I. L. Leeds1, J. E. Efron1, N. Ahuja1, E. R. Haut1, F. M. Johnston1 1Johns Hopkins University School Of Medicine,Department Of Surgery,Baltimore, MD, USA
Introduction:
There is a growing consensus that common surgical prediction models do not adequately predict surgical outcomes for cancer surgery patients. The purpose of this study was to examine the comparative performance of these models in benign versus malignant gastrointestinal surgery populations to assess their utility as a risk-counseling tool for providers and patients.
Methods:
The American College of Surgeons’ National Surgical Quality Improvement Program’s (NSQIP) 2014 participant use file was queried for patients undergoing elective surgery with a primary procedure code of a major colon, pancreas, or stomach resection. All patients with a diagnosis of gastrointestinal cancer were compared against those without using Chi-square and Wilcox-Mann-Whitney tests. Multivariable logistic regression was performed to identify independent predictors of mortality and morbidity.
Results:
We identified 37,809 cases (16,322 malignant, 21,487 benign; 30,789 colectomies, 5,963 pancreatectomies, 1,057 gastrectomies). All of the following reported differences were statistically significant. The gastrointestinal cancer surgery population was disproportionately male, older than 65, non-white, and less functionally independent. Major comorbidities more prevalent in the non-cancer surgery group included chronic steroid use, preoperative sepsis, and active smoking. Comorbidities more common in the cancer surgery population included disseminated malignancy, diabetes, hypertension, dyspnea, and COPD.
Cancer surgery patients had a longer length of stay (+0.9 days), higher mortality rate (0.9% vs. 0.5%), higher complication rate (26.9% vs. 21.2%), higher total number of complications per surgery (0.5 vs. 0.3), broader variation in total number of complications (σ2= 1.0 vs. 0.8), higher readmission rate (11.4% versus 10.6%), higher pneumonia incidence (2.5% vs. 1.9%), higher reintubation rate (2.1% versus 1.3%), extended ventilator-days (1.6% vs. 1.1%), more kidney injuries (1.4% vs. 0.9%), more in-hospital cardiac arrests (0.7% vs. 0.4%), more blood transfusions (12.0% vs. 6.4%), and more blood clots (1.4% vs. 1.1%).
NSQIP probability variables performed similarly for complications in cancer and non-cancer patients (r= 0.32 vs. 0.31) but underperformed for mortality (r= 0.24 vs. 0.14). Multivariable regression controlling for predictors and procedure type demonstrated that a diagnosis of cancer requiring surgery independently confers an increased risk of death of 34.4% (p=0.034) and 12.0% (p<0.001) increased risk of additional complications as compared to patients with benign disease.
Conclusion:
NSQIP prediction models less effectively evaluate the risk of death in cancer patients as compared to patients undergoing similar resections for benign disease. Furthermore, a diagnosis of cancer independently increases the risk of complication and death from surgical resection. Additional counseling of short-term outcomes uncertainty may be needed.