Describing Successful Multimodal Prehabilitation Before Colorectal Cancer Resection: A Recursive Partitioning Analysis of Pooled Trial Data.

BACKGROUND: Multimodal prehabilitation has been shown to improve outcomes after colorectal cancer resection. However, it remains unclear which patient subgroups exhibit more favorable postoperative outcomes following multimodal prehabilitation and which characteristics characterize these subgroups. We sought to identify and describe clinically interpretable subgroups of prehabilitation participants, defined by combinations of baseline characteristics, within which differing rates of 30-day postoperative complications were observed following colorectal cancer resection. METHODS: Using a sample of participants randomized to the intervention arms of six trials of multimodal prehabilitation, recursive partitioning (classification tree) models were generated to identify subgroups with reduced 30-day complication rates. Candidate classifiers were selected for modifiability through prehabilitation and their established relevance to surgical outcomes. Two models were generated, one using baseline patient characteristics and another using post-prehabilitation characteristics. RESULTS: Among 135 (58.4%) men and 96 (41.6%) women, 34.2% (79/221) experienced 30-day postoperative complications. In the baseline tree model, those subgroups associated with the lowest complication rates had a six-minute walk distance (6MWD) ≥519 m with a RAND Short Form 36-Item Physical Functioning Score (SF-36 PFI) ≥56 (14.9%; 95% confidence interval [CI], 8.3%-25.3%) or had a combination of SF-36 PFI ≥56, 6MWD 314 to 518 m, and low risk of malnutrition (as defined by Patient-Generated Subjective Global Assessment Score <3) (6.2%; 95% CI, 1.1%-28.3%). In the post-multimodal prehabilitation tree, the lowest 30-day complication rate was observed among patients who achieved a 6MWD ≥487 m at multimodal prehabilitation completion (22.2%; 95% CI, 15.4%-30.9%). For those who could not reach this distance, ≥41 days of prehabilitation offset an otherwise elevated rate, reducing it from 46.4% to 31.8% (95% CI, 19.5%-45.7%). CONCLUSIONS: This study identifies and describes clinically interpretable subgroups of prehabilitation participants defined by combinations of baseline characteristics, within which differing rates of 30-day postoperative complications were observed following colorectal cancer resection. These subgroup profiles are intended as hypothesis-generating summaries of potential response patterns to multimodal prehabilitation, rather than as predictive tools or estimates of independent variable effects.