Critical review · May 7, 2026
ViV-TMVR versus redo-SMVR: the Sá et al. meta-analysis
Sá MP, Neves G, Consoli L, Iqbal A, Dexheimer L, da Hora DAB, et al. Valve-in-Valve Transcatheter Mitral Valve Replacement Versus Redo Surgical Mitral Valve Replacement: Meta-Analysis of Early and Late Outcomes. J Am Heart Assoc. 2026;15(9):e050299.
doi:10.1161/JAHA.126.050299 · Systematic review and meta-analysis of 13 observational studies, with individual data reconstructed from Kaplan–Meier curves, n = 15,941
Summary
The meta-analysis included 13 observational studies with 15,941 patients (ViV-TMVR 5465; redo-SMVR 10,476).
| Early outcome (ViV-TMVR vs redo-SMVR) | RR (95% CI) |
|---|---|
| In-hospital mortality | 0.72 (0.57 to 0.90) |
| Stroke | 0.49 (0.29 to 0.83) |
| Bleeding | 0.43 (0.20 to 0.94) |
| Acute kidney injury | 0.57 (0.42 to 0.77) |
| Pacemaker | 0.30 (0.19 to 0.49) |
Length of stay was 5 days shorter with ViV-TMVR. Thirty-day mortality did not differ.
At 5 years, survival did not differ: HR 0.92 (0.81 to 1.05). The proportional hazards assumption was violated (Schoenfeld p = 0.002). The authors used a flexible parametric B-spline model and a 6-month landmark: HR 0.69 (0.58 to 0.83) before and HR 1.47 (1.20 to 1.79) after.
Strengths
- Meta-analysis done properly for the most part, searching four databases.
- Coherent early results: a less invasive procedure with less acute morbidity.
- Proportional hazards violation tested and handled with a flexible model.
- The authors acknowledge the lack of adjustment and are cautious in their conclusions.
Methodological issues
- ViV-TMVR patients were frailer. They were older, with more diabetes, COPD and atrial fibrillation, and a higher STS score in 5 of the 7 studies that reported it. In-hospital mortality favored ViV-TMVR, but 30-day mortality did not. This implies excess deaths between discharge and day 30 in the ViV-TMVR arm, as expected in frailer patients.
- Post hoc cutoff. The 6-month landmark was chosen by visual inspection of the HR(t) curve generated by the same data. The reversal is real; the exact split is post hoc.
- No patient-level adjustment. Individual data reconstructed from published curves carry no covariates, propensity matching or E-value. The late HR of 1.47 fits at least three causal models: lower ViV-TMVR durability (underexpansion, residual gradients, hypoattenuated leaflet thickening); shorter baseline life expectancy; or a mix of both.
- Low E-value. For HR 1.47, the E-value at the point estimate is about 2.3. Residual risk, frailty index and severity of bioprosthetic failure can all clear that bar.
Opinion
The study shows that ViV-TMVR is a procedure with less acute morbidity, in the populations currently selected for it.
The study does not show that redo surgery gives better late survival in comparable patients. The reversal at 6 months fits confounding by indication as well as prosthesis durability. Valve deterioration driving mortality as early as 6 months seems unlikely to me; confounding by indication explains it better.
The authors are cautious in the text. To me, the problem is how the study will be cited: as proof that redo surgery offers better late durability.