A prospective ascertainment of cancer incidence in sub- Saharan Africa: The case of Kaposi sarcoma

Abstract

In resource- limited areas, such as sub-Saharan Africa, problems in accurate cancer case ascertainment and enumeration of the at-risk population make it difficult to estimate cancer incidence. We took advantage of a large well- enumerated healthcare system to estimate the incidence of Kaposi sarcoma (KS), a cancer which has become prominent in the HIV era and whose incidence may be changing with the rollout of antiretroviral therapy (ART). To achieve this, we evaluated HIV- infected adults receiving care between 2007 and 2012 at any of three medical centers in Kenya and Uganda that participate in the East Africa International Epidemiologic Databases to Evaluate AIDS (IeDEA) Consortium. Through IeDEA, clinicians received training in KS recognition and biopsy equipment. We found that the overall prevalence of KS among 102,945 HIV- infected adults upon clinic enrollment was 1.4%; it declined over time at the largest site. Among 140,552 patients followed for 319,632 person-years, the age- standardized incidence rate was 334/100,000 person-years (95% CI: 314– 354/100,000 person- years). Incidence decreased over time and was lower in women, persons on ART, and those with higher CD4 counts. The incidence rate among patients on ART with a CD4 count >350 cells/mm3 was 32/100,000 person- years (95% CI: 14–70/100,000 person-years). Despite reductions over time coincident with the expansion of ART, KS incidence among HIV-infected adults in East Africa equals or exceeds the most common cancers in resource- replete settings. In resource-limited settings, strategic efforts to improve cancer diagnosis in combination with already well-enumerated at- risk denominators can make healthcare systems attractive platforms for estimating cancer incidence.Cancer incidence is one of the most fundamental param- eters in cancer epidemiology. Incidence encompasses both the natural history of a malignancy and the effects of interventions to reduce occurrence [1]. Accurate estimates of cancer incidence are vital elements in ascertaining the etiology of cancers, planning for public health burden, and monitoring the effects of interventions. In resource- rich settings, given the better equipped medical infrastruc- ture, virtually all instances of cancer are diagnosed and recorded. These diagnoses are then placed into contextof the underlying denominator of persons at risk by the creation of incidence rates [2]. The denominators are typically derived from municipally funded complete enu- merations (i.e., a census) of geographic populations. In contrast, in resource-limited settings, such as sub-Saharan Africa, there is limited infrastructure for cancer diagnosis, and even when diagnosed, not all cancers are formally recorded [3]. Further, there are challenges in enumerating the denominator from which cancers arise. The WHO- sponsored Cancer Incidence in Five (“CI5”) Continents project deemed only 4 out of 25 registries from countries in sub-Saharan Africa to have sufficient quality [4, 5], and even within these countries, there are issues in both ascertainment of total cancer cases and the underlying denominator. Kaposi sarcoma (KS) is an example of a malignancy in a resource- limited setting which would benefit from knowledge about incidence. From a perspective of percentage of all recorded cancers, KS was among the most common cancers in sub-Saharan Africa even before the human immunodeficiency virus (HIV) epidemic [6, 7], and it experienced explosive growth as HIV infection spread [8, 9]. The clinical relevance of KS includes both cosmetic disfigurement and considerable morbidity and mortality. In persons untreated for HIV, 1-year mortality after KS diagnosis in sub-Saharan Africa is 60% to 70% [10, 11]. Even among persons treated with antiretroviral therapy (ART), those with KS have about a fourfold higher rate of death [12]. In resource-rich settings, ART has substantially reduced KS incidence, but because of the lack of robust sources of incidence data, the status in sub- Saharan Africa is less clear aside from an initial report from South Africa [13]. As is true for many cancers, changes in KS incidence in resource- replete settings cannot necessarily be extrapolated to resource-limited ones. Differences between settings regarding the strain of the etiologic viral agent (Kaposi sarcoma-associated herpesvi- rus, KSHV), ambient HIV strains, human host, and poten- tially other environmental cofactors dictate that KS incidence must be directly measured in Africa for it to be relevant. To overcome the challenges inherent in a resource- limited setting, we used a newly assembled collection of healthcare system-derived databases, the International Epidemiological Databases to Evaluate AIDS (IeDEA) Consortium in East Africa, to derive a well-substantiated (in terms of numerator and denominator) estimate of cancer incidence in a large representative population of HIV-infected adults in sub-Saharan Africa. We focused on KS, not only because of its ease of measurement and clinical relevance, but also to demonstrate how add- ing the selective measurements to an already well- enumerated healthcare system-based population has the potential to be a powerful platform for the study of other cancers.