In Reply We thank Zhu and Wang for their interest in the International TNT Study. In our series, 12.1% of patients were offered a watch and wait strategy, and 31.9% of these had salvage surgery for local regrowth, which occurred at a median of 34 (IQR, 24-46) weeks after treatment completion. In only 1.7% and 22.0% of the cases, respectively, local regrowth was accompanied or followed by tumor progression/recurrence elsewhere. While increasingly offered in clinical practice, nonoperative management mostly remains an opportunistic strategy for select patients rather than a planned approach primarily aiming at organ preservation. Accordingly, if followed by an R0/R1 salvage surgery, isolated local regrowth during watch and wait suggests failure of complete response assessment rather than treatment failure and true recurrence after a curative-intent strategy. In line with international consensus recommendations, we excluded isolated, R0/R1 resected, local regrowth during watch and wait from the censoring events for either local progression or event-free survival (EFS), and we do not share the concerns about this handling underestimating local events and inflating EFS. Regarding multiple imputation for missing data, we did acknowledge this as a limitation of our study. Notably, the proportion of patients with no available information for each prognostic variable was reported, overall and by treatment regimen, in the supplementary material. Of the 61.2% of patients tested for mismatch repair deficiency (dMMR)/microsatellite instability (MSI-H), 3.6% had dMMR/MSI-H tumors. Collecting information on the subsequent use of immune checkpoint inhibitors on disease progression was out of scope. Outcomes of our TNT-treated population by mismatch repair/microsatellite status will be reported in a separate article. We agree that covariate balance diagnostics after propensity vector matching are informative. Balance before matching can be assessed from the table in the manuscript. We assessed postmatching balance for categorical baseline variables by computing the proportion in each group at each level of the variable, then computing the absolute between-group difference in proportions for each pair of groups. For age at diagnosis, we compared means and 25th, 50th, and 75th percentiles across groups. The largest differences in proportions were observed for cN stage; cN2 was disproportionately represented in the RAPIDO-like and OPRA induction-like groups (52% and 53%, respectively) relative to the PRODIGE23-like (33%) and OPRA consolidation-like groups (41%), while cN1 disproportionately represented in the RAPIDO-like group relative to the PRODIGE23-like group (37% vs 52%). The next largest difference observed was in low tumor grade between the PRODIGE23-like and OPRA induction-like groups (42% vs 28%). All other differences were 11 percentage points or fewer. For age at diagnosis, the maximum between-group differences in the 25th, 50th, and 75th percentiles were 2, 3.5, and 3 years, respectively. Finally, regarding the maturity of survival estimates, we agree that the similarity of results across treatment regimens should be interpreted cautiously, and this is the reason why we emphasized 3-year rather than 5-year EFS and overall survival probabilities in the article. We do not think that interpretation would be improved by using restricted mean survival times that are less intuitive for clinicians despite their worth in settings marked by nonproportional hazards, which does not appear to be the case in our study.
