.8)	−0.10 (0.7)	0.04 (0.8)	0.011	0.982	0.011	<0.001

Note. WRAT-4 = Wide Range Achievement Test-4^th Edition.

^aBecause of limited cell sizes, chi-square analysis compares 3 levels: alcohol non-users, use < 1–2 a week, and use ≥1–2 a week.

^bFactor scores from a bifactor model (Calkins et al., 2015) reflecting orthogonal dimensions of psychopathology, with higher scores indicating more symptoms.