Diagnostic value of tumor markers in lung adenocarcinoma-associated cytologically negative pleural effusions
The first 2 authors contributed equally to this article.
Cytology fails to detect neoplastic cells in approximately 40% to 50% of malignant pleural effusions (PEs), which commonly accompany lung adenocarcinomas. The diagnostic accuracy of various tumor markers in lung adenocarcinoma-associated cytologically negative pleural effusions (LAC-CNPEs) has been poor. The current study attempted to maximize diagnostic efforts in distinguishing LAC-CNPEs from benign PEs.
PE samples were collected from 74 patients with lung adenocarcinoma with associated cytologically positive (41 patients) and negative (33 patients) PEs, and from 99 patients with benign conditions including tuberculosis (26 patients), pneumonia (28 patients), congestive heart failure (25 patients), and cirrhosis (20 patients). The authors evaluated the diagnostic sensitivity and optimal cutoff points for the tumor markers HER2/neu, CYFRA 21-1, and carcinoembryonic antigen (CEA) to distinguish LAC-CNPEs from benign PEs.
Mean levels of HER2/neu, CYRFA 21-1, and CEA were found to be significantly higher in LAC-CNPEs compared with benign PEs (P = .0050, P = .0039, and P < .0001, respectively). The cutoff points for HER2/neu, CYFRA 21-1, and CEA were optimally set at 3.6 ng/mL, 60 ng/mL, and 6.0 ng/mL, respectively. Their sensitivities ranged from 12.1%, to 30.3%, to 63.6%, respectively. CEA combined with CYFRA 21-1 increased diagnostic sensitivity to 66.7%. The false-positive rates of these markers in benign PEs were 6.1%, 2.0%, and 0%, respectively.
The combination of CEA with CYFRA 21-1 appears to provide the best differentiation between LAC-CNPEs and benign PEs to date using 2 tumor markers, and allows for the early diagnosis and early treatment of approximately two-thirds of affected patients. Cancer (Cancer Cytopathol) 2013;121:483–8. © 2013 American Cancer Society.
Malignant neoplasms are a main cause of pleural effusion (PE), but the diagnosis of malignant PE is often a vexing problem because PE cytology findings are positive in only an average of 60% of cases.[1-3] A blind needle pleural biopsy adds little to false-negative cytology. Consequently, various tumor markers found in PEs have been investigated for their ability to differentiate malignant PEs from benign ones. These tumor markers include carcinoembryonic antigen (CEA), cancer antigen (CA) 125, CA 15-3, CA 19-9, CA 72-4, CYFRA 21-1, neuron-specific enolase, and squamous cell carcinoma (SCC) antigen.[1, 5] However, the majority of these studies were based on definite malignant PEs. Cancer cells had already been cytologically detected in the PEs.[5-7] Therefore, the results of these studies are not able to provide any benefit with regard to early diagnosis or early treatment.
Malignant PEs commonly accompany lung adenocarcinoma,[8, 9] the most common histologic type of lung cancer diagnosed in Taiwan.[10, 11] In Taiwan, patients with lung adenocarcinoma-associated malignant PEs (LAC-MPEs) have become increasingly more common. Studies regarding lung adenocarcinoma-associated cytologically negative PEs (LAC-CNPEs) have been poor.[1, 5, 12-14]
To the best of our knowledge, there has been no consensus regarding which tumor markers are the most effective at detecting lung adenocarcinoma, or even with regard to what their optimal cutoff points should be. To clarify this situation, we performed the current study with the objective of evaluating the diagnostic sensitivity, specificity, and optimal cutoff points of HER2/neu, CYFRA 21-1, and CEA to discriminate LAC-CNPEs from benign PEs, and to provide early diagnosis and early treatment for these patients.
MATERIALS AND METHODS
Sample Collection and Processing
After obtaining informed consent from all subjects, a total of 1005 residual PE samples were collected for medical diagnosis from January 2005 to March 2007 from patients hospitalized at the Buddhist Dalin Tzu Chi General Hospital in Chiayi County, Taiwan. Of these, 173 samples were analyzed. Analyzed PEs came from patients with LAC-MPEs (41 cases) or LAC-CNPEs (33 cases), as well as patients with various benign conditions (99 cases) (tuberculosis [26 cases], parapneumonic PE [28 cases], congestive heart failure [25 cases], and cirrhosis [20 cases]).
LAC-MPE was categorized if adenocarcinomatous cells were found cytologically by 2 pathologists in the PE sample from patients with a primary lung adenocarcinoma. LAC-CNPE was diagnosed in patients with a known primary lung adenocarcinoma and a negative PE cytology finding after ruling out benign causes of the PE. Afterward, adenocarcinoma was demonstrated in PE cytology or pleural biopsy in these patients within 3 months. Diagnosis of tuberculous pleurisy was confirmed by isolation of Mycobacterium tuberculosis through culture of the PE sample or tissue or a pleural biopsy specimen demonstrating typical caseating granulomas. When a pleural biopsy was not performed, we considered patients to have tuberculous pleurisy if they met the following 2 conditions: 1) isolation of M. tuberculosis from the sputum and 2) if the effusion cleared in response to antituberculous therapy. Parapneumonic PE indicated a nonempyemic effusion accompanying an acute febrile illness with purulent sputum, pulmonary infiltrates, and responsiveness to antibiotic treatment. Patients were found to have no evidence of tuberculosis, malignancy, or other major disease. Congestive heart failure was considered in the instance of an enlarged heart noted on chest radiography or echocardiography; pulmonary venous congestion on radiography; peripheral edema; response to optimal medical treatment; and an absence of malignancy, pulmonary infiltrates, purulent sputum, or pleuritic chest pain. Cirrhosis was diagnosed based on biopsy or on the presence of clinical and laboratory features of portal hypertension (esophageal varices at endoscopy and/or collateral circulation noted on ultrasonography).
Patients with empyema, other primary sites of malignancies, PE specimens that tested positive for bacterial culture, or clinical courses that were complicated with other major diseases were excluded from the current study. The study protocol was approved by the Institutional Review Board of the Buddhist Dalin Tzu Chi General Hospital in Chiayi County, Taiwan.
PE samples obtained by thoracentesis were collected in sterile heparinized tubes and centrifuged immediately at 1500 g. Cell-free supernatant fluid was collected, and aliquots were stored frozen at −70 °C until the assay was performed.
Immunoassay for Tumor Markers
PE samples were assayed for HER2/neu using an enzyme-linked immunoadsorbent assay kit (Oncogene Research Products, Cambridge, Mass) according to manufacturer's instructions. The interassay coefficients of variation of replicate samples for the HER2/neu were 1.5% at 1.9 ng/mL and 3.4% at 2.9 ng/mL.
Levels of CYFRA 21-1 and CEA in PEs were determined using electrochemiluminescence immunoassay on Roche Elecsys 2010 analyzers (Roche Diagnostics Corporation, Indianapolis, Ind) according to the manufacturer's instructions. If the levels of CYFRA 21-1 and CEA were >500 ng/mL and 1000 ng/mL (the upper limit of instrumental linearity), respectively, the specimens were not diluted further and reassayed, and 500 ng/mL and 1000 ng/mL were used as the final results. The interassay coefficients of variation of replicate samples for the CYFRA 21-1 were 2.2% at 2.9 ng/mL and 2.0% at 26.6 ng/mL; the coefficients of variation for CEA were 3.4% at 5.4 ng/mL and 2.2% at 47.2 ng/mL. All tumor marker assays were performed blinded to clinical information.
The mean and standard errors were used to express the distribution of tumor markers in PEs. Differences in means of continuous variables without normal distribution between 2 groups of patients were analyzed using the Wilcoxon–Mann-Whitney test. Results from patients with definite malignant and benign PEs were used to select cutoff values for sensitivity and specificity for all markers. Sensitivity and specificity patterns were studied by cumulative distribution analysis. Cutoff points were determined by the maximum sum of sensitivity and specificity. Statistical significance was set at P < .05 (2-tailed). Statistical analyses were performed using SPSS statistical software for Windows (version 12.0; SPSS Inc, Chicago, Ill).
Of the 41 patients with LAC-MPEs, 21 (51.2%) were men and 20 (48.8%) were women, with a mean age of 69.4 years (range, 48 years-89 years). Of the 33 patients with LAC-CNPEs, 19 (57.6%) were men and 14 (42.4%) were women, with a mean age of 68.6 years (range, 43 years-96 years). Of the 99 patients with benign PEs, 64 (64.6%) were men and 35 (35.4%) were women, with a mean age of 68.6 years (range, 25 years-97 years).
Table 1 shows the means and the standard errors of the mean of the tumor markers HER2/neu, CYFRA 21-1, and CEA in all groups of patients by diagnosis. The mean levels of these 3 markers were found to be significantly greater in patients with LAC-MPEs than in those with benign PEs (Wilcoxon–Mann-Whitney test: P = .0157, P < .0001, and P < .0001, respectively). The mean levels of HER2/neu, CYFRA 21-1, and CEA were significantly greater in patients with LAC-CNPEs than in those with benign PEs (Wilcoxon–Mann-Whitney test: P = .0050, P = .0039, and P < .0001, respectively). The mean levels of HER2/neu, CYFRA 21-1, and CEA were significantly greater in patients with LAC-MPEs compared with those with LAC-CNPEs (Wilcoxon–Mann-Whitney test: P = .0013, P < .0001, and P = .0219, respectively).
Table 1. Levels of HER2/neu, CYFRA 21-1, and CEA in Various Groups of Pleural Effusions
|LAC-MPE (n = 41)a||4.6 ± 0.7||247.5 ± 30.6||297.1 ± 58.0|
|LAC-CNPE ( n = 33)a||2.3 ± 0.3||112.0 ± 29.8||148.3 ± 47.0|
|Benign (n = 99)a||2.4 ± 0.1||20.1 ± 2.0||1.6 ± 0.1|
Table 2 shows the diagnostic sensitivity of the 3 tumor markers for diagnosing lung adenocarcinoma in PE. We used cutoff values of 3.6 ng/mL, 60 ng/mL, and 6.0 ng/mL for PE HER2/neu, CYFRA 21-1, and CEA, respectively. For LAC-MPEs, diagnostic sensitivities ranged from 43.9% for HER2/neu to 87.8% for CEA. CEA combined with CYFRA 21-1 was found to elevate the diagnostic sensitivity in patients with LAC-MPEs to 97.6%.
Table 2. Measures of Diagnostic Sensitivity for Tumor Marker Assays in LAC-MPE and LAC-CNPE
|HER2/neu > 3.6 ng/mL|| |
|Globala||29.7 ± 5|
|LAC-MPE||43.9 ± 8|
|LAC-CNPE||12.1 ± 4|
|CYFRA 21-1 > 60 ng/mL|| |
|Globala||58.1 ± 6|
|LAC-MPE||80.5 ± 8|
|LAC-CNPE||30.3 ± 8|
|CEA > 6.0 ng/mL|| |
|Globala||77.0 ± 6|
|LAC-MPE||87.8 ± 7|
|LAC-CNPE||63.6 ± 9|
|CEA and CYFRA 21-1|| |
|Globala||83.8 ± 5|
|LAC-MPE||97.6 ± 1|
|LAC-CNPE||66.7 ± 9|
|CEA, CYFRA 21-1, and HER2/neu|| |
|Globala||83.8 ± 5|
|LAC-MPE||97.6 ± 1|
|LAC-CNPE||66.7 ± 9|
For patients with LAC-CNPEs, diagnostic sensitivities ranged from 12.1% for HER2/neu to 63.6% for CEA. CEA combined with CYFRA 21-1 was found to elevate the diagnostic sensitivity in LAC-CNPEs to 66.7%. Further combination with HER2/neu was not found to increase diagnostic sensitivity. False-positive rates of HER2/neu, CYFRA 21-1, and CEA in benign effusions were 6.1%, 2.0%, and 0%, respectively.
If the cutoff value of HER2/neu was set at 5.5 ng/mL, as the previous authors had,[16, 17] the diagnostic sensitivity of HER2/neu that was obtained for LAC-CNPEs in the current study would be 9.1% (3 of 33 cases).
The results of the current study demonstrated that the mean values of HER2/neu, CYFRA 21-1, and CEA in LAC-MPEs were all significantly greater than those in benign PEs (P = .0157, P < .0001, and P < .0001, respectively). For LAC-MPEs, CEA combined with CYFRA 21-1 was found to elevate diagnostic sensitivity to 97.6%. However, this could not provide an earlier diagnosis than that with PE cytology because cancer cells had already been found in PEs cytologically.
To the best of our knowledge, the quality of studies published to date regarding LAC-CNPEs has been poor. The case numbers in each study were too small to be representative.[1, 5, 12-14] In 1999, Miedouge et al combined 4 tumor markers (CEA, CA 15-3, CYFRA 21-1, and neuron-specific enolase) for a discriminant assay in 19 adenocarcinoma-associated cytologically negative malignant PEs and reported a sensitivity of 94.7%. They did not specify how many cases of the cytologically negative malignant fluids were found to be associated with lung adenocarcinoma.
In the study by Porcel et al, the cutoff points of CEA and CYFRA 21-1 in PEs were set at ≥ 50 ng/mL and ≥ 175 ng/mL, respectively. The positivities for the tumor markers CEA and CYFRA 21-1 among the patients with LAC-MPEs who had negative cytology findings of their pleural fluids were 35.7% and 7.1%, respectively. The low sensitivities are likely to be associated with the high cutoff points. Furthermore, only 14 cases were included in that study.
The study by Trape et al included only 10 cases of lung cancer. They combined 3 tumor markers (CEA, CA 15-3, and CA 19-9) and reported a sensitivity of 79% in malignant fluids with negative cytology; when combining 5 tumor markers (including CA 125 and CYFRA 21-1), a sensitivity of 86% was obtained in these cases. The so-called “malignant fluids” included malignant PEs and malignant ascites as well as cytology-negative and cytology-positive cases; however, there are some basic differences between PE and ascites.[18-20] Furthermore, in their study, Trape et al did not specify how many cases of malignant fluids were found to be negative on cytology. There was also no sensitivity of individual markers demonstrated in cytology-negative malignant fluids.
Paraneoplastic PEs may have negative cytology due to the fact that cancer cells may be degenerated, rare, or undetectable. The level of tumor markers in these PEs should be less than that of malignant effusions as the results of the current study demonstrate. The results of studies by Ferrer et al and Gaspar et al also validated this point.[1, 12] Both studies contained only 18 lung-associated paraneoplastic PEs. In the former study, the median pleural concentrations of tumor markers were lower in paraneoplastic than in malignant effusions: 2.5 ng/mL and 5.2 ng/mL for CEA, 23.7 IU/L and 40.8 IU/L for CA 15.3, and 0.6 IU/L and 8.7 IU/L for TAG 72 [tumor-associated glycoprotein 72]. In the latter, the median pleural concentrations of tumor markers were lower in paraneoplastic compared with malignant effusions: 1.7 ng/mL and 5.7 ng/mL for CEA and 22.5 ng/mL and 90.5 ng/mL for CYFRA 21-1. Therefore, the sensitivities of tumor markers in paraneoplastic effusions should be less than in malignant effusions. In the study by Gaspar et al, the sensitivity of the combination of the 3 markers in paraneoplastic effusions (60.0%) was less than that in malignant effusions (87.5%). In the study by Ferrer et al, no combined diagnostic sensitivity for paraneoplastic effusions was shown. In the current study, the combined sensitivity of CEA and CYFRA 21-1 in paraneoplastic effusions (66.7%) was also found to be less than that in malignant effusions (97.6%).
To the best of our knowledge, reports concerning the usefulness of HER2/neu in LAC-CNPEs have not been published to date. In the current study, the mean value of HER2/neu in LAC-CNPEs was not greater, and was even less, than that of the benign group, although the mean values of both CYFRA 21-1 and CEA were significantly greater than those in benign PEs (P = .0039 and P < .0001, respectively). The mean values of HER2/neu, CYFRA 21-1, and CEA in LAC-CNPEs were apparently less than those in LAC-MPEs (P = .0013, P < .0001, and P = .0219, respectively). This might be because of the lower amount of tumor markers released by fewer tumor cells in LAC-CNPEs compared with LAC-MPEs. This could explain why only a few tumor cells could be found in LAC-CNPEs.
To the best of our knowledge, the current study is the largest known study describing 33 cases of LAC-CNPE. Of the 3 markers, HER2/neu was not found to be optimal for LAC-CNPE, because it provided the worst sensitivity (12.1%). Conversely, CEA was found to be the best marker, and resulted in the highest diagnostic sensitivity. When CEA was combined with CYFRA 21-1, the diagnostic sensitivity was elevated to 66.7% for LAC-CNPEs. This appeared to provide the best diagnostic sensitivity with 2 tumor markers reported to date.[5, 12-14] In other words, approximately two-thirds of LAC-CNPEs could be distinguished from benign PEs using this combination when no cancer cells were detected cytologically, and affected patients could obtain early diagnosis and early treatment.
An elevated tumor marker level usually indicates a malignant effusion. However, the consequence of incorrectly diagnosing a pleural malignancy is devastating. Elevated levels of tumor markers in the PE can warrant advancing to a more invasive procedure (eg, repeated pleural biopsy for pathology study or thoracentesis for a cytology study) to establish a malignant diagnosis, but should not in itself be used to make a definite diagnosis.
In conclusion, to the best of our knowledge, the current study is the largest known study of its kind published to date, describing 33 cases of LAC-CNPE. HER2/neu was found to have the lowest diagnostic sensitivity, whereas CEA had the highest diagnostic sensitivity and appeared to be a good marker for differentiating between LAC-CNPEs and benign PEs. CEA combined with CYFRA 21-1 would increase diagnostic sensitivity in LAC-CNPE up to 66.7%. To our knowledge, this is the best result reported to date using 2 tumor markers. Approximately two-thirds of these patients would be provided with early diagnosis and early treatment.
Supported by grant DTCRD94(2)-08 from the Buddhist Dalin Tzu Chi General Hospital, Chiayi, Taiwan.
CONFLICT OF INTEREST DISCLOSURES
The authors made no disclosures.