Brief Intervention Decreases Drinking Frequency in HIV-Infected, Heavy Drinking Women: Results of a Randomized Controlled Trial

Abstract

Objective: Hazardous alcohol use by HIV-infected women is associated with poor HIV outcomes and HIV transmission risk behaviors. We examined the effectiveness of brief alcohol intervention (BI) among hazardous drinking women receiving care in an urban HIV clinic.

Methods: Women were randomized to a 2-session BI or usual care. Outcomes assessed at baseline, 3, 6, and 12 months included 90-day frequency of any alcohol use and heavy/binge drinking (≥4 drinks per occasion), and average drinks per drinking episode. Secondary outcomes included HIV medication and appointment adherence, HIV-1 RNA suppression, and days of unprotected vaginal sex. We examined intervention effectiveness using generalized mixed-effect models and quantile regression.

Results: Of 148 eligible women, 74 were randomized to each arm. In mixed-effects models, 90-day drinking frequency decreased among intervention group compared with control, with women in the intervention condition less likely to have a drinking day (odds ratio: 0.42; 95% confidence interval: 0.23 to 0.75). Heavy/binge drinking days and drinks per drinking day did not differ significantly between groups. Quantile regression demonstrated a decrease in drinking frequency in the middle to upper ranges of the distribution of drinking days and heavy/binge drinking days that differed significantly between intervention and control conditions. At follow-up, the intervention group had significantly fewer episodes of unprotected vaginal sex. No intervention effects were observed for other outcomes.

Conclusions: BI reduces frequency of alcohol use and unprotected vaginal sex among HIV-infected women. More intensive services may be needed to lower drinks per drinking day and enhance care for more severely affected drinkers.

Figure 1

Consort Diagram * Lost to Follow-Up defined as missing all the 3, 6, and 12 month visits.

Figure 2

Frequency of a) of drinking days ) binge drinking days and c) number of drinks per drinking day by intervention and control conditions at baseline and 3, 6 and 12 month follow-up. The 0 and + symbols indicate the mean of the outcome in the control and intervention conditions, respectively. The horizontal short lines in the boxes are the medians. In panel 1, at baseline, mean drinking days in the intervention group was 34 days. At follow-up, the mean was 14–16 drinking days (depending on visit). Thus drinking days were reduced by 52–59% in the intervention group. The baseline mean drinking days for the control group was 30 and reduced to 20–24 drinking days at follow-up, a reduction of 20–33%. In panel 2, heavy drinking days, the baseline mean in the intervention group was 25, and was reduced to 8 to 13 drinks at follow-up, a 48–68% reduction. In the control group the mean at baseline was 27, and was reduced to 13 to 16 heavy drinking days at follow-up, a reduction between 41–52%. In panel 3, the mean drinks per drinking day at baseline was 10 in both groups, and reduced to 6–7 in both groups, a reduction of 30–40%.

Figure 3

Quantile Regression is used to estimate the treatment effects across a range of the distribution (i.e., quantile such as 10, 25, 50 [median], 75, and 90^th percentiles) of the outcomes. For a given percentile (horizontal axis), the vertical axis represents the A) number of drinking days, B) number of heavy/binge drinking days, and C) number of drinks per drinking day. The solid black line represents the treatment effect difference between the control and intervention arms (95% CI represented by gray shaded area). For reference the dashed long line shows the value of the outcome among the controls for a given percentile. For example, at the 80^th percentile panel A shows that the 80% of the controls were below 30 drinking days, whereas the intervention effect is 10 drinks less and therefore the 80% of the intervention group were below 20 drinking days. The light grey lines represent the actual estimated values from the quantile regression to which a non-parametric smoother was applied to provide guidance for interpretation.