Prognostic Value of Combined Detection of Serum and Pleural Fluid Tumor Markers in Non-small Cell Lung Cancer With Malignant Pleural Effusion

Background: Serum tumor markers are considered a negative prognostic factor in non-small cell lung cancer (NSCLC). Combined detection of serum and pleural uid (PF) tumor markers is rarely reported. The aim of this study was to analyze the relationship between combined detection of serum and pleural uid (SPF) tumor markers and prognosis in advanced NSCLC. Results: High levels of SPF CEA (P = 0.001) , SPF CYFRA21-1 (P = 0.001) and SPF CA-125 (P = 0.023) were adversely prognostic factors for overall survival (OS). High levels of SPF CEA (P = 0.003) , SPF CYFRA21-1 (P = 0.001) and SPF NSE (P = 0.019) were related to worse progression-free survival (PFS). In multivariate analysis high levels of SPF CYFRA21-1 was a independent predictor of OS and high levels of SPF CEA and SPF CYFRA21-1 were independent predictors of PFS. Conclusions: High levels of SPF CEA and SPF CYFRA21-1 are correlated with worse survival in advanced NSCLC patients with MPE. The identication of prognostic factors maybe useful in stratifying high-risk populations and will assist the choice of treatment planning.

Background Non-small cell lung cancer (NSCLC) remains a leading cause of cancer-related death worldwide, representing about 80% of lung cancer [1,2]. Despite signi cant advances in public awareness and medical care, more than 50% of patients diagnosed with NSCLC have distant metastases and only 16% of cases are diagnosed at eariy stages [3]. Several prognostic factors of patients with NSCLC have been reported; however, many factors require invasive examination or surgery [4]. The measurement of serum or pleural uid (PF) tumor markers are inexpensive method and routinely available.
For lung cancer diagnosis, the related markers are carcinoembryonic antigen (CEA), soluble fragment of cytokeratin 19 (CYFRA21-1) and neuron-speci c enolase (NSE). Recent studies have reported that high levels of CEA, CYFRA21-1, NSE and carbohydrate antigen (CA)-125 are associated with poor prognosis in NSCLC [5][6][7]. CEA is a glycoprotein in carcinoembryonic cell surface, and it has been a widely applied tumor marker in serum and PF. In NSCLC, CEA has been considered a marker for the adenocarcinoma histological subtype [8]. CYFRA21-1 is a soluble fragment of cytokeratin 19 and a primitively expressed epithelial cytokeratin. CA-125 is a mucinous glycoprotein originated from fetal lumen epithelium and it is used for the follow-up of ovarian cancer [9]. NSE is an isozyme of enolase, existed in neuroendocrine tumors. NSE is a sensitive marker in the diagnosis of small cell lung cancer (SCLC), but increased serum NSE has been reported in 11.7% to 28% of patients with NSCLC [5].
However, the ndings regarding the prognosis of serum tumor markers have been studied in lung cancer with con icting results and the relationship between combined detection of serum and PF tumor markers and prognosis in advanced NSCLC is unclear [10][11][12]. Therefore, the aim of this study is to assess the prognostic value of combined detection of serum and PF tumor markers CEA, CYFRA21-1, CA-125 and NSE in advanced NSCLC.

Results
Baseline characteristics of patients A total of 335 patients with MPE were histologically com rmed NSCLC. Ninety-one patients were excluded for some reasons, including 45 patients incomplete data, 18 patients lost to follow-up and 28 patient no treatment. Finally, 244 patients were enrolled in this study (Fig 1).
Characteristics of the patients are outlined in Table 1

CEA analysis
High levels of serum CEA was adversely prognostic factors for PFS and OS. The PFS was 6.6 months versus 5.0 months in patients with normal and elevated level of serum CEA (P = 0.003) and the OS was 15.0 months versus 12.0 months in patients with normal and high levels of tumor marker (P = 0.030). We analyzed the prognostic value of combination of serum and pleural uid CEA. High levels of SPF CEA were related to worse PFS and OS. The PFS was 6.6 months versus 5.0 months in patients with normal and elevated level of SPF CEA (P = 0.003) and the OS was 15.3 months versus 10.6 months in patients with normal and high levels of tumor marker (P = 0.001) (Fig 3).

CYFRA21-1 analysis
We found that abnormal levels of serum CYFRA21-1 and SPF CYFRA21-1 were negative prognostic factors. The PFS was 6.5 months versus 5.3 months in patients with normal and elevated level of serum CYFRA21-1 (P < 0.001) and the OS was 16.1 months versus 11.7 months in patients with normal and high levels of tumor marker (P = 0.004). The PFS was 6.5 months versus 5.3 months in patients with normal and elevated level of SPF CYFRA21-1 (P = 0.001) and the OS was 15.8 months versus 11.1 months in patients with normal and high levels of tumor marker (P = 0.001) (Fig 4).

CA-125 analysis
Serum CA-125 and SPF CA-125 were negative prognostic factors for OS. The OS was 15.0 months versus 12.0 months in patients with normal and elevated level of serum CA-125 (P = 0.043) and 12.7 months versus 12.0 months in patients with normal and high levels of SPF CA-125 (P = 0.023) (Fig 5).

NSE analysis
High levels of SPF NSE was a negative prognostic factor for PFS. The PFS was 6.4 months versus 5.5 months in patients with normal and elevated level of SPF NSE (P = 0.019) (Fig 6).
Serum CEA, CYFRA21-1, CA-125 and NSEcombination We analyzed the prognostic value of combination of serum tumor markers. For patients with 4 tumor markers elevated PFS was 4.0 months versus 5.5 months in patients with 3 tumor markers, 6.4 months in patients with 2 tumor markers, 7.0 months in patients with 1 tumor marker and 6.8 months in patients with all the tumor markers negative (P = 0.001). We found the same results for OS in patients with a combination of serum tumor markers (P = 0.018) (Fig 7).

Dissusion
With the progress of molecular biology in recent years, the role of tumor markers in predicting outcome and detecting recurrence of patients with lung cancer has been the subject of extensive researches [13,14]. The results of our study indicated that high levels of SPF CEA and SPF CYFRA21-1 may be used as biomarkers to predict outcomes in NSCLC with MPE; this conclusion was based on the signi cant differences in OS and PFS observed in these patients. To the best of our knowledge, this study is the rst to analyze the relationship between combined detection of SPF tumor markers and prognosis in advanced NSCLC.
Tumor marker CEA and CYFRA21-1 has been reported as poor prognostic factors in NSCLC. In patients with stage IIIB-IV NSCLC, Arrieta et al. showed that patients with elevated serum CEA had shorter OS than patients who had normal levels [15]. In patients with stage I-III NSCLC, Tomita et al. observed that elevated CEA in pleural lavage uid (PLF) was a signi cant prognostic factor for OS [16]. In the meta-analysis of Zhang et al. with 1990 patients, serum CYFRA21-1 and CEA were independent prognosis factors of NSCLC patients [6]. We found that high levels of serum CEA and CYFRA21-1 were unfavorable prognostic factors for survival, but this results were inconsistent with Tokito et al [13]. Interestingly, when combining use of serum and pleural uid tumor markers, we found that SPF CEA and CYFRA21-1 correlated with prognosis in a signi cant and independent manner. In other words, high levels of SPF CEA and CYFRA21-1 may be useful noninvasiveness markers to identify advanced NSCLC risk. It is not surprising that the relationship between elevated CEA or CYFRA21-1 levels and prognosis remains controversial. In fact, many studies reporting on the prognosis of tumor markers levels were based on different stages or subtypes. Further prospective validation of our data would be necessary.

Previously, Muley et al. introduced the use of a Tumor Marker Index (TMI)
, based on the combined use of serum CEA and CYFRA21-1 [17]. Muley et al. showed that TMI was the prognostic signi cance in operate early stage of NSCLC, and this is substantiated by Tomita et al [18]. Barlèsi et al. analyzed the prognotic value of combination of CEA, CYFRA21-1 and NSE in stage IIIB-IV NSCLC and found that combination of the three markers was a poor prognostic factor [19]. In contrast, Blankenburg et al. reported that TMI was not associated with a worse outcome in stage I NSCLC [20]. In the present study, the use of tumor marker indexs including serum CEA, CYFRA21-1, CA-125 and NSE evaluated the prognosis and progression of patients with stage IV NSCLC. We found that combining of the four markers was a poor prognostic factor.
The simultaneous use of several serum tumor markers may increase the reliability of prognostic value.
Several reports have been published about the prognostic value of serum CA-125 and NSE. In a random survival forest prognostic model, serum CA-125 was related to a poor outcome in lung adenocarcinoma patients with brain metastasis [21]. Li et al. reported that high levels of serum NSE was correlated with worse survival in resected lung adenocarcinoma patients harboring anaplastic lymphoma kinase (ALK) fusion gene rearrangements [7]. We found that serum CA-125 and SPF CA-125 were prognostic risk factors for OS. In our study, serum NSE was not a prognostic factor, but SPF NSE was associated with worse PFS.
We aslo found that PF levels of CEA, CYFRA21-1, CA-125 and NSE were signi cantly higher than those in serum. Because of the large molecular weight of tumor markers, they are easier to accumulate in the pleural cavity than the blood. Also, tumor markers in the pleural cavity are not easily inactivated by liver. Hackner et al. showed that ratio of CEA in PF and serum was a useful tool for diagnosis of MPE [22].
There are several limitations in our retrospective study. The populations enrolled in the study were suitable for undergoing thoracoscopy and the selection bias could not be completely avoided. In addition, we did not perform the trials evaluating the use of consecutive measurenment of tumor markers during treatment and follow-up. Finally, further prospective study in this area are warranted.

Conclusions
In