Serum CEA concentrations are usually measured for rectal cancer patients. Measurement of serum CEA levels is inexpensive, standardized for normal limit, widely used and easily performed, as compared with other potential prognostic markers for rectal cancer patients, such as CA 19–9, p53, ras expression, thymidine synthase, dihydropyrimidine dehydrogenase, 18q loss of heterozygosity, deleted in colon cancer (DCC) protein, DNA ploidy or micro-RNA signature [11, 17, 18]. The micro-RNA signature may be of value in predicting pCR , but will need further validation. Several studies had focused on the predictive value of pre- and post-CRT CEA levels in patients with rectal cancer receiving pre-operative CRT [12–15, 19, 20]. In this study, we found not only pre-CRT CEA levels had prognostic significance, “normalization” of these values and CEA ratio also predicted tumor response and may be helpful in the design of individualized treatment for rectal cancer with high CEA levels before treatment.
There is some controversy as to the role of pre-CRT CEA in rectal cancer patients. It was reported that pre-CRT CEA levels >2.5 or >5 ng/ml were associated with poor pCR rates and poor disease-free survival on univariate analysis, but not for both on multivariate analysis [12, 13]. However, in other studies, pre-CRT CEA levels were a common predictor of downstaging, pCR and tumor response on multivariate analysis [14, 15]. As the pre-CRT CEA levels increase, the rates of good response might decrease. In a recent study, for pre-CRT CEA <3, 3–6, 6–9 and >9 ng/ml, the rates of good response were 36%, 24%, 16% and 8%, respectively . On the other hand, post-CRT CEA level >5 ng/ml was associated with decreased pCR rates and disease-free survival on univariate and multivariate analysis [12, 19]. The literature mentioned above suggests that pre- or post-CRT CEA level 3–6 ng/ml may serve as a valuable threshold for prognosis and prediction of pathologic tumor response. The change of CEA levels before and after CRT were also investigated in some Korean studies [16, 20]. Park et al. categorized locally advanced rectal cancer patients into 3 groups according to pre- and post-CRT CEA levels (> or ≦3 ng/ml). They concluded that these groupings could be of clinical value as a predictor of response (TRG) to preoperative CRT and as an independent prognostic factor. However, the idea of CEA reduction ratio (like our CEA ratio) had not been previously mentioned until Kim’s study. They also classified rectal cancer patients into 3 groups according to pre- and post-CRT CEA levels, though with different cut off values (> or ≦6 ng/ml), For patients with pre-CRT CEA >6 mg/ml, they were further divided by whether post-CRT CEA was ≥70% lower than pre-CRT CEA. Similarly, they had a better 5-year DFS for the lower pre-CRT group and higher pre-CRT group with a CEA reduction ratio ≥ 70% than the other group. However, they could not definitely explain the pathological basis behind the CEA reduction ratio; and it was unknown whether pCR were related to this ratio.
Our study implied that pre-CRT CEA levels may be prognostic of local tumor control but may not be predictive of pCR. Higher pre-CRT CEA levels could be related to more advanced locoregional spread and thus associated with poorer local control, but not necessarily reflect sensitivity to CRT. The 5-year DFS rates of groups L, H-L and H-H among our patients were compatible with the two Korean studies mentioned above. Group H-H carried significantly higher risks of ypN1-2 disease and pathologic stage III-IV than groups L and H-L. This could explain why group H-H had the worst 5-year DFS rate of the three groups. Both ypN stage and pCR were parameters representing tumor response to CRT and, in our multivariate analysis, were independent prognostic factors for DFS. Though CEA change groups were not, this grouping made before surgery was obviously related to these two parameters (Table 1). Perhaps, the small number of patients may also make CEA change group statistically less prognostic. Furthermore, both ypN stage and pCR are pathologic features, limiting their prognostic value before operation.
We tried to find a clinical parameter predicting the pathologic response. CEA change groups were relevant to pCR, but may not be significant enough. Interestingly, we found male gender was a negative predictor for pCR in univariate analysis, although not anymore in multivariate analysis. Post-CRT CEA level was the only independent predictor for pCR in our data. By using ROC curves, post-CRT CEA level was a better predictor than pre-CRT CEA level or CEA ratio (Figure 3A), and the optimal cutoff value of post-CRT CEA was 2.61 ng/ml. However, when restricting to those with pre-CRT CEA <6 ng/ml, this observation was weakened and not statistically valuable (Figure 3B). On the other hand, for those with pre-CRT CEA ≥6 ng/ml, the AUC was even larger for post-CRT CEA level to predict pCR, and CEA ratio had value as good as post-CRT CEA level did (Figure 3C). This suggested that the pathologic tumor response was related to post-CRT CEA levels rather than to pre-CRT CEA levels, especially for those with high pre-CRT CEA levels.
RT-surgery interval is a well-known predictor for tumor response to pre-operative CRT, but we did not identify this even in univariate analysis, which could be due to less variance of the interval in our patients. The mean RT-surgery intervals of the patients with post-CRT CEA <2.61 ng/ml and those with post-CRT CEA ≥2.61 ng/ml were 44.6 days (range, 27–87) and 44.8 days (range, 24–79), respectively (p = 0.946). The comparable results supported that post-CRT CEA groups could predicted pCR independently. Besides, when dividing all patients into two groups according to the post-CRT CEA cutoff value 2.61 mentioned above, lower post-CRT CEA group correlated with earlier pathologic stages (p = 0.031; not shown in the result section) and better overall survival (p = 0.005).
The change of serum CEA levels before and after CRT seems to be more obvious in patients with pre-CRT serum CEA ≥6 ng/ml than those with normal levels (mean CEA ratio ± standard deviation, 0.36 ± 0.25 vs. 0.87 ± 0.30; p < 0.001); accordingly, a low post-CRT CEA level or CEA ratio may represent a marked reduction of tumor burden after CRT, especially when pre-CRT CEA levels are higher than the normal limit. We observed pCR may be even more predictable (larger AUC) when using post-CRT CEA level or CEA ratio in those with pre-CRT CEA levels ≥6 ng/ml; the optimal cutoff values of post-CRT CEA level and CEA ratio were 2.87 ng/ml and 0.22, respectively (not shown in the result section), with the same sensitivity 87.5%, specificity 76.7% and maximum Youden index (0.642). But, it was still not reasonable to conclude that surgery could be totally omitted for those with elevated pre-CRT CEA and low post-CRT CEA levels (i.e. low CEA ratios). We hypothesized that significant CEA reduction may play a role in supporting significant tumor regression for patients with both high pre-CRT CEA and clinical good response after CRT. For some highly selected cases like these, more conservative surgery may be used to preserve the sphincter without compromising local control .
In this study, all the patients were treated according to the same treatment protocol of pre-operative CRT, which had been proved to be effective and tolerable . However, several limitations to this study exist, including retrospective design, relatively small patient numbers, imbalance of case numbers in different groups, variation in RT-surgery intervals, and inconsistent principle and regimen for adjuvant chemotherapy. Many patients were excluded due to absence of serum CEA levels or due to different laboratory techniques for measurement.