Criteria for considering studies for this review:
Types of studies:
Randomized controlled trials(RCTs) comparing Chinese herbal medicine with placebo or other active interventions will be included, regardless of language and publication status.
Types of participants:
People with acute or chronic HIV/AIDS-associated diarrhea, regardless of race, age, gender or economic status will be included.
People with diarrhea caused by dysentery, cholera, poisoningand systemic disease will be excluded.
Types of intervention:
Experimental interventions are Chinese herbal medicine, including extracts from herb(s), single herbs or compoundformula, administered orally, taken either alone or in combination with other active treatment.
Control intervention can be no treatment, placebo, or other active treatment.
Co-intervention is allowed as long as it is applied in both groups.
Types of outcome measures:
- Diarrhea frequency (per day or per week);
- Fecal character;
- Recovery durationof diarrhea;
- Other symptom such as fatigue and anorexia;
- Length of hospital stay;
- Body weight;
- Nutrition status;
- Quality of life (measured by validated tool or scale);
- Adverse effect: the incidence of alladverse events or adverse effects reported,includingwithdrawal from trial due to some adverse effects and modification of treatment plan. Based on the available data, we will classify the adverseevents as serious and non-serious. Any external medical event thatwas life-threatening, led to deathorsignificant or persistentdisability, or any medical occurrence which may have endangered patients or required intervention to prevent it belongs tothe serious adverse events . Additionally, other adverse eventswill be considered non-serious.
For the composite outcome index, only the results with the same components will be used for data analyses. If the composite outcome index is an ordered polytaxic variable, such as cure/significant effect/effective/ineffective, it will be dichotomized into effective/ineffective.
Search methods for identification of studies:
We will attempt to identify all relevant trials regardless of language or publication status (published, unpublished, in press, or in progress).
We will use the search terms and strategy described in Additional file 2 to search the following databases : The Cochrane Central Register of Controlled Trials (CENTRAL)in the Cochrane Library; PubMed;China National Knowledge Infrastructure (CNKI);Wanfang Data; Chinese Biomedical Literature Database (SinoMed); China Science and Technology Journal database (VIP).
Searching other resources:
We will examine all retrieved studies’reference lists and relevant reviews for further studies.
We will searchrelevant Chinese journals that aremostly likely to publish the trials aboutAIDS-related diarrhea by hand.
We will also contact researchers in the field to identify unpublished and ongoing studies.
Data collection and analysis:
For selection of studies and data extraction,two review authors (BLC and MZZ) will independently conduct each step, and examine agreement between them. We will resolve any disagreements through discussion with a third senior author (JPL).
Selection of studies:
Two review authors (ZWH and BLC) will check titles and abstracts retrieved by the searches of databases mentioned above independently. To make sure that we only include the same trial from multiple publications at once, we will examine each trial report. We will further evaluate the fulltext, if a record cannot be rejected with complete certainty for its title and abstract. We will contact the authors to clarify if the eligibility of a trial is unclear. We will resolve any disagreements through discussion,and involving a third author when necessary. We will also list the excluded studies as well as the reasons for their exclusion in a 'Characteristics of excluded studies' table. We will use a PRISMA diagram to display the study selection process.
Data extraction and management:
Retrieved citations from the various search engines will be imported into EndNote and checked for duplicates. Two review authors (BLC and MZZ) will use a pre‐piloted data extraction form to extract data on general information (study ID, study author and title); methodological information (study design, the number of groups, sequence generation, allocation sequence concealment, blinding, selective outcome reporting, baseline comparability); participant characteristics (diagnostic criteria, inclusion criteria, exclusion criteria, acute/chronic, total number of intervention groups, number lost during follow-up, age, sex, country,setting and disease duration); intervention (the name of therapeutic drug, its form, dosage and regimen, other treatments given, drug combination, duration of treatment), and outcome measures.We will record relevant datausing a pre‐defined data extraction form.We will resolve disagreements through discussion, and contact the corresponding trial author in case of unclear and missing data.
Assessment of risk of bias in included studies:
Two review authors (CX and HRZ) will assess methodological quality using the Cochrane 'Risk of bias' tool and report the results in a 'Risk of bias' table. We will use the criteria for judging risk of bias in the 'Risk of bias' assessment tool in the Cochrane Handbook for Systematic Reviews ofInterventions,considering the following:
- Sequence generation.
- Allocation concealment.
- Blinding of participants, personnel and outcomes assessors.
- Incomplete outcome data.
- Selective outcome reporting. We hope that trials report adverse event data, survival and disease progression.
- Other possible sources of bias.We will assess vested interests depending on authors of thetrials and the source of funding. If a conflict of interests is obvious,the risk of bias will be judged If there is a clear declaration of a lack of conflict of interests, and there is no clear vested interest in the author or the funding,the risk of bias will be judged low.In the absence of a declaration, the risk of bias will be judged unclear.Additionally, we will assess baseline differences between groups. If the differences are not significant or exist, but the impact on the results has been investigated and found to be insignificant, the risk of bias will be judged low. If the differences are obvious and not addressed, the risk of bias will be judged high.In other cases, the risk of bias will be judged unclear.
If information is unclear or not specified, we will also attempt to contact the trial authors. We will resolve any disagreements by discussion between the review authorsor, if necessary, by consulting another author.
Measures of treatment effect:
We will report dichotomous outcomes as risk ratios (RR) and 95% confidence intervals (CIs). We will present continuous outcomes as mean differences (MD) and 95% CIs if the outcomes have been measured in the same way across all included trials. We will use the standardized mean difference (SMDs) and 95% CI as the effect measureif included trials measured continuous outcomes in different ways. We will present time‐to‐event outcomes as hazard ratios and 95% CIs. We will also pay attention to the time points at which results were collected and reported.
Unit of analysis issues:
The unit of analysis will be individual patient in the case of individual trials. The unit of analysis will be the cluster in the case of a cluster randomized study design. If we identifycluster‐randomized studies meeting the inclusion criteria, we will carry out appropriate analysis adjusting for the effect of cluster randomization before including effects estimates in a meta‐analysis. If available, we will extract adjusted measures of effect from the trial reports. If only unadjusted data are available, we will adjust this data ourselves using the intracluster correlation coefficient (ICC). If the ICC is not reported, we will contact the study authors to obtain this, or borrow an ICC value from a similar study, or estimate the ICC. If the ICC is estimated, we will explore the robustness of analyses by conducting sensitivity analyses.
Dealing with missing data:
We will contact the trail authors to obtain the missing data. If it isnot successful, we will perform sensitivity analyses to investigate the impact of missing data on the primary outcomes. Data must lose credibility, at some degree of loss of follow‐up. However, wedetermined to include studies with more than 50% drop‐out preliminarily due to insufficient evidence base up to now.If the trial reported the number of patients for whomthe primary outcomedata is missing, we will adopt a worst‐case scenario approach toanalyze data.That is to say, patients with missing data in the control group will be considered treatment successes, while patients with missing data in the treatment group will be considered treatment failure. We will indicate where studies have used imputation (and which methods) in this review.We will compare each protocol analysis of the available data with the one that contains missing data. We will conclude with more confidence in the case that the effect estimatesare the same and there are significant differences between groups. We will interpret the results more carefully and make more conservative conclusions about the effect of the treatment in the case that the effect estimates of the two analyses differ.
Assessment of heterogeneity:
Heterogeneity will be assessed by visual examination of the forest plots, to determine the closeness of the point estimates to each other and the overlap of CIs. Chi2 test with a P value of 0.10 will be used to indicate statistical significance, and I2 statistic will be used to assess the heterogeneity with a value of 50% indicatingsubstantial heterogeneity. If there is significant heterogeneity, we will use the random effects model, or we will choose the fixed effects model. The analyses will be carried out using Statistical analysis of the Cochrane software.
Assessment of reporting biases:
We will assess the potential impact of small studies such as publication bias by examining the funnel plots.If an adequate number of trials (more than 10 trials) are included in the meta-analysis, we will examine the symmetry of funnel plot. If the funnel plot is asymmetric, there may be publication bias.
We will use RevMan Web to analyze the data. We will use RR with 95% CIs and a random‐effects model to pool their results in the meta‐analyses if there are sufficient clinically similar studies available. For dichotomous outcomes, we will calculate theRR for each study and then aggregate the data. For continuous outcomes, if all the trials measured results on the same scale, we will pool MDs between the treatment arms at the end of the follow‐up, otherwise we will pool SMDs. Where a meta‐analysis is inappropriate, we will summarize data in tables.
Subgroup analysis and investigation of heterogeneity:
We will perform the following subgroup analyses for people with AIDS-related diarrhea to investigate heterogeneity:
- Acute or chronic;
- Diarrhea cause by etiology;
- Type of interventions including but not limited to individual prescriptions of Chinese herbal medicine; a single herb or compound; dosage regimen;
- Sex, age and ethnicity.
We will conduct sensitivity analyses to explore the impact of losses to follow‐up on the effect estimates for the primary outcomes and to exclude studies considered to be high risk of bias. For dichotomous outcomes, we will vary the event rate within the missing patients from intervention and control groups within plausible limits. For continuous data, we will perform sensitivity analyses using methods described by Ebrahim 2013 and Ebrahim 2014 .
Summary of findings and assessment of the certainty of the evidence:
We will use the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach to evaluate the quality of evidence and present it in 'Summary of Findings' tables. We willassess the certainty of a range of evidence which associated with the studies providing data for meta‐analyses of prespecified outcomes by using the five GRADE considerations (study limitations, effectconsistency, imprecision, indirectness and publication bias). We will use methods and recommendations described in Chapter 14 of the Cochrane Handbook for Systematic Reviews of Interventions, and use GRADEpro software. We will use footnotes to justify all decisions that downgrade the certainty of evidenceandmake comments where necessaryto help readers understand the review.