Fax: +358-9-471-71403; +358-9-471-75550
Predictive Markers and Cancer Prevention
Serum HCGβ, CA 72-4 and CEA are independent prognostic factors in colorectal cancer
Version of Record online: 27 AUG 2002
Copyright © 2002 Wiley-Liss, Inc.
International Journal of Cancer
Volume 101, Issue 6, pages 545–548, 20 October 2002
How to Cite
Louhimo, J., Carpelan-Holmström, M., Alfthan, H., Stenman, U.-H., Järvinen, H. J. and Haglund, C. (2002), Serum HCGβ, CA 72-4 and CEA are independent prognostic factors in colorectal cancer. Int. J. Cancer, 101: 545–548. doi: 10.1002/ijc.90009
- Issue online: 17 SEP 2002
- Version of Record online: 27 AUG 2002
- Manuscript Accepted: 15 JUL 2002
- Manuscript Revised: 8 JUL 2002
- Manuscript Received: 15 APR 2002
- Finska Läkaresällskapet
- Medicinska Understödsföreningen Liv och Hälsa
- Foundation for the Finnish Cancer Institute
- CA 19-9;
- CA 242;
- CA 72-4;
- colorectal cancer;
- tumour marker
In colorectal cancer, stage is considered to be the strongest prognostic factor, but also serum tumour markers have been reported to be of prognostic value. The aim of our study was to investigate the prognostic value of serum carcinoembryonic antigen (CEA), CA 19-9, CA 242, CA 72-4 and free β subunit of human chorionic gonadotropin (hCGβ) in colorectal cancer. Preoperative serum samples were obtained from 204 colorectal cancer patients, including 31 patients with Dukes' A, 70 with Dukes' B, 49 with Dukes' C and 54 with Dukes' D cancer. The serum levels of CEA, CA 19-9, CA 242 and CA 72-4 were measured with commercial kits with cut-off values of 5 μg/L for CEA, 37 kU/L for CA 19-9, 20 kU/L for CA 242 and 6 kU/L for CA 72-4. The serum hCGβ was quantitated by an immunofluorometric assay (IFMA) with 2 pmol/L as a cut-off value. Survival analyses were performed with Kaplan-Meier life tables, log-rank test and Cox proportional hazards model. The sensitivity was 44% for CEA, 26% for CA 19-9, 36% for CA 242, 27% for CA 72-4 and 16% for hCGβ. The overall 5-year survival was 55%, and in Dukes' A, B, C and D cancers the survival was 89%, 77%, 52% and 3%, respectively. Elevated serum values of all markers correlated with worse survival (p < 0.001). In Cox multivariate analysis, the strongest prognostic factor was Dukes' stage (p < 0.001), followed by tumour location (p = 0.002) and preoperative serum markers hCGβ (p = 0.002), CA 72-4 (p = 0.003) and CEA (p = 0.005). In conclusion, elevated CEA, CA 19-9, CA 242, CA 72-4 and hCGβ relate to poor outcome in colorectal cancer. In multivariate analysis, independent prognostic significance was observed with hCGβ, CA 72-4 and CEA. © 2002 Wiley-Liss, Inc.
Colorectal cancer is the most common gastrointestinal cancer in the Western world and it is an important cause of cancer-related death. Tumour stage is generally considered to be the strongest prognostic factor in colorectal cancer.1 With the development of more effective chemo- and radiotherapies, additional prognostic factors, such as tumour markers, could have an important role in identifying patients who would benefit the most from pre- or postoperative adjuvant therapies. In colorectal cancer, the most widely applied marker in follow-up of the disease is carcinoembryonic antigen (CEA), and its preoperative serum level has been shown to predict adverse prognosis.1, 2, 3 Elevated levels of cancer associated antigens CA 19-9, CA 242 and CA 72-4 have also been reported to be associated with poor prognosis.4, 5, 6, 7
Human chorionic gonadotropin (hCG) is a glycoprotein hormone produced at high concentrations by the placenta during pregnancy. It is also a very sensitive marker for trophoblastic tumours. HCG consists of an α and a β subunit, and the free β subunit (hCGβ) has been shown to be expressed in many nontrophoblastic cancers, including gastrointestinal ones.8, 9, 10
We have previously shown in another patient material that CA 242, specific tissue polypeptide antigen (TPS) and hCGβ have independent prognostic significance in colorectal cancer.6, 11 In our current study, we investigated the preoperative serum concentrations of CEA, CA 19-9, CA 242, CA 72-4 and hCGβ and their relationship with prognosis in colorectal cancer patients.
MATERIAL AND METHODS
From January 1990 to June 1995, 354 patients with histologically verified primary colorectal cancer were operated on at the Department of Surgery of the Meilahti Hospital, Helsinki University Central Hospital. Preoperative serum samples (taken within 30 days prior to surgery) were available from 243 patients. Of these, 38 were excluded because they had had another cancer previously or during the follow-up period, and 1 patient because of lack of follow-up and survival data. Clinical characteristics of the 204 patients included are shown in Table I. Stage of the tumours was determined according to the modified Dukes' classification.12 Clinical data was obtained from patient records, survival data until August 2001 from the Population Register Centre of Finland and cause of death data from the Statistics Finland. The minimum follow-up period was 6 years or until death; median follow-up time was 4.66 years (range 0.02–11.18 years).
|Variables||No. of patients|
|Age||22.7–92.5 years, median 66.6 years|
|Tumour location and Dukes' stage|
|Mucinous carcinoma||14 (7%)|
The ethics committee of the hospital approved our study.
Serum samples were stored at −20°C before assays. The serum levels of CEA, CA 19-9, CA 242 and CA 72-4 were measured with commercial assays. CEA was quantitated with a time-resolved immunofluorometric assay (AutoDELFIA®; Wallac, Turku, Finland). The detection limit of the assay is 0.2 μg/L and the interassay coefficient of variation (CV) is <3% in the concentration range 3–90 μg/L (total CV <4%). CA19-9 and CA72-4 were quantitated with immunoenzymometric assays (Immuno1®; Bayer, Tarrytown, NY). Detection limits of the assays are 0.8 kU/L and 0.3 kU/L, respectively. For CA19-9 the total CV is <3% in the concentration range 15–230 kU/L and for CA72-4 <6% in the range 6–51 kU/L. CA242 was quantitated with a manual immunoenzymometric assay (CanAg, Gothenburg, Sweden). The detection limit is 1 kU/L and interassay CV is <7% in the concentration range 13–134 kU/L. The cut-off values were 5 μg/L for CEA, 37 kU/L for CA 19-9, 6 kU/L for CA 72-4 and 20 kU/L for CA 242. The concentration of hCGβ in serum was quantitated with an in-house time-resolved immunofluorometric assay (IFMA) based on free hCGβ-specific monoclonal antibodies.13 The cut-off value for hCGβ (based on the 97.5 percentile of healthy blood donors) was 2 pmol/L.14
The association between marker positivity (serum value above the specified cut-off value) and stage was assessed by the χ2 test. The Mann-Whitney U-test was applied to determine the significance of the difference in marker concentrations between colonic and rectal tumours and between sexes. The correlation between serum levels of various markers was assessed by the Spearman-rank correlation test. Life tables were calculated by the Kaplan-Meier method. The log-rank test was applied to test the significance of differences in survival between groups with marker values above or below the cut-off values. Patients alive at the end of the follow-up period (August 2001) and patients who died from unrelated causes or postoperatively (within 30 days from operation) were censored. Multivariate survival analysis was performed with the Cox proportional hazards model. The covariates included in the model were: age (as a continuous variable), gender (female, 0; male, 1), tumour location (colon, 0; rectum, 1), Dukes' stage (categorical variable), histologic type (adenocarcinoma, 0; mucinous carcinoma, 1) and serum levels of tumour markers CEA, CA 19-9, CA 242, CA 72-4 and hCGβ. The tumour marker values were entered either as continuous variables as such or after logarithmic transformation. For selection of variables, the multivariate regression analysis was performed with a backward stepwise procedure using the likelihood ratio test with a limit of p = 0.05 for inclusion of a variable. Statistical analyses were performed with SPSS 10.07 software.
CEA was found to be the most sensitive marker (Table II). High serum tumour marker levels were associated with advanced disease, and the association between marker positivity and stage was significant with all markers (p ≤ 0.001). Median serum CEA level was 3.7 μg/L (range 0.6–6,536 μg/L), the median CA 19-9 was 11.0 kU/L (range 0.01–34,800 kU/L), the median CA 242 12.3 kU/L (range 0.01–13,654 kU/L), the median CA 72-4 1.8 kU/L (range 0.02–729 kU/L) and the median hCGβ 0.7 pmol/L (range 0.06–2,600 pmol/L). Between sexes, we observed no significant difference in tumour marker values (p > 0.279), except for hCGβ, which was significantly higher in females (p = 0.002). No significant difference in tumour marker values was observed between colonic and rectal tumours (p > 0.063), except for CA 72-4, which was significantly higher in colonic tumours (p = 0.019). The strongest correlation between 2 markers was observed between CA 19-9 and CA 242 (rs= 0.859, p < 0.001). Between all other markers, the correlations between 2 markers were modest (rs= 0.178–0.464, p < 0.011).
|n||CEA >5 μg/L||CA 19-9 >37 kU/L||CA 242 >20 kU/L||CA 72-4 >6 kU/L||hCGβ >2 pmol/L|
The overall 5-year survival was 55%, and with Kaplan-Meier life tables and log-rank test, all markers were found to be prognostic factors in colorectal cancer (p < 0.0001, Fig. 1). In the CEA-positive group (>5 μg/L), the 5-year survival was 30% and in the negative group 73%; in the CA 19-9-positive (>37 kU/L) and -negative groups the figures were 20% and 66%; in the CA 242-positive (>20 kU/L) and -negative groups 29% and 69%; and in the CA 72-4-positive (>6 kU/L) and -negative groups 27% and 65%, respectively. In the hCGβ-positive group (>2 pmol/L), the 5-year survival was below 16% (4 patients had survival over 5 years but were censored because they were alive at the end of the follow-up period) and in the negative group it was 62%.
In individual Dukes' stages, the overall 5-year survival was 89% for Dukes' A, 77% for Dukes' B, 52% for Dukes' C and 3% for Dukes' D. Within individual stages, CEA approached significance as a prognostic factor in Dukes' B (p = 0.056), while it was highly significant in Dukes' C (p = 0.0001). CA 19-9 was a significant prognostic marker in Dukes' A (p = 0.007) and in Dukes' B cancers (p = 0.014). CA 242 approached significance in Dukes' A, B and C cancers (p = 0.061–0.096), as did CA 72-4 in Dukes' B cancer (p = 0.070). However, CA 72-4 was a highly significant prognostic factor in Dukes' D cancer (p = 0.0007). HCGβ was a significant prognostic factor in Dukes' B cancer (p = 0.010) and nearly significant in Dukes' D cancer (p = 0.050).
In Cox multivariate analysis, the best model was achieved with entering the tumour marker values as continuous variables after logarithmic transformation. The strongest prognostic factor was found to be Dukes' stage, followed by rectal tumour location. Other independent prognostic factors were preoperative serum markers CEA, CA 72-4 and hCGβ (Table III). When only curatively treated patients were included (Dukes' A, B and C cancers), the same independent prognostic factors emerged, the strongest being CEA, followed by rectal tumour location, hCGβ, stage and CA 72-4, in the order of importance. Other factors were insignificant (data not shown).
|Rectal tumour location||0.002||1.98||1.28–3.05|
|Ln (CA 19-9)||NS|
Our results confirm that CEA is the most sensitive tumour marker for colorectal cancer. The concentrations of markers were associated with stage, thus it was not surprising that all markers had prognostic value in univariate analysis. However, in Cox multivariate regression survival analysis, only markers hCGβ, CA 72-4 and CEA had independent prognostic significance; other prognostic factors were Dukes' stage and rectal tumour location.
Stage has been established to be the strongest prognostic factor in colorectal cancer.1 From other possible prognostic factors, tumour markers have also been extensively studied with the objective of finding additional tools to help in the patient selection for possible adjuvant therapies. Since it was first described in 1965,15 CEA has been the most used and best documented tumour marker for colorectal cancer. However, its prognostic value in multivariate analysis has been controversial. A recent study showed CEA to be an independent prognostic factor,3 whereas in other studies this has not been the case.6, 16, 17 Previous studies have suggested that CA 19-9 and CA 72-4 correlate with adverse survival,4, 5 and studies with multivariate survival analysis have indicated that CA 19-9, CA 242 and hCGβ provide independent prognostic information in colorectal cancer.6, 7, 18
In our previous studies on the prognostic value of CEA, CA 242, tissue polypeptide antigen (TPA), specific tissue polypeptide antigen (TPS) and hCGβ, we found CA 242, TPS and hCGβ to be independent prognostic factors.6, 11 In this study on another series of colorectal cancer patients, we set out to evaluate the prognostic value of 2 additional tumour markers, CA 19-9 and CA 72-4, and to compare them with CEA, CA 242 and hCGβ. Our results are in agreement with the previous data.3–6, 11, 18 In univariate analysis, all studied markers were significant prognostic factors; the survival was significantly better in patients who had serum marker values below the cut-off values. In multivariate survival analysis, independent prognostic factors were Dukes' stage, rectal tumour location and preoperative marker concentrations of hCGβ, CA 72-4 and CEA, with stage as the strongest prognostic factor. These markers may provide prognostic information also within Dukes' stages, but no definite conclusions can be drawn because the number of patients in each stage group was rather small. Contrary to results from our previous studies on the tissue expression of hCGβ in colorectal cancer,11 the serum levels of hCGβ were significantly higher in females than in males. The significance of this finding is uncertain, but it may be related to the small number of patients with elevated hCGβ serum level in our study.
We conclude that elevated preoperative serum levels of tumour markers CEA, CA 19-9, CA 242, CA 72-4 and hCGβ are related to poor outcome in patients with colorectal cancer. Dukes' stage is the strongest prognostic factor, but tumour markers CEA, CA 72-4 and hCGβ are also independent prognostic factors. We suggest that the measurement of preoperative serum tumour markers is beneficial in clinical practice. Further studies are warranted to establish the possible role of these markers in selection of patients for adjuvant therapies.
The skillful technical assistance of Ms. A. Ahmanheimo, Ms. T. Grönholm and Ms. E. Malkki is greatly acknowledged. The authors thank Wallac, Bayer and CanAg for kindly supplying the kits for the serum measurements.
- 1Digestive system tumors. In: I.U.A.C. UICC ed. Prognostic factors in cancer, 2nd ed. New York: Wiley-Liss, 2001. 235–347., , , , , .
- 6Independent prognostic value of preoperative serum markers CA 242, specific tissue polypeptide antigen and human chorionic gonadotrophin beta, but not of carcinoembryonic antigen or tissue polypeptide antigen in colorectal cancer. Br J Cancer 1996;74: 925–9., , , , , .