Association of High-Potency Cannabis Use With Mental Health and Substance Use in Adolescence

Key Points Question Does use of high-potency cannabis (compared with use of low-potency cannabis) increase risks for problems resulting from cannabis use, common mental disorders, and psychotic experiences after controlling for early-life mental health symptoms and frequency of use? Findings In this cohort study of 1087 participants who reported cannabis use in the previous year, after adjusting for frequency of cannabis use and early adolescent mental health, use of high-potency cannabis was associated with a significant increase in the frequency of cannabis use, likelihood of cannabis problems, and likelihood of anxiety disorder. Those using high-potency cannabis had a small increase in the likelihood of psychotic experiences; however, this risk was attenuated after adjustment for frequency of cannabis use. Meaning Risks for cannabis use problems and anxiety disorders are higher among those reporting use of high-potency cannabis; provision of public health messaging regarding the importance of reducing both frequency of cannabis use and the potency of the drug, as well as limiting the availability of high-potency cannabis, may be effective for reducing these risks.

This supplementary material has been provided by the authors to give readers additional information about their work.

Study population
Pregnant women in the former Avon Health Authority in south-west England who had an estimated date of delivery between 1 April 1991 and 31 December 1992 were invited to take part, resulting in a cohort of 14 541 pregnancies and 13,988 children alive at 1 year of age. When the oldest children were approximately 7 years of age, an attempt was made to bolster the initial sample with eligible cases who had failed to join the study originally. The total sample size for analyses using any data collected after the age of seven is therefore 15,454 pregnancies, resulting in 15,589 foetuses. Of these 14,901 were alive at 1 year of age. 1 Ethical approval for this study was obtained from the ALSPAC Law and Ethics Committee and the Local Research Ethics Committees. The ALSPAC study website contains details of all the data available through a fully searchable data dictionary (http://www.bristol.ac.uk/alspac/researchers/our-data/). Study data were collected and managed using REDCap electronic data capture tools hosted at University of Bristol. 2,3

Cannabis use frequency at age 24
Participants were asked "in the last 12 months, how often have you used cannabis?" with the options "not in the last 12 months", "once or twice", less than monthly", "monthly", "weekly", "daily or almost daily".

Problematic cannabis use at age 24
Those who endorsed two or more of the following Cannabis Abuse Screening Test (CAST) 6 items within the past year were classified as having recently experienced problems as a result of their cannabis use: using cannabis before midday, using cannabis alone, having memory problems when using cannabis, having friends or family telling them to reduce their cannabis use, experiencing problems such as arguments or fights as a result of cannabis use

Other substance use and dependencies at age 24
Participants reported any use in the past 12 months of the following: powder cocaine, crack cocaine, amphetamines, nitrous oxide, inhalants, sedatives, hallucinogens, opiates, or injected drugs. Reporting use of any one of these drugs was categorised as recent other illicit drug use.

Prospective Measures from Early Childhood and Adolescence
Childhood socioeconomic position was assessed through measures from maternal questionnaires completed during pregnancy; variables were maternal educational attainment (university degree/A level or advanced level/O level, or less than O level, which includes any other qualifications of a lower academic standard or having no qualifications), and parents occupation class (i/ii/iii or iv/v).

Missing data and imputation
As outcomes and exposures were collected at the same time point, the majority of missing data were in the covariates assessed at earlier ages (see Table 1). Missing data in all analysis variables (exposures, outcomes, covariates) were addressed through multiple imputation using chained equations, which uses a series of univariate regression models to impute each incomplete variable sequentially. Each model included all other analysis variables as predictors, along with the following auxiliary variables: experiencing bullying between ages 0-16, parental separation ages 0-16, parent mental health problems age 0-16, parent substance use age 0-16, MFQ score at age 16 and 18, number of self-reported psychotic-like experiences at age 14, and conduct disorder symptoms to age 13. Estimates were obtained by pooling results across 40 imputed datasets using Rubin's rules, and assessment of Monte Carlo variability confirmed this as a suitable number of imputations. 7

eAppendix 2. Sensitivity Test Using Propensity Score Models in Complete Case Data
We have generated a propensity score in the complete-case data using logistic regression (outcome: reported cannabis potency) and the predict command in Stata 15.1. The variables and interactions included in the score are listed in eTable 1.

Individual variables
Interaction terms Gender PLIKS score age 12 x MFQ score age 12 Parent's socioeconomic status Gender x MFQ score age 13 Maternal education Gender x PLIKS score age 12 Frequency of cannabis use Age of cannabis onset x MFQ score age 13 MFQ score at age 13 Age of cannabis onset x PLIKS score age 12 Number of psychotic-like experiences at age 12 Age of cannabis onset x gender Age of cannabis onset Age of cannabis onset x Parent's socioeconomic status Class of conduct problems 8 Parent's socioeconomic status x MFQ score age 13 Being bullied age 0-16 9 Parent's socioeconomic status x PLIKS score age 12 Parent substance problems when child age 0-16 9 Parent's socioeconomic status x gender Caregiver mental health problems age 0-16 9 Parent's socioeconomic status x age of cannabis onset Parental separation age 0-16 9 This resulted in a model with R 2 of 0.26, and _hat predictor P Value of ≤0.001 (indicating good specification of the model). The score was applied to the data using Inverse Probability Weights (IPW). As IPW is sensitive to very large numbers, the IPWs were truncated to the 95% percentile