Clinical significance of discordance between carcinoembryonic antigen levels and RECIST in metastatic colorectal cancer

Abstract

Purpose The purpose of this study was to investigate the prognostic implications of carcinoembryonic antigen (CEA) levels that are inconsistent with Response Evaluation Criteria in Solid Tumor (RECIST) responses in metastatic colorectal cancer patients. Materials and Methods We retrospectively evaluated 360 patients with at least one measurable lesion who received first-line palliative chemotherapy. CEA-response was defined as CEA-complete response (CR; CEA normalization), CEA-partial response (PR; ≥ 50% decrease in CEA levels), CEA-progressive disease (PD; ≥ 50% increase in CEA levels), and CEA-stable disease (SD; non-CR/PR/PD). Overall survival (OS) and progression-free survival (PFS) were evaluated according to CEAresponse. Results In RECIST-PR patients, poorer CEA-response was associated with disease progression at the subsequent evaluation. In RECIST-SD patients, CEA-CR and -PR were associated with lower disease progression rates than CEA-PD at the subsequent evaluation. Correlations between survival outcome and CEA-response in same-category RECIST patients were assessed. In RECIST-PR patients, discordant CEA-response (CEA-PD/SD) was associated with poorer survival than CEA-CR/PR (median OS and PFS, 44.0 and 15.4 [CEA-CR], 28.9 and 12.5 [CEA-PR], 21.0 and 9.8 [CEA-SD], and 13.0 and 7.0 [CEA-PD] months, respectively; all p < 0.001). In RECIST-SD patients, favorable CEA-response produced better survival (median OS and PFS, 26.8 and 21.0 [CEA-CR], 21.0 and 11.0 [CEA-PR], 16.1 and 8.2 [CEASD], and 12.2 and 6.0 [CEA-PD] months, respectively; all p < 0.001). RECIST-PD patients with CEA-CR showed longer OS than those with CEA-PD. Multivariate analysis demonstrated that discordant CEA-response is a powerful prognostic factor for RECIST-PR and RECIST-SD patients. Conclusion Among patients of the same RECIST-response categories, CEA-response patterns are significantly prognostic and strongly predictive of subsequent evaluation outcomes.