Prognostic value of postoperative CEA clearance in rectal cancer patients with high preoperative CEA levels

Abstract

Purpose: We determined the prognostic value of carcinoembryonic antigen (CEA) clearance after tumor resection with serial evaluation of postoperative CEA levels in rectal cancer.

Methods: Between 1994 and 2004, we retrospectively reviewed 122 patients with rectal cancer whose serum CEA levels were measured on the preoperative day and postoperative days 7 and 30. Patients with preoperative CEA levels <5.0 ng/ml were excluded. An exponential trend line was drawn using the three CEA values. Patients were categorized into three groups based on R(2) values calculated through trend line, which indicates the correlation coefficient between exponential graph and measured CEA values: exponential decrease group (group 1: 0.9 < R(2) < or = 1.0), nearly exponential decrease group (group 2: 0.5 < R(2) < or = 0.9), and randomized clearance group (group 3: 0.5 < or = R(2)). We then analyzed the CEA clearance pattern as a prognostic indicator.

Results: With a median follow-up of 57 months, the 5-year overall survival was 62.3% vs. 48.1% vs. 25% and the 5-year disease-free survival was 58.6% vs. 52.7% vs. 25% among groups 1, 2, and 3 (P = 0.014, P = 0.027, respectively) in patients with stage III rectal cancer. For those with stage II rectal cancer, the 5-year overall survival rate of group 1 was significantly better than groups 2 and 3 (88.8% vs. 74.1%, respectively, P = 0.021).

Conclusions: The postoperative pattern of CEA clearance is a useful prognostic determinant in patients with rectal cancer. Patients with a randomized pattern of CEA clearance after tumor resection should be regarded as having the possibility of a persistent CEA source and may require consideration of intensive follow-up or adjuvant therapy.

Fig. 1

CEA values (preoperative, postoperative day 7, postoperative day 30) over time (a). An exponential trend line was drawn using each CEA value. R² values were calculated as the deviation between calculated curves and the measured CEA value. The function of the exponential curve also was calculated (Y = pX^−q). The value “q” represents the gradient; it refers to the acute or slow decrease in CEA following tumor resection (b)

Fig. 2

The survival curves according to the R² value in stage II rectal cancer. Group I represents the exponential decrease group compared with group II, the nonexponential decrease group. a There was a significant difference between the two groups in OS (P = 0.021). b There was no statistically significant difference between the groups in DFS (P = 0.167)

Fig. 3

The overall survival (OS) curves according to the R² value in stage III rectal cancer. Group I was the exponential decrease group compared with group II, the nonexponential decrease group. a There was a significant difference between the two groups (P = 0.039). The group also was divided to three subgroups: group 1 was the exponential decrease group; group 2 was the nearly exponential decrease group; and group 3 was the randomized clearance group. b Group 3 had significantly worse OS than group 1 (P = 0.014)

Fig. 4

The disease-free survival (DFS) curves according to the R² value in stage III rectal cancer. Group I was the exponential decrease group compared with group II, the nonexponential decrease group. a There was no significant difference between the two groups (P = 0.098). The group was further divided into three groups: group 1 was the exponential decrease group; group 2 was the nearly exponential decrease group; and group 3 was the randomized clearance group. b Group 3 had significantly worse DFS than group 1 (P = 0.027)

Fig. 5

Overall survival curves according to gradients with categorized. Using gradients, patients were divided into three groups: gradient ≤ 0.4 (slow rate of decline); 0.4 < gradient ≤ 0.7 (moderate rate of decline); 0.7 < gradient (rapid rate of decline). The categorized three groups were significantly related to the preoperative CEA value (8.03 ± 1.25 ng/ml vs. 30.70 ± 4.25 ng/ml vs. 45.20 ± 1.25 ng/ml, respectively; P < 0.001); however, there was no significant difference in OS among them (P = 0.537)