Background Postoperative upper-limb morbidity following breast cancer treatment is common and clinically significant, encompassing pain, functional impairment, range-of-motion limitation, and lymphedema. While traditional risk models have focused primarily on surgical and treatment-related factors, less is known about the relative contribution of psychosocial and behavioral determinants. The Arm Morbidity following Breast Cancer Treatment (ARM-BCT) tool was developed as a clinically practical risk stratification instrument; however, its performance and underlying predictors require further evaluation in larger and more heterogeneous populations. Objectives To identify multidimensional predictors of postoperative upper-limb morbidity following breast cancer treatment and evaluate the relative contribution of psychosocial, behavioral, clinical, and treatment-related factors across complementary analytical approaches. Methods This pooled analysis included 1,602 women assessed 0–36 months following breast cancer surgery across three prospective cohorts. Postoperative morbidity was defined as a composite outcome including pain, range-of-motion limitation, functional impairment, or lymphedema. Multivariable logistic regression and complementary machine-learning approaches were applied to evaluate the relative contribution of clinical, treatment-related, psychosocial, and lifestyle-related factors to postoperative morbidity. Model discrimination was assessed using the area under the receiver operating characteristic curve (AUC). Results Postoperative morbidity was observed in 61% of participants. Across analytical approaches, psychosocial and behavioral factors demonstrated consistent associations with postoperative morbidity alongside several clinical and treatment-related variables. Insomnia (OR 2.91, 95% CI 2.19–3.87) and emotional distress (OR 1.18 per unit increase, 95% CI 1.12–1.25) were independently associated with increased morbidity, along with comorbidity (OR 1.80, 95% CI 1.37–2.36) and chemotherapy exposure (OR 2.01, 95% CI 1.39–2.89). Higher physical activity levels during survivorship follow-up were also associated with morbidity, although this relationship may partly reflect reverse causation. Several traditional surgical variables demonstrated relatively limited independent associations after multivariable adjustment compared with the multidimensional psychosocial, behavioral, and treatment-related framework. The ARM-BCT score demonstrated modest discrimination (AUC 0.68), whereas multivariable regression and complementary machine-learning approaches demonstrated improved discrimination (AUC 0.79–0.81). Conclusions Postoperative arm morbidity following breast cancer treatment appears to reflect a multidimensional survivorship outcome influenced by psychosocial, behavioral, clinical, and treatment-related factors. While machine-learning approaches provided modest improvements in predictive performance, their primary contribution was in refining the interpretation of complex interactions between predictors. Integrating psychosocial and behavioral dimensions into postoperative risk assessment may support more personalized and proactive rehabilitation-oriented survivorship care strategies.
Klein et al. (Mon,) studied this question.