The burden of HIV-associated cryptococcal disease

In this issue, Park et al. [1] from the Centers for Disease Control, University of Alabama, and Johannesburg, South Africa, have made the first attempt to estimate the global burden of HIV-associated cryptococcal infection. As clearly stated by the authors, the estimates are exactly that – estimates, which rely on the quality of the data on which they are based. Nevertheless, the results are striking, especially for sub-Saharan Africa, where the estimated number of deaths associated with cryptococcal disease, at half a million per year, is comparable with the number attributed to tuberculosis, a much more frequent infection, but one for which we have much more effective and widely available therapy. So are these results credible? The data are sparse, with just three incidence studies driving the estimates for sub-Saharan Africa (ruling out the possibility of evaluating or taking account of possible intraregional differences in incidence); and the reported incidences are variable, resulting in wide ranges for the estimates of cases and associated deaths. However, biases are as possible on the downside as on the upside. As the authors acknowledge, the relatively low incidence estimate from a South African study with which some of the same authors were involved [2] was likely affected by incomplete case ascertainment. Regarding the assumptions over case fatality rates, the estimated 70% 3-month case fatality in Africa is not unreasonable as an overall estimate. Recent prospective studies, inevitably selective to some degree, in Uganda have reported over 50% 6-month mortality with either amphotericin B or high dose fluconazole therapy [3,4]. In many centers, fluconazole, available through a donation programme, is the only affordable therapy. Retrospective cohorts with incomplete follow-up have suggested a minimum 50% 10-week mortality with fluconazole at the previous conventional dosage of 400 mg/day [5]. However, it should be remembered that not all the deaths in these studies are as a result of the cryptococcal infection. Causes of death are hard to ascertain in patients with late-stage HIV infection in resource-limited settings. Deaths within the first 2 weeks of diagnosis are predominantly a direct result of cryptococcal infection, but increasingly over time of follow-up, other HIV-related problems contribute. Although acknowledging the wide range and uncertainty of the estimates, some independent data suggest that the authors' conclusions are broadly correct, and the burden of HIV-associated cryptococcal infection lies well within the lower limit of the ranges provided by this study. Prospective studies have suggested that 10–20% of all deaths in HIV-infected patients in Africa are attributable to cryptococcal infection [6,7], even if one leaves aside a study of South African miners with very high cryptococcal infection rates as reflecting a special population [8]. Multiplication by estimates for total HIV/AIDS deaths in sub-Saharan Africa does not give dissimilar conclusions from the authors' original analysis – as the authors point out in their discussion. Interestingly, each of these cause-of-death studies also concluded that cryptococcal infection accounted for more deaths in HIV-infected individuals than tuberculosis [6–8]. There is little doubt that the contribution of cryptococcal disease to HIV-related mortality in Africa, in particular, has been underestimated. But just as important as the scale of the problem, is the fact that at least in some areas of Africa, there is as yet no evidence for a decrease in incidence despite expanding antiretroviral therapy (ART) programmes. The increasing numbers of patients on ART is only just keeping pace with the increasing number of patients progressing to stage IV disease and hence, becoming vulnerable to late-stage complications such as cryptococcal meningitis [9]. The difference between the numbers requiring and numbers on therapy, or the ‘treatment gap’, needs to be narrowed, and the median CD4 cell count at diagnosis raised further in order for cryptococcal disease to be impacted. In addition, however, this important study emphasizes the need to specifically address the problem of cryptococcal disease. Primary fluconazole prophylaxis may work and is under investigation, but an alternative, attractive strategy, might be to target preemptive fluconazole therapy to those screening positive for cryptococcal antigen prior to starting ART [10]. Such a strategy could prevent one-third of the cases of cryptococcal meningitis that now present after the start of ART at some centers [11]. Many patients in Africa simply present too late for current antifungal therapy to be effective, and efforts are needed to facilitate earlier diagnosis and treatment. Trials are needed to optimize and compare amphotericin B-based and oral antifungal regimens, and to define the best time to start ART after a diagnosis of cryptococcal infection. Increased access to antifungal drugs and the facilities to monitor therapy are needed, including manometers to enable measurement and management of the common complication of raised cerebrospinal fluid pressure. Finally, further, ongoing, prospective epidemiology studies are needed to monitor the impact of such interventions on the burden of cryptococcal infection.