Consistency in treatment effects and statistical significance between early-stage pregnancy endpoints and live birth in randomized controlled trials of infertility: a systematic meta-epidemiological study.

IMPORTANCE

Live births are the "gold standard" for assessing infertility treatments but are less frequently reported than clinical pregnancy. Identifying approaches to interpret randomized controlled trials (RCTs) without live birth data is essential.

OBJECTIVE

To evaluate the correlation between treatment effects and the consistency of statistical significance in conclusions drawn from live birth versus early-stage pregnancy endpoints (an umbrella term defined as biochemical, clinical, or ongoing pregnancy) in RCTs reporting both outcomes.

DATA SOURCES: We systematically searched EMBASE, MEDLINE, and CENTRAL for RCTs in infertility from January 1, 2012 to August 30, 2023.

STUDY SELECTION AND SYNTHESIS: Randomized controlled trials involving subfertile women reporting contingency tables for live birth and at least one early-stage pregnancy endpoint were eligible. Contingency tables on pregnancy or live births were manually extracted from the trial publications. We calculated Spearman's rho for treatment effects between early-stage pregnancy endpoints and live birth and compared their statistical significance using Chi-square tests. The above analyses were then conducted in prespecified subgroups.

MAIN OUTCOMES: The correlation between treatment effects and the consistency of statistical significance in conclusions drawn from live birth versus early-stage pregnancy endpoints.

RESULTS

Among 8,757 records, 1,281 infertility RCTs were eligible. The Spearman's rho for treatment effects was 0.78 (95% confidence interval [CI]: 0.69-0.86, P<.001; 169 RCTs) between biochemical pregnancy and live birth, 0.87 (95% CI: 0.83-0.90, P<.01; 429 RCTs) between clinical pregnancy and live birth, and 0.96 (95% CI: 0.92-0.98, P<.01; 138 RCTs) between ongoing pregnancy and live birth. Statistical significance between early-stage pregnancy endpoints and live birth was consistent in above 88% of trials. The correlation of treatment effects between clinical or ongoing pregnancy and live birth remained robust in subgroups of infertility treatments, including ovarian stimulation for medically assisted reproduction treatment.

CONCLUSION AND RELEVANCE: A strong correlation was observed between clinical or ongoing pregnancy and live birth, supporting the rationale for using clinical or ongoing pregnancy data to assess treatment effectiveness when live birth data are limited or unavailable. However, the risk of intervention-related pregnancy loss must be considered during the assessment.