Background: We used community demographic data with information on HIV status, diagnosis and treatment from studies that comprise the Network for Analysing Longitudinal, Population-based HIV/AIDS data on Africa (ALPHA) to estimate mortality among people who had started ART by duration of infection, measures of morbidity and clinic attendance.
Methods: We used data on adults aged 15-59 from Kenya (Kisumu), Malawi (Karonga), South Africa (uMkhanyakude), Tanzania (Kisesa), Uganda (Masaka and Rakai) and Zimbabwe (Manicaland).
We estimated population-based mortality rates by time since ART initiation for people who initiated treatment whilst resident in one of the study areas, based on self-report in surveys or clinic data linked to individual-level demographic surveillance system (DSS) data. Deaths were observed in the DSS. Data on clinic visits, ART prescriptions and refill dates were used to describe timeliness of clinic attendance (on ART and on time for last refill; on ART but late for last refill; not on ART).
We fitted piecewise exponential models for men and women with time since initiation split into < 6 months, 6-11 months, 12-23 months, 24-35 months and >36 months. Time-varying covariates included age, calendar year, clinical stage, timeliness of clinic attendance, and for a subset of observed sero-convertors, duration of infection. Other covariates were sex, CD4 count at initiation and study.
Results: ART initiation was observed for 2,811 men and 7,153 women. 405 men and 519 women died (mean exposure 2.7 person-years).
Between 2005 and 2014, the mortality among people who had initiated ART fell (Table 1). This decline were seen at all times following ART initiation, in all studies and at all ages.
In three studies, we could adjust for CD4 at initiation and (time-varying) timeliness of clinic attendance and clinical stage. For men, the effect of calendar year was stronger after adjustment (HR 0.82 (p=0.001) versus 0.85 (p=0.002)).
Conclusions: Mortality among adults post-ART initiation declined substantially between 2005 and 2014. These declines are not immediately explained by more timely presentation or better attendance for treatment. There was rapid expansion and decentralisation of services in these resource-poor rural settings in this period. Treatment may have improved and contributed to improved survival.