Is it possible to apply trial outcomes to a real-world population? A novel approach to External Validity Analysis

Translation of findings from randomised controlled trials (RCT), the foundation of evidence-based medicine, into clinical practice requires an understanding of relationships between patient characteristics, treatment practices and outcomes. We propose a novel technique, External Validity Analysis (EVA), to evaluate applicability of findings from a large RCT, comparing baseline characteristics, interventions and outcomes between the RCT and a large clinical database.

To perform EVA of the findings of a randomised controlled trial (ESTHER-1) to a population in an Australian clinic setting. To demonstrate this method, we evaluated the discordance in first cycle follicle-stimulating hormone (FSH) exposure and outcomes between the two populations, to inform clinical practice.

In this retrospective, descriptive analysis, we compared practices and outcomes between the follitropin alfa 'conventional' dosing arm of the ESTHER-1 trial and a selected comparable clinic subpopulation of patients who underwent controlled ovarian stimulation using FSH.

Mean FSH exposure was 34% higher in the clinic subpopulation than in the trial subpopulation, resulting in higher average ovarian response without improving the likelihood of clinical pregnancy or live birth.

EVA allowed for the comparison of a trial population with a selected clinic population with similar characteristics. With respect to FSH consumption, this analysis revealed higher exposure to FSH in the clinic setting without a corresponding benefit. The comparison reveals population differences as well as the potential to improve clinical outcomes through a reappraisal of current practices and objectives in gonadotropin dose selection.