From RCTs to RWE in ovarian stimulation: the latest French evidence comparing gonadotropins

Annalisa Racca's lecture addresses how far the conclusions of randomised controlled trials (RCTs) can be transferred into everyday clinical practice when choosing a gonadotropin for controlled ovarian stimulation. The classic RCT remains the gold standard of evidence, yet its strict inclusion criteria produce a study population that often differs from the real mix of patients seen in a reproductive clinic.

This gap between the ideal conditions of a trial and routine practice is precisely what real-world evidence (RWE) sets out to close. Unlike an RCT, an RWE analysis draws on large observational datasets that capture the full diversity of patients, protocols and clinical decisions. The talk sets out exactly how these two sources of evidence differ and why they should be seen as complementary rather than competing.

Central to the lecture is the analysis of one of the largest national databases in France, which aggregates data on stimulation cycles across the whole country. This volume of observations makes it possible to compare the gonadotropins in use under conditions close to real practice and to assess outcomes in a sample no single randomised trial could reach.

The comparison of products is examined through the lens of both the biological differences between recombinant and urinary gonadotropins and the clinical outcomes of stimulation. The lecture discusses how the shift from RCT to RWE changes the interpretation of product comparability and which endpoints — from the number of oocytes retrieved to cumulative outcome measures — are most informative when assessed in a real-world population.

At the same time, the observational nature of RWE calls for cautious interpretation: without randomisation, comparisons between products are shaped by systematic differences between groups arising from clinician choice and patient characteristics. The lecture stresses that the value of large real-world datasets emerges only with sound statistical handling and awareness of potential bias, and that such data complement rather than replace the controlled experiment.