The postrelease mortality rate was extraordinarily high (71.1 deaths per 1,000 PY) in this sample of men with OUD and HIV. Although few studies provide an opportunity for direct comparison, studies in populations with one or more of these risk factors have documented far lower mortality rates. For comparison, mortality among released prisoners globally is estimated at about 10 deaths per 1,000 person-years [9, 58]. Mortality rates lower than ours also have been observed in people with HIV who use opioids (28.6 deaths per 1,000 PY), released prisoners with untreated OUD in Australia (36.7 deaths per 1,000 PY)  and Taiwan (26 deaths per 1,000 PY) , and even in released prisoners with HIV in French Guiana (33.8 deaths per 1,000 PY)  and the U.S. (28.6 deaths per 1,000 PY) . Only one study, a cohort of men with HIV released from prisons in Indonesia, documented higher mortality rates (215 deaths per 1,000 PY), but was based on a relatively small sample . Elevated mortality rates in our study and the Indonesian cohort are unsurprising, perhaps, given the combined risk of OUD, HIV, and prison release in these cohorts.
The few existing studies of prerelease methadone that have examined postrelease mortality have generally shown some survival benefit, which stands in contrast to our study that failed to detect a postrelease survival benefit. Two large cohort studies in Australia and England concluded that prerelease OAT increased survival, but only in the first postrelease month [14, 40]. A three-arm randomized trial in the US showed one postrelease death in each of the two groups treated with methadone (immediately before or after prison release) compared to six deaths in a group that received neither . In one other trial in the U.S. examining within-prison methadone, postrelease mortality was negligible .
Our study differs from prior research on prerelease methadone and postrelease mortality in that it included only people with HIV. This difference in inclusion criteria alongside critical differences in social and epidemiological context in Malaysia versus higher-income settings may explain in part why we did not detect any postrelease survival benefit from prerelease methadone. The benefits of prerelease methadone have mainly been demonstrated in high-income countries where life expectancy for people with HIV is higher and OAT effectively prevents relapse to opioids and overdose [31–35], which is the leading cause of postrelease mortality in these settings . By contrast, our sample was drawn from a prison population with numerous HIV-related health risks and perhaps fewer risks related to opioid use. In Malaysia, while non-fatal overdose among people who use opioids is common , rates of fatal overdose are unknown, as are the relative contributions of HIV and opioid-related mortality in this population. For men with OUD and HIV released from prison in Malaysia, the risk of death associated with advanced HIV disease may supersede the risks associated with opioid use disorder, treated or not.
We cannot exclude overdose as a cause of death due to possible underreporting, yet no cases of fatal overdose were reported in our data. Our findings are consistent, however, with the theory that inadequately treated HIV, not overdose, is the primary contributor to postrelease mortality in this population. As further evidence, baseline CD4+ T-lymphocyte count strongly predicted mortality, and most causes of death recorded by pronouncing physicians identified some underlying infectious etiology, many of which represented AIDS-defining illnesses. Finally, we did not observe markedly higher mortality rates in the period immediately after release, as would be expected if deaths were caused by opioid overdose .
While evidence from other settings has shown that OAT can improve HIV treatment outcomes , this may not have occurred in our sample. Reasons may include that despite initiating methadone in prison, participants may not have adhered to it or may have discontinued it postrelease. Additionally, linkage to HIV care postrelease may have been poor due to structural barriers. In some high-income settings, postrelease linkage to ART has been found to be low , and this is also likely to be the case in Malaysia, where the number of people with HIV receiving treatment is low despite ART being fully subsidized . Provider discrimination may play a role: more than half of sampled Malaysian HIV care providers in one study were unwilling to start ART in patients who injected drugs or were recently released from prison, even at very low CD4+ T-cell counts [63, 64]. Improving access to quality HIV care for people who use drugs in Malaysia will be a crucial step to reducing health disparities in this population .
Factors related to methadone implementation also may explain why our estimates of the impact of prerelease methadone on postrelease mortality were consistent with a null effect. First, participants allocated to methadone may not have reached an optimal dose before release for reasons including inability to tolerate the titration schedule or being released early before completion of the titration . An as-treated analysis defining treatment as receipt of at least 60mg of methadone prerelease, however, also was consistent with a null effect. Second, some participants allocated to receive methadone may not have been linked successfully to methadone treatment after release, and some participants not allocated to receive prerelease methadone may have initiated community methadone treatment postrelease. A main study limitation is the absence of complete data on methadone utilization postrelease. Fourth, because the study was not powered to detect survival differences, absence of evidence for a treatment effect could be due to low power. Fifth, selection bias among non-randomized participants due to factors omitted in our adjustment strategy could have biased our estimates toward a null treatment effect. We did not, however, detect an effect in the subsample of randomized participants, where selection bias was not an issue, though our power to detect an effect in this subsample was lower. Finally, we were unable to distinguish deaths that occurred in the community from deaths that may have occurred in prison among participants who were reincarcerated during the postrelease observation period. It should be noted that methadone could have positively impacted other health and social outcomes in our sample including potentially by reducing injection drug use, but this was not the focus of this analysis.
Of interest, pre-incarceration alcohol use, reported by 19.2% of participants, was predictive of postrelease mortality. In people with HIV, alcohol use is associated with increased mortality . Alcohol use also has been implicated in overdose deaths in people who use opioids [24, 66], and in the deaths of people receiving methadone for opioid use disorder [67, 68]. One promising intervention is extended-release naltrexone, which has been associated with improved postrelease HIV treatment outcomes separately in people with opioid and alcohol use disorders [69, 70].