Nationwide Heterogeneity in Hospital-Specific Probabilities of Rectal Cancer Understaging and Its Effects on Outcomes.

BACKGROUND: Rectal cancer patients who are understaged may not be offered the highest quality treatment modalities, which are based on an accurate assessment of preoperative staging. The objective of this study was to evaluate heterogeneity in the probability of being understaged at Commission on Cancer hospitals in the United States and to assess how this variation affects outcomes. METHODS: The 2006-2013 National Cancer Data Base was queried for clinical stage I-III rectal cancer patients who underwent resection. The initial clinical stage was compared with the "gold standard," pathological stage. A Bayesian multilevel logistic regression model was used to characterize variation in hospital-specific probabilities of being understaged (clinical stage < pathologic stage). Separate analyses assessed the impact of being understaged on positive circumferential resection margins (CRM), receipt of adjuvant chemotherapy, and 5-year overall survival. RESULTS: Among 12,684 patients who did not receive neoadjuvant chemoradiation and treated at 1176 hospitals, 3044 (24%) were understaged. After patient level risk-adjustment, a 24-fold difference in the probability of being understaged was observed between hospitals (range 3-72%, median = 15%). Understaging was independently associated with positive CRM [odds ratio (OR) 1.59, 95% confidence interval (CI) 1.39, 1.92] and receipt of adjuvant chemotherapy (OR 14.22, 95% CI 13.55, 18.88). Despite an increase in the delivery of systemic therapy after surgical resection, understaging was associated with worse survival (hazard ratio = 1.61, 95% CI 1.48, 1.95). CONCLUSIONS: Deficiencies in high-quality rectal cancer management begin with incorrect clinical staging. The risk-adjusted probability of understaging varied widely between hospitals. This institutional failure to provide optimal oncological management at the start of care was associated with worse long-term survival.