When an association remains after accounting for a proposed mediator, it is tempting to declare a direct biological effect. A historical HIV cohort shows why that step requires restraint. Measuring medication adherence imperfectly is different from experimentally excluding adherence as an explanation.
Who was studied?
Feelemyer and colleagues analyzed 645 HIV-positive people who injected drugs in Hai Phong, Vietnam, with follow-up over 24 months. Approximately 95% were men. Nearly the entire population injected heroin; fewer than 1% reported injecting methamphetamine. Methamphetamine use was predominantly through smoking, so the cohort label must not be mistaken for its methamphetamine route.
Analysis required reported antiretroviral therapy, adherence data and laboratory viral-load measurements. Recent methamphetamine exposure was self-reported days of use during the preceding 30 days: none, 1–19 days, or at least 20 days. It was not a controlled-dose exposure.
What did viral suppression mean here?
The investigators defined suppression as HIV viral load below 1,000 copies/mL, measured from banked blood samples. About 76% were suppressed at baseline. The threshold is essential: this study’s definition should not be silently replaced with an undetectable result or another clinical cutoff.
During follow-up, recent methamphetamine use was associated with lack of suppression in the adjusted random-effects analysis: odds ratio 1.84, 95% confidence interval 1.06–3.17. That is an odds ratio for the specified outcome, not a direct multiplication of every participant’s probability.
How was adherence measured?
Participants rated adherence over the preceding six months on a 1–10 visual scale. Scores 9–10 were classified as high or complete adherence; 1–8 formed the lower category. This was self-report, rather than continuous observation of ingestion or a drug-concentration measure.
The adherence and exposure windows also differed: six months for adherence versus 30 days for recent methamphetamine use. Their relationship therefore cannot be interpreted as a precisely synchronized day-by-day experiment.
What did the mediation analysis establish?
In this 2020 analysis, self-reported adherence did not statistically explain the methamphetamine–viral-load association. The association remained after the adherence measure was included. That result is specific to the variables, measurement windows and model used.
It does not prove that adherence played no role. A broad self-rating may miss interruptions, timing differences or other medication-taking patterns. Residual association could also reflect unmeasured factors or other pathways. Naming a direct effect in a statistical model does not, by itself, identify a molecular mechanism.
What limits the interpretation?
The authors identified possible underreporting of drug use and adherence difficulties, uncertainty about generalizability, and approximately 16% loss to follow-up. Losses may matter when retention relates to exposure or health outcomes. This selected cohort does not represent every person with HIV or every pattern of methamphetamine use.
The paper discussed biological hypotheses, but did not experimentally test methamphetamine’s effect on antiretroviral metabolism in these participants. It supplies a historical clinical association and a constrained mediation result, not a current treatment recommendation or proof that an individual patient’s virological result was drug-caused.
Source
Feelemyer J and colleagues, DRIVE Study Team. Associations between methamphetamine use and lack of viral suppression among a cohort of HIV-positive persons who inject drugs in Hai Phong, Vietnam. AIDS. 2020;34(13):1875–1882. DOI: 10.1097/QAD.0000000000002680. Full text; bibliographic record.