Efficacy of induction regimens for cryptococcal meningitis in HIV-infected adults: a systematic review and network meta-analysis

Cryptococcal meningitis (CM) is the most fatal adult meningitis in patients with human immunodeficiency virus (HIV). There is no conclusive evidence for the superiority of 1-week amphotericin B deoxycholate (AmphB) + flucytosine (5-FC) regimen over other antifungals in the management of HIV patients with CM (HIV–CM patients). We aimed to evaluate the differences in efficacy 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 2642 participants. A regimen indicated a possibly lower early mortality rate, namely, AmphB + 5-FC + Azole (OR = 1.1E−12, 95% CIs = 1.3E−41 to 0.06) comparing to AmphB + 5-FC. The current NMA provides evidence that AmphB + 5-FC + Azole are superior to all the investigated treatments for induction regimen in HIV–CM patients.

www.nature.com/scientificreports/ non-affordable in resource-limited settings that experience heavy cryptococcal burden, such as Africa 11 . Another NMA that compared the efficacy of AmphB + 5-FC and AmphB + fluconazole 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 efficacy 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 efficacy 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) 16 .
Search strategy and selection criteria. Several  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) all-cause mortality rates by weeks 2 and 10; (5) mycological suppression; (6) hepatic adverse reaction (Supplementary Tables 5 and 6).
Outcome definitions. Two primary outcomes and two secondary outcomes were identified. Two primary outcomes were early mortality rate and late mortality rate. www.nature.com/scientificreports/ 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 based on quantitative CSF cultures or the rate of change in colony-forming-units (CFU) of fungal cultures during the initial 2 weeks of induction. The mycological suppression improvement was defined as the mycological suppression decline at a rate of ≤ 0.33 log CFU/day or the equal effect during the first 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 defined as being of 1-week duration. The azole regimen defined as fluconazole and voriconazole in current study. The high dose of azole (azole_H) was defined as fluconazole being larger than 800 mg/ day 5 . Characteristics of the early-and late-mortality rates of the included studies are listed in the Supplementary  Table 5. Characteristics of the microbiological eradication and hepatic adverse reactions in the included studies are shown in the Supplementary Table 6. Supplementary Table 7 details the key findings of the included studies.
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% confidence 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 Win-BUGS (version 1.4.3, Medical Research Council Biostatistics Unit, Cambridge, Massachusetts) and R version 3.6.1 (http:// www.r-proje ct. org/). Because the pooled estimates of these pairs are different, funnel plots for NMA center the effect size of each pair of treatments. And, we centered the effect sized values according to previous methodology 21 We evaluated the potential inconsistency between direct and indirect evidence by using the deign-by-treatment interaction model, loop inconsistency model and node-splitting model 22,23 . 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 efficacy is closer to that of a perfect treatment which is always the best and has a SUCRA value of 1.
Ethical approval. The study was approved by the institutional review board of Changhua Christian Hospital (CCH IRB No. 180801).

Results
Characteristics and description of the included studies. In total, 46 publications were considered for full-text review, and 27 were excluded (Supplementary Table 3). 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 2642 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 Figure 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 + 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.
Risk of bias, publication bias, inconsistency assessment, and GRADE ratings. We found that 61.7%, 14.3%, and 24.0% of the enrolled 19 studies showed a low, unclear, and high risk of bias, respectively. www.nature.com/scientificreports/ 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 Figure 3). The overall quality of direct and indirect evidence in the overall NMA was low to medium based on GRADE evaluation (Supplementary Figure 3). We found no evidence of inconsistencies by using either loop-specific approach, node-splitting approach, or design-by-treatment approach (Supplementary Table 10).
Funnel plots of the publication bias (Supplement Figure 4) showed general symmetry. No significant publication bias among the included studies was evaluated by Egger's test. Because intercept significantly is close zero, small study bias is not significant.

Discussion
Our NMA summarizes the current evidence on the efficacy and tolerability of the induction therapy for individual antifungals in HIV-CM patients. And our results showed that AmphB + 5-FC + Azole as induction regimen 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 mycological suppression, 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 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 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 1-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 significantly lower early mortality rate (Fig. 1), but, the result came from only a single study 24 . The evidence is not conclusive due to the small sample bias. In addition, there was no statistically significant difference in mortality rates and hepatic adverse events between AmphB + azole_H and AmphB + 5-FC for HIV-CM patients. Briefly, our study shows 1-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 25,26 . Mycological suppression could achieve fungal eradication in CSF as a prognostic factor. In previous studies, mycological suppression was independently associated with mortality 17,27,28 . Our result is similar to a previous report 14 showing no statistically significant difference in MS between the investigated antifungals. Our current NMA revealed that all the investigated antifungals were associated with a similar mycological suppression compared to AmphB + 5-FC in participants with mycological suppression, but the azole_H alone presented with poorly fungicidal activity (OR = 3.8, 95% CIs = 0.62-23). The results of our analysis is in line with those of a previous report 29 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 mycological suppression 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 [30][31][32] . The most frequent hepatic adverse events were due to an increased hepatic injury related to azole. Guidelines issued in 2010 recommended AmphB plus fluconazole (800 mg/day) as the induction therapy 5 . According to the current NMA, azole_H did not achieve a significantly cumulative hepatic toxicity effect. Moreover, AmphB plus fluconazole (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 confidence (or credible) intervals for some treatment comparisons, although we also adopted the Bayesian model to obtain more robust estimates 33,34 . 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 efficacy and safety of individual antifungals in patients with HIV-CM for early mortality. We found that AmphB + 5-FC + Azole are superior to all the investigated treatments as induction regimens for HIV-CM. Our NMA contributes compelling evidence the impactful evidence with AmphB + 5-FC + Azole for treating HIV-CM to alleviate the burden of 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 efficacy 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. www.nature.com/scientificreports/