Re-Appraisal of Effective, and Tolerable Induction Regimens for Managing Cryptococcal Meningitis in HIV-Infected Adults: A Systematic Review and Network Meta-Analysis

Cryptococcal meningitis (CM) has become the most fatal adult meningitis in patients with human immunodeciency virus (HIV). There is no conclusive evidence for the superiority of one-week amphotericin B deoxycholate (AmphB) + ucytosine (5-FC) regimen over other antifungals in the management of HIV patients with CM (HIV-CM patients). We aimed to evaluate the differences in ecacy and tolerability of different antifungal agents in HIV-CM patients by conducting a current network meta-analysis NMA. Overall, 19 randomized controlled trials were included with 2,642 participants. A regimen indicated a possibly lower early mortality rate, namely, AmphB + 5-FC + High dose azole (Azole_H) (OR = 1.1*10 − 12 , 95% CIs = 1.3*10 − 41 to 0.06) comparing to AmphB + 5-FC. Azole_H was possibly associated with the lowest late-mortality rate (OR = 0.27, 95% CIs = 0.08 to 0.99). The current NMA provides evidence that AmphB + 5-FC + Azole_H are superior to all the investigated treatments for induction regimen in HIV-CM patients. AmphB, amphotericin B; AmphB_S, short course (only 1week) amphotericin B; Azole_H, high dose ( > 800 mgday ); Cryptococcal meningitis; CM ; ES: effect size; uconazole; 5-FC , ucytosine; human immunodeciency virus; HIV ; HIV-associated CM: HIV-CM ; Lip AmphB, liposomal amphotericin B; NMA: network meta-analysis; PRISMA: preferred reporting items for systematic reviews and meta-analyses; RCT: randomized controlled trial; OR: Odds ratio; SUCRA: surface under the cumulative ranking curve.


Introduction
Cryptococcal meningitis (CM) has become the most fatal meningitis in adult patients with human immunode ciency virus (HIV) [1][2][3] , particularly in those with acquired immunode ciency syndrome 4 . An effective anti-cryptococcal regimen is needed to treat HIV-associated CM (HIV-CM) 5,6 . HIV-CM treatment can be divided into three phases: induction, consolidation, and maintenance 5,7 . During the induction phase, antifungal treatment needs to decrease cryptococcal burden in the cerebrospinal uid (CSF) to increase patient survival 5 , 8 . Amphotericin B deoxycholate (AmphB)-based regimen is widely used during the 2-week induction period for HIV-CM, with uconazole or ucytosine (5-FC) being used for synergistic effects 5,9 . Research suggests that uconazole has good tissue penetration 10 ; moreover, 5-FC has a fungicidal effect against Cryptococcus neoformans 5 . The updated WHO (2018) guidelines for the preferred induction regimens to treat HIV-CM 6 recommended a change from 2 weeks of AmphB + 5-FC or 2 weeks AmphB + uconazole into 1 week of AmphB + 5-FC followed by 1 week of uconazole, and this offers a lower risk of medication toxicity and a reduced risk of nosocomial sepsis 6,11 .
The WHO recommendation was based largely on the ndings of meta-analyses reporting that 1 week of AmphB + 5-FC was possibly superior to other regimens 6,11 12 . However, the amount of outcome data used in these analysis was insu cient for direct comparisons between regimens, resulting in an imprecise evaluation of treatment effects 12 . A new study of a novel single-dose 10 mg/kg liposomal AmphB (LipAmphB) showed no statistically signi cant compared to 14 days of 3 mg/kg/day LipAmphB 13 ; furthermore, additional information on this topic has now become available 14 . A network meta-analysis (NMA) showed similar e cacy between the 1-week and 2-week induction regimens of AmphB + 5-FC 12 . In addition, 5-FC is both non-accessible and non-affordable in resource-limited settings that experience heavy cryptococcal burden, such as Africa 11 . Another NMA that compared the e cacy of AmphB + 5-FC and AmphB + uconazole showed similar outcomes of late mortality rate 15 . These reports demonstrated that the most effective and tolerable induction regimen for HIV-CM has not been completely elucidated as yet. Therefore, we conducted a systematic review and NMA of randomized controlled trials (RCTs) to compare the e cacy and safety of induction regimens of anti-cryptococcal agents in HIV-CM patients.

Materials And Methods
A systematic review and NMA were performed to evaluate the effectiveness and safety of different induction regimens in HIV patients with CM (HIV-CM patients). This protocol was approved by the institutional review board of Changhua Christian Hospital (CCH IRB No. 180801). The current study compared the e cacy and safety of induction regimens between AmphB + 5-FC and other evailable regimens in treating HIV-CM patients (Supplementary Table 1). The current NMA was performed according to the preferred reporting items for systematic reviews and meta-analyses (PRISMA) extension guideline for NMAs (Supplementary Table 2, Fig. 1

Data collection and inclusion criteria
Following the PRISMA guideline 16 , two reviewers (CHC, CYM) screened separately the titles and abstracts of identi ed articles. Discrepancies or issues of study methodology and quality were resolved by consulting with the third reviewer (CYJ). Studies were included if (1) patients were randomly assigned to different treatments; (2) they were published as full-length articles in peer-reviewed journals, and (3) the e cacy or adverse events of antifungals in treating CM were reported. Excluded studies are listed in the Supplementary

Data extraction
An information-extraction form was created and the following data were collected: (1) details of study design and publication; (2) baseline patient information; (3) the total number of recruited patients; (4) allcause mortality rates by weeks 2 and 10; (5) mycological suppression (MS); (6) hepatic adverse reaction (Supplementary Tables 5 and 6).

Outcome de nitions
Two primary outcomes and two secondary outcomes were identi ed. Two primary outcomes were early mortality rate and late mortality rate. The de nition of early mortality rate was all-cause mortality rate within two weeks of diagnosing CM. The de nition of late mortality rate was all-cause mortality rate occurring more than 6 weeks after diagnosing CM. Two secondary outcomes were microbiological eradication and hepatic adverse reaction. The microbiological-eradication outcome was derived from measuring mycological suppression (MS) based on quantitative CSF cultures or the rate of change in colony-forming-units (CFU) of fungal cultures during the initial two weeks of induction. The MS improvement was de ned as the MS decline at a rate of ≤ 0.33 log CFU/day or the equal effect during the rst 14 days of treatment 17 . The adverse reaction data were derived from the reported rate of liver damage events. The short course of AmphB was de ned as being of one-week duration. The azole regimen de ned as uconazole and voriconazole in current study. The high dose of azole (azole_H) was de ned as uconazole being larger than 800 mg/day 5 . Characteristics of the early-and late-mortality rates of the included studies are listed in the Supplementary Cochrane risk-of-bias tool and GRADE ratings Two independent reviewers (CHC, CYM) evaluated the risk of bias for each domain described in the Cochrane risk-of-bias tool 18 . The study evaluated the certainty of the evidence according to the GRADE framework 19 .

Network Meta-Analysis
The odds ratio (OR) with 95% con dence interval (CI) was summarized as the effect size for measuring all outcomes. We undertook the frequentist approach to NMA by using the mvmeta command 20 written for the statistical software package Stata (version 16.0, StataCorp LLC, Texas 77845 USA). When the numbers of event were small, we adopted the Bayesian approach by using the software package WinBUGS (version 1.4.3, Medical Research Council Biostatistics Unit, Cambridge, Massachusetts) and R version 3.6.1 (http://www.r-project.org/). We evaluated the potential inconsistency between direct and indirect evidence by using the deign-by-treatment interaction model, loop inconsistency model and nodesplitting model 21 22 . We also computed the ranking probabilities of treatments which were then summarized by the surface under the cumulative ranking area (SUCRA) ranging from 0 to 1. A treatment with a greater SUCRA value indicates that its e cacy is closer to that of a perfect treatment which is always the best and has a SUCRA value of 1.

Characteristics and description of the included studies
In total, 46 publications were considered for full-text review, and 27 were excluded (Supplementary Table 3). The current NMA was conducted according to the PRISMA guideline for NMAs (Fig. 1). Finally, 19 articles were included in our NMA (Supplementary Table 5). The quality of the included studies and their risk of bias were rated. Figure 1 depicts the entire geometric distribution of the treatment arms. A total of 2,642 participants were included; the baseline characteristics of the included participants are summarized in Supplementary Tables 4 and 5. In brief, 19 studies reported the early-mortality rate and 5 were multi-arm trials; 18 reported the late-mortality rate and 5 were multi-arm trials; 10 reported mycological suppression and 4 were multi-arm trials; and 11 reported hepatic adverse reaction involving 3 multi-arm trials. Figure 2 and Supplementary Fig. 1 depict the entire geometric distribution of the treatment arms for four different outcomes.

Primary Outcomes
The NMA showed that all the investigated antifungals associated with early mortality rate were similar to those seen in the AmphB + 5-FC-treated participants with CM (Fig. 3A). According to the forest plot, three regimens were possibly related to lower early mortality rate, namely AmphB + 5-FC + High dose azole  (Supplementary Table 8A), and the SUCRA evaluation (Supplementary Table 9), AmphB + 5-FC + azole was associated with the lowest risk of early mortality rate, followed by short-course AmphB (AmphB_S) and AmphB_S + 5-FC.

Co-primary outcome: late mortality rate
The NMA revealed that all the investigated antifungals associated with the late mortality rate were similar to the AmphB + 5-FC in participants with CM (Fig. 3B). According to the forest plot, the Azole_H was possibly associated with the lowest late-mortality rate (OR = 0.27, 95% CIs = 0.08 to 0.99) followed by AmphB_S (OR = 0.53, 95% CIs = 0.01 to 37.37) and 5-FC + AmphB_S (OR = 0.60, 95% CIs = 0.21 to 1.97) comparing to AmphB plus 5-FC. According to the League table (Supplementary Table 8B), Azole_H alone presented to higher early mortality rate than most regimens. According to SUCRA (Supplementary   Table 9), azole_H was associated with the lowest risk of the late-mortality rate followed by AmphB_S + 5-FC + azole_H.
The NMA revealed that all the investigated antifungals were associated with a hepatic-adverse-event rate similar to that of AmphB and 5-FC in participants with CM ( Supplementary Fig. 2B, Supplementary   Table 8D). According to the forest plot, the azole regimen was probably associated with the lowest adverse event (OR = 0.075, 95% CIs = < 0.001 to 1.6) followed by LipAmphB (OR = 0.50, 95% CIs = 0.018 to 11) and azole_H + 5-FC (OR = 0.55, 95% CIs = 0.059 to 3.9) comparing to AmphB with 5-FC.

Risk of bias, publication bias, inconsistency assessment, and GRADE ratings
We found that 61.7%, 14.3%, and 24.0% of the enrolled studies showed a low, unclear, and high risk of bias, respectively. Unclear reporting of the allocation procedures and blinding of the participants or research personnel was the most often encountered reason for the high risk of bias (Supplement Fig. 3).
The overall quality of direct and indirect evidence in the overall NMA was low to medium based on GRADE evaluation (Supplement Table 7).
We found no evidence of inconsistencies by using either loop-speci c approach, node-splitting approach, or design-by-treatment approach (Supplementary Table 10).
Funnel plots of the publication bias (Supplement Fig. 4) showed general symmetry. No signi cant publication bias among the included studies was evaluated by Egger's test.

Discussion
Our NMA summarizes the current evidence on the e cacy and tolerability of individual antifungals in HIV-CM patients showed that AmphB + 5-FC + Azole_H yielded a lower mortality rate than other antifungal regimens. All the investigated antifungals were associated with a similar early mortality rate and also with a similar late mortality rate in HIV-CM patients. Based on MS, the azole_H alone was not recommended to be induction regimen in HIV-CM patients due to poor fungicidal activity. In summary, AmphB + 5-FC + Azole_H are superior to all investigated regimens for induction and the azole_H alone was not recommended to be induction regimen in HIV-CM patients due to poor fungicidal activity in HIV-CM patients. Our NMA also found that the AmphB + 5-FC + Azole_H are superior to all the investigated treatments for induction regimens in HIV-CM patients. In contrast to Tenforde et al., our study did not found that one-week AmphB + 5-FC-based therapy was superior to other regimens used to treat HIV-CM 11 . As shown in a previous meta-analysis 12 , the combination of AmphB + 5-FC + azole showed mostly signi cantly lower mortality rate (Fig. 1), but, the result came from only a single study 23 . The evidence is not conclusive due to the small sample bias. In addition, there was no statistically signi cant difference in mortality rates and hepatic adverse events between AmphB + azole_H and AmphB + 5-FC for HIV-CM patients. Brie y, our study shows two-week AmphB + 5-FC remains the best preferred regimen than others in HIV-CM patients.
The goal of treating CM is to attain cryptococcal eradication and to protect from neurological damage 24 25 . MS could achieve fungal eradication in CSF as a prognostic factor. In previous studies, MS was independently associated with mortality 17 26 27 . Our result is similar to a previous report 14 showing no statistically signi cant difference in MS between the investigated antifungals. Our current NMA revealed that all the investigated antifungals were associated with a similar MS compared to AmphB + 5-FC in participants with CM, but the azole_H alone presented with poorly fungicidal activity (OR = 3.8, 95% CIs = 0.62 to 23). The results of our analysis is in line with those of a previous report 28 that the azole_H alone was not recommended to be an induction regimen in HIV-CM patients due to poor fungicidal activity. Due to disconnection of network from the reporting of MS events, the secondary outcome needs further evaluation to elucidate this viewpoint.
Anti-fungal agents have high toxicity and cause hepatotoxicity, although the effectiveness of AmphB for HIV-CM patients is outstanding. The administration of azole with AmphB is recommended for the induction treatment of CM 29 30 31 . The most frequent hepatic adverse events were due to an increased hepatic injury related to azole. Guidelines issued in 2010 recommended AmphB plus uconazole (800 mg/d) as the induction therapy 5 . According to the current NMA, azole_H did not achieve a signi cantly cumulative hepatic toxicity effect. Moreover, AmphB plus uconazole (800 mg/day) could be adopted as the standard induction antifungal regimen after Cryptococcal Optimal ART Timing Trial for HIV-CM patients 14 .

Limitations
There are several limitations to be acknowledged in the current NMA. First, some data analyzed in this study were limited by under-powered statistics, namely in heterogeneity between and within studies and a small number of trials for some treatment arms. Second, in the current NMA, we did not exclude trials with small case numbers in both the intervention and control arms because most RCTs had zero of relapse cases; this yielded a relatively large con dence (or credible) intervals for some treatment comparisons, although we also adopted the Bayesian model to obtain more robust estimates 32 33 . Lastly, in spite of comparing different antifungals in our NMA, future large RCTs are required to evaluate the effectiveness and safety of different induction regimens to determine the best regimen for the management of HIV-CM patients.

Conclusions
Our NMA provides synthesized current evidence on the e cacy and safety of individual antifungals in patients with HIV-CM. We found that AmphB + 5-FC + Azole_H are superior to all the investigated treatments as induction regimens for HIV-CM. Furthermore, azole_H alone is not recommended as an induction regimen for HIV-CM owing to its poor fungicidal activity. However, given the small number of studies included in the current analysis, future large-scale RCTs focusing on the e cacy of different dosages and treatment durations of antifungals in patients with HIV-CM should be conducted to support or refute the results of the current NMA.  Network structure of network meta-analysis of different outcomes for Cryptococcal meningitis in HIV patients 2A Network structure of network meta-analysis of early mortality rate for Cryptococcal meningitis in HIV patients 2B Network structure of network meta-analysis of late mortality rate for Cryptococcal meningitis in HIV patients Figure 3 Forest plot of network meta-analysis of two major outcomes for Cryptococcal meningitis in HIV patients 3A forest plot of network meta-analysis of early mortality rate for Cryptococcal meningitis in HIV patients 3B forest plot of network meta-analysis of early mortality rate for Cryptococcal meningitis in HIV patients

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. AppendixTables.docx