BACKGROUND: Alcohol use is measured in diverse ways across settings. Harmonization of measures is necessary to assess effects of alcohol use in multi-cohort collaborations, such as studies of people with HIV (PWH).

METHODS: Data were combined from 14 HIV cohort studies (nine European, five North American) participating in the Antiretroviral Therapy Cohort Collaboration. We analyzed data on adult PWH with measured alcohol use at any time from 6 months before starting antiretroviral therapy. Five cohorts measured alcohol use with AUDIT-C and others used cohort-specific measures. We harmonized alcohol use as grams/day, calculated using country-level definitions of a standard drink. For Alcohol Use Disorders Identification Test (AUDIT-C), we used Items 1 (frequency) and 2 (number of drinks on a typical day). Where alcohol was measured in categories, we used the mid-point to calculate grams/day. We used multivariable Cox models to estimate associations of alcohol use with mortality.

RESULTS: Alcohol use data were available for 83,424 PWH, 22,447 (27%) had AUDIT-C measures and 60,977 (73%) recorded the number of drinks/units per week/day. Of the sample, 19,150 (23%) were female, 54,006 (65%) had White ethnicity, and median age was 42 years. Median alcohol use was 0.3 g/day (interquartile range [IQR] 0-4.8) and 0 g/day (IQR 0-20) for those with and without AUDIT-C. There was a J-shaped relationship between grams/day and mortality, with higher mortality for PWH reporting no alcohol use (adjusted hazard ratio [aHR] 1.46; 95% CI: 1.23-1.72) and heavier (>61.0 g/day) alcohol use (aHR 1.92; 1.41-2.59) compared with 0.1-5.5 g/day among those with AUDIT-C measures. Associations were similar among those with non-AUDIT-C measures.

CONCLUSIONS: Grams/day is a useful metric to harmonize diverse measures of alcohol use. Magnitudes of associations of alcohol use with mortality may differ by setting and measurement method. Higher mortality among those with heavier alcohol use strengthens the case for interventions to reduce drinking.