Selection Criteria, Data Collection, and Judgements
Scoping questions and a list of beneficial and harmful outcomes of the intervention were sent by e-mail to an international panel of experts. Members of the panel were invited to comment on the relevance of the questions, to modify them if required, and to add additional relevant questions. Panel members were asked to rate each beneficial and harmful outcome on a scale of 1 to 9. A "critical outcome" was defined as an outcome that scored on average between 7 and 9. Those outcomes that scored between 4 and 6 on average were considered "important but not critical," while those scoring less than 4 were considered "not important."
Systematic reviews were used to summarize the evidence from randomized trials related to interventions for prevention of postpartum haemorrhage (PPH). Titles identified from the searches for reviews and assessed for the quality of relevant reviews were screened by two reviewers using checklists. For each question, data were extracted for all of the outcomes that were judged to be important, beginning with the most recent review of good quality and supplementing that with additional data from other good quality reviews that addressed the same question.
Evidence profiles were created using the GRADE approach. Using this approach, assessments of the quality of evidence for each important outcome take into account the study design, limitations of the studies, consistency of the evidence across studies, the directness of the evidence and the precision of the estimate. A liberal approach to assessment of study limitations was taken. Three main criteria were used for assessing trial limitations: concealment of allocation, blinding and follow-up. If most of the evidence for an outcome (based on the weight given to each study in the meta-analysis) came from trials that did not have serious limitations, the overall assessment for that outcome was that there were no important limitations.
If data were available as continuous outcomes, such as mean blood loss, absolute differences were presented as weighted mean difference (WMD). All estimates of effect size were expressed as relative risk if it was possible to calculate it from the data provided, with absolute risk estimates included where appropriate. In order to provide the panel with a broad and informative set of measures of effect, the numbers needed to treat (NNTs) and numbers needed to harm (NNHs) were calculated for each outcome. In systematic reviews, for each outcome, the lowest and highest baseline risks were extrapolated from control groups across the studies. The minimum and maximum NNTs and NNHs were therefore calculated, providing a range of values for these measures.
One reviewer extracted data from the reviews and prepared drafts of the evidence profiles with detailed footnotes explaining the judgements that were made. These were checked by at least one other member of the team and discussed with the team that prepared the background documentation.
All of the evidence profiles and additional tables were sent to the members of the Technical Consultation Group for review prior to the technical consultation.
Summary of Findings Tables
The key findings for each question were summarized in tables with the most important findings from the systematic reviews together with additional information from randomized clinical trials.