Mark W Tenforde, Adrienne E Shapiro, Benjamin Rouse, Joseph N Jarvis, Tianjing Li, Ingrid Eshun‐Wilson, Nathan Ford

Cryptococcal meningitis is a severe fungal infection that occurs primarily in the setting of advanced immunodeficiency and remains a major cause of HIV-related deaths worldwide. The best induction therapy to reduce mortality from HIV-associated cryptococcal meningitis is unclear, particularly in resource-limited settings where management of drug-related toxicities associated with more potent antifungal drugs is a challenge. To evaluate the best induction therapy to reduce mortality from HIV-associated cryptococcal meningitis; to compare side effect profiles of different therapies. We searched the Cochrane Infectious Diseases Group Specialized Register, CENTRAL, MEDLINE (PubMed), Embase (Ovid), LILACS (BIREME), African Index Medicus, and Index Medicus for the South-East Asia Region (IMSEAR) from 1 January 1980 to 9 July 2018. We also searched the World Health Organization International Clinical Trials Registry Platform (WHO ICTRP), ClinicalTrials.gov, and the ISRCTN registry; and abstracts of select conferences published between 1 July 2014 and 9 July 2018. We included randomized controlled trials that compared antifungal induction therapies used for the first episode of HIV-associated cryptococcal meningitis. Comparisons could include different individual or combination therapies, or the same antifungal therapies with differing durations of induction (less than two weeks or two or more weeks, the latter being the current standard of care). We included data regardless of age, geographical region, or drug dosage. We specified no language restriction. Two review authors independently screened titles and abstracts identified by the search strategy. We obtained the full texts of potentially eligible studies to assess eligibility and extracted data using standardized forms. The main outcomes included mortality at 2 weeks, 10 weeks, and 6 months; mean rate of cerebrospinal fluid fungal clearance in the first two weeks of treatment; and Division of AIDS (DAIDS) grade three or four laboratory events. Using random-effects models we determined pooled risk ratio (RR) and 95% confidence interval (CI) for dichotomous outcomes and mean differences (MD) and 95% CI for continuous outcomes. For the direct comparison of 10-week mortality, we assessed the certainty of the evidence using the GRADE approach. We performed a network meta-analysis using multivariate meta-regression. We modelled treatment differences (RR and 95% CI) and determined treatment rankings for two-week and 10-week mortality outcomes using surface under the cumulative ranking curve (SUCRA). We assessed transitivity by comparing distribution of effect modifiers between studies, local inconsistency through a node-splitting approach, and global inconsistency using design-by-treatment interaction modelling. For the network meta-analysis, we applied a modified GRADE approach for assessing the certainty of the evidence for 10-week mortality. We included 13 eligible studies that enrolled 2426 participants and compared 21 interventions. All studies were carried out in adults, and all but two studies were conducted in resource-limited settings, including 11 of 12 studies with 10-week mortality data.In the direct pairwise comparisons evaluating 10-week mortality, one study from four sub-Saharan African countries contributed data to several key comparisons. At 10 weeks these data showed that those on the regimen of one-week amphotericin B deoxycholate (AmBd) and flucytosine (5FC) followed by fluconazole (FLU) on days 8 to 14 had lower mortality when compared to (i) two weeks of AmBd and 5FC (RR 0.62, 95% CI 0.42 to 0.93; 228 participants, 1 study), (ii) two weeks of AmBd and FLU (RR 0.58, 95% CI 0.39 to 0.86; 227 participants, 1 study), (iii) one week of AmBd with two weeks of FLU (RR 0.49, 95% CI 0.34 to 0.72; 224 participants, 1 study), and (iv) two weeks of 5FC and FLU (RR 0.68, 95% CI 0.47 to 0.99; 338 participants, 1 study). The evidence for each of these comparisons was of moderate certainty. For other outcomes, this shortened one-week AmBd and 5FC regimen had similar fungal clearance (MD 0.05 log10 CFU/mL/day, 95% CI -0.02 to 0.12; 186 participants, 1 study) as well as lower risk of grade three or four anaemia (RR 0.31, 95% CI 0.16 to 0.60; 228 participants, 1 study) compared to the two-week regimen of AmBd and 5FC.For 10-week mortality, the comparison of two weeks of 5FC and FLU with two weeks of AmBd and 5FC (RR 0.92, 95% CI 0.69 to 1.23; 340 participants, 1 study) or two weeks of AmBd and FLU (RR 0.85, 95% CI 0.64 to 1.13; 339 participants, 1 study) did not show a difference in mortality, with moderate-certainty evidence for both comparisons.When two weeks of combination AmBd and 5FC was compared with AmBd alone, pooled data showed lower mortality at 10 weeks (RR 0.66, 95% CI 0.46 to 0.95; 231 participants, 2 studies, moderate-certainty evidence).When two weeks of AmBd and FLU was compared to AmBd alone, there was no difference in 10-week mortality in pooled data (RR 0.94, 95% CI 0.55 to 1.62; 371 participants, 3 studies, low-certainty evidence).One week of AmBd and 5FC followed by FLU on days 8 to 14 was the best induction therapy regimen after comparison with 11 other regimens for 10-week mortality in the network meta-analysis, with an overall SUCRA ranking of 88%. In resource-limited settings, one-week AmBd- and 5FC-based therapy is probably superior to other regimens for treatment of HIV-associated cryptococcal meningitis. An all-oral regimen of two weeks 5FC and FLU may be an alternative in settings where AmBd is unavailable or intravenous therapy cannot be safely administered. We found no mortality benefit of combination two weeks AmBd and FLU compared to AmBd alone. Given the absence of data from studies in children, and limited data from high-income countries, our findings provide limited guidance for treatment in these patients and settings.