Response

We thank Xia et al. for raising some interesting points. Firstly, we recognize the peculiarity of rectal tumors, and therefore an exploratory analysis differentiating rectal vs descending or sigmoid colon was conducted and is available in the Supplementary Materials. Secondly, an increasing number of prognostic variables in metastatic colorectal cancer (CRC) has been identified in the last years, but many of them were not collected in clinical studies conducted a decade ago. All available data with a possible prognostic implication were included in the multivariable model to estimate the independent role of tumor location. Thirdly, the relative impact of the addition of bevacizumab was estimated through an adjusted interaction test between tumor location and treatment effect.

We read with interest data from the Netherlands Cancer Registry reported by van Erning et al. As compared with our analysis of PROVETTA, AVF2107g, and NO16966, results by van Erning et al. confirm in a large set of 7104 patients the worse prognosis of right- vs left-sided metastatic CRC. On the other hand, this finding is not replicated in a large cohort (n = 43 482) of nonmetastatic CRC patients, where no statistically significant difference in overall survival (OS) is evidenced between right- and left-sided tumors. Notably, in a subgroup of 5457 patients with available data on recurrence, similar OS and shorter relapse-free survival (RFS) are reported for left-sided tumors. Recent data from other registries also have shown that right-sided nonmetastatic CRCs may have better outcome, with discordant results according to histopathological stage: right-sided stage II CRCs, but not stage I and III, seem to have better prognosis, thus suggesting that stage II disease may have a different molecular makeup (1–4).

How can these conflicting results be explained? Recent advances in the field of molecular characterization of CRC evidenced a highly statistically significant association of right location with both BRAF mutation and microsatellite instability (MSI-high). The negative prognostic impact of BRAF mutation in metastatic CRC is now well established (5,6), and it is retained also in the event of MSI-high (7), occurring in around 5% of metastatic CRCs (8). Conversely, the retrospective analysis of adjuvant trials demonstrated a favorable impact of MSI-high on nonmetastatic CRC patients’ prognosis (9), while no association of BRAF mutation with RFS was evidenced (10,11). The protective role of MSI-high in nonmetastatic patients is confirmed also in the presence of BRAF mutation (12).

When looking at right-sided tumors, the prevalence of MSI-high is around 20% to 25% when compared with 5% to 7% in left-sided tumors, thus making crucial the estimation of the relative weight of MSI-high on the good prognosis of right-sided tumors in the nonmetastatic population. In our opinion, this molecular information would add substantial knowledge to results from the Netherlands Cancer Registry. The replication of findings by van Erning et al. in independent series would be needed in order to deepen the potential contribution of primary tumor location for clinical decision-making and trial design in nonmetastatic patients.