Abstract
Background/Aim: Pleural effusion (PE) has a heterogeneous aetiology, and differential diagnosis between benign and malignant disease may require invasive procedures in up to 60% of cases. The sensitivity of pleural cytology is limited, and several strategies have been tested to reduce the need of invasive diagnostic approaches. The aim of this study was to evaluate the usefulness of pleural fluid cytology, compared to, and combined with, carcinoembryonic antigen (CEA), C reactive protein (CRP), and lactate dehydrogenase (LDH) assay of pleural fluid (PF) in patients with a history of cancer, exudative non-purulent PE, and suspicion of malignant PE on imaging studies. Patients and Methods: The medical records of 40 patients with pulmonary metastases and malignant PE, and 57 controls with benign exudative PE were reviewed. All the patients underwent pleural cytology and CEA, CRP, and LDH assay before VATS-guided biopsy. Results: The sensitivity and specificity were 55.0% and 98.2% (cytology), 35.0% and 98.2% (CEA), 92.5% and 71.9% (CRP), 70.0% and 54.4% (LDH). The multivariate analysis excluded LDH, and the final AUC (cytology+CEA+CRP) was 0.894. Conclusion: In all patients with a history of cancer and PE of uncertain origin, the combination of PF cytology plus pleural CEA and CRP assay together should be suggested to recognize malignant plural effusion (MPE), minimising the use of unnecessary invasive investigations.
Malignant pleural effusion (PE) is common in patients with several advanced malignancies, including lung cancer and pulmonary metastases (PMs), usually from breast and ovarian cancer, while non-malignant pleural exudates are relatively uncommon (1, 2). However, the differential diagnosis between benign and malignant PE becomes important especially when a metastatic disease is suspected. In any case, in all patients with PE the evaluation of pleural fluid (PF) specific gravity, pH, differential cell counts, glucose, protein and lactate dehydrogenase (LDH) is routinely performed, and the Light's criteria are usually applied to differentiate exudates from transudates (1, 3). Unfortunately, the reported diagnostic accuracy of such approaches remains modest to confirm (or rule out) the diagnosis of malignant PE (MPE), also testing further biochemical or tumour markers, including C-reactive protein (CRP) and carcinoembryonic antigen (CEA). Moreover, compared to fine-needle aspiration samples of solid tissues, PF cytology typically transfers fewer cells, offering high specificity (near 100%), but low sensitivity (50-60%), and thus more invasive investigations are usually required (4).
The aim of the study was to evaluate the usefulness of PF cytology, compared to, and combined with, CEA, CRP, and LDH assay of PF in patients with a history of cancer, exudative PE, and the suspicion of PMs on imaging studies (high-resolution CT scan).
Results of univariate and multivariate analysis (95% CI).
Patients and Methods
The medical records of hospitalised patients with non-purulent pleural exudate, who underwent video-assisted thoracoscopic (VAT) thoracentesis and VAT-guided biopsy, were retrospectively reviewed. Part of these data have already been analysed for different purposes, evaluating talc pleurodesis for palliation of patients with MPE (5).
The final diagnosis, based on histopathology of VAT specimens, showed 40 patients with malignant PE (cases) due to PMs, mainly from breast cancer (n=18, 45%) and urinary tract cancer (n=8, 20%). The controls were 57 sex- and age-matched randomly selected patients, with histologically confirmed benign PE. Overall, there were 56 (58.7%) males and 41 (42.3%) females (median age=71 years; range=28-86 years). Patients with transudative (e.g., heart or renal failure, liver cirrhosis) or indeterminate PE according to the Light's criteria, as well as those with relapse of PE after previous thoracentesis, were excluded from the study. After giving their informed consent, all patients (n=97) underwent PF cytology, and complete standard analysis (specific gravity, pH, glucose and protein content) of the PF, including CEA, CRP, and LDH assay. Pleural adenosine deaminase (ADA) was not routinely measured. Samples of at least 40-45 ml of PF were obtained, and a standard staining method for cytological examination was used, as previously reported (6). Inadequate or unclear specimens were repeated (when possible), and when still inconclusive were definitively considered a false-negative result. CEA, CRP, and LDH were measured using a chemiluminescent (CLIA) immunoassay (sandwich CLIA with native CEA coated to nanomagnetic microbeads), a human sandwich (quantitative) enzyme-linked immunosorbent assay (ELISA), and a colorimetric quantification assay, respectively (7).
Assuming that the data were not normally distributed, the non-parametric Mann-Whitney U-test was used to evaluate the statistical significance of continuous variables between the groups. In the univariate analysis, the parameters (independent variables) significantly related to the presence of a malignant PE (dependent variable) were used in the forward stepwise multivariate logistic regression analysis (including the parametric statistical Wald test), useful in constructing a predictive model. Odds ratio (OR) estimates and associated 95% confidence intervals (CI) were calculated. The receiver operating characteristic (ROC) curve was used to represent the diagnostic accuracy of the model, and the area under the curve (AUC) was also obtained. The AUC is a single scalar value that measures the overall performance of a binary classifier (8). The higher the AUC (near to the 1), the better the model is in distinguishing between patients with malignant or benign PE. A p-value <0.05 was considered statistically significant. Statistical calculations were performed with the Statistica software (StatSoft, Tulsa, OK, USA) (version 2012). A written consent was obtained from all patients for both the clinical procedures and the study. Because the present study is a retrospective review of anonymized clinical records, ethical permission was not required.
Results
The age of patients (68.025±10.169 vs. 68.018±14.292 years; p=0.998) and the gender distribution (p=0.172) were not statistically different. Also specific gravity (1,026.9±6.3 vs. 1,027.4±5.5; p=0.702), pH (7.605±0.167 vs. 7.588±0.202; p=0.652), glucose (105.5±35.1 vs. 107.2±41.6 mg/dl; p=0.835) and protein (4.04±1.08 vs. 4.23±1.11 g/dl; p=0.386) content of PF did not differ (p=not significant) between groups. A significant difference between pleural levels (cases vs. controls) of CEA (37.2±100.3 vs. 1.6±1.4 ng/ml; p=0.0084), CRP (5.8±8.5 vs. 11.9±7.4 mg/l; p=0.0008), and LHD (414.6±327.2 vs. 250.6±222.2 U/l; p=0.0017) was found. The optimal cut-off values used to dichotomise the results and obtain categorical variables were the following: CEA=5 ng/ml (sensitivity=35.0%, specificity=98.2%), CRP=8 mg/l (sensitivity=92.5%, specificity=71.9%), LDH=180 U/l (sensitivity=72.5%, specificity=52.6%). The results of univariate analysis for each significantly useful diagnostic tool (pleural cytology, CEA, CRP, LDH) and the subsequent logistic regression, which excluded LDH (p=not significant), are shown in Table I. Figure 1 shows the ROC curve of the final predictive model obtained with the combination of all the three tools. The complete function correctly classified 82.5% of the population, and the relative AUC was 0.894 (95% CI=0.830-0.958).
ROC (receiver operating characteristic) curve and AUC (area under the ROC curve) of the proposed predictive model.
Discussion
Pleural effusion is a relatively common disease, with an incidence of more than 3,000 cases/million in the general population. Aetiological diagnosis is crucial to determine the correct treatment, particularly when there is a suspicion of malignancy (9). A pleural effusion may be associated with cancer in 15-30% of cases, and 40-60% of patients require an invasive diagnostic approach to confirm or rule out the suspect (10). In clinical practice, patients with suspected MPE undergo several diagnostic procedures. Unfortunately, especially in patients with malignant PE not related to primary lung cancer (including PMs), the accuracy of PF cytology is inadequate as the sole diagnostic tool (4, 7, 11-12). Several enzymes and tumour markers (TMs), other than CEA, have been tested in the PF of patients with MPE, to increase sensitivity of cytological results, but their role is questionable (13-15).
Nguyen et al., in a recent meta-analysis, suggested to use a combination of TMs, to increase accuracy (16). Moreover, it has been argued that indiscriminate testing of a panel of markers in all patients with PE may not be a cost-effective procedure, and strategies based on the risk of malignancy have been suggested. Furthermore, the sensitivity and specificity of pleural CEA varies widely, according to the heterogeneity of the studied population, the cut-off used, and (in part) the method of detection (17). We have previously shown that the measurement of CEA in PF is more accurate than PC alone, and suggested to test both in PE, before further procedures (7).
It has been shown that pleural CRP levels were significantly lower in patients with malignant PE (reported AUC: between 0.752 and 0.810), and that both pleural LDH and serum LDH: PF ADA ratio were higher in patients with lung cancer and PE, compared with controls (AUC=0.814) (18-20). If cytology and immunochemistry are inconclusive, a blind pleural biopsy adds little value, due to the low sensitivity, and a pleural biopsy with thoracoscopic approach is needed, but the risk of complications is not negligible (7, 21). Thus, developing a predictive model is crucial, to reduce the heterogeneity in approach and avoid invasive procedures in the management of patients with a suspicion of MPE (22). The usefulness of tree diagrams including clinical, radiological and PF analytical findings, despite their simplicity, has been reported in usual clinical practice (23). We found a significant difference between levels of CEA, CRP and LHD in the pleural fluid of patients with MPE.
After univariate analysis and logistic regression, LDH was excluded, and a predictive model was obtained with the combination of PC, CEA and CRP. This simple predictive model was able to correctly identify 33 out of 40 MPE (82.5%), and the relative AUC was 0.894 (95% CI=0.830-0.958).
Our study has several limitations, mainly represented by the small number of patients, from a single center, and the retrospective analysis. The strength of the study is the presence of a definitive histological diagnosis, obtained in all patients, both those with benign and malignant disease. The presence of a certain diagnosis of malignancy guarantees the reliability of the statistical evaluation, accurately calculating the sensitivity and specificity of our predictive model. Furthermore, these preliminary data may represent a starting point for future research, on predictive models to minimise the use of an invasive diagnostic approach for patients with PE.
In conclusion, in all patients with a history of cancer and PE of uncertain origin, PF cytology should be performed together with pleural CEA and CRP assay. The association cytology+CEA+CRP has been found to be reliable and accurate in diagnosing a malignant PE, minimising the use of unnecessary invasive investigations. In our study, the AUC obtained with the combination of all the tools was relevant, and higher than that of other previously suggested non-invasive protocols. This means that our simple predictive model is effective in differentiating malignant PE from benign one, without the use of invasive procedures, as VAT-guided biopsy. However, the validity of our hypothesised predictive model should be verified by multicenter studies with more patients.
Footnotes
Authors' Contributions
SMMB contributed to study design, manuscript writing, literature search and data analysis. FL contributed to study design, manuscript writing and data analysis. ADC, UZ, GF, FM and PU contributed to acquisition of data and literature research. SCS performed pleural fluid cytology and contributed to acquisition of the data. ME contributed to study design and manuscript revision and performed the statistical analysis. All Authors discussed the results and approved the final version of the manuscript.
Conflicts of Interest
The Authors declare no conflicts of interest.
- Received July 18, 2020.
- Revision received August 2, 2020.
- Accepted August 3, 2020.
- Copyright© 2020, International Institute of Anticancer Research (Dr. George J. Delinasios), All rights reserved
