Competing risks analysis of the association between perioperative blood transfusion and long-term outcomes after resection of colorectal cancer.

Abstract

Despite numerous reports over three decades, the association between perioperative blood transfusion and long-term outcomes after resection of colorectal cancer remains controversial. This cohort study used competing risks statistical methods to examine the association between transfusion and recurrence and colorectal cancer-specific death after potentially curative and noncurative resection. A hospital database provided prospectively recorded clinical, operative and follow-up information. All surviving patients were followed for at least 5 years. Data were analysed by multivariable competing risks regression. From 2575 patients in the period 1995-2010 inclusive, after exclusions, 2334 remained for analysis. Among 1941 who had a potentially curative resection and 393 who had a noncurative resection the transfusion rates were 24.9% and 33.6%, respectively. After potentially curative resection there was no significant bivariate association between transfusion and recurrence (HR 0.93, CI 0.74-1.16, P = 0.499) or between transfusion and colorectal cancer-specific death (HR 1.04, CI 0.82-1.33, P = 0.753). After noncurative resection there was no significant association between transfusion and cancer-specific death (HR 0.93, CI 0.73-1.19, P = 0.560). Multivariable models showed no material effect of potential confounder variables on these results. The competing risks findings in this study showed no significant association between perioperative transfusion and recurrence or colorectal cancer-specific death.