One summary measure from each test was chosen a priori as the best estimate of the function of that test. We factor analyzed the test battery to reduce the number of variables. Supplementary text and Table S1 provide extensive detail on the battery. We examined missing data prior to implementing multiple imputation. From a sample of 1043, 953 received baseline neurocognitive testing . Of the CHR sample that transitioned to psychosis during the two-year follow-up , 89 received testing . Overall data completeness for the tested sample was 96.6% for 19 test variables. After MI, we conducted a FA of the 19 neurocognitive variables . All analyses were done with SPSS, version 23.Groups were HCs, CHR converters and non-converters . T-tests, Kolmogorov-Smirnov Z and Chi Square tests were used to assess demographic comparability. Due to differences in age and maternal education, we controlled for both using MANCOVA and also controlled for site as a random effects factor with a linear mixed model. We covaried for estimated and premorbid IQ to test the role of general intellectual ability in cognitive dysfunctions. We compared medicated vs. Non-medicated groups of CHR +C vs HC, and CHR+C vs CHR-NC by conducting MANOVA with planned comparisons using residualized factor scores generated from the linear mixed models. To examine group cognitive profiles we residualized out age and maternal education from all neurocognitive indices . Area under the curve was calculated by the ROC program in SPSS. Prediction of conversion to psychosis and time to conversion was assessed by logistic and Cox regression. Covariates were selected based on similar prediction analyses conducted in NAPLS-164 and NAPLS-245 and entered into the model if they were associated with survival time and predicted conversion status in logistic regression. Survival time was time to the last SIPS interview or conversion, ebb and flow rolling benches whichever occurred first. Candidate covariates were added to the model as a block then subjected to backward selection with a criterion p value of 0.10. Candidates that survived at p ≤ .05 within domain were entered into an omnibus model.

ESs were calculated with Cohen’s d. Bonferroni corrected significance for mean comparisons was set for individual tests at p<. 00263 and for factors at p< .0125 .In the largest and most detailed study of CHR prodromal cases, using a multi-site, case control design and standardized assessments, we demonstrated that individuals at CHR were impaired in virtually all neurocognitive dimensions compared to controls, and this could not be accounted for by premorbid or current general cognitive ability, current depression, medications, alcohol or cannabis abuse. ESs in comparison to HCs for Declarative Memory and Attention/WM were large for CHR+C participants. Compared to CHR-NC, CHR+C participants were significantly impaired in Attention/WM and Declarative Memory, the latter significantly predicting conversion to psychosis and time to event in concert with positive symptoms. Comparable impairments were observed in never-medicated and currently unmedicated CHR-NCs and CHR+C’s. These data demonstrate the sensitivity of neurocognitive function as a component risk marker for psychosis.Our findings support theoretical models hypothesizing Attention/WM impairments, and even more strongly, impaired Declarative Memory, as central to the CHR stage.The results are consistent with NAPLS-1, in which Declarative Memory had the largest ES decrement and roughly the same magnitude in CHR+C.The distinct profile of performance across domains, especially in CHR+C, suggests that at the incipient psychotic phase, specific forms of neurocognition are affected and are predictive of later psychosis. Within CHR participants, there was considerable variability in neurocognitive performance. CHR-NC’s impairments , were on the order of other psychiatric disorders in young people, such as attention-deficit/hyperactivity disorder. CHR+Cs impairments were approximately 57% larger, although smaller than those observed in first episode schizophrenia . Analyses of individual variability and longitudinal analyses are needed to identify how profile and severity differ according to comorbid disorders, final diagnoses and pre- and post-conversion. A key question was how neurocognitive deficits are associated with medication status. Psychotropic-naive and unmedicated subgroups had significant impairments comparable to the overall CHR subgroups. Treated groups, including with antipsychotic medications, were largely comparable to those without treatment, except they had somewhat greater Attention/WM impairment.

These observations emphasize the essential nature of neurocognitive impairment in the CHR stage and de-emphasize the role of medications as confounders in our results. Our design precludes conclusions about causality and future work should study the effect of medications on neurocognition in CHR populations in a prospective design. There were a number of other potentially important observations. The unexpectedly higher Verbal score that was retained in logistic and Cox regressions in concert with impaired Declarative Memory was not a significant predictor in univariate comparisons. This pattern of high verbal premorbid ability and impaired memory, coupled with P1/P2 composite appears to be a pernicious combination predicting conversion and needs replication. Importantly, the BVMT-R showed comparably large impairments as the two verbal memory tasks, highlighting that Declarative Memory deficits in CHR are not solely verbal, and that Declarative Memory impairments are key neurocognitive risk markers . Neurocognitive tests used in concert with other clinical and psychobiological measures may enhance prediction of psychosis or functional outcome. For example, in analyses limited to two tests selected from literature review14 prior to these neuropsychological analyses, NAPLS-2 investigators found that the HVLT-R and BACS Symbol Coding added modest but significant independent predictive power above the clinical measures in a risk calculator algorithm for psychosis conversion and this was replicated in an independent non-NAPLS sample. Similar results have been observed in other studies. In this study, we showed that other tests, including BVMT-R, PAM, and ACPT QA Vigil added significant independent variance beyond P1-P2 symptoms, augmenting the importance of neurocognitive markers. NAPLS-2, because of its large sample from diverse geographical areas, extensive neurocognitive coverage, remarkably complete neurocognitive dataset, and large never medicated sample, allowed for a strong confirmation of neurocognitive hypotheses. The NAPLS-2 study built upon and improved the NAPLS-1 assessment, confirming and expanding prior results . This broad range of measures expanded the scope of what is known about CHR neurocognition. Limitations include the fact that most of these tests and factors are complex. Thus, while Declarative Memory is clearly affected, the tasks tapping this domain cannot parse the specific mechanisms underlying the deficits.

Further research with more molecular measures of cognition, such as those developed by CENTRACS, may allow specification of the cognitive processes underlying the deficits. We did not randomize or counterbalance the order of tests, so we cannot rule out order effects. However, the most impaired tasks were spread out across the battery from the sixth to the last tests in the battery so there is no obvious fatigue effect. Adolescence is a developmental period between childhood and adulthood characterized by marked physiological, psychological, and behavioral changes. Adolescents experience rapid physical growth, sexual maturation,rolling grow benches and advances in cognitive and emotional processing . These changes coincide with increases in substance use, with alcohol being the most widely used illegal substance among adolescents . National survey data indicate that 33% of 8th grade students have tried alcohol, and this percentage increases to 70% among 12th graders . Of greater concern is the increase in heavy episodic drinking where prevalence rates increase from 6% to 22% for 8th and 12th grades, respectively , as heavy episodic drinking during adolescence is associated with numerous negative effects on adolescent health and well being, including risky sexual behaviors , hazardous driving , and alterations in adolescent brain development . During adolescence, the brain undergoes significant changes, and a recent longitudinal neuroimaging study suggests that heavy episodic drinking during this developmental period alters brain functioning . Squeglia and colleagues examined the effects of heavy episodic drinking on brain function during a visual working memory task, comparing brain activity in adolescents at baseline and again at follow-up to compare brain activity in those who transitioned into heavy drinking during adolescence to demographically matched adolescents who remained nondrinkers. Adolescents who initiated heavy drinking exhibited increasing brain activity in frontal and parietal brain regions during a visual working memory task compared to adolescents who remained nondrinkers through follow-up, who showed decreasing frontal activation, consistent with studies in typical development . Thus, adolescent heavy episodic drinking may alter brain functioning involved in working memory; however, additional longitudinal studies are needed to explore the effects of alcohol on neural correlates of other vital cognitive processes, such as response inhibition.

Response inhibition refers to the ability to withhold a prepotent response in order to select a more appropriate, goal-directed response . The neural circuitry underlying response inhibition develops during adolescence , and as such, brain response during inhibition changes during adolescence . Briefly, cross-sectional research indicates that brain activation during response inhibition transitions from diffuse prefrontal and parietal activation to localized prefrontal activation . Longitudinal studies report that atypical brain responses during response inhibition, despite comparable performance, is predictive of later alcohol use , substance use and dependence symptoms , and alcohol-related consequences . Together, these findings indicate that neural substrates associated with response inhibition change over time and abnormalities in development may contribute to later substance use. To this end, the current longitudinal fMRI study examined the effects of initiating heavy drinking during adolescence on brain activity during response inhibition. We examined blood oxygen level dependent response during a go/no-go response inhibition task prior to alcohol initiation , then again on the same scanner approximately 3 years later, after some adolescents had transitioned into heavy drinking. Based on our previous findings , we hypothesized that adolescents who transition into heavy drinking would show reduced BOLD response during response inhibition prior to initiating heavy drinking followed by increased activation after the onset of heavy episodic drinking, as compared to adolescents who remained non-users. By identifying potential neurobiological antecedents and consequences of heavy episodic drinking, this study will extend previous research on the effects of alcohol on brain function and point to risk factors for heavy episodic drinking during adolescence.Table 1 provides baseline and follow-up descriptive information for Heavy Drinkers and Controls. Of note, the ability to inhibit prepotent responses improved with age with no differences in this improvement between groups . The no-go versus go contrast at baseline revealed activations consistent with meta-analyses of response inhibition showing significant clusters of activation in inferior, superior, and medial frontal gyri, and in parietal, temporal, cerebellar, and subcortical areas . Because Heavy Drinkers reported significantly more substance use than Controls at followup, a lifetime substance use composite and biological sex were included as covariates. A repeated measures ANCOVA revealed significant group × time interactions in 5 regions: the bilateral middle frontal gyri, right inferior parietal lobule, left putamen, and left cerebellar tonsil . At baseline, Heavy Drinkers showed significantly less no-go BOLD contrast than Controls in all 5 clusters . Across adolescence, Heavy Drinkers exhibited increasing response inhibition BOLD contrast, and Controls showed attenuated response in clusters. At follow-up, Heavy Drinkers showed significantly greater response inhibition activity than Controls in 4 brain regions : bilateral middle frontal gyri, right inferior parietal lobule, and left cerebellar tonsil. Exploratory post-hoc analyses examined whether BOLD response contrast change over time correlated with subsequent alcohol involvement in Heavy Drinkers . BOLD response contrast during no-go relative to go trials over time in the right middle frontal gyrus positively correlated with follow-up lifetime number of drinks . Follow-up hierarchical linear regressions revealed BOLD response contrast at baseline did not predict follow-up alcohol consumption after controlling for baseline alcohol, biological sex, and follow-up age at our conservative, corrected threshold . The present longitudinal neuroimaging study examined the effects of initiating heavy drinking during adolescence on brain responses during response inhibition. We hypothesized, based on previous findings , that adolescents who transition into heavy drinking would show reduced BOLD response during response inhibition prior to initiating heavy drinking followed by increased activation after the onset of heavy episodic drinking, as compared to adolescents who remained non-drinkers. Examining a longitudinal neuroimaging sample of youth both preand post-alcohol use initiation allowed us to address the etiology of neural pattern differences. Although group × time effect sizes were small, our findings suggest that differential neural activity patterns predate alcohol initiation and also arise as a consequence of heavy drinking.