Balancing Immediate Surgical Risks with Deferred Benefits in Preventive Surgery

Preventive surgeries, such as carotid endarterectomy (CEA), involve immediate procedure risks for every patient, yet only survivors experience benefit. Conventional analyses such as Kaplan-Meier estimates and Cox models fail to adequately address competing risks and the timing of events, limiting their ability to assess the risk benefit trade off. This study aimed to evaluate alternative methodologies, specifically competing risk analysis and restricted mean survival time (RMST), using a landmark CEA trial as an example.

Methodological re-evaluation was performed on reconstructed individual patient data from the ACST-1 trial, a randomised clinical trial of 3 120 patients comparing immediate vs. deferred CEA for asymptomatic carotid stenosis. Data were derived from published Kaplan-Meier curves and supplementary materials, with mortality modelled from a comparable study. Key parameters for the reconstructed data fell within 1% of the original data. The Aalen-Johansen estimator and RMST were used to quantify stroke free days and life years gained or lost.

Accounting for competing risks reduced Kaplan-Meier based ten year stroke risk estimates by 20% and lowered the estimated risk reduction for immediate CEA from 6.1% to 5.6%. Alternative analysis strategies did not indicate a difference in stroke free days between the two arms at 10 years: difference, 40 stroke free days (95% confidence interval [CI] -52 - 129 days; p = .38). Across mortality scenarios, the overall balance between prevented stroke related deaths and procedure deaths resulted in a small net disadvantage for immediate CEA ranging from 1 (95% CI -21 - 39; p = .53) to 16 (95% CI -39 - 4; p = .12) life days lost over 10 years.

Competing risk analysis and RMST overcome the limitations of traditional analyses, enabling comprehensive evaluation of the surgical trade off in preventive surgery and potentially altering study conclusions based on traditional approaches. By providing more tangible outcome measures, this framework facilitates intertrial comparisons and medical decision making.