A modeling-based evaluation of the evidential basis for and cost effectiveness of intensive post-diagnosis extra-colonic surveillance of non-metastatic colorectal cancer patients

Abstract

Between 70-80% of colorectal cancer (CRC) patients present with non-metastatic disease and can potentially be cured with surgical resection. However, between 5-60% of these patients will suffer a recurrence, generally in the form of late-occurring metastatic disease. For this reason, most professional-society guidelines recommend intensive extra-colonic-focused surveillance (CT scans and routine testing for tumor-markers) of these patients for 3-5 years post-diagnosis with the aim of detecting recurrence at an earlier stage when it is more likely to be amenable to salvage surgery with a curative intent. Until recently, this practice was corroborated by the results of meta-analyses of randomized control trials (RCTs) comparing more intensive with less intensive (or no) surveillance. However, the negative results of two large recently-published RCTs – the UK FACS trial and the Italian GILDA trial - and of subsequently updated meta-analyses have cast doubt on the value of aggressive follow-up and ultimately the value of aggressive treatment of recurrent CRC. In this dissertation I use a modeling analysis to argue that the results of these two trials have been misinterpreted. Accordingly, the conclusions of the most recent meta-analyses are misguided and calls to throw in the towel on intensive follow-up are premature. The negative trial results are not surprising given the low recurrence rates of contemporary practice and thus the small proportion of patients who could potentially benefit from aggressive follow-up. I show that, if aggressive follow-up were to confer a survival advantage in virtue of increasing the chances of salvage therapy with a curative intent, the average benefit would be very small. Moreover, the two trials would have had essentially no chance to detect an effect of that size, and this problem of insufficient power was likely exacerbated in at least one of the trials by a sizable chance recurrence imbalance. I further show that it is unlikely that a RCT with adequate power could ever or will ever be possible. However, I argue there is reason to take seriously the hypothesis that aggressive use of follow-up testing and subsequent salvage therapy can offer a small survival advantage on average. Finally, I report the results of a modeling-based cost-effectiveness analysis to identify follow-up strategies that would be cost-effective if this hypothesis is correct.