Lives saved by Global Fund-supported HIV/AIDS, tuberculosis and malaria programs: estimation approach and results between 2003 and end-2007

Only in some five years, the Global Fund to Fight AIDS, Tuberculosis and Malaria (Global Fund) has become a major player in international health development. In 2005, the Global Fund provided 21% of international funding for HIV/AIDS programs in low- and middle-income countries, 67% for tuberculosis (TB), and 64% for malaria[1]. By December 2007, it has approved a total of US$ 10.1 billion proposals, in more than 550 grants in 136 countries, of which US$ 4.8 billion had been disbursed to recipients in 134 countries[2]. As control effort of the three diseases is scaled up, there is now considerable interest to show the scale of health benefits from interventions supported.

For evaluation of disease impact, Global Fund recipients collect epidemiological data on relevant changes in mortality and morbidity. Given the nature of the diseases and interventions and data collection mechanisms, it may take several years before impact becomes detectable. For example, reductions in new HIV infections are inferred retrospectively from prevalence trends over preceding years. WHO and UNAIDS regularly publish regional-level HIV incidence and prevalence estimates, using country surveillance and survey data that often require two years to be released [3‐7]. Similarly, all-cause under-five mortality, a key impact indicator for malaria in areas of high endemicity, is often measured in population-based surveys with 3-5 year intervals[8]. Malaria-related mortality reductions become apparent only 3-5 years after they start, with the 'birth history' method[9]. The consensus TB estimation model, applied to high-burden countries by WHO with national TB programs[10, 11], relies on patient cohort data about treatment outcomes from two preceding years.

In view of these data limitations, it is still early for the impact of Global Fund support to be fully measured. As an interim approach, we estimated the scale of lives saved, by summarizing deaths averted from key service delivery results reported by recipient programs. Interventions considered are antiretroviral treatment (ARV), Directly Observed TB Treatment, short-course (DOTS) and insecticide-treated mosquito nets (ITN). Estimates of their efficacies against mortality are available[7, 12‐17]. This paper presents and discusses the methodology for this approach with estimates of lives saved through service deliveries reported to the Global Fund between 2003 and end-2007.

Lives saved and, for ARV, life-years gained, were estimated using mortality estimation models and assumptions used by UNAIDS, WHO Stop TB and Roll Back Malaria. Models were adapted to use as main input service delivery results available from recipient countries (Table 1 and below).

The presented estimates of lives saved and life-years gained by Global Fund-supported ARV, DOTS and ITNs indicate the magnitude of mortality reductions achieved by recipient programs. While not being an actual measurement or formal evaluation of program impact, the estimations allow the Global Fund to gauge progress as service delivery reports from programs accumulate.

The estimates do not reflect the whole spectrum of interventions supported. Notably, HIV infections averted by prevention services will in future save lives from AIDS, but cannot readily be estimated from routine delivery reports. Effective HIV prevention often results from a multi-sectoral package of services[33‐36], and infections prevented for a given level of intervention coverage vary among countries depending on local HIV incidence rates and transmission modes. Methods to estimate HIV infections averted need urgent development[37].

For malaria, artemisinin-based combination treatments (ACTs) probably have important mortality effects. By December 2007, Global Fund-supported programs had delivered over 24 million ACT courses, across 49 countries. The effectiveness of ACTs against mortality has not been established through randomized trials and will depend on the timeliness of and compliance with treatment; proportions of ACTs for true malaria episodes (rather than misused for non-malarial fevers); possible effects of ACTs on preventing repeat episodes due to recrudescent infection; and the comparison scenario (no treatment or ineffective mono-therapy). The presented model furthermore excluded indoor residual spraying (IRS). As more knowledge becomes available, ACT and IRS should be included in lives saved estimation.

The gap-filling nature of Global Fund financing makes it difficult to define a specific contribution as distinct from other donors and (pre-)existing in-country efforts. Some programs, particularly HIV and TB, report program-wide results covering all contributions from multiple domestic and international funding sources to the Global Fund, which promotes program supports rather than a project approach. Lives saved result from collective efforts of in-country programs, to which the Global Fund contributes alongside other partners. For malaria, on the other hand, total service deliveries covering all partners' contributions considerably outnumber results reported to the Global Fund: e.g. in 2006 alone, 63 million ITNs were produced worldwide[31]. For ARV, the 1,439,000 life-years saved represents around 33% of a total 3.2 million estimated life-years saved in low- and middle-income countries by end-2007[24]. The 33% share is somewhat less than approximately 50% share in ARV results (a total of 2.9 million as of end-2007[24]), because some large ARV programs with support from Global Fund (e.g., Brazil) had provided people on ARV much longer, with larger effects.

Several limitations and caveats concerning input data and assumptions must be acknowledged. Errors may occur at any stages. Misclassification of smear positivity for TB is one example. While service delivery numbers may be imprecise and performance-based funding may provide incentives for over-reporting, the errors are not systematically in one direction, so that aggregated portfolio results should roughly reflect what was delivered[20]. In reality, results at mid- and end-year released by the Global Fund secretariat are significantly conservative because results always have reporting delays but only actual results available at the time of data compilation are aggregated without projecting to release dates. Aggregation at country level is done also in a conservative manner to avoid double counting. Data quality issues nevertheless highlight an urgent need to strengthen national monitoring and health information systems.

The presented methodology was designed exclusively to produce a portfolio-aggregate indication of global-level health gain, although we used certain country-specific input assumptions and presented estimates by regional break-down. Country-specific evaluations would, in contrast, require more refined methods and inputs.

The largest uncertainties in these estimates lie in the epidemiological models and their basic (cross-country) assumptions. The 2007 UNAIDS model for estimating HIV-related deaths[25] may be possibly optimistic in assuming 5-15% annual mortality on ARV, since cohort studies in resource-poor settings found up to 27% mortality within the first year, and substantial losses to follow-up which may include some additional mortality[38, 39]. Given that ARV reporting to Global Fund does not include an explicit dimension of treatment quality or patient retention, some proportion of treatments delivered are likely to be of less good quality and clinical outcome than is typical in western settings or in research cohorts in resource-poor settings.

For TB, lives saved depend critically on the comparison scenario (Figure 1). While a simplistic comparison against 'no-treatment' is most consistent across the services studied, this scenario may be less appropriate for DOTS (compared to ARV and ITNs), given the relatively high effectiveness and potential availability of non-DOTS treatment[12, 27, 28]. The purpose of the paper is to indicate the overall scale of programs rather than additional impact due to scale-ups.

Our estimate of lives saved by Global Fund-supported ITNs depends critically on assumptions to convert ITN deliveries into child-years of protection. The assumption that each ITN distributed provides an average 1.1 child-year of protection (see Methods) is simplistic and probably conservative, because: (i) some 38% of ITNs distributions reported by supported programs were long-lasting insecticidal nets (LLINs) that remain effective throughout their lifetime of typically 3-4 years[40]; (ii) even for conventional ITNs, the assumed 1.5-year lifespan may be short in view of insecticide re-impregnations supported through Global Fund grants[41], and (iii) ITN usage was estimated based on 2000-06 surveys, but increased funds also support behavior change communication activities. Estimates of child protection by ITNs should be validated and improved based on country-specific coverage data, forthcoming from household surveys in many recipient countries. Since recent surveys demonstrated significant increases in ITN usage among under-fives[31, 42], lives saved from malaria are expected to increase rapidly over the coming years.

There is reasonable consensus about the efficacy of ITNs against mortality in children exposed to endemic falciparum malaria, for given levels of coverage, in all recent malaria impact models[14], including a consensus model under development by UNICEF and WHO[43]. Malaria effect estimates were, nonetheless, conservative because we did not consider mortality averted outside sub-Saharan Africa and Papua New Guinea. In any case, considerable uncertainties about the size and geographical distribution of populations exposed to stable endemic malaria[30] contribute to relatively wide uncertainty ranges, compared to ARV and DOTS (error bars in Figure 1).

The presented uncertainty ranges, particularly for ARV and DOTS, may be minimum intervals, which do not capture the complex chain of factors that intervene between service delivery counts, population coverage and impact, such as quality of services including (mis-)diagnosis, drug quality, patient counselling and support to promote patient retention and treatment adherence. Periodic evaluation of quality of services must therefore be added to program progress tracking and performance evaluation.

Longer-term, improved approaches to measuring disease impact are essential in low- and middle-income countries. To enable this and to evaluate its institutional performance and partners context, the Global Fund recommends that programs spend 5-10% of grant budgets on strengthening monitoring and evaluation systems as well as supporting vital registration. Furthermore, the Global Fund Board has commissioned an independent Five-Year Evaluation including impact evaluations reinforcing quality data collection and analysis in 20 countries[20, 44].

The presented estimates indicate the scale of mortality effects that Global Fund-supported programs may have achieved over recent years. While the numbers carry great margins of uncertainty and cover only selected program components, they reflect the best information available to date. Evidence-based evaluation of the impact of HIV/AIDS, TB and malaria programs supported by the Global Fund and other partners is ongoing and must be further strengthened to maximize health gain. Great improvements will be made by triangulating service delivery reports with population-level data on intervention coverage and demographic outcomes, information on quality of services, and trends in health facility disease burdens recorded in national health information and management systems.