Our findings show that three years after the start of ART clinics in rural, northern Malawi there has been a substantial reduction in all-cause and AIDS mortality among people aged 15–59 years old. For all-cause mortality, our best estimate is that mortality has fallen by one-third during the second and third years of ART roll-out in the district, compared with the pre-ART period, and for AIDS mortality our best estimate is that it has fallen by around 57%. These dramatic reductions have occurred with the adoption of a “public health” approach to ART delivery, which has enabled a rapid increase in treatment coverage and accessibility of care while still achieving good treatment outcomes
It is striking that the overall mortality reduction is much larger since an ART clinic opened within the study area, making treatment more accessible than it was during the first year of ART roll-out. From the start of the third year of ART roll-out in the district, uptake of ART was also enabled by the sero-survey: 65% of adults aged 15–59 years old learnt their HIV status, several hundred were referred to HIV care services, and 52 initiated ART between the end of January and the end of September 2008. The cohort study provision of consistently available CD4 counts at the local clinic may also have enhanced treatment access compared with other areas of Malawi. Even with the opening of the ART clinic in the DSS area, the reduction in all-cause and AIDS mortality remained greater in areas that were closer to the tarmac road, providing additional evidence of the importance of making services as accessible as possible. The relatively low HIV prevalence in the more remote areas also contributes to the lower reduction in all-cause mortality in these areas.
Our estimates of the reduction in all-cause mortality are compatible with estimates of ART need, known levels of ART uptake, and known levels of survival on ART among individuals registered at the clinic that provides ART, in the study area. We have estimated that around 355 individuals were taking ART, with around 508 individuals in need of ART, during the time period covered by the 2007/8 sero-survey. Given that 63% of deaths were attributed to AIDS during the period 2002-5, then if we assume treatment coverage of 70% (355/508), and that the “efficacy” of ART during the first 2–3 years of ART roll-out is to avert 75% of AIDS deaths among individuals on ART (which is broadly compatible with published data on retention in care in Malawi and with data from the clinic in the DSS area), then the proportion of deaths in the population averted by ART provision (the population attributable fraction) would be 33% (i.e. 0.63×0.70×0.75), very similar to the observed reduction.
Other evidence that observed reductions in mortality are due to ART is that there was little change in either non-AIDS mortality or in mortality among individuals aged 60+ years old during the period following ART roll-out. The slight increase in non-AIDS mortality during the period following ART roll-out may be due to misattribution of cause of death. Verbal autopsy reviewers were more likely to have access to information on HIV status in the time period following ART roll-out; as a consequence, the number of HIV-negative individuals whose cause of death was wrongly assigned to AIDS (because they had an AIDS-like clinical condition) will have fallen. Nonetheless, our estimates of 6.4 per 1000 person-years for the pre-ART AIDS mortality rate (corresponding to 6.4 per 100 person-years among HIV-positive individuals if we assume HIV prevalence of 10%), and 3.8 per 1000 person-years for the non-AIDS mortality rate, are broadly in line with published data from Southern and East Africa on the mortality of HIV-positive and HIV-negative individuals
. This provides confidence that, at the population-level, verbal autopsies are a useful tool for estimating the proportion of deaths that are due to AIDS.
Our findings are also in line with other studies of the population-level impact of ART on AIDS mortality in the early years of ART roll-out. The estimated 57% (95% CI 44–67%) reduction in AIDS mortality with ART coverage of around 70% found in our study is consistent with a 50% reduction in Addis Ababa during 2 years in which ART uptake was apparently very high
as well as the approximately 25% reduction in a rural area of South Africa with ART uptake averaging around 40% of those in need during the first 3 years of ART roll-out
, and a 37% fall in registered deaths in a district of southern Malawi with treatment coverage estimated to be around 80%
. Overall, our findings support the roll-out of ART treatment to health centres and clinics with unsophisticated facilities: we believe that increased access to ART is the primary reason for the sustained fall in all-cause and AIDS mortality, although the sero-survey conducted in 2007/8 made a contribution and other temporal changes in health-seeking behaviour might also have made a difference.
A key question is whether the reduction in all-cause and AIDS mortality can be sustained or even increased. ART coverage in the study area is already around 70% using current treatment criteria in Malawi, but recently adjusted criteria will see the numbers of HIV-infected adults who are eligible for ART increase
, while increased access to HIV testing and increased confidence in the health care system will enlarge the numbers seeking care. Set against this, enhancing access may be constrained because many clinics in Malawi are already operating at high capacity
, and further task-shifting will probably be required
. Roll-out of ART to the next layer of health clinics and health posts seems justified in order to further increase access and limit the impact of the ART programme on hospital facilities, whilst increasing the capacity to manage the rising number of HIV-infected individuals who need treatment. Localising care is also expected to enhance adherence to therapy and retention in treatment programmes.
It has been recognised that the early years of ART roll-out are a “honeymoon” period during which it is possible to avert a high proportion of AIDS deaths
. Dramatic mortality reductions at the population level can be sustained only if the survival of individuals on ART is prolonged for as long as possible, rather than for only a few years, and if HIV prevention measures are implemented at the same time since HIV prevalence will increase due to ART provision unless HIV incidence simultaneously falls.
In conclusion, our study shows that ART can have a dramatic effect on mortality in a resource-constrained setting in Africa, at least in the early years of treatment provision. Continued monitoring of this effect is essential, to confirm if it can be sustained and to anticipate and plan changes to continue this successful intervention. Continued funding to maintain and further scale-up treatment provision will bring large benefits in terms of saving lives.