The Effects of Rifaximin in Hepatic Encephalopathy
The Effects of Rifaximin in Hepatic Encephalopathy
This review was carried out and reported based on a registered (Prospero no CRD42013005776) protocol developed using the methods described in the Cochrane Handbook for Systematic Reviews of Interventions and the PRISMA Statement for Reporting Systematic Reviews and Meta-analyses. The literature searches were performed with help from the Cochrane Hepato-Biliary Group (Table S1). The electronic searches included The Cochrane Library, Medline, Embase, The Cochrane Hepato-Biliary Group Controlled Trials Register and Science Citation Index Expanded. Manual searches included scanning of bibliographies in relevant articles and annual conference proceedings from the American Association for the study of Liver Diseases (AASLD), European Association for the Study of the Liver (EASL), the Asian Pacific Association for the study of the Liver (APASL), the International Society for Hepatic Encephalopathies and Nitrogen Metabolism (ISHEN) and associations hereunder. Electronic searches were performed in December 2013 and updated April 1, 2014. Additional trials were identified through electronic registers including the World Health Organization International Clinical Trials Registry Platform. All corresponding authors as well as the pharmaceutical companies sponsoring or assisting in conduction of trials were contacted for additional data. Information was also sought retrieved through the Food and Drug Administration (www.fda.gov) and European Medicines Agency websites (www.ema.europa.eu).
Randomised clinical trials were included irrespective of language, publication status, dose, type or duration of therapy. We included RCTs on prevention of HE and trials on patients with overt or minimal HE. We included RCTs on patients with TIPSS in a sensitivity analysis.
Three authors (NK, AK and LG) extracted data in an independent manner. The extracted data included year of publication, study design and population, dose and duration of treatment, number of patients in each group, number of dropouts, patient age and sex. The primary outcomes were HE (prevention, recovery and improved manifestations), mortality and adverse events. Secondary outcome measures included surrogate markers of HE (psychometric tests and ammonia levels).
Two authors (NK and LG) performed bias assessment independently. The control of bias control was assessed based on the recommendations specified in the Cochrane Handbook for Intervention Reviews. Trials were classed as having a low, unclear or high risk of bias based on the individual components: allocation sequence generation, concealment of allocation, blinding, incomplete outcome data, selective reporting and 'other bias'.
Statistical analyses were performed using Revman version 5 (Nordic Cochrane Centre, Copenhagen, Denmark), stata version 13 (STATA Corp, College Station, TX, USA) and TSA version 9 (Copenhagen Trial Unit, Copenhagen, Denmark). The meta-analyses were performed using random effects models with results presented as risk ratios (RR) or mean differences (MD) with 95% confidence interval (CI), I and χ (presented as Q and P values) as markers of heterogeneity. We defined I values between 30% and 60% as moderate heterogeneity, 60–75% as considerable heterogeneity and values >75% as substantial heterogeneity. Values below 30% were considered unimportant. For the χ, a P value below 0.1 was considered to indicate significant heterogeneity. To determine the influence of between trial heterogeneity and the robustness of the result, the analyses were repeated using fixed effect models. The results were reported if the conclusions of the two models differed. The meta-analyses were performed with outcome measures that were recalculated based on individual patient data when available or based on data extracted from published reports or online information (the FDA, Food and Drug Administration website as described below). We calculated the number needed to treat (NNT) for the primary outcome measures that were statistically significant (the 95% CI did not cross 1) in trials with a low risk of bias and confirmed in sequential analyses. The NNT was calculated as based on the RR and the average control group risk (ACR) in trials with a low risk of bias (based on the formula NNT = 1/(ACR × (1−RR) and in trials with different control groups. We performed subgroup, sensitivity and regression analyses to evaluate sources of heterogeneity and bias. Subgroup analyses evaluated the effect of rifaximin in relation to the type of HE, control interventions (placebo, disaccharides or antibiotics), the type of data (outcomes recalculated based on individual patient data or extracted from published trial reports) and bias control. Differences between subgroups were expressed as P values (test for subgroup differences) with values below 0.05 considered statistically significant. We also performed a sensitivity analysis that included RCTs on patients randomised after TIPSS. The risk of bias and other small study effects were evaluated through regression analyses (using Harbord's test for dichotomous and Egger's test for continuous outcomes) and funnel plots for meta-analyses with at least 10 RCTs. The sequential analyses were based on random effects models, alpha 5%, power 80% and model-based heterogeneity (diversity). The sequential analysis was considered as confirmatory of the primary meta-analyses when the z-curve (cumulative results of included trials) crossed the trial monitoring curve and the required information size.
Methods
Search Strategy and Data Extraction
This review was carried out and reported based on a registered (Prospero no CRD42013005776) protocol developed using the methods described in the Cochrane Handbook for Systematic Reviews of Interventions and the PRISMA Statement for Reporting Systematic Reviews and Meta-analyses. The literature searches were performed with help from the Cochrane Hepato-Biliary Group (Table S1). The electronic searches included The Cochrane Library, Medline, Embase, The Cochrane Hepato-Biliary Group Controlled Trials Register and Science Citation Index Expanded. Manual searches included scanning of bibliographies in relevant articles and annual conference proceedings from the American Association for the study of Liver Diseases (AASLD), European Association for the Study of the Liver (EASL), the Asian Pacific Association for the study of the Liver (APASL), the International Society for Hepatic Encephalopathies and Nitrogen Metabolism (ISHEN) and associations hereunder. Electronic searches were performed in December 2013 and updated April 1, 2014. Additional trials were identified through electronic registers including the World Health Organization International Clinical Trials Registry Platform. All corresponding authors as well as the pharmaceutical companies sponsoring or assisting in conduction of trials were contacted for additional data. Information was also sought retrieved through the Food and Drug Administration (www.fda.gov) and European Medicines Agency websites (www.ema.europa.eu).
Randomised clinical trials were included irrespective of language, publication status, dose, type or duration of therapy. We included RCTs on prevention of HE and trials on patients with overt or minimal HE. We included RCTs on patients with TIPSS in a sensitivity analysis.
Three authors (NK, AK and LG) extracted data in an independent manner. The extracted data included year of publication, study design and population, dose and duration of treatment, number of patients in each group, number of dropouts, patient age and sex. The primary outcomes were HE (prevention, recovery and improved manifestations), mortality and adverse events. Secondary outcome measures included surrogate markers of HE (psychometric tests and ammonia levels).
Assessment of Risk of Bias
Two authors (NK and LG) performed bias assessment independently. The control of bias control was assessed based on the recommendations specified in the Cochrane Handbook for Intervention Reviews. Trials were classed as having a low, unclear or high risk of bias based on the individual components: allocation sequence generation, concealment of allocation, blinding, incomplete outcome data, selective reporting and 'other bias'.
Statistical Analysis
Statistical analyses were performed using Revman version 5 (Nordic Cochrane Centre, Copenhagen, Denmark), stata version 13 (STATA Corp, College Station, TX, USA) and TSA version 9 (Copenhagen Trial Unit, Copenhagen, Denmark). The meta-analyses were performed using random effects models with results presented as risk ratios (RR) or mean differences (MD) with 95% confidence interval (CI), I and χ (presented as Q and P values) as markers of heterogeneity. We defined I values between 30% and 60% as moderate heterogeneity, 60–75% as considerable heterogeneity and values >75% as substantial heterogeneity. Values below 30% were considered unimportant. For the χ, a P value below 0.1 was considered to indicate significant heterogeneity. To determine the influence of between trial heterogeneity and the robustness of the result, the analyses were repeated using fixed effect models. The results were reported if the conclusions of the two models differed. The meta-analyses were performed with outcome measures that were recalculated based on individual patient data when available or based on data extracted from published reports or online information (the FDA, Food and Drug Administration website as described below). We calculated the number needed to treat (NNT) for the primary outcome measures that were statistically significant (the 95% CI did not cross 1) in trials with a low risk of bias and confirmed in sequential analyses. The NNT was calculated as based on the RR and the average control group risk (ACR) in trials with a low risk of bias (based on the formula NNT = 1/(ACR × (1−RR) and in trials with different control groups. We performed subgroup, sensitivity and regression analyses to evaluate sources of heterogeneity and bias. Subgroup analyses evaluated the effect of rifaximin in relation to the type of HE, control interventions (placebo, disaccharides or antibiotics), the type of data (outcomes recalculated based on individual patient data or extracted from published trial reports) and bias control. Differences between subgroups were expressed as P values (test for subgroup differences) with values below 0.05 considered statistically significant. We also performed a sensitivity analysis that included RCTs on patients randomised after TIPSS. The risk of bias and other small study effects were evaluated through regression analyses (using Harbord's test for dichotomous and Egger's test for continuous outcomes) and funnel plots for meta-analyses with at least 10 RCTs. The sequential analyses were based on random effects models, alpha 5%, power 80% and model-based heterogeneity (diversity). The sequential analysis was considered as confirmatory of the primary meta-analyses when the z-curve (cumulative results of included trials) crossed the trial monitoring curve and the required information size.
Source...