PARTNER 2A at 10 years: TAVI versus surgery at intermediate risk

Summary

Between 2011 and 2013, PARTNER 2A randomized patients with severe aortic stenosis at intermediate surgical risk to TAVI with SAPIEN XT or surgery. In the valve implant population, 974 received TAVI and 936 surgery. Mean age 81.6 years, mean STS score 5.8%.

Ten-year follow-up required reconsent at 5 years. Mortality was completed with vital status sweeps of public records, reaching about 90% completeness.

Mortality: HR 1.13 (95% CI 1.02 to 1.25). The excess was confined to transapical/transaortic access; transfemoral mortality was similar (83.9% vs 82.1%). Kaplan–Meier reintervention: HR 4.26.

Strengths

• Randomized TAVI-versus-surgery trial followed for a decade, in an elderly intermediate-risk cohort.
• Reintervention analyzed as a cumulative incidence function with Gray’s test, treating death as a competing risk. This is the right estimand in a cohort with about 83% ten-year mortality.
• The naive Kaplan–Meier was reported as a sensitivity analysis, flagged as overstating risk.
• Mortality recovered to about 90% completeness through vital status sweeps.
• The authors describe their own follow-up limitations.

Methodological issues

• Nonfatal outcomes outside the vital status sweep. Reintervention relies on the on-study cohort alone. At year 6, mortality has about 380/396 patients at risk per arm; reintervention has 174/159. After the 5-year reconsent, durability is estimated on less than half the randomized population.
• Differential and possibly informative dropout. Among surgical patients alive at 5 years, 59.3% (270/455) declined reconsent. They left during years 5 to 10, the window in which bioprosthetic failure is most likely. Dropout was higher with surgery. If reconsent tracks health, censoring is informative and can bias the cumulative incidence itself, in favor of surgery.
• Few events in the surgical arm. The 1.6% comes from eight events, three of them in the late window.
• Fragile result. I reconstructed the published curves (Aalen–Johansen, reproducing 6.3% / 1.6% and a p-value close to the reported Gray’s test). About ten unobserved surgical reinterventions erase significance; about 25 reach parity. Across the roughly 425 surgical study-leavers, ten events is a reintervention rate of 2% to 4%.
• Surgical rate probably underestimated. About 15% of the surgical valves were Mitroflow and Trifecta, both with known early failure; Mitroflow reintervention at 10 years reached about 22%. In NOTION at 10 years, with near-complete follow-up, surgical severe structural deterioration was about 10% and bioprosthetic valve failure about 14%. Reintervention also undercounts failure, because frail octogenarians are often not reoperated, as the authors note.