We also conducted sensitivity analyses for different classification of marijuana use

Participants were invited to participate in 9 in-person clinic examinations over the study period . We used data from the first seven visits, up to year 20, as ECG measures were available at baseline, visit Year 7, and 20. Each participant who attended the examination received non-monetary gifts and monetary reimbursement to cover expenses. All study protocols were approved by the institutional review boards at each site. Multiple marijuana use variables are available for all visits . Current marijuana use was assessed by the following survey question: ‘During the last 30 days, on how many days did you use marijuana?’. We defined daily use as 30 days of use in the last 30 days. Direct self-reported lifetime exposure was assessed by the question: ‘About how many times in your lifetime have you used marijuana?’ We used current use and baseline lifetime use to compute marijuana-years, with 1 year of exposure equivalent to 365 days of marijuana use. We assumed that current use at each visit reflected the average number of days of use during the months before and after each visit. We estimated the cumulative lifetime use by adding the total number of days using marijuana during followup. We adjusted our estimate upwards whenever participants self-reported higher lifetime use than we compute for each visit. Marijuana use was not legal in the cities at this time. Standard 12-lead electrocardiogram was recorded at baseline, Year 7 and Year 20 visits,weed growing systems as described extensively elsewhere. All abnormalities were coded according to Minnesota Code Manual of Electrocardiographic Findings . The MC is used for population research and clinical trials and standardizes coding of ECG abnormalities.

We classified abnormalities according to MC . We also built composite categories of major and minor abnormalities: If an individual’s ECG contained any abnormality on the major list , the ECG was classed as composite major. If it contained only abnormalities on the minor list , we classed the ECG as composite minor. This allowed us to study both composite major and minor abnormalities and specific ECG abnormalities.Tobacco cigarette smoking behavior was evaluated at each in-person CARDIA examination and at yearly phone follow-up between CARDIA examinations. We used these data to estimate cumulative lifetime exposure to tobacco cigarettes in terms of pack-years. We estimated alcohol consumption as drink-years . Education was the highest educational grade reached by the participant by examination Year 20. We measured physical activity at every visit with the CARDIA physical activity history questionnaire. Our cardiovascular risk factor measurements included blood pressure, blood cholesterol, body mass index , binge drinking and diagnosis of diabetes mellitus, which were collected at each CARDIA examination . We observed prevalent abnormalities at visit 0, 7, and 20, but focused on Year 20 since major ECG abnormalities are expectedly more prevalent later in life and since cumulative marijuana exposure rises with time. Based on the number of computed cumulative marijuana-years and data on current use, we divided participants into four categories: never used marijuana; past use and moderate cumulative lifetime marijuana use, up to or 0.5 marijuana-years; past use and higher cumulative lifetime marijuana use, above 0.5 marijuana-years; and, current users , no matter the level of their cumulative use. We first analyzed the association between marijuana use and ECG abnormalities in unadjusted logistic regression models, separately at visit 0, 7, and 20. We then adjusted for demographic variables and then further adjusted for potential confounders such as tobacco cigarette smoking, alcohol consumption, and physical activity and BMI .. We decided to restrict main multi-variable adjusted models to the previously mentioned variables because of low event number in specific ECG abnormalities, but still performed exploratory analyses with fully adjusted models including cardiovascular health variables , presented in the online supplement.

To account for deaths and informative censoring in later examinations , we used inverse probability of attrition weights . We separately fit a model for loss to follow-up caused by the death, and a separate model for censoring due to reasons other than death, computed in one score. We used last-value-carried-forward and backward imputation for missing covariables and verified results using multiple imputation. The first considered only cumulative use and not current use; the second compared daily marijuana use to less frequent, past use, and never use. Finally, we stratified our results to see if they varied by sex and race because prevalence of ECG abnormalities, and distribution of exposure and covariables differs between Black and white, and male and female participants. We restricted regression analyses to ECG abnormalities that occurred in at least 50 of each race-sex stratum. We also fitted models with marijuana-years modelled as a restricted cubic spline as covariable for state of marijuana use . We further modelled incident abnormalities between Year 0 and Year 20. We included specific major and minor abnormalities at Year 20 that were not already identified in these categories in Year 0. For example, a specific minor abnormality detected at Year 0 that evolved into a major abnormality detected at Year 20 was coded as an incident major abnormality at Year 20. We applied a series of unadjusted and multi-variable adjusted models to analyze the association between current and lifetime cumulative marijuana use on incident ECG abnormalities. Tests of statistical significance were two-tailed; alpha level was 0.05. All statistical analyses were performed on Stata version 14.2 . We hypothesized that cumulative marijuana use was not associated with ECG abnormalities, but that current use might be associated with unspecific changes in ECG.

Various small experimental studies suggested immediate effects after using marijuana, with measured parameters returning to pre-exposure levels after ceasing marijuana use. The primary research question and analysis plan were submitted to the CARDIA Presentation & Publication Committee, before obtaining and analysing the data. However, they were not pre-registered on a publicly available platform and the results should thus be considered exploratory. Table 2 shows that at the Year 20 examination, 173 participants had composite major abnormalities and 944 had composite minor abnormalities . Composite and specific major abnormalities at Year 20 did not vary with status of marijuana use, but showed a tendency towards fewer events in current marijuana users: when we compared current marijuana use to never use,indoor farming systems the unadjusted OR for composite major ECG abnormalities was 0.77 . After multi-variable adjustment, the OR was 0.55 . Tables 2 and 3 show that in the unadjusted model, composite minor abnormalities and some specific minor abnormalities were more common among current marijuana users . These differences were attenuated after adjustment for demographic variables . The odds ratios stayed similar between categories of marijuana use after multi-variable adjustment and use of IPAW . Figure 1 illustrates that after multi-variable adjustment, no specific ECG abnormality or composite major or minor abnormalities was significantly associated with marijuana use. Current marijuana users had a multi-variable adjusted OR of 0.34 for major ST-T abnormalities, with a p-value across categories of 0.17. Past users with a cumulative use of over 0.5 marijuana-years had a multi-variable adjusted OR of 2.06 for minor ST-T abnormalities, with a pvalues across categories of 0.044. At Year 0 and Year 7, no ECG abnormality was significantly associated with marijuana use . Further adjusting for cardiovascular risk factors did not change results . We found no association between alternative categorizations of marijuana use and prevalent major or minor ECG abnormality after multi-variable adjustment. When we compared >2 marijuana-years of cumulative use to never use, the multi-variable adjusted OR for composite major ECG abnormalities was 0.70 and the multi-variable adjusted OR for composite minor ECG abnormalities was 1.03 . When we compared daily marijuana use to never use, the proportion of composite major abnormalities was no higher . Because few daily users had major abnormalities , we did not fit logistic regression models. For composite minor ECG abnormalities, the multi-variable adjusted OR was 1.72 . We found no association between current marijuana use and ECG abnormalities after adjusting for cumulative marijuana use . In stratified analyses by sex and race, black women with 0.5 to 2 marijuana-years of cumulative exposure had a multi-variable adjusted OR of minor ST-T abnormalities of 2.40 , with a p-value across categories of 0.10 . We found no abnormality associated with cumulative marijuana use in stratified analyses by sex and race at baseline or Year 7 . Current use at Year 20 was not associated with prevalent or incident ECG abnormalities in stratified analyses . Whether we applied IPAW or not, and used LVCBF or multiple imputation, results were virtually unchanged.

We found no evidence that current or lifetime cumulative use of marijuana was associated with a higher prevalence or incidence of major or minor ECG abnormalities in this cohort including black and white participants, although major ECG abnormalities seemed to be less frequent in current marijuana users. In this population, we also observed the tendency towards more minor ECG abnormalities compared to never marijuana users. Whether participants used marijuana daily, in the last 30 days or intermittently over a lifetime, marijuana use was not associated with an increase in prevalent or incident specific ECG abnormalities by middle-age. Applying different classifications of marijuana use did not change our results. Our findings did not vary by sex and race. Unlike some small experimental studies from the USA in the late 1970’s that, in samples of around 10 people, suggested marijuana was associated with some specific ECG abnormalities, we found these were just as frequent in current or cumulative marijuana users as in never users. The small sample size, brief exposure of participants to THC, short follow-up, and inclusion only of young, healthy men makes it difficult to draw useful conclusions on the population level from these experimental studies. We found multiple case-reports from the early 2010’s about ECG abnormalities following marijuana use. In our large biracial 20-year cohort of women and men participants who reported a broad variety of marijuana use habits, from never users to daily users, we found no evidence to support an association between marijuana and any ECG abnormality, incident abnormalities in new marijuana users, or abnormalities that would indicate past, ongoing, or future myocardial infarction. Our findings align with earlier epidemiological research on thousands of participants from Europe and the USA, including participants of the same cohort, that found no association between marijuana and CVD, mortality or other measure of sub-clinical atherosclerosis. When we stratified results by sex and race, we noticed that black participants presented with a higher proportion of ECG abnormalities than white participants, regardless of their marijuana use habits, but black marijuana users had no more ECG abnormalities than black never users; likewise, white marijuana users had no more ECG abnormalities than white never users. The assessment of exposure to marijuana was not validated by biological markers, so we rely on participant self-reports. Marijuana use was illegal during the whole course of the study and we cannot exclude social desirability bias, but because participants were queried about marijuana and other illegal drug use at each clinical visit, we could track past exposures. With this method, 84% percent of participants reported any past marijuana, suggesting that possible social desirability bias was mitigated by the trust participants had in the study personnel to report their true exposure. The low number of daily marijuana users at Year 20 in CARDIA limited our ability to fit multi-variable adjusted models, and our results should be carefully interpreted in this population. Future studies with higher sample size will be better equipped to assess the association between daily marijuana use and ECG abnormalities. Also, immediate effects of marijuana might not be reflected on resting ECGs performed hours or days after its use. We rely on marijuana use information provided by participants every 2 to 5 years, and participants only reported on how many days they had used marijuana within the last 30 days. Our analyses cannot inform on the acute effects of marijuana use on ECG. Previous experiments suggested acute effects of marijuana use on ECGs, with conflicting results . Further experimental studies, especially among people with underlying risk of CVD, are needed to test the effects of acute marijuana use on ECG abnormalities. Though the cohort we studied was racially diverse and spanned 20 years, our analysis was limited mainly to a middle-aged population, where CVD is not yet common.