ORIGINAL ARTICLE Comparison of Diagnostic Accuracy Between CellprepPlusVR and Liquid-Based Preparations in Effusion Cytology Yong-Moon Lee, M.D., 1 Ji-Yong Hwang, 1 Seung-Myoung Son, M.D., 1 Song-Yi Choi, M.D., Ph.D., 1 Ho-Chang Lee, M.D., Ph.D., 1 Eun-Joong Kim, 2 Hye-Suk Han, M.D., Ph.D., 3 Jin young An, M.D., 3 Joung-Ho Han, M.D., Ph.D., 3 and Ok-Jun Lee, M.D., Ph.D. 1 * Liquid-based cytology (LBC) is being increasingly used for body fluid specimens and has improved diagnostic accuracy when compared to conventional smears. We compared the diagnostic accuracy and cellular morphologic features between CellprepPlusVR LBC and LBC in effusion cytology. One hundred and eighty body fluid specimens, consisting of 119 pleural fluid specimens, 59 peritoneal fluid specimens, and 2 pericardial fluid specimens, were obtained from 166 patients. Equal volumes of body fluid from each specimen were used in the CellprepPlusVR and preparations. Sensitivity, specificity, and positive and negative predictive values were evaluated. In addition, we selected 16 specimens from patients with metastatic adenocarcinoma, confirmed them by both LBC preparations, and measured the size of the nucleus in the tumor cells in these specimens. The sensitivity of the CellprepPlusVR and methods was 73.1% and 50.0%, respectively. The specificity and positive This study was approved by the institutional review board of the Chungbuk National University Hospital (CBNU-IRB-2009-G03). Yong-Moon Lee and Ji-Yong Hwang contributed equally to this work. Additional Supporting Information may be found in the online version of this article. 1 Department of Pathology, Chungbuk National University College of Medicine, Heungdeok-gu, Cheongju, Korea 2 Department of Clinical Laboratory Science, Chungbuk Health and Science University, Cheongwon, Korea 3 Department of Internal Medicine, Chungbuk National University College of Medicine, Heungdeok-gu, Cheongju, Korea This work was supported by the research grant of Chungbuk National University in 2012. *Correspondence to: Ok-Jun Lee, M.D., Ph.D., Department of Pathology, Chungbuk National University College of Medicine, 52 Naesudong-ro, Heungdeok-gu, Cheongju 361-763, Korea. E-mail: ok5218@hanmail.net Received 25 January 2013; Accepted 28 August 2013 DOI: 10.1002/dc.23041 Published online 00 Month 2013 in Wiley Online Library (wileyonlinelibrary.com). predictive values were 100% for both LBC methods, and the negative predictive values of the CellprepPlusVR and methods were 90.9% and 83.3%, respectively. The average nuclear size of the tumor cells was calculated as 20.87 lm using the CellprepPlusVR method and 15.08 lm using the method (P < 0.05). The CellprepPlusVR method provided better diagnostic accuracy of effusion cytology compared to the method and revealed the characteristic morphological features of tumor cells, including large and hypochromatic nuclei, prominent nucleoli, distinct nuclear membranes, and high cellularity. Diagn. Cytopathol. 2013;00:000 000. VC 2013 Wiley Periodicals, Inc. Key Words: effusion; body fluid; cytology; CellprepPlusVR ; The appearance of malignant effusions in body cavities is a common complication of cancer affecting the pleural, peritoneal, and, less often, pericardial space. Around onefifth of body cavity effusions detected each year are malignant. Lung cancer accounts for the largest portion of malignant pleural effusions, followed by breast cancer. 1 Ovarian cancer is the most common cause of malignant peritoneal effusions, which are detected at diagnosis in two-thirds of patients. 2 On the other hand, malignant pericardial effusions, which have the same etiology as pleural effusions, are less common and are only detected in 2 30% of cancer patients at autopsy. 1 Cytologic body fluid samples are commonly encountered in routine practice; however, distinguishing malignant from benign cells remains a challenge. The liquidbased cytology (LBC) processor (Cytyc Co., Boxborough, MA) has been approved by the US Food and Drug Administration (FDA) since 1996 for processing gynecological VC 2013 WILEY PERIODICALS, INC. Diagnostic Cytopathology, Vol. 00, No 00 1
LEE ET AL. Table I. Summary of Cases Sample type Cases (%) Patients (%) Pleural fluid 119 (66.1) 106 (63.9) Peritoneal fluid 59 (32.8) 58 (34.9) Pericardial fluid 2 (1.11) 2 (1.20) Total 180 (100) 166 (100) specimens. Similar to LBC, CellprepPlusVR LBC (Biodyne, Seongnam, Korea) is based on a filtration process and has its own special blowing technology. Many studies have compared liquid-based preparation methods, such as LBC, and the conventional smear, 3 11 and some reports have also examined CellprepPlusVR LBC. 12,13 However, to our knowledge, only one study has compared the accuracy and unsatisfactory rates in cytology between CellPrepVR LBC and LBC. 14 Therefore, this study compared the sensitivity and specificity of the CellprepPlusVR and LBC methods in effusion cytology and investigated the differences in cytomorphological features between samples prepared using both methods. Materials and Methods Sample Collection One hundred and eighty body fluid specimens were obtained from 166 patients, consisting of 119 pleural fluid specimens, 59 peritoneal fluid specimens, and 2 pericardial fluid specimens (Table I). All samples were collected from January 2010 to May 2011 and submitted to the Chungbuk National University Hospital Department of Pathology. Sample Processing Equal volumes from each sample were used to simultaneously prepare a CellprepPlusVR and slide, and each method followed the detailed manufacturer s protocol for sample preservation and slide preparation. The detailed processing steps for preparation of the CellprepPlusVR slide are as follows: between 5 and 10 ml of body fluid was prepared by cytocentrifugation for 5 minutes at 1,800 rpm, the supernatant was removed, and 20 ml of preservation solution was added to the remaining pellet. The mixture was vortexed for 10 seconds, poured into the CellprepPlusVR vial, and placed onto the fully automated CellprepPlusVR processor. Once a bottle containing cells in preservation solution is placed into the CellprepPlusVR processor, the CellprepPlusVR processor transfers the cells onto a slide through a two-step process: (1) it filters the cells out by negative pressure and (2) blows the cells from the micromembrane to the slide using air pressure, resulting in their transfer to a 20-mm diameter circular area on a glass slide. Contrast to the process, CellprepPlusVR processor filtered out cells toward gravity direction, and blew them to the slide using air pressure. Once smearing was completed, the slide was fixed in 95% ethanol for at least 30 minutes, and Papanicolaou staining was applied in the same manner as the process. Cytological Diagnosis Two independent pathologists simultaneously reviewed all cases on an Olympus BX51 (Olympus Corp., Tokyo, Japan) multi-head microscope without information concerning previous pathological and clinical findings. For each sample, the CellprepPlusVR slide was examined first, followed by the slide, without consideration of the results of the CellprepPlusVR slide. Cytologic diagnoses were considered adequate or inadequate, and adequate diagnoses were further categorized as: (1) negative, (2) atypical, (3) suspicious, or (4) malignant. We defined diagnosis confirmed malignant as: (1) Malignant cells present in CellprepPlusVR and/or slides and a primary lesion pathologically and/or clinically confirmed as malignant ; and (2) malignant cells detected on cell block slides produced at the same time as the preparations of the CellprepPlusVR and methods and confirmed by immunocytochemical staining. We defined diagnosis confirmed benign as no malignant cells observed on a cell block or LBC of samples from patients who had no clinical and/or radiological evidence of malignancy. We excluded cases where primary cancer was histologically confirmed but effusion cytology identified no malignant cells. Comparison of Diagnostic Accuracy and Cytomorphologic Features We investigated the sensitivity, specificity, positive predictive value, and negative predictive value of the CellprepPlusVR LBC and LBC methods. We also reviewed all 30 confirmed malignant effusion cases to compare the general cytological features observed in CellprepPlusVR and LBC preparations, including cellularity, cell distributional uniformity, cohesiveness, cell size, nuclear size, nuclear hyperchromasia, and nuclear detail. Comparison of Tumor Cell Nuclear Size Sixteen samples, which showed cells ranging from atypical cells to adenocarcinoma, diagnosed by both methods were selected, and digital images of the tumor cells were captured using an excope X3 digital camera (DIXI Optics Co., Daejeon, Korea). The sizes of the tumor cell nuclei on the slides prepared by the CellprepPlusVR and methods were measured by a computerized image analyser (DIXI excope Image Software 5.0.1, DIXI Optics Co., Daejeon, Korea). For each sample, the longest line profiles were obtained from 10 tumor cells, and the mean and standard deviation (SD) between the 2 Diagnostic Cytopathology, Vol. 00, No 00
COMPARISON OF DIAGNOSTIC ACCURACY Fig. 1. Slides prepared with CellprepPlusVR (A) and (B). A: CellprepPlusVR LBC slides exhibited a well-defined, circular area of a thin, evenly distributed cell layer with a clean background. B: LBC slides exhibited a similar cell distribution pattern. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.] Table II. Sample Adequacy of Body Fluids Using CellprepPlusVR and Diagnostic category CellprepPlusVR (%) (%) Adequate 179 (99.5) 170 (94.5) Inadequate 1 (0.56) 10 (5.56) Thick smear Scanty cellularity 1 (0.56) 9 (5.00) Obscuring inflammation Obscuring blood 1 (0.56) Total 180 (100) 180 (100) CellprepPlusVR and methods were compared. We used an independent Student t-test for statistical analysis, and a P value <0.05 was considered statistically significant (SPSS 19.0, IBM Inc., NY). Results Most slides prepared by both CellprepPlusVR and methods exhibited a thin layer of cells uniformly distributed along a well-defined circular area measuring 20 and 22 mm in diameter, respectively (Fig. 1). One of the 180 samples prepared using the CellprepPlusVR LBC method was inadequate for evaluation due to the absence of cells, and 10 samples prepared using the method were inadequate for evaluation due to the absence of cells in 9 samples and obscuring blood in 1 sample (Table II). The cytologic diagnoses of 140 CellprepPlusVR LBC preparations (77.8%) were classified as negative. Of the LBC preparations, 129 (71.7%) were classified as negative, 1 as atypical, 1 as suspicious, and 9 as inadequate. Twenty-seven of the CellprepPlusVR LBC preparations were diagnosed as malignant, while 17 of the LBC preparations were classified as malignant, 2 as suspicious, 5 as atypical, and 3 as negative (Table III). One hundred and three of the 180 cases (57.2%) were classified as diagnosis confirmed. Of these, 30 cases were further classified as diagnosis confirmed malignant, and the rest were classified as diagnosis confirmed benign, including four cases of Mycobacterium tuberculosis infection confirmed by polymerase chain reaction (PCR, BioSewoom, Inc., Korea). All 30 confirmed malignant effusion cases were caused by metastatic tumors that included lung adenocarcinomas (17 cases), squamous cell carcinoma from the lung (1 case), invasive ductal carcinomas from the breast (2 cases), gastric adenocarcinomas (2 cases), lymphoma (1 case), and adenocarcinomas of various organs including the ovary, pancreas, common bile duct, endometrium, colorectum, and of unknown origin (7 cases). Of the 30 diagnosis confirmed malignant cases, 18 were interpreted as malignant, 1 as suspicious, 4 as atypical, and 7 as negative using CellprepPlusVR LBC, whereas 13 were interpreted as malignant, 4 as atypical, and 13 as negative using LBC. Of the 73 diagnosis confirmed benign cases, 70 were interpreted as negative and 3 as atypical using the CellprepPlusVR method, whereas 65 were interpreted as negative, 3 as atypical, and 5 as inadequate using the LBC method (Table IV). To compare the sensitivity, specificity, and positive and negative predictive values between both methods, we reclassified the cytological diagnoses, except for cases classified as atypical, as follows: negative and inadequate slides were categorized as benign, while suspicious and malignant slides were categorized as malignant. Sensitivity and specificity were 73.1% and 100%, respectively, for CellprepPlusVR LBC preparations, compared to 50.0% and 100%, respectively, for LBC preparations. The positive and negative predictive values were estimated as 100% and 90.9%, respectively, for the CellprepPlusVR method, compared to 100% and 83.3%, respectively, for the LBC method (Table V). Diagnostic Cytopathology, Vol. 00, No 00 3
LEE ET AL. Table III. Comparison of Cytologic Diagnoses Between CellprepPlusVR and Preparations CellprepPlusVR Inadequate Negative Atypical Suspicious Malignant Total (%) Inadequate 1 1 (0.55) Negative 9 129 1 1 140 (77.8) Atypical 7 4 11 (6.11) Suspicious 1 1 (0.55) Malignant 3 5 2 17 27 (15.0) Total (%) 10 (5.56) 140 (77.8) 10 (5.56) 3 (1.67) 17 (9.44) 180 (100) Table IV. Correlation Between Cytologic and Confirmed Diagnosis in CellprepPlusVR and Preparations Diagnosis confirmed Benign Malignant Cytologic diagnosis CellprepPlusVR CellprepPlusVR Inadequate 0 5 Negative 70 65 7 13 Atypical 3 3 4 4 Suspicious 0 0 1 Malignant 0 0 18 13 Total 73 30 We evaluated all 30 cases of confirmed malignant effusion and compared the preparations from both methods by examining cytological features, including cellularity (Supporting Information Table 1; the number of cell clumps counted in both methods), cell distributional uniformity, cohesiveness, cell size, nuclear size, nuclear hyperchromasia, and nuclear detail. Analysis of cell distribution revealed fewer cells on the slides compared to the CellprepPlusVR slides in 23 of the 30 cases and more peripherally distributed cells on the slides (Fig. 2). Cell size evaluation revealed that cells on CellprepPlusVR slides were larger than those on slides in 20 of the 30 cases. High-magnification observations revealed larger nuclei in the cells on the CellprepPlusVR slides compared to the slides in 20 of the 30 cases. Moreover, the nuclei of the cells on the CellprepPlusVR slides were more hypochromatic in 18 of the 30 cases, such that tumor cell nucleoli were more easily detectable on the CellprepPlusVR slides compared to the slides in 20 of the 30 cases. The characteristic cytological features observed in samples prepared using the CellprepPlusVR and methods are summarized in Table VI (Fig. 3). The nuclear size of the tumor cells was measured in 16 cases of metastatic adenocarcinoma, which showed a statistically significant increase (P < 0.05) in the tumor cells in CellprepPlusVR LBC compared to LBC preparations (Table VII). These findings were also consistent when evaluating the nuclear size in reactive mesothelial cells using 30 benign cases (Supporting Information Table II). Table V. Comparison of Diagnostic Accuracy Between CellprepPlusVR and Methods Diagnosis confirmed Negative CellprepPlusVR Suspicious or malignant Negative Suspicious or malignant Benign 70 0 65 0 Malignant 7 19 13 13 Total 77 19 78 13 Sensitivity (%) 73.1 50.0 Specificity (%) 100 100 Positive 100 100 predictive value (%) Negative predictive value (%) 90.9 83.3 Discussion LBC preparations allow cell enrichment through the reduction of background inflammatory and blood cells. The presentation of cells in a uniform layer through an automated and standardized process enhances the identification of malignant cells. In addition, residual specimens can be used in immunocytochemical and molecular studies. 5 LBC is widely used for preparing gynecological and nongynecological cytology specimens. Koo et al. 13 reported that preparation of body fluids using CellprepPlusVR LBC yielded better sensitivity and negative predictive values, as well as a higher slide quality, than the conventional smear. The use of CellprepPlusVR LBC in preparing urine cytology specimens also resulted in improved slide quality and diagnostic accuracy compared to the conventional smear. 12 Additionally, CellPrepVR LBC produced results comparable to those of LBC in terms of smear quality and diagnostic efficacy in cervicovaginal smear, body fluid, and urine specimen preparations. 14 It is thought that the increased diagnostic accuracy of the CellprepPlusVR method over the method is due to the high cellularity and characteristic cytomorphological features produced by CellprepPlusVR. These results possibly originate from the different preparation steps and preservation solutions used by the two methods. First, the CellprepPlusVR processor filters the cells toward gravity 4 Diagnostic Cytopathology, Vol. 00, No 00
COMPARISON OF DIAGNOSTIC ACCURACY Fig. 2. Slides prepared by LBC. A: Low-magnification view of slides prepared by LBC (312). Some cell clusters were observed at the peripheral rim (between arrows), while an acellular area was observed at the center. B: High-magnification view of slides prepared by LBC (3100). [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.] Table VI. Characteristic Cytologic Features Using CellprepPlusVR and CellprepPlusVR vs. No. of cases (%) Cellularity > 23 (76.7) 5 7 (23.3) Distributional uniformity > 15 (50.0) 5 15 (50.0) Cohesiveness < 9 (30.0) 5 21 (70.0) Cell size > 20 (66.7) 5 10 (33.3) Cytomorphology Nuclear size > 20 (66.7) 5 10 (33.3) Nuclear hyperchromasia < 18 (60.0) 5 12 (40.0) Nuclear detail > 20 (66.7) 5 10 (33.3) Total 30 (100) direction otherwise the antigravity direction of processor. This different relationship between the glass slide and the gravitational forces may contribute to the lower number of cells on the slide because large cell clumps are likely to be detached. Second, no instrument is introduced into the specimen vial in the CellprepPlusVR processor, while in the processor, a cylinder with a thin, polycarbonate filter attached to one end is introduced into the vial and rotated to create a current that disaggregates mucus, blood, and other debris. This direct contact may affect the cellular, morphological differences between the two methods. Third, the CellprepPlusVR processor blows the cells onto the slide, making a very thin, monolayered, 20-mm circle. The cylinder containing the cells in the method is directly pressed against the slide to transfer the cells, causing the slides to exhibit a less cellular central area with a more cellular peripheral rim. In our study, we also observed a greater tumor cell distribution around the periphery of the circular area in the LBC preparations, particularly in cases with low cellularity. Similarly, Choi et al. 15 reported that 67 of 474 cases exhibited more cells in the periphery of the slide. Consequently, they concluded that cytological diagnosis based on cells in the peripheral area should be made with caution, since the peripheral compression artifact affects the size of the cells and their degree of chromatin compaction, which are key criteria in the diagnosis of malignant cells. Finally, the preservation solution used for the CellprepPlusVR and methods is different in that the former is ethanol-based and the latter is methanol-based. Nuclear swelling and cytoplasmic shrinking in the methanol-based preservation solution is well known, while the ethanol-based preservation solution provides cellular details more familiar to the cytologist than the methanol-based method. 16 Consistent with previous studies, 12,14 we found that CellprepPlusVR LBC yielded larger tumor cell nuclei in 20 of the 30 cases and more hypochromatic tumor cell nuclei in 18 of the 30 cases compared to LBC. Therefore, the nucleoli and nuclear membrane of the tumor cells were more prominent in CellprepPlusVR LBC preparations compared to LBC preparations in 20 of the 30 cases. Furthermore, we measured the size of tumor cell nuclei in CellprepPlusVR and preparations; the mean diameter of the tumor cell nuclei was 20.87 6 5.67 lm in the former and 15.08 6 4.94 lm in the latter, indicating a statistically significant difference in tumor cell nuclear size between both methods (P < 0.01). Overall, the CellprepPlusVR LBC method produced more slides that contained larger cells and nuclei than the LBC method, making recognition of nuclear details such as nucleoli or chromatin changes easier. However, these characteristics Diagnostic Cytopathology, Vol. 00, No 00 5
LEE ET AL. Table VII. Measurements of Tumor Nuclei of Metastatic Adenocarcinoma in CellprepPlusVR and Preparations Case No. CellprepPlusVR Diameter (lm) 1 19.35 6 2.94 16.06 6 3.44 0.034 2 20.13 6 3.60 11.79 6 3.24 <0.01 3 20.84 6 3.42 15.95 6 4.60 0.015 4 26.60 6 2.30 13.91 6 3.39 <0.01 5 26.62 6 4.21 18.69 6 2.80 <0.01 6 17.87 6 1.69 16.52 6 4.61 0.39 7 30.76 6 6.37 21.59 6 6.93 <0.01 8 26.95 6 2.02 20.73 6 4.30 <0.01 9 16.23 6 2.89 14.81 6 3.63 0.34 10 20.29 6 2.68 14.20 6 2.76 <0.01 11 14.53 6 1.56 15.16 6 2.72 0.53 12 17.33 6 2.07 15.16 6 3.93 0.14 13 19.57 6 4.56 15.20 6 2.91 0.02 14 24.56 6 4.62 15.59 6 3.53 <0.01 15 17.05 6 2.23 8.08 6 2.34 <0.01 16 15.21 6 2.55 9.51 6 3.08 <0.01 Average 20.87 6 5.67 15.08 6 4.94 <0.01 of CellprepPlusVR LBC could cause cells to appear more tumor-like, especially mesothelial cells; therefore, cytopathologists should be aware of the morphological characteristics of the LBC method used to avoid an erroneous diagnosis. p In summary, CellprepPlusVR LBC had good diagnostic accuracy and allowed the display of tumor cell characteristics, including large and hypochromatic nuclei, prominent nucleoli, distinct nuclear membranes, and high cellularity. References 1. Rosalyn AJ, Alex IS, Julie RB. Effusion. In: Abeloff MD, Armitage GO, Niederhuber JE, Kastan MB, McKenna WG, editors. Abeloff s clinical oncology. 4th ed. Philadelphia: Churchill Livingstone; 2008. p 925 944. 2. Ben D. Malignant effusion: From diagnosis to biology. Diagn Cytopathol 2004;31:246 254. 3. Kim S, Owens CL. Analysis of ThinPrep cytology in establishing the diagnosis of small cell carcinoma of lung. Cancer Cytopathol 2009;117:51 56. 4. Zhao FH, Hu SY, Bian JJ, et al. Comparison of ThinPrep and SurePath liquid-based cytology and subsequent human papillomavirus DNA testing in china. Cancer Cytopathol 2011;119: 387 394. 5. Moriarty AT, Schwartz MR, Ducatman BS, et al. A liquid concept-do classic preparations of body cavity fluid perform differently than ThinPrep cases? Arch Pathol Lab Med 2008;132: 1716 1718. 6. Waugh MS, Guy CD, Maygarden SJ, Livasy CA, Jones CK, Volmar KE. Use of method in bile duct brushing: Analysis of morphologic parameters associated with malignancy and determination of interobserver reliability. Diagn Cytopathol 2008; 36:651 656. 6 Diagnostic Cytopathology, Vol. 00, No 00
COMPARISON OF DIAGNOSTIC ACCURACY 7. Shin BK, Lee YS, Jeong H, et al. Detecting malignant urothelial cells by morphometric analysis of liquid-based urine cytology specimens. Korean J Cytopathol 2008;19:136 143. 8. Hoda RS. Non-gynecologic cytology on liquid-based preparation: A morphologic review of facts and artifacts. Diagn Cytopathol 2007; 35:621 634. 9. Lu DY, Nassar A, Siddiqui MT. High-grade urothelial carcinoma: Comparison of SurePath TM liquid-based processing with cytospin processing. Diagn Cytopathol 2009;37:16 20. 10. Lee S, Park JH, Do SI, et al. Diagnostic value of urine cytology in 236 cases: A comparison of liquid-based preparation and conventional cytospin method. Korean J Cytopathol 2007;18:119 125. 11. Yim H, Joo HJ, Kim YB, Hong SW. Comparison of conventional smear, cell block and liquid-based preparation in evaluation of bronchial washing specimen in lung cancer patients. Korean J Pathol 2011;45:296 302. 12. Son SM, Koo JH, Choi SY, et al. Evaluation of urine cytology in urothelial carcinoma patients: A comparison of CellprepPlusVR liquid-based cytology and conventional smear. Korean J Pathol 2012;46:68 74. 13. Koo JH, Lee HC, Song HG, et al. Comparison of cytologic evaluation between conventional method and CellprepPlusVR liquid-based cytology in body fluid. Korean J Pathol 2009;43: 448 452. 14. Koh JS, Cho SY, Ha HJ, Kim JS, Shin MS. Cytologic evaluation of CellPrepVR liquid-based cytology in cervicovaginal, body fluid, and urine specimens: Comparison with. Korean J Cytopathol 2007;18:29 35. 15. Choi J, Shim HS, Song JW, et al. Image analysis of peripheral compression artefacts of liquid-based cytology preparations. Cytopathology, in press. 16. Yoshiaki N, Shingo S, Hiroyuki O, et al. Cytologic features of the endometrial adenocarcinoma: Comparison of ThinPrep and BD SurePath preparations. Diagn Cytopathol, in press. 17. Park M, Baek T, Baek J, et al. Nuclear image analysis study of neuroendocrine tumors. Korean J Pathol 2012;46:38 41. Diagnostic Cytopathology, Vol. 00, No 00 7