Abstract
Background: Carcinoembryogenic antigen (CEA) is useful in the evaluation of chemotherapy response of metastatic colorectal cancer (CRC). We studied weekly CEA during one fluorouracil-based chemotherapy cycle, correlated with long-term (8-12 week interval) computed tomography (CT) and CEA responses. Patients and Methods: CEA, liver function tests and inflammatory parameters were measured prospectively at baseline, day 7, day 14, and after the cycle (day 21/28), in 60 patients with metastatic CRC. Results: CEA non-significantly decreased at day 7 and was increased on day 14. In progressive disease, CEA increased significantly during the evaluation cycle (55.4 μg/l vs. 148.2 μg/l; p=0.024), but the level was stable in patients with disease control (10.6 μg/l vs. 17.8 μg/l; p=0.58). CEA fluctuation correlated neither with liver function test nor with inflammatory parameters. Correlation of long-term response was most evident in progressive disease. Conclusion: CEA should not be measured during 5-fluorouracil-based oral chemotherapy nor within two weeks from intravenous chemotherapy administration
Colorectal cancer (CRC), the third most common malignancy in the Western world, represents the second leading cause of cancer-related mortality. Worldwide in 2008, of some 1,234 million new patients with colorectal cancer 609,000 died from metastatic disease (1).
Carcinoembryonic antigen (CEA), a tumour-associated rather than a tumour-specific marker, is the only widely recommended serum tumour marker for CRC (2). CEA is of limited value for primary diagnosis of CRC (3, 4), but should be determined before treatment to obtain a baseline value (5-9). CEA adds valuable information when evaluating prognosis (10, 11), in postoperative surveillance (12, 13), and when monitoring treatment efficacy during systemic therapy for metastatic CRC (14).
During treatment for metastatic CRC, CEA should be measured at the start of treatment and then every 8 to 12 weeks during active treatment according to the American Society of Clinical Oncology (ASCO) guidelines (14). Two values above baseline are adequate evidence of progressive disease, even in the absence of corroborating radiographs. Two reports question the ASCO definition of progressive disease (15, 16), both showing that initial elevation of CEA during chemotherapy for metastatic CRC does not always indicate disease progression. This transient CEA elevation is considered a tumour flare reaction or CEA surge, defined as an increase from baseline of >20% followed by a more than 20% drop in CEA levels (15). The ASCO GI Tumour Marker Guideline Update 2006 recommends caution when interpreting CEA elevations during the first 4 to 6 weeks of a new therapy, since spurious early rises may occur, especially after oxaliplatin use (2). In the guidelines no recommendation on the timing of CEA measurement during chemotherapy administration is given.
5-FU chemotherapy is generally administered every one, two or four weeks as intravenous boluses (Roswell Park, Nordic FLv, Mayo), every one to two weeks in continuous infusion (AIO, LV5FU2) and every three weeks in oral administration (capecitabine, carmofur, UFT). Modern combination chemotherapy with irinotecan, oxaliplatin, bevacizumab, cetuximab and panitumumab is based on these 5-FU chemotherapy backbones.
Tumour marker levels are often measured in the middle of the chemotherapy administration, in conjunction with nadir blood counts, in order to have relevant values available for evaluation at follow-up visits. To the best of our knowledge, the pattern of CEA fluctuation and optimal timing of measurement during a 3- to 4-week 5-FU-based chemotherapy cycle has received little attention and has not been correlated with repeated CEA or CT responses at 8- to 12-week intervals.
Patients and Methods
CEA evaluation was a secondary endpoint in an open, prospective non-randomized study, which included 60 consecutive patients with metastatic CRC patients at a single institution (Helsinki University Central Hospital), the primary end-point being the inflammatory response, which has been published previously (17). The protocols of the study were approved by the local Ethics Committee and the National Agency for Medicines, Helsinki, Finland, and informed consent was required from all patients. The patients were divided into three groups depending on the type of chemotherapy they received, with 20 patients in each group (Figure 1). The first group received a combination of raltitrexed and carmofur; the second group raltitrexed as a single-agent therapy; and the third group a 5-FU-based therapy, with 5-FU in combination with leucovorin (either the Mayo 5-day bolus injection regimen or the De Gramont infusional regimen), or the 5-FU prodrug carmofur.
Chemotherapy regimens. Raltitrexed at 3.0 mg/m2 was given as a 15- to 30-min infusion three times weekly. When given in combination with carmofur, raltitrexed at 1.5 to 3.0 mg/m2 was administered on cycle day 1 and carmofur at 300 to 400 mg/m2 orally was divided into three daily doses on cycle days 2-14, followed by a week of rest (18). The Mayo regimen was administered as bolus injections of leucovorin at 20 mg/m2 and at 5-FU 425 mg/m2 on days 1 to 5 of the cycle, repeated every four weeks. The simplified De Gramont regimen was given every two weeks with leucovorin at 400 mg/m2 as a 2-h infusion, followed by 5-FU at 400 mg/m2 given as a bolus, followed by a continuous infusion of 5-FU at 3.6 g/m2 for 48 h using a portable pump. Single-agent carmofur was administered in 3-week cycles, in which carmofur at 300 mg/m2, divided into three daily doses, was given orally for 14 days, followed by one week of rest.
Data collection. Collected data included type of treatment, treatment response, patient demographics, location of primary tumour, location of metastatic sites, performance status according to WHO, and serial weekly levels of laboratory tests, including CEA (AutoDELFIA®, HUSLAB, Helsinki, Finland), C-reactive protein (CRP), aspartate aminotransferase (AST), alanine aminotransferase (ALT), alkaline phosphatase (ALP), bilirubin (BIL), gamma-glutamyltransferase (GT), interleukin-6 (IL-6), interleukin-8 (IL-8), and tumour necrosis factor-α (TNF-α). The laboratory tests were run during one cycle of chemotherapy treatment and were taken at baseline (on day 0), day 7, day 14, and after the treatment cycle (generally on day 21, n=52; but in the Mayo and De Gramont regimen on day 28, n=8).
Whole-body CT and CEA were performed before the chemotherapy treatment cycle and repeated after 8 to 12 weeks. CT response was coded according to WHO criteria (Figure 1) (19). In this study, CT response served as the gold standard for evaluation of treatment response. Tumour marker progression was defined as a 20% or greater increase in CEA level. CEA values were missing for two patients on day 0 (n=58), for four on day 7 (n=56), five on day 14 (n=55), one on day 21 (n=59), and for six on both pre- and post-therapy (n=54).
Statistical analysis. All analyses were performed by the StatView software, version 5.0.1 (SAS Institute, Abacus Concepts Inc., Berkeley, CA, USA) or SPSS (PASW statistics version 18.0 Inc.; Chicago, IL, USA). Descriptive data are given as the median (range) for skewed distributions. CEA alteration was calculated as the difference (Δ) between days 21 and 0, as well as between post-therapy and pre-therapy values. The χ-squared test served for analysis of categoric data. The Wilcoxon signed-rank test served for paired comparisons and the Mann Whitney U-test or the Kruskall Wallis test for non-paired comparisons. Non-parametric correlation was calculated with Spearman's rho (Rs). A p-value less than 0.05 was regarded as statistically significant.
Results
Patients' characteristics. Patients' characteristics were generally well-balanced (Table I). Regarding the location of the primary tumour, there was a higher frequency of patients with rectal cancer in the group treated with raltitrexed and carmofur. A slight imbalance with more patients with WHO performance status 2 was noted in the group treated with single-agent raltitrexed.
CEA fluctuation during one chemotherapy cycle. Weekly CEA measurements often showed a wavelike variation, with lower values at day 7 and increased values at day 14, when compared with CEA on baseline day 0 or at day 21 (Table II). We found the CEA level to be significantly higher at day 21 than at day 0 (p=0.037), whereas the differences for the other comparisons were non-significant (CEA day 0 to day 7 p=0.36, CEA day 0 to 14 p=0.59). At cycle day 0, 84% of patients had an elevated CEA level (median 60 μg/l, range 5.2-2960 μg/l).
Chemotherapy regimens. The CEA levels categorized by chemotherapy regimen are presented in Table II. No significant CEA difference was apparent between chemotherapy regimens. The five largest fluctuation quotients of the CEA levels during one chemotherapy cycle according to chemotherapy regimens are presented in Figure 2A.
Early or late timing of the evaluation cycle. Dividing the evaluation cycles into two groups, early within CEA surge period (chemotherapy cycles 1 to 2, n=38) and later (cycles 3 to 15, n=22), showed no significant difference in the wavelike pattern (Table II).
CEA in relation to treatment efficacy. Median CEA levels were nearly four-fold higher in patients with progressive disease (PD) on CT than in patients with stable disease (SD) or partial response (PR) (Table II). The wavelike pattern was not as evident in the PD group. When the CEA level before and at the end of the cycle was compared according to treatment response, a correlation was found for the PD group (p=0.024), but no correlation was apparent for the PR (p=0.72) and the SD groups (p=0.72). Comparing the CEA level on day 0 with day 7 in the three response groups, the differences were not significant, nor were they significant between day 0 and day 14. Figure 2B shows data for five patients from each treatment response group, with the largest quotient between the highest and the lowest CEA values during the chemotherapy cycle.
In 19 patients with a true PD response on CT evaluation, 10 had a more than 20% increase in CEA levels from baseline to the end of the cycle. Thus, sensitivity for CEA was 53%. In 39 patients with disease control (PR or SD) on CT evaluation, 28 did not have a more than 20% increase in the CEA level from before the cycle to the end of the cycle. Thus, specificity for CEA was 72%.
Correlation of long-term CEA values with the one-cycle CEA value. Long-term CEA values were available for 54 patients in conjunction with CT evaluation, performed at 8- to 12-week intervals (Table II). The median pre-therapy CEA was 35.2 μg/l and that post-therapy CEA was 39.6 μg/l (p=0.079). CEA increased significantly in patients with PD response (77.0 μg/l vs. 190.0 μg/l; p=0.004), but was stable in SD (17.5 μg/l vs. 17.8 μg/l; p=0.21) and decreased non-significantly in PR (26.4 μg/l vs. 11.4 μg/l; p=0.20). In 17 patients with a true PD response on CT evaluation, 14 had a more than 20% increase in long-term CEA levels. Thus, sensitivity for CEA was 82%. In 37 patients with disease control (PR or SD) on CT evaluation, 23 did not have a more than 20% increase in long-term CEA level. Thus, specificity for CEA was 62%.
Correlation of single-cycle CEA alteration (Δ CEA day 21 to 0) with long-term CEA alteration (Δ CEA post-therapy and pre-therapy at 8- to 12-week intervals) was highly significant for patients with PD (median Δ was 19.5 μg/l vs. 83 μg/l, R=0.804; p<0.001), significant for SD (0.2 μg/l vs. 0.6μg/l, R=0.677; p=0.006), and non-significant for PR (0.3 μg/l vs. −1.6 μg/l, R=0.402; p=0.063).
Explanatory factors. Table III shows explanatory factors for treatment response in the radiological response evaluation and a CEA increase of at least 20%. Baseline WHO performance status, primary location, sex, chemotherapy regimen, early or later evaluation cycle and number of metastatic sites were not explanatory. Patients with liver metastases more often had a radiologically PD and at least a 20% CEA increase from baseline to the end of the cycle.
Correlation between CEA and liver function tests and inflammatory parameters. Liver function tests had a similar wavelike pattern to that of CEA, with a decrease at day 7 and an increase at day 14 and normalization at the end of the cycle. Median and 95% confidence intervals (CI 95%) for CEA and liver function tests are presented in Figure 3A. Pairwise day 0 to day 21 comparisons show no significant correlations between CEA and liver function test (Rs=0.046-0.225). No liver function tests predicted radiological response.
Inflammatory parameters showed a wave-like pattern different from that of CEA, with an increase at day 7 and a decrease at day 14 (Figure 3B). Pairwise day 0 to day 21 comparisons show a weak but statistically significant correlation between CEA and CRP (Rs=0.276, p=0.034), but no other significant correlations were found (Rs-0.051-0.074). No inflammatory parameters predicted treatment response.
Discussion
Two successive elevated CEA values, at least two months apart may be indicative of tumour progression, but radiology remains the gold standard for determining tumour status/response. When response evaluation with CEA measurement is planned every 8 to 12 weeks, very little is known about the optimal timing of CEA measurement during the actual 3- to 4-week chemotherapy administration cycle.
In most tumour marker studies, the measurements have been repeated at 4- to 15-week intervals, showing decent correlation with chemotherapy response (14). The study with the most frequent CEA measurement, every 14 days, was performed by Sorbye and Dahl but only for the first two months of chemotherapy, with oxaliplatin combined with bolus 5-FU (n=27) (15). Sorbye found a transient CEA increase in four patients (15%) and the time from start of chemotherapy to the CEA peak ranged from two to eight weeks. In another study, 10 out of 89 patients with metastatic CRC (nine with oxaliplatin and 5-FU), had a CEA surge, with the median peak at four weeks (range=2-10 weeks) from the start of chemotherapy (16). A CEA surge with initial tumour marker elevation has been evidenced for both metastatic CRC and other tumour types, namely non-seminomatous germ cell tumours (18), and in prostate and breast cancer with bone metastasis (20, 21).
In our study, the wavelike pattern of the CEA alteration during one cycle was similarly independent of whether the assessment was carried out during early treatment cycles (within the CEA surge period of six weeks from chemotherapy initiation) or during later cycles (6-45 weeks). The initial surge phenomenon, thus, does not explain the CEA fluctuation pattern.
The evaluation was carried out for patients receiving three different chemotherapy regimens for metastatic CRC with an anti-metabolite mode of action. Forty patients received 5-FU-based regimens (the Mayo, infusional and oral 5-FU prodrug), which are still the cornerstone in CRC chemotherapy. The wavelike patterns of variation were similar under all 5-FU regimens, as well as with the antimetabolite raltitrexed. This phenomenon has not been studied in patients receiving modern combination regimens, but since such regimens are all based on 5-FU, either intravenously or orally, it is probable that a similar CEA fluctuation will occur using these combinations.
Tumour markers are measured because they are considered to reflect chemotherapy efficacy. In this study, CEA alterations were correlated with response based on radiology, and in the four weekly CEA measurements during the chemotherapy cycle, only the difference between CEA at the end of each therapy cycle (day 21 or 28), when compared to baseline (day 0) correlated with treatment response. The 53% sensitivity and 72% specificity were disappointingly low. The CEA at days 7 and 14, however, did not predict response. Patients with SD and PR could not be assessed reliably within this short evaluation period. A longer time-frame for CEA assessment with the ASCO recommended 8- to 12-week interval was also performed but showed similar correlation only in patients with PD, with 82% sensitivity; specificity was dissappointingly low (62%).
Baseline-elevated ALP and lactate dehydrogenase (LDH) are associated with worse prognosis in many studies (22, 23). We did not analyze LDH here, but ALP in conjunction with other liver function tests was performed weekly. ALP was the only marker that correlated with CEA in patients with PD. No other liver function tests correlated either with treatment response or with CEA alterations.
Inflammatory parameters, especially CRP, have been evaluated in response prediction and prognostic evaluation (24). IL-6 and -8, together with TNF-α, reflect inflammation, liver damage, and tumour necrosis, but did not correlate with treatment response in our limited analysis, nor with alterations in the CEA level. Cytostatic agents, especially drugs that are inflammatory-reaction inducers, such as bleomycin, raltitrexed, oxaliplatin, and cytarabine, may affect tumour marker levels.
CEA is metabolized in the liver and it is well-known that increased CEA levels may occur in different liver diseases (25). Since chemotherapy is frequently associated with an increase in liver function test levels, one possibility could be that a transient increase in CEA level reflects liver damage. It is also possible that behind the tumour marker fluctuation lay shedding of antigen or tumour necrosis. The reason for tumour marker variation during single-chemotherapy cycle remains unknown.
Conclusion
The timing for tumour marker assessment is crucial in evaluation of response to therapy of metastatic CRC. CEA should not be measured during 5-FU-based oral chemotherapy, nor within two weeks of intravenous chemotherapy administration, because significant variation may occur. This variation showed a wavelike pattern in patients with disease control. Significantly increasing values at the end of the cycle were seen in patients whose disease was progressing in both short- (3-week) and in long-term (8- to 12-week) evaluation intervals. CEA alteration was not significant in patients with PR or SD.
Acknowledgements
This work was supported by grants from Finska Läkaresällskapet, The Kurt och Doris Palander Foundation, The Dorothea Olivia, Karl Walter och Jarl Walter Perklén Foundation, The Sigrid Jusélius Foundation and Medicinska understödsföreningen Liv och Hälsa.
Special thanks are due to Antti Hermunen for assistance with graphics and to Carol Norris for language editing.
Footnotes
-
↵* These Authors contributed equally to this study.
- Received October 15, 2012.
- Revision received November 9, 2012.
- Accepted November 12, 2012.
- Copyright© 2013 International Institute of Anticancer Research (Dr. John G. Delinassios), All rights reserved