Combining participant-level data can answer a focused question, but it does not create a third independent trial. This report integrated two randomized IVF studies. Primary report
What was analyzed
The analysis included 986 treated participants undergoing IVF with a long down-regulation protocol and excluded ICSI cycles. Live birth per initiated cycle was 26.5% with highly purified hMG and 20.8% with recombinant FSH. The reported odds ratio was 1.36, with a 95% confidence interval of 1.01–1.83. Primary report
Our appraisal: pooling needs a lineage map
The two source trials, their pooled analysis, and later economic reports may describe overlapping participants. Counting every publication as a separate replication would inflate the evidence base. A useful library should identify the shared data rather than hiding that relationship behind different paper titles.
The cycle denominator also matters. A rate per initiated cycle differs from a rate per embryo transfer because the latter excludes people who never reach transfer. Neither can be substituted for the other without changing the question.
An odds ratio is not a relative probability. The reported group rates provide the more direct starting point for understanding the absolute contrast. The interval's proximity to the null should remain visible alongside the point estimate, rather than presenting the result as unusually precise.
Finally, the protocol and ICSI exclusion limit transportability. This analysis does not establish the same contrast in every stimulation strategy, patient group, or laboratory setting. Full review should examine trial consistency, selection into the integrated dataset, missing outcomes, and the planned statistical model before extending the conclusion beyond this comparison.
Frequently asked questions
Was this a newly recruited third trial?
No; it combined data from two trials.
Were ICSI cycles included?
No.
What denominator was used for the cited rates?
Initiated IVF cycles.
Is the odds ratio a risk ratio?
No.
Can related publications be counted as independent cohorts?
Not when they reuse participants.
Does the result establish superiority in every IVF protocol?
No.
What should reviewers check next?
Dataset lineage, eligibility, trial consistency, and outcome completeness.