The cure rate after different treatments for mucosal leishmaniasis in the Americas: A systematic review

Background Mucosal or mucocutaneous leishmaniasis is the most severe form of tegumentary leishmaniasis due to its destructive character and potential damage to respiratory and digestive tracts. The current treatment recommendations are based on low or very low-quality evidence, and pentavalent antimonial derivatives remain strongly recommended. The aim of this review was to update the evidence and estimate the cure rate and safety profile of the therapeutic options available for mucosal leishmaniasis (ML) in the Americas. Methodology A systematic review was conducted in four different databases and by different reviewers, independently, to evaluate the therapeutic efficacy and toxicity associated with different treatments for ML. All original studies reporting cure rates in more than 10 patients from American regions were included, without restriction of design, language, or publication date. The risk of bias was assessed by two reviewers, using different tools according to the study design. The pooled cure rate based on the latest cure assessment reported in the original studies was calculated grouping all study arms addressing the same intervention. The protocol for this review was registered at the International Prospective Register of Systematic Reviews, PROSPERO: CRD42019130708. Principal findings Twenty-seven original studies from four databases fulfilled the selection criteria. A total of 1,666 patients with ML were treated predominantly with pentavalent antimonials in Brazil. Other interventions, such as pentamidine, miltefosine, imidazoles, aminosidine sulfate, deoxycholate and lipidic formulations of amphotericin B (liposomal, lipid complex, colloidal dispersion), in addition to combinations with pentoxifylline, allopurinol or sulfa were also considered. In general, at least one domain with a high risk of bias was identified in the included studies, suggesting low methodological quality. The pooled cure rate based on the latest cure assessment reported in the original studies was calculated grouping all study arms addressing the same intervention. It was confirmed that antimony is still the most used treatment for ML, with only moderate efficacy (possibly increased by combining with pentoxifylline). There is already evidence for the use of miltefosine for ML, with a cure rate similar to antimony, as observed in the only direct meta-analysis including 57 patients (OR: 1.2; 0.43–3.49, I2 = 0). It was possible to gather all descriptions available about adverse events reported during ML treatment, and the toxicity reflected the pattern informed in the manufacturers’ technical information. Conclusions This study provides an overview of the clinical experience in the Americas related to ML treatment and points out interventions and possible combinations that are eligible to be explored in future well-designed studies.

Introduction Mucosal (ML) or mucocutaneous (MCL) leishmaniasis is considered the most severe form of tegumentary leishmaniasis (TL) due to its destructive character and potential for functional loss in the respiratory and digestive tracts. Mucous involvement characteristically occurs years after the appearance of skin lesions, although it can also be detected simultaneously with or without apparent cutaneous involvement [1,2]. Although the species L. panamensis, L. guyanensis and L.amazonensis [3,4] have also been implicated with mucosal involvement, in addition to L. infantum and L. major [5], according to the World Health Organization, the term 'mucocutaneous leishmaniasis' should be applied only to the New World disease, caused by L. braziliensis [6].
Relative to the magnitude of the problem, approximately half of the cases of the American TL cases are registered in Brazil [7], which has a robust national epidemiological surveillance system based on the mandatory notification for all forms of leishmaniasis. Based on the Notifiable Diseases Information System (SINAN) of the Ministry of Health, between 2007 and 2020, 288,133 cases of TL were registered in Brazil, and in 17,079 (6%) ML was the clinical presentation indicated in the notification form [8]. In the Americas, the Pan American Health Organization has made efforts to gather notification data from all countries through the SisLeish platform. According to this database, between 3.4% and 4.3% of ML or MCL cases have been observed in relation to the total number of CL cases reported in recent years, estimates likely underreported [7].
The established therapeutic options available for ML require parenteral administration and have a high toxicity profile [6]. According to the treatment guidelines for the Americas, the current recommendations are based on low or very low-quality evidence and the pentavalent antimonial (Sb v ) derivatives (with or without oral pentoxifylline) remain strongly recommended, given the scarcity of studies evaluating other alternatives [9].
Other systematic reviews have been published on the subject [1,10,11], but they are outdated or have included only the few randomized controlled studies (RCTs) available. Assuming that the available evidence in RCTs is insufficient to support treatment recommendations, we decided to carry out a more flexible search and analyze all the accumulated experience in ML treatment. Therefore, the aim of this review was to update the evidence and estimate the cure rate with the therapeutic options available for LM in the Americas based on a comprehensive and critical literature search. The safety profile related to each of the therapeutic options was also evaluated, when reported, and considered the secondary outcome of this review.

Protocol and registration
The protocol was registered at the International Prospective Register of Systematic Reviews (PROSPERO) under protocol CRD42019130708. This review was conducted following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) [12] statement (S1 Table).

Eligibility criteria
The eligibility criteria were established according to a guiding question defined based on the acronym PICO: (P) Population: subjects with leishmaniasis acquired in the Americas and mucosal involvement; (I) Intervention: any pharmaceutical drug treatment; (C) Comparator, if applicable: any other therapy, placebo, or no treatment; (O) Outcome: cure rate. Secondary outcomes: adverse events and relapse rates. RCTs, nonrandomized clinical trials, and observational studies (cohorts or case reports) were considered eligible.

Exclusion criteria
The exclusion criteria were nonoriginal studies (literature review, letters, replies, editorials, guidelines, publications presenting the same patients previously described in another article); in vitro studies, or those addressing nonhuman participants or specific populations (immunosuppressed or failed patients) and studies in which, for each intervention evaluated, fewer than 10 patients were treated. Studies that did not present the number of ML patients cured, when there were CL patients in the sample, were also excluded. There were no language or publication date restrictions.

Information sources and search strategies
The literature search was concluded on December 15th, 2021, and the sources were MEDLINE (PubMed), Embase, LILACS, and Web of Science. Supporting information (S2 Table) describes in detail the search strategy for each database and the number of studies retrieved. A manual search was also performed in the references included in the previously published systematic reviews.

Study selection and data extraction
The titles and abstracts of the retrieved articles were independently evaluated by two reviewers (JPC and CSAS) using Rayyan software [13]. Disagreements were resolved by consensus or by a third reviewer (MLF, GC). After the initial selection, four authors (JPC, SNS, MLF, and LLA) performed a new screening in pairs of the selected full texts to confirm eligibility or identify exclusion criteria.
Four review authors (JPC, SNS, MLF and LLA) independently extracted data from the included studies using a standard data extraction form to collect characteristics of the studies, population, outcomes, and adverse events. The data extracted were then confirmed by a second reviewer (JPC, SNS and GC). If more than one publication reported the same patients, the study with the largest number of patients was selected and the other studies were excluded to avoid duplicate cases. When the same patients were described in more than one publication but for the presentation of different outcomes (e.g., cure and relapse rates), they were considered, but patients were counted only once.

Outcomes
The outcome of interest was cure rate at any moment after the end of treatment, using the intention-to-treat approach, regardless of how the author described it in the original publication and expressed as the number of cases cured by the total number of cases treated with a given intervention. Thus, the losses observed during the follow up were considered therapeutic failures. In order to maintain consistency among the definitions of cure used in the studies, the cure criterion assumed in the analyzes was as strict as possible, that is, even for authors who considered a re-epithelialization above 90% as cure, if the data on complete healing were available, these were the ones adopted as cure. Considering D1 as the first day of treatment, and using the definitions proposed by Olliaro (2013) [14], we arbitrarily adopted the following intervals to standardize the cure rate data extraction at different times after treatment: cure assessment performed between D60 and D135 was assumed to be D90; cure assessment performed between D136 and D270 was assumed to be D180; and cure assessment performed between D271 and D390 was assumed to be D360. For relapse rate estimation, only patients considered cured previously were considered. The safety of antileishmanial therapy was captured in each study as the number of adverse events (AE) per total patients (or treatments, alternatively, in case it is the only information available) evaluated. The rates were expressed as percentages and presented together with their respective 95% confidence intervals (CIs) calculated using the Mantel Haenszel random effects model. The adopted AE classification originally used (severity, intensity or other) was also obtained and, when available, was described for each study.

Data synthesis and statistical analysis
Comprehensive Meta-Analysis software v.3.0 was used to perform a one-group meta-analysis of study arms using a given treatment (pooled rates) based on the latest cure assessment reported in the original studies. Clinical cure rates were calculated according to the intentionto-treat analysis: the analysis was based on the total number of randomly assigned participants, irrespective of how the original study's authors analysed the data. These unadjusted indirect comparisons were compared with direct comparisons, when available. We used the inconsistency (I 2 ) statistic to evaluate heterogeneity. If significant heterogeneity was found, the results from the random effects model were emphasized and summary measures were analyzed as limited information, looking for differences in studies. A random effects model is a strategy that allows the interstudy heterogeneity to be incorporated through a broad CI, generating a more conservative estimate of the measure of the effect. For a global analysis within a given therapy with subgroups, a mixed effects analysis was used: a random effects model was used to combine studies within each subgroup and a fixed effect model was used to combine subgroups and yield the overall effect. The study-to-study variance (tau-squared) was assumed to be the same for all subgroups; this value was computed within subgroups and then pooled.
The Medical Dictionary for Regulatory Activities terminology (MedDRA), version 25.0, an international medical terminology developed under the auspices of the International Council for Harmonization of Technical Requirements for Pharmaceuticals for Human Use (ICH) [15], was used to standardize the safety data in this review. In brief, each AE reported in the primary study was recorded and classified in accordance with the Preferred Term (PT), High Level Group Terms (HLGL), and System Organ Classes (SOC). In the same hierarchical order of the stratification by system-organ, the MedDRA system includes the term "investigations" to designate alterations of complementary diagnostic tests referring to several systems.

Quality of evidence assessment
Two pairs of independent researchers (MLF/SNS and MLF/LLA) assessed the risk of bias using specific tools according to the study design. RCTs were evaluated using the Cochrane risk of bias score (RoB 2) [16,17] and nonrandomized clinical trials and prospective cohort studies were evaluated using the Newcastle Ottawa Scale (NOS) [18]. In addition, the modified NOS (Murad, 2018 [19]) was used to assess bias in retrospective observational studies (mostly case series), as previously performed by others [20][21][22][23][24]. RoB 2 is based on five domains: (1) randomization process; (2) deviations from intended interventions; (3) missing outcome data; (4) measurement of the outcome; and (5) selection of the reported result. In turn, the NOS tool includes nine items that can be categorized into three dimensions: 1) selection of study groups, 2) comparability of groups, and 3) determination of the results of interest. The adopted tool for case report assessment is based on four domains (selection, ascertainment, causality and report) and eight items. This tool was adapted, as suggested by Murad (2018), resulting in five items. Three of these five items, related to report and outcome aspects, received double weight. To standardize the risk of bias assessment, it was defined that the minimum follow-up time of 90 days [14] and a loss to follow-up of less than 20% would be the parameters for a low risk of bias study. For all domains, the risk was assumed to be high if there was not enough information to assess the quality.

Studies and population characteristics
The search identified 1,104 records from the Embase (350), MEDLINE (323), Web of Science (231) and LILACS (200) databases. The process of study selection and reasons for exclusion are summarized in a PRISMA flow diagram (Fig 1). After removal of duplicates, 822 records had their titles and abstracts evaluated and 679 were excluded. From the 143 selected studies, 142 were read in full to confirm their eligibility and to extract the data. After numerous attempts, we were unable to obtain Fernandes's (1990) [25] study for full text evaluation. After 115 exclusions, 27 articles were included for quantitative synthesis.
All studies included symptomatic patients, that is, individuals with clinical manifestation of mucosal involvement, and most of them with confirmation of Leishmania infection either by direct examination, culture, polymerase chain reaction (PCR) or presence of amastigote in the histological examination. Some studies also considered a Montenegro skin test (MST), serological tests, epidemiological exposition, and a therapeutic test, as alternative confirmatory criteria.
Most studies defined cure as complete epithelialization of all lesions, associated with the disappearance of inflammatory signs (infiltration, oedema, redness). For some authors, based on the assessment of a lesion severity score, a clinical improvement above 90% was considered a cure [32, 48,52]. For these studies, the total number of patients with complete healing was also reported, allowing us to extract this as the study cure outcome, to maintain consistency with the other studies. In one study, the definition adopted for cure was not informed [26]. Some series described more than one course of treatment for the same patient who did not achieve cure after the first treatment attempt. In these cases, the outcome considered was the number of patients cured with the first course of a given intervention.
In general, the authors used the terms "relapse" or "recurrence" indistinctly. In 11 studies, the condition was defined as the reappearance of the activity of an already healed mucosa [27,28,31,33,36,41,42,45,46,50,51]. For three studies, relapse could occur after incomplete healing [42, 45,50], and for the other five studies, relapse could be new lesion onset [27,28,31,33,45]. Time for cure assessment and follow-up varied widely among studies and the initial date from which the follow-up accounting starts was not a consensus among the authors, some counting from the first and others from the last day of treatment. Most studies presented cure assessment within one year from the end of treatment but at different moments. The follow-up time in general was also variable, ranging from one month [40] to 10 years [36]. When the follow-up time was not previously set but the patients' follow-up average was presented, this information was considered ( Table 1).
As shown in Table 2, the population gathered in this review was mostly made up of adults, with an average age varying from 30.7 to 74 years. Only five studies [26,27,33,38,49] included children, and in four studies, the authors did not report the patient's ages [34,36,37,48]. Based on studies providing the sample size stratified by sex, the male:female ratio was 1077:437, as expected due to the recognized predominance of men affected by ML. The nasal mucosa was the most affected site, and the average duration of symptoms before treatment varied from 3 to 272 months. ML severity was classified using different criteria across the studies. The Llanos-Cuentas (1997) classification [42], which combines mucosal lesion extension and severity of symptoms, was used in five studies [31,33,39,42,50]. In turn, Soto's (2007) [48] classification was adopted in three studies [32, 48,52] and Lessa's classification (2012) [54] was used in two others [30,40]. Several other criteria were used for classifying the disease severity, including the presence of

PLOS NEGLECTED TROPICAL DISEASES
The cure rate for mucosal leishmaniasis in the Americas: A systematic review

PLOS NEGLECTED TROPICAL DISEASES
The cure rate for mucosal leishmaniasis in the Americas: A systematic review

PLOS NEGLECTED TROPICAL DISEASES
The cure rate for mucosal leishmaniasis in the Americas: A systematic review

PLOS NEGLECTED TROPICAL DISEASES
The cure rate for mucosal leishmaniasis in the Americas: A systematic review

PLOS NEGLECTED TROPICAL DISEASES
The cure rate for mucosal leishmaniasis in the Americas: A systematic review

Outcomes
The cure and relapse rates described by each study at different times after the start of treatment are shown in Table 3. Taking as a model the outcomes harmonization for CL studies [14,55] (Fig 2). No significant difference was observed between patients treated with standard MA therapeutic doses (15-20 mg/kg for 10-30 days) in comparison to MA low doses (5 mg/kg/day): 65.1% (CI: 52.8-75.6%; I 2 = 55.8) and 67.2% (CI: 42.1-85.2%; I 2 = 0), respectively. The cure rate of the 299 patients treated with MA derivatives (66.2%, CI: 57.7-73.9%) was significantly higher than that observed for the 156 patients treated with SSG (51.8%, CI: 39.5-64%), p = 0.00. The combination of MA with pentoxifylline apparently increased the cure rate based on an indirect comparison between patients treated with and without this adjuvant drug (77.4% and 65.1%, respectively, p = 0.00). On the other hand, the cure rate of patients treated with other combinations (MA plus allopurinol or sulfa) was similar to that observed for MA standard dose therapy (64.0% and 65.1%, respectively).
Two RCTs [30,32] compared the same interventions, MA and miltefosine, allowing a direct comparison in the meta-analysis (Fig 9). Involving 57 patients and considering the cure rate assessed in setpoints as close as possible, no difference was observed between these interventions (OR: 1.2; 0.43-3.49, I 2 = 0).

Adverse events
The occurrence of AEs was reported in 21 of the 27 studies (S3 Table). For most primary studies, the reporting of AE did not follow a systematic and active methodology. The analysis was generally incomplete (causality and expectation were not assessed), and different AE designations and toxicity classification systems were used. Only four studies explicitly reported the use of an AE classification system: three of them adopted the Common Terminology Criteria for Adverse Events (CTCAE) [32, 48,52], and one study used the Division of AIDS Considering only studies that reported AEs, we identified 1,401 AEs among 1,008 treated patients. For most studies, AEs were reported in absolute numbers, and not as the proportion  &: for all cure rates, the information available was based on the number of cures in relation to the number of treatments performed, which does not rule out that the same patient may have received more than one treatment. � Unlike the criterion adopted by the author in the study, only patients with 100% epithelialization were counted to maintain alignment with the definition of cure

PLOS NEGLECTED TROPICAL DISEASES
The cure rate for mucosal leishmaniasis in the Americas: A systematic review

PLOS NEGLECTED TROPICAL DISEASES
The cure rate for mucosal leishmaniasis in the Americas: A systematic review of treated patients with a given intervention. Few authors described serious or severe AEs; however, it was possible to identify treatment discontinuation due to AE for 37 patients (3.6%): 6 patients in therapy with MA (1 cardiac event, 2 laboratory alterations, 2 leukopenia, and 1 Jarisch-Herxheimer reaction); 1 patient under MF therapy (abdominal pain and elevated serum amylase); 5 patients using L-AMB (acute renal failure); 3 patients using d-AMB (vomiting or increased creatinine level); 10 patients using SSG (5 thrombocytopenia, 3 liver function test abnormality, 1 thrombocytopenia and 1 aspartate aminotransferase increase); 11 patients using SSG plus allopurinol (8 thrombocytopenia, 2 bradycardia, 1 liver function test abnormality, and 1 infectious condition), and 1 patient using AS (hearing loss for high-frequency sounds) (S3 Table).

PLOS NEGLECTED TROPICAL DISEASES
The cure rate for mucosal leishmaniasis in the Americas: A systematic review The detailed description of AEs by SOC for each intervention is presented in Table 4. Overall, 64% of the AEs occurred among patients treated with antimony derivatives, and the therapy received by 45% (744 of 1,666) of patients gathered in this review. Among patients treated with antimonial drugs, musculoskeletal and connective tissue disorders were the AEs most identified, followed by laboratory abnormalities, named the "investigations" SOC term (S4 Table).
Among the 173 patients treated with one of the amphotericin formulations, we identified 182 AEs, which corresponds to 13% of the events accounted for in this review. "Injury, poisoning and procedural complications", such as infusion-related reactions and "Renal and urinary disorders" were the AEs most reported for both groups of amphotericin B formulations (lipid formulations and deoxycholate).
The most frequent AEs reported among patients treated with AS were "General disorders and administration site conditions" (such as fever, chest pain, chills, malaise, etc.), while gastrointestinal alterations were the most described among those treated with MF.

PLOS NEGLECTED TROPICAL DISEASES
The cure rate for mucosal leishmaniasis in the Americas: A systematic review

Risk of bias in included studies
The risk of bias assessment is presented in Figs 10 and S1. For RCTs, overall, the risk of bias was considered moderate. The highest risk domain was outcome assessment (mainly due to lack of blinding), followed by the absence of allocation concealment during randomization. As expected, for cohort and nonrandomized studies, the major concern, present for all studies, was the absence of comparator or comparability between groups with different interventions, followed by risk of bias in the selection of the population. Finally, for retrospective studies, the risk of bias was present in all evaluated domains, with greater intensity for the ascertainment bias.

Discussion
The main observation of this review is the identification of a relatively small number of published therapeutic interventions for ML and, in general, a small number of clinical trials addressing the disease. As expected, antimony derivatives account for most of the treatments gathered here, including the two pentavalent antimony derivatives, meglumine and

PLOS NEGLECTED TROPICAL DISEASES
The cure rate for mucosal leishmaniasis in the Americas: A systematic review

PLOS NEGLECTED TROPICAL DISEASES
The cure rate for mucosal leishmaniasis in the Americas: A systematic review stibogluconate, in addition to antimony combinations with sulfa, allopurinol or pentoxifylline, used in several therapeutic regimens (different doses, durations and scheduled administrations). The other therapies retrieved for ML were different formulations of amphotericin B, aminosidine, pentamidine, miltefosine, and imidazoles.

PLOS NEGLECTED TROPICAL DISEASES
The cure rate for mucosal leishmaniasis in the Americas: A systematic review The absence of controlled trials prevented the performance of a classical meta-analysis with direct comparison between treatments. Furthermore, due to the high heterogeneity and methodological weakness of the studies gathered, and mainly due to the impossibility of producing an effect measure for comparisons between interventions, the certainty assessment using

PLOS NEGLECTED TROPICAL DISEASES
The cure rate for mucosal leishmaniasis in the Americas: A systematic review Grading of Recommendations Assessment, Development and Evaluation (GRADE) [56] was considered inappropriate. For this reason, this review does not intend to support recommendations for ML treatment. Even so, this comprehensive review of literature focusing on the typical American mucosal disease and the pooled estimations of the cure rates observed with different interventions are hypotheses generators for future clinical trials. The scarcity and low quality of studies evaluating treatment alternatives for ML are explained by several factors. As a disease that affects neglected populations in underdeveloped countries, the chronic lack of investment in research is widely recognized [57]. In addition, there is an intrinsic difficulty in identifying new drugs against Leishmania [58]. Finally, ML is a relatively infrequent complication of cutaneous leishmaniasis, accounting for approximately 3-6% of cases with skin lesions [8], preventing the execution of large clinical trials.
Other systematic reviews addressing interventions for different clinical forms of leishmaniasis have already been published [1,10,11]. The first, published 15 years ago (2007), also used a comprehensive inclusion criterion in terms of study design, however, only 8 of the 22 previously included studies [1] were selected, due to our definition of a minimum of 10 patients treated in each intervention arm. In 2018, a new review focusing on combined therapies with pentavalent antimonial derivatives was published, but only two studies addressing ML were included [10]. The most recent review was conducted by Pinart et al, 2020 [11], updating a review published in 2009 [59]. Based exclusively on RCTs, eight studies addressing ML were retrieved [11].
Considering that the available evidence has not yet been able to answer the clinical questions surrounding the treatment of ML, this study follows the more recent trend of reframing the classic evidence hierarchy pyramid [60]. For conditions not studied in controlled studies, it is important to look for alternatives to optimally promote the production of knowledge and guide practice. Using a broad and systematic search strategy and focusing on critical analysis, aware of the limitation of grouping intrinsically different studies, this approach intends to provide the best available knowledge to support clinical decisions and identify the most promising interventions to be studied in clinical trials.
Despite the high risk of bias involving the selection of population in the retrospective and noncomparative studies, the profile of the population gathered here, based on the age average reported by authors, reflects the patient typically affected by ML, adult male, between the 5th and 6th decade of life, although, there is a wide age range (from 1 month to 90 years). The low representation of children in this review, while it corresponds to the expected-as a secondary complication of the cutaneous form, also alerts to the diversity of a disease that can involve different clinical variations and consequently require different therapeutic approaches. The large variation in disease duration supports this statement, as well as the diversity of affected mucosal sites and of clinical severity patterns. In general, the degree of confidence in the diagnosis of ML was high, and most studies included patients with parasitological or molecular confirmation. On the other hand, few studies have presented the species of Leishmania involved. In addition to the heterogeneity of the disease itself, the studies also present great methodological heterogeneity (variation in study design and risk of bias). The high heterogeneity between the studies was confirmed by the wide confidence intervals and moderate to high I 2 index observed for almost all intervention groups. Based on that, we chose to critically analyse the clinical and methodological differences among the studies to explain the discrepancies. This review clearly demonstrates that there is a lack of standardization in the definitions of cure, failure, and relapse, including the assessment criteria itself and the time defined for this evaluation, both compromising data gathering. The existence of an attempt to standardize criteria and outcomes in CL trials [14,55], although representing a useful model, minimally this should be validated for use in ML and this first compilation allows scaling current practice in relation to these definitions. The interventions evaluated in RCTs were MA, MF, SSG, AS, and associations of SSG with allopurinol and of MA with pentoxifylline; however, most studies presented some bias risk. The only low-risk RCT study evaluated the association of MA with pentoxifylline [34] in a limited number of participants (23 patients in total). For all antimonial therapeutic schemes evaluated (Fig 2), only moderate cure rates were observed, varying from 52.6% (CI: 37.9-66.8%) to 77.4% (CI: 51.4-91.7%). As a factor affecting all interventions evaluated, losses observed during follow-up were considered therapeutic failures, which may have underestimated the cure rate estimation. The pooled cure rates reported by studies using low or standard MA doses were quite similar. It is important to note that the two studies addressing low MA doses [38,39] were small and noncomparative studies with evident bias in the outcome assessment. Regarding the association of MA with pentoxifylline, the findings suggest some benefit on the cure rate. Despite the small sample size of the only trial addressing this association [34], the observations provided by another uncontrolled but larger study reinforce this effect [40]. On the other hand, the cure rates using other combinations (MA plus allopurinol [42] or sulfa [26]) were similar to those observed with MA monotherapy in standard doses. It is important to note that the study describing the highest cure rate with the combination of MA plus sulfa [26] represents a retrospective description of cases with significant bias in all domains, selection, treatment, and outcome assessment. In turn, the apparent lower effectiveness of SSG in relation to MA reinforces that observed in other studies [61] and may suggest that meglumine is the most suitable pentavalent derivative for the Leishmania species prevalent in the Americas. Although it was based on noncontrolled studies, difference can be observed in the cure rate between patients treated with lipid (Fig 3) and deoxycholate amphotericin B formulations (p = 0.0001) (Fig 4). However, the analysis of the studies describing the use of amphotericin deoxycholate clearly indicates that most treatment failures were related to patients who were not able to complete the treatment due to AEs.
In turn, the pooled cure rate grouping the patients treated with pentamidine ( Fig 5) represents the experience in only one Brazilian centre. Although cases were reported in two different retrospective studies [28,33], there is some overlap in the observation period, allowing the assumption that some patients were reported more than once, which may have overestimated the cure rate. On the other hand, confidence in the estimated cure rate of miltefosine is higher since the rates observed in randomized and uncontrolled studies were similar.
In addition, the only direct meta-analysis involving two RCTs and comparing MA and MF showed no difference between the interventions. Furthermore, a higher cure rate in four years (80%) compared to D180 (55%) was reported by Sampaio (2019), one of the RCTs addressing miltefosine. Even so, to keep alignment with the cure assessment criterion defined previously, the rate used in the grouping of patients was that observed at D180, which may have underestimated the miltefosine's efficacy in long-term evaluations.
Among the studies that evaluated the treatment based on imidazole drugs, high methodological fragility and variation in the cure rate were observed (Fig 7). The study reporting the highest cure rate [26] consists of a retrospective report of the individual experience in a single service, without details of the diagnostic and cure criteria, classified as very poor-quality evidence.
Finally, the experience with aminosidine consists of two small studies [50,51], both prospective and one of them controlled. The cure rate was uniformly low (Fig 8), less than 50%, suggesting the ineffectiveness of this intervention.
The safety profile for the therapies for ML presented in this review needs to be evaluated with caution, considering the lack of systematic and harmonized safety monitoring procedures in the primary studies. Thus, some level of underreporting is expected [62], making it difficult to estimate the true frequency of occurrence. In addition, the absence of a complete analysis of the AEs, including seriousness, severity, intensity, causality and expectation limits the toxicity impact assessment. It is important to note that the lack of homogeneity between studies in the definition of AE significantly affects the comparison of the event rate between interventions. Furthermore, differences in the management of the follow-up losses and treatment interruptions due to AE, possibly impacted the observed cure rates. In general, for all interventions, the AEs identified reflect the known toxicity pattern already informed in the manufacturers' technical information and literature [63]. The suspension of treatment due to the occurrence of an AE can be understood as the severity criterion adopted by some authors, but the result evaluated (with only 3.6% of serious AE or that caused the interruption) does not reflect the known concern with the toxicity of therapies for leishmaniasis [6].
The present review has several limitations. However, one of its main contributions was to gather, for the first time, in a systematic and exhaustive way, all descriptions of the occurrence of adverse events available among therapeutic studies addressing ML. Thus, in addition to effectiveness, safety profiles can also be used to support recommendations for the management of ML. For research and public polices, our observations confirm the urgent need for harmonization of clinical outcomes in leishmaniasis. Additionally, the poor, non-systematized and incomplete description of adverse events in terms of frequency, intensity, seriousness and causality also confirms the lack of reliable pharmacovigilance data. These points represent weaknesses to be faced also by health surveillance systems, in order to improve the usefulness of the compulsory notification data collected in endemic countries.
Finally, this review brings important reflections on the treatment of ML. First, there is a lack of quality evidence supporting clinical decisions. Based on the experience gathered, we confirmed an overall only moderate cure rate and a large discrepancy in the published rates, mainly reflecting the lack of standardization in the outcome evaluation criteria but also possibly the existence of prognostic factors related to the different clinical forms, characteristics of patients, and possibly to the parasite, which should be the subject of future studies.
The only possible direct comparison does not suggest the presence of a difference between the interventions, confirming what was observed in the indirect comparison between MA and miltefosine using pooled cure rates. This observation reinforces those biases related to small samples, and uncontrolled studies should be assumed to be present in the effect estimates presented. Thus, the main contribution of this review is to provide a comprehensive overview of the clinical experience in the treatment of ML in the Americas and to point out interventions and possible combinations that are eligible to be compared in a large pragmatic clinical trial.