Methods
We estimated associations between retrospective report of maternal cannabis use during pregnancy (only before maternal knowledge of pregnancy [BK-PCE], before and after maternal knowledge of pregnancy [BAK-PCE], and no exposure [NE]) and longitudinal assessments (baseline [June 1, 2016, to October 15, 2018], 1-year follow-up, and 2-year follow-up) of psychopathology (Child Behavior Checklist3 subscales, n = 20; total reported psychoticlike experiences on the Prodromal Questionnaire–Brief Child Version4). PCE groups had greater attrition (χ2 = 34.2, P < .001). Participants provided assent and caregivers provided written informed consent to a protocol approved by the institutional review board of each data collection site. We followed the STROBE reporting guideline for cohort studies.

Associations between PCE groups (BK-PCE, BAK-PCE, and NE) and child psychopathology were estimated using mixed models. In addition to main associations of exposure and age, interactions (ie, [age + age2] × [BK-PCE + BAK-PCE]) modeled age-associated change. χ2 Tests of log likelihood compared models with and without predictors of interest (ie, PCE group, PCE group × age interaction) to determine significance.

Covariates included family and child variables (Table). False discovery rate (FDR) multiple comparison correction was used (n = 42; exposure main associations and interactions with age). Secondary analyses tested whether associations were robust to the additional inclusion of pregnancy-associated covariates with high levels of missingness in the entire sample and to polygenic risk scores for cannabis use disorder and proximal outcomes of interest (eg, polygenic risk for schizophrenia, depression) in the European ancestry subsample (n = 5110; genetic methodological details available elsewhere2).

Results
A total of 391 individuals were in the BK-PCE group, 208 were in the BAK-PCE group, and 10 032 were in the NE group. Of those, 2379 (22%) self-reported as African American; 709 (7%), Asian/Asian American; 766 (7%), Hispanic; 378 (4%), Native American; 69 (0.6%), Pacific Islander; 8593 (81%), White; and 766 (7%), other. There were 10 631 individuals and 30 091 longitudinal assessments (baseline: n = 10 624; mean [IQR] age, 9.9 [8.9-11.1] years; 1-year follow-up: n = 10 094; mean [IQR] age, 10.9 [9.7-12.4] years; 2-year follow-up: n = 9373; mean [IQR] age, 12.0 [10.6-13.8] years). PCE was associated with persisting vulnerability to psychopathology throughout early adolescence (Figure and Table). These associations did not change with age (FDR-corrected P > .11). Significant findings were primarily driven by exposure following maternal knowledge of pregnancy (Figure and Table). Results remained FDR-significant when including covariates with high missingness (ie, pregnancy-associated covariates) with the exception of psychoticlike experiences (FDR-corrected P = .13; influential covariates were maternal age at birth and planned pregnancy). Associations remained directionally consistent and of similar magnitude in the BAK-PCE group after accounting for polygenic risk in the European ancestry subsample; 4 scales (ie, sluggish cognitive tempo, social problems, rule-breaking behavior, and Diagnostic and Statistical Manual of Mental Disorders [Fifth Edition] conduct problems) remained nominally significant.

Discussion
PCE is associated with persisting vulnerability to broad-spectrum psychopathology as children progress through early adolescence. Increased psychopathology may lead to greater risk for psychiatric disorders and problematic substance use as children enter peak periods of vulnerability in later adolescence.5 Larger associations in the BAK-PCE group may be attributable to the timing of cannabinoid receptor neural expression, which onsets in rodents at the equivalent of 5 to 6 weeks.6 Limitations include the small sample of prenatal cannabis-exposed offspring, potential underreporting of use during pregnancy, imprecise data on the timing/frequency/potency of cannabis exposure, and the lack of data on some potential confounders (eg, maternal stress during pregnancy). Evidence that the impact of PCE on psychopathology does not ameliorate as children enter adolescence further cautions against cannabis use during pregnancy.