1 Introduction

Colorectal cancer (CRC) is one of the most common cancer-related causes of morbidity and mortality in developed countries [1, 2]. Targeted therapies based on monoclonal antibodies against vascular endothelial growth factor (VEGF) or epidermal growth factor receptor (EGFR) are widely used in the treatment of metastatic CRC (mCRC). Bevacizumab is a humanized monoclonal antibody that blocks angiogenesis by inhibition of the most important angiogenic growth factor, VEGF. So far, no reliable biochemical or molecular predictors of response to bevacizumab have been validated.

The measurement of serum tumor markers is a simple and non-invasive method for assessing the response to systemic therapies in mCRC and, in some cases, the prognosis [3, 4]. Molecules used as serum tumor markers play a role in several cancer-related processes including cell adhesion, proliferation, and tumor angiogenesis and hypothetically could be useful as a readily available surrogate marker for molecular characteristics of a tumor [5–11].

The aim of our retrospective study was to evaluate the association of baseline serum levels of carcinoembryonic antigen (CEA), carbohydrate antigen 19-9 (CA 19-9), thymidine kinase (TK), and tissue polypeptide specific antigen (TPS) with outcomes of patients with mCRC treated with bevacizumab.

2 Patients and Methods

Patients and Treatment

We retrospectively analyzed clinical data of 152 adult patients with histologically confirmed mCRC treated with bevacizumab-based therapy between years 2005 and 2014 at Department of Oncology and Radiotherapy, Medical School and Teaching Hospital Pilsen, Czech Republic. Bevacizumab (Avastin, F. Hoffman-La Roche Ltd., Basel, Switzerland) was administered in combination with chemotherapy or as a single agent in a standard approved doses (5.0 mg/kg every 14 days or 7.5 mg every 21 days). The chemotherapy consisted of the following schedules: fluorouracil and leucovorin in combination with oxaliplatin (FOLFOX) or irinotecan (FOLFIRI), or alone (FUFA); capecitabine alone; oxaliplatin alone; and irinotecan alone. None of the patients had previously received antiangiogenic therapies.

Data Source

Data were obtained from the clinical registry CORECT. The clinical registry CORECT (http://corect.registry.cz/) is a non-interventional post-registration population-based database of epidemiological and clinical data of patients with mCRC treated with targeted therapies in the Czech Republic. The registry contains anonymized individual patient data including demographic parameters, initial staging and disease characteristics, baseline patient information at the start of targeted therapy, as well as data on survival and adverse events. The entries are updated at least twice a year.

The protocol of our study was approved by the independent ethics committee of the University Hospital Pilsen and complied with the International Ethical Guidelines for Biomedical Research Involving Human Subjects, Good Clinical Practice guidelines, the Declaration of Helsinki, and local laws. Data from patients treated at the Medical School and Teaching Hospital Pilsen were extracted from the CORECT registry for the purposes of our study. Clinical data from the registry were validated against hospital medical records. Data on serum tumor markers were extracted from the hospital information system and merged to the registry data. The data have been analyzed retrospectively.

Clinical Monitoring of Data in the CORECT Registry and Statistics

The treatment was prospectively monitored, and the clinical course of patients was continuously assessed at pre-specified time points. Physical examination and routine laboratory tests were performed every 2 weeks; computed tomography (CT) or positron emission tomography—(PET)-CT was performed every 3 months of the treatment. The objective tumor response was assessed by the attending physician using Response Evaluation Criteria in Solid Tumors (RECIST) [12].

Tumor Marker Measurement

Serum samples were collected, and the measurement was performed within 1 month before the initiation of bevacizumab treatment. Serum levels of CEA and CA 19-9 were measured using chemiluminescent method on a DxI 800 analyzer (BeckmanCoulter, Brea, CA, USA). Serum levels of TK were measured using radioenzymatic assay (REA) on an Stratec 300 analyzer (Immunotech, Czech Republic). Serum levels of TPS were measured using immunoradiometric assay (IRMA) on a Stratec 300 analyzer (IDL Biotech, Sweden). The measurements were performed in the Central Immunoanalytic Laboratory at the Department of Nuclear Medicine, Pilsen University Hospital, using the following cutoff values: CEA 3 μg/l; CA 19-9 28 μg/l; TK 8 U/l, and TPS 90 μg/l. These are the upper normal values for the tumor markers measured by the used tests.

Statistical Analysis

Standard frequency tables and descriptive statistics were used to characterize sample data set. Progression-free survival (PFS) and overall survival (OS) were estimated using the Kaplan-Meier method, and all point estimates were accompanied by 95 % confidence intervals. PFS was determined from the date of bevacizumab initiation until the date of first documented progression or death. OS was determined from the date of bevacizumab initiation until the date of death. Statistical significance of the differences in Kaplan-Meier estimates according to tumor marker levels was assessed using the log-rank test. Based on the results from univariate Cox proportional hazards models, multivariable Cox regression model was used to adjust results in terms of number of metastatic sites and line of therapy as the most clinical relevant potential confounders. As a level of statistical significance, alpha = 0.05 was used.

3 Results

Patient Characteristics

The study included 152 patients. The median age was 61.1 years (range 32.6–83.1 years). One hundred and four (68.4 %) patients were male, 86 (56.6 %) had a primary tumor localized in the colon, 101 (66.4 %) had metastatic disease at diagnosis, and 131 (86.2 %) received the bevacizumab-containing regimen in the first line. Bevacizumab was combined with FOLFOX in 118 (77.6 %) patients, FOLFIRI in 16 (10.5 %) patients, FUFA in 6 (3.9 %), capecitabine in 5 (3.3 %) patients, irinotecan in 3 (2.0 %) patients, oxaliplatin in 2 (1.3 %) patients, and 2 (1.3 %) patients received bevacizumab in monotherapy. The baseline patient characteristics are summarized in Table 1. The median follow-up of the cohort was 18.9 months, the median OS was 35.5 months (95 % CI 24.1–46.9), and the median PFS was 11.6 months (95 % CI 9.2–14.1).

Table 1 Baseline patient characteristics
Full size table

Relation Between Baseline Levels of Serum Tumor Markers and Survival

Baseline serum levels of CEA, CA 19-9, TK, and TPS are shown in Table 2. The median PFS and OS for patients with high CEA was 9.7 and 35.1 compared to 21.0 and 56.0 months for patients with low CEA (p < 0.001 and p = 0.107) (Fig. 1a, b). The median PFS and OS for patients with high CA 19-9 was 9.7 and 29.0 compared to 13.5 and 56.0 months for patients with low CA 19-9 (p = 0.034 and p = 0.003) (Fig. 1a, b). The median PFS and OS for patients with high TK was 9.9 and 33.8 compared to 12.9 and 45.5 months for patients with low TK (p = 0.735 and p = 0.179). The median PFS and OS for patients with high TPS was 10.1 and 35.1 compared to 13.5 and 42.5 months for patients with low TPS (p = 0.797 and p = 0.563). The PFS and OS data are summarized in Table 2, and survival curves for CEA and CA 19-9 are shown in Fig. 1.

Table 2 Progression-free survival and overall survival according to baseline levels of serum tumor markers
Full size table
Fig. 1
figure 1

Kaplan-Meier estimates of progression-free survival (PFS) and overall survival (OS) according to baseline levels of carcinoembryonic antigen (CEA) (a, c) and carbohydrate antigen 19-9 (CA 19-9) (b, d)

Full size image

Baseline clinical parameters were assessed together with serum tumor marker levels in univariate and multivariate models. The univariate Cox proportional hazards model revealed that the number of metastatic sites (HR = 1.75, p = 0.003), lines of therapy (HR = 1.80, p = 0.020), CEA (HR = 2.70, p < 0.001), and CA 19-9 (HR = 1.49, p = 0.035) were significantly associated with PFS, whereas CA 19-9 (HR = 2.33, p = 0.004) was significantly associated with OS (Table 3). The multivariate Cox proportional hazards model revealed that the number of metastatic sites (HR = 2.00, p = 0.001), and CEA (HR = 2.80, p < 0.001) were significantly associated with PFS, whereas CA 19-9 (HR = 2.25, p = 0.008) was the only studied parameter associated with OS (Table 4).

Table 3 Hazard ratios from univariate Cox proportional hazard model for progression-free survival and overall survival
Full size table
Table 4 Hazard ratios from multivariable Cox proportional hazard model for progression-free survival and overall survival (n = 140 [only patients in which all characteristics are known are included in the model])
Full size table

4 Discussion

Randomized phase III clinical trials as well as observational studies have provided evidence for the efficacy and safety of bevacizumab in the treatment of patients with mCRC [13–17]. Despite the rapid developments in the field of predictive oncology in recent years, there is still no available biomarker predicting treatment efficacy of bevacizumab-based therapy. Several candidate predictive biomarkers including angiopoietin-2, circulating levels of short VEGF-A isoforms, soluble VEGFR-1, and intramural expression of VEGFR-2 or neuropilins have been studied but not sufficiently validated for routine clinical use [18–20]. CEA, CA 19-9, TK, and TPS are serum tumor markers used as a part of the diagnostic work-up and monitoring of patients with CRC. In the present study, we focused on their association with the efficacy of bevacizumab-based therapy in patients with mCRC.

CEA plays a role in cell-to-cell adhesion and has a dominant effect in blocking cell differentiation [5], and it cooperates with Myc and Bcl-2 in cellular transformation [6]. Additionally, several proangiogenic effects of CEA have been described recently. Bramswig et al. reported that soluble CEA activates endothelial cells and tumor angiogenesis via paracrine manner. They observed that endothelial cell adhesion, spreading, migration, and proliferation were enhanced in the presence of CEA, and the CEA-induced endothelial cell activation was independent of the VEGF-VEGFR 1/2 system [7]. This finding suggests that CEA bypass VEGF signaling and may result in resistance to VEGF-targeting drugs. However, the prognostic value of CEA in the setting of locoregional CRC or after metastasectomy has been previously clearly reported [21–25], it is controversial whether it is an independent prognostic parameter in metastatic CRC [26–28]. In our study, we observed a significantly shorter PFS (9.7 vs. 21.0 months; p  < 0.001) for patients with high pre-treatment CEA levels (>3 μg/l) compared to those with normal pre-treatment CEA levels (≤3 μg/l) although no association of CEA levels with OS was detected. The multivariate Cox proportional hazards model confirmed that high CEA was independently associated with shorter PFS (HR = 2.80; p < 0.001). Similar results have been recently published in a retrospective study by Prager et al., who reported significantly lower disease control rate (60.0 vs. 84 %; p = 0.0005) and shorter PFS (6.4 vs. 8.5 month; p = 0.023) with bevacizumab-based therapies in patients with high baseline CEA levels. Moreover, they also reported the lack of association between pre-treatment CEA levels and the treatment efficacy of cetuximab-based therapy [29]. In comparison with our study, they used different cutoff value (26.8 μg/l), which was obtained as a median value of pretreatment CEA levels. In order to facilitate the clinical interpretation of the results, we used the upper normal values for all the studied tumor markers as a cutoff.

CA 19-9 is an antigen expressed by the glycosylated extracellular MUC1 protein. CA 19-9 plays an important role in cancer invasion by enhancing cell adhesion and promoting angiogenesis indirectly [8]. In our study, we observed a significantly shorter PFS and OS for patients with high CA 19-9 (>28 μg/l), and the multivariate Cox proportional hazards model confirmed that high CA 19-9 was independently associated with shorter OS (HR = 2.25; p = 0.008) but not PFS (HR = 1.37; p = 0.111). The baseline serum level of CA19-9 has been previously reported as a significant prognostic indicator for mCRC (18), which is in agreement with our results. Interesting data on the predictive role of CA 19-9 have been provided by Formica et al. and Narita et al. [30, 31]. In observational studies, they found that only patients with high baseline levels of CA 19-9 benefited significantly from the administration of bevacizumab in comparison with chemotherapy alone [30, 31]. Narita et al. observed significantly longer OS for the group treated with bevacizumab and chemotherapy compared to those treated with chemotherapy alone among patients with high CA 19-9 levels (27.8 vs. 15.3 months; p = 0.0021), however, the OS was not different with or without bevacizumab among patients with normal CA 19-9 levels (36.5 vs. 38.0 months; p = 0.9515) [31]. Similarly to our study, they used normal values as a cutoff. All patients included in our study received treatment with bevacizumab; therefore, it is impossible to confirm or refute the results of these studies including also chemotherapy-only treated patients.

TK is an enzyme present in most cells, correlating with their proliferative characteristics. It has two isoforms, TK I and TK II that differ in chemical structure and biological function. TK I is expressed during cell division in the G1 and S phase while it is absent in resting cells [9]. The association between TK I expression and angiogenesis in NSCLC has been recently reported by Brockenbrough et al. [32]. In our study, we did not demonstrate any significant association between PFS or OS and the baseline level of TK, although several authors have reported on the putative role of TK as a prognostic factor [33, 34].

The TPS assay detects the M3 epitope of cytokeratin 18 or of tissue polypeptide antigen. Cytokeratin 18 is an acid-type cytosolic protein expressed in simple epithelial cells and also by tumor cells [10, 11]. TPS has been shown to indicate tumor cell proliferative activity. TPS has been mostly studied as a biomarker for monitoring of treatment response to palliative chemotherapy in patients with gastrointestinal tumors including mCRC [35–37]. No impact of baseline level of TPS and survival has been detected in our study.

The principal limitations of the present study are its retrospective design and relatively small number of patients with resulting heterogeneity, especially regarding backbone chemotherapy regimens. Nevertheless, it is the largest study published so far to use a comprehensive tumor marker panel in patients with mCRC treated with bevacizumab.

In conclusion, the results of the conducted retrospective study suggest that the baseline level of CEA was independently associated with PFS in bevacizumab-based therapy. CA 19-9 was independently associated with OS in these patients. Both CEA and CA 19-9 are commonly used serum tumor markers which are simple and easy to detect, and thus, they are feasible for the use in the routine clinical practice. We have not demonstrated association of baseline levels of TK and TPS with patients’outcome. The differential association of CEA and CA19-9 with PFS and OS suggests that their levels reflect intrinsic molecular properties of the tumors rather than solely tumor volume. Prospective studies on the predictive role of serum tumor markers should be performed to confirm these results.