We hypothesized that higher levels of current marijuana use and greater cumulative marijuana exposure would be associated with worsening cognitive processing speed and flexibility and the magnitude of effects will be greater in HIV+ compared to HIV- men. The MACS is an ongoing prospective cohort study of the natural and treated history of HIV infection among Men who have Sex with Men in the United States. 6,972 men were enrolled during the history of the study in three waves: 4,954 men in 1984–1985, 668 in 1987–1991, and 1350 in 2001–2003 and at 4 centers located in Baltimore/Washington DC, Chicago, Los Angeles, and Pittsburgh. The study design of the MACS has been described previously and only the design relevant to the current analyses are described here. Participants return every 6 months for physical examinations, HIV testing, laboratory testing, structured clinical interviews, collection of data on cigarette smoking, alcohol use, illicit drug use and cognitive function assessments. The institutional review boards at the respective study centers approved the MACS study protocols and all participants provided informed consent. Participants were eligible for the current study if they had two or more cognitive function assessments over the study period . Furthermore, HIV+ individuals were eligible if they initiated highly active antiretroviral therapy and reported continuous use for one year. Exclusion criteria for all participants included history of: a learning disorder , stroke, seizures, peripheral neuropathy, multiple sclerosis,flood table and head injury with loss of consciousness greater than 1 hour. Socio-demographic characteristics included participant’s selfreported age, race/ethnicity status and educational attainment. Study participants were classified according to the MACS study center and MACS cohort status .
Depressive symptoms were evaluated at every study visit with the Center for Epidemiologic Studies Depression scale . Current alcohol use was self-reported at every study visit and categorized. We computed cumulative exposure to alcohol in drink-years with 1 drink-year equivalent to consuming a standard drink of alcohol every day for a year. We calculated the average number of drinks consumed per week for each participant by multiplying the average number of drinking-days per week by the average number of drinks consumed per drinking-day. Alcohol drink-years was computed by adding the total average number of drinks consumed during all follow-up visits . Cigarette use was self-reported at every study visit. We categorized current smoking status at every visit into three groups: never former and current smoker. Cumulative exposure to cigarettes was computed and defined in pack-years, with one-packyear of exposure equivalent to 7300 cigarettes . Stimulant/recreational drug use was selfreported at every study visit. Participants were considered to be users of stimulant drugs if they reported the use of: crack cocaine, other forms of cocaine and methamphetamines and ecstasy. Participants self-reported their frequency of use of poppers using similar response options as for marijuana use, but categorized as any use in the past six months. We used similar approach for marijuana use-years to compute cumulative exposure to stimulants and poppers in use-years. Participants were categorized as having a history of injection drug use , if they self-reported ever injecting any substance. Hypertension was assessed at every visit and was defined as systolic blood pressure greater than 140 mmHg, or diastolic blood pressure greater than 90 mmHg or diagnosed with hypertension and use of medications. Diabetes status was classified using a combination of HgA1C values ≥6.5 and diagnosed with diabetes and use of medication.
Participants were classified at each MACS study visit as HCV positive if they were found to be in the process of seroconversion, acute infection, chronic infection, clearing , or previously HCV positive, but now clear of HCV RNA. HIV-serostatus was assessed using enzyme-linked immunosorbent assay with confirmatory Western blot tests on all participants at each participant’s initial study visit and at every semiannual visit thereafter for participants who were initially HIV– to confirm their serostatus. Plasma HIV RNA concentrations were measured using the COBAS Ultrasensitive Amplicor HIV-1 monitor assay for HIV RNA , with a sensitivity of 50 copies of HIV per RNA/mm3 . Standardized flow cytometry was used by each MACS center to quantify CD4+ T-lymphocyte subset levels . Antiretroviral therapy and ART adherence was self-reported at every study visit. ART was classified as none, nucleoside reverse transcriptase inhibitors , protease inhibitors , and non-NRTIs . Adherence to ART was assessed in the MACS beginning from October 1998 using a scale measuring four levels of adherence, which has been described previously . ART use prior to October 1998 was considered 100% adherent. We computed cumulative years of each class of ART at each study visit, weighted for self-reported adherence. The weights for the four levels of adherence were 1, 0.975, 0.85, 0.375, 0 for adherence levels of 100%, 95–99%, 75–94% and less than 75%. History of clinical AIDS was determined according to the 1993 CDC definition of AIDS . Data analyses was conducted from April 1, 1996, to September 30, 2013. April 1, 1996 was chosen as the baseline because that was when most men in the MACS initiated HAART . We used linear mixed effects models to test associations between current and cumulative exposure to marijuana and changes in cognitive function measures using SAS PROC MIXED, to account for correlations between repeated cognitive function measures over time, from the same participants. We specified an unstructured covariance matrix for the repeated outcomemeasures as this achieved the best model fit compared to other covariance structures . We used robust standard errors from the robust empirical covariance estimator. Time since baseline was used as the longitudinal metric for time.
Models included both linear and nonlinear time trends. We fit linear mixed effects models over the 17-year follow-up period, using maximum likelihood estimation and allowing for random intercepts and random slopes to account for individual differences in baseline cognitive function and to allow for subject-specific rates of cognitive change. We performed stratified analysis by HIV-serostatus. We modeled each cognitive function outcome separately on current marijuana use, as well as on cumulative exposure to marijuana. The primary coefficients of interest were interactions between current and cumulative marijuana-use-years with time. The model for the HIV- men adjusted for time-stable covariates and time-varying covariates . The models for the HIV + men additionally adjusted for time-stable and time-varying HIV-specific parameters . The models assessing cumulative exposure to marijuana use and cognitive function included time-varying cumulative exposure variables including pack-years of smoking, alcohol drink-years,indoor plant table stimulant and popper use-years. We log transformed test scores from the TMTA and TMTB to approximate a normal distribution. To facilitate interpretation, we transformed the regression coefficients of the TMTA and TMTA models using the formula, 100 where β is the regression coefficient . Because the longitudinal metric for time was measured in years, the coefficients can be interpreted as annual percent change in test scores across time. We used inverse probabilities of attrition weights to adjust for selective attrition . Cohen’s f 2 effect sizes was calculated to understand the magnitude of the associations between current and cumulative exposure to marijuana use in all models. Cohen’s f 2 was calculated using SAS PROC MIXED procedures introduced by Selya et al , with f 2 ≥ 0.02, f 2 ≥ 0.15, and f 2 ≥ 0.35 representing small, medium, and large effect sizes, respectively . All statistical analyses were conducted in SAS version 9.4 . Statistical tests for significance was defined as p <0.05.Among HIV+ men only, there were no statistically significant association between cumulative marijuana use-years and changes in any cognitive function domain . Conversely, in the HIV- men only, each additional 5 marijuana use-years was significantly associated with a decline in TMTA scores by 0.18 percent annually . Similarly to the findings for current marijuana use, all effect sizes were very small, falling below Cohen’s f 2 criteria for a small effect size and were similar in the HIV+ and HIV- men. In exploratory analysis, we tested for interactions between history of AIDS, detectable viral load and CD4 counts by current and cumulative marijuana use but the results were not significant . In all models, the most consistent set of covariates that were associated with increased rate of decline across cognitive function tests was advancing age, non-white race and lower education. Supplemental tables 1–4 show the full model estimates stratified by HIV-serostatus. In addition, we conducted a series of sensitivity analyses to assess the impact of multiple imputation on our findings. Specifically, we re-ran all our models without imputing missing marijuana values and compared it to our extant results; and the results for current and cumulative marijuana-use-years remained relatively consistent .
In this analysis of HIV+ MSM in the MACS followed for 17-years, we found current monthly and daily marijuana use to be significantly associated with slowed cognitive processing speed, but not cognitive flexibility. Additionally, we found no significant associations between cumulative exposure to marijuana and changes in cognitive processing speed and flexibility. Among the HIV- MSM, we found no statistically significant association between current marijuana use across all cognitive function measures, although, each additional 5 marijuana-use-years was associated with significant decline in one measure of cognitive processing speed. Our findings of significant associations between current monthly and daily marijuana use with slowed cognitive processing speed differ from other studies of HIV+ individuals that found no significant associations . For instance, Thames et al. in a small crosssectional study of 89 HIV+ and HIV- subjects found that HIV+ subjects with moderate-toheavy marijuana use demonstrated no significant associations with slowed processing speed than none-users . In a more recent study, Thames et al. found no significant difference on tasks of processing speed among HIV+ subjects when levels of marijuana use increased over 1.4grams/week . These studies were cross-sectional with modest sample sizes compared to our study which used cognitive function assessments at multiple time-point from a large sample. However, our study found no significant associations between current marijuana use and decline in cognitive flexibility, which is consistent with other studies of HIV+ individuals that assessed this cognitive domain . Our findings of no significant association with current marijuana use and processing speed and flexibility in HIV- men mirror findings from other longitudinal studies of adults conducted in the general population . For example, one longitudinal study of 2,404 adults, 22 years of age at baseline followed for 8-years in an Australian cohort found no significant differences in performance on tasks of processing speed in some marijuana using groups versu none-using groups . Similarly, another longitudinal study of 1,897 adults, mean age at baseline of 42 years followed for nearly 8 years found no significant differences in performance on tasks of cognitive processing speed between weekly or more and less than weekly marijuana use in the past year compared to nonuse . In addition, we note that all of the coefficients from our models, were of very small magnitude, falling below Cohen’s f 2 criteria for a small effect size suggesting that our findings, likely do not represent clinically meaningful declines in cognitive function. Our study is among the first to longitudinally assess the impact of cumulative exposure to marijuana and changes in cognitive function performance for a 17-year follow-up period. Among the HIV- men, our study found cumulative exposure to marijuana was associated with statistically significant decline in one measure of cognitive processing speed and not in the other . This is inconsistent with studies in the general population that have found no significant associations with cumulative or chronic marijuana use and decline in cognitive processing speed . For example, in a recent cohort study of 3,385 men and women in the Coronary Artery Risk Development in Young Adults study, cumulative marijuana use for 25 years was not statistically significantly associated with worse cognitive processing speed . Although that study assessed cognitive function at a single time-point when participants mean age was ~50 years compared to our study, which had cognitive function assessments at multiple time-points beyond 50 years – when cognitive decline may be more apparent. However, in one study that comprised 1,037 individuals in a New Zealand birth cohort study followed-up for 20 years found that diagnosis with cannabis dependence at 3 or more study waves was associated with widespread declines in cognitive domains including memory, executive function and cognitive processing speed in adulthood .