12 The assessment was based on published reports and information provided by the authors of included trials. Following

the implications of empirical evidence,12–14 the methodological quality of the trials was assessed based on sequence generation allocation concealment, blinding of outcome assessors, incomplete outcome data (lost to follow-up and adherence to intention-to-treat analysis), and early stopping for benefit. The analyses were performed using Review Manager 5.0 and Trial Sequential Analysis version 0.8. Dichotomous data were expressed as the risk ratio (RR) with 95% confidence interval (CI). Furthermore, the number needed to treat was derived from the RR in meta-analyses where MI-503 price the 95% CI (or the RR) did not include zero. Heterogeneity was explored using a chi-square test, and the quantity of heterogeneity

was measured using the I2 statistic.15 Sources of heterogeneity were assessed with subgroup analysis and meta-regression whenever possible. Subgroup analyses were performed only when data from at least two trials were available for each subgroup. Meta-regression was performed only for meta-analyses including more than 10 trials. Suitable sensitivity analysis was identified during the review process. When patients were lost to follow-up, data were analyzed according to the intention-to-treat principle. Intention-to-treat learn more analysis was performed assuming poor outcome in both groups, where dropouts were considered failures and the total number

of patients was used as the denominator. We used the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) approach16 to present the summary of findings for the patient important outcomes. To assess the reliability of pooled inferences from our meta-analysis on SVR, we calculated the optimum information size (OIS)—that is, the required meta-analysis sample size—to detect a 10% relative risk reduction in SVR, assuming an average event rate of 50% in the two treatment arms, assuming that 30% of the variation in the meta-analysis would be explained by variation across trials, and using statistical error levels of alpha = 5% and beta = 10% (90% power). www.selleck.co.jp/products/AG-014699.html Meta-analyses conducted before surpassing their OIS are analogous to interim analyses in single RCTs, and thus necessitate adjustment of the threshold for statistical significance to maintain the predetermined maximum risk of obtaining a false positive results (set to alpha 5% in our analysis). We therefore substituted the conventional 5% threshold for statistical significance with those of Lan-DeMets alpha-spending monitoring boundaries.8, 17–19 Figure 1 shows the results of the study screening. Twelve trials, including a total number of 5,008 participants,3, 20–30 that met our inclusion criteria31 were retrieved.