School of Public Health

Mark Tenforde

The Epidemiology, Outcomes, and Prevention of HIV-associated Meningitis in Southern Africa

Central nervous system infection are a primary cause of HIV-associated morbidity and mortality worldwide. In this dissertation, I discuss results from research conducted in Botswana, a country in southern Africa with an HIV prevalence that is among the highest in the world, including the first comprehensive meningitis national incidence estimates in a high HIV prevalence country and temporal trends over 16 years with national scale-up of ART. Linking cerebrospinal fluid (CSF) laboratory records of patients evaluated with meningitis to HIV and vital status records from national electronic registries, I provide novel insights into HIV correlates and outcomes from suspected and microbiologically-confirmed meningitis. I compare local findings in Botswana to other settings in sub-Saharan Africa through a systematic review and meta-analysis. Finally, I discuss estimates of the impact and cost-effectiveness of preventing cryptococcal meningitis - the most common cause of meningitis we observed in Botswana - through cryptococcal antigen (CrAg) screening and targeted pre-emptive antifungal therapy for individuals with advanced HIV. In the first two studies (Chapters 2-3), I discuss results from the Botswana National Meningitis Survey, a 16-year national cross-sectional audit of patients evaluated for meningitis by lumbar puncture with CSF analysis from 2000-2015, including complete national records for 2013 and 2014. Forty-one percent (8759/21,560) of cases of suspected meningitis in adults had abnormal cerebrospinal fluid (CSF) findings, half of these receiving a diagnosis of cryptococcal meningitis. Other causes of meningitis were rarely identified. Most evaluations for meningitis were in individuals with known HIV (73%), almost half (47%) of these with a preceding history of ART use. Using UNAIDS population data, we estimated a 2013-2014 national cryptococcal meningitis incidence of 17.8 cases / 100,000 person-years (PYO), or 96.8 cases / 100,000 PYO in HIV-positive, nearly identical to estimates from Gauteng Province, South Africa in the pre-ART era of 2002-2004. Incidence was two-fold higher in HIV-positive men than women. In the third study (Chapter 4), I evaluated outcomes of “culture-positive” (including cryptococcal, pneumococcal, or TB) and culture-negative suspected meningitis (stratified by CSF white cell count [WCC]) in a retrospective cohort study. This was done deterministic linkage of CSF to vital status records in a national death registry with a unique national identification number (“Omang”). Estimated mortality for cryptococcal meningitis under routine care settings in Botswana was 50% at 10 weeks and 65% at one year with amphotericin B-based therapy, similar to mortality estimated in sub-Saharan Africa with less potent fluconazole therapy. Missed doses of amphotericin were common and independently predicted poor outcomes, while adherence to interventions usually associated with improved outcomes such as therapeutic lumbar punctures was poor. One-year mortality for culture-negative, pneumococcal, and TB meningitis were 49%, 49%, and 56%, respectively. For HIV-associated cases without a cause identified, CSF inflammation (CSF WCC >20 cells/µL vs. ≤20 cells/µL) was not a predictor of mortality in the setting of advanced immune-deficiency (median CD4 136 cells/µL). In the fourth study (Chapter 5), I describe results from a systematic review and meta-analysis evaluating published mortality estimates for common HIV-associated meningitides in sub-Saharan Africa, under routine care conditions and “clinical trial” settings subject to more intensive clinical management. Pooled short-term mortality (defined as ≤2 weeks or in-hospital) for cryptococcal meningitis under routine care settings was 44% (95%CI: 39-49%, 40 studies) with no difference in mortality when stratified by treatment (amphotericin vs. fluconazole monotherapy). This contrasts with pooled mortality of 21% (95%CI: 17-26%, 17 studies) from controlled trial settings. Short-term mortality from pneumococcal and TB meningitis under routine settings were estimated at 55% (95%CI: 44-66%, 8 studies) and 48% (95%CI:35-61%, 10 studies), respectively. Few studies reported long-term outcomes. In the fifth study (Chapter 6), I discuss results from a modeling analysis. I estimated the impact and cost-effectiveness of CrAg screening and targeted pre-emptive antifungal treatment for CrAg-positive patients for the prevention of HIV-associated cryptococcal meningitis. Using local CD4 count, CrAg prevalence, outcomes, costing, and national incidence estimates, implementation of CrAg screening in patients with a CD4 T-cell count <100 cells/µL was found to be highly cost-effective and estimated to prevent one-quarter of cryptococcal meningitis deaths. I explored treatment of ART-experienced patients, not the target population in guidelines but increasingly recognized with cryptococcal infection through laboratory-based CrAg screening. Screening and pre-emptive treatment in this population remained highly cost-effective and prevented additional deaths. CrAg screening at a higher CD4 count thresholds (100-200 cells/µL), now conditionally recommended by the World Health Organization, had little impact and was significantly less cost-effective, with an estimated cost of nearly US$15,000 per death averted in Botswana (although still cost effective per World Health Organization standards). This research has important public health implications. First, the findings highlight that cryptococcal meningitis remains a key cause of HIV-associated mortality well into the ART era in southern Africa and now commonly occurs in patients previously established in HIV care services. Despite widespread ART availability, incidence remained high through 2013-2014 and almost half of cases occurred in patients with previous ART use. Improved strategies for care and retention in HIV-treatment services are warranted, particularly for key at-risk populations (e.g. working-age men). Secondly, these findings support nationwide scale-up of CrAg screening for individuals with advanced HIV in Botswana with CD4 count <100 cells/µL, as well as new guideline recommendations for treating the emerging population of ART-experienced patients with early cryptococcal infection increasingly recognized in mature ART programs. Given the high cost and modest impact of CrAg screening at higher CD4 counts, policy decisions will need to weigh the benefits against other health system priorities. Thirdly, under routine care settings I showed similar high mortality with amphotericin B based treatment for cryptococcal meningitis compared to less potent antifungal treatment with fluconazole. These findings suggest that simplified cryptococcal meningitis treatment strategies, such as with less toxic, shorter-course regimens, may be necessary to reduce mortality and to complement prevention efforts through CrAg screening and other upstream measures to prevent incident disease. Finally, a majority of suspected meningitis cases have no cause identified through laboratory testing. Typical prognostic markers, e.g. CSF WCC, are poor predictors of mortality in the context of advanced HIV, and mortality is estimated around 50% at one year. These novel findings support use of enhanced meningitis diagnostic strategies, particularly routine use of available TB molecular diagnostics (e.g. Xpert Ultra), and a need to better define etiologies and treatment strategies to improve survival.