Prediction of postoperative mortality in elderly patients with colorectal cancer

PURPOSE: This study was designed to develop a model for predicting postoperative mortality in elderly patients undergoing surgery for colorectal cancer. METHODS: This multicenter study was conducted by using routinely collected clinical data, assessing patients older than aged 80 years, with 30-day operative mortality as the primary end point. Data were collected from The Association of Coloproctology of Great Britain and Ireland database, encompassing 8,077 newly diagnosed colorectal cancer patients undergoing resectional surgery in 79 hospitals between April 2000 to March 2002, The Association of Coloproctology Malignant Bowel Obstruction Study, encompassing 1,046 patients with malignant bowel obstruction in 148 hospitals, between April 1998 to March 1999, and The Wales-Trent audit, encompassing 3,522 newly diagnosed colorectal cancer patients, between July 1992 to June 1993. A multilevel logistic regression model was developed to adjust for case-mix and to accommodate the variability of outcomes between the three study populations. The model was internally validated using a Bayesian resampling technique and tested using measures of discrimination, calibration, and subgroup analysis. RESULTS: A total of 2,533 patients satisfied the inclusion criteria, with a 30-day mortality of 15.6 percent. Multivariate analysis identified the following independent risk factors: age (odds ratio for 85-90, 90-95, > 95 vs. 80-85 = 1.1, 1.8, 2.9), American Society of Anesthesiology grade (odds ratio for Grade III, IV vs. I-II = 2.7, 6.1), operative urgency (odds ratio for emergency vs. elective = 1.9), no cancer excision vs. resection (odds ratio = 1.2), and metastatic disease (odds ratio for metastases vs. no metastases = 1.9). The model offered adequate discrimination (area under receiver operator curve = 0.732) and excellent agreement between observed and predicted outcomes during eight colorectal procedures (P = 0.885). CONCLUSIONS: The elderly colorectal cancer model can accurately estimate 30-day mortality in patients older than aged 80 years undergoing surgery for colorectal cancer. Because the mortality can be considerable, this may have important implications when determining management for this group of patients.