Combination of squamous cell carcinoma-antigen, carcinoembryonic antigen, and carbohydrate antigen 19-9 predicts positive pelvic lymph nodes and parametrial involvement in early stage squamous cell carcinoma of the uterine cervix
Dr Shun-ichi Ikeda, Division of Gynecology, National Cancer Center Hospital, 5-1-1 Tsukiji, Chuo-ku, 104-0045 Tokyo, Japan. Email: firstname.lastname@example.org
Aim: We examined the correlations between the pretreatment values of four tumor markers (squamous cell carcinoma [SCC]-antigen, carcinoembryonic antigen [CEA], carbohydrate antigen [CA]19-9, and CA125) and postsurgical high-risk factors (parametrial involvement and positive pelvic lymph nodes) in women with SCC of the uterine cervix who had International Federation of Gynecology and Obstetrics clinical stage IB and IIA disease and underwent radical hysterectomy.
Material and Methods: In this retrospective study, we reviewed 291 patients between April 1989 and December 2008. The first 200 subjects, studied between 1989 and 2001, served as the training set, and another 91 subjects, studied between 2002 and 2008, comprised the test set. To evaluate the correlations between pretreatment tumor markers and postsurgical high-risk factors, the χ2-test and logistic regression analysis were used for univariate and multivariate analysis, respectively.
Results: Multivariate analysis with receiver–operator curves showed that the combination of SCC-antigen, CEA, and CA19-9 strongly predicted postsurgical high-risk factors. Analysis of the training set showed that 66.7% (95% confidence interval, 52.6–84.8%) of patients who tested positive for at least two of these three tumor markers had postsurgical high-risk factors. Similar results were obtained with the test set.
Conclusions: Preoperative levels of SCC-antigen, CEA, and CA19-9 are useful for predicting the status of postsurgical high-risk factors in women with SCC of the uterine cervix who undergo radical hysterectomy.
Various pathological prognostic factors have been proposed in women with early stage uterine cervical cancer who undergo radical hysterectomy. Parametrial involvement and positive pelvic lymph nodes are considered to indicate a particularly high postsurgical risk.1–3 The ability to more accurately predict postsurgical high-risk factors before therapy would facilitate prognosis in individual patients. Several circulating tumor markers have been proposed as predictors of outcomes.4–6 The usefulness of these markers remains uncertain; however, squamous cell carcinoma (SCC)-ag levels are widely used to monitor recurrence in women with SCC of the uterine cervix.7 We studied the correlation between the pretreatment status of four tumor markers and the postsurgical status of high-risk factors in women with uterine cervical SCC who underwent radical hysterectomy.
From April 1989 through December 2008, 301 patients with International Federation of Gynecology and Obstetrics (FIGO) stage IB or IIA SCC of the uterine cervix (245 with FIGO stage IB disease and 46 with stage IIA disease) underwent radical hysterectomy with pelvic lymphadenectomy (Class III of the Piver–Rutledge classification) at the Division of Gynecology, National Cancer Center Hospital, Tokyo, Japan. Systemic lymphadenectomy, including inguinal, external iliac, internal iliac obturator and common iliac lymph nodes, was performed and all lymph nodes were submitted for pathological examination. Ten (3.0%) of these patients were excluded from analysis because three or fewer tumor markers were measured, and the remaining 291 patients were retrospectively evaluated. These patients were divided into a training set (200 patients; 162 with FIGO stage IB disease and 38 with stage IIA disease), studied from April 1989 through December 2001, and a test set (91 patients; 83 with FIGO stage IB disease and eight with stage IIA disease), studied from January 2002 through December 2008. In accordance with the treatment policy of our hospital, patients with histologically proven positive pelvic lymph nodes or parametrial involvement additionally received postoperative adjuvant radiotherapy. All tumors were staged according to the FIGO clinical staging system (1994) for cervical cancer, based on the results of physical examination, chest radiography, drip infusion pyelography, and cystoscopy. To examine the influence of tumor markers on postsurgical high-risk factors, serum levels of SCC-antigen (SCC-ag), carcinoembryonic antigen (CEA), carbohydrate antigen (CA)19-9, and CA125 were measured in the 200 patients in the training set.
Postsurgical risk was classified as high in patients who had parametrial involvement, positive pelvic lymph nodes, or both, and low in patients who had none of these prognostic factors.
All serum samples were routinely analyzed before operation by the central laboratory in our hospital. In patients who underwent conization, peripheral blood samples were also obtained before the procedure. CEA, CA125, CA19-9, and SCC-ag were measured by enzyme immunoassay. The cut-off levels recommended by the manufacturers of the test kits were used, that is, 5 ng/mL for CEA, 35 U/mL for CA125, 37 U/mL for CA19-9, and 1.5 ng/mL for SCC-ag.
Tissue sections stained with hematoxylin–eosin were examined to establish the histopathological diagnosis. Parametrial invasion was defined as tumor in parametrial tissues, including direct invasion and lymphovascular spread. Lymph node metastasis in the parametrium was included in positive pelvic lymph nodes in the present study. Lymphovascular involvement was defined as the presence of tumor cells in the luminal space lined by flattened endothelial cells. The depth of cervical stromal invasion was classified as ≤two-thirds or >two-thirds.
To assess the correlation between tumor markers and postsurgical risk, the χ2-test and logistic regression analysis were used for univariate and multivariate analyses, respectively. Significant factors on univariate analysis were included in multivariate analysis. To evaluate the predictive value of tumor markers, the areas under receiver–operator curves (AUC) were estimated using a logistic regression model, in which tumor markers were analyzed as variables according to the cut-off levels recommended by the manufacturers. The combination of tumor markers with highest AUC was retested in the test set.
In the training set, 5-year survival rates and 95% confidence intervals (CI) were calculated by the Kaplan–Meier method. Deaths from any other causes and loss to follow-up were treated as censored data for survival analysis. The log–rank test was used to detect the differences in survival curves among different subgroups.
All statistical analyses were performed using sas, version 9.1.3. All P-values were two-sided, and P-values of less than 0.05 were considered to indicate statistical significance.
Table 1 shows the patients' characteristics in the training set and the test set. In the training set, 53 (26.5%) patients had postsurgical high-risk factors: five (2.5%) had only parametrial invasion, 31 (15.5%) had only positive pelvic lymph nodes, and 17 (8.5%) had parametrial invasion plus positive pelvic lymph nodes. In the test set, 25 (27.5%) patients had postsurgical high-risk factors: seven (7.7%) had only parametrial invasion, 10 (11.0%) had only positive pelvic lymph nodes, and eight (8.8%) had parametrial invasion plus positive pelvic lymph nodes.
Table 1. Characteristics of patients in the training set and test set
|FIGO stage|| || || || |
|Age|| || || || |
|Pathological stage|| || || || |
| pT|| 1b||132||66||62||68|
| || 2a||46||23||14||15|
| || 2b||22||11||15||17|
|Tumor diameter|| || || || |
| ≤4 cm||151||75.5||79||86.8|
| >4 cm||49||24.5||12||13.2|
|Cervical stromal invasion|| || || || |
|Lymphovascular space involvement|| || || || |
|Postsurgical high-risk factors||53||26.5||25||27.5|
| Only parametrial invasion||5||2.5||7||7.7|
| Only positive pelvic lymph nodes||31||15.5||10||11|
| Parametrial invasion and positive pelvic lymph nodes||17||8.5||8||8.8|
Univariate and multivariate analyses of the training set showed that the four tumor markers (SCC-ag, CEA, CA19-9, and CA125) significantly correlated with postsurgical high-risk factors (Tables 2, 3). Four tumor markers (SCC, CEA, CA19-9, and CA125) were analyzed on multivariate analysis in Table 3.
Table 2. Preoperative tumor markers according to high-risk-factor status in the training set
|SCC (ng/mL)|| || || || || ||<0.001|
| ≤1.5||109||91||(83.5)||18||(16.5)|| |
| >1.5||91||55||(60.4)||36||(39.6)|| |
|CEA (ng/mL)|| || || || || ||<0.001|
| ≤5.0||168||133||(79.2)||35||(20.8)|| |
| >5.0||32||13||(40.6)||19||(59.4)|| |
|CA19-9 (U/mL)|| || || || || ||<0.001|
| ≤37||175||136||(77.7)||39||(22.3)|| |
| >37||25||10||(40.0)||15||(60.0)|| |
|CA125 (U/mL)|| || || || || ||0.02|
| ≤35||182||137||(75.3)||45||(24.7)|| |
| >35||18||9||(50.0)||9||(50.0)|| |
Table 3. Multivariate logistic regression analysis of preoperative predictors of high-risk factors in the training set
|SCC||≤1.5 vs >1.5||2.258||1.101–4.631||0.03|
|CEA||≤5.0 vs >5.0||3.512||1.465–8.417||0.005|
|CA19-9||≤37 vs >37||3.437||1.315–8.987||0.01|
|CA125||≤35 vs >35||3.077||1.053–8.997||0.04|
Table 4 shows the AUC values of preoperative SCC-ag, CEA, CA19-9, and CA125 for predicting postsurgical high-risk factors in the training set. The combination of SCC-ag, CEA, and CA19-9 was related to postsurgical high-risk factors as strongly as was the combination of SCC-ag, CEA, CA19-9, and CA125 (Table 4). The estimated logistic regression model on ROC analysis was as follows: log odds = −5.15 + 0.80*I(SCC-ag ≥ 1.5 ng/mL) + 1.27*I(CEA ≥ 5.0 ng/mL) + 1.24*I(CA19-9 ≥ 37 U/mL), which gives the predicted probability of postsurgical high risk, where I(A ≥ a) = 1 if A ≥ a and I(A ≥ a) = 0 if A < a. We used c-statistics to estimate the AUC of the ROC.
Table 4. AUC of different combinations of tumor markers for predicting postsurgical high-risk factors in the training set
| CEA, CA19-9||0.69|
| SCC, CA19-9||0.68|
| SCC, CEA||0.70|
| CEA, CA19-9, CA125||0.70|
| SCC, CA19-9, CA125||0.69|
| SCC, CEA, CA125||0.70|
| SCC, CEA, CA19-9||0.73|
| SCC, CEA, CA19-9, CA125||0.73|
Postsurgical high-risk factors were present in 22 (66.7%) of 33 patients with at least two positive tumor markers, as compared with 19 (26.4%) of 72 patients with only one positive tumor marker and 13 (13.7%) of 95 patients in whom all three tumor markers were negative (Table 5). The value of these three markers for the prediction of postsurgical high-risk factors was validated in 92 patients with stage IB and IIA disease who underwent radical hysterectomy from January 2002 through December 2008 at the National Cancer Center Hospital, Tokyo, Japan. In the test set, postsurgical high-risk factors were present in 11 (68.8%) of 16 patients with at least two positive tumor markers, as compared with 11 (37.9%) of 29 patients with one positive tumor marker and three (6.4%) of 47 patients in whom all three tumor markers were negative (Table 6). The results obtained with the test set were similar to those obtained with the training set.
Positive postoperative high-risk factors according to the combination of SCC, CEA, and CA19-9 in the training set
Validation of positive postoperative high-risk factors according to the combination of SCC, CEA, and CA19-9 in the test set
In the training set, the tumor markers SCC-ag, CEA, and CA19-9 were found to be significantly related to the 5-year survival rate, that is, patients in whom two or three markers were positive had poorer survival (93.6% in patients in whom all three markers were negative, 91.7% in those with one positive marker, and 60.6% in those with two or three positive markers, P < 0.001, log–rank test).
Parametrial involvement and positive pelvic lymph nodes have been established to be important prognostic factors in early stage uterine cervical cancer.1–3 However, these prognostic factors cannot be accurately evaluated preoperatively, even on exhaustive studies with computed tomography, magnetic resonance imaging, and positron emission tomography. In particular, microscopic parametrial metastases, positive pelvic lymph nodes, or both cannot be detected on currently available imaging techniques.
Several studies have shown by multivariate analysis that elevated preoperative SCC-ag levels are associated with lymph node metastasis among high-risk factors. Lin et al. found on multivariate analysis that only lymph node metastasis was significantly related to SCC-Ag levels exceeding 8 ng/mL.8 Takeshima et al. also reported that SCC-Ag levels greater than 4 ng/mL were associated with nodal metastasis.9 Only one study performed by Takeda et al. examined the usefulness of pretreatment levels of multiple tumor markers for predicting the status of high-risk factors. A combination of SCC-ag and CA125 levels was shown to be related to lymph node status.10 However, these studies did not assess parametrial involvement.
Reesink-Peters et al. investigated the correlation between preoperative SCC-ag levels and postsurgical high-risk factors in early stage SCC of the uterine cervix. They proposed that an SCC-ag level of >1.9 ng/mL could be used to identify patients who required adjuvant radiotherapy because postsurgical high-risk factors were present in 73 (66.4%) of 110 patients with stage IB and IIA disease who had elevated SCC-ag levels;11 however, their study excluded from analysis 184 patients who underwent previous loop excision or conization. The exclusion of these patients might have resulted in the high proportion of patients with postsurgical high-risk factors.
The present study was designed to identify prognostic variables related to postsurgical high-risk factors in early stage SCC of the uterine cervix. We performed multivariate analysis of ROC curves to assess the value of various combinations of four tumor markers for the prediction of postsurgical high-risk factors. Our results showed that positivity for at least two of three markers (i.e., elevated preoperative levels of SCC-ag, CEA, and CA19-9 in serum) is strongly related to high-risk factors in women with SCC of the uterine cervix.
One important question is why the combination of tumor markers correlates with high-risk factors. Unfortunately, the answer to this question is beyond the scope of the present investigation. The biologic functions of tumor markers are largely unknown and must be elucidated to answer this question in the future.
Using a combination of SCC-antigen, CEA, and CA19-9 as determined by the training set, we confirmed that 11 (68.8%; 95%CI, 46.0%–91.5%) patients with at least two positive tumor markers had postoperative high-risk factors in the test set. A combination of at least two positive tumor markers had a sensitivity of 40% and a specificity of 92.5% in the training set. This combination of tumor markers cannot be used for screening because of its low sensitivity and high specificity, but confirms our results. Physicians want to know the probability of high-risk factors in a patient with positive test results. One way of expressing this probability is the positive predictive value (PPV). PPV is influenced by the prevalence of disease in the population being tested. Few studies have assessed the incidences of pelvic lymph node metastasis, parametrial involvement, or both in patients undergoing radical hysterectomy for early stage uterine cervical cancer. The incidence of pelvic lymph node metastasis, however, is estimated to range between 20% and 30%.12,13 Because our combination of tumor markers had a sensitivity of 40% and a specificity of 92.5%, the PPV was estimated to be 57% given a 20% prevalence of high-risk factors and 70.6% given a 30% prevalence. On the basis of the literature,12,13 the proportion of patients with stage IB–II disease who have high-risk factors is estimated to be 20–30%. Historically in Japan, radical hysterectomy was aggressively performed to treat stage IB-II cervical cancer. Internationally, surgery and radiotherapy have been demonstrated to have similar effectiveness in stage IB–IIA cervical cancer. If a method was available to identify patients with stage IB–II disease who have high-risk factors with a probability of 60–70% before treatment, such a method might be used to decide whether patients should receive surgery or radiotherapy.
Varying the cut-off value of the tumor markers, we analyzed sensitivity and specificity as continuous levels using ROC, and the AUC for the combination of the three and four tumor markers were 0.8. On multivariate analysis of the ROC in the present study, it was difficult to determine the cut-off level for each tumor marker. Therefore, we used the cut-off levels recommended by the manufacturers of the test kits. These cut-off levels could identify patients with high-risk factors, and reproducibility was confirmed in the test set.
Our study also showed that survival was poorer in patients in whom two or three markers were positive. This finding is attributed to the close correlation between two or three positive markers and the presence of high-risk factors.
Our results suggest that the combination of three tumor markers (SCC-antigen, CEA, and CA19-9) might be useful for identifying subgroups of patients with early stage cervical cancer associated with different risk levels before treatment.
In conclusion, the combination of the preoperative serum levels of SCC-antigen, CEA, and CA19-9 is useful for predicting the status of postsurgical high-risk factors in women with SCC of the uterine cervix who undergo radical hysterectomy. Our study is limited by its retrospective nature; however, our findings will hopefully contribute to more accurate prediction of clinical outcomes before treatment.
The authors make no disclosures.