The prognostic value of serum CA 19-9 for patients with advanced lung adenocarcinoma

Background This study aimed to assess the prognostic accuracy of serum CA 19-9 in patients with advanced lung adenocarcinoma. Methods We retrospectively reviewed data of 246 patients who were diagnosed at our institute with advanced (stage IIIB or IV) lung adenocarcinoma between March 2006 and December 2012. We excluded patients who received no chemotherapy, or for whom we had no data on pre-treatment tumor markers. We also evaluated 116 consecutive resected specimens from patients with clinical stage I lung adenocarcinoma pathologically. Results The 76 (31 %) patients who were CA 19-9+ had shorter overall survival (OS) than CA 19-9− group (12.5 vs 26.2 months, P = 0.005). Cox’s multivariate regression analysis identified Eastern Cooperative Oncology Group Performance Status 0 or 1 (P < 0.001), mutated epidermal growth factor receptor (EGFR) status (P < 0.001), stage IIIB (P < 0.001), CYFRA 21-1− (P < 0.001), CA 19-9− (P = 0.005) and use of platinum doublet therapy (P = 0.034) as independent predictors of longer OS. We stratified patients by CA 19-9 and CYFRA 21-1 as double positive (CA 19-9+/CYFRA 21-1+, n = 59), single positive (either CA19-9+ or CYFRA 21-1+, n = 113), or double negative (CA 19-9−/CYFRA 21-1−, n = 74). Their respective OS were 10.0, 23.3 and 31.8 months (P < 0.001). Pathological analysis also correlated CA 19-9 expression with malignant features such as vessel invasion, pleural invasion, cancer invasive factors and mucin production. Conclusions CA 19-9 and CYFRA 21-1 are independent prognostic markers in patients with advanced lung adenocarcinoma. Combined use of CA 19-9 and CYFRA 21-1 provides further prognostic information in patients with advanced lung adenocarcinoma. Electronic supplementary material The online version of this article (doi:10.1186/s12885-016-2897-6) contains supplementary material, which is available to authorized users.


Background
Lung cancer is the leading cause of cancer death worldwide. Unfortunately, most lung cancers are already unresectable and metastatic at initial diagnosis [1,2]. Overall survival (OS) of patients with advanced lung adenocarcinoma (ALAD) is still very poor, despite progress in treatment and chemotherapy. ALAD prognosis can be assessed through various factors, such as pathologic characteristics, imaging features, and oncogenes, but identification of more accurate prognostic markers is imperative.
Measurement of tumor markers is a non-invasive means to predict prognosis, and is therefore used in daily clinical practice [3]. Earlier investigations of the relationships between prognosis and serum cytokeratin 19 fragments (CYFRA 21-1), carcinoembryonic antigen (CEA) or neuron-specific enolase (NSE) in ALAD patients found only CYFRA 21-1 to be an independent prognostic marker among them [4][5][6]. Therefore, identification of another independent prognostic tumor marker would have great value in managing these patients.
Although patients with ALAD who show extremely high serum levels of CA 19-9 are reportedly have poor prognoses, the relationship between serum CA 19-9 and prognosis in lung adenocarcinoma has not been studied yet [16,17]. We hypothesized that CA 19-9 is a prognostic marker for ALAD patients.
The purpose of this study was to investigate the clinical utility of CA 19-9 as a prognostic marker in ALAD patients, and to improve its prognostic accuracy.

Study subjects
We retrospectively analyzed 433 patients diagnosed advanced (stage IIIB or IV) lung adenocarcinoma at Kobe City Medical Center General Hospital between March 2006 and December 2012. We excluded patients who received no chemotherapy (n = 71), or for whom no data on tumor markers (CEA, CYFRA 21-1 and CA 19-9) before receiving chemotherapy was available (n = 116). Patients who reported never having smoked were defined as never-smokers, those who had smoked within 1 year of the diagnosis were categorized as current smokers, and the rest were considered to be former smokers. All patients were classified by clinical stage according to the 7 th edition TNM (tumor, node, metastasis) classification [18]. OS was measured from the diagnosis of lung cancer until death from any cause or the end of the follow-up. We isolated tumor DNA from various specimens and analyzed epidermal growth factor receptor gene (EGFR) mutation status at exons 18-21, using the peptide nucleic acid-locked nucleic acid PCR clamp methods, as described previously [19]. We retrospectively analyzed the presence of intestinal pneumonia, non-tuberculous mycobacteriosis (NTM) infection, bronchiectasis, and diffuse panbronchiolitis by reviewing patients' charts and radiological records.

Determination of tumor markers concentration
Serum samples were obtained to determine tumor markers CEA, CYFRA 21-1 and CA 19-9 as a part of routine evaluations within 28 days before starting chemotherapy. The concentration of each tumor makers was measured using LumiPulse Presto kit (Fujirebio Inc., Tokyo, Japan). It utilizes the chemiluminescent enzyme immunoassay (CLEIA) principle and is a fully automated assay [20]. The CLEIA method uses 1.6 ng/mL as the upper limit of normal (ULN) serum CYFRA 21-1 level in healthy individuals [21]. For this study, we set the cutoff value for CYFRA 21-1 at 2.2 ng/mL, which was used in a previous study that showed the prognostic impact of CYFRA 21-1 in patients with ALAD, and is the mean value for healthy subjects +3 SD (standard deviation) [4]. The cutoff values for serum CEA and CA 19-9 were set at 5.0 ng/mL and 37.0 U/mL, which are their respective ULNs [22,23]. This testing was performed at the Department of Laboratory Medicine at our hospital.

Statistical analysis
Continuous variables were analyzed using Student's t-test. Dichotomous variables were analyzed using Fisher's exact test. Correlations between CA 19-9 levels and CYFRA 21-1 levels were assessed using Spearman's rank-based correlation test. The Kaplan − Meier method was used to estimate survival outcomes; groups were compared using the log-rank test. Cox's proportional hazard models were fitted to determine associations between patient characteristics and survival outcomes. A multivariate Cox proportional hazard model was developed on all clinically important factors (age, sex, smoking status, ECOG PS, EGFR status, stage, positivity of serum CEA, CYFRA 21-1 and CA 19-9, and platinum doublet therapy administration). The results are expressed as hazard ratios (HRs) with 95 % confidence intervals (CI). All tests were two-tailed. A value of P < 0.05 was considered to indicate significance. We conducted statistical analyses on JMP software (11 th version; SAS Institute, Cary, NC, USA).
Test of correlation between CA 19-9 levels and CYFRA 21-1 levels showed no significant relationship between these tumor markers (r = 0.006). The scatter plot is shown in (Additional file 2: Figure S1).

Survival analysis according to the tumor markers
The OS of patients included in this study was 21.4 months (interquartile range, 18.9 − 25.0 months; Table 2). Overall survival curves of CA 19-9 + and CA 19-9 − patients are shown in Fig. 3a. The OS of serum CA 19-9 + patients  OS by combined serum levels of CA 19-9 and CYFRA 21-1

Multivariate analysis of overall survival time
We initially divided the patients into 4 groups by serum CA 19-9 and CYFRA 21-1 level: (a) both CA 19-9 + and ). Although the MSTs of (a) and (b) (P = 0.001) and that of (c) and (d) (P = 0.003) differed statistically, the difference between the two single-positive groups, (b) and (c), was not significant (P = 0.14). Therefore, we combined groups (b) and (c) into one single positive group. We also defined group (a) as double positive and group (d) as double negative (Fig. 2). Survival curves for these 3 groups are shown in Fig. 3b.

Pathological findings and recurrence-free survival analysis
We pathologically evaluated 116 consecutive surgically resected specimens from patients with clinical stage I lung adenocarcinoma. Of those, 54 (47 %) were CA 19-9 + . Relationships between CA 19-9 expression and patients' clinicopathological characteristics are summarized in Table 3. Comparison of clinical profiles of CA 19-9 + and CA 19-9 − patients showed the CA 19-9 + group to include significantly higher proportions of patients with vessel invasion (P = 0.032), pleural invasion (P = 0.023), cancer invasive factors (P = 0.005) and positive PAS stain (P = 0.001). CA 19-9 + patients had significantly shorter recurrence-free survival than CA 19-9 − patients (P = 0.030). The Kaplan-Meier curve is shown in Fig. 4. We also investigated the association between EGFR status and survival in the 116 patients with stage I lung adenocarcinoma; for whom 110 (95 %) had EGFR mutation analyses available, which showed 61 (53 %) to have EGFR mutations. Log-lank analysis revealed that EGFR status had no prognostic effect on recurrence-free survival (P = 0.569), or OS (P = 0.171).

Discussion
In the present study, we showed that both serum CA 19-9 and CYFRA 21-1 were independent prognostic markers in ALAD patients, and their combined use improves prognostic accuracy. We have shown serum CA 19-9 to be an independent predictive factor for OS, according to multivariate analysis of possible prognostic factors that included serum CYFRA 21-1. To the best of our knowledge, this is the first report to show the correlation between positive CA 19-9 levels and shorter OS in patients with ALAD, although this correlation has been reported in adenocarcinomas of other organs such as pancreas, colon, and stomach [8][9][10][11][12][13][14][15]. The consistency of this pattern among adenocarcinomas of different organs implies that serum CA 19-9 might be a prognostic marker in all kinds of adenocarcinoma.
The major advantage of CA 19-9 is that it can be measured quickly at low cost. Additionally, CA 19-9 is a standard biomarker for gastrointestinal cancers, such as pancreatic, colon, and gastric cancers. Therefore we speculate that its application to lung cancer would be relatively easy.
Our results also showed that 31 % of ALAD patients had positive serum CA 19-9. We believe that this positive rate is common in patients with ALAD, although it was much lower than in studies of patients with advanced pancreatic adenocarcinoma (for which CA 19-9 is a prognostic marker), who were reportedly 50 % − 84 % positive [8,10,30].
The precise reason for high CA 19-9 levels is unclear. However, large studies have shown that healthy volunteers did not have high serum CA 19-9 levels [31,32]. High CA 19-9 elevation has been reported in some chronic inflammatory lung diseases, such as intestinal pneumonia, NTM infection, bronchiectasis, and diffuse panbronchiolitis [33]. In the present study, CA 19-9 positivity and presence of inflammatory disease showed no correlation (P = 0.147); thus inflammatory disease did not cause CA 19-9 elevation. In addition, our pathological analysis demonstrated that the lung cancer cells generated CA 19-9. Therefore, we speculated the elevated serum CA 19-9 was associated with the CA 19-9 generated by cancer cells.
Although CA 19-9 expression is also related to unfavorable prognosis in some kinds of cancer, why high CA 19-9 predict shorter OS is not understood [34,35]. We therefore investigated the relationship between CA 19-9 IHC positivity and pathological findings in patients with stage I lung adenocarcinoma. We analyzed the presence of cancer invasive factors (vessel and pleural invasion) and mucin production, which were reportedly associated with highly malignant features such as shorter recurrence-free survival and OS in lung adenocarcinoma patients [24,29,36]. In our pathological analysis of clinical stage I patients, CA 19-9 positive lung adenocarcinoma had more histologically malignant features (P = 0.005) and shorter recurrence-free survival (P = 0.030) than CA 19-9 negative lung adenocarcinoma (Fig. 4). These findings indicate that CA 19-9 positive lung adenocarcinoma is highly malignant. We speculate that these malignant features caused the elevated serum CA 19-9, as cancer cell invasion to the blood could cause the elevated serum tumor markers. Our findings are of special interest, but there were some limitations to this study. First, as this study was conducted in a single institute, it included limited number of patients. Second, a considerable number of patients were excluded from this analysis due to missing tumor marker data before their initial therapy, because measurement of tumor markers was at the discretion of the attending physician. Third, subject selection in this study was confined to Japanese patients, and racial differences may need to be considered in the interpretation of this study. Fourth, we did not examine Lewis antigen status [37]. Patients who are Lewis antigen-negative cannot synthesize CA 19-9, and therefore present as falsely negative [38]. However, as they comprise only 5-7 % of the general population, we assume that it did not affect the results [39].

Conclusions
In conclusion, our study showed serum CA 19-9 to be an independent prognostic indicator in patients with ALAD, and combined use of CA 19-9 and CYFRA 21-1 to provide more accurate prognostic information.