Do predictive scoring systems have any role in determining treatment strategy for colorectal liver metastases at the individual patient level?

Abstract

e15104 Background: the literature contains four prognostic scoring systems from large volume series predicting survival after resection for colorectal liver metastases. Our aim was to directly compare these systems in terms of accuracy in predicting survival, and evaluate the role of such systems in determining therapeutic strategies on intention to treat at the individual patient level. Methods: the original published data were reviewed and the median survivals plotted against each point in the scoring system. The best fit line was drawn and the resultant plots enabled calculation of slope of the linear regression line and the square of the Pearson product moment correlation coefficient through data points allowing all four prognostic scoring systems to be directly compared. Results: when the prognostic scores were compared on a common scale, the R-squared values of all the scoring systems were similar with R-squared values of greater than 0.94 in all cases. The Basingstoke Predictive Index however produced the steepest gradient suggesting it can better distribute patients along a wider range of survival. Conclusions: none of the four scoring systems was perfect in their predictive value. It is not surprising that the risk score from MSKCC and Leeds are not as good as the others since they are meant to be preoperative prognostic scores and postoperative scoring system by their nature of having more information should be better. Clearly, future developments in this field must incorporate advances in our understanding of tumor biology. [Table: see text]