Estimands for Clinical Effectiveness of Risk-Reducing Early Salpingectomy in Women With High Risk of Ovarian Cancer
Importance
Risk-reducing early-salpingectomy (RRES) and delayed oophorectomy (DO) is a novel 2-stage alternative prevention strategy to risk-reducing salpingo-oophorectomy (RRSO) that avoids detrimental consequences of premature menopause. However, direct data on the clinical effectiveness for ovarian cancer (OC) risk reduction are lacking.
Objective
To explore how to define clinical effectiveness from prospective cohort studies using the estimand framework and sample size requirements.
Design, Setting, and Participants
In this comparative effectiveness research study, estimand and analysis options were considered to evaluate the clinical effectiveness of RRES with DO by extending the UK PROTECTOR cohort study, a multicenter, prospective, observational, national cohort study (N = 1250 recruited from January 1, 2019, to December 31, 2024) evaluating RRES and DO for OC surgical prevention. Eligibility criteria for participants were broadly premenopausal women 30 years or older at increased OC risk due to BRCA1/BRCA2 pathogenic variants. Participants could choose RRES, RRSO, or no surgery at entry. Sample-size requirements to extend recruitment used initial data (eg, age and BRCA1/BRCA2 distribution) from PROTECTOR (analysis undertaken from January 1, 2024, to December 31, 2025).
Main Outcomes and Measures
Incidence of OC after (not at) RRES and before or at DO in women with normal histologic analysis findings at surgery. The proportion of cancers prevented was estimated as the completement of the observed (O) to expected (E; assuming no preventive effect of surgery) number of cancers detected (1 – O/E).
Results
Initial data were obtained from 889 women in PROTECTOR (overall mean [SD] age, 39 [4.9] years), with 255 (28.7%) choosing RRSO (mean [SD] age, 42 [4.4] years), 405 (45.5%) choosing RRES (mean [SD] age, 38 [4.4] years), and 229 (25.7%) choosing no surgery (mean [SD], 38 [4.6] years). The preferred estimand outcome was OC incidence after surgery (RRES or RRSO) with a “while on intervention” strategy to account for intercurrent events. The primary target measure was the proportion of cancers prevented for RRES vs no surgery with superiority testing. A secondary target measure was noninferiority of RRES vs RRSO. An estimated 1150 RRES participants with 8 to 10 years of follow-up would provide approximately 92% power to show that 20% or more of cancers are prevented using a 1-sample binomial test of the O:E risk (external reference) at the 5% level under a range of assumptions and at least the same power for a noninferiority margin for the proportion of cancers prevented by RRES of those prevented by RRSO. Estimands based on incidence ratios had an infeasible sample size.
Conclusions and Relevance
For this comparative effectiveness study of UK BRCA carriers, the recommended estimand differed from other ongoing clinical-effectiveness studies of RRES and DO. Advantages include direct use of expected risk at baseline (unknown at design stage), easier interpretation across cohorts than absolute risk differences, and providing a feasible recruitment target for PROTECTOR to evaluate clinical effectiveness.