Highlight
– A National Cancer Database (NCDB) analysis of 53,651 patients undergoing proctectomy for stage I–III rectal adenocarcinoma identified an overall 30-day mortality of 1.1% and eight independent predictors of 30-day death.
– The authors converted these predictors into a three-tiered risk score with a stepwise increase in 30-day mortality (0.8%, 1.9%, 4.5%) and very high specificity (≈99.5–99.6%) and accuracy (≈98.7–98.8%).
– Key actionable predictors include age, male sex, Black race, high comorbidity burden, pathologic stage III, prolonged hospital stay, and conversion to open surgery; neoadjuvant systemic therapy was associated with lower odds of 30-day mortality.
Background
Rectal cancer surgery—particularly proctectomy—remains a complex procedure with substantial perioperative risk in selected patients. Although 30-day mortality after elective colorectal surgery has fallen over decades, short-term mortality remains a critical quality metric and driver of perioperative decision-making, informed consent, and resource allocation. Predictive models to identify patients at elevated short-term mortality risk can guide preoperative optimization, postoperative care intensity, and shared decision-making. The study by Emile et al. (Surgery. 2025) uses the National Cancer Database (NCDB) to derive and validate a simple, registry-based score for 30-day mortality after proctectomy for localized rectal adenocarcinoma.
Study design
This is a retrospective case-control analysis using the NCDB from 2010–2017 to develop the model with external validation in an NCDB cohort from 2018–2019. Inclusion criteria were patients with stage I–III rectal adenocarcinoma who underwent proctectomy. Cases were defined as patients who died within 30 days of surgery and were compared to patients alive at 30 days across patient demographics, comorbidity (Charlson score), tumor characteristics (pathologic TNM stage), and treatment variables (neoadjuvant therapy, surgical approach including conversions, and hospitalization length).
Multivariable binary logistic regression identified independent predictors of 30-day mortality. Significant predictors were transformed into a point-based risk score that defined three risk groups. The same scoring rules were applied to the validation cohort to confirm performance metrics.
Key findings
Population and outcome
– The development cohort included 53,651 patients (60.9% male). The overall 30-day mortality rate was 1.1%.
Independent predictors (adjusted odds ratios [OR] and 95% confidence intervals reported by authors)
– Increasing age (OR per year: 1.07; 95% CI: 1.05–1.08). This indicates a strong continuous effect of age on short-term mortality.
– Male sex (OR 2.19; 95% CI: 1.61–2.98).
– Black race (OR 2.16; 95% CI: 1.44–3.25).
– Charlson comorbidity index ≥3 (OR 1.86; 95% CI: 1.05–3.30).
– Pathologic TNM stage III (OR 1.66; 95% CI: 1.12–2.46).
– Receipt of neoadjuvant systemic therapy was associated with lower 30-day mortality (OR 0.523; 95% CI: 0.296–0.925).
– Prolonged hospitalization (OR 1.02 per day; 95% CI: 1.01–1.03), reflecting that longer length of stay is both a marker and potential contributor to risk.
– Conversion to open surgery (OR 1.59; 95% CI: 1.13–2.23).
Score development and performance
– The regression model was simplified into a score that stratified patients into three risk groups. Observed 30-day mortality rose linearly across groups: 0.8% (low risk), 1.9% (intermediate), and 4.5% (high risk) (P < .001 for trend).
– In the development and validation cohorts, score specificity was extraordinarily high (99.6% and 99.5%, respectively), and overall accuracy was 98.7% and 98.8%.
Interpretation of metrics
– Very high specificity and accuracy are driven in part by the low overall event rate (1.1%). High specificity implies the score rarely misclassifies survivors as high risk; however, sensitivity (probability of detecting those who will die) was not highlighted in the abstract, and with a low event rate such models often trade sensitivity for specificity.
Clinical implications
Risk communication and informed consent
– The score provides a straightforward, registry-derived way to quantify short-term mortality risk. It may augment preoperative counseling by providing objective risk strata rather than solely relying on clinician gestalt.
Perioperative management
– Patients in the higher risk group (observed 30-day mortality 4.5%) warrant consideration of enhanced perioperative pathways: focused prehabilitation, optimization of medical comorbidities, geriatric assessment when appropriate, early ICU stepdown planning, and more intensive early postoperative surveillance to detect and treat complications promptly.
Quality improvement and resource use
– The score could inform allocation of perioperative resources (e.g., higher nurse-to-patient ratios, early physiotherapy, low threshold for imaging/diagnostics) and help identify populations for targeted interventions to reduce short-term mortality.
Health equity considerations
– The independent association between Black race and higher 30-day mortality after adjustment raises concerns about health system factors, access to timely care, differences in case mix not captured by registry variables, and potential implicit bias. The score could highlight populations requiring further process-level interventions.
Expert commentary and critical appraisal
Strengths
– Large, national registry dataset provides statistical power to detect relatively rare events and enables internal development with temporal validation.
– The model uses variables readily available in administrative/oncologic registries, facilitating application to similar datasets.
Limitations
– Registry data limitations: The NCDB does not include key physiologic and perioperative variables such as ASA class, functional status, frailty indices, detailed intraoperative physiology (blood loss, operative time), or postoperative complications timing and cause of death. Residual confounding is therefore likely.
– Event rarity and performance metrics: With a 1.1% event rate, predictive tools often achieve high specificity and accuracy by classifying most patients as low risk; sensitivity and positive predictive value (PPV) are crucial but were not prominent in the summary. A score that is highly specific but insensitive could miss many at-risk patients.
– Generalizability: Although validated temporally in the NCDB, external validation in clinically granular registries (e.g., ACS NSQIP) and prospective cohorts is necessary before routine clinical deployment.
– Interpretation of protective effect of neoadjuvant systemic therapy: This may reflect patient selection (fitter patients selected for multimodality therapy) or the benefit of tumor downstaging; causal inference cannot be assumed in a retrospective registry analysis.
Mechanistic and health-systems considerations
– The association of conversion to open surgery with higher mortality is plausibly related to increased physiologic stress, higher complication rates, or the circumstances that prompt conversion (difficult dissections, bleeding, obese/hostile pelvis). Prolonged hospital stay may be a marker of complications that themselves increase mortality risk rather than an independent causal factor.
Recommendations from an expert perspective
– Use the score as a screening tool to flag patients for more detailed, multidisciplinary preoperative assessment rather than as a sole determinant of care pathways.
– Prioritize external validation in data sources that capture physiologic variables and postoperative complications timing, and evaluate incremental value over existing risk calculators (for example, the ACS NSQIP risk calculator) and clinician judgment.
Conclusions and next steps
Emile et al. provide a pragmatic, registry-derived 30-day mortality risk score for patients undergoing proctectomy for stage I–III rectal adenocarcinoma with clear stratification of low-, intermediate-, and high-risk groups. The model’s high specificity and accuracy are promising, but limitations inherent to NCDB data and the low event rate temper enthusiasm for immediate clinical application without further validation.
Priority next steps
– External validation in surgical quality registries that capture physiologic and perioperative details (e.g., ACS NSQIP) to measure sensitivity, PPV, and net reclassification improvement compared with current tools.
– Prospective evaluation of whether identifying high-risk patients with this score and applying targeted interventions (prehabilitation, geriatric assessment, enhanced monitoring) reduces 30-day mortality.
– Investigation into drivers of racial disparities and system-level interventions to mitigate them.
In practice, clinicians should view the score as an adjunct to comprehensive preoperative assessment that may help prioritize optimization and monitoring in patients at higher predicted short-term risk.
Funding and clinicaltrials.gov
Funding: Not reported in the provided manuscript citation.
ClinicalTrials.gov: Not applicable (retrospective registry analysis).
References
1. Emile SH, Horesh N, Garoufalia Z, Gefen R, Zhou P, Wexner SD. Development and validation of a predictive score of 30-day mortality following proctectomy for rectal cancer: A National Cancer Database analysis. Surgery. 2025 Dec;188:109718. doi: 10.1016/j.surg.2025.109718. Epub 2025 Sep 29. PMID: 41027396.
2. Ljungqvist O, Scott M, Fearon KC. Enhanced Recovery After Surgery: A review. JAMA Surg. 2017;152(3):292–298.
3. Bilimoria KY, Liu Y, Paruch JL, et al. Development and evaluation of the ACS NSQIP surgical risk calculator: A decision aid and informed consent tool for patients and surgeons. J Am Coll Surg. 2013;217(5):833–842.e1–3.
Author note
This article is a structured, clinically focused interpretation of the NCDB analysis by Emile et al. It aims to help clinicians understand the strengths, limitations, and potential applications of a registry-derived 30-day mortality score after proctectomy for rectal cancer.
