Brief alcohol interventions in South African HIV care reduce drinking but leave stigma untouched

Introduction

Drinking is not a side issue in South African HIV care. It turns up in who takes up prevention, in which days pills get missed, in how women describe the trip to the clinic. In a population-based analysis of 8,528 women aged 15 to 49, Sileo et al. (2023) found that alcohol use, intimate partner violence and unprotected sex tended to cluster in the same places, the kind of pattern that makes people talk about a syndemic. Among young men in rural KwaZulu-Natal, 146 of the 325 analysed (45%) reported hazardous drinking (Castle et al., 2025). Numbers like these are why alcohol reduction has been added to HIV services, usually as a short counselling intervention delivered inside the clinic (Parry et al., 2023a).

My argument is that these brief interventions do work, but less than the enthusiasm around them suggests, and that the reasons are visible in the South African studies themselves. Drinking falls by a modest amount. The outcomes that matter most to patients and clinicians, such as distress and adherence, are barely moved or barely tested. Self-reported drinking misses a good deal of actual drinking (Parry et al., 2023b), and stigma, which predicts heavier drinking, has not shifted under the interventions tried so far (Belus et al., 2025). The paper looks first at where alcohol and HIV risk meet, then at adherence, then at what brief interventions achieve, and then at measurement and stigma, before saying what services should ask of the next round of programmes.

Where alcohol and HIV risk meet

The population picture is messier than the syndemic label suggests. Sileo et al. (2023) used the 2016 South African Demographic and Health Survey and found geographic clustering of alcohol use, intimate partner violence and unprotected sex among women. HIV status, though, did not cluster with those three. It clustered by itself. One way to read this is that HIV is a slow outcome that reflects years of partner networks and treatment coverage, while drinking and violence are current behaviours that sit in particular neighbourhoods. The authors themselves say more work is needed to link these factors to HIV from a place-based angle, which is a fair admission that the syndemic story is stronger as a prevention rationale than as a proven causal chain. Clinic-based work points the same way for women already in care, with drinking, abuse and HIV treatment sitting in the same lives (Ndirangu et al., 2022; Washio et al., 2021). Population surveys and clinic samples are different animals, of course, so I treat the match as suggestive only.

Men tell a different story, and a more hopeful one in its first half. Castle et al. (2025) followed young men offered oral pre-exposure prophylaxis after referral to sexual health services in KwaZulu-Natal. Uptake was highest among the hazardous drinkers (40%) and lowest among men who did not drink (17%), and after adjustment hazardous drinkers were still about twice as likely to start (adjusted odds ratio 1.96). So drinking did not keep men away from prevention. The second half is the worry. Only 27% of those who started (24 of 88) were still collecting PrEP at four months, and continuation looked much the same whatever their drinking level. Alcohol, then, was not what separated the men who stayed from the men who left, which is a useful corrective for anyone tempted to treat alcohol as the main explanation for poor retention in prevention. Sileo et al. (2023) studied women and Castle et al. (2025) studied men, so the two cannot simply be merged into one South African story.

The picture for pregnant women is thinner still. Washio et al. (2021) reported on 33 women who had a positive pregnancy test at baseline in an implementation study of the Women’s Health CoOp. Some 55% reported drinking in the past month, 27% had a history of physical abuse and 12% a history of sexual abuse. Only five were still pregnant at six months, and those five reported better adherence and less drinking. That is a brief report, a handful of women and no comparison group, so it cannot carry any weight about effectiveness. What it does show, together with Sileo et al. (2023), is that women’s drinking in this setting rarely sits alone. It travels with violence and with HIV care, and any intervention that addresses only the glass is probably addressing the smaller part of the problem.

Drinking days and missed doses

Most of the evidence linking alcohol to non-adherence is aggregate, comparing people who drink heavily with people who do not. Rose et al. (2022) did something more specific. They followed 53 adults in HIV care in Cape Town who had both adherence difficulties and harmful drinking, and looked at alcohol use and antiretroviral adherence on the same day and the next. Adherence was worse on weekends and holidays in general. Next-day adherence was worse after weekend drinking, and it was worse among men. The detail matters because it moves the problem from a person to a calendar. A patient who takes every dose Monday to Thursday and then loses Saturday and Sunday looks fairly adherent in a monthly clinic review, and that is exactly the pattern a quick adherence question at a visit would miss. Sileo et al. (2023) described alcohol as something that clusters in places and populations, while Rose et al. (2022) show it also clusters in time, so both place and calendar may matter when services decide where to put their effort (Ndirangu et al., 2022).

The sample was small and was chosen because of trouble with both adherence and drinking, so I would not stretch these results to people on treatment in general (Rose et al., 2022). Still, they sit comfortably beside the focus group findings of Ndirangu et al. (2022), who spoke with 69 women in Cape Town six months after a Women’s Health CoOp workshop series. These women reported drinking a mean of 13 drinks a day. Their list of obstacles to taking treatment started with substance use but also included money, hunger, anticipated and enacted stigma, community violence, and strained relationships with providers and clinic procedures. The authors concluded that interventions have to address these contextual barriers alongside the behavioural work if gains are going to last. Washio et al. (2021) make a similar case for pregnant women, whose drinking they place beside histories of abuse.

Put side by side, the two studies pull in slightly different directions. Rose et al. (2022) make drinking itself look like the proximate cause of a missed dose, and a time-specific one. Ndirangu et al. (2022) make drinking look like one barrier among several that tend to arrive together. Both can be true, and they probably are, but they imply different clinical responses. The first points to weekend-focused planning, reminders and perhaps peer contact on the days most at risk. The second points to food parcels, safer transport, a respectful reception at the clinic, none of which a counselling session at a hospital can deliver. South Africa’s poverty and violence do not sit outside the HIV programme. They are part of what a brief intervention is competing against.

What brief interventions actually achieve

The strongest local trial is that of Parry et al. (2023a), which recruited 623 of 756 eligible patients on antiretroviral therapy at six public hospital clinics in Tshwane. Participants were randomised to four modules of motivational interviewing and problem-solving therapy, delivered over two sessions, or to usual care. At baseline they had on average 25.2 drinks in the previous 30 days. At six months the intervention group was drinking about 34% less than the comparison group, a difference unlikely to be chance. Of the 305 people assigned to the intervention, 225 (74%) completed all the modules. For a two-session intervention in busy public clinics, that is a credible result and I do not want to talk it down. Retention at six months was 88% in the control arm and 83% in the intervention arm, and a sensitivity analysis restricted to the 299 patients with AUDIT scores of 8 or more gave similar findings (Parry et al., 2023a). The trial leans on a single type of measure, though, which I come back to below (Parry et al., 2023b).

It helps to put the size of the effect in context, though. Hahn et al. (2025) pooled individual participant data from 16 alcohol-intervention trials with people living with HIV (3,559 participants, about 73% of them men) and found lower odds of unhealthy drinking in the intervention groups, with an odds ratio of 0.69. The effect sizes were small in every analysis. When the outcome was the blood biomarker phosphatidylethanol (PEth) alone, the odds ratio was 0.81 and Cohen’s d was 0.12. For self-report alone, the odds ratio was 0.67, with d at 0.22 and noticeably more heterogeneity. The certainty of the evidence was rated moderate. So brief interventions work, as the authors say, but the pooled gain is smaller than most clinicians would guess from a headline such as a 34% reduction, and the biomarker result is smaller than the self-report one. The pooled trials were mostly of men, whereas the Tshwane trial enrolled 57.5% women (Parry et al., 2023a), so the two evidence bases describe somewhat different groups.

The next question is whether less drinking buys anything else. Bantjes et al. (2024) went back to the Tshwane trial data, 622 people, to ask whether the intervention eased depression and psychological distress. Symptoms fell at three and six months in both groups, intervention and control, and the differences between groups were not statistically meaningful. Within the trial, however, people who cut down their drinking did tend to have lower depression and distress scores, which suggests that the link between alcohol and low mood is real but that this particular brief package did not exploit it. This whole result is a little awkward for the usual sales pitch. If the intervention reduces drinking and drinking is tied to depression, one would expect the intervention to reduce depression, and it did not, at least not more than usual care did. Regression to the mean, the attention both groups received, or the intervention simply being too light could each explain it, and the data cannot sort them out. A monthly drinks total could also fall without touching the weekend days that matter for pills (Rose et al., 2022), and persistent stigma might keep mood low whatever the drinking does (Belus et al., 2025). Those are guesses, flagged as such.

Measurement and stigma

Measurement is the quiet weakness under much of this literature. Parry et al. (2023b) compared self-reported drinking with PEth among 309 participants in the Tshwane trial. At six months 51% had PEth levels of 50 ng/mL or above, which is the biomarker threshold they used for hazardous drinking. Only 38% scored as hazardous on the AUDIT and just 13% reported heavy drinking in the past seven days. The AUDIT-C picked up 83% of the biomarker-positive participants, but past-week heavy drinking picked up only 20%. Women had higher odds of underreporting than men at six months (odds ratio about 3.5). A trial whose main outcome is self-reported drinks is working with a ruler that stretches differently for different people, and the primary outcome in the Tshwane trial was exactly that, the number of drinks reported for the past 30 days (Parry et al., 2023a). Hahn et al. (2025) saw the same weakness across trials, with more heterogeneity when self-report was used alone.

This matters for what we believe about the 34% reduction. I am not suggesting the Tshwane result is an artefact, since the people assessing outcomes were masked to group (Parry et al., 2023a) and the pooled biomarker analysis in Hahn et al. (2025) still found benefit. But when the effect shrinks from d of 0.22 to 0.12 once PEth enters the picture, it seems reasonable to treat self-reported reductions as an upper estimate. For South African clinics, where clinic time is tight, the practical question is whether PEth testing is affordable at scale. I do not think the literature answers that yet, and a screening instrument that misses so many of the drinkers is poor ground for deciding who needs help.

Stigma is the other gap, and Belus et al. (2025) put numbers on it. They recruited 60 people with unhealthy alcohol use and suboptimal adherence at two public HIV care sites and randomised them to a peer-delivered behavioural intervention or enhanced usual care. Higher enacted alcohol stigma and higher internalised HIV stigma at baseline separately predicted higher PEth three months later, and internalised alcohol stigma predicted higher PEth at six months. The intervention did not change stigma, and stigma did not explain any effect of the intervention. The sample is small, and these are lagged associations in a trial that was not built around stigma, so the claims have to stay modest. Even so, they match what the women in Ndirangu et al. (2022) described, anticipated and enacted stigma in the community and in the clinic.

Taken together, the studies suggest a loop that brief interventions have mostly ignored. People who drink in HIV care carry two stigmas at once, one for the virus and one for the drinking. The stigma seems to predict more drinking, and more drinking makes adherence harder (Rose et al., 2022; Belus et al., 2025). Women in Cape Town described much the same thing when they talked about anticipated and enacted stigma (Ndirangu et al., 2022). A counselling session about drinking goals cannot touch the loop if no one in the clinic or the neighbourhood changes how they treat the person. Whether peer delivery, community work or staff training would loosen it is not something these studies have tested, which I think is the most important thing left open.

Conclusion

Brief alcohol interventions have earned their place in South African HIV services. The Tshwane trial shows a clear fall in reported drinking after two sessions (Parry et al., 2023a), and the pooled evidence agrees on direction while shrinking the size (Hahn et al., 2025). What the evidence does not yet show is that this fall carries through to distress, to adherence on the weekend days that go wrong, or to retention in prevention (Bantjes et al., 2024; Rose et al., 2022; Castle et al., 2025).

Three things follow for the next round of programmes. First, outcomes should be checked with a biomarker such as PEth in at least a subsample, because self-report overstates the gain and understates the problem, particularly among women (Parry et al., 2023b). Second, the intervention should be timed and tailored, with weekend and holiday support for the people who struggle on those days, and with the hunger, violence and clinic experiences that women describe taken as part of the treatment rather than as background (Ndirangu et al., 2022; Sileo et al., 2023). Third, stigma has to be measured and targeted directly, because it predicts drinking and the current interventions leave it where it was (Belus et al., 2025). Programmes that do those three things might deliver more than the modest effects we have so far. Programmes that do not will probably keep producing the same modest, hard-to-verify reduction in drinks per month, and a patient who is still missing Sunday’s pills.

Reference list

Bantjes, J., Morojele, N. K., Myers, B., Swanevelder, S., & Parry, C. (2024). Does reducing alcohol use among people with HIV alleviate psychological distress and symptoms of depression? A randomized controlled trial in Tshwane, South Africa. AIDS and Behavior, 28(3), 985-992. https://doi.org/10.1007/s10461-023-04205-x

Belus, J. M., Anvari, M. S., Ke, H., Regenauer, K. S., Ma, T., Myers, B., Andersen, L. S., Joska, J. A., & Magidson, J. F. (2025). Effects of HIV and alcohol stigma on biomarker-confirmed alcohol use following a peer-delivered intervention in South Africa. Addictive Behaviors Reports, 22, Article 100640. https://doi.org/10.1016/j.abrep.2025.100640

Castle, A. C., Busang, J., Dreyer, J., Herbst, C., Okesola, N., Chimbindi, N., Zuma, T., Jarolimova, J., Hahn, J. A., Psaros, C., Shenoi, S. V., Shahmanesh, M., & Siedner, M. J. (2025). Alcohol use and the pre-exposure prophylaxis continuum of care among men in rural South Africa: Results from a longitudinal study. AIDS and Behavior, 29(7), 2312-2323. https://doi.org/10.1007/s10461-025-04694-y

Hahn, J. A., Kane, J. C., Fatch, R., Espinosa da Silva, C., Emenyonu, N. I., Scheffler, A., Chirayil, P., So-Armah, K., Kahler, C. W., Conroy, A. A., Edelman, E. J., Woolf-King, S., Parry, C. D., Kiene, S. M., Chamie, G., Muyindike, W. R., Adong, J., Go, V. F., Cook, R. L., Morojele, N. K., Fiellin, D. A., Santos, G. M., Tahir, P., Samet, J., Krupitsky, E., & Allen, I. E. (2025). Alcohol interventions for persons with HIV: Meta-analysis of randomized controlled trials using phosphatidylethanol and self-report. Drug and Alcohol Dependence, 276, Article 112879. https://doi.org/10.1016/j.drugalcdep.2025.112879

Ndirangu, J. W., Gichane, M. W., Browne, F. A., Bonner, C. P., Zule, W. A., Cox, E. N., Smith, K. M., Carney, T., & Wechsberg, W. M. (2022). ‘We have goals but [it is difficult]’: Barriers to antiretroviral therapy adherence among women using alcohol and other drugs living with HIV in South Africa. Health Expectations, 25(2), 754-763. https://doi.org/10.1111/hex.13422

Parry, C. D. H., Myers, B., Londani, M., Shuper, P. A., Janse van Rensburg, C., Manda, S. O. M., Nkosi, S., Kekwaletswe, C. T., Hahn, J. A., Rehm, J., Sorsdahl, K., & Morojele, N. K. (2023a). Motivational interviewing and problem-solving therapy intervention for patients on antiretroviral therapy for HIV in Tshwane, South Africa: A randomized controlled trial to assess the impact on alcohol consumption. Addiction, 118(11), 2164-2176. https://doi.org/10.1111/add.16278

Parry, C. D. H., Myers, B., Londani, M., Shuper, P. A., Nkosi, S., Hahn, J. A., Kekwaletswe, C., & Morojele, N. K. (2023b). Self-reported alcohol use versus phosphatidylethanol in behavioral trials: A study of people living with HIV in Tshwane, South Africa. Alcohol: Clinical and Experimental Research, 47(5), 940-950. https://doi.org/10.1111/acer.15062

Rose, A. L., Belus, J. M., Ma, T., Lee, J. S., Wan, C., De Los Reyes, A., Joska, J. A., Andersen, L. S., Myers, B., & Magidson, J. F. (2022). The relationship between harmful alcohol use and antiretroviral non-adherence in people accessing HIV treatment in Cape Town, South Africa: An event-level analysis. AIDS and Behavior, 26(6), 2055-2066. https://doi.org/10.1007/s10461-021-03552-x

Sileo, K. M., Sparks, C. S., & Luttinen, R. (2023). Spatial analysis of the alcohol, intimate partner violence, and HIV syndemic among women in South Africa. AIDS and Behavior, 27(4), 1339-1349. https://doi.org/10.1007/s10461-022-03870-8

Washio, Y., Browne, F. A., Ndirangu, J., Kline, T. L., & Wechsberg, W. M. (2021). Antiretroviral therapy (ART) adherence and prenatal alcohol use among women who are pregnant with HIV in South Africa. International Journal of Environmental Research and Public Health, 18(14), Article 7446. https://doi.org/10.3390/ijerph18147446