Brain metastasis development and poor survival associated with carcinoembryonic antigen (CEA) level in advanced non-small cell lung cancer: a prospective analysis

Between March 2005 and June 2007, patients with recent histological diagnosis of advanced NSCLC without previous treatment and referred to the Lung cancer and Thoracic Tumors Clinic (Instituto Nacional de Cancerología and Instituto Nacional de Enfermedades Respiratorias, Mexico City, Mexico) were enrolled in this prospective study. This work was approved by bioethical and research committees of each institution, and patient informed consent was obtained before enrollment. Histological diagnosis of primary NSCLC was established according to the revised classification of lung tumors of World Health Organization and the International Association for Lung Cancer Study. During pre-treatment clinical evaluation, we focused on physical examination, weigh loss, and performance status (according to the Eastern Cooperative Oncologic Group scale, ECOG). Thoracic, upper abdominal computed tomography (CT) and bone scintillography were used in disease staging. Only patients with IIIB (T4 N0-1-2 M0 or any T N3 M0: tumor of any size that invades mediastinum, heart, great vessels trachea, esophagus, vertebral body, carina; or tumor with a malignant pleural or pericardial effusion, or with satellite tumor nodule(s) within the ipsilateral primary-tumor lobe of the lung [T4], no regional lymph node metastasis [N0], metastasis to ipsilateral peribronchial and/or ipsilateral hiliar lymph nodes, and intrapulmonary lymph nodes involved by direct extension of the primary tumor [N1], metastasis to ipsilateral mediastinal and/or subcarinal lymph node(s) [N2], or metastasis to contralateral mediastinal, contralateral hiliar, ipsilateral or contralateral scalene, or supraclavicular lymph node(s) [N3], and no distant metastasis [M0]) and IV clinical stage (any T any N M1: distant metastasis present) were enrolled [31, 32]. Chemotherapy was based on platin or EGFR inhibitors. Six different schedules were instituted, including gemcitabine, vinorelbine, or paclitaxel combined with cisplatin or carboplatin and erlotinib.

Analyzed variables comprised smoking history, gender, general condition, histology, brain metastasis development, serum CEA levels, and EGFR and HER2 tissue expression.

CEA levels in serum were evaluated prior to chemotherapy. CEA was measured by solid-phase, two-site sequential chemiluminescent immunometric assay (IMMULITE 2000-CEA Analyzer, Siemens, Inc., Los Angeles, CA) with an analytical sensibility of 0.15 ng/mL and a high-dose Hook effect at levels of >250,000 ng/mL. Serum samples (15 μL serum) were collected using the same method for each patient and stored if necessary at -20°C before processing. Cut-off points for resuming tables were selected according to the previously reported normal value for CEA serum level (< 10 ng/mL) [10].

EGFR and HER2 tissue expression were evaluated by immunohistochemistry (IHC). IHC was performed on formalin-fixed, paraffin-embedded tissue sections from primary lung tumors. 5 μm sections were placed on chemically charged slides. Sections were deparaffinized and rehydrated in a series of alcohols and xylene according to established procedures. Following deparaffination, sections were immersed in an antigen-retrieval solution (Dako Corporation) for 40 min at 95°C. Endogen peroxidase was blocked with 3% H₂O₂ in absolute methanol for 5 min. Slides were incubated with rabbit antibody human HER2 (Hercep Test, Dako Corporation) for 30 min, and staining was completed using Dako Rabbit Evision Plus Kit (Dako Corporation). The antibody binding site was visualized using diaminobenzidine reagent for 5 min. The slides were counterstained with Mayer's hematoxylin. Immunostainig for EGFR was accomplished using the monoclonal mouse antibody to EGFR (1:100, Zymed Laboratories, San Francisco, CA). The slides were incubated with peroxide block for 10 min, this followed by incubation for 10 min with proteinase K. After incubation of slides with primary antibody for 30 min, staining was completed by incubation with monoclonal-labeled polymer for 10 min. The antibody binding site was visualized using diaminobenzidine reagent. Finally, the slides were counterstained with Mayer's hematoxylin.

EGFR and HER2 were examined using light microscopy. For EGFR, we considered negative tissue expression if the stain was heterogeneous in ≤ 25% of the sample, and positive tissue expression if stain was homogeneous >25% of sample. For semi-quantitative evaluation of HER2, all slides were scored following the guidelines for scoring HercepTest: 0, completely negative or membrane staining in fewer than 10% of tumor cells; 1+, faint membranous staining in >10% of the tumor cells; 2+, weak or moderate complete staining in >10% of tumor cells, and 3+, strong complete membrane staining in >10% of the tumor cells. IHC was graded by a single pathologist (AAS), who was blinded to clinical characteristics and outcomes.

In presence of neurological symptoms (persistent headache, neurological focalization, motor deficits, or abnormal behaviour), brain magnetic resonance imaging (MRI) was performed. Monitoring of disease (including primary endpoints, e.g., OS and brain metastasis development) was carried out by means of clinical follow-up.

With a descriptive purpose, we resumed each continuous variable as arithmetic mean, median and standard deviation (SD), and categorical variables as proportion with 95% confidence interval (95% CI). For inferential comparisons, we employed Student t or Mann-Whitney U test, according to data distribution (normal or non-normal, determined with Kolmogorov-Smirnov test). To calculate statistical significance between categorical variables, we used chi-square or Fisher exact test. Statistically significant and borderline results (p < 0.1) were included in multivariate logistic regression analysis. CNS metastasis development and OS were defined as the period from date of histological diagnosis to date of confirmed diagnosis of brain metastasis by MRI and to date of death, respectively; both were analyzed with Kaplan-Meier method, and sub-groups were compared with log-rank test. To analyze survival curves, each variable was dichotomized. HER2 tissue expression was considered negative if IHC was classified as 0 or +1, and positive if classified as +2 and +3. For EGFR, tissue expression was negative or positive if IHC was positive in <25% or ≥ 25% of tumor cells, respectively. Statistically significant or borderline (p < 0.1) variables on univariate analysis were included in multivariate analysis utilizing Cox proportional hazards model. Statistical significance was determined with a p < 0.05 in a two-sided test. SPSS software package (version 14.0; SPSS, Inc., Chicago, IL) was employed for data analysis.

Two hundred ninety three consecutive patients were prospectively studied. Patient characteristics are shown in Table 1. Mean for age was 60.7 ± 0.7 years. Only 54% of patients had smoking history. The majority of tumors were adenocarcinoma (65%) classified as moderate or high histological grade (36 and 54%, respectively). Seventy one percent of patients were diagnosed in a metastatic stage, and 29% were in IIIB clinical stage. Of patients with metastasis at time of diagnosis, 18.6% had CNS metastasis and 7.5%, liver metastasis. Most patients were referred to our Institute without paraffin blocks and in others the diagnosis was based on cytopathological samples; even, when tissue samples were available, some of them were insufficient to perform the immunohistochemistry. Because of these reasons, we only analyzed 85 biopsies of primary tumor to determine tissue expression of EGFRs types 1 and 2, of which, 59% were positive for EGFR and only 7% for HER2 tissue expression.

In 42.8, 32.3, 22.2, and 21.4% of patients, basal serum CEA level was ≥ 10, 20, 40, and 50 ng/mL at diagnosis, respectively (median ± standard deviation, 6.3 ± 1,021 ng/mL; range, 0.2–15,475 ng/mL). Age, gender, positive smoking history, status performance, histological-grade of differentiation and presence of liver metastasis at diagnosis were not associated to basal CEA serum levels ≥ 40 ng/mL. However, at the bi- and multivariate analysis, factors associated with basal CEA serum levels ≥ 40 ng/mL were adenocarcinoma histological type (frequency of 41.3% compared with 14.1% of patients with squamous or large-cell histologies, RR 1.6; 95% CI, 1.4–3.4; p = 0.005) and presence of CNS metastasis at diagnosis (frequency of 77.7% compared with 12.3% of patients without CNS metastasis at diagnosis, RR 14.05; 95% CI, 5.7–34.4; p < 0.001).

At 1 year of diagnosis, 27% (95% CI, 23–30%) of patients developed brain metastasis, and at 2 years, 32% (95% CI, 21–43%). As treatment for brain metastasis, only 4 patients achieved criteria to undergo radiosurgery, none of the patients was submitted to brain surgery, and all other patients received whole brain irradiation. Independent associated factors comprised adenocarcinoma histological type (RR 5.2; 95% CI, 1.002–29; p = 0.0002) and CEA serum levels ≥ 40 ng/mL (RR 11.4; 95% CI, 1.7–74; p < 0.001) (Table 2). In the subgroup analysis of patients with adenocarcinoma histologycal type, frequency ± standard error of CNS metastasis development at 12 months of diagnosis were 16.4 ± 0.03% (95% CI, 16.34 – 16.46%) and 67 ± 0.09% (95% CI, 66.91 – 67.09), in patients with CEA serum levels < 40 ng/mL, and ≥ 40 ng/mL, respectively. And at 24 months of diagnosis, frequency ± standard error of CNS metastasis development were 20.2 ± 0.05% (95% CI, 20.19 – 20.21%) and 67 ± 0.09% (95% CI, 66.82 – 67.18%) in patients with CEA serum levels < 40 ng/mL and ≥ 40 ng/mL, respectively. EGFR tissue expression was not statistically significant in terms of CNS metastasis development.

OS was 7 ± 0.48 months. Factors associated with statistical significance at univariate analysis with poor OS were male gender (p = 0.02), age ≥ 60 years (p = 0.028), poor performance status (ECOG III, p < 0.001), serum CEA levels ≥ 40 ng/mL (p = 0.002), and EGFR-positive tissue expression (p = 0.023). Clinical stage IV demonstrated solely a tendency toward significance (p = 0.056) (Table 3). On multivariate analysis, male gender (RR = 1.4; 95% CI, 1.002–1.9; p = 0.048) [Fig 1a], poor performance status (RR 1.8; 95% CI, 1.5–2.3; p = 0.002) [Fig 1b], clinical stage IV (RR 1.4; 95% CI, 1.02–2; p = 0.04) [Fig 1c], serum CEA levels ≥ 40 mg/mL (RR 1.5; 95% CI, 1.09–2.2; p = 0.014) [Fig 1d], and EGFR tissue expression (RR 1.6; 95% CI, 1.4–19; p = 0.012) [Fig 1e] were statistically significant.