ABSTRACT: Carcinoembryonic antigen (CEA) monitoring in patients with stage I-IV colorectal cancer has been, and remains, a controversial issue in oncology practice. Recommendations vary from bimonthly monitoring to no monitoring in the surveillance setting (for stage I-III disease). In the metastatic setting, there are no clear guidelines for CEA follow-up, although continued monitoring in such patients is common in the oncology community. This manuscript reviews the accuracy of CEA testing, its value as a prognostic indicator, and its role in surveillance and response assessment. The limitations of the test in the adjuvant and metastatic settings are illustrated through several case reports from the Colorectal Oncology Clinic at Roswell Park Cancer Institute. Guidelines for CEA monitoring are provided, based on a detailed literature review and institutional experience.
In 1965, Gold and Freedman described an oncofetal antigen expressed in human fetal colonic tissues and in colonic carcinomas but not in adult colon; they named it carcinoembryonic antigen (CEA).[1,2] CEA was subsequently characterized as a glycosylated cell surface glycoprotein with a molecular weight of 180,000 daltons. Variance in CEA glycosylation between normal and tumor tissue explains the difference in reported molecular weights. CEA is detectable in the serum through a radiommunoassay technique first developed by Thomson et al in 1969. It is present at very low serum concentrations in healthy adults and at high concentrations in a variety of cancers, particularly epithelial tumors.
Sensitivity and Specificity in Colorectal Cancer
A lower limit of normal varies according to the performing laboratory but usually ranges from 2.5 to 5 ng/mL. Serum CEA was < 2.5 ng/mL in more than 87% and < 5 ng/mL in more than 95% of 1,020 subjects attending primary prevention clinics. Elevated CEA levels are more common in smokers and in patients with inflammatory conditions but rarely exceed 10 ng/mL.[7-9] The test can also be elevated in a variety of other carcinomas, including lung, breast, gastrointestinal, and gynecologic cancers.
Detecting Primary Colorectal Cancer
The sensitivity and specificity of CEA depends on the immunoassay in place, the tested patient population, and the performing facility.[11-13] The sensitivity of CEA for early colon cancer patients is low and increases with an increasing stage of the disease. In a study of 358 patients who presented to surgery with a diagnosis of colon cancer, only 4% of patients with stage I disease had an elevated CEA (> 5 ng/mL), whereas 25%, 44%, and 65% of patients with stage II, III, and IV disease, respectively, had abnormal levels. In another study of 319 surgical patients, CEA was elevated in only 26% of resectable patients and in 72% of patients with unresectable or metastatic disease.
Fletcher reviewed the sensitivity and specificity of this test at various stages of disease. The sensitivity was 36% with a specificity of 87% for a CEA > 2.5 ng/mL in patients with stage I and II disease. The sensitivity for stage III and IV disease at similar CEA levels was 74% and 83%, respectively. The sensitivity decreased while the specificity increased for higher CEA cutpoints (5 ng and 10 ng). The poor reliability of this test in early colorectal cancer and the imperfect specificity in the normal population makes this test unsuitable for primary colorectal cancer screening.
Detecting Disease Recurrence
CEA has been studied extensively as a marker for disease recurrence in patients undergoing curative intent resection of a colorectal cancer primary. Patients with resected colorectal cancer should normalize their CEA levels within weeks from surgery, and failure to do so by 4 months is highly suspicious for systemic disease.[17,18] This time frame of a few weeks to several months is consistent with an estimated CEA half-life of 3 to 13 days; longer half-lives are associated with higher preoperative CEA levels.
Specificity and sensitivity in detecting disease recurrence depends largely on the definition of abnormal CEA levels (cutoff CEA). The higher the cutoff for abnormal CEA levels, the higher the specificity and the lower the sensitivity.[20,21] Using a CEA cutoff of 5 ng/mL, Moertel and colleagues reported a sensitivity of 34%, specificity of 84%, and median lead time of 4.5 months from detection of clinical recurrence. Interestingly, the sensitivity of CEA depends on the site of recurrence, with sensitivities exceeding 70% for liver and retroperitoneal metastases and lower than 50% for lung, peritoneal, and locoregional recurrences. Solitary lung recurrence was detectable by CEA in only 15% of instances.
Other studies that have imposed strict CEA monitoring guidelines reported higher sensitivity rates. In a follow-up of 311 patients with potentially curative resections of colorectal cancer, CEA was measured every 3 months for 2 years, then every 6 months for 3 years. The sensitivity of CEA in detecting recurrence was 58%, with a median lead time of 6 months. The sensitivity was the highest for the hepatic metastases group (80%). The specificity, positive predictive value, and negative predictive value were 93%, 79%, and 83%, respectively.
Other investigations have confirmed sensitivities and specificities exceeding 60%.[23-25] The highest sensitivities reported (89% and 91%) were associated with studies implementing frequent CEA monitoring (4 to 8 weeks) with a low abnormal cutoff of 2.5 to 3 ng/mL.[21,26] It is important to note that most studies that evaluated the sensitivity and specificity of CEA in detecting recurrence did not stratify patients according to their preoperative CEA levels (before resection of the original primary). However, there is no evidence at this time that CEA would be a less effective test in detecting recurrences in patients with normal CEA at the time of original diagnosis. In fact, one study evaluated the frequency of elevated CEA upon recurrence in patients with a history of CEA-negative stage III disease (CEA < 5 ng/mL at time of diagnosis). In this report, 44% of patients had an elevated CEA at the time of recurrence, supporting the usefulness of the test in the surveillance of patients with a resected normal- and high-CEA primary.
CEA monitoring is more sensitive in detecting any disease recurrence than liver function tests, ultrasound of the liver, chest x-ray, or colonoscopy.[28-30] Sensitivity was shown to be comparable to computed tomography (CT) scan, but the combination was more effective in detecting recurrences than either modality alone.[21,31] In a surveillance study of 530 patients with resected stage II/III colon cancer, patients underwent a CT scan of the chest, abdomen, and pelvis at baseline, 12 months, and 24 months after initiation of adjuvant chemotherapy. The same patients underwent CEA testing every 3 months during the first year, every 6 months in the second year, and then annually. Overall, 32% of recurrences were detected by CT scan and 29% by CEA. Only 28% of recurrences detected by CT scan showed an abnormal CEA, suggesting that CT scan may offer an added advantage in improving detection of recurrences when added to CEA monitoring.
CEA as a Prognostic Marker
Given the positive association between an elevated CEA and the stage of colorectal cancer at time of initial presentation, it is natural to expect a higher risk of relapse in patients with an elevated preoperative CEA level. However, an association between CEA levels and disease outcome has been reported even when controlling for disease stage. A linear inverse correlation between preoperative levels and estimated mean time to recurrence in patients with stage II and III disease has been demonstrated.
Patients with stage III disease had a median time to recurrence of 13 months if preoperative levels were > 5 ng/mL, and 28 months if < 5 mg/mL. Patients with stage II disease were evaluated for outcome based on preoperative CEA level. Patients with CEA < 5 ng/mL had the best outcome, followed by patients with CEA of 5 to 10 ng/mL, and patients with CEA > 10 ng/mL had the worst outcome. A multivariate analysis accounting for nodal metastases and depth of invasion found preoperative CEA elevation (> 5 ng/mL) to be an independent prognostic factor for survival after curative-intent surgery. Another study suggested a similar prognostic value for patients with stage IV disease. Others failed to show a prognostic value for CEA when stage was controlled for.[34,35]
CEA has also been investigated as a prognostic factor in patients undergoing resection of hepatic metastases of colorectal origin. Several large series have shown that an elevated CEA prior to hepatic metastases resection is associated with a higher risk of disease recurrence and decreased overall survival, independent from other prognostic factors such as nodal stage, synchronous disease, and size and number of metastases.[36-41] Others have failed to confirm an association; however, those series included a small number of patients and were likely underpowered to adequately investigate CEA as a prognostic variable.[42-44]
Limited data are available regarding the cost-effectiveness of CEA in detecting potentially curable recurrences. In the follow-up of patients receiving adjuvant therapy on a large randomized study of 1,356 patients, 2.2% of resectable recurrences were attributed to CEA screening. The cost of CEA monitoring per detected resectable recurrence was $5,696, comparing favorably to colonoscopies and chest x-rays. A recent study suggests a cost of $25,289 for each surviving patient after surgery for CEA detected recurrence. Others have estimated that CEA costs are $500,000 per patient cured secondary to CEA screening, while some have suggested a range of $22,963 to $4,888,208 per quality-adjusted life-year saved.[46,47] Most estimates would be considered cost-effective when taking into consideration the current costs of treating metastatic colorectal cancer.