Search strategy and study selection
The study was conducted following the accepted methodology recommendations of the PRISMA checklist for systematic reviews. On December 18, 2019, we searched PubMed, Cochrane Central, Scopus, Web of Science, and Ovid databases for pertinent English articles using search terms ("Schistosoma haematobium" OR "schistosomiasis haematobia" OR "Bilharzia haematobium" OR "urinary schistosomiasis" OR "urogenital schistosomiasis" OR "vesical schistosomiasis" OR "Bilharziasis haematobium" OR "Bilharzia haematobium") AND (Metrifonate OR "Praziquantel"OR "Artesunate" OR anthelmin* OR treat* OR therapy*). This search strategy was developed and implemented in collaboration with a medical librarian experienced in systematic review and database searching. A manual search was done by searching for relevant publications in reference list of included articles and excluded systematic reviews; relevant papers in PubMed and Google Scholar; and primary studies that had cited the included papers. We also hand-searched using each keyword to avoid missing any relevant publications.
Study selection
Three independent reviewers scanned the titles and abstracts to select potentially-relevant articles. We included all original studies reporting treatment of urinary schistosomiasis. There was no restriction on country, language or publication date. Studies were included if they i) were children (less than 18 years old) and diagnosed with urinary schistosomiasis, ii) were clinical trial or any comparative observational study (cross sectional, case control, and cohort studies), iii) investigating the effect of PZQ, artesunate, and metrifonate comparing with other drugs or controls with no restrictions on the dose, frequency, and duration. iv) reporting the efficacy of cure rates, and egg reduction rates, v) were reported in English. We excluded papers if they met the following exclusion criteria: i) in vitro or animal studies, ii) data duplication, overlapping or unreliably extracted or incomplete data, iii) abstract only articles, reviews, theses, books, conference papers or articles with inavailable full texts (editorials, author responses, letters, and comments), and iv) any previous systematic reviews, meta-analyses and literature reviews on our topic of interest. Three reviewers independently performed an initial eligibility assessment of the retrieved titles and abstracts. Full texts of eligible articles were then retrieved and reviewed for inclusion. In both screening steps, inclusion or exclusion of a study by all three reviewers was considered conclusive. Conflicts were resolved through discussion among the authors. When necessary, the authors sought the opinion of senior reviewers on disagreements and discrepancies. All the references in this study were managed using Endnote X9, a reference manager software program.
Data extraction
Based on a pilot scan and extraction, two authors prepared a data extraction sheet using Microsoft Excel, and three reviewers independently extracted data from included studies. For accuracy, two different authors revised the data, and a third reviewer rechecked them. Disagreements or discrepancies were resolved through discussion and reaching consensus. Papers by the same research group and those studying the same factors were checked for potential duplicate data based on recruitment year, recruitment place, and confirmation from study authors. The outcome of interest was the efficacy of the medications of concern (PZQ, artesunate, and metrifonate), and outcome measurements were opted premised on the most commonly reported data in the included papers, namely cure rates, and egg reduction rates. Not only were these two measures endorsed for being frequently reported but also because they were feasible to assess pre and post-therapy using diagnostic tests. Two independent reviewers extracted and recapitulated data entailing study ID, last name of the first author, publication year, country, total number of participants, percentage of males, age touted as mean (SD), and the administered medication/s (name, dose, number of participants assigned to each drug). Regarding tools of outcome measurement, the extracted data encompassed drug name, the quantitative mean of efficacy, the total number of patients, the dose, the length of the therapeutic course in days, and the diagnostic test used for assessment. Whenever any article reported multiple checkpoints, only the last point was analyzed.
Quality assessment
Two reviewers have independently assessed the risk of bias of the included studies using the National Institute of Health (NIH) Quality Assessment Tool for Observational Cohort and Cross-sectional Studies and Case-control Studies (24). If the ratings were different, then the reviewers discussed to reach consensus. If consensus was not obtained, the study was forwarded to a third reviewer to settle the conflict. The elements of the quality assessment were indexed either Yes (1), No (0), or others including CD (Cannot Determine), NA (Not Applicable), and NR (Not Reported). Eventually, papers were rated fair, good, or high based on the final score.
Having four randomized controlled trials included the Cochrane collaboration risk of bias tool was used for its assessment (25). Five types of risks were evaluated selection (the adequacy of randomization), performance (the adequacy of blinding participants and study personnel), detection (the adequacy of blinding outcome assessors), attrition (the influence of missing data on the study results), and reporting (the adequacy of full, open reporting of results). participants and personnel (detection bias), incomplete outcome. The risk of each type of bias was graded “low,” “high,” or” unclear”.
Statistical analysis
Odds Ratios (ORs) with 95% confidence interval (CI) for direct comparisons or 95% credible intervals (CrI) for indirect comparisons were used. A network meta-analysis was performed using Stata software (version 14.2, StataCorp, College Station, TX) with random-effects models. To rank the treatments, the surface under the cumulative ranking probabilities (SUCRA) were opted to show which treatment was the best. Heterogeneity was considered significant with either I2>50 or P<0.05 with a subsequent adjustment of the model to small-study effects (incorporation of the heterogeneity) in this case. Inconsistency was also considered significant with either P<0.05, which was further investigated using node splitting methods (of direct and indirect comparisons) whenever indicated.