Our study compared risk factors for early and late mortality in PLWHIV commencing ART in sub-Saharan Africa. Our findings suggest that previously established risk factors for mortality, including low CD4 count, WHO stage III or IV, and reduced renal function, may not identify patients at risk of death after more than one year of ART.
WHO stage III and IV are well established predictors of mortality [2, 8,9,10]. While these stages were associated with more than a doubling of early mortality risk in our study, we did not find an increased risk after 1 year of ART. A similar trend was noted by Moore et al.  in a cohort from Uganda comparing patients who died before or after 3 months of ART. There, WHO stage III or IV was associated with a nine-fold increased risk of early mortality, but only a three-fold increase in mortality after 3 months. This apparent waning of the predictive value of WHO stage may relate in part to immune reconstitution inflammatory syndrome (IRIS), which is more common in patients with more advanced immunodeficiency on commencement of ART [12,13,14]. Haddow et al.  reported a cumulative incidence of IRIS of 22.9% in their cohort of adult PLWHIV initiating ART in South Africa, with IRIS accounting for nearly one quarter of total clinical events and deaths within the first six months of treatment. This suggests a meaningful contribution of IRIS to early mortality. However, other research in sub-Saharan Africa has shown that IRIS did not largely contribute to mortality in adult PLWHIV undergoing ART treatment—one study in Uganda demonstrated that 7% of total HIV-related deaths within the first three months of ART were related to IRIS , whereas another study in Johannesburg observed that among IRIS patients, there was a 4.5% IRIS-related mortality rate within 6 months of starting ART . Therefore, the contribution of IRIS in the present study is unclear. Whether late mortality risk could be predicted by restaging patients after a period of ART remains a possibility, however, and is an important avenue for further research.
A baseline eGFR < 90 mL/min/1.73 m2 was associated with a 2.5-fold increase in mortality risk of early, but not late mortality. Reports of other cohorts from the region have also identified eGFR as a significant risk factor for mortality, with lower levels of renal function being associated with greater mortality risk [5, 6]. We were unable to determine if more severe renal impairment was associated with a late mortality risk as few patients in our cohort had a baseline eGFR < 60 mL/min/1.73 m2. As our study used a single measure of creatinine to determine baseline renal function, we were also unable to differentiate patients with acute kidney injury from those with chronic kidney disease.
The presence of proteinuria at baseline was associated with a doubling of both early and late mortality risk. As with serum creatinine, baseline proteinuria status was assessed by a single measure. It may be that dipstick proteinuria is less affected than serum creatinine by factors such as hydration status or muscle wasting and so provides a more reliable assessment than a single measure of serum creatinine. Proteinuria status is also important to identify patients with potential HIV-associated nephropathy, which presents as proteinuria with rapidly progressive renal failure .
The apparent protective effect of hypertension in the early mortality group was likely artefactual, a consequence of normotensive and hypotensive patients being grouped together. Hypotension has previously been shown to increase mortality risk in a cohort of PLWHIV from Kenya, with increasing severity of hypotension linked to increasing mortality risk. However, this association was not observed in the subset of patients with advanced HIV disease . While hypertension was not associated with late mortality risk in our cohort, this may have been due to a median follow-up time of only 2.9 years. However, deleterious changes from hypertension have been shown to occur within this timeframe. In previously published data from this same cohort, hypertension predicted renal function loss over a period of only six months . These findings suggest the clinical importance of identifying those with hypotension, as this may contribute to elevated early mortality risk.
A higher body mass index was a protective factor for early mortality, as noted in other cohorts of PLWHIV commencing ART in sub-Saharan Africa [2, 3]. A lower baseline BMI has also been associated with increased mortality within the first 3 months [21,22,23]. Zachariah et al.  observed a linear trend between increased mortality risk within the first 3 months and reductions in BMI below 18.5–16 kg/m2, with a marked increase in mortality risk once BMI falls below 16 kg/m2. This is likely a result of those with lower BMI representing a complex interaction between undernourishment in addition to HIV-associated wasting, as sub-Saharan Africa is disproportionately affected by both food scarcity and HIV infection . Together, these factors increase disease progression, likely contributing to the exaggerated early mortality observed in PLWHIV with low BMI initiating ART in sub-Saharan Africa [21,22,23]. One study from Zambia indicated a significant prevalence of wasting in a cohort of people with severe HIV initiating ART, with 34% of the cohort having a BMI of < 18.5 kg/m2 at baseline . The increased early mortality observed in sub-Saharan PLWHIV initiating ART is believed to be a culmination of immune dysfunction secondary to malnutrition, metabolic derangements resulting from an increased basal metabolic rate and/or selective skeletal muscle atrophy, as well as increased burden of opportunistic infections .
The main limitation of this study is that it examined ambulatory patients at a single clinic in Zimbabwe, and so may not be generalizable to other populations in sub-Saharan Africa. Reports of other cohorts have found similar protective benefits of a higher BMI, as well as increased risk of mortality associated with reduced eGFR, CD4 count, and WHO Stage III and IV [2, 3, 27]. These studies also had similar baseline BMI, mean CD4 count and eGFR compared to the present study. Therefore, it is likely that the present findings are generalizable to the larger sub-Saharan African PLWHIV population.