Is neuron-specific enolase useful for diagnosing malignant pleural effusions? evidence from a validation study and meta-analysis

Neuron-Specific enolase (NSE) has been used as a typical tumor marker and shows a potential to diagnose malignant pleural effusion (MPE). The ability of NSE in diagnosing MPE has been investigated in many studies, but with inconsistent conclusions. This study sought to investigate the diagnostic accuracy of NSE for MPE through a clinical study and together with a meta-analysis. Pleural effusion samples from 136 patients with MPE and 102 patients with benign pleural effusion (BPE) were collected, and NSE levels were measured by electrochemiluminescence immunoassay. Receiver operating characteristic (ROC) curve analysis was performed to assess the ability of NSE to differentiate MPE from BPE. Literature search was conducted to identify suitable publications, data were extracted and diagnostic indexes including sensitivity, specificity, positive/negative likelihood ratio (PLR/NLR), and diagnostic odds ratio (DOR) were pooled. Summary ROC curve was generated to determine the overall diagnostic accuracy of NSE for MPE. Levels of NSE were significantly increased in pleural effusion from patients with MPE than that from BPE (18.53 ± 27.30 vs. 6.41 ± 6.95 ng/ml, p < 0.001). With a cut-off value of 8.92 ng/ml, pleural NSE had a sensitivity of 59.56% and a specificity of 83.33% in diagnosing MPE. A total of 14 studies with 1896 subjects were included for meta-analysis. The diagnostic parameters of NSE were listed as follows: sensitivity, 0.53 (95% CI: 0.38–0.67); specificity, 0.85 (95% CI: 0.75–0.91); PLR, 3.54 (95% CI: 2.33–5.39); NLR, 0.56 (95% CI: 0.42–0.73); and DOR, 6.39 (95% CI: 3.72–10.96). The area under the summary ROC curve was 0.78. The role of pleural NSE measurement in diagnosing MPE is limited and with a low sensitivity. The clinical utility of NSE assay should be combined with the results of other tumor markers examination and the detail clinical information of patient. Further studies are needed to confirm the role of NSE in diagnosing MPE.


Background
Neuron-specific enolase (NSE), which localized predominately in the cytoplasm of neurons, is a cell specific isoenzyme of the glycolytic enzyme enolase [1]. During normal condition, NSE is not secreted. While NSE is up-regulated to maintain homeostasis when axons are injured, thus, NSE is a classical biomarker that directly evaluates functional damage to neurons [2], and lots of studies have found that NSE is a biomarker of neurological disorders [3]. Considering NSE as a specific biomarker for neurons and peripheral neuroendocrine tissues, the increased expression of NSE in both tissues and circulations may be presented with malignant proliferation of neuroendocrine tissues, and thus could be of potential value in the diagnosing, staging and guiding treatment of such cancers [1,4].
Small-cell lung cancer (SCLC), a malignant disease associated with neuroendocrine differentiation, is characterized by its rapid doubling time, high growth fraction, and early propensity for metastases [5,6]. Non-small-cell lung cancer (NSCLC) also presented with neuroendocrine properties, since both SCLC and NSCLC originate from a common cell lineage and differentiated lately for oncogenetic development, studies reported that about 11.7-28% of patients with NSCLC presented with increased serum NSE levels [7,8]. Thus, neuroendocrine marker like NSE has been proved to be useful in immunohistochemically differentiating NSCLC and SCLC, which released into the blood and body fluid, can be used as tumor marker [1].
Malignant pleural effusion (MPE) is caused by lung cancer and other malignant diseases. The presence of pleural effusion also suggests metastases of tumor, indicating an unoptimistic prognosis [9]. Thus, to diagnose MPE early and accurately may benefit patient with timely and effective treatments [10]. Many studies have reported that NSE levels increased significantly in MPE, NSE may be a biomarker for MPE [11,12]. However, the results of these studies are so different, and there is no definite conclusion on the diagnostic value of NSE for MPE. The present study sought to validate the diagnostic accuracy of NSE for MPE in Chinese patients, and summarize the overall diagnostic accuracy of NSE for MPE through a metaanalysis based on current available literatures.

Patient inclusion
Ethics Committee of West China Hospital of Sichuan University approved this study protocol. This study was performed based on the principles expressed in the Declaration of Helsinki. Written informed consents were collected from all patients for the collection of clinical samples and subsequent analysis at admission.
From February 2011 to August 2013, 238 patients with undiagnosed pleural effusion admitted to our hospital for further investigation were included this retrospective clinical study. Among them, 136 patients were diagnosed as MPE, which was diagnosed by experienced pathologists based on identification of malignant cells in pleural fluid as detected using cytological tests or biopsy analysis on pleura or lung tissues. 102 patients with benign pleural effusion (BPE) were also recruited as controls.

Sample collection and measurement
All include patients underwent a standard thoracocentesis before the treatment, during which pleural effusion samples were collected. When multiple thoracenteses were performed on the same patient, only the first sample was analyzed. For serum sample collection, after fast overnight from 21:00, venous blood samples from patients were collected and serum was separated immediately. Both pleural effusion and serum samples were collected and sent for biochemical analysis in the department of laboratory medicine. Serum and pleural NSE levels were measured by an electrochemiluminescence immunoassay (Roche Cobas 8000 modular analyser series; Roche Diagnostics, USA). Pleural glucose, total protein, lactate dehydrogenase levels were examined simultaneously. Technicians processing pleural effusion samples for NSE measurement and biochemical assays were blinded to patient details.

Statistical analysis
Data were presented as the Means ±standard deviation. Difference in MPE and BPE groups was analyzed by the nonparametric Mann-Whitney U-test. Differences among multiple groups were detected with analysis of variance (ANOVA). Receiver operating characteristic (ROC) curves were constructed, and areas under the curve (AUC) were measured to quantify the accuracy of NSE to discriminate MPE from BPE. The optimal cut-off value was set to obtain the best sensitivity and specificity for diagnosing MPE. Statistical analysis was performed using SPSS 18.0 software (Chicago, IL, USA). A value of p < 0.05 was set as significant.

Meta-analysis
This meta-analysis was carried out based on the standard method that recommend for meta-analysis of diagnostic studies and the guidelines of the Preferred Reporting Items for Systematic Reviews [13].
We searched in PubMed and EMBASE for eligible articles published up to March 2016, the following search terms were used as Medical Headings and/or text words: "Neuron specific enolase or NSE" AND "Malignant pleural effusion or malignant pleural fluid or malignant hydrothorax" AND "sensitivity or specificity or accuracy". Potential related studies were also checked from the reference lists of the included original and review articles. Studies were included if: they measured the accuracy of pleural NSE for differentiating MPE and BPE in humans; they presented sufficient data to calculate true positive (TP), false positive (FP), false negative (FN), and true negative (TN) rates, and they were published in English. Data were retrieved and formed a 2 × 2 table of diagnostic performance. A 14-items Quality Assessment of Diagnostic Accuracy Studies (QUADAS) list was used to evaluate the quality of included studies [14].
The meta-analysis was carried out using a bivariate regression model [15,16], with which we calculated pooled sensitivity, specificity, positive/negative likelihood ratios (PLR/NLR), and diagnostic odds ratios (DOR). We also generated summary receiver operating characteristic (SROC) curves to summarize the diagnostic accuracy performance of NSE [17]. Heterogeneity was evaluated using the I 2 inconsistency test, I 2 > 50% suggested substantial heterogeneity. Potential publication bias was detected by Deeks's funnel plot test [18]. All statistical analysis was conducted using STATA 12.0 (Stata Corp., College Station, TX). All statistical analysis was two-sided, a p value <0.05 was set as statistically significant.

General clinical data of patients
There were 136 patients with MPE, including 74 males and 62 females with mean ages of 58 years. In MPE patients, Cytology examinations were positive in 56 cases, corresponding to a positive rate of 41.17%. Among patients with MPE, 101 had NSCLC (90, lung adenocarcinoma; 11, lung squamous cell carcinoma); 11, small cell lung carcinoma; 18, metastatic carcinoma; 5, lymphoma, and 1, malignant mesothelioma.
There were 102 patients with BPE as controls, including 68 males and 34 females, with mean ages of 56 years. These patients had been diagnosed with the following conditions: tuberculous pleurisy, 49; parapneumonic effusion, 26; heart failure, 25; liver cirrhosis, 1; and chylothorax, 1. The MPE and BPE groups didn't differ significantly on age or gender. The clinical information and pleural fluid characteristics of both MPE and BPE group are listed in Table 1.

Levels of NSE
The levels of NSE in both serum and pleural effusion were significantly increased in patients with MPE than in patients with BPE (serum 19.51 ± 16.54 vs. 13.77 ± 13.33 ng/ ml, p = 0.004; pleural effusion 18.53 ± 27.30 vs. 6.41 ± 6.95 ng/ml, p < 0.001) ( Table 1). In patients with MPE, the SCLC patients showed the highest levels of NSE in both serum and pleural effusion when compared with other causes of MPE (both P < 0.001), as shown in Fig. 1. After adjusted by pleural protein, the patients with MPE remained have a higher levels of NSE in serum and pleural effusion than patients with BPE (Additional file 1: Fig. S1).

Diagnostic accuracy of NSE
Next, we evaluated the diagnostic accuracy of NSE for MPE with ROC curves. At a cut off value of 8.92 ng/ml, the diagnostic sensitivity and specificity of pleural NSE for MPE were 59.56% and 83.33%, respectively, and the AUC was 0.76. At a cut off value of 12.29 ng/ml, serum levels of NSE play a role in diagnosing MPE with the sensitivity and specificity of 66.91% and 62.75%, respectively, but the AUC was only 0.65, as shown in Fig. 2. Meanwhile, the AUC of pleural/serum NSE ratio in diagnosing MPE was 0.68 (Fig. 2).
We also noticed that in 11 SCLC patients, the serum/ pleural levels of NSE were the highest among all causes of MPE. When compared with BPE patients, at a cut-off value of 17.42 ng/ml, pleural NSE plays a valuable role in diagnosing MPE with the sensitivity and specificity of 100% and 92.16%, respectively, and the AUC was 0.99. The diagnostic summary of serum and pleural levels of NSE for MPE and SCLC related MPE was listed in Table 2.

Discussion
To diagnose MPE accurately remains a clinical challenge, and the searching for useful biomarkers for MPE is still on the way. NSE is typical marker for cancers with neuroendocrine characteristic, especially for SCLC. Growing studies suggested that NSE is increased in MPE, and it may be a biomarker for MPE [32]. However, these studies gave different results. This study validated the diagnostic accuracy of NSE for MPE in 238 patients, which included the second largest patients that evaluated the diagnostic utility of NSE for MPE. In addition, we next performed a meta-analysis with 1896 subjects to make a full judgment of NSE for diagnosing MPE based on current available publications.
In this study, we enrolled 136 MPE patients, and we observed that both serum and pleural levels of NSE were higher in patients with MPE than in patients with BPE, even after adjustment by pleural protein. Pleural NSE  shows a better diagnostic performance than serum NSE and pleural/serum NSE ratio, and its sensitivity and specificity were 59.56% and 83.33%, respectively. Pleural NSE showed a low sensitivity and a high rate of missed diagnoses, which may be due to only a limited proportion of NSCLC patients with neuroendocrine characteristic [7,8]. Thus, the clinical value of NSE alone in screening MPE is limited. It may be more appropriate to use the combination of NSE and other tumor markers for diagnosing MPE [22][23][24][25][26].
The diagnostic performance of a serial of tumor markers for MPE, such as carcino-embryonic antigen, carbohydrate antigen 19-9, carbohydrate antigen 15-3, has been summarized by two meta-analysis, and studies suggested that one tumor marker alone doesn't have not enough sensitivity to diagnose MPE, the combination of two or more tumor markers may increase the sensitivity and play more role in MPE diagnosis and management [33,34]. In clinical utility, the results of NSE test should be used in conjugation with other tumor markers tests, and clinical information of patients, such as previous medical history, radiological findings.
For MPE with multiple causes (Lung adenocarcinoma, lung squamous cell carcinoma, SCLC, other etiologies), we noticed that both serum and pleural levels of NSE were highest in patients with SCLC, and NSE show a high diagnostic accuracy for SCLC-related MPE. Both serum and pleural NSE reach a sensitivity of 100% for SCLC-related MPE. Such results were also supported by Miédougé's report [23]. These findings suggest the diagnostic performance of NSE may be tumor-subtype specific. Based on above findings, the NSE may not be used for screening MPE at the first choice. But for patients who were highly suspected for SCLC or neuroendocrine tumors, the examination of NSE may provide more valuable information.
To make a systemic assessment of the diagnostic performance of NSE for MPE, we performed an updated meta-analysis. A recent published meta-analysis has discussed the diagnostic role of NSE for MPE [35], however, it included only seven studies, and missed several studies. Thus, we made a more systemic literature search and updated this meta-analysis. In our meta-analysis, there were 1896 cases of patients, and the pooled sensitivity and specificity of NSE were 0.53 and 0.85, respective, confirmed   Our study had several limitations. First, we only recruited 238 patients, and our meta-analysis only included 1896 patients, such limited number of patients may be not adequate for building final conclusions on the ability of NSE in diagnosing MPE. Second, only articles published in English were included, and there may be language bias exist, we may also miss the studies that not in the searched databases, both may bias the results. Further studies should include more patients from different centers to confirm the diagnostic role of NSE for MPE. The current NSE assay is with low sensitivity, it may be helpful to develop a novel method to examine NSE and increase the diagnostic accuracy. In addition, we found substantial heterogeneity among included studies. However, we didn't investigate potential covariates with meta-regression analysis due to limited included studies.

Conclusions
Taken together, the role of pleural NSE examination in diagnosing MPE is limited with low sensitivity. Our study suggests that the interpretation of NSE results should be in combination with the results of other tumor markers, and clinical data of patients. Further studies are needed to confirm our findings.

Additional file
Additional file 1: Fig. S1. Serum and pleural levels of neuron specific enolase in patients standardized by pleural protein levels. After standardized by pleural protein levels, both serum and pleural levels of neuron specific enolase were higher than that in patient with benign pleural effusion.