Introduction
Malignant mesothelioma (MM) is a tumor that develops from the serous membranes that line the body cavities and it may arise in the pleura, peritoneum and pericardium; in addition, although extremely rare, it may also develop in the tunica vaginalis testis. The most common form of this disease is the malignant pleural mesothelioma (MPM). MM was previously considered as being extremely rare; however, its incidence and associated mortality rate exhibited a sharp increase worldwide over the last 50 years, due to the close association of MM with asbestos exposure. The prognosis of MPM is poor, with a median survival of ∼9–17 months (1). However, in selected patients with epithelioid tumor histology, early-stage disease, who undergo trimodality treatment (combination of chemotherapy, postoperative radiotherapy and extrapleural pneumonectomy), median overall survival of 51 months and 5-year survival rates of 46% have been reported (2). Recent phase II trials reported a median survival of ∼30 months for the patients who completed the trimodality treatment (3,4). Therefore, early diagnosis may play a vital role in the improvement of therapeutic outcomes. Together with the advances in imaging studies and endoscopic examinations, the development of biomarkers useful for serum or effusion diagnosis is crucial for the early diagnosis of MM. Currently known biomarkers for diagnosing MM include cytokeratin 19 fragment (CYFRA) (5–7), tissue polypeptide antigen (TPA) (5,6,8), hyaluronic acid (8), carbohydrate antigen (CA125) (8,9) and osteopontin (10–15). However, these markers have low specificity for MM.
Mesothelin is a 40-kDa cell surface glycoprotein that is overexpressed in cells of pancreatic and ovarian cancer, mesothelioma and other malignancies. The mesothelin gene encodes a 69-kDa glycoprotein, the mesothelin precursor protein, which is cleaved by a furin-like protease and its N-terminal region is released in the blood as a 31-kDa protein, the megakaryocyte potentiating factor (MPF). The 40-kDa C-terminal region of this glycoprotein binds to the cell membrane as mesothelin. Three distinct variants of mesothelin have been identified, one of which has a modified C-terminus and becomes detached from the cell membrane since it lacks a glycosylphosphatidylinositol (GPI) anchor. This soluble isoform corresponds to the soluble mesothelin-related peptide (SMRP) (16). SMRP and MPF may be highly specific biomarkers for MM and have an equivalent diagnostic performance (17–19). SMRP is currently the most extensively investigated and is considered to be the best available blood protein biomarker of MM (20).
However, the diagnostic performance of SMRP alone is not considered to be sufficiently high, as it appears to exhibit insufficient sensitivity for MM (20,21). In diagnosing malignant tumors, such as ovarian or prostate cancer, the diagnostic performance of individual serum biomarkers was improved by combining data obtained using multiple biomarkers (22,23).
In the present study, we evaluated the performance of serum SMRP levels in the diagnosis of MM and investigated whether its diagnostic value could be improved through its combination with other biomarkers.
Materials and methods
Study design
The subjects of this study were patients who satisfied the following inclusion criteria: i) age ≥20 years; ii) pathologically proven MM or lung cancer; and iii) except for ii), individuals with asbestos exposure proven on the basis of their history or from the medical viewpoint. Only patients who personally provided written informed consent for the measurement of their serum biomarkers were enrolled in this study. Subjects who satisfied the above inclusion criteria during the study period were retrospectively enrolled. The pathological diagnosis was based on standard histological and immunohistochemical criteria (24,25). The subjects were classified into three groups: individuals with a history of asbestos exposure, patients with lung cancer and patients with MM. This study was approved by the Institutional Review Board of the Hyogo College of Medicine.
Measurement of serum biomarker levels
At the time of confirmation of the diagnosis, blood samples were collected from the subjects and, following prompt separation of the serum, the samples were stored at −80°C. The serum SMRP levels were measured using an ELISA kit (Mesomark™; Fujirebio Diagnostics Inc., Malvern, PA, USA) according to the manufacturer’s instructions. The serum levels of CYFRA and carcinoembryonic antigen (CEA) were measured using commercially available immunoassay systems according to the manufacturer’s instructions: the serum CEA levels were determined using a chemiluminescent immunoassay (Abbott Japan Co., Ltd., Tokyo, Japan) and the serum levels of CYFRA were determined using a solid-phase sandwich immunoradiometric assay (CIS Bio International, Gif-sur-Yvette, France). The manufacturer suggests 3.5 ng/ml for CYFRA and 5.0 ng/ml for CEA as the cut-off values to differentiate between non-malignant disease and malignant tumors.
Statistical analysis
Summary statistics were used (median and 25th and 75th percentiles) to evaluate the distribution of serum SMRP levels. The Steel’s test, a non-parametric form of the Dunnett’s test, was used for comparing MM to the other groups. The sensitivity and specificity of SMRP for diagnosing MM were calculated, along with the corresponding 95% exact confidence intervals (CIs). The above analyses were also performed for CYFRA and its performance was compared to that of SMRP by using the McNemar’s test. To compare the serum SMRP levels between each histological subtype of MM, the Steel-Dwass test, a non-parametric form of the Tukey’s test, was performed. Subsequently, a stepwise logistic regression analysis was used to select marker combinations (MCs) that were more effective for diagnosing MM. The criterion for assessing whether a difference was significant in the variable selection was 5%. The diagnostic performance of SMRP and the MC was assessed by constructing a receiver operating characteristic (ROC) curve and calculating the area under the curve (AUC). The AUC for SMRP and that for the MC were compared using the theory on generalized U-statistics to generate an estimated covariance matrix and the χ2 test (26). For each test, two-sided P<0.05 was considered to indicate a statistically significant difference. Data were analyzed using the statistical software SAS, version 9.1.3 (SAS Institute Inc., Cary, NC, USA) and Stata, version 11.0 (StataCorp College Station, TX, USA). The GraphPad Prism software, version 4.00 for Windows (GraphPad Software, San Diego, CA, USA) was used to prepare the figures.
Results
Patient characteristics
A total of 190 subjects were enrolled in this study. A summary of the clinical characteristics of these subjects, together with a breakdown of each group by age, gender, history of asbestos exposure and presence of effusion (pleural or peritoneal) is presented in Table I. Among the 39 individuals with asbestos exposure, pleural plaque was present in 16, benign asbestos pleurisy in 7, asbestosis in 3 patients, asbestosis plus benign asbestos pleurisy in 5, round atelectasis in 2 and no imaging abnormalities in 6 patients. The histological subtype in the 55 patients with lung cancer was adenocarcinoma in 24, squamous cell carcinoma in 14 and small-cell carcinoma in 17 patients. Among the 96 patients with MM, the primary tumor site was the pleura in 91 and the peritoneum in 5 patients (Table II). The histological subtype was epithelioid in 57 patients, sarcomatoid in 12, biphasic in 6, desmoplastic in 4 and unspecified in the remaining 7 patients (Table II). Of the 91 patients with MPM, 74 were diagnosed with clinical stage IV disease according to the staging classification proposed by the International Mesothelioma Interest Group (IMIG). Only 5 patients had either stage I or II disease (Table II).
Performance of serum SMRP in diagnosing MM
Fujirebio Diagnostics, Inc., the developer of the Mesomark assay, recommends a cut-off value of 1.5 nM, which was the 99th percentile of the normal serum SMRP concentration in a population of 409 healthy Americans (27). An investigation in a population of healthy Germans revealed a cut-off value of 1.5–1.6 nM, which was the 95th percentile of the serum SMRP concentration (28). In our study, we performed a preliminary investigation of the distribution of serum SMRP levels among 72 healthy individuals without a history of asbestos exposure. Since this investigation revealed that 69 individuals (96%) had serum SMRP levels of <1.5 nM, we selected 1.5 nM, the 96th percentile, as the cut-off value.
The distributions of serum SMRP levels in each group are shown in Fig. 1. The serum SMRP levels in MM patients were significantly higher compared to those in the other groups (P<0.001) (Table III). The sensitivity of SMRP for diagnosing MM was 56% (95% CI: 46–66%) and its specificity for MM vs. lung cancer and individuals with asbestos exposure was 87% (95% CI: 76–95%) and 92% (95% CI: 79–98%), respectively (Table IV). By contrast, the sensitivity of CYFRA for diagnosing MM was 63% (95% CI: 52–72%) and its specificity for MM vs. lung cancer was 49% (95% CI: 35–63%) (Table IV). The sensitivity of SMRP and CYFRA did not differ significantly (P= 0.157), although the specificity of SMRP for MM vs. lung cancer was significantly higher compared to that of CYFRA (P<0.001). The serum SMRP levels in epithelioid disease [median, 2.47 nM; interquartile range (IQR): 0.97–4.86] were significantly higher compared to those in sarcomatoid disease (median, 0.8 nM; IQR: 0.38–1.15) (P=0.04). However, there were no significant differences when compared to the other histological subtypes. There was no significant association between the serum SMRP levels and MPM stages (data not shown).
The diagnostic performance of SMRP was evaluated using ROC curves (Fig. 2). For the differentiation between MM and lung cancer, the AUC was 0.76 (95% CI: 0.68–0.83) (Fig. 2A) and for the differentiation between MM and individuals with asbestos exposure, the AUC was 0.78 (95% CI: 0.71–0.86) (Fig. 2B). For CYFRA, the AUC for the differentiation between MM and lung cancer was 0.55 (data not shown). Therefore, the diagnostic performance of SMRP for differentiating between MM and lung cancer was superior to that of CYFRA.
Investigation of MCs and their performance in diagnosing MM
To improve the performance of serum biomarkers in diagnosing MM, we investigated the optimal MCs. The measured variables common to patients with MM and lung cancer were age, gender, presence of effusion, clinical stage and the levels of SMRP, CYFRA and CEA. The measured variables common to patients with MM and individuals with a history of asbestos exposure were age, presence of effusion and the levels of SMRP, CYFRA and CEA. Since the distributions of all the biomarkers were significantly skewed to the right, the variables were logarithmically transformed using common logarithms. A stepwise logistic regression analysis was used to select the variables. To differentiate between MM and lung cancer, SMRP levels, presence of effusion and CEA levels were selected (Table V). From the signs of the estimates, we determined that the probability of a diagnosis of MM was higher for elevated SMRP levels, presence of pleural effusion and lower CEA levels. It was concluded that the selected markers were reasonable from the clinical standpoint. Subsequently, the markers selected to differentiate between MM and individuals with a history of asbestos exposure were age and CYFRA (data not shown). However, this model was composed of a single marker rather than multiple markers. Therefore, it was excluded from further investigation.
To further evaluate the models in Table V, the association between SMRP and CEA was analyzed using scatter diagrams (Fig. 3). The scatter diagrams demonstrated that the majority of patients with high CEA levels were those with lung cancer. In addition, the majority of patients with high SMRP levels were those with MM. Therefore, the combination of SMRP and CEA resulted in only a minor overlap of the diagnostic findings of MM and lung cancer, suggesting that the diagnostic performance for MM was improved. By contrast, since the combination of SMRP and CYFRA resulted in a significant overlap of the diagnostic findings of MM and lung cancer, it was inferred that the diagnostic performance was scarcely improved (data not shown).
The MC was composed using the results of Table V. Since the ratio of the estimates for SMRP, presence of effusion and CEA was ∼3:1:5, the following MC was selected: MC=1×I(presence of effusion) + 3 × log10(SMRP) − 5 × log10(CEA), where I (presence of effusion) was defined as an indicator function with a value of 1 when effusion was present and 0 when effusion was absent. Wherein -1 was selected as the cut-off value to maximize the sum of the sensitivity and specificity, the sensitivity of MC for diagnosing MM was 76% (95% CI: 64–85%) and its specificity for MM vs. lung cancer and individuals with asbestos exposure was 88% (95% CI: 74–96%) and 90% (95% CI: 68–99%), respectively. While the specificity of MC was comparable to SMRP alone, its sensitivity was ∼20% higher compared to that of SMRP alone. In addition, three of the five MPM patients with stage I–II disease were above the cut-off value, although none exhibited elevated serum levels of SMRP alone. The ROC curves for MC are shown in Fig. 2. The AUC for the differentiation between MM and lung cancer was 0.92 (95% CI: 0.88–0.97), which was significantly higher compared to that for SMRP alone (P=0.0001) (Fig. 2A). The AUC for the differentiation between MM and individuals with a history of asbestos exposure was 0.93 (95%CI: 0.87–1.0), which was also significantly higher compared to that for SMRP alone (P= 0.0058) (Fig. 2B). These results indicate that combining CEA with SMRP improves the performance of SMRP alone in diagnosing MM and may facilitate early detection of MPM.
Discussion
The recent development of Mesomark, a quantitative ELISA kit using two monoclonal antibodies (OV569 and 4H3) that recognize SMRP, has enabled the measurement of serum SMRP levels. The findings of key studies on the performance of SMRP in diagnosing MM by using the Mesomark kit demonstrated that serum SMRP levels were significantly higher in MM patients compared to those in controls, such as healthy individuals, subjects with a history of asbestos exposure, or patients with asbestos-related benign pleural disease or lung cancer (9,11–21,27–35). In the present study, also undertaken using the Mesomark kit, the serum SMRP levels were found to be significantly higher in MM patients compared to those in lung cancer patients and individuals with asbestos exposure. These findings are consistent with those first reported by Robinson et al (36), suggesting that the use of serum SMRP levels for diagnosing MM has excellent universality and reproducibility. Based on previous studies, including our own, SMRP is considered to be a highly specific biomarker for MM; however, its sensitivity, ranging from 48–80%, is moderate (9,11–21,27–35). To improve the performance of SMRP in diagnosing MM, there is a need to increase the sensitivity while maintaining a high degree of specificity.
One way of improving the sensitivity may be by lowering the cut-off value; however, this is not recommended, since it may result in a simultaneous reduction of specificity (26,28). Another approach may be to improve the diagnostic performance by combining data obtained using multiple biomarkers. The accuracy of the histopathological diagnosis of MM has markedly improved. One reason for this improvement has been the introduction of immunohistochemical analysis involving the combination of a positive marker that is highly expressed in MM and a negative marker that has a low frequency of expression in MM (37,38). A systemic review of markers for diagnosis of MM demonstrated that positive staining for CEA and epithelial antigen (clone Ber-EP4) and negative staining for epithelial membrane antigens and calretinin may confirm that a patient does not have MM (21). In addition, based on biomarker measurements in the pleural effusion, algorithms for the diagnosis of malignant pleural diseases were established. The CEA level achieved a greater accuracy in the differential diagnosis of MPM through its combination with other markers. For example, an elevated CYFRA level with a low CEA level in pleural effusion was shown to be highly suggestive of MPM (7).
To date, whether the combination of blood biomarkers, including SMRP, is able to improve the performance of SMRP alone in diagnosing MM remains controversial. A previous study by van den Heuvel et al (34) reported that the combination of two serum markers (CEA and SMRP) was the most accurate in differentiating MPM from non-small-cell lung cancer. The AUC of this marker combination demonstrated a significant improvement compared to the inverse levels of CEA alone. However, in that study, a direct comparison of diagnostic performance between this combination and SMRP alone was not performed.
Amati et al (31) evaluated the combination of two hematological biomarkers: 8-hydroxy-2′-deoxyguanosine (8-OHdG), an indicator of oxidative DNA damage and vascular endothelial growth factor β (VEGFβ), an angiogenic molecule. The results of that study indicated that the diagnostic performance of this combination in differentiating between healthy individuals and those with a history of asbestos exposure was superior to that of each biomarker alone. Although it was also mentioned that a combination of SMRP, 8-OhdG and VEGFβ was optimal for distinguishing between individual groups, including the MM group, that study provided no specific measures of diagnostic performance or any further details.
Several previous studies evaluated the diagnostic performance of combined SMRP and osteopontin measurements in MM. Creaney et al (12) demonstrated that the combination of SMRP, serum osteopontin and MPF did not exhibit increased sensitivity for detecting MM compared to that of SMRP alone. A recent study investigated serum SMRP and plasma osteopontin levels in 66 patients with MPM, 47 patients with non-malignant asbestos-related lung or pleural diseases, 42 patients with other benign pleural and lung diseases and 21 patients with lung cancer, as plasma osteopontin was proven to be more stable compared to serum osteopontin (14). A logistic regression analysis revealed that the combined marker model had an AUC of 0.912 and a sensitivity of 76%, with a 95% specificity (14). The AUC for this marker combination did not differ from that for serum SMRP alone. In previous studies, the majority of osteopontin-positive MM patients were also found to be positive for SMRP. This high degree of concordance may result in the finding that a combination of these two markers does not improve the performance of SMRP alone in diagnosing MM (12,14). Cristaudo et al (15) also measured serum SMRP and plasma osteopontin levels in 93 healthy subjects, 111 individuals with benign respiratory disease and 31 patients with MPM. That study was the first to demonstrate that a combination of these two markers was more efficient in MPM diagnosis compared to each marker used alone by means of the combined risk index, a new statistical approach of a logistic regression analysis. In that study, however, a small number of patients with MPM were enrolled and its histological subtype was limited to the epithelioid type. To confirm those findings, larger-scale studies are required. The combination of SMRP with CA125 (9), or MPF (12,18) has also been investigated. However, none of those studies demonstrated that the diagnostic performance of SMRP in combination with other markers outperformed that of SMRP alone.
The present study demonstrated that combining SMRP and CEA improved the diagnostic performance of SMRP alone, since these two markers act in a complementary manner. However, since we used the same data for selecting and assessing the performance of MC, it is possible that our evaluation of the MC may have been optimistic. Furthermore, in our study, data were collected from a single center; validation of the diagnostic performance of this particular MC by a multicenter study is recommended in the future.
It is difficult to determine whether pleural effusion developing in individuals with a history of asbestos exposure represents benign asbestos pleurisy or is an initial symptom of MPM and misdiagnosis at this stage may hinder the early detection of MPM. Future prospective research is required to confirm whether a combination of serum biomarkers, including SMRP, may be useful in diagnosing early-stage MPM.