Polygenic prediction of cannabis-related outcomes over time: Evidence from the ALSPAC longitudinal cohort and the EU-GEI case-control study

Polygenic prediction of cannabis-related outcomes over time: Evidence from the ALSPAC longitudinal cohort and the EU-GEI case-control study | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Polygenic prediction of cannabis-related outcomes over time: Evidence from the ALSPAC longitudinal cohort and the EU-GEI case-control study Edoardo Spinazzola, Isabelle Austin-Zimmerman, Giulia Trotta, and 14 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8359813/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Polygenic risk scores (PRSs) for cannabis use disorder (CUD), major depressive disorder (MDD), insomnia, and chronic pain might contribute to patterns of cannabis use, including problematic cannabis use. We examined PRSs for the above traits and different patterns of cannabis use across developmental stages using the ALSPAC longitudinal cohort (N = 8,224 participants with genotype data) and the EU-GEI case-control study (N = 994 [56.9%] controls and 752 [43.1%] first-episode psychosis cases with genotype data) for replication. We fitted regression models to test associations between PRSs and patterns of cannabis use at different ages. An interaction term was included to test whether the association between early cannabis initiation and heavy use is moderated by PRSs. Genetic liability to CUD, MDD, and pain was consistently associated with heavier use of cannabis. In the ALSPAC sample, linear regression models showed that CUD PRS was associated with CAST score at 20, and 24 years. MDD and pain PRSs were associated with CAST scores at 17, 20, and 24 years. In the EU-GEI, CUD PRS and pain PRS were associated with “weekly-to-daily” use of cannabis. Similarly, in the cases-only sample, CUD PRS and pain PRS were associated with weekly-to-daily use. In the controls-only sample, only CUD PRS was associated with weekly-to-daily use. This reflects an underlying vulnerability that may lead some people to use cannabis more heavily as a coping mechanism. However, age at initiation (ALSPAC = 7.1%, EU-GEI = 4.53%) explained a greater proportion of the variance of problematic cannabis use than polygenic liability, indicating atime-sensitiveintervention target. Psychiatry Polygenic risk score Cannabis use Gene x environment interaction ALSPAC EU-GEI Figures Figure 1 Figure 2 Figure 3 Figure 4 Introduction Lifetime cannabis use, cannabis use disorder, and psychotic disorders have shared genetic liability( 1 ), suggesting shared underlying biological mechanisms( 2 ). Previous studies have focused on three possible explanations for the association between cannabis and psychosis: ( 1 ) the association is entirely causal (i.e., heavy cannabis use is a risk factors for psychosis), ( 2 ) it is partly causal and partly confounded by shared genetic liability, or ( 3 ) it is completely noncausal and driven genetically by horizontal pleiotropy (i.e., heavy cannabis use does not cause psychosis, but genetic risk factors for SZ lead to cannabis use)( 3 ). We recently reported that SZ PRS was neither associated with nor interacted with the pattern of cannabis use; instead, SZ PRS and cannabis use acted independently in increasing risk for psychotic disorders( 4 ). Liability to depression, pain or insomnia/sleep may influence how much people use cannabis, whether they use it over developmental stages, and how early they initiate it. Therefore, investigating these liabilities, could bring some clarity to the underlying genetic underpinnings of cannabis use( 5 ). Both in the US( 6 ) and elsewhere( 7 ) medical cannabis use appears to be a major drive for frequent cannabis use, with the majority reporting to use to help with pain, sleep, anxiety, and depression( 6 ). Concerningly, those who report initiating cannabis use to get relief from either physical (i.e. ‘pain’) or psychological (i.e. ‘anxiety or depression’) discomfort were more likely to engage in harmful patterns of cannabis use and develop more psychopathology( 8 ). The advances in genome-wide association studies (GWASs) across medicine have made it possible to calculate polygenic risk profiles (PRSs) for schizophrenia (SZ)( 9 ), cannabis use disorder (CUD)( 10 ), major depressive disorder (MDD)( 11 ), insomnia( 12 , 13 ), and chronic pain( 14 ). Given the above, we used genetic data alongside detailed information on cannabis use from two independent cohorts—the Avon Longitudinal Study of Parents and Children (ALSPAC) and for replication, the European Network of National Schizophrenia Networks Studying Gene-Environment Interactions (EU-GEI) first-episode psychosis (FEP) patients-control sample. We examined the associations between PRSs for CUD, MDD, pain and insomnia and age at cannabis initiation and patterns of cannabis use from adolescence to adulthood and ( 2 ) tested whether the association between early cannabis initiation and heavy cannabis use is moderated by these PRSs. Methods We analysed data from two independent samples, the ALSPAC cohort and the EU-GEI case-control study. The former is a UK population-based cohort. Data for this study were collected through self-report questionnaires at ages 16 (lifetime-cannabis use only) and 20 and an in-person data collection clinic at the University of Bristol at ~ 17 and 24. ALSPAC participants were included if they had available genome-wide genotyping data. Data for this study were collected and managed utilising REDCap electronic data capture tools held at the University of Bristol. REDCap is a safe, web-based software platform designed to support data capture for research( 15 ). The ALSPAC Ethics and Law Committee and the local Research Ethics Committees provided ethical approval for the study. Participants provided written and informed consent to use the data collected following the recommendations of the ALSPAC Ethics and Law Committee at the time( 16 – 18 ). Participants can retrospectively withdraw their consent for their data to be used at any time by contacting the study team. Study participation is voluntary and during all data collection sweeps, information was given on the intended use of the data. For biological samples, consent was obtained according to the Human Tissue Act (2004). For questionnaires and clinic data, either on paper or online, consent was obtained following committee recommendations. Further information can be found at https://www.bristol.ac.uk/alspac/ . We also conducted a comparative analysis using data from the EU-GEI ( http://www.eu-gei.eu ), a multi-centre incidence and case-control study of genetic and environmental determinants of psychotic disorders. First-episode psychosis patients (FEPp) and population-based controls were recruited between 2010 and 2015 in 17 catchment areas in England, France, the Netherlands, Italy, Spain, and Brazil( 19 ). The study was approved by the local research ethics committee at each participating site. Participants ALSPAC Participants were eligible to be included in the ALSPAC study if the expected delivery for each pregnancy in the selected catchment area (Avon County) was between 1st April 1991 and 31st December 1992. More detailed recruitment information can be found in the ALSPAC study core paper( 16 ). The study was established to examine the genetic and environmental determinants of health and development (Bristol.ac.uk/alspac/about). Details on all the data is available in the study website through a fully searchable data dictionary and variable search tool ( http://www.bristol.ac.uk/alspac/researchers/our-data/ ). Phenotypic data for the present study were collected at ages 16, 17, 20, and 24. The present analyses are restricted to individuals who reported on cannabis use at these stages. EU-GEI Replication analyses were conducted using the EU-GEI sample. Participants were recruited across 17 sites in six European countries (England, the Netherlands, France, Spain, and Italy) and Brazil. FEP cases were included if they were ( 1 ) aged 18–64, ( 2 ) resident within the catchment area, and ( 3 ) presented with a clinical diagnosis for an untreated FEP. FEPp were excluded if they had ( 1 ) evidence of organic psychosis or ( 2 ) transient symptoms resulting from an acute intoxication. Controls were excluded if they had received a diagnosis and/or treatment for psychosis. In the EU-GEI, controls were selected to be as representative as possible of the general population. Further details on the study design and sampling strategy can be found in Gayer-Anderson et al, 2019( 19 ). Main outcome Measures of cannabis use In the ALSPAC sample, we employed the following measures of cannabis use. Firstly, a measure of lifetime cannabis (Yes or No) use was collected at 16, 17, and 24 years of age. Participants also reported at what time they started using cannabis. This latter measure was used to derive a variable of age at first cannabis use( 20 ). Our main analyses were performed using the Cannabis Abuse Screening Test (CAST)( 21 ) as it provided a reliable and validated measure of problematic cannabis use at different time points, 17, 20, and 24 years. The CAST score is a 6-item short screening tool that assesses the presence of past-year problematic cannabis use( 22 ). The use of a continuous variable (score 0 to 6) allowed us to distinguish variation in severity of cannabis use problems. To compare this data with the EU-GEI sample, we conducted additional sensitivity analyses using the frequency of cannabis use at the same ages. Additional information is available in the Supplementary Materials section. In the EU-GEI, data on cannabis use were collected using the Cannabis Experiences Questionnaire modified version for the EU-GEI (CEQ EUGEI )( 23 ). For this study, all measures of cannabis use were not taken at specific points, but rather, they indicated lifetime consumption. More precisely, we used the following measures: ( 1 ) lifetime cannabis use, ( 2 ) age at cannabis initiation, and ( 3 ) highest lifetime frequency of cannabis use. Regarding ( 3 ), to be as consistent as possible with the ALSPAC data, we created a composite measure of cannabis use frequency by merging the two highest consumption categories into a single ‘weekly-to-daily’ use category. We also merged the two lowest consumption categories (‘only once or twice’ and ‘a few times each year’) into a single category ‘a few times (each year)’. We therefore distinguished the following categories: ( 1 ) never used cannabis; ( 2 ) A few times (each year); ( 3 ) A few times (each month); ( 4 ) Weekly-to-daily use. In the supplements, we reported additional sensitivity analyses employing the original variable. Genotyping and PRS In ALSPAC, participants were genotyped using Illumina arrays, and imputation was performed with the Haplotype Reference Consortium (HRC) panel. Data from 8,224 genetically confirmed European-ancestry participants were analysed. PRS construction was done using PRS-CS and PLINK. Principal components were imputed independently to adjust for population stratification. In the EU-GEI, DNA samples were genotyped at the MRC Centre for Neuropsychiatric Genetics and Genomics, using the IlluminaCoreExome-24 BeadChip. Standard quality control procedures were applied as described in Quattrone et al( 24 ). Similar to the ALSPAC sample, analyses were limited to European individuals. This was due to the predominance of European ancestry in the EU-GEI sample, to optimise PRS validity and alignment with discovery GWAS. We used the most recent and largest publicly available GWAS summary to calculate PRSs for the following traits: cannabis use disorder (CUD), major depressive disorder (MDD), chronic pain (pain), and insomnia. We also calculated sensitivity PRSs, short sleep duration and long sleep duration PRSs. In the ALSPAC cohort, we included the schizophrenia (SZ) PRS as an additional sensitivity PRS. All PRSs were standardised (z-scores, mean = 0, SD = 1) and adjusted for the first 10 principal components. Statistical analyses All statistical analyses were performed using R Studio, version 4.2.1. Descriptive statistics were used to summarise participants’ characteristics in both cohorts. Analyses were conducted separately for the ALSPAC and the EUGEI samples. In the ALSPAC cohort, logistic regressions tested the association between each PRS and lifetime cannabis use at 16, 17, and 24 years. Linear regressions were conducted to test the association between each PRS and CAST score at each time point. We conducted multinomial logistic regressions adjusted for sex and PCs as sensitivity analyses to test the association between each PRS and frequency of cannabis use at 17, 20, and 24 years. To inform subsequent interaction models between PRSs, age at cannabis initiation, and CAST scores at 24 years, we first examined whether different PRSs were associated with age at first cannabis use. Age at cannabis initiation was used both as a dichotomous (≤ 15 years vs > 15 years) and a continuous measure (years) in separate regression models. Analyses were conducted in the genotyped subsample. Finally, we fitted multiplicative interaction terms to the logistic models to test if different PRSs modified the effect of age at first use on the levels of CAST score at 24 years. Sensitivity analyses were performed to test whether different PRSs modified the effect of age at first use on the probability of daily/almost daily cannabis use at 24 years. We tested additive interaction models with RERI (Relative Excess Risk due to Interaction) using a dichotomous CAST at 24 years outcome. All the above analyses were adjusted for sex and 10 principal components (PCs). In the EU-GEI cohort, analyses were conducted in four groups: controls only, cases only, combined cases and controls, and combined cases and controls with adjustment for case-control status. In addition to the fact that the control group was representative of the general population, this analytical approach helped minimise potential bias arising from the use of a case-control sample to investigate an outcome different from the one originally intended. Also, all analyses were adjusted for age, sex, site, and PCs. Binomial logistic regressions tested the association between each PRS and lifetime cannabis use. Multinomial logistic regressions tested the association between each PRS and frequency of cannabis use. Like for the ALSPAC cohort, we examined whether different PRSs were associated with age at first cannabis use. Age at cannabis initiation was used both as a dichotomous (≤ 15 years vs > 15 years) and a continuous measure in separate regression models. Binomial logistic regression models were run separately for each PRS. We then fitted multiplicative interaction terms to the logistic models to test if different PRSs modified the effect of age at first use on the probability of weekly-to-daily cannabis use. Finally, we tested additive interaction models with RERI (relative excess risk due to interaction) using a dichotomous weekly-to-daily use of cannabis outcome. In the ALSPAC sample, we fitted a series hierarchical linear regression model to quantify the variance in CAST score at age 24 explained by PRSs for CUD, SZ, MDD, and insomnia and pain. Each PRS was added to a base model including sex and the first 10 PCs, followed by age at cannabis initiation. Variance explained (R2 and adjusted R2) was plotted for each nested model. For the EU-GEI, we computed McFadden R-squared by a hierarchical model. Both our variance-explained models are reported in the Supplementary Materials section. We applied False Discovery Rate (FDR) correction (Benjamini-Hochberg) across all PRS-cannabis use tests within each cohort. We considered statistically significant results at FDR-adjusted p < 0.05. In the ALSPAC sample, missing data in outcomes (namely, all measures of cannabis use) were addressed through multiple imputation, consistent with previous ALSPAC studies( 20 , 25 ). We employed multiple imputation by chained equations (MICE) under a missing at random assumption. Cannabis outcomes were imputed via regression models (20 datasets). Regression results were pooled using Rubin’s rules. In the supplementary materials section, results are reported both in the imputed and non-imputed datasets with additional information on imputation methods. Results ALSPAC sample characteristics The primary analyses for this study were conducted using the ALSPAC cohort. Of 8,224 participants of European ancestry with genotyped data, 895 (10.9%) participants completed the Cannabis Abuse Screening Test (CAST) questionnaire at 17 years, 1,587 (19.3%) at 20 years, and 836 (10.2%) at 24 years. See Figure S1 for additional details on the sample ascertainment for the ALSPAC sample. See also Table 1 and Table S1 for detailed sample characteristics. Additional analyses to address attrition bias showed that the SZ PRS distribution was not associated with the missing data and the available sample for the CAST score at 17, 20, and 24 years and between CUD PRS and CAST score at 17 and 24 years. EU-GEI sample characteristics The total sample with available data on cannabis use comprised 994 (56.9%) controls and 752 (43.1%) FEP cases ( Table S1 ). All the analyses reported in the manuscript were conducted in the subsample of people of European descent with available genetic data ( Figure S2 ). Table 1 shows differences in sociodemographic factors and cannabis use measures across both working samples. [Insert Table 1 here] Polygenic Risk scores, and cannabis use over time ALSPAC: Using binomial logistic regression models, adjusted for sex and 10 principal components (PCs), we found that CUD, MDD, and pain PRSs were associated with lifetime cannabis use at 16, 17, and 24 years. In Table 2, we reported the results. Pairwise correlation coefficients between each PRS are reported in the supplementary materials ( Figure S3, S4 ). All coefficients were in the range of low to moderate correlation. Table 2 PANEL A: Output of the binomial logistic regressions between PRSs (both main and sensitivity PRSs), and lifetime cannabis use at 16, 17, and 24 years a in the ALSPAC sample; PANEL B: linear regressions between PRSs (both main and sensitivity PRSs), and CAST scores at 17, 20, and 24 years in the ALSPAC sample b PANEL A Lifetime use 16 years Lifetime use 17 years Lifetime use 24 years OR (95% CI); p FDR -value OR (95% CI); p FDR -value OR (95% CI); p FDR -value CUD PRS 1.25 (1.16–1.34); <0.001 1.19 (1.11–1.29); <0.001 1.15 (1.07–1.23); <0.001 MD PRS 1.2 (1.13–1.28); <0.001 1.2 (1.11–1.29); <0.001 1.19 (1.1–1.28); <0.001 Pain PRS 1.13 (1.07–1.2); <0.001 1.1 (1.03–1.18); 0.007 1.08 (1.00-1.18); 0.007 Insomnia PRS 1.01 (0.95–1.09); 0.7 1.01 (0.95–1.07); 0.78 1.02 (0.93–1.11); 0.74 Short sleep PRS 1.08 (1.01–1.15); 0.028 1.07 (1.00-1.15); 0.076 1.04 (0.96–1.12); 0.4 Long sleep PRS 1.07 (1.00-1.14); 0.06 1.04 (0.97–1.12); 0.28 1.03 (0.96–1.12); 0.4 SCZ PRS 1.22 (1.11–1.34); <0.001 1.18 (1.1–1.26); <0.001 1.12 (1.00-1.25); 0.13 PANEL B CAST score 17 years CAST score 20 years CAST score 24 years β (95% CI); p FDR -value β (95% CI); p FDR -value β (95% CI); p FDR -value CUD PRS 0.02 (-0.04, 0.08); 0.5 0.04 (0.01, 0.07); 0.03 0.11 (0.04, 0.18); 0.005 MD PRS 0.1 (0.04, 0.15); 0.002 0.06 (0.02, 0.1); 0.01 0.09 (0.02, 0.16); 0.01 Pain PRS 0.09 (0.04, 0.14); 0.002 0.07 (0.04, 0.1); <0.001 0.09 (0.03–0.14); 0.005 Insomnia PRS 0.04 (-0.02, 0.11); 0.23 0.02 (-0.02, 0.06); 0.28 0.04 (-0.02, 0.1); 0.2 Short sleep PRS 0.06 (0.01, 0.12); 0.08 0.03 (-0.00, 0.06); 0.17 0.04 (-0.03, 0.1); 0.4 Long sleep PRS 0.04 (-0.02, 0.1); 0.29 0.02 (-0.02, 0.06); 0.41 0.04 (-0.01, 0.1); 0.3 SCZ PRS 0.03 (-0.03, 0.08); 0.33 0.02 (-0.03, 0.06); 0.41 0.01 (-0.05, 0.07); 0.72 a Each model is adjusted for sex and genetic ancestry (PC1-PC10). Odds Ratios (OR), 95% confidence intervals (95% CI), and FDR-adjusted p-values (FDR p-values) are reported in the table. Significant results after the FDR-adjustment are reported significant results after FDR-adjustment. b Each model is adjusted for sex and genetic ancestry (PC1-PC10). Beta coeff. (β), 95% confidence intervals (95% CI), and FDR-adjusted p-values (FDR p-values) are reported in the table. In linear regressions, adjusted for sex and PCs we found that CUD PRS was associated with CAST score at 20 (β = 0.04, [95% CI 0.01–0.07] p FDR = 0.03), and 24 years (β = 0.11, [95% CI 0.04–0.18] p FDR = 0.005). MDD PRS was associated with CAST scores at 17 (β = 0.1, [95% CI 0.04–0.15] p FDR = 0.002), 20 years (β = 0.06, [95% CI 0.02–0.1] p FDR = 0.01), and 24 years (β = 0.09, [95% CI 0.02–0.16] p FDR = 0.01). Pain PRS was associated with CAST scores at 17 (β = 0.09, [95% CI 0.04–0.14] p FDR = 0.002), 20 years (β = 0.07, [95% CI 0.04–0.1] p FDR < 0.001), and 24 years (β = 0.09, [95% CI 0.03–0.14] p FDR = 0.005). See Table 2 (panel b). All our sensitivity PRSs, namely the short and long sleep duration PRSs, and the SZ PRS, were not associated with CAST scores at any time points. All the analyses were conducted in the imputed and in non-imputed datasets (Table 2, Tables S5-S8 ). Additional sensitivity analyses investigated the association between each PRS, and the frequency of cannabis use at 24 years (see Tables S9-S12 ). See also Table S26 for additional adjustment for CUD PRS in the SZ PRS model. We calculated adjusted R 2 statistics for each of the predictors of the CAST score at 24 years. A model including sex and 10 principal components (step 1) explained around 1% of the variance in the CAST score at 24 years. Adding CUD PRS (step 2) increased this to around 3%. When age at cannabis initiation was included (Step 3), the model predicted just under 10% of the variance. Alternative models with PRSs for MDD (step 2 R 2 = 2.38%), pain (step 2 R 2 = 2.41%), and insomnia (step 2 R 2 = 1.09%) explained lower variance. Age at cannabis initiation added the largest proportion of the variance in all models, and alone, it explained 7.1% of the variance (See Figure S7, S7b ). [Insert Table 2 here] In regression models, adjusted for sex, age, site, and 10 PCs, using data from EU-GEI, we found that CUD PRS was associated with lifetime cannabis use (OR 1.47, 95% CI 1.14–1.89), but not when adjusting for case-control status, or in cases or controls separately. MDD PRS was associated with lifetime cannabis use in cases and controls combined (OR 1.33, 95% CI 1.15–1.54), even when adjusted for case-control status (OR 1.28, 95% CI 1.1—1.48) and in the cases-only sample (OR 1.41, 95% CI 1.1–1.82), but not in the controls-only sample. Insomnia PRS was associated with lifetime cannabis use in cases only (OR 1.52, 95% CI 1.11–2.1). No PRS was associated with lifetime cannabis use in the controls-only sample ( Table S4 ). A multinomial regression model (with ‘never users’ as the baseline group), additionally adjusted for case-control status, found that CUD PRS (RRR 1.82, 95% CI 1.38–2.41; p FDR = 0.001) and pain PRS (RRR 1.27, 95% CI 1.11–1.46; p FDR = 0.007) were associated with “weekly-to-daily” use of cannabis. Similarly, in the cases-only sample, CUD PRS (RRR 1.84, 95% CI 1.24–2.74; p FDR = 0.01) and pain PRS (RRR 1.49, 95% CI 1.21–1.82; p FDR = 0.002) were associated with weekly-to-daily use. In the controls-only sample, only CUD PRS (RRR 1.93, 95% CI 1.27–2.94; p FDR = 0.013) was associated with weekly-to-daily use. See Fig. 1, Table S15 . See Tables S13 , S14 , and S16 for sensitivity analyses. In the EU-GEI, Step 1 (sex, site and 10 PCs) explained 4.56% of the variance. When age was included (step 2), this rose to 6.78%. Adding CUD PRS (step 3) increased this to 7.28% and adding age at cannabis initiation (step 4) increased this to 9.41%. Age at cannabis initiation explained 4.53% of the variance alone. The models with PRSs for MDD (Step 3 R 2 = 6.77%), pain (Step 3 R 2 = 6.89%), and insomnia (Step 3 R 2 = 6.8%) explained a similar but lower proportion of the variance. (See supplementary for models restricted to cases and controls separately). Figure S11-S11b . [Insert Fig. 1 here] The association between PRSs and age at cannabis initiation In the ALSPAC cohort, CUD (β=-0.26, [95% CI -0.4, -0.11]; p FDR 0.004), MDD (β=-0.16, [95% CI -0.3, -0.05]; p FDR = 0.01), and SZ PRSs (β=-0.18, [95% CI -0.3, -0.08]; p FDR 0.002) were consistently associated with earlier age at cannabis initiation in both the linear and the logistic models ( Table S18 ). By contrast, we found no clear evidence of association between the other PRSs and age at first cannabis use and therefore were not included in further analyses. In the EU-GEI sample, in the linear regression model, MDD (β = 0.42, SE: 0.16, [95% CI 0.1, 0.73]; p FDR = 0.02) and insomnia (β = 0.87, SE: 0.2, [95% CI 0.47, 1.27]; p FDR < 0.001) PRSs were associated with age at cannabis initiation ( Table S20, Table S19) . Independent and combined effects of CUD PRS and age at first use on problematic use In the ALSPAC cohort, we fitted negative binomial regression models to examine the relationship between CUD PRS and CAST scores at 24 years, adjusting for principal components and sex. Higher CUD PRS (IRR = 1.5, 95% CI [1.16, 1.93], p FDR = 0.004) and higher MDD PRS (IRR = 1.44, 95% CI [1.19, 1.75], p FDR < 0.001) were associated with increased CAST scores. Age at first cannabis use was associated with CAST scores at 24 years in all three models. Neither CUD PRS (IRR = 0.95, 95% CI [0.82, 1.11], p FDR = 1) or MDD PRS (IRR = 0.95, 95% CI [0.84, 1.08], p FDR = 1) moderate the relationship between early age at first use and CAST score. SZ PRS was not associated with CAST scores at 24 years nor moderated its relationship with age at first use ( Table S21, Figs. 2–4) . Results were consistent across the imputed and non-imputed datasets and in sensitivity analyses in which the outcome was the probability of ‘daily or almost daily use’ at 24 years ( Table S22 ). Secondary analyses, adjusted for sex and 10 PCs, suggested an additive interaction between CUD PRS and early age at first use (≤ 15 years) on CAST score at age 24, RERI 2.4 (95% CI [0.82, 3.99]). MDD and SZ PRSs did not show evidence of additive interaction. Table S23 . In the EU-GEI sample, logistic regressions, adjusted for sex, site, 10 PCs, and case-control status, were conducted to test for evidence of interaction between CUD and MDD PRSs x age at cannabis initiation on the outcome “weekly-to-daily” frequency of use category. No clear multiplicative or additive interactions between CUD or MDD PRSs and early cannabis initiation (≤ 15 years) were found in the EU-GEI sample, in the combined, cases-only, or controls-only group. Figures S17-S18 , Table S24 . [Insert Fig. 2 here] [Insert Fig. 3 here] [Insert Fig. 4 here] Discussion To our knowledge, this is the first study to find that genetic liability for different psychiatric and behavioural phenotypes is associated with problematic cannabis use across different ages. We also found that early initiation impacts cannabis use patterns. We selected the PRSs for CUD, MDD, pain, and insomnia (alongside the sensitivity PRSs for SZ, short and long sleep duration) to explore factors that could explain ( 1 ) self-medication and ( 2 ) medical cannabis use, both increasing worldwide( 6 , 26 ). Overall, two main novel findings emerge from our analyses: (i) an independent effect of age at first use and PRSs on later problematic cannabis use with earlier initiation, contributing to problematic use later in life. (ii) The influence of PRSs for CUD, pain, and MDD on cannabis use, suggesting that in some people self-medication might lead to heavier cannabis use. In ALSPAC, CUD, MDD, and pain PRSs were associated with lifetime cannabis use at all time points (16, 17, and 24 years). MDD and pain PRSs were also associated with problematic cannabis use (CAST scores) at all time points (17, 20, and 24 years). Interestingly, CUD PRS was associated with CAST scores at 20 and 24 years, but not at 17. SZ PRS was associated with lifetime use but not with CAST scores. Younger age at cannabis initiation (≤ 15 years) was associated with higher genetic risk for CUD, MDD, and SZ. Insomnia, short and long sleep duration PRSs had minimal/no association with all our outcomes (see Supplementary). In the EU-GEI sample, while we could not test for the developmental timing of patterns of cannabis use, we found that CUD PRS was associated with weekly-to-daily use of cannabis in the full sample and in cases and controls independently. Pain PRS was associated with weekly-to-daily use in the full sample and in cases only. MDD PRS was only associated with weekly-to-daily use in the full sample. Lastly, in both cohorts, we did not find evidence that either CUD or MDD PRSs moderated the relationship between age at first use and later problematic cannabis use. However, in ALSPAC, additive interaction was found between CUD PRS and early initiation (≤ 15 years). Although Figs. 2 and 3 show that those who started using cannabis at age ≤ 15, with high CUD and MDD PRS have a sharper increase in CAST score at age 24 compared with those who started at a later age, overall these results, combined with our variance explained models, suggest that age at cannabis initiation and polygenic liability for CUD and MDD act mostly independently from each other. Indeed, the former appears to play a more important role in the risk of developing heavy cannabis use later in life. These novel findings have important public health implications. Firstly, they suggest that PRSs for CUD, MDD, and pain are associated with more frequent and problematic cannabis use disorder. Therefore, these results could contribute to the early identification of individuals at higher risk of consumption. Earlier age at first use emerged as a robust, independent predictor of problematic cannabis use. Lastly, the lack of moderation effects across all analyses suggests that early cannabis use is a risk factor irrespective of the genetic loading we explored. Our as other studies found that the schizophrenia PRS is not associated with patterns of cannabis use nor with CUD( 4 , 27 , 28 ), thus speaking against the hypothesis that shared genetic liability explains the cannabis-psychosis association reported in epidemiology. Consistent with Austin-Zimmerman et al( 4 ), we found that, after controlling for CUD PRS, the associations between SZ PRS and some of the measures of cannabis use were either markedly reduced or no longer significant, again indicating a substantial confounding by CUD PRS ( Table S26 ). This less evident effect of SZ PRS on frequency of use might be explained by SZ GWAS including a substantial proportion of cases with both SZ and CUD, given the higher prevalence of cannabis use among people with schizophrenia than among controls( 29 ). Furthermore, the difference in outcome between lifetime cannabis use and CUD, have been confirmed by GWAS studies pointing at differences in their genetic architecture (Table 2 )( 1 ). Few studies have been conducted on the association between MDD, pain, and insomnia genetic liability with problematic cannabis use. MDD GWAS correlates with both cannabis initiation and with CUD GWAS( 1 ). In our analyses, we found an association between MDD PRS, and different cannabis use variables, which was consistent across datasets ( Tables S3 - S4 , S10 , S13 , S15 ). Liability to pain is genetically correlated with level of substance use and substance use disorders, including CUD( 30 ). Mendelian randomisation suggests a causal effect of chronic pain on CUD with important public health consequences( 10 ). Here, we provide evidence that genetic liability to chronic pain is also associated with some cannabis use phenotypes. Lastly, genomic structural equation modelling has shown that long sleep duration shares genetic liability with CUD, as well as other traits, suggesting that these traits may represent a common underlying genetic factor( 5 ). Our analyses were less consistent in showing such an association. An important strength of our study is the longitudinal nature of the large and well-phenotyped ALSPAC sample. Moreover, the replication on the multi-centre EUGEI sample adds generalisability. The richness of data from both cohorts and several sensitivity analyses have allowed to account for multiple confounding factors. Among the study limitations, we need to consider the potential recall bias due to the cannabis variables that in the EU-GEI were based on self-reported retrospective information. However, prior studies have shown that self-reported information on cannabis well-correlated with biological measures, the latter only being able to capture recent use rather than the cumulative harmful effect of cannabis overtime( 31 ). Secondly, the information on the type of cannabis used in the ALSPAC sample was only collected at 24 years( 20 ), preventing comparison of data on cannabis potency, across the two cohorts. In the EU-GEI, controls are representative of each site’s population at risk( 19 ). We adjusted for case-control status, besides conducting separate analyses in the cases and controls-only samples, which possibly reduced power as evidenced by wider confidence intervals. Additionally, our sensitivity analyses showed that SZ and CUD PRS scores did not differ between participants with observed vs missing CAST outcome. This suggests that our findings are unlikely to be explained by attrition bias or poor data quality ( Figures S8-S13 ). Because of the differences between our two samples, we could not make direct data replication. We conducted a series of sensitivity analyses employing measures like frequency of cannabis use common to both datasets. Lastly, our analyses were restricted to individuals of European ancestry which limits the generalisability of our findings. The available PRSs were built from the largest available GWASs, still predominantly from European descent, a limitation still affecting the whole field. Conclusions Although genetic predisposition contributes to problematic cannabis use later in life, our results suggest that early age at first use explains a greater proportion of the variance in problematic use, highlighting the importance of early detection and targeted intervention. Public health strategies and primary care intervention should aim at delaying first cannabis use, particularly during adolescence, when both neurodevelopmental processes and environmental influences may shape long-term outcomes. Furthermore, the association we found between genetic liability to CUD, MDD, and pain with heavier use of cannabis may reflect an underlying vulnerability in some of those who use cannabis to cope with physical or psychological discomfort, underscoring the importance of identifying self-medication as early as possible and offering monitored and tailored intervention strategies. Declarations Acknowledgments We express our gratitude to all the families who took part in this study, the midwives for their help in recruiting team, and the ALSPAC team, which includes staff involved in data collection, data and administration staff, technical managers and the technical staff working at the Bristol Bioresource Laboratory, based within the University of Bristol. The EU-GEI WP2 group collected or supervised the data collection. Funding The UK Medical Research Council and Wellcome (MR/Z505924/1) and the University of Bristol provide core support for ALSPAC. The EU-GEI project was funded by the European Community’s Seventh Framework Programme under grant agreement No. HEALTH-F2-2009-241909 (Project EU-GEI). The UK Medical Research Council (MRC) under the grant awarded to MDF reference MR/T007818/1 funded the salaries of IAZ, MDF, GT, and ZL. The Lord Leverhulme’s Charitable Trust and the Velvet Foundation funded the salary of ES. EV is supported by the National Institute for Health Research (NIHR) Maudsley Biomedical Research Centre at South London and Maudsley NHS Foundation Trust and King’s College London. Genome-wide genotyping data was obtained by Sample Logistics and Genotyping Facilities at Wellcome Sanger Institute and LabCorp (Laboratory Corporation of America) with support from 23 and Me. The views expressed are those of the authors and not necessarily those of the NHS, the NIHR, or the Department of Health and Social Care in the UK. Data availability statement The ALSPAC participants’ informed consent does not allow the data to be made available through any third party maintained public repository. Supporting data are available from ALSPAC upon request under the approved proposal number (number B4002). Full instructions for applying for data access can be found here: http://bristol.ac.uk/alspac/researchers/access/. The ALSPAC study website contains details of all available data (http://www.bristol.ac.uk/alspac/researchers/our-data/). References Johnson EC, Demontis D, Thorgeirsson TE, Walters RK, Polimanti R, Hatoum AS et al (2020) A large-scale genome-wide association study meta-analysis of cannabis use disorder. Lancet Psychiatry 7(12):1032–1045 Greco LA, Reay WR, Dayas CV, Cairns MJ (2022) Pairwise genetic meta-analyses between schizophrenia and substance dependence phenotypes reveals novel association signals with pharmacological significance. Transl Psychiatry 12(1):403 Gillespie NA, Kendler KS (2021) Use of genetically informed methods to clarify the nature of the association between cannabis use and risk for schizophrenia. JAMA Psychiatry 78(5):467–468 Austin-Zimmerman I, Spinazzola E, Quattrone D, Wu-Choi B, Trotta G, Li Z et al (2024) The impact of schizophrenia genetic load and heavy cannabis use on the risk of psychotic disorder in the EU-GEI case-control and UK Biobank studies. Psychol Med 54(15):4160–4172 Austin-Zimmerman I, Levey DF, Deak JD, Galimberti M, Adhikari K, Coleman JR et al (2025) Multi-ancestry genetic architecture of sleep duration and its relationship to other sleep and psychiatric phenotypes. medRxiv Leung J, Chan G, Stjepanović D, Chung JYC, Hall W, Hammond D (2022) Prevalence and self-reported reasons of cannabis use for medical purposes in USA and Canada. Psychopharmacology 239(5):1509–1519 Hallinan CM, Bonomo YA (2022) The rise and rise of medicinal cannabis, what now? Medicinal cannabis prescribing in Australia 2017–2022. Int J Environ Res Public Health 19(16):9853 Spinazzola E, Degen H, Austin-Zimmerman I, Trotta G, Chesney E, Li Z et al (2025) Are reasons for first using cannabis associated with subsequent cannabis consumption (standard THC units) and psychopathology? BMJ Mental Health. ;28(1) Trubetskoy V, Pardiñas AF, Qi T, Panagiotaropoulou G, Awasthi S, Bigdeli TB et al (2022) Mapping genomic loci implicates genes and synaptic biology in schizophrenia. Nature 604(7906):502–508 Levey DF, Galimberti M, Deak JD, Wendt FR, Bhattacharya A, Koller D et al (2023) Multi-ancestry genome-wide association study of cannabis use disorder yields insight into disease biology and public health implications. Nat Genet 55(12):2094–2103 Adams MJ, Streit F, Meng X, Awasthi S, Adey BN, Choi KW et al (2025) Trans-ancestry genome-wide study of depression identifies 697 associations implicating cell types and pharmacotherapies. Cell 188(3):640–652 e9 Lane JM, Jones SE, Dashti HS, Wood AR, Aragam KG, van Hees VT et al (2019) Biological and clinical insights from genetics of insomnia symptoms. Nat Genet 51(3):387–393 Jansen PR, Watanabe K, Stringer S, Skene N, Bryois J, Hammerschlag AR et al (2019) Genome-wide analysis of insomnia in 1,331,010 individuals identifies new risk loci and functional pathways. Nat Genet 51(3):394–403 Johnston KJ, Adams MJ, Nicholl BI, Ward J, Strawbridge RJ, Ferguson A et al (2019) Genome-wide association study of multisite chronic pain in UK Biobank. PLoS Genet 15(6):e1008164 Harris PA, Taylor R, Thielke R, Payne J, Gonzalez N, Conde JG (2009) Research electronic data capture (REDCap)—a metadata-driven methodology and workflow process for providing translational research informatics support. J Biomed Inf 42(2):377–381 Boyd A, Golding J, Macleod J, Lawlor DA, Fraser A, Henderson J et al (2013) Cohort profile: the ‘children of the 90s’—the index offspring of the Avon Longitudinal Study of Parents and Children. Int J Epidemiol 42(1):111–127 Fraser A, Macdonald-Wallis C, Tilling K, Boyd A, Golding J, Davey Smith G et al (2013) Cohort profile: the Avon Longitudinal Study of Parents and Children: ALSPAC mothers cohort. Int J Epidemiol 42(1):97–110 Northstone K, Lewcock M, Groom A, Boyd A, Macleod J, Timpson N et al (2019) The Avon Longitudinal Study of Parents and Children (ALSPAC): an update on the enrolled sample of index children in 2019. Wellcome Open Res 4:51 Gayer-Anderson C, Jongsma HE, Di Forti M, Quattrone D, Velthorst E, De Haan L et al (2020) The European network of national schizophrenia networks studying gene–environment interactions (EU-GEI): Incidence and first-episode case–control programme. Soc Psychiatry Psychiatr Epidemiol 55(5):645–657 Hines LA, Freeman TP, Gage SH, Zammit S, Hickman M, Cannon M et al (2020) Association of high-potency cannabis use with mental health and substance use in adolescence. JAMA psychiatry 77(10):1044–1051 Legleye S, Karila L, Beck F, Reynaud M (2007) Validation of the CAST, a general population Cannabis Abuse Screening Test. J Subst Use 12(4):233–242 Lees Thorne R, Hines LA, Burke C, Jones HJ, Freeman TP (2025) Association of childhood mental health and cognition with longitudinal patterns of cannabis problems in adolescence. Psychol Med 55:e129 Di Forti M, Quattrone D, Freeman TP, Tripoli G, Gayer-Anderson C, Quigley H et al (2019) The contribution of cannabis use to variation in the incidence of psychotic disorder across Europe (EU-GEI): a multicentre case-control study. Lancet Psychiatry 6(5):427–436 Quattrone D, Reininghaus U, Richards AL, Tripoli G, Ferraro L, Quattrone A et al (2021) The continuity of effect of schizophrenia polygenic risk score and patterns of cannabis use on transdiagnostic symptom dimensions at first-episode psychosis: findings from the EU-GEI study. Transl Psychiatry 11(1):423 Hines LA, Heron J, Zammit S (2024) Incident psychotic experiences following self-reported use of high‐potency cannabis: Results from a longitudinal cohort study. Addiction 119(9):1629–1634 Dawson D, Chan G, Stjepanović D, Lorenzetti V, Hall WD, Leung J (2025) The prevalence and correlates of self-reported cannabis use for medicinal, dual and recreational motives in Australia: Findings from the National Drug Strategy Household Survey 2022/2023. Drug Alcohol Rev Hjorthøj C, Uddin MJ, Wimberley T, Dalsgaard S, Hougaard DM, Børglum A et al (2021) No evidence of associations between genetic liability for schizophrenia and development of cannabis use disorder. Psychol Med 51(3):479–484 Guloksuz S, Pries LK, Delespaul P, Kenis G, Luykx JJ, Lin BD et al (2019) Examining the independent and joint effects of molecular genetic liability and environmental exposures in schizophrenia: results from the EUGEI study. World Psychiatry 18(2):173–182 Johnson EC, Austin-Zimmerman I, Thorpe HH, Levey DF, Baranger DA, Colbert SM et al (2024) Cross-ancestry genetic investigation of schizophrenia, cannabis use disorder, and tobacco smoking. Neuropsychopharmacol 49(11):1655–1665 Toikumo S, Vickers-Smith R, Jinwala Z, Xu H, Saini D, Hartwell E et al (2023) The genetic architecture of pain intensity in a sample of 598,339 US veterans. medRxiv Bharat C, Webb P, Wilkinson Z, McKetin R, Grebely J, Farrell M et al (2023) Agreement between self-reported illicit drug use and biological samples: a systematic review and meta‐analysis. Addiction 118(9):1624–1648 Table 1 Table 1 is available in the Supplementary Files section. Additional Declarations The authors declare no competing interests. Supplementary Files SUPPLEMENTARYMATERIALSTOACCOMPANY.docx Table1.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8359813","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":560145163,"identity":"57cabc56-1d58-410e-916b-019acf86d5ac","order_by":0,"name":"Edoardo Spinazzola","email":"data:image/png;base64,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","orcid":"https://orcid.org/0000-0002-8070-0618","institution":"King's College London","correspondingAuthor":true,"prefix":"","firstName":"Edoardo","middleName":"","lastName":"Spinazzola","suffix":""},{"id":560147348,"identity":"e4a37904-b120-47ec-9b15-a61cf2c14eaf","order_by":1,"name":"Isabelle Austin-Zimmerman","email":"","orcid":"","institution":"King's College London","correspondingAuthor":false,"prefix":"","firstName":"Isabelle","middleName":"","lastName":"Austin-Zimmerman","suffix":""},{"id":560147349,"identity":"f9b0e516-2306-40bd-926a-fa6e512a45de","order_by":2,"name":"Giulia Trotta","email":"","orcid":"","institution":"King's College London","correspondingAuthor":false,"prefix":"","firstName":"Giulia","middleName":"","lastName":"Trotta","suffix":""},{"id":560147350,"identity":"1ab5efe9-b0dd-413b-85c8-bb481583e5da","order_by":3,"name":"Zhikun Li","email":"","orcid":"","institution":"King's College London","correspondingAuthor":false,"prefix":"","firstName":"Zhikun","middleName":"","lastName":"Li","suffix":""},{"id":560147351,"identity":"792ec4e1-be2f-4104-8b65-0986ddc764d2","order_by":4,"name":"Perry BM Leung","email":"","orcid":"","institution":"King's College London","correspondingAuthor":false,"prefix":"","firstName":"Perry","middleName":"BM","lastName":"Leung","suffix":""},{"id":560147352,"identity":"bb90955c-fa4a-4a2d-b1c8-da6f2408adb4","order_by":5,"name":"Qiang Li","email":"","orcid":"","institution":"King's College London","correspondingAuthor":false,"prefix":"","firstName":"Qiang","middleName":"","lastName":"Li","suffix":""},{"id":560147353,"identity":"63e6b126-204b-4471-8791-e3b1e60b119a","order_by":6,"name":"Luis Alameda","email":"","orcid":"","institution":"King's College London","correspondingAuthor":false,"prefix":"","firstName":"Luis","middleName":"","lastName":"Alameda","suffix":""},{"id":560147354,"identity":"66dfc37a-c07a-4059-91a8-98b48d18e9d1","order_by":7,"name":"Celso Arango","email":"","orcid":"","institution":"Hospital Universitario La Paz, IdiPAZ, School of Medicine Universidad Autónoma de Madrid","correspondingAuthor":false,"prefix":"","firstName":"Celso","middleName":"","lastName":"Arango","suffix":""},{"id":560147355,"identity":"75ebdb46-fc8b-46ba-92fb-27cbda2cc642","order_by":8,"name":"Michael O'Donovan","email":"","orcid":"","institution":"Centre for Neuropsychiatric Genetics and Genomics, Division of Psychological Medicine and Clinical Neurosciences","correspondingAuthor":false,"prefix":"","firstName":"Michael","middleName":"","lastName":"O'Donovan","suffix":""},{"id":560147356,"identity":"d30a822b-9c4a-49ce-83c5-335831f165cb","order_by":9,"name":"Jim van Os","email":"","orcid":"","institution":"Department of Psychiatry, University Medical Center Utrecht Brain Center, University Medical Center Utrecht","correspondingAuthor":false,"prefix":"","firstName":"Jim","middleName":"van","lastName":"Os","suffix":""},{"id":560147357,"identity":"8b126b81-ae2a-49c0-bd5a-2d8dbcdf9481","order_by":10,"name":"Pak C Sham","email":"","orcid":"","institution":"Department of Psychiatry, LKS Faculty of Medicine, The University of Hong Kong","correspondingAuthor":false,"prefix":"","firstName":"Pak","middleName":"C","lastName":"Sham","suffix":""},{"id":560147358,"identity":"1344ed23-0bd3-4080-a889-c501f6fdda0f","order_by":11,"name":"Craig Morgan","email":"","orcid":"","institution":"King's College London","correspondingAuthor":false,"prefix":"","firstName":"Craig","middleName":"","lastName":"Morgan","suffix":""},{"id":560147359,"identity":"3dfef935-6cac-4b49-bf67-66b3ea423aa0","order_by":12,"name":"Evangelos Vassos","email":"","orcid":"","institution":"King's College London","correspondingAuthor":false,"prefix":"","firstName":"Evangelos","middleName":"","lastName":"Vassos","suffix":""},{"id":560147360,"identity":"6e852e9a-1204-4780-81f5-48876d168509","order_by":13,"name":"Diego Quattrone","email":"","orcid":"","institution":"King's College London","correspondingAuthor":false,"prefix":"","firstName":"Diego","middleName":"","lastName":"Quattrone","suffix":""},{"id":560147361,"identity":"1e5ffab3-a40d-4aa6-a6c0-b9421e312110","order_by":14,"name":"Tom Freeman","email":"","orcid":"","institution":"Psychology, University of Bath","correspondingAuthor":false,"prefix":"","firstName":"Tom","middleName":"","lastName":"Freeman","suffix":""},{"id":560147362,"identity":"6d0b3efc-28b7-49d4-908a-3a3072bee699","order_by":15,"name":"Marta Di Forti","email":"","orcid":"","institution":"King's College London","correspondingAuthor":false,"prefix":"","firstName":"Marta","middleName":"Di","lastName":"Forti","suffix":""},{"id":560147363,"identity":"6714663a-0878-48c4-bc87-ffa3af418b27","order_by":16,"name":"Robin M Murray","email":"","orcid":"","institution":"King's College London","correspondingAuthor":false,"prefix":"","firstName":"Robin","middleName":"M","lastName":"Murray","suffix":""}],"badges":[],"createdAt":"2025-12-14 18:35:04","currentVersionCode":1,"declarations":{"humanSubjects":true,"vertebrateSubjects":false,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":true,"humanSubjectConsent":true,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-8359813/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8359813/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":98434309,"identity":"3e16840a-e9c4-4086-9aaf-9fe55635ecec","added_by":"auto","created_at":"2025-12-17 16:51:51","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":87178,"visible":true,"origin":"","legend":"","description":"","filename":"ManuscriptPRS.docx","url":"https://assets-eu.researchsquare.com/files/rs-8359813/v1/36c949a1b8f770a858451e86.docx"},{"id":98293991,"identity":"a6af6f28-1ce4-48a0-a4ab-9aa19132d081","added_by":"auto","created_at":"2025-12-16 09:04:29","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":342,"visible":true,"origin":"","legend":"","description":"","filename":"rs8359813.json","url":"https://assets-eu.researchsquare.com/files/rs-8359813/v1/45d15de72d20a6ed02fb3570.json"},{"id":98293998,"identity":"4b87dd28-e147-469d-bc5a-b1c68ac8a4d9","added_by":"auto","created_at":"2025-12-16 09:04:30","extension":"xml","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":126854,"visible":true,"origin":"","legend":"","description":"","filename":"rs83598130enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-8359813/v1/b214451b3681be01a874cff7.xml"},{"id":98293996,"identity":"63d35026-09d0-4227-95ed-ae76ad3e5b6d","added_by":"auto","created_at":"2025-12-16 09:04:29","extension":"xml","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":121897,"visible":true,"origin":"","legend":"","description":"","filename":"rs83598130structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-8359813/v1/55b036ac7a596adb289bfec4.xml"},{"id":98294003,"identity":"49d36d56-016d-4e00-a2dd-c0046880ee37","added_by":"auto","created_at":"2025-12-16 09:04:30","extension":"html","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":141802,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-8359813/v1/38e34e97b07ac13875e90a6c.html"},{"id":98293986,"identity":"9ac79c1f-5f96-4ee6-9fb9-0e0832740ec9","added_by":"auto","created_at":"2025-12-16 09:04:29","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":1392756,"visible":true,"origin":"","legend":"\u003cp\u003eRegression plot displaying the outcome of the multinomial regressions to test the associations between MAIN PRSs and frequency of cannabis use (simplified 4-categories variable) in the EU-GEI sample\u003c/p\u003e","description":"","filename":"Picture1.png","url":"https://assets-eu.researchsquare.com/files/rs-8359813/v1/eb93563769f29cedeac0b476.png"},{"id":98293994,"identity":"84f69a72-d0ae-449f-97fc-ef9bfd4b18be","added_by":"auto","created_at":"2025-12-16 09:04:29","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":991531,"visible":true,"origin":"","legend":"\u003cp\u003eInteraction CUD PRS x age at cannabis initiation.\u003c/p\u003e","description":"","filename":"Picture2.png","url":"https://assets-eu.researchsquare.com/files/rs-8359813/v1/35283e71fd18b9c4212c5b58.png"},{"id":98294001,"identity":"a6500668-faef-4b64-8b49-c8cb68b8f353","added_by":"auto","created_at":"2025-12-16 09:04:30","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":1005082,"visible":true,"origin":"","legend":"\u003cp\u003eALSPAC: Interaction MDD PRS x age at cannabis initiation.\u003c/p\u003e","description":"","filename":"Picture3.png","url":"https://assets-eu.researchsquare.com/files/rs-8359813/v1/031c5678b3fb74f447c7f843.png"},{"id":98294000,"identity":"53a9a21a-3d46-4a63-8b73-e202e798011b","added_by":"auto","created_at":"2025-12-16 09:04:30","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":660563,"visible":true,"origin":"","legend":"\u003cp\u003eALSPAC: Interaction SZ PRS x age at cannabis initiation.\u003c/p\u003e","description":"","filename":"Picture4.png","url":"https://assets-eu.researchsquare.com/files/rs-8359813/v1/bb70506cdc55b40bfc9cb0c7.png"},{"id":98445322,"identity":"bd05d6eb-9ec4-4b75-962d-7a160f016137","added_by":"auto","created_at":"2025-12-17 17:19:31","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":4606589,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8359813/v1/9b5a66a5-dd61-4d73-87c3-fd8b13c1c06c.pdf"},{"id":98435971,"identity":"6609372a-6872-486c-90e5-9d14e661e108","added_by":"auto","created_at":"2025-12-17 16:54:42","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":2807253,"visible":true,"origin":"","legend":"","description":"","filename":"SUPPLEMENTARYMATERIALSTOACCOMPANY.docx","url":"https://assets-eu.researchsquare.com/files/rs-8359813/v1/42e02bd36f69a6c8012202ac.docx"},{"id":98435794,"identity":"212ba6b9-899f-46ee-9372-ea3a024e74f0","added_by":"auto","created_at":"2025-12-17 16:54:26","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":18270,"visible":true,"origin":"","legend":"","description":"","filename":"Table1.docx","url":"https://assets-eu.researchsquare.com/files/rs-8359813/v1/d216634eea151770d17eb108.docx"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003ePolygenic prediction of cannabis-related outcomes over time: Evidence from the ALSPAC longitudinal cohort and the EU-GEI case-control study\u003c/p\u003e","fulltext":[{"header":"Introduction","content":"\u003cp\u003eLifetime cannabis use, cannabis use disorder, and psychotic disorders have shared genetic liability(\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e), suggesting shared underlying biological mechanisms(\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e). Previous studies have focused on three possible explanations for the association between cannabis and psychosis: (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e) the association is entirely causal (i.e., heavy cannabis use is a risk factors for psychosis), (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e) it is partly causal and partly confounded by shared genetic liability, or (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e) it is completely noncausal and driven genetically by horizontal pleiotropy (i.e., heavy cannabis use does not cause psychosis, but genetic risk factors for SZ lead to cannabis use)(\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e). We recently reported that SZ PRS was neither associated with nor interacted with the pattern of cannabis use; instead, SZ PRS and cannabis use acted independently in increasing risk for psychotic disorders(\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eLiability to depression, pain or insomnia/sleep may influence how much people use cannabis, whether they use it over developmental stages, and how early they initiate it. Therefore, investigating these liabilities, could bring some clarity to the underlying genetic underpinnings of cannabis use(\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e). Both in the US(\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e) and elsewhere(\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e) medical cannabis use appears to be a major drive for frequent cannabis use, with the majority reporting to use to help with pain, sleep, anxiety, and depression(\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e). Concerningly, those who report initiating cannabis use to get relief from either physical (i.e. ‘pain’) or psychological (i.e. ‘anxiety or depression’) discomfort were more likely to engage in harmful patterns of cannabis use and develop more psychopathology(\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe advances in genome-wide association studies (GWASs) across medicine have made it possible to calculate polygenic risk profiles (PRSs) for schizophrenia (SZ)(\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e), cannabis use disorder (CUD)(\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e), major depressive disorder (MDD)(\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e), insomnia(\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e), and chronic pain(\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eGiven the above, we used genetic data alongside detailed information on cannabis use from two independent cohorts—the Avon Longitudinal Study of Parents and Children (ALSPAC) and for replication, the European Network of National Schizophrenia Networks Studying Gene-Environment Interactions (EU-GEI) first-episode psychosis (FEP) patients-control sample. We examined the associations between PRSs for CUD, MDD, pain and insomnia and age at cannabis initiation and patterns of cannabis use from adolescence to adulthood and (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e) tested whether the association between early cannabis initiation and heavy cannabis use is moderated by these PRSs.\u003c/p\u003e\n\n \u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003cdiv id=\"Sec4\" class=\"Section3\"\u003e \u003c/div\u003e \u003c/div\u003e\n\n\n\n \n\n "},{"header":"Methods","content":"\u003cp\u003eWe analysed data from two independent samples, the ALSPAC cohort and the EU-GEI case-control study. The former is a UK population-based cohort. Data for this study were collected through self-report questionnaires at ages 16 (lifetime-cannabis use only) and 20 and an in-person data collection clinic at the University of Bristol at ~ 17 and 24. ALSPAC participants were included if they had available genome-wide genotyping data. Data for this study were collected and managed utilising REDCap electronic data capture tools held at the University of Bristol. REDCap is a safe, web-based software platform designed to support data capture for research(\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e). The ALSPAC Ethics and Law Committee and the local Research Ethics Committees provided ethical approval for the study. Participants provided written and informed consent to use the data collected following the recommendations of the ALSPAC Ethics and Law Committee at the time(\u003cspan additionalcitationids=\"CR17\" citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e–\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e). Participants can retrospectively withdraw their consent for their data to be used at any time by contacting the study team. Study participation is voluntary and during all data collection sweeps, information was given on the intended use of the data. For biological samples, consent was obtained according to the Human Tissue Act (2004). For questionnaires and clinic data, either on paper or online, consent was obtained following committee recommendations. Further information can be found at \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.bristol.ac.uk/alspac/\u003c/span\u003e\u003cspan address=\"https://www.bristol.ac.uk/alspac/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/p\u003e\u003cp\u003eWe also conducted a comparative analysis using data from the EU-GEI (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://www.eu-gei.eu\u003c/span\u003e\u003cspan address=\"http://www.eu-gei.eu\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e), a multi-centre incidence and case-control study of genetic and environmental determinants of psychotic disorders. First-episode psychosis patients (FEPp) and population-based controls were recruited between 2010 and 2015 in 17 catchment areas in England, France, the Netherlands, Italy, Spain, and Brazil(\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e). The study was approved by the local research ethics committee at each participating site.\u003c/p\u003e\u003ch2\u003eParticipants\u003c/h2\u003e\u003ch2\u003eALSPAC\u003c/h2\u003e\u003cp\u003eParticipants were eligible to be included in the ALSPAC study if the expected delivery for each pregnancy in the selected catchment area (Avon County) was between 1st April 1991 and 31st December 1992. More detailed recruitment information can be found in the ALSPAC study core paper(\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e). The study was established to examine the genetic and environmental determinants of health and development (Bristol.ac.uk/alspac/about). Details on all the data is available in the study website through a fully searchable data dictionary and variable search tool (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://www.bristol.ac.uk/alspac/researchers/our-data/\u003c/span\u003e\u003cspan address=\"http://www.bristol.ac.uk/alspac/researchers/our-data/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). Phenotypic data for the present study were collected at ages 16, 17, 20, and 24. The present analyses are restricted to individuals who reported on cannabis use at these stages.\u003c/p\u003e\u003ch3\u003eEU-GEI\u003c/h3\u003e\u003cp\u003eReplication analyses were conducted using the EU-GEI sample. Participants were recruited across 17 sites in six European countries (England, the Netherlands, France, Spain, and Italy) and Brazil. FEP cases were included if they were (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e) aged 18–64, (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e) resident within the catchment area, and (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e) presented with a clinical diagnosis for an untreated FEP. FEPp were excluded if they had (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e) evidence of organic psychosis or (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e) transient symptoms resulting from an acute intoxication. Controls were excluded if they had received a diagnosis and/or treatment for psychosis. In the EU-GEI, controls were selected to be as representative as possible of the general population. Further details on the study design and sampling strategy can be found in Gayer-Anderson et al, 2019(\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e).\u003c/p\u003e\u003ch3\u003eMain outcome\u003c/h3\u003e\u003ch2\u003eMeasures of cannabis use\u003c/h2\u003e\u003cp\u003eIn the ALSPAC sample, we employed the following measures of cannabis use. Firstly, a measure of lifetime cannabis (Yes or No) use was collected at 16, 17, and 24 years of age. Participants also reported at what time they started using cannabis. This latter measure was used to derive a variable of age at first cannabis use(\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e). Our main analyses were performed using the Cannabis Abuse Screening Test (CAST)(\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e) as it provided a reliable and validated measure of problematic cannabis use at different time points, 17, 20, and 24 years. The CAST score is a 6-item short screening tool that assesses the presence of past-year problematic cannabis use(\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e). The use of a continuous variable (score 0 to 6) allowed us to distinguish variation in severity of cannabis use problems. To compare this data with the EU-GEI sample, we conducted additional sensitivity analyses using the frequency of cannabis use at the same ages. Additional information is available in the Supplementary Materials section.\u003c/p\u003e\u003cp\u003eIn the EU-GEI, data on cannabis use were collected using the Cannabis Experiences Questionnaire modified version for the EU-GEI (CEQ\u003csub\u003eEUGEI\u003c/sub\u003e)(\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e). For this study, all measures of cannabis use were not taken at specific points, but rather, they indicated lifetime consumption. More precisely, we used the following measures: (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e) lifetime cannabis use, (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e) age at cannabis initiation, and (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e) highest lifetime frequency of cannabis use. Regarding (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e), to be as consistent as possible with the ALSPAC data, we created a composite measure of cannabis use frequency by merging the two highest consumption categories into a single ‘weekly-to-daily’ use category. We also merged the two lowest consumption categories (‘only once or twice’ and ‘a few times each year’) into a single category ‘a few times (each year)’. We therefore distinguished the following categories: (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e) never used cannabis; (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e) A few times (each year); (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e) A few times (each month); (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e) Weekly-to-daily use. In the supplements, we reported additional sensitivity analyses employing the original variable.\u003c/p\u003e\u003ch2\u003eGenotyping and PRS\u003c/h2\u003e\u003cp\u003eIn ALSPAC, participants were genotyped using Illumina arrays, and imputation was performed with the Haplotype Reference Consortium (HRC) panel. Data from 8,224 genetically confirmed European-ancestry participants were analysed. PRS construction was done using PRS-CS and PLINK. Principal components were imputed independently to adjust for population stratification.\u003c/p\u003e\u003cp\u003eIn the EU-GEI, DNA samples were genotyped at the MRC Centre for Neuropsychiatric Genetics and Genomics, using the IlluminaCoreExome-24 BeadChip. Standard quality control procedures were applied as described in Quattrone et al(\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e). Similar to the ALSPAC sample, analyses were limited to European individuals. This was due to the predominance of European ancestry in the EU-GEI sample, to optimise PRS validity and alignment with discovery GWAS.\u003c/p\u003e\u003cp\u003eWe used the most recent and largest publicly available GWAS summary to calculate PRSs for the following traits: cannabis use disorder (CUD), major depressive disorder (MDD), chronic pain (pain), and insomnia. We also calculated sensitivity PRSs, short sleep duration and long sleep duration PRSs. In the ALSPAC cohort, we included the schizophrenia (SZ) PRS as an additional sensitivity PRS. All PRSs were standardised (z-scores, mean = 0, SD = 1) and adjusted for the first 10 principal components.\u003c/p\u003e\u003ch3\u003eStatistical analyses\u003c/h3\u003e\u003cp\u003eAll statistical analyses were performed using R Studio, version 4.2.1. Descriptive statistics were used to summarise participants’ characteristics in both cohorts. Analyses were conducted separately for the ALSPAC and the EUGEI samples.\u003c/p\u003e\u003cp\u003eIn the ALSPAC cohort, logistic regressions tested the association between each PRS and lifetime cannabis use at 16, 17, and 24 years. Linear regressions were conducted to test the association between each PRS and CAST score at each time point. We conducted multinomial logistic regressions adjusted for sex and PCs as sensitivity analyses to test the association between each PRS and frequency of cannabis use at 17, 20, and 24 years.\u003c/p\u003e\u003cp\u003eTo inform subsequent interaction models between PRSs, age at cannabis initiation, and CAST scores at 24 years, we first examined whether different PRSs were associated with age at first cannabis use. Age at cannabis initiation was used both as a dichotomous (≤ 15 years vs \u0026gt; 15 years) and a continuous measure (years) in separate regression models. Analyses were conducted in the genotyped subsample. Finally, we fitted multiplicative interaction terms to the logistic models to test if different PRSs modified the effect of age at first use on the levels of CAST score at 24 years. Sensitivity analyses were performed to test whether different PRSs modified the effect of age at first use on the probability of daily/almost daily cannabis use at 24 years. We tested additive interaction models with RERI (Relative Excess Risk due to Interaction) using a dichotomous CAST at 24 years outcome. All the above analyses were adjusted for sex and 10 principal components (PCs).\u003c/p\u003e\u003cp\u003eIn the EU-GEI cohort, analyses were conducted in four groups: controls only, cases only, combined cases and controls, and combined cases and controls with adjustment for case-control status. In addition to the fact that the control group was representative of the general population, this analytical approach helped minimise potential bias arising from the use of a case-control sample to investigate an outcome different from the one originally intended. Also, all analyses were adjusted for age, sex, site, and PCs. Binomial logistic regressions tested the association between each PRS and lifetime cannabis use. Multinomial logistic regressions tested the association between each PRS and frequency of cannabis use.\u003c/p\u003e\u003cp\u003eLike for the ALSPAC cohort, we examined whether different PRSs were associated with age at first cannabis use. Age at cannabis initiation was used both as a dichotomous (≤ 15 years vs \u0026gt; 15 years) and a continuous measure in separate regression models. Binomial logistic regression models were run separately for each PRS. We then fitted multiplicative interaction terms to the logistic models to test if different PRSs modified the effect of age at first use on the probability of weekly-to-daily cannabis use. Finally, we tested additive interaction models with RERI (relative excess risk due to interaction) using a dichotomous weekly-to-daily use of cannabis outcome.\u003c/p\u003e\u003cp\u003eIn the ALSPAC sample, we fitted a series hierarchical linear regression model to quantify the variance in CAST score at age 24 explained by PRSs for CUD, SZ, MDD, and insomnia and pain. Each PRS was added to a base model including sex and the first 10 PCs, followed by age at cannabis initiation. Variance explained (R2 and adjusted R2) was plotted for each nested model. For the EU-GEI, we computed McFadden R-squared by a hierarchical model. Both our variance-explained models are reported in the Supplementary Materials section.\u003c/p\u003e\u003cp\u003eWe applied False Discovery Rate (FDR) correction (Benjamini-Hochberg) across all PRS-cannabis use tests within each cohort. We considered statistically significant results at FDR-adjusted p \u0026lt; 0.05.\u003c/p\u003e\u003cp\u003eIn the ALSPAC sample, missing data in outcomes (namely, all measures of cannabis use) were addressed through multiple imputation, consistent with previous ALSPAC studies(\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e, \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e). We employed multiple imputation by chained equations (MICE) under a missing at random assumption. Cannabis outcomes were imputed via regression models (20 datasets). Regression results were pooled using Rubin’s rules. In the supplementary materials section, results are reported both in the imputed and non-imputed datasets with additional information on imputation methods.\u003c/p\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec11\"\u003e\n \u003ch2\u003eALSPAC sample characteristics\u003c/h2\u003e\n \u003cp\u003eThe primary analyses for this study were conducted using the ALSPAC cohort. Of 8,224 participants of European ancestry with genotyped data, 895 (10.9%) participants completed the Cannabis Abuse Screening Test (CAST) questionnaire at 17 years, 1,587 (19.3%) at 20 years, and 836 (10.2%) at 24 years. See \u003cstrong\u003eFigure S1\u003c/strong\u003e for additional details on the sample ascertainment for the ALSPAC sample. See also Table 1 and \u003cstrong\u003eTable S1\u003c/strong\u003e for detailed sample characteristics.\u003c/p\u003e\n \u003cp\u003eAdditional analyses to address attrition bias showed that the SZ PRS distribution was not associated with the missing data and the available sample for the CAST score at 17, 20, and 24 years and between CUD PRS and CAST score at 17 and 24 years.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec12\"\u003e\n \u003ch2\u003eEU-GEI sample characteristics\u003c/h2\u003e\n \u003cp\u003eThe total sample with available data on cannabis use comprised 994 (56.9%) controls and 752 (43.1%) FEP cases (\u003cstrong\u003eTable S1\u003c/strong\u003e). All the analyses reported in the manuscript were conducted in the subsample of people of European descent with available genetic data (\u003cstrong\u003eFigure S2\u003c/strong\u003e). Table\u0026nbsp;1 shows differences in sociodemographic factors and cannabis use measures across both working samples.\u003c/p\u003e\n \u003cp\u003e[Insert Table\u0026nbsp;1 here]\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec13\"\u003e\n \u003ch2\u003ePolygenic Risk scores, and cannabis use over time\u003c/h2\u003e\n \u003cp\u003eALSPAC: Using binomial logistic regression models, adjusted for sex and 10 principal components (PCs), we found that CUD, MDD, and pain PRSs were associated with lifetime cannabis use at 16, 17, and 24 years. In Table\u0026nbsp;2, we reported the results. Pairwise correlation coefficients between each PRS are reported in the supplementary materials (\u003cstrong\u003eFigure S3, S4\u003c/strong\u003e). All coefficients were in the range of low to moderate correlation.\u003c/p\u003e\n \u003cdiv\u003e\n \u003ctable id=\"Tab2\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv\u003eTable 2\u003c/div\u003e\n \u003cdiv\u003e\n \u003cp\u003ePANEL A: \u003cem\u003eOutput of the binomial logistic regressions between PRSs (both main and sensitivity PRSs), and lifetime cannabis use at 16, 17, and 24 years\u003c/em\u003e\u003csup\u003e\u003cem\u003ea\u003c/em\u003e\u003c/sup\u003e \u003cem\u003ein the ALSPAC sample; PANEL B: linear regressions between PRSs (both main and sensitivity PRSs), and CAST scores at 17, 20, and 24 years in the ALSPAC sample\u003c/em\u003e\u003csup\u003e\u003cem\u003eb\u003c/em\u003e\u003c/sup\u003e\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003ePANEL A\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\u0026nbsp;\u003c/th\u003e\n \u003cth align=\"left\"\u003e\u0026nbsp;\u003c/th\u003e\n \u003cth align=\"left\"\u003e\u0026nbsp;\u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" rowspan=\"2\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eLifetime use 16 years\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eLifetime use 17 years\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eLifetime use 24 years\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eOR (95% CI); p\u003csub\u003eFDR\u003c/sub\u003e-value\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eOR (95% CI); p\u003csub\u003eFDR\u003c/sub\u003e-value\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eOR (95% CI); p\u003csub\u003eFDR\u003c/sub\u003e-value\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eCUD PRS\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003e1.25 (1.16–1.34); \u0026lt;0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003e1.19 (1.11–1.29); \u0026lt;0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003e1.15 (1.07–1.23); \u0026lt;0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMD PRS\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003e1.2 (1.13–1.28); \u0026lt;0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003e1.2 (1.11–1.29); \u0026lt;0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003e1.19 (1.1–1.28); \u0026lt;0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePain PRS\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003e1.13 (1.07–1.2); \u0026lt;0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003e1.1 (1.03–1.18); 0.007\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.08 (1.00-1.18); 0.007\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eInsomnia PRS\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.01 (0.95–1.09); 0.7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.01 (0.95–1.07); 0.78\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.02 (0.93–1.11); 0.74\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eShort sleep PRS\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003e1.08 (1.01–1.15); 0.028\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.07 (1.00-1.15); 0.076\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.04 (0.96–1.12); 0.4\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eLong sleep PRS\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.07 (1.00-1.14); 0.06\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.04 (0.97–1.12); 0.28\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.03 (0.96–1.12); 0.4\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSCZ PRS\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003e1.22 (1.11–1.34); \u0026lt;0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003e1.18 (1.1–1.26); \u0026lt;0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.12 (1.00-1.25); 0.13\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePANEL B\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" rowspan=\"2\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eCAST score 17 years\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eCAST score 20 years\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eCAST score 24 years\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eβ (95% CI); p\u003csub\u003eFDR\u003c/sub\u003e-value\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eβ (95% CI); p\u003csub\u003eFDR\u003c/sub\u003e-value\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eβ (95% CI); p\u003csub\u003eFDR\u003c/sub\u003e-value\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eCUD PRS\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.02 (-0.04, 0.08); 0.5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.04 (0.01, 0.07); 0.03\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.11 (0.04, 0.18); 0.005\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMD PRS\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.1 (0.04, 0.15); 0.002\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.06 (0.02, 0.1); 0.01\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.09 (0.02, 0.16); 0.01\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePain PRS\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.09 (0.04, 0.14); 0.002\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.07 (0.04, 0.1); \u0026lt;0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.09 (0.03–0.14); 0.005\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eInsomnia PRS\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.04 (-0.02, 0.11); 0.23\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.02 (-0.02, 0.06); 0.28\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.04 (-0.02, 0.1); 0.2\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eShort sleep PRS\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.06 (0.01, 0.12); 0.08\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.03 (-0.00, 0.06); 0.17\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.04 (-0.03, 0.1); 0.4\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eLong sleep PRS\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.04 (-0.02, 0.1); 0.29\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.02 (-0.02, 0.06); 0.41\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.04 (-0.01, 0.1); 0.3\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSCZ PRS\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.03 (-0.03, 0.08); 0.33\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.02 (-0.03, 0.06); 0.41\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.01 (-0.05, 0.07); 0.72\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003ctfoot\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"4\"\u003e\u003csup\u003ea\u003c/sup\u003eEach model is adjusted for sex and genetic ancestry (PC1-PC10). Odds Ratios (OR), 95% confidence intervals (95% CI), and FDR-adjusted p-values (FDR p-values) are reported in the table. Significant results after the FDR-adjustment are reported significant results after FDR-adjustment.\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"4\"\u003e\u003csup\u003eb\u003c/sup\u003eEach model is adjusted for sex and genetic ancestry (PC1-PC10). Beta coeff. (β), 95% confidence intervals (95% CI), and FDR-adjusted p-values (FDR p-values) are reported in the table.\u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tfoot\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n \u003cp\u003eIn linear regressions, adjusted for sex and PCs we found that CUD PRS was associated with CAST score at 20 (β = 0.04, [95% CI 0.01–0.07] p\u003csub\u003eFDR\u003c/sub\u003e = 0.03), and 24 years (β = 0.11, [95% CI 0.04–0.18] p\u003csub\u003eFDR\u003c/sub\u003e = 0.005). MDD PRS was associated with CAST scores at 17 (β = 0.1, [95% CI 0.04–0.15] p\u003csub\u003eFDR\u003c/sub\u003e = 0.002), 20 years (β = 0.06, [95% CI 0.02–0.1] p\u003csub\u003eFDR\u003c/sub\u003e = 0.01), and 24 years (β = 0.09, [95% CI 0.02–0.16] p\u003csub\u003eFDR\u003c/sub\u003e = 0.01). Pain PRS was associated with CAST scores at 17 (β = 0.09, [95% CI 0.04–0.14] p\u003csub\u003eFDR\u003c/sub\u003e = 0.002), 20 years (β = 0.07, [95% CI 0.04–0.1] p\u003csub\u003eFDR\u003c/sub\u003e \u0026lt; 0.001), and 24 years (β = 0.09, [95% CI 0.03–0.14] p\u003csub\u003eFDR\u003c/sub\u003e = 0.005). See Table\u0026nbsp;2 (panel b). All our sensitivity PRSs, namely the short and long sleep duration PRSs, and the SZ PRS, were not associated with CAST scores at any time points. All the analyses were conducted in the imputed and in non-imputed datasets (Table\u0026nbsp;2, \u003cstrong\u003eTables S5-S8\u003c/strong\u003e). Additional sensitivity analyses investigated the association between each PRS, and the frequency of cannabis use at 24 years (see \u003cstrong\u003eTables S9-S12\u003c/strong\u003e). See also \u003cstrong\u003eTable S26\u003c/strong\u003e for additional adjustment for CUD PRS in the SZ PRS model.\u003c/p\u003e\n \u003cp\u003eWe calculated adjusted R\u003csup\u003e2\u003c/sup\u003e statistics for each of the predictors of the CAST score at 24 years. A model including sex and 10 principal components (step 1) explained around 1% of the variance in the CAST score at 24 years. Adding CUD PRS (step 2) increased this to around 3%. When age at cannabis initiation was included (Step 3), the model predicted just under 10% of the variance. Alternative models with PRSs for MDD (step 2 R\u003csup\u003e2\u003c/sup\u003e = 2.38%), pain (step 2 R\u003csup\u003e2\u003c/sup\u003e = 2.41%), and insomnia (step 2 R\u003csup\u003e2\u003c/sup\u003e = 1.09%) explained lower variance. Age at cannabis initiation added the largest proportion of the variance in all models, and alone, it explained 7.1% of the variance (See \u003cstrong\u003eFigure S7, S7b\u003c/strong\u003e).\u003c/p\u003e\n \u003cp\u003e[Insert Table\u0026nbsp;2 here]\u003c/p\u003e\n \u003cp\u003eIn regression models, adjusted for sex, age, site, and 10 PCs, using data from EU-GEI, we found that CUD PRS was associated with lifetime cannabis use (OR 1.47, 95% CI 1.14–1.89), but not when adjusting for case-control status, or in cases or controls separately. MDD PRS was associated with lifetime cannabis use in cases and controls combined (OR 1.33, 95% CI 1.15–1.54), even when adjusted for case-control status (OR 1.28, 95% CI 1.1—1.48) and in the cases-only sample (OR 1.41, 95% CI 1.1–1.82), but not in the controls-only sample. Insomnia PRS was associated with lifetime cannabis use in cases only (OR 1.52, 95% CI 1.11–2.1). No PRS was associated with lifetime cannabis use in the controls-only sample (\u003cstrong\u003eTable S4\u003c/strong\u003e).\u003c/p\u003e\n \u003cp\u003eA multinomial regression model (with ‘never users’ as the baseline group), additionally adjusted for case-control status, found that CUD PRS (RRR 1.82, 95% CI 1.38–2.41; p\u003csub\u003eFDR\u003c/sub\u003e = 0.001) and pain PRS (RRR 1.27, 95% CI 1.11–1.46; p\u003csub\u003eFDR\u003c/sub\u003e = 0.007) were associated with “weekly-to-daily” use of cannabis. Similarly, in the cases-only sample, CUD PRS (RRR 1.84, 95% CI 1.24–2.74; p\u003csub\u003eFDR\u003c/sub\u003e = 0.01) and pain PRS (RRR 1.49, 95% CI 1.21–1.82; p\u003csub\u003eFDR\u003c/sub\u003e = 0.002) were associated with weekly-to-daily use. In the controls-only sample, only CUD PRS (RRR 1.93, 95% CI 1.27–2.94; p\u003csub\u003eFDR\u003c/sub\u003e = 0.013) was associated with weekly-to-daily use. See \u003cstrong\u003eFig.\u0026nbsp;1, Table S15\u003c/strong\u003e. See \u003cstrong\u003eTables S13\u003c/strong\u003e, \u003cstrong\u003eS14\u003c/strong\u003e, and \u003cstrong\u003eS16\u003c/strong\u003e for sensitivity analyses.\u003c/p\u003e\n \u003cp\u003eIn the EU-GEI, Step 1 (sex, site and 10 PCs) explained 4.56% of the variance. When age was included (step 2), this rose to 6.78%. Adding CUD PRS (step 3) increased this to 7.28% and adding age at cannabis initiation (step 4) increased this to 9.41%. Age at cannabis initiation explained 4.53% of the variance alone. The models with PRSs for MDD (Step 3 R\u003csup\u003e2\u003c/sup\u003e = 6.77%), pain (Step 3 R\u003csup\u003e2\u003c/sup\u003e = 6.89%), and insomnia (Step 3 R\u003csup\u003e2\u003c/sup\u003e = 6.8%) explained a similar but lower proportion of the variance. (See supplementary for models restricted to cases and controls separately). \u003cstrong\u003eFigure S11-S11b\u003c/strong\u003e.\u003c/p\u003e\n \u003cp\u003e[Insert Fig.\u0026nbsp;1 here]\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec14\"\u003e\n \u003ch2\u003eThe association between PRSs and age at cannabis initiation\u003c/h2\u003e\n \u003cp\u003eIn the ALSPAC cohort, CUD (β=-0.26, [95% CI -0.4, -0.11]; p\u003csub\u003eFDR\u003c/sub\u003e 0.004), MDD (β=-0.16, [95% CI -0.3, -0.05]; p\u003csub\u003eFDR\u003c/sub\u003e = 0.01), and SZ PRSs (β=-0.18, [95% CI -0.3, -0.08]; p\u003csub\u003eFDR\u003c/sub\u003e 0.002) were consistently associated with earlier age at cannabis initiation in both the linear and the logistic models (\u003cstrong\u003eTable S18\u003c/strong\u003e). By contrast, we found no clear evidence of association between the other PRSs and age at first cannabis use and therefore were not included in further analyses.\u003c/p\u003e\n \u003cp\u003eIn the EU-GEI sample, in the linear regression model, MDD (β = 0.42, SE: 0.16, [95% CI 0.1, 0.73]; p\u003csub\u003eFDR\u003c/sub\u003e = 0.02) and insomnia (β = 0.87, SE: 0.2, [95% CI 0.47, 1.27]; p\u003csub\u003eFDR\u003c/sub\u003e \u0026lt; 0.001) PRSs were associated with age at cannabis initiation (\u003cstrong\u003eTable S20, Table S19)\u003c/strong\u003e.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec15\"\u003e\n \u003ch2\u003eIndependent and combined effects of CUD PRS and age at first use on problematic use\u003c/h2\u003e\n \u003cp\u003eIn the ALSPAC cohort, we fitted negative binomial regression models to examine the relationship between CUD PRS and CAST scores at 24 years, adjusting for principal components and sex. Higher CUD PRS (IRR = 1.5, 95% CI [1.16, 1.93], p\u003csub\u003eFDR\u003c/sub\u003e = 0.004) and higher MDD PRS (IRR = 1.44, 95% CI [1.19, 1.75], p\u003csub\u003eFDR\u003c/sub\u003e \u0026lt; 0.001) were associated with increased CAST scores. Age at first cannabis use was associated with CAST scores at 24 years in all three models. Neither CUD PRS (IRR = 0.95, 95% CI [0.82, 1.11], p\u003csub\u003eFDR\u003c/sub\u003e = 1) or MDD PRS (IRR = 0.95, 95% CI [0.84, 1.08], p\u003csub\u003eFDR\u003c/sub\u003e = 1) moderate the relationship between early age at first use and CAST score. SZ PRS was not associated with CAST scores at 24 years nor moderated its relationship with age at first use (\u003cstrong\u003eTable S21, Figs.\u0026nbsp;2–4)\u003c/strong\u003e. Results were consistent across the imputed and non-imputed datasets and in sensitivity analyses in which the outcome was the probability of ‘daily or almost daily use’ at 24 years (\u003cstrong\u003eTable S22\u003c/strong\u003e).\u003c/p\u003e\n \u003cp\u003eSecondary analyses, adjusted for sex and 10 PCs, suggested an additive interaction between CUD PRS and early age at first use (≤ 15 years) on CAST score at age 24, RERI 2.4 (95% CI [0.82, 3.99]). MDD and SZ PRSs did not show evidence of additive interaction. \u003cstrong\u003eTable S23\u003c/strong\u003e.\u003c/p\u003e\n \u003cp\u003eIn the EU-GEI sample, logistic regressions, adjusted for sex, site, 10 PCs, and case-control status, were conducted to test for evidence of interaction between CUD and MDD PRSs x age at cannabis initiation on the outcome “weekly-to-daily” frequency of use category. No clear multiplicative or additive interactions between CUD or MDD PRSs and early cannabis initiation (≤ 15 years) were found in the EU-GEI sample, in the combined, cases-only, or controls-only group. \u003cstrong\u003eFigures S17-S18\u003c/strong\u003e, \u003cstrong\u003eTable S24\u003c/strong\u003e.\u003c/p\u003e\n \u003cp\u003e[Insert Fig.\u0026nbsp;2 here]\u003c/p\u003e\n \u003cp\u003e[Insert Fig.\u0026nbsp;3 here]\u003c/p\u003e\n \u003cp\u003e[Insert Fig.\u0026nbsp;4 here]\u003c/p\u003e\n\u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eTo our knowledge, this is the first study to find that genetic liability for different psychiatric and behavioural phenotypes is associated with problematic cannabis use across different ages. We also found that early initiation impacts cannabis use patterns. We selected the PRSs for CUD, MDD, pain, and insomnia (alongside the sensitivity PRSs for SZ, short and long sleep duration) to explore factors that could explain (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e) self-medication and (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e) medical cannabis use, both increasing worldwide(\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eOverall, two main novel findings emerge from our analyses: (i) an independent effect of age at first use and PRSs on later problematic cannabis use with earlier initiation, contributing to problematic use later in life. (ii) The influence of PRSs for CUD, pain, and MDD on cannabis use, suggesting that in some people self-medication might lead to heavier cannabis use.\u003c/p\u003e \u003cp\u003eIn ALSPAC, CUD, MDD, and pain PRSs were associated with lifetime cannabis use at all time points (16, 17, and 24 years). MDD and pain PRSs were also associated with problematic cannabis use (CAST scores) at all time points (17, 20, and 24 years). Interestingly, CUD PRS was associated with CAST scores at 20 and 24 years, but not at 17. SZ PRS was associated with lifetime use but not with CAST scores. Younger age at cannabis initiation (\u0026le;\u0026thinsp;15 years) was associated with higher genetic risk for CUD, MDD, and SZ. Insomnia, short and long sleep duration PRSs had minimal/no association with all our outcomes (see Supplementary).\u003c/p\u003e \u003cp\u003eIn the EU-GEI sample, while we could not test for the developmental timing of patterns of cannabis use, we found that CUD PRS was associated with weekly-to-daily use of cannabis in the full sample and in cases and controls independently. Pain PRS was associated with weekly-to-daily use in the full sample and in cases only. MDD PRS was only associated with weekly-to-daily use in the full sample. Lastly, in both cohorts, we did not find evidence that either CUD or MDD PRSs moderated the relationship between age at first use and later problematic cannabis use. However, in ALSPAC, additive interaction was found between CUD PRS and early initiation (\u0026le;\u0026thinsp;15 years).\u003c/p\u003e \u003cp\u003eAlthough \u003cb\u003eFigs.\u0026nbsp;2 and 3\u003c/b\u003e show that those who started using cannabis at age\u0026thinsp;\u0026le;\u0026thinsp;15, with high CUD and MDD PRS have a sharper increase in CAST score at age 24 compared with those who started at a later age, overall these results, combined with our variance explained models, suggest that age at cannabis initiation and polygenic liability for CUD and MDD act mostly independently from each other. Indeed, the former appears to play a more important role in the risk of developing heavy cannabis use later in life.\u003c/p\u003e \u003cp\u003eThese novel findings have important public health implications. Firstly, they suggest that PRSs for CUD, MDD, and pain are associated with more frequent and problematic cannabis use disorder. Therefore, these results could contribute to the early identification of individuals at higher risk of consumption. Earlier age at first use emerged as a robust, independent predictor of problematic cannabis use. Lastly, the lack of moderation effects across all analyses suggests that early cannabis use is a risk factor irrespective of the genetic loading we explored.\u003c/p\u003e \u003cp\u003eOur as other studies found that the schizophrenia PRS is not associated with patterns of cannabis use nor with CUD(\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e, \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e, \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e), thus speaking against the hypothesis that shared genetic liability explains the cannabis-psychosis association reported in epidemiology. Consistent with Austin-Zimmerman et al(\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e), we found that, after controlling for CUD PRS, the associations between SZ PRS and some of the measures of cannabis use were either markedly reduced or no longer significant, again indicating a substantial confounding by CUD PRS (\u003cb\u003eTable S26\u003c/b\u003e). This less evident effect of SZ PRS on frequency of use might be explained by SZ GWAS including a substantial proportion of cases with both SZ and CUD, given the higher prevalence of cannabis use among people with schizophrenia than among controls(\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e). Furthermore, the difference in outcome between lifetime cannabis use and CUD, have been confirmed by GWAS studies pointing at differences in their genetic architecture (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e)(\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e). Few studies have been conducted on the association between MDD, pain, and insomnia genetic liability with problematic cannabis use. MDD GWAS correlates with both cannabis initiation and with CUD GWAS(\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e). In our analyses, we found an association between MDD PRS, and different cannabis use variables, which was consistent across datasets (\u003cb\u003eTables S3\u003c/b\u003e-\u003cb\u003eS4\u003c/b\u003e, \u003cb\u003eS10\u003c/b\u003e, \u003cb\u003eS13\u003c/b\u003e, \u003cb\u003eS15\u003c/b\u003e). Liability to pain is genetically correlated with level of substance use and substance use disorders, including CUD(\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e). Mendelian randomisation suggests a causal effect of chronic pain on CUD with important public health consequences(\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e). Here, we provide evidence that genetic liability to chronic pain is also associated with some cannabis use phenotypes. Lastly, genomic structural equation modelling has shown that long sleep duration shares genetic liability with CUD, as well as other traits, suggesting that these traits may represent a common underlying genetic factor(\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e). Our analyses were less consistent in showing such an association.\u003c/p\u003e \u003cp\u003eAn important strength of our study is the longitudinal nature of the large and well-phenotyped ALSPAC sample. Moreover, the replication on the multi-centre EUGEI sample adds generalisability. The richness of data from both cohorts and several sensitivity analyses have allowed to account for multiple confounding factors.\u003c/p\u003e \u003cp\u003eAmong the study limitations, we need to consider the potential recall bias due to the cannabis variables that in the EU-GEI were based on self-reported retrospective information. However, prior studies have shown that self-reported information on cannabis well-correlated with biological measures, the latter only being able to capture recent use rather than the cumulative harmful effect of cannabis overtime(\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e). Secondly, the information on the type of cannabis used in the ALSPAC sample was only collected at 24 years(\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e), preventing comparison of data on cannabis potency, across the two cohorts. In the EU-GEI, controls are representative of each site\u0026rsquo;s population at risk(\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e). We adjusted for case-control status, besides conducting separate analyses in the cases and controls-only samples, which possibly reduced power as evidenced by wider confidence intervals. Additionally, our sensitivity analyses showed that SZ and CUD PRS scores did not differ between participants with observed vs missing CAST outcome. This suggests that our findings are unlikely to be explained by attrition bias or poor data quality (\u003cb\u003eFigures S8-S13\u003c/b\u003e).\u003c/p\u003e \u003cp\u003eBecause of the differences between our two samples, we could not make direct data replication. We conducted a series of sensitivity analyses employing measures like frequency of cannabis use common to both datasets.\u003c/p\u003e \u003cp\u003eLastly, our analyses were restricted to individuals of European ancestry which limits the generalisability of our findings. The available PRSs were built from the largest available GWASs, still predominantly from European descent, a limitation still affecting the whole field.\u003c/p\u003e"},{"header":"Conclusions","content":"\u003cp\u003eAlthough genetic predisposition contributes to problematic cannabis use later in life, our results suggest that early age at first use explains a greater proportion of the variance in problematic use, highlighting the importance of early detection and targeted intervention. Public health strategies and primary care intervention should aim at delaying first cannabis use, particularly during adolescence, when both neurodevelopmental processes and environmental influences may shape long-term outcomes. Furthermore, the association we found between genetic liability to CUD, MDD, and pain with heavier use of cannabis may reflect an underlying vulnerability in some of those who use cannabis to cope with physical or psychological discomfort, underscoring the importance of identifying self-medication as early as possible and offering monitored and tailored intervention strategies.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAcknowledgments\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe express our gratitude to all the families who took part in this study, the midwives for their help in recruiting team, and the ALSPAC team, which includes staff involved in data collection, data and administration staff, technical managers and the technical staff working at the Bristol Bioresource Laboratory, based within the University of Bristol.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe EU-GEI WP2 group collected or supervised the data collection.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe UK Medical Research Council and Wellcome (MR/Z505924/1) and the University of Bristol provide core support for ALSPAC. The EU-GEI project was funded by the European Community\u0026rsquo;s Seventh Framework Programme under grant agreement No. HEALTH-F2-2009-241909 (Project EU-GEI). The UK Medical Research Council (MRC) under the grant awarded to MDF reference MR/T007818/1 funded the salaries of IAZ, MDF, GT, and ZL. The Lord Leverhulme\u0026rsquo;s Charitable Trust and the Velvet Foundation funded the salary of ES. EV is supported by the National Institute for Health Research (NIHR) Maudsley Biomedical Research Centre at South London and Maudsley NHS Foundation Trust and King\u0026rsquo;s College London. Genome-wide genotyping data was obtained by Sample Logistics and Genotyping Facilities at Wellcome Sanger Institute and LabCorp (Laboratory Corporation of America) with support from 23 and Me. The views expressed are those of the authors and not necessarily those of the NHS, the NIHR, or the Department of Health and Social Care in the UK.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003cstrong\u003eData availability statement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe ALSPAC participants\u0026rsquo; informed consent does not allow the data to be made available through any third party maintained public repository. Supporting data are available from ALSPAC upon request under the approved proposal number (number B4002). Full instructions for applying for data access can be found here: http://bristol.ac.uk/alspac/researchers/access/. The ALSPAC study website contains details of all available data (http://www.bristol.ac.uk/alspac/researchers/our-data/).\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eJohnson EC, Demontis D, Thorgeirsson TE, Walters RK, Polimanti R, Hatoum AS et al (2020) A large-scale genome-wide association study meta-analysis of cannabis use disorder. Lancet Psychiatry 7(12):1032\u0026ndash;1045\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGreco LA, Reay WR, Dayas CV, Cairns MJ (2022) Pairwise genetic meta-analyses between schizophrenia and substance dependence phenotypes reveals novel association signals with pharmacological significance. Transl Psychiatry 12(1):403\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGillespie NA, Kendler KS (2021) Use of genetically informed methods to clarify the nature of the association between cannabis use and risk for schizophrenia. JAMA Psychiatry 78(5):467\u0026ndash;468\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAustin-Zimmerman I, Spinazzola E, Quattrone D, Wu-Choi B, Trotta G, Li Z et al (2024) The impact of schizophrenia genetic load and heavy cannabis use on the risk of psychotic disorder in the EU-GEI case-control and UK Biobank studies. Psychol Med 54(15):4160\u0026ndash;4172\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAustin-Zimmerman I, Levey DF, Deak JD, Galimberti M, Adhikari K, Coleman JR et al (2025) Multi-ancestry genetic architecture of sleep duration and its relationship to other sleep and psychiatric phenotypes. medRxiv\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLeung J, Chan G, Stjepanović D, Chung JYC, Hall W, Hammond D (2022) Prevalence and self-reported reasons of cannabis use for medical purposes in USA and Canada. Psychopharmacology 239(5):1509\u0026ndash;1519\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHallinan CM, Bonomo YA (2022) The rise and rise of medicinal cannabis, what now? Medicinal cannabis prescribing in Australia 2017\u0026ndash;2022. Int J Environ Res Public Health 19(16):9853\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSpinazzola E, Degen H, Austin-Zimmerman I, Trotta G, Chesney E, Li Z et al (2025) Are reasons for first using cannabis associated with subsequent cannabis consumption (standard THC units) and psychopathology? BMJ Mental Health. ;28(1)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTrubetskoy V, Pardi\u0026ntilde;as AF, Qi T, Panagiotaropoulou G, Awasthi S, Bigdeli TB et al (2022) Mapping genomic loci implicates genes and synaptic biology in schizophrenia. Nature 604(7906):502\u0026ndash;508\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLevey DF, Galimberti M, Deak JD, Wendt FR, Bhattacharya A, Koller D et al (2023) Multi-ancestry genome-wide association study of cannabis use disorder yields insight into disease biology and public health implications. Nat Genet 55(12):2094\u0026ndash;2103\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAdams MJ, Streit F, Meng X, Awasthi S, Adey BN, Choi KW et al (2025) Trans-ancestry genome-wide study of depression identifies 697 associations implicating cell types and pharmacotherapies. Cell 188(3):640\u0026ndash;652 e9\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLane JM, Jones SE, Dashti HS, Wood AR, Aragam KG, van Hees VT et al (2019) Biological and clinical insights from genetics of insomnia symptoms. Nat Genet 51(3):387\u0026ndash;393\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJansen PR, Watanabe K, Stringer S, Skene N, Bryois J, Hammerschlag AR et al (2019) Genome-wide analysis of insomnia in 1,331,010 individuals identifies new risk loci and functional pathways. Nat Genet 51(3):394\u0026ndash;403\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJohnston KJ, Adams MJ, Nicholl BI, Ward J, Strawbridge RJ, Ferguson A et al (2019) Genome-wide association study of multisite chronic pain in UK Biobank. PLoS Genet 15(6):e1008164\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHarris PA, Taylor R, Thielke R, Payne J, Gonzalez N, Conde JG (2009) Research electronic data capture (REDCap)\u0026mdash;a metadata-driven methodology and workflow process for providing translational research informatics support. J Biomed Inf 42(2):377\u0026ndash;381\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBoyd A, Golding J, Macleod J, Lawlor DA, Fraser A, Henderson J et al (2013) Cohort profile: the \u0026lsquo;children of the 90s\u0026rsquo;\u0026mdash;the index offspring of the Avon Longitudinal Study of Parents and Children. Int J Epidemiol 42(1):111\u0026ndash;127\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFraser A, Macdonald-Wallis C, Tilling K, Boyd A, Golding J, Davey Smith G et al (2013) Cohort profile: the Avon Longitudinal Study of Parents and Children: ALSPAC mothers cohort. Int J Epidemiol 42(1):97\u0026ndash;110\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNorthstone K, Lewcock M, Groom A, Boyd A, Macleod J, Timpson N et al (2019) The Avon Longitudinal Study of Parents and Children (ALSPAC): an update on the enrolled sample of index children in 2019. Wellcome Open Res 4:51\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGayer-Anderson C, Jongsma HE, Di Forti M, Quattrone D, Velthorst E, De Haan L et al (2020) The European network of national schizophrenia networks studying gene\u0026ndash;environment interactions (EU-GEI): Incidence and first-episode case\u0026ndash;control programme. Soc Psychiatry Psychiatr Epidemiol 55(5):645\u0026ndash;657\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHines LA, Freeman TP, Gage SH, Zammit S, Hickman M, Cannon M et al (2020) Association of high-potency cannabis use with mental health and substance use in adolescence. JAMA psychiatry 77(10):1044\u0026ndash;1051\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLegleye S, Karila L, Beck F, Reynaud M (2007) Validation of the CAST, a general population Cannabis Abuse Screening Test. J Subst Use 12(4):233\u0026ndash;242\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLees Thorne R, Hines LA, Burke C, Jones HJ, Freeman TP (2025) Association of childhood mental health and cognition with longitudinal patterns of cannabis problems in adolescence. Psychol Med 55:e129\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDi Forti M, Quattrone D, Freeman TP, Tripoli G, Gayer-Anderson C, Quigley H et al (2019) The contribution of cannabis use to variation in the incidence of psychotic disorder across Europe (EU-GEI): a multicentre case-control study. Lancet Psychiatry 6(5):427\u0026ndash;436\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eQuattrone D, Reininghaus U, Richards AL, Tripoli G, Ferraro L, Quattrone A et al (2021) The continuity of effect of schizophrenia polygenic risk score and patterns of cannabis use on transdiagnostic symptom dimensions at first-episode psychosis: findings from the EU-GEI study. Transl Psychiatry 11(1):423\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHines LA, Heron J, Zammit S (2024) Incident psychotic experiences following self-reported use of high‐potency cannabis: Results from a longitudinal cohort study. Addiction 119(9):1629\u0026ndash;1634\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDawson D, Chan G, Stjepanović D, Lorenzetti V, Hall WD, Leung J (2025) The prevalence and correlates of self-reported cannabis use for medicinal, dual and recreational motives in Australia: Findings from the National Drug Strategy Household Survey 2022/2023. Drug Alcohol Rev\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHjorth\u0026oslash;j C, Uddin MJ, Wimberley T, Dalsgaard S, Hougaard DM, B\u0026oslash;rglum A et al (2021) No evidence of associations between genetic liability for schizophrenia and development of cannabis use disorder. Psychol Med 51(3):479\u0026ndash;484\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGuloksuz S, Pries LK, Delespaul P, Kenis G, Luykx JJ, Lin BD et al (2019) Examining the independent and joint effects of molecular genetic liability and environmental exposures in schizophrenia: results from the EUGEI study. World Psychiatry 18(2):173\u0026ndash;182\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJohnson EC, Austin-Zimmerman I, Thorpe HH, Levey DF, Baranger DA, Colbert SM et al (2024) Cross-ancestry genetic investigation of schizophrenia, cannabis use disorder, and tobacco smoking. Neuropsychopharmacol 49(11):1655\u0026ndash;1665\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eToikumo S, Vickers-Smith R, Jinwala Z, Xu H, Saini D, Hartwell E et al (2023) The genetic architecture of pain intensity in a sample of 598,339 US veterans. medRxiv\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBharat C, Webb P, Wilkinson Z, McKetin R, Grebely J, Farrell M et al (2023) Agreement between self-reported illicit drug use and biological samples: a systematic review and meta‐analysis. Addiction 118(9):1624\u0026ndash;1648\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"},{"header":"Table 1","content":"\u003cp\u003eTable 1 is available in the Supplementary Files section.\u003c/p\u003e\n"}],"fulltextSource":"","fullText":"","funders":[{"identity":"315231cc-417e-476c-af5b-ac01bc46fc18","identifier":"10.13039/100010269","name":"Wellcome Trust","awardNumber":"MR/Z505924/1","order_by":0},{"identity":"959d1114-00c5-4bc1-9f3c-fda7d88e2851","identifier":"10.13039/100011102","name":"Seventh Framework Programme","awardNumber":"HEALTH-F2-2009-241909 ","order_by":1},{"identity":"e2ee5af1-4e23-4f09-a620-ae977fad5155","identifier":"10.13039/501100000265","name":"Medical Research Council","awardNumber":"MR/T007818/1 ","order_by":2}],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"King's College London","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Polygenic risk score, Cannabis use, Gene x environment interaction, ALSPAC, EU-GEI","lastPublishedDoi":"10.21203/rs.3.rs-8359813/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8359813/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003ePolygenic risk scores (PRSs) for cannabis use disorder (CUD), major depressive disorder (MDD), insomnia, and chronic pain might contribute to patterns of cannabis use, including problematic cannabis use. We examined PRSs for the above traits and different patterns of cannabis use across developmental stages using the ALSPAC longitudinal cohort (N = 8,224 participants with genotype data) and the EU-GEI case-control study (N = 994 [56.9%] controls and 752 [43.1%] first-episode psychosis cases with genotype data) for replication. We fitted regression models to test associations between PRSs and patterns of cannabis use at different ages. An interaction term was included to test whether the association between early cannabis initiation and heavy use is moderated by PRSs. Genetic liability to CUD, MDD, and pain was consistently associated with heavier use of cannabis. In the ALSPAC sample, linear regression models showed that CUD PRS was associated with CAST score at 20, and 24 years. MDD and pain PRSs were associated with CAST scores at 17, 20, and 24 years. In the EU-GEI, CUD PRS and pain PRS were associated with “weekly-to-daily” use of cannabis. Similarly, in the cases-only sample, CUD PRS and pain PRS were associated with weekly-to-daily use. In the controls-only sample, only CUD PRS was associated with weekly-to-daily use. This reflects an underlying vulnerability that may lead some people to use cannabis more heavily as a coping mechanism. However, age at initiation (ALSPAC = 7.1%, EU-GEI = 4.53%) explained a greater proportion of the variance of problematic cannabis use than polygenic liability, indicating atime-sensitiveintervention target.\u003c/p\u003e","manuscriptTitle":"Polygenic prediction of cannabis-related outcomes over time: Evidence from the ALSPAC longitudinal cohort and the EU-GEI case-control study","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-12-16 09:04:20","doi":"10.21203/rs.3.rs-8359813/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"ab4b987e-dedb-41aa-aaa2-6c375c06fc9c","owner":[],"postedDate":"December 16th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":59628681,"name":"Psychiatry"}],"tags":[],"updatedAt":"2025-12-16T09:04:20+00:00","versionOfRecord":[],"versionCreatedAt":"2025-12-16 09:04:20","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8359813","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8359813","identity":"rs-8359813","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}