Explain Conflicting Studies Without Counting Votes
When studies appear to disagree, first check whether they asked the same question. Counting favourable and unfavourable papers can hide differences in design, outcome, precision, and context.
The worked examples are teaching aids. Check them against your own source material and the requirements of your specific task.
Start with the task
Cochrane's interpretation guidance treats disagreements as something to explain within the evidence context.
Align the questions
Put the population, intervention, comparator, outcome, and follow-up of each study beside one another.
In an invented assignment, one study reports better immediate recall after supported teaching; another finds an uncertain difference in later self-reported understanding after written information alone. These findings are not necessarily contradictory.
If the questions do align, compare the estimates and uncertainty rather than just the authors' conclusion words.
Examine the methods
Different risks of bias, missing data, measurement choices, or implementation may help explain the results. Identify what the reports actually support; do not invent a methodological flaw to dismiss an inconvenient study.
Two estimates can point in the same direction while only one meets a significance threshold. Conversely, apparently similar summary labels can conceal materially different effects.
The number of papers on each side is not a substitute for this comparison.
Write the remaining uncertainty
A useful synthesis states where the studies agree, where they differ, and what that means for the specific rationale. You may conclude that the available evidence does not support a strong claim.
Do not resolve uncertainty by choosing the source with the most confident abstract or the most convenient conclusion. Where appropriate, use a credible review that has examined the body of evidence.
The rationale guide helps keep the argument explicit. Nursevi can preserve conflicting findings and limitations instead of turning them into an unsupported single recommendation.