Pleural biomarkers in diagnostics of malignant pleural effusion: a narrative review.

Although cytology and pleural biopsy of pleural effusion (PE) are the gold standards for diagnosing malignant pleural effusion (MPE), these tools' diagnostic accuracy is plagued by some limitations such as low sensitivity, considerable inter-observer variation and invasiveness. The assessment of PE biomarkers may hence be seen as an objective and non-invasive diagnostic alternative in MPE diagnostics. In this review, we summarize the characteristics and diagnostic accuracy of available PE biomarkers, including carcinoembryonic antigen (CEA), neuron-specific enolase (NSE), carbohydrate antigens 125 (CA125), carbohydrate antigen 19-9 (CA19-9), carbohydrate antigen 15-3 (CA15-3), a fragment of cytokeratin 19 (CYFRA 21-1), chitinase-like proteins (CLPs), vascular endothelial growth factor (VEGF) and its soluble receptor, endostatin, calprotectin, cancer ratio, homocysteine, apolipoprotein E (Apo-E), B7 family members, matrix metalloproteinase (MMPs) and tissue-specific inhibitors of metalloproteinases (TIMPs), reactive oxygen species modulator 1 (Romo1), tumor-associated macrophages (TAMs) and monocytes, epigenetic markers (e.g., cell-free microRNA and mRNA). We summarized the evidence from systematic review and meta-analysis for traditional tumor markers' diagnostic accuracy. According to the currently available evidence, we conclude that the traditional tumor markers have high specificity (around 0.90) but low sensitivity (around 0.50). The diagnostic accuracy of novel tumor markers needs to be validated by further studies. None of these tumor biomarkers would have sufficient diagnostic accuracy to confirm or exclude MPE when used alone. A multi-biomarker strategy, also encompassing the use of artificial intelligence algorithms, may be a valuable perspective for improving the diagnostic accuracy of MPE. 2021 Translational Lung Cancer Research. All rights reserved.

Introduction

Pleural effusion (PE) is a relatively common sign in clinical practice. It can be caused by various conditions, including malignant pleural effusion (MPE), tuberculous pleural effusion (TPE), parapneumonic effusion (PPE), heart failure (HF), and others (1). PE cytology and thoracentesis with pleural biopsy are the gold standards for diagnosing MPE (1). However, both of these techniques have limitations. Low sensitivity is perhaps the major drawback of PE cytology, though its specificity is very close to 100% (2). Like PE cytology, the pleural biopsy also has very high specificity (i.e., approximating 100%), but is invasive, may be vulnerable to sampling error, is highly subjective, and has the inherent risk of operation-related complications (3). According to this evidence, development of non-invasive and objective diagnostic tools for MPE seems a promising perspective.

The measurement of some PE biomarkers has been proposed as a valuable approach for diagnosing MPE. Compared with PE cytology and pleural biopsy, diagnostic testing is mini-invasive, provides more objective results and is relatively less expensive (4). Furthermore, reliable and non-invasive testing allows early specialist referral, minimizes diagnostic delays, and optimizes clinical trials access (5).

To date, a number of biomarkers have been tested, and their diagnostic accuracy for MPE has been evaluated in a growing number of studies. Here, we provide a narrative review based on current scientific literature, aimed to summarize the current evidence of using pleural diagnostic biomarkers in MPE diagnostics. We present the following article in accordance with the Narrative Review reporting checklist (available at http://dx.doi.org/10.21037/tlcr-20-1111).

Conventional cancer biomarkers

To date, numerous studies have investigated the diagnostic accuracy of conventional cancer biomarkers for MPE, including carcinoembryonic antigen (CEA) (6-8), neuron-specific enolase (NSE) (6,9), carbohydrate antigens 125 (CA125) (6,10), carbohydrate antigen 19-9 (CA19-9) (6,10), carbohydrate antigen 15-3 (CA15-3) (6,10,11) and a fragment of cytokeratin 19 (CYFRA 21-1) (6,8,10). summarizes the evidence from meta-analyses investigating the diagnostic accuracy of these cancer biomarkers. Generally, the overall sensitivity of traditional cancer biomarkers is unsatisfactory (around 0.50), while their specificities seem globally higher (>0.90). Notably, nearly all the meta-analyses highlighted the presence of publication bias, thus indicating that the diagnostic accuracy of these conventional cancer biomarkers may be overestimated. A meta-analysis performed by Yang et al. in 2017 investigated the diagnostic accuracy for MPE of combined tumor markers (12) and reported that the combinations of positive pleural CEA + CA 19-9 and CEA + CA 15-3 had extremely high specificities for MPE (i.e., 0.99), but the sensitivity was low (around 0.65).

Table 1

Diagnostic accuracy of traditional tumor markers for malignant pleural effusion: evidence from meta-analysis

Author	Year	Tumor marker	MPE/BPE	N	Model	AUC	Sensitivity (95% CI)	Specificity (95% CI)	PB
Nguyen (6)	2015	CEA	NR	36	REM	0.81	0.55 (0.49–0.61)	0.96 (0.94–0.98)	Yes
Shi (7)	2008	CEA	2,834/3,251	45	REM	0.77	0.54 (0.52–0.55)	0.94 (0.93–0.95)	Yes
Gu (8)	2007	CEA	1,404/1,543	15	DLM	0.77	0.46 (0.43–0.49)	0.97 (0.96–0.98)	NR
Nguyen (6)	2015	NSE	NR	7	REM	0.84	0.61 (0.26–0.96)	0.88 (0.73–1.00)	Yes
Zhu (9)	2017	NSE	1,093/803	14	BVM	0.78	0.53 (0.38–0.67)	0.85 (0.75–0.91)	No
Nguyen (6)	2015	CA125	NR	10	REM	0.79	0.58 (0.33–0.82)	0.93 (0.78–1.00)	Yes
Liang (10)	2008	CA125	512/801	10	REM	0.88	0.48 (0.44–0.53)	0.85 (0.83–0.88)	Yes
Nguyen (6)	2015	CA15-3	NR	11	REM	0.78	0.51 (0.40–0.62)	0.98 (0.96–1.00)	Yes
Wu (11)	2015	CA15-3	1,492/1,414	21	REM	0.84	0.58 (0.56–0.61)	0.91 (0.90–0.93)	Yes
Liang (10)	2008	CA15-3	819/966	11	REM	0.73	0.51 (0.47–0.54)	0.96 (0.95–0.97)	Yes
Nguyen (6)	2015	CA19-9	NR	8	REM	0.91	0.38 (0.08–0.68)	0.98 (0.95–1.00)	Yes
Liang (10)	2008	CA19-9	598/488	7	REM	0.78	0.25 (0.21–0.28)	0.96 (0.94–0.98)	Yes
Nguyen (6)	2015	CYFRA 21-1	NR	16	REM	0.84	0.63 (0.48–0.77)	0.93 (0.86–1.00)	Yes
Liang (10)	2008	CYFRA 21-1	1,152/1,122	18	REM	0.83	0.55 (0.52–0.58)	0.91 (0.90–0.93)	Yes
Gu (8)	2007	CYFRA 21-1	890/730	12	DLM	0.82	0.47 (0.44–0.51)	0.92 (0.90–0.94)	NR
Yang (12)	2017	CEA+CA125	107/167	3	REM	0.86	0.65 (0.56–0.74)	0.98 (0.94–0.99)	No
Yang (12)	2017	CEA+CA153	272/443	4	REM	0.88	0.64 (0.58–0.70)	0.98 (0.97–0.99)	No
Yang (12)	2017	CEA+ CA19-9	264/355	4	REM	0.96	0.58 (0.51–0.64)	0.98 (0.96–0.99)	No
Yang (12)	2017	CEA+CYFRA 21-1	325/415	7	REM	0.95	0.82 (0.77–0.86)	0.92 (0.89–0.95)	No
Yang (12)	2017	CA15-3+CYFRA 21-1	139/159	4	REM	0.98	0.89 (0.82–0.93)	0.94 (0.89–0.97)	Yes

CEA, carcinoembryonic antigen; NSE, neuron-specific enolase; CA125, carbohydrate antigens 125; CA19-9, carbohydrate antigen 19-9; CA15-3, carbohydrate antigen; CYFRA 21-1, fragment of cytokeratin 19 (CYFRA 21-1); MPE, malignant pleural effusion; BPE, benign pleural effusion; N, number of included studies; NR, not reported; AUC, area under summary receiver operating characteristic curve; DLM, DerSimonian Laird method; REM, random-effects model; BVM, bivariate model; PB, publication bias.

In another meta-analysis (13), the diagnostic accuracy of CEA, CA19-9, and CYFRA 21-1 for lung adenocarcinoma-associated MPE has been investigated. Among these tumor markers, CEA displayed the highest area under the curve (AUC; 0.93), and its sensitivity and specificity were 0.75 and 0.96, respectively. CA19-9 had 0.58 sensitivity, and 0.84 specificity, whilst CYFRA 21-1 exhibited 0.70 sensitivity and 0.88 specificity, respectively. These results indicate that the diagnostic accuracy of CEA, CA19-9, and CYFRA 21-1 for lung adenocarcinoma-associated MPE shall be considered overall modest.

Along with conventional cancer biomarkers earlier mentioned, some others have been investigated. Human epididymis protein 4 (HE4) has been for long considered a specific ovarian cancer biomarker, though a recent meta-analysis emphasized its moderate diagnostic accuracy for lung cancer (14). To date, two studies have investigated the diagnostic accuracy of HE4 for MPE (15,16). In the first, the authors prospectively enrolled 32 patients with MPE and 54 patients with benign pleural effusion (BPE). They found that the AUC of pleural HE4 was 0.89, with 0.85 sensitivity and 0.91 specificity at a diagnostic threshold of 1,675 pmol/L (16). In the second study, the authors investigated the diagnostic accuracy of HE4 for lung cancer-associated MPE (15) and found that pleural HE4 had an AUC of 0.83 for diagnosing MPE. At the 652.2 pmol/L threshold, the sensitivity and specificity of HE4 were 0.78 and 0.75, respectively.

Four studies investigated the diagnostic accuracy of squamous cell carcinoma antigen (SCC) for diagnosing MPE (17-20). Two of these concluded that SCC concentration was not significantly increased in MPE (17,20), whilst in two other studies, the diagnostic accuracy was found to be poor (18,19). Notably, two of these studies also found that patients with squamous-cell lung cancer had considerably increased SCC values (19,20). Taken together, these results suggest that the diagnostic accuracy of SCC is low for MPE, whilst this biomarker may have high diagnostic accuracy for squamous-cell lung cancer-associated MPE.

Progastrin-Releasing peptide (ProGRP) was proven as a useful diagnostic biomarker for small cell lung cancer (SCLC), with 0.64 sensitivity and 0.94 specificity, respectively (21). Only one study has analyzed the potential diagnostic utility of proGRP for MPE (22). In this study, the proGRP level in PE was found to be significantly higher in patients with SCLC, but its diagnostic accuracy was not assessed by statistical methods. In addition, no correlation was found between proGRP and NSE values, thus indicating that the combination of these two biomarkers may improve the overall diagnostic accuracy.

The ratio between PE and serum of cancer biomarkers can also be used for diagnosing MPE. In a prospective study including 98 patients with MPEs and 103 with BPEs, the authors investigated the diagnostic accuracy of pleural effusion CEA (pCEA) to serum CEA (sCEA) ratio (23). Surprisingly, it was found that the ratio displayed an AUC as high as 0.90, with 0.85 sensitivity and 0.92 specificity at a threshold of 1.0, thus indicating that the pCEA/sCEA ratio may represent a promising diagnostic tool.

Vascular endothelial growth factor (VEGF) and its soluble receptor

VEGF is a glycoprotein that acts as a promoter of angiogenesis, and is also critically involved in MPE formation. The first study investigating the diagnostic accuracy of VEGF for MPE has been published in 1993 (24). The authors found that the diagnostic sensitivity and specificity of VEGF in PE were 0.72 and 0.74, respectively (24). Several other studies have then addressed the diagnostic role of PE VEGF in MPE, providing contradictory results (25,26). In 2012, a meta-analysis pooled the results of ten available studies, concluding that the diagnostic sensitivity and specificity of VEGF were 0.75 and 0.72, respectively (27). Some other studies on this topic have since been published (28-31), so that an updated meta-analysis may be needed to re-assess the diagnostic accuracy of VEGF in MPE.

VEGF has three specific tyrosine kinase receptors, called VEGF receptor-1 (VEGFR1), VEGF receptor-2 (VEGFR2), and VEGF receptor-3 (VEGFR3) (32). VEGFR1, also known as Flt-1, is a negative angiogenesis regulator (32). The soluble form of VEGFR-1 (sVEGFR-1) has been identified in PE (33), with exudate displaying significantly higher sVEGFR-1values than transudate (33). To date, two studies have investigated the diagnostic accuracy of sVEGFR-1 in MPE. In the earlier study, including 40 patients with MPEs and 15 with BPEs, higher sVEGFR-1 value has been found in MPE. The AUC of sVEGFR-1 was as high as 0.93 (34), whilst sensitivity and specificity were 0.92 and 0.93, respectively, using a 852 pg/mL diagnostic threshold (34). Nevertheless, the high diagnostic accuracy of sVEGFR-1 was not confirmed in a subsequent study (35), including 44 patients with MPEs and 36 with BPEs, whereby the AUC was found to be 0.79, with 0.88 sensitivity and 0.58 specificity at a threshold of 3.95 ng/mL (35).

Endostatin

Unlike VEGF, which promotes angiogenesis, endostatin is an endogenous angiogenesis inhibitor, whose concentration has been associated with enhanced risk of developing various malignant diseases (36). The first report on the diagnostic accuracy of endostatin for diagnosing MPE has been published in 2003. In this study cohort, encompassing 38 patients with MPE and 29 with BPE, the sensitivity and specificity of this biomarker were 0.68 and 0.55, respectively (37). In 2013, a meta-analysis was performed (38), concluding that the pooled sensitivity and specificity of endostatin for diagnosing MPE were 0.69 and 0.78, respectively, thus underscoring that the overall diagnostic accuracy of this biomarker seems rather poor.

Calprotectin

In 2010, Rodríguez-Piñeiro et al. carried out a proteomic analysis for verifying the significance of some proteins for diagnosing MPE (39). Interestingly, S100-A8 and S100-A9, which form the non-covalent heterodimer, named calprotectin, were found to be decreased in MPE (39). An ensuing clinical trial was planned for evaluating the diagnostic accuracy of calprotectin in MPE, including 67 patients with MPE and 89 with BPE (40). It was finally found that calprotectin showed a considerably high diagnostic accuracy for MPE (AUC, 0.96). However, these results were not confirmed by three subsequent studies (41-43), which showed that the diagnostic accuracy (i.e., the AUC) of calprotectin in MPE was 0.68 (41), <0.50 (42), and 0.85 (43), respectively.

This results inconsistency may be attributable to the wide disease spectrum in the different study cohorts, as well as to the use of different calprotectin assays. Interestingly, in a recently published work, the authors found that calprotectin in transudate (defined by Light’s criteria) was significantly lower than that in MPE (43), thus indicating that the proportion of transudate may have an impact on the diagnostic accuracy of this biomarker. In summary, more evidence would be needed to ascertain the diagnostic utility of calprotectin in PE.

Cancer ratio and cancer ratio plus

Cancer ratio, which is defined as serum lactate dehydrogenase (LDH) to PE adenosine deaminase ratio (ADA) ratio, has been firstly proposed by Verma et al., as a potential diagnostic parameter in MPE (44). The authors studied 100 patients with MPEs, 40 with TPE and 14 with PPE, reporting that the cancer ratio was increased in MPE, displaying an AUC of 0.81 for diagnosing MPE (44). This initial finding has then been validated in subsequent studies (45-47). A meta-analysis, published in 2019, concluded that sensitivity and specificity of cancer ratio were 0.97 and 0.89, respectively (48). Nonetheless, all of the available studies were only based on PPE and/or TPE as control, whilst transudate was not included in the studied cohorts (48). Therefore, further studies using transudate would be necessary for better evaluating the diagnostic value of cancer ratio in MPE. Although HF due to PE is usually categorized into transudate by Light’s criteria, misclassification as exudate could occur due to diuretic therapy (49), whilst nearly 5% of MPE can be misclassified into transudate (4). It is hence essential to include HF into the study cohort when investigating the diagnostic accuracy of MPE biomarkers.

Cancer ratio plus, which has also been proposed by Verma et al., is another diagnostic MPE parameter (45), defined as the ratio between cancer ratio and pleural lymphocyte count. To present, only one study has evaluated the diagnostic accuracy of cancer ratio plus in MPE. In this investigation, including 87 patients with MPEs and 34 with TPEs, the authors found an AUC of 0.86 for cancer ratio plus, which was higher than that of cancer ratio (AUC, 0.81). Using a cut-off of 50, the sensitivity and specificity of cancer ratio plus were 0.89 and 0.94, respectively, thus indicating that cancer ratio plus may also be seen as a useful diagnostic parameter in MPE. However, these results would also need to be validated in further studies, with larger sample sizes and more representative patients’ cohorts.

Homocysteine

Homocysteine is an amino acid that is currently considered a risk factor for the development of cardiovascular and Alzheimer’s diseases (50). In 2015, Santotoribio et al. firstly reported that homocysteine was increased in PE of patients with MPE (51). In a study cohort including PPE, MPE, TPE, and transudate, the AUC of homocysteine for diagnosing MPE was 0.83, with 1.00 sensitivity and 0.57 specificity at 9.4 µmol/L threshold. These findings were replicated and validated in a subsequent study (52).

Apolipoprotein E (Apo-E)

Apo-E is a plasma lipoprotein that is usually considered a risk factor for cardiovascular disease, and some of its genetic polymorphisms have also been associated with increased risks of developing Alzheimer’s disease (53). In a proteomic study performed in 2005, Apo-E was found to be higher in lung adenocarcinoma-induced MPE (54), indicating that this protein may retain a potential diagnostic value for MPE. In 2013, a prospective study investigated the diagnostic accuracy of Apo-E in MPE (55). In this study, including 160 patients with MPE and 40 with BPE, the AUC of Apo-E was 0.75, whilst sensitivity and specificity were 0.88 and 0.86, respectively, at 105 ng/ml threshold (55). Notably, since all MPE patients in this study were non-small-cell lung cancer (NSCLC), whether Apo-E may have similar diagnostic accuracy for MPE in patients with other etiology remains to be elucidated. Recently, Xue et al. enrolled a more representative patent cohort for investigating the diagnostic accuracy of Apo-E in MPE (56), obtaining a lower sensitivity (i.e., 0.79) and specificity (i.e., 0.74).

Chitinase-like proteins (CLPs)

CLPs play critical roles in various pulmonary and cardiovascular diseases such as asthma, chronic obstructive pulmonary disease (COPD) and HF (57,58). Citinase-3-like protein 1 (YKL-40) and Citinase-3-like protein 2 (YKL-39) are the most investigated CLPs. Some studies explored the diagnostic accuracy of YKL-40 and YKL-39 for MPE, but the results are almost contradictory. One study found that the AUC of YKL-40 for MPE was 0.90 (59), though such a good diagnostic accuracy could not be replicated in other studies. For example, the AUC of YKL-40 for MPE in the study performed by Kayhan et al. was only 0.77 (60). In other two studies, YKL-40 was not even significantly increased in MPE (61,62). The sources of heterogeneity among available studies were largely unknown. Taken together, these results would suggest that CLPs, including YKL-39 and YKL-40, may not be really useful diagnostic biomarkers in MPE, and this may be due to the fact that increased CLPs can be observed in many other diseases.

B7 family

The B7 family, which comprises a number of regulators of T-cell inhibition (63), has eleven members: B7-1 (CD80), B7-2 (CD86), B7-H1 (PD-L1), B7-DC (PD-L2), B7-H2 (ICOSL), B7-H3 (CD276), B7-H4 (VTCN1), B7-H5, BTNL2, B7-H6, and B7-H7. By binding to its ligand, the proteins of the B7 family regulate T-cell proliferation and cytokine release. Some members are present in two forms, soluble and membrane-bound. The former was proven to be a diagnostic marker in various types of cancer. Some studies also investigated the diagnostic accuracy of soluble B7 family members for MPE.

Chen (64) et al. explored the diagnostic accuracy of soluble B7-H3 (sB7-H3) for MPE in 52 patients with NSCLC-derived MPE and 47 with BPE (28 exudates and 19 transudates), reporting an AUC of 0.85 for sB7-H3. The values of sB7-H3 in serum and PE were found to be positively correlated (r=−0.784), thus suggesting that PE sB7-H3 may be passively diffused from blood to pleural cavity.

The diagnostic accuracy of soluble B7-H4 (sB7-H4) has also been investigated in two studies. In the first of these, encompassing 90 lung cancer patients with MPE and 58 with BPE (48 TPE and 10 PPE), the AUC of sB7-H4 was 0.862 (65). In the other study, Jing (66) enrolled a more representative cohort, consisting of 55 patients with MPEs due to various types of cancer and 42 with BPEs (25 TPEs, 7 PPEs, and 10 HFs), finally calculating an AUC of 0.884 for sB7-H4. A positive correlation was noted between the pleural values of sB7-H4 and CEA, thus arguing that sB7-H4 may provide incremental diagnostic value beyond CEA.

Matrix metalloproteinase (MMPs) and tissue-specific inhibitors of metalloproteinases (TIMPs)

The MMPs are a family of extracellular matrix (ECM) degradation enzymes, consisting of 23 members, whose activity is counterbalanced by the TIMPs (67). MMPs and TIMPs are involved in various physiological and pathological processes, including the development of PE.

Among all the various members of the MMPs family, MMP-9 is the most studied. Notably, MMP-9 values were found to be increased in MPE in some studies (34,68-72), whilst no variation was found in another (73). Some studies also concluded that MMP-9 be increased in TPE (74-77) or PPE (78), but not in MPE. The possible sources of heterogeneity among these investigations may be attributable to the assay used for measuring MMP-9, to different disease spectrum of the study cohort, as well as to heterogeneous study design.

Two studies investigated the diagnostic accuracy of MMP-3 in MPE (68,79). In one study, including 52 patients with malignant pleural mesothelioma (MPM), 33 with BPE, and 34 with lung cancer-induced MPE, MMP-3 was found to be increased in MPM, rather than in lung cancer-induced MPE (79). In another study, including 19 patients with lung cancer-induced MPE and 22 with BPE, MMP-3 was found to be increased in those with MPE (68). MMP-1 has also been studied in two studies, with conflicting results (73,80). One study found that MMP-1 was higher in TPE than in transudate and MPE (80), while the other did not observe significant differences (73).

In addition to MMP-9, MMP-1, and MMP-3, the diagnostic accuracy of MMP-2 (73,74,78), MMP-7 (71), MMP-8 (73), TIMP-1 (69,77), and TIMP-2 (77) has also been studied, and available results indicate that their diagnostic accuracy seems unsatisfactory, thus confirming that the diagnostic role of MMPs and TIMPs is probably very limited in MPE.

Tumor-associated macrophages (TAMs) and monocytes

There are many types of immune cells in the MPE environment, including lymphocytes, monocytes, and macrophages (81). Macrophages can be categorized into M1 and M2 types. M1 is a classically activated phenotype, and can produce pro-inflammatory and immunostimulatory cytokines that promote the clearance of tumor cell (81). M2 is instead an alternatively activated phenotype, which produces enzymes and cytokines suppressing the immune response against the tumor, thus potentially promoting cancer development (81). Two studies have investigated the diagnostic value of monocytes and macrophages in MPE.

In 2015, Wang et al. reported that the frequency of CD14+D163+ macrophages in PE of MPE is approximately 10-fold higher than that in BPE (82). The AUC of CD14+D163+ macrophages was considerably high (i.e., 0.941). At a threshold of 3.65%, the diagnostic sensitivity and specificity of CD14+D163+ macrophages were 0.86 and 1.00, respectively (82). In another study, the authors investigated the diagnostic accuracy of CD206+CD14+ macrophages for lung cancer-induced MPE (83), reporting an AUC as high as 0.98, with 0.88 sensitivity and 1.00 specificity at a 39.8% threshold.

Despite the results of these two studies are promising, their sample size was relatively modest (i.e., 60 and 100), whilst the diagnostic technique (i.e., flow cytometry) needs special training, is expensive, time-consuming, and poorly standardized.

Reactive oxygen species modulator 1 (Romo1)

Romo1 is a mitochondrial membrane protein involved in the production of intracellular reactive oxygen species (ROS) (84). It actively interplays with the development of various cancers, and increased serum Romo1 has been observed in lung cancer patients (85). Three studies have investigated the diagnostic accuracy of PE Romo1 for NSCLC-associated MPE so far (86-88), yielding AUCs between 0.71 and 0.84, and thus underpinning a modest diagnostic accuracy. Notably, the MPE group in these studies only included NSCLC patients, so that future investigations with various types of MPE would be needed to evaluate the diagnostic accuracy of this putative biomarker.

Epigenetic markers

mRNA

In addition to protein and cellular markers, epigenetic biomarkers such as cell-free nucleic acid have been evaluated in MPE. To date, many studies have investigated the diagnostic accuracy of cell-free mRNA for MPE, and the results are summarized in . The diagnostic accuracy of available cell-free mRNAs is variegated. Overall, it seems that LUNX mRNA (89), MN/CA9 mRNA (91), MUC1 mRNA (90), and EpCAM mRNA (90) are more promising because, since their AUCs were >0.90, and their sensitivity and specificity were also >0.90.

Table 2

Summary of studies investigating the diagnostic accuracy of cell-free mRNA for MPE

Author	Year	Country	Markers	Design	MPE/non-MPE	Non-MPE	AUC	Sensitivity	Specificity
Tang (89)	2013	China	LUNX mRNA	NR	106/103	PPE, TPE, HF, O	0.92	0.96	0.85
Sun (90)	2014	China	MUC1 mRNA	NR	58/40	PPE, TPE, HF	0.92	0.67	0.95
			EpCAM mRNA	NR	58/40	PPE, TPE, HF	0.92	0.71	0.95
Li (91)	2007	France	MN/CA9 mRNA	Prospective	59/12	NR	NR	0.90	0.92
Shu (92)	2007	China	hTERT mRNA	Prospective	41/55	PPE, TPE, HF	NR	0.80	0.95
Tang (93)	2015	China	Foxm mRNA	NR	23/15	PPE, TPE, HF	0.88	0.83	0.87
Li (94)	2013	China	p16 mRNA loss	Prospective	58/46	PPE, TPE, HF, O	NR	0.48	0.90
Bao (95)	2014	China	LUNX mRNA	Prospective	76/60	PPE, TPE, HF, LC, O	0.78	0.52	0.95
			VEGF mRNA	Prospective	76/60	PPE, TPE, HF, LC, O	0.80	0.68	0.95
Jiang (96)	2008	China	TTF-1 mRNA	NR	56/44	PPE, TPE	NR	0.73	1.00
Yu (97)	2001	China	MUC1 mRNA	Prospective	54/35	PPE, TPE	NR	0.65	0.96
			MUC5AC mRNA	Prospective	54/35	PPE, TPE	NR	0.72	0.96
Jeon (98)	2012	Korea	MAGE mRNA	Prospective	44/23	PPE, TPE, HF, O	NR	0.61	0.96
Li (99)	2012	China	XAIP mRNA	NR	56/42	PPE, TPE, HF, LC, O	0.76	0.66	0.86

NR, not reported; MPE, malignant pleural effusion; PPE, parapneumonic effusion; TPE, tuberculous pleural effusion; HF, heart failure; LC, liver cirrhosis; O, others; AUC, area under curve.

DNA methylation

Some studies investigated the diagnostic accuracy of DNA methylation in MPE. In a study performed by Katayama et al. (100) using methylation-specific polymerase chain reaction, the authors assessed the aberrant hypermethylation of DNA repair gene ras association domain family 1A (RASSF1A), O6-methylguanine-DNA methyltransferase (MGMT), p16INK4a, apoptosis-related genes, retinoic acid receptor b (RARb), death-associated protein kinase (DAPK). Among these five epigenetic biomarkers, DAPK was unrelated to MPE. The incorporation of the remaining four biomarkers into a logistic regression model yielded sensitivity and specificity of 0.60 and 0.79, respectively. Benlloch et al. (101) studied the diagnostic accuracy of these four epigenetic markers with a parallel diagnostic approach (at least one gene methylated in pleural fluid) and obtained 1.00 specificity, counterbalanced by 0.59 sensitivity. Some studies have also investigated the role of methylation of SHOX2 (102), SEPT9 (102), and WIF-1 promoter region (103) in MPE diagnostics. Although their specificity was as high as 1.00, sensitivities were almost unsatisfactory, thus raising doubts as to whether DNA methylation may be a useful diagnostic tool in MPE.

Cell-free microRNA

MicroRNAs are short non-coding RNAs that regulate gene expression at the post-transcriptional level (104). microRNAs are sufficiently stable in the circulation and have been suggested as potential diagnostic markers for various diseases, including cancer (105) and tissue injury (106,107). The first report concerning the diagnostic accuracy of pleural microRNA in MPE diagnostics has been published in 2010 (108). The authors measured the level of 22 cell-free miRNAs, which were found to be dysregulated in lung cancer patients’ serum. miR-24, miR-26a, and miR-30d were increased in MPE and ascites. The diagnostic accuracy of miR-24 and miR-30d in MPE was also evaluated (109), with AUCs of 0.71 and 0.75, respectively.

In 2013, a study exploring a miRNA array revealed that miR-98 was a possible diagnostic biomarker for lung adenocarcinoma-associated malignant pleural effusion (LA-MPE) (110). Unlike traditional cancer biomarkers, which are increased in PE, miR-98 was decreased in LA-MPE. The AUC of miR-98 for LA-MPE was 0.887, which is comparable to the diagnostic performance of CEA and CYFRA 21-1. A logistic regression model, incorporating CEA, CYFRA 21-1, and miR-98, displayed a considerably high AUC (i.e., 0.926), thus indicating that miR-98 can improve the diagnostic accuracy of CEA and CYFRA 21-1. In a subsequent study, PE miR-134, miR-185, and miR-22 were found to be decreased in LA-MPE patients (111). The AUCs of these three microRNAs were between 0.72 and 0.83, and their combination with CEA was found to improve the diagnostic accuracy of LA-MPE (AUC, 0.942). In 2018, a study based on 20 MPEs and 20 BPEs analyzed the diagnostic value of miR-21 and miR-24 in MPE. The AUCs of these two microRNAs were approximately 0.87 (112).

Exosomal microRNA

The source of cell-free microRNA in PE remains largely unknown. For circulating microRNA, three hypotheses have been proposed to explain their origin (113), including passive released by injured cells, active release in the form of a complex with proteins such as high-density lipoproteins (HDL), and release in extracellular vesicles (EVs) such as exosomes and microvesicles. It hence seems that these three hypotheses can also be used to explain the source of cell-free microRNA in PE. The cell-free microRNAs can be directly released into the pleural cavity by tumor cells or passively diffuse from blood to the pleural cavity.

Exosomes are EVs secreted by all types of cells, with a diameter between 30 and 100 nm (114). Because miRNA signatures in circulating exosomes are very similar to those originating from tumor cells (115), they appear to be more specific cancer biomarkers. Moreover, exosomal microRNAs seem more stable than the other two forms of cell-free microRNA, because their encapsulation within the exosome prevents their possible degeneration (116).

Two studies with microRNA array have addressed the differentially expressed exosomal microRNAs in MPE and BPE (117,118). Some microRNAs could be identified, such as miR-375 and miR-200. The diagnostic accuracy of exosomal microRNAs was assessed with receiver operating characteristics (ROC) curve analysis in the former investigation (118), and the AUCs of miR-375, miR-200b, miR-200c, and miR-141 were found to be al >0.90. The second study explored the diagnostic accuracy of exosomal miR-21, miR-31, miR-182, and miR-210 for MPE (119). Among these, the diagnostic accuracy of miR-182 (AUC, 0.87) and miR-210 (AUC, 0.81) was the most promising. Unfortunately, the MPE group in these three studies only included lung adenocarcinoma and the sample size was small, so that the diagnostic accuracy of exosomal microRNAs in MPE would need to be evaluated in future clinical investigations.

Tumor markers for MPM

PE is relatively common in MPM patients. It is reported that nearly 3% of undiagnosed PE are MPM, and PE can be observed in about half of MPM patients (120,121). Therefore, PE biomarkers represent a promising diagnostic tool for MPM. Soluble mesothelin-related peptides (SMRP), osteopontin, and fibulin-3 are the most widely studied biomarkers for MPM (122), and the diagnostic role of their serum or plasma concentrations have been explored in several meta-analyses (123-125). However, the diagnostic value of their PE concentrations remains largely unknown. Only two studies have investigated the possible diagnostic role of PE SMRP for MPM (126,127), providing contradictory findings. The sensitivity, specificity and AUC in one study were 0.79 (95% CI: 0.75–0.83), 0.85 (95% CI: 0.83–0.87) and 0.89 (95% CI: 0.85–0.93), respectively. In the second study, sensitivity specificity and AUC were 0.69 (95% CI: 0.64–0.72), 0.90 (95% CI: 0.85–0.94) and 0.76 (95% CI: 0.72–0.80), respectively. The proteomic technology also shows high diagnostic accuracy for MPM (128), with one study finding an AUC as high as 0.99. However, additional evidence would be needed to support these preliminary data.

Recent studies have also highlighted that cell-free microRNA in PE may be seen as a novel diagnostic marker for MPM. However, according to a recently published literature review (129), the diagnostic accuracy of PE microRNA is limited, and more studies will be needed to validate.

Conclusions

In this review, we have summarized several cancer biomarkers, whose assessment may be useful in MPE diagnosis. Some of these, such as cytotoxic T lymphocyte-associated antigen-4 (CTLA-4) (130), Her 2 (131), total sialic acid (TSA) (132), carbohydrate antigen 549 (CA 549) (133), tissue polypeptide-specific (TPS) (134), midkine (46), syndecan-1 and osteopontin (135), endothelial-cell-specific molecule-1 (ESM-1) (136) and heat shock protein 90β (HSP90β) (137) have been investigated in a limited number of studies, which also failed to provide encouraging results. TAMs seem to have the highest diagnostic accuracy, but their assessment is expensive, time-consuming, requires skilled personnel and test results are poorly standardized. Cancer ratio represents another promising parameter in MPE (48), but future studies would be needed to validate its diagnostic accuracy.

Notably, some of the above-mentioned studies may be plagued by a patients’ selection bias. For example, only lung cancer patients were enrolled in the MPE group, and thus the disease spectrum may not be representative enough of the common malignancies causing PE in clinical practice. Moreover, subjects’ enrollment in some studies was not consecutive, and this may have led to a selection bias (138).

We can hence conclude that the diagnostic accuracy of each single cancer biomarker is relatively modest so that it seems reasonable to suggest that a multi-marker strategy may be a much better approach in MPE diagnostics. Although the development of such algorithms is indeed challenging, artificial intelligence approaches could be an option (139,140). Further studies, such as SIMPLE (141) and DIAPHRAGM (5), are attempting to use this approach for improving the diagnostic accuracy of cancer biomarkers in MPE. In the promising era of targeted cancer treatment, therapeutic strategies are determined by therapy-related genetic alterations, such as EGFR mutation. Unfortunately, the current evidence does not support PE soluble markers for predicting the genetic alteration. Under such a circumstance, invasive biopsy using an appropriate specimen is unavoidable. Therefore, exploring soluble biomarkers that can predict the therapy-related genetic alteration is an attractive object for future studies.

The article’s supplementary files as