Predicting Post-Operative Mortality in Colorectal Cancer Surgery: a Systematic Review of the Accuracy of Possum, P-Possum and Cr-Possum
Colin Richards*, Fiona Leitch, Paul G. Horgan, Donald C. Mcmillan
University Department of Surgery, Glasgow Royal Infirmary, Glasgow, United Kingdom
IntroductionThe Physiological and Operative Severity Score for the enUmeration of Mortality and morbidity (POSSUM) model and its Portsmouth (P-POSSUM) and colorectal (CR-POSSUM) modifications are used extensively monitor post-operative outcome. This systematic review assessed the predictive accuracy of each POSSUM model after colorectal cancer resection. MethodsThe review was undertaken according to a predefined protocol. Major electronic databases, including Medline, Embase, Cochrane Library and Pubmed were searched using appropriate MeSH and free text terms. Abstracts were scanned and full text obtained for potentially relevant articles. The reference list of each article was then hand searched. Each study was entered into an electronic database using a data extraction tool. Two independent reviewers assessed each study against explicit inclusion criteria (colorectal cancer-specific data, actual and predicted mortality rates). Predictive accuracy for mortality was assessed by calculating weighted observed:expected (O:E) ratios with 95% confidence intervals (CI) for each POSSUM model. Results345 abstracts were scanned before 48 articles were progressed to data extraction. After applying inclusion criteria, 18 studies, published between 1991 and 2009 were included in the final review. 10 studies (4799 patients) reported data on POSSUM, 17 studies (6576 patients) reported data on P-POSSUM and 13 studies (4998 patients) reported data on CR-POSSUM. The majority of operations were elective open colorectal cancer resections. Pooling of the data returned a weighted O:E ratio of 0.31 (CI 0.17-0.45, range 0.11-0.80) for POSSUM, 0.90 (CI 0.52-1.28, range 0.20-2.36) for P-POSSUM and 0.62 (CI 0.43-0.80, range 0.21-1.11) for CR-POSSUM.ConclusionThere was also significant heterogeneity in the reported O:E ratio for each POSSUM model, in particular P-POSSUM. Pooling of the data demonstrated P-POSSUM as the most accurate with a weighted O:E ratio nearest to 1. POSSUM had a weighted O:E ratio furthest from 1. CR-POSSUM, despite development specifically for use within colorectal surgery, was not as accurate as P-POSSUM and consistently over-predicted mortality.
Back to Program
