Mediators in pleural effusions of different origin: a two-step diagnostic study

Background

In recent publications we investigated the value of the determination of cytokines and proteases in the differential diagnosis of lung disorders by assessing pleural effusions of different origins. Increased levels of interleukin (IL)-6 and IL-8 but decreased levels of the soluble form of the interleukin 6 receptor (sIL-6R) were found in pleural effusion fluid of patients with lung cancer and malignant pleural effusions (CA) and could be used for differential diagnosis against effusions of tuberculous origins (TB) [1, 2]. In another study matrix metalloproteinases (MMP) and tissue inhibitors of metalloproteinases (TIMP) were examined. MMP-2, TIMP-1, and TIMP-2 concentrations were increased in CA pleural fluid vs. plasma, in contrast to MMP-9 being higher in plasma, but only pleural fluid MMP-1 and MMP-8 were significantly increased in CA vs. congestive heart failure (CHF) effusions [3]. In patients with TB pleural effusions compartmentalized MMP-1, MMP-2, TIMP-1, and TIMP-2 and, compared to CHF, a surplus of MMP-1, MMP-2, MMP-8, and MMP-9 in the pleural space was found [3]. The determination of vascular endothelial growth factor (VEGF) in pleural effusions of patients with CA, tumours with secondaries to the lung (TM), congestive heart failure (CHF), TB, and acute infections (INF) showed VEGF to be higher in effusions of patients with CA as well as TM in comparison to INF, TB, or CHF [4]. In another study we investigated immunoreactive concentrations of IL-4, IL-6, IL-11, IL-15, IL-17, IL-18, and tumor necrosis factor-α (TNF-α) in pleural effusions from patients with TB, CA, TM, CHF, and INF. Statistical analysis based on singular cytokines did not provide evidence for diagnostic relevance. However, fuzzy-logic analysis was able to assign patients to the correct diseases with an 80% accuracy using IL-6 and IL-15 measurements [4].

Various parameters have been studied by other groups for better differentiation of pleural effusions, including high molecular weight proteins, acute phase reactants, and cytokines [5]. TIMP-2 is considered as an interesting marker in malignant pleural effusions due to a primary tumour elsewhere in the body [6]. VEGF is another important mediator in exudative pleural effusions [7]. VEGF has been studied in human cancer cell lines and is considered to play a critical role in the production of malignant pleural effusions [8, 9]. VEGF was shown to be useful as an adjunct of the conventional algorithm in the diagnosis of malignant pleural effusions [10]. The comparison of VEGF with adenosine deaminase (ADA) and interferon-γ (IFN-γ) suggested that high amounts of pleural fluid IFN-γ may indicate an increased rate of residual pleural thickening in tuberculous pleurisy [11]. IFN-γ, IL-12p40, and IL-6 proved useful in differentiating tuberculous and malignant effusions [12]. C-reactive protein (CRP), IL-6, and TNF-α provided useful information for the differentiation of parapneumonic, tuberculous and malignant effusions in clinical practice [2]. Applying a decision tree analysis that contained age, temperature, pleural fluid ADA, and LDH was helpful in the differential diagnosis of tuberculous and malignant pleural effusions [13]. Finally, the determination of ADA, INF-γ, CRP, carcinoembryonic antigen (CEA), IL-6, TNF-α and VEGF in pleural effusions of different origin and in addition, the application of receiver operating characteristic curve (ROC) analysis and further statistical methods showed that the combination of ADA and CRP levels might be sufficient for discriminating between malignant, tuberculous and parapneumonic pleural effusions [14].

The combination of various parameters and the application of advanced statistical tools seemed therefore promising in achieving a higher diagnostic yield, improved diagnostic accuracy, and differential diagnostic capability. This observation encouraged us to expand and combine previous investigations. We therefore, in a first step, intended to find out whether the determination of the so far most promising parameters IL-6, IL-15, VEGF, and TIMP-2 would confirm previous results in a new set of patients. In a second step we intended to find out whether the use of a combination of these parameters would lead to a higher diagnostic yield.

Materials and methods

Patients were consecutively enrolled into the study (series I) after informed consent was obtained following the ethical standards of our institutional guidelines. In order to test the validity of parameters measured and the diagnostic accuracy obtained a second series of patients (series II) was examined. In series I altogether 38 patients (19 women, 19 men, age 67±19 years) and in series II altogether 50 patients (18 women, 32 men, age 64±13 years) were enrolled. Twelve patients with primary lung cancer and malignant pleural effusion (CA) in series I (six women, six men, age 75±8 years) and 16 patients in series II (five women, eleven men, age 66±8 years) were included; the diagnosis was based on radiological, clinical, histological, and/or cytological evidence, and on malignant cells in pleural fluid or pleural biopsy specimens. Seven patients with secondaries to the lung from a primary tumour (two breast, two gastric, two colon and one kidney cancer; TM) in series I (six women, one man, age 69±19 years) and 14 patients in series II (nine women, five men, age 60±10 years) were included; the diagnosis was based on radiological, clinical, histological, and/or cytological evidence, and on malignant cells in pleural fluid or pleural biopsy specimens. Twelve patients with tuberculous pleurisy (TB) in series I (eight women, four men, age 52±24 years) were included, in series II no TB cases were available; the diagnosis was based on the clinical and radiological appearance, microbiological proof of acid fast bacilli and/or growth of M. tuberculosis in sputum or pleural effusion fluid, or on histological proof of granuloma formation in pleural biopsy specimens. Two patients with pneumonia and parapneumonic effusions (PNEU) in series I (two men, age 76±29 years) and seven patients in series II (two women, five men, age 62±25 years) were included; the diagnosis was based on radiological, clinical, and/or microbiological evidence. Finally, five patients with congestive heart failure and pleural effusion (CHF) in series I (three women, two men, age 76±9 years) and 13 patients in series II (two women, eleven men, age 69±12 years) were included; the diagnosis was based on radiological and clinical findings, the transudate nature of the effusion, and exclusion of other etiologies.

Pleural effusion samples were collected during thoracocentesis performed for diagnostic or therapeutic purposes. Plain tubes were used to collect samples for determination of lactate dehydrogenase (LDH) and total protein by standardised automated methods; Light’s criteria were calculated as described before [15]. EDTA tubes (final concentration: 1 mg/mL) were used to collect and store samples at –70°C for later determination of VEGF, IL-15, IL-6, and TIMP-2 by Enzyme Linked Immunosorbent Assays (ELISA) using commercially available kits (human VEGF, IL-15: Quantikine™ Immunoassays, R&D Systems, Wiesbaden, Germany; human IL-6: Biosource International, Camarillo, CA, USA; TIMP-2: Biotrak™, Amersham, Freiburg, Germany) as described before [1, 4, 15]. The lower detection limits were: 5 pg/mL (IL-6), 1 pg/mL (IL-15), 3.13 ng/mL (TIMP-1) and 2.5 pg/mL (VEGF).

The results of series I and II were used for determination of differential diagnosis accuracy. Box-plot-format was chosen to present data (boxes show median and 25th/75th percentiles, horizontal bars represent 5th/95th percentiles). Diagnostic characteristics were calculated as ROC curves by an Excel application (ROC-tools, details see http://www.acomed-statistik.de). Per study, the occurrence of tumour disease (yes/no) in dependence on all markers was investigated in logistic regression models (complete models without selection processes).

Results

The results of series I in comparison to series II show differences to an unexpected degree (Figure 1). The values obtained in series II can no longer differentiate the entities examined as sharply as in series I. For VEGF lower values for CA, TM, and PNEU were obtained in contrast to higher values for CHF effusions in series I. As for IL-15, a totally deviating level of results was obtained with barely detectable values in series II. The values of TIMP-2, however, were at a higher level in series II than in series I for all entities examined. Only for IL-6 the distribution pattern of results in series I was comparable to that obtained in series II. Based on these observations, the diagnostic accuracies of IL-6 and to a lesser degree of VEGF seem both high when comparing the two series. The diagnostic accuracy of TIMP-2, however, seems low, while that of IL-15 was only minimal.

Figure 1

Levels of VEGF (A), IL-15 (B), IL-6 (C), and TIMP-2 (D) in pleural effusion fluid separated by disease entities.

For each parameter two values are given: series I and II. Circles and stars indicate outliers. CA, primary lung cancer and malignant pleural effusion; TM, secondaries to the lung from a primary tumour; TB, tuberculous pleurisy; PNEU, pneumonia and parapneumonic effusions; CHF, congestive heart failure and pleural effusion.

Table 1 shows the results for the ROC curves. Especially, the results of series II demonstrate a diagnostic potential for IL-6 and VEGF. The data of VEGF and IL-6 were therefore used to create a classifying tool by applying the method of logistic regression. TB cases were excluded, as samples were only available in series I. The classifying tool calculated in this way based on the data of series I shows a high degree of diagnostic accuracy (AUC=0.94); whereas the calculation based on the data of series II does not (AUC=0.56). This could further be demonstrated by presenting the classification values (values >0 and <1) in a box-plot format (Figure 2A). It should be noted, however, that the number of values for development and for testing was rather low (Table 2). Therefore, it seemed appropriate to use series II as the development data base and series I as the testing data base, with the result of an AUC=0.866 (development) and 0.78 (testing), respectively. The results shown in box-plot format are depicted in Figure 2B.

Figure 2

Classification values presented in box-plot format.

Calculation Figure 2A based on the data of series I, and classification values in Figure 2B were developed on the basis of series II. The appearance is more homogenous in comparison to the presentation in Figure 2A. CA, primary lung cancer and malignant pleural effusion; TM, secondaries to the lung from a primary tumour; PNEU, pneumonia and parapneumonic effusions; CHF, congestive heart failure and pleural effusion.

Combining data, VEGF is the best parameter discriminating between underlying pathologies (Figure 3).

Figure 3

ROC analysis of cytokines for detection of malignant pleural effusions in comparison to non-malignant ones.

VEGF is shown to be superior to other parameters IL-15, IL-6 or TIMP-2.

Discussion

Results of numerous publications in pulmonology and immunology describing cytokines and other markers in pleural effusions suggest a diagnostic benefit of such parameters in patients’ care. A multitude of observations is available. However, it seems important not only to confirm previous findings but, in addition, to search for tools improving the diagnostic accuracy based on results already obtained. We therefore decided to investigate IL-6, IL-15, TIMP-2, and VEGF in pleural effusions repeatedly and evaluated results by application of expanded calculatory methods. We aimed 1) to confirm pre-existing data and 2) to check whether a combination of these data could improve the process of differential diagnosis.

This study shows that IL-6 has a high diagnostic accuracy when determined repeatedly in different series of patients. IL-6 has a wide spectrum of biological activities and was found to be increased in malignant pleural effusions of patients with pulmonary carcinoma and also tuberculosis [16–19]. IL-6 also seems to play an important anti-inflammatory role in pleural effusions, breaking with the traditional idea of IL-6 being a pro-inflammatory acute phase reaction cytokine only [20]. IL-6, IL-8, and VEGF also play a role in the pathogenesis of prenatal chylothoraces in human fetuses [21]. The diagnosis of tuberculous pleural effusions by determination of CRP, IL-1α, IL-6, tumour necrosis factor (TNF)-α, and interferon (IFN)-γ has been the focus of numerous studies [2, 22, 23].

This study further shows that IL-15 has a low diagnostic accuracy, with deviating results when tested in different series of patients. This deviation was considered as not being related to the methods or tests used but rather to the unstable nature of IL-15 itself. Recent studies report the occurrence of IL-15 in autoimmune as well as in malignant diseases, including pleural effusions [24–26]. IL-15 is an interesting parameter and should be studied further, however, seems not suitable as a candidate to improve diagnostic accuracy in the differential diagnosis of pleural effusions.

This study demonstrated that the determination of TIMP-2 is of low diagnostic accuracy. Although MMP and their endogenous inhibitors TIMP play an important role in the homeostasis of the pleural space in various disease states, their contribution to differential diagnosis of pleural effusions of different origins seems limited [27, 28].

VEGF demonstrated a high diagnostic accuracy in this study similar to that of IL-6. VEGF has potent angiogenic, mitogenic, and vascular permeability enhancing properties specific for endothelial cells, was found at high levels in exudative inflammatory and neoplastic pleural effusions [28].

A limitation of this study was that TB cases were not available in both series. The results for TB could therefore only demonstrate that the respective parameters could be detected in the entities studied. The classifying tool could only be used when TB cases were excluded. With regards to the calculation of the classifying tool the number of values for development and for testing was low. In general, 10 to 20 values are necessary per group and per parameter. The study demonstrates that pilot studies for investigation of diagnostic potential of biomarkers as currently widely undertaken in terms of “individualized medicine” might be of limited validity and need confirmation in independent trials.

It should be noted and has been shown strikingly by our results that discriminative power of biomarkers highly depends on clinical question, compared groups, and patients’ entities. Inhomogeneous sample distribution in different studies makes comparisons very hard.

In summary, VEGF was not only confirmed as a valid marker for malignant diseases. Furthermore it was shown that the results of VEGF and IL-6 could be used to create a classifying tool. The diagnostic relevance of IL-15 could not be confirmed in the second dataset. In studies based on a small case numbers, however, such effects can be expected. The previously proposed relevance of TIMP-2 as a diagnostic parameter could, similarly to IL-15, not be confirmed when the two series were compared. Although its functions stay controversely discussed, VEGF stays the best suitable diagnostic parameter [29].

Conclusions

VEGF as well as IL-6 are both parameters demonstrating a diagnostic potential for differentiation between non-malignant effusions (e.g., CHF, parapneumonic) vs. malignant effusions (e.g., lung cancer and secondary cancer type effusions). Both parameters could be used in combination. Whether an accuracy of AUC=0.8 with a sensitivity of 75% and a specificity of 75% is clinically relevant or not depends on the clinical situation of the individual patient. However, this degree of diagnostic accuracy is comparable to that of other known tumour markers. IL-15 and TIMP-2 demonstrated detrimental results in the two consecutive series and were therefore not considered to be useful for differential diagnosis.