An Investigation Into False-Negative Transthoracic Fine Needle Aspiration and Core Biopsy Specimens Douglas M. Minot, C.T., M.B. (A.S.C.P.),1 Elizabeth A. Gilman, M.D.,2 Marie-Christine Aubry, M.D.,1 Jesse S. Voss, C.T., M.B. (A.S.C.P.),1 Sarah G. Van Epps, C.T. (A.S.C.P.),1 Delores J Tuve, C.T., (H.E.W.),1 Andrew P. Sciallis, M.D.,1 Michael R. Henry, M.D.,1 Diva R. Salomao, M.D.,1 Peter Lee, M.D., Ph.D.,3 Stephanie K. Carlson, M.D.,4 and Amy C. Clayton, M.D.1*
Transthoracic fine needle aspiration (TFNA)/core needle biopsy (CNB) under computed tomography (CT) guidance has proved useful in the assessment of pulmonary nodules. We sought to determine the TFNA false-negative (FN) rate at our institution and identify potential causes of FN diagnoses. Medical records were reviewed from 1,043 consecutive patients who underwent CT-guided TFNA with or without CNB of lung nodules over a 5-year time period (2003–2007). Thirty-seven FN cases of “negative” TFNA/CNB with malignant outcome were identified with 36 cases available for review, of which 35 had a corresponding CNB. Cases were reviewed independently (blinded to original diagnosis) by three pathologists with 15 age- and sexmatched positive and negative controls. Diagnosis (i.e., nondiagnostic, negative or positive for malignancy, atypical or suspicious) and qualitative assessments were recorded. Consensus diagnosis was suspicious or positive in 10 (28%) of 36 TFNA cases and suspicious in 1 (3%) of 35 CNB cases, indicating potential interpretive errors. Of the 11 interpretive errors (including both suspicious and positive cases), 8 were adenocarcinomas, 1 squamous cell carcinoma, 1 metastatic renal cell carcinoma, and 1 lymphoma. The remaining 25 FN cases (69.4%) were considered sampling errors and consisted of 7 adenocarcinomas, 3 nonsmall cell carcinomas, 3 lymphomas, 2 squamous cell carcinomas, and 2 renal cell carcinomas. Inter-
1 Department of Laboratory Medicine and Pathology, Mayo Clinic, Rochester, Minnesota 2 Department of Laboratory Medicine and Pathology, Mayo Clinic, Scottsdale, Arizona 3 Department of Radiology, Consulting Radiologists, Minneapolis, Minnesota 4 Department of Radiology, Mayo Clinic, Rochester, Minnesota *Correspondence to: Amy C. Clayton, M.D., Mayo Clinic, 200 First Street SW, Hilton Building 11th Floor, Rochester, MN 55905. E-mail: [email protected] Received 5 July 2013; Accepted 2 May 2014 DOI: 10.1002/dc.23169 Published online 28 May 2014 in Wiley Online Library (wileyonlinelibrary.com).
C 2014 WILEY PERIODICALS, INC. V
pretive and sampling error cases were more likely to abut the pleura, while histopathologically, they tended to be necrotic and air-dried. The overall FN rate in this patient cohort is 3.5% (1.1% interpretive and 2.4% sampling errors). Diagn. Cytopathol. 2014;42:1063–1068. VC 2014 Wiley Periodicals, Inc. Key Words: lung; neoplasm; computed tomography; diagnosis; cytopathology
Advances in the field of diagnostic radiology, including the use of computed tomography (CT), have led to the visualization and attempted assessment of smaller and smaller indeterminate pulmonary nodules. The use of the transthoracic needle biopsy technique on these pulmonary nodules has proved to be a safe, highly accurate, and minimally invasive method to determine the benign or malignant nature of these lesions. Specimens from these transthoracic fine needle aspiration (TFNA) and core needle biopsy (CNB) procedures are submitted for analysis by clinicians, oftentimes without knowledge of their laboratory false-negative (FN) rates. Some early institutional studies have reported relatively high FN rates on biopsy specimens collected using this technique, ranging from 9.6 to 45.6%.1–5 In more recent studies in which biopsies are primarily collected under CT-guidance, FN rates have ranged from 4.1 to 15.2%.6–9 The investigators’ predominant explanation for these FNs is that they are mostly a result of sampling error, or failure to collect adequate biologic material for accurate pathologic assessment. These findings have led some experts to conclude that the majority of FN diagnoses are usually attributed to sampling, rather than interpretive errors.10 As the sensitivity to detect pulmonary lesions via radiologic methods has improved, the Diagnostic Cytopathology, Vol. 42, No 12
1063
Diagnostic Cytopathology DOI 10.1002/dc
MINOT ET AL. Table I. Lung Lesion Characteristics and Specimen Qualitative Features
Mean size (cm) Lesion abuts pleura (%) Mean distance from pleura (cm) Necrosis (%) Bloody (%) Air-drying artifact (%) Precluding definitive diagnosis (%) Number of core biopsies, mean Mean total size of cores (length; mm)a a
Interpretive error
Sampling error
Negative control
Positive control
Overall
2.88 73 1.28 45 0 45 15 4.8 8.64
3.01 64 1.02 16 8 32 30 4.0 7.67
3.55 43 0.83 18 16 47 20 5.1 13.11
2.83 47 1.47 31 4 44 22 4.5 8.63
2.89 50 1.19 – – – – – –
Width of cores were mostly 0.5 mm.
debate becomes about what impact, if any, these methods have on the sensitivity of the pathologic assessment of these smaller nodules. A previous article from our group11 noted a small improvement in sensitivity (94 vs. 91%) and negative predictive value (80 vs. 74%) when we examined our pulmonary practice between the years 1996–1998 and 2003–2005 as a transition took place from a predominance of fluoroscopic-guided procedures to CT-guided procedures. As this transition has occurred within our practice, we determined that it was important to investigate our pulmonary TFNA practice by calculating our FN rate. We also sought to study the primary cause(s) by reviewing each FN case, blindly, to ascertain if there are interpretive and/or sampling issues and also any potential quality control issues. For the present study, we used a case-control method using multiple observers with the goal of categorizing the errors as either sampling or interpretive, as well as gathering qualitative data about the cases themselves.
Materials and Methods The Mayo Clinic Institution Review Board approved the use of patient records and slides for this study. Medical histories were reviewed from 1,043 consecutive patients who underwent a CT-guided TFNA, CNB, or both, to assess pulmonary nodule(s) from a 5-year time period, January 1, 2003 to December 31, 2007. Mean patient age was 66.2 years (range, 15–94) and the male to female ratio was 53:47. Over this period, the total number of cases diagnosed as positive for malignancy, suspicious for malignancy, atypical, negative for malignancy, and nondiagnostic was 633 (60.6%), 39 (3.7%), 50 (4.8%), 290 (27.8%), and 31 (3.0%), respectively. In our study, FN cases were defined as those having a pathologic diagnosis of negative for malignancy with a malignant outcome related to the initial lesion assessment. The malignant outcome was determined through either a follow-up histologic specimen diagnosed as positive for malignancy or clinical evidence of malignant disease. Such clinical evidence of disease was defined as radio1064
Diagnostic Cytopathology, Vol. 42, No 12
logic, clinical evidence of metastasis or patient management consistent with a malignant diagnosis, solely or in combination. A total of 37 cases met these criteria for inclusion in our study. Of these cases, 36 were available for review with 35 of the 36 cases also having corresponding CNB material. The clinical and/or radiologic impression for 34/35 cases listed carcinoma (either primary or metastatic) in the differential. The remaining case was suspected to be benign. Although a majority of our TFNA case volume had both a cytology smear (without cell blocks) and CNB specimen submitted together under the same case number, for the present study, each specimen was assessed individually and the worst outcome diagnosis from either used for our analysis. Information was also gathered through review of the patients radiologic images attained at the time of the TFNA procedure from our in-house developed clinical radiologic system (QREADS, current version 5.1). The lesion size (cm) and its proximity to the pleura (cm) were measured and recorded using tools available in QREADS. Measurements were further aggregated to determine whether there were notable differences between interpretive and sampling errors when comparing these metrics (Table I). Matched age and gender controls, 15 positive for malignancy cases and 15 negative for malignancy cases, were also identified and included among the review cases with QREADS data retrieved. In addition, two cytotechnologists (S.G.V. and D.J.T.), with experience screening TFNAs of 5 years and 38 years, respectively, aided the study pathologists by performing a prescreen assessment (i.e., marking and providing preliminary qualitative and diagnostic assessments) of all cytology smears in the study. These cases were circulated among three pathologists (M.C.A., M.R.H., and D.R.S.), with levels of experience diagnosing TFNAs ranging from 15 to 25 years, who examined the slides in an independent and blinded fashion to case status and original diagnosis.
Pathologist Review Pathologists reviewed the cytology smear slides and the CNB slides separately and recorded information about the
Diagnostic Cytopathology DOI 10.1002/dc
FALSE-NEGATIVE PULMONARY FNA SPECIMENS
Fig. 1. Screening and data collection algorithm for core biopsy and cytology smear samples. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]
individual cases on separate worksheets (Fig. 1). Any slides from immunohistochemical studies performed as part of the pathologic work-up of the case were also retrieved and made available for pathologist review. Worksheets with consensus and individual data from all pathologist assessments (both FN and control cases) were then analyzed. An FN case was determined to be an interpretive error if at least 2 of the 3 pathologists independently documented and agreed that the diagnosis was either suspicious or positive for malignancy. A sampling error was determined if the same proportion of pathologists (2/3) agreed that there was an absence of malignant cells on the pathology slides (TFNA, CNB, or both).
Results Thirty-six cases (3.5% of our total of 1,043 TFNA case volume) with an FN diagnosis were available for review (specimens collected between January 1, 2003 and December 31, 2007). Interpretive errors accounted for 11 (30.6%) of our total FN volume, with 10 of the 11 diagnosed from the cytology smear material and the other case diagnosed from the CNB material. The other 25 FN cases (69.4%) were considered sampling errors as they were found to be absent of malignant cells by the reviewing pathologists. All positive and negative controls were appropriately assessed, with no FN and no false-positive diagnoses on pathologist review. Interpretive errors consisted of mostly adenocarcinomas (n 5 8), with five primary to the lung and three of metastatic origin. The other three cancer types were primary squamous cell carcinoma, metastatic renal cell carcinoma, and lymphoma. Sampling errors consisted of 15 primary and 10 metastatic lesions. The primary cancers were three adenocarcinomas, three nonsmall cell carcinomas, three lymphomas, two lung cancers (not otherwise specified), one
small cell carcinoma, one bronchioloalveolar carcinoma, one mesothelioma, and one pulmonary artery intimal sarcoma. The metastatic lesions were four adenocarcinomas, two squamous cell carcinomas, two renal cell carcinomas, a myxofibrosarcoma, and a myxopapillary ependymoma. Qualitative data from these cases are detailed in Table I. Sampling error cases tended to be slightly larger (mean; 3.01 cm) than the interpretive error cases (mean; 2.88 cm) and the overall cohort average (mean; 2.89 cm). Interpretive error cases tended to be about the same distance away from the pleura as the entire patient cohort (mean; 1.28 cm vs. mean; 1.19). Sampling error cases were closer to the pleura (mean; 1.02) than both the interpretive errors and the overall. Interpretive error and sampling error cases were much more likely to abut the pleura, 73 and 64%, respectively, than the overall cohort (50%). Necrosis and air-drying artifact were present more frequently in the interpretive error cohort but obscuring blood and air-drying artifact that could preclude a definitive diagnosis were more frequently found in the sampling error cases. We also found minimal differences between our error and control cases when looking at the mean number of core biopsies taken. Sampling error cases seemed more likely to have more abundant pulmonary macrophages than interpretive error cases and the controls. (Fig. 2) However, there was no clear difference between our cases and controls when we looked at the number of bronchial epithelial cells present. Presence of inflammation also did not seem to play a role in whether a case became an interpretive or sampling error (data not shown). Almost all of the interpretive cases (10/ 11) had at least 25 abnormal cells present on the slide but almost all of those (9/10) had less than 100 cells. Table II shows the consensus diagnoses for all of the cases evaluated in our study. Of 36 FN cases, the FNA Diagnostic Cytopathology, Vol. 42, No 12
1065
Diagnostic Cytopathology DOI 10.1002/dc
MINOT ET AL.
Fig. 2. Presence/absence of pulmonary macrophages and bronchial epithelial cells. Abbreviations: PM, pulmonary macrophage; BEC, bronchial epithelial cells. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.] Table II. FNA/CNB Consensus Diagnosis of False-Negative, Positive-Control, and Negative-Control Cases False negative, no. (%) Diagnosis Nondiagnostic Negative for malignancy Atypical Suspicious for malignancy Positive for malignancy Total number of cases
Positive control, no. (%)
Negative control, no. (%)
FNA
CNB
FNA
CNB
FNA
CNB
1 (3) 20 (56) 5 (14) 7 (19) 3 (8) 36
0 (0) 33 (94) 1 (3) 1 (3) 0 (0) 35
0 (0) 0 (0) 0 (0) 2 (13) 13 (87) 15
0 (0) 2 (14) 1 (7) 2 (14) 9 (64) 14
4 (27) 10 (67) 1 (7) 0 (0) 0 (0) 15
0 (0) 14 (100) 0 (0) 0 (0) 0 (0) 14
Abbreviations: CNB, core biopsy; FNA, fine-needle aspiration biopsy.
sample yielded more abnormal cells (10; 28%) with samples diagnosed as either suspicious (7; 19%) or positive for malignancy (3; 8%). Pathologists’ re-review of the 35 CNB samples generated only one interpretive FN result, a suspicious for malignancy diagnosis (1; 3%). An atypical diagnosis was given in 5 (14%) of the 36 FNA samples and 1 (3%) of the 35 CNB samples. FNA samples were either suspicious (2; 13%) or positive for malignancy (13; 87%) in all positive control samples. Fourteen CNB samples from the same positive control cases were suspicious for malignancy (2; 14%), positive for malignancy (9; 64%), atypical (1; 7%), and negative for malignancy (2; 14%). All FNA and CNB negative control samples were negative for malignancy on re-review by our study pathologists.
Discussion An investigation of our pulmonary FNA practice revealed an overall FN rate of 3.5% (including both positive and suspicious cases on review). Of note, in our practice the results of both the FNA and CNB are combined into one report. This combination allows for a convenient way to visualize results of both procedures but can make it a challenge to resolve discrepancies (FNA vs. CNB). Therefore, our FN rate is cal1066
Diagnostic Cytopathology, Vol. 42, No 12
culated from a combined diagnosis, with the worst diagnosis of either the FNA or CNB being the final diagnosis. Our FN rate is in the low range in comparison to recent studies in which suspicious pulmonary lesions are sampled with the assistance of CT-guidance. A limitation to comparing interlaboratory study data is the inconsistent or unclear definition of a “FN” TFNA specimen. Two studies, both performed on similar cohorts of patients in which CT fluoroscopy was used to attain specimens from pulmonary lesions, used a similar follow-up gold standard (histological, clinical, or both) and achieved similar FN rates of 4.1%6 and 4.8%.7 A study by Tsukada et al.9 of CT-guided FNA’s showed a higher FN rate (15.2%); however, the investigators required radiologic confirmation of an increase in lesion size or proven metastasis for follow-up—a more strict definition than in our study and others. One study8 had an FN rate of 6.4%, or almost twice as high as our study, but had a less clear definition of follow-up: “proof of malignancy or obvious clinical outcome.” We considered a diagnosis of “suspicious for malignancy” to be a malignant diagnosis; a high number of our interpretive error cases (8; 73%) received this diagnosis. In our previous study,11 we determined that cases with a
Diagnostic Cytopathology DOI 10.1002/dc
FALSE-NEGATIVE PULMONARY FNA SPECIMENS
suspicious diagnosis had a high likelihood of being malignant on follow-up (a high-positive predictive value), thus we applied the same standard in the present study. Conversely, we also determined that an atypical diagnosis was likely to be negative for malignancy on patient follow-up (a high-negative predictive value), so we were comfortable in considering them to be negative in the present study. Certainly, in other institutions, the same reasoning may not apply. Previous studies have been mixed in regard to how these cases are handled. Some investigators have clearly excluded these cases from analysis altogether,4,12 others have considered them as malignant cases because of a high-positive predictive value.13 Yet, interestingly, in another study they were considered to be negative for malignancy despite having a highpositive predictive value of 87.5%.2 But by far, investigators seem to be more inclined to be unclear as to whether any suspicious cases existed or where they fit into their statistical calculations.3,6,7,9,14–16 This area clearly warrants further investigation with more clear and consistent reporting. To our knowledge, few investigators have attempted to systematically and in a case-control fashion, determine the true source of an FN pulmonary FNA diagnosis. The Memorial Sloan-Kettering Cancer Center performed a study on 45 FN cases obtained from 176 fluoroscopically or CT-guided pulmonary lesion samples.1 Both a cytotechnologist and a pathologist rescreened and reclassified these cases independently and without prior knowledge of the outcome. They found a total FN rate of 46.7% (21/ 45) in that cohort and 95.2% (20 of the 21) were determined to be sampling errors (calculations made from published raw data). Other studies seem to corroborate the assertion that sampling errors seem to be much more common than interpretive errors.2,3,12,14 It was not clear whether negative and positive controls were included to help blind the study authors to case status. Carefully matching negative and positive controls to the FN cases in our study, we believed was the best way to blind the screeners to case status and determine whether features present in FN cases are unique. In our study, we found that the lesion sizes of interpretive error cases (2.88 cm) were no different than the overall average (2.89 cm). However, these cases were more likely to abut the pleura than the overall cohort and the controls, again bringing up the possible difficulty in accessing these lesions. But, we are limited by having 11 interpretive errors making it impossible to draw any far reaching conclusions based on these low numbers. In general, these cases were paucicellular with only rare tumor cells (less than 100 cells) present oftentimes with extensive necrosis. Diagnoses on these cases can vary significantly depending on an individual pathologists experi-
ence and confidence in handling such limited material. Low-grade malignancies as well as unusual metastatic presentations mimicking benign components can also cause an increased chance of misinterpretation. Sampling errors tended to be slightly larger radiographically than the overall average lesion size and also tended to be closer to and more frequently abut the pleura possibly causing them to be difficult to sample. Also, as expected, pathologically there were very few abnormal cells on these slides with the majority of cases containing less than 25 cells. They were more likely to have an abundance (over 50 cells) of pulmonary macrophages (9/ 30) than interpretive error cases (1/11), negative controls (2/15), and positive controls (2/15). The significance of this finding is unclear. Presence of inflammation also did not seem to play a factor in our sampling error or interpretive error cases. Recently, investigators have explored the use of rapid on-site evaluation and the area shows promise.17–19 This could have potentially helped some of the FN cases where cellularity was a confounding factor. Currently, at our institution we regularly perform adequacy assessments on our TFNAs, however, even when informed of a nondiagnostic or negative aspirate, our clinicians rarely rebiopsy patients due to a number of contributing factors such as difficult to access lesions, lesions too small to sample, low suspicion of malignancy or satisfaction with quality of tissue biopsy. Having a second pathologist review pulmonary FNA’s could potentially reduce lab error rates. This would significantly increase workload if all negative or nondiagnostic lung FNAs were reviewed by a second pathologist (over 300 specimens of the 1045 study set representing almost one third of lung FNAs). It is not known if the small number of interpretive errors (11 of 300) would be picked up in a nonstudy clinical practice setting. This has been investigated more in FNAs of the thyroid and with little published in regard to lung FNAs. Some practices might consider a second review, but it is unknown if such a review would be of benefit, considering the fact that paucity of material played such a large role in these errors, and the large number of second reviews that would be required (all negative and nondiagnostic cases). A study by Montaudon et al.8 in which the investigators evaluated the 5-year performance and complication rate of their CT-guided lung biopsy practice, revealed a 6.4% (39/ 605) FN rate. On the basis of multivariate analysis, they determined that lesion size equal to or smaller than 10 mm in diameter was the only variable to reach statistical significance. They found that both lesion distance from the pleura and pleura contact were not statistically significant variables. Other reasons for their FNs included abundant necrosis and the existence of a benign abscess or pseudoinflammatory tumor coexisting with a malignant Diagnostic Cytopathology, Vol. 42, No 12
1067
Diagnostic Cytopathology DOI 10.1002/dc
MINOT ET AL.
tumor in the same lesion. In a smaller, 3-year study in which biopsies were collected under CT-guidance, a FN rate of 15.2% was found, with the discovery that lesion size, too, had a significant role in the FN rate.9 We found in the present study that of the 11 interpretive errors, 8 were from either primary or metastatic adenocarcinoma and the other 3 were from primary squamous cell carcinoma, metastatic renal cell carcinoma, and lymphoma. A diagnosis of adenocarcinoma can sometimes be difficult to determine, particularly when necrosis is abundant, which was present in 45% of our interpretive error cases. Reported mimics of adenocarcinoma include reactive type II pneumocytes or mesothelial cells,10 bronchial metaplasia as well as atypical adenomatous hyperplasia.20 It has also been reported that squamous cell cancers can sometimes lead to a FN diagnosis due to the presence of necrosis and sometimes severe, acute inflammation.21 Mimics of squamous cell carcinoma can sometimes lead to false-positive diagnoses, making these cancers particularly challenging.20,22,23 A limitation to our study is its retrospective design. Information gleaned from reviews of patient demographic, clinical, and diagnostic records can sometimes be incomplete leading to erroneous assumptions. Furthermore, our reviewing pathologists were limited to the material at hand. It is possible that if they were allowed to order additional ancillary studies (i.e., immunohistochemical stains) the diagnoses or even diagnostic categories could have changed for some of the cases. In summary, we conclude that in our pulmonary FNA practice and with our relatively few errors, inadequate tumor sampling seems to be a larger issue than errors related to incorrect interpretations. The use of CT has led to the biopsying of smaller lesions, many of which prove to be benign. However, in an early primary malignancy or metastasis in which the malignant lesions are very small, adequate material can be difficult to collect particularly when the tumors abut the pleura. Diagnostically, it seems necrosis can be a factor leading to incorrect diagnoses. As radiological methods continue to advance, clinicians need to be cognizant of the anatomic limitations to the collection of biologic material as well as the presence of extracellular material that can hinder pathologic diagnoses.
References
3. Winning AJ, McIvor J, Seed WA, Husain OA, Metaxas N. Interpretation of negative results in fine needle aspiration of discrete pulmonary lesions. Thorax 1986;41:875–879. 4. Calhoun P, Feldman PS, Armstrong P, et al. The clinical outcome of needle aspirations of the lung when cancer is not diagnosed. Ann Thorac Surg 1986;41:592–596. 5. Johnston WW. Percutaneous fine needle aspiration biopsy of the lung. A study of 1,015 patients. Acta Cytol 1984;28:218–224. 6. Hiraki T, Mimura H, Gobara H, et al. CT fluoroscopy-guided biopsy of 1,000 pulmonary lesions performed with 20-gauge coaxial cutting needles: Diagnostic yield and risk factors for diagnostic failure. Chest 2009;136:1612–1617. 7. Matsui Y, Hiraki T, Mimura H, et al. Role of computed tomography fluoroscopy-guided cutting needle biopsy of lung lesions after transbronchial examination resulting in negative diagnosis. Clin Lung Cancer 2011;12:51–55. 8. Montaudon M, Latrabe V, Pariente A, Corneloup O, Begueret H, Laurent F. Factors influencing accuracy of CT-guided percutaneous biopsies of pulmonary lesions. Eur Radiol 2004;14:1234–1240. 9. Tsukada H, Satou T, Iwashima A, Souma T. Diagnostic accuracy of CT-guided automated needle biopsy of lung nodules. AJR Am J Roentgenol 2000;175:239–243. 10. Saad RS, Silverman JF. Respiratory cytology: Differential diagnosis and pitfalls. Diagn Cytopathol 2009;38:297–307. 11. Minot DM, Jaben E, Aubry MC, et al. Evolution of transthoracic fine needle aspiration and core needle biopsy practice: A comparison of two time periods, 1996–1998 and 2003–2005. Diagn Cytopathol 2012;40:876–881. 12. Afify A, Davila RM. Pulmonary fine needle aspiration biopsy. Assessing the negative diagnosis. Acta Cytol 1999;43:601–604. 13. Khouri NF, Stitik FP, Erozan YS, et al. Transthoracic needle aspiration biopsy of benign and malignant lung lesions. AJR Am J Roentgenol 1985;144:281–288. 14. Cagle PT, Kovach M, Ramzy I. Causes of false results in transthoracic fine needle lung aspirates. Acta Cytol 1993;37:16–20. 15. Gobien RP, Valicenti JF, Paris BS, Daniell C. Thin-needle aspiration biopsy: Methods of increasing the accuracy of a negative prediction. Radiology 1982;145:603–605. 16. Priola AM, Priola SM, Cataldi A, et al. Diagnostic accuracy and complication rate of CT-guided fine needle aspiration biopsy of lung lesions: A study based on the experience of the cytopathologist. Acta Radiol 2010;51:527–533. 17. Collins BT, Chen AC, Wang JF, Bernadt CT, Sanati S. Improved laboratory resource utilization and patient care with the use of rapid on-site evaluation for endobronchial ultrasound fine-needle aspiration biopsy. Cancer Cytopathol 2013;121:544–551. 18. Ecka RS, Sharma M. Rapid on-site evaluation of EUS-FNA by cytopathologist: An experience of a tertiary hospital. Diagn Cytopathol 2013;41:1075–1080. 19. Mondoni M, Carlucci P, Di Marco F, et al. Rapid on-site evaluation improves needle aspiration sensitivity in the diagnosis of central lung cancers: A randomized trial. Respiration 2013; 86:52–58. 20. Butnor KJ. Avoiding underdiagnosis, overdiagnosis, and misdiagnosis of lung carcinoma. Arch Pathol Lab Med 2008;132:1118–1132.
1. Zakowski MF, Gatscha RM, Zaman MB. Negative predictive value of pulmonary fine needle aspiration cytology. Acta Cytol 1992;36: 283–286.
21. Silverman JF. Inflammatory and neoplastic processes of the lung: Differential diagnosis and pitfalls in FNA biopsies. Diagn Cytopathol 1995;13:448–462.
2. Caya JG, Clowry LJ, Wollenberg NJ, Tieu TM. Transthoracic fineneedle aspiration cytology. Analysis of 82 patients with detailed verification criteria and evaluation of false-negative cases. Am J Clin Pathol 1984;82:100–103.
22. Idowu MO, Powers CN. Lung cancer cytology: Potential pitfalls and mimics—A review. Int J Clin Exp Pathol 2010;3:367–385.
1068
Diagnostic Cytopathology, Vol. 42, No 12
23. Crapanzano JP, Zakowski MF. Diagnostic dilemmas in pulmonary cytology. Cancer 2001;93:364–375.
Transthoracic fine needle aspiration (TFNA)/core needle biopsy (CNB) under computed tomography (CT) guidance has proved useful in the assessment of pulmonary nodules. We sought to determine the TFNA false-negative (FN) rate at our institution and identify potential causes of FN diagnoses. Medical records were reviewed from 1,043 consecutive patients who underwent CT-guided TFNA with or without CNB of lung nodules over a 5-year time period (2003–2007). Thirty-seven FN cases of “negative” TFNA/CNB with malignant outcome were identified with 36 cases available for review, of which 35 had a corresponding CNB. Cases were reviewed independently (blinded to original diagnosis) by three pathologists with 15 age- and sexmatched positive and negative controls. Diagnosis (i.e., nondiagnostic, negative or positive for malignancy, atypical or suspicious) and qualitative assessments were recorded. Consensus diagnosis was suspicious or positive in 10 (28%) of 36 TFNA cases and suspicious in 1 (3%) of 35 CNB cases, indicating potential interpretive errors. Of the 11 interpretive errors (including both suspicious and positive cases), 8 were adenocarcinomas, 1 squamous cell carcinoma, 1 metastatic renal cell carcinoma, and 1 lymphoma. The remaining 25 FN cases (69.4%) were considered sampling errors and consisted of 7 adenocarcinomas, 3 nonsmall cell carcinomas, 3 lymphomas, 2 squamous cell carcinomas, and 2 renal cell carcinomas. Inter-
1 Department of Laboratory Medicine and Pathology, Mayo Clinic, Rochester, Minnesota 2 Department of Laboratory Medicine and Pathology, Mayo Clinic, Scottsdale, Arizona 3 Department of Radiology, Consulting Radiologists, Minneapolis, Minnesota 4 Department of Radiology, Mayo Clinic, Rochester, Minnesota *Correspondence to: Amy C. Clayton, M.D., Mayo Clinic, 200 First Street SW, Hilton Building 11th Floor, Rochester, MN 55905. E-mail: [email protected] Received 5 July 2013; Accepted 2 May 2014 DOI: 10.1002/dc.23169 Published online 28 May 2014 in Wiley Online Library (wileyonlinelibrary.com).
C 2014 WILEY PERIODICALS, INC. V
pretive and sampling error cases were more likely to abut the pleura, while histopathologically, they tended to be necrotic and air-dried. The overall FN rate in this patient cohort is 3.5% (1.1% interpretive and 2.4% sampling errors). Diagn. Cytopathol. 2014;42:1063–1068. VC 2014 Wiley Periodicals, Inc. Key Words: lung; neoplasm; computed tomography; diagnosis; cytopathology
Advances in the field of diagnostic radiology, including the use of computed tomography (CT), have led to the visualization and attempted assessment of smaller and smaller indeterminate pulmonary nodules. The use of the transthoracic needle biopsy technique on these pulmonary nodules has proved to be a safe, highly accurate, and minimally invasive method to determine the benign or malignant nature of these lesions. Specimens from these transthoracic fine needle aspiration (TFNA) and core needle biopsy (CNB) procedures are submitted for analysis by clinicians, oftentimes without knowledge of their laboratory false-negative (FN) rates. Some early institutional studies have reported relatively high FN rates on biopsy specimens collected using this technique, ranging from 9.6 to 45.6%.1–5 In more recent studies in which biopsies are primarily collected under CT-guidance, FN rates have ranged from 4.1 to 15.2%.6–9 The investigators’ predominant explanation for these FNs is that they are mostly a result of sampling error, or failure to collect adequate biologic material for accurate pathologic assessment. These findings have led some experts to conclude that the majority of FN diagnoses are usually attributed to sampling, rather than interpretive errors.10 As the sensitivity to detect pulmonary lesions via radiologic methods has improved, the Diagnostic Cytopathology, Vol. 42, No 12
1063
Diagnostic Cytopathology DOI 10.1002/dc
MINOT ET AL. Table I. Lung Lesion Characteristics and Specimen Qualitative Features
Mean size (cm) Lesion abuts pleura (%) Mean distance from pleura (cm) Necrosis (%) Bloody (%) Air-drying artifact (%) Precluding definitive diagnosis (%) Number of core biopsies, mean Mean total size of cores (length; mm)a a
Interpretive error
Sampling error
Negative control
Positive control
Overall
2.88 73 1.28 45 0 45 15 4.8 8.64
3.01 64 1.02 16 8 32 30 4.0 7.67
3.55 43 0.83 18 16 47 20 5.1 13.11
2.83 47 1.47 31 4 44 22 4.5 8.63
2.89 50 1.19 – – – – – –
Width of cores were mostly 0.5 mm.
debate becomes about what impact, if any, these methods have on the sensitivity of the pathologic assessment of these smaller nodules. A previous article from our group11 noted a small improvement in sensitivity (94 vs. 91%) and negative predictive value (80 vs. 74%) when we examined our pulmonary practice between the years 1996–1998 and 2003–2005 as a transition took place from a predominance of fluoroscopic-guided procedures to CT-guided procedures. As this transition has occurred within our practice, we determined that it was important to investigate our pulmonary TFNA practice by calculating our FN rate. We also sought to study the primary cause(s) by reviewing each FN case, blindly, to ascertain if there are interpretive and/or sampling issues and also any potential quality control issues. For the present study, we used a case-control method using multiple observers with the goal of categorizing the errors as either sampling or interpretive, as well as gathering qualitative data about the cases themselves.
Materials and Methods The Mayo Clinic Institution Review Board approved the use of patient records and slides for this study. Medical histories were reviewed from 1,043 consecutive patients who underwent a CT-guided TFNA, CNB, or both, to assess pulmonary nodule(s) from a 5-year time period, January 1, 2003 to December 31, 2007. Mean patient age was 66.2 years (range, 15–94) and the male to female ratio was 53:47. Over this period, the total number of cases diagnosed as positive for malignancy, suspicious for malignancy, atypical, negative for malignancy, and nondiagnostic was 633 (60.6%), 39 (3.7%), 50 (4.8%), 290 (27.8%), and 31 (3.0%), respectively. In our study, FN cases were defined as those having a pathologic diagnosis of negative for malignancy with a malignant outcome related to the initial lesion assessment. The malignant outcome was determined through either a follow-up histologic specimen diagnosed as positive for malignancy or clinical evidence of malignant disease. Such clinical evidence of disease was defined as radio1064
Diagnostic Cytopathology, Vol. 42, No 12
logic, clinical evidence of metastasis or patient management consistent with a malignant diagnosis, solely or in combination. A total of 37 cases met these criteria for inclusion in our study. Of these cases, 36 were available for review with 35 of the 36 cases also having corresponding CNB material. The clinical and/or radiologic impression for 34/35 cases listed carcinoma (either primary or metastatic) in the differential. The remaining case was suspected to be benign. Although a majority of our TFNA case volume had both a cytology smear (without cell blocks) and CNB specimen submitted together under the same case number, for the present study, each specimen was assessed individually and the worst outcome diagnosis from either used for our analysis. Information was also gathered through review of the patients radiologic images attained at the time of the TFNA procedure from our in-house developed clinical radiologic system (QREADS, current version 5.1). The lesion size (cm) and its proximity to the pleura (cm) were measured and recorded using tools available in QREADS. Measurements were further aggregated to determine whether there were notable differences between interpretive and sampling errors when comparing these metrics (Table I). Matched age and gender controls, 15 positive for malignancy cases and 15 negative for malignancy cases, were also identified and included among the review cases with QREADS data retrieved. In addition, two cytotechnologists (S.G.V. and D.J.T.), with experience screening TFNAs of 5 years and 38 years, respectively, aided the study pathologists by performing a prescreen assessment (i.e., marking and providing preliminary qualitative and diagnostic assessments) of all cytology smears in the study. These cases were circulated among three pathologists (M.C.A., M.R.H., and D.R.S.), with levels of experience diagnosing TFNAs ranging from 15 to 25 years, who examined the slides in an independent and blinded fashion to case status and original diagnosis.
Pathologist Review Pathologists reviewed the cytology smear slides and the CNB slides separately and recorded information about the
Diagnostic Cytopathology DOI 10.1002/dc
FALSE-NEGATIVE PULMONARY FNA SPECIMENS
Fig. 1. Screening and data collection algorithm for core biopsy and cytology smear samples. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]
individual cases on separate worksheets (Fig. 1). Any slides from immunohistochemical studies performed as part of the pathologic work-up of the case were also retrieved and made available for pathologist review. Worksheets with consensus and individual data from all pathologist assessments (both FN and control cases) were then analyzed. An FN case was determined to be an interpretive error if at least 2 of the 3 pathologists independently documented and agreed that the diagnosis was either suspicious or positive for malignancy. A sampling error was determined if the same proportion of pathologists (2/3) agreed that there was an absence of malignant cells on the pathology slides (TFNA, CNB, or both).
Results Thirty-six cases (3.5% of our total of 1,043 TFNA case volume) with an FN diagnosis were available for review (specimens collected between January 1, 2003 and December 31, 2007). Interpretive errors accounted for 11 (30.6%) of our total FN volume, with 10 of the 11 diagnosed from the cytology smear material and the other case diagnosed from the CNB material. The other 25 FN cases (69.4%) were considered sampling errors as they were found to be absent of malignant cells by the reviewing pathologists. All positive and negative controls were appropriately assessed, with no FN and no false-positive diagnoses on pathologist review. Interpretive errors consisted of mostly adenocarcinomas (n 5 8), with five primary to the lung and three of metastatic origin. The other three cancer types were primary squamous cell carcinoma, metastatic renal cell carcinoma, and lymphoma. Sampling errors consisted of 15 primary and 10 metastatic lesions. The primary cancers were three adenocarcinomas, three nonsmall cell carcinomas, three lymphomas, two lung cancers (not otherwise specified), one
small cell carcinoma, one bronchioloalveolar carcinoma, one mesothelioma, and one pulmonary artery intimal sarcoma. The metastatic lesions were four adenocarcinomas, two squamous cell carcinomas, two renal cell carcinomas, a myxofibrosarcoma, and a myxopapillary ependymoma. Qualitative data from these cases are detailed in Table I. Sampling error cases tended to be slightly larger (mean; 3.01 cm) than the interpretive error cases (mean; 2.88 cm) and the overall cohort average (mean; 2.89 cm). Interpretive error cases tended to be about the same distance away from the pleura as the entire patient cohort (mean; 1.28 cm vs. mean; 1.19). Sampling error cases were closer to the pleura (mean; 1.02) than both the interpretive errors and the overall. Interpretive error and sampling error cases were much more likely to abut the pleura, 73 and 64%, respectively, than the overall cohort (50%). Necrosis and air-drying artifact were present more frequently in the interpretive error cohort but obscuring blood and air-drying artifact that could preclude a definitive diagnosis were more frequently found in the sampling error cases. We also found minimal differences between our error and control cases when looking at the mean number of core biopsies taken. Sampling error cases seemed more likely to have more abundant pulmonary macrophages than interpretive error cases and the controls. (Fig. 2) However, there was no clear difference between our cases and controls when we looked at the number of bronchial epithelial cells present. Presence of inflammation also did not seem to play a role in whether a case became an interpretive or sampling error (data not shown). Almost all of the interpretive cases (10/ 11) had at least 25 abnormal cells present on the slide but almost all of those (9/10) had less than 100 cells. Table II shows the consensus diagnoses for all of the cases evaluated in our study. Of 36 FN cases, the FNA Diagnostic Cytopathology, Vol. 42, No 12
1065
Diagnostic Cytopathology DOI 10.1002/dc
MINOT ET AL.
Fig. 2. Presence/absence of pulmonary macrophages and bronchial epithelial cells. Abbreviations: PM, pulmonary macrophage; BEC, bronchial epithelial cells. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.] Table II. FNA/CNB Consensus Diagnosis of False-Negative, Positive-Control, and Negative-Control Cases False negative, no. (%) Diagnosis Nondiagnostic Negative for malignancy Atypical Suspicious for malignancy Positive for malignancy Total number of cases
Positive control, no. (%)
Negative control, no. (%)
FNA
CNB
FNA
CNB
FNA
CNB
1 (3) 20 (56) 5 (14) 7 (19) 3 (8) 36
0 (0) 33 (94) 1 (3) 1 (3) 0 (0) 35
0 (0) 0 (0) 0 (0) 2 (13) 13 (87) 15
0 (0) 2 (14) 1 (7) 2 (14) 9 (64) 14
4 (27) 10 (67) 1 (7) 0 (0) 0 (0) 15
0 (0) 14 (100) 0 (0) 0 (0) 0 (0) 14
Abbreviations: CNB, core biopsy; FNA, fine-needle aspiration biopsy.
sample yielded more abnormal cells (10; 28%) with samples diagnosed as either suspicious (7; 19%) or positive for malignancy (3; 8%). Pathologists’ re-review of the 35 CNB samples generated only one interpretive FN result, a suspicious for malignancy diagnosis (1; 3%). An atypical diagnosis was given in 5 (14%) of the 36 FNA samples and 1 (3%) of the 35 CNB samples. FNA samples were either suspicious (2; 13%) or positive for malignancy (13; 87%) in all positive control samples. Fourteen CNB samples from the same positive control cases were suspicious for malignancy (2; 14%), positive for malignancy (9; 64%), atypical (1; 7%), and negative for malignancy (2; 14%). All FNA and CNB negative control samples were negative for malignancy on re-review by our study pathologists.
Discussion An investigation of our pulmonary FNA practice revealed an overall FN rate of 3.5% (including both positive and suspicious cases on review). Of note, in our practice the results of both the FNA and CNB are combined into one report. This combination allows for a convenient way to visualize results of both procedures but can make it a challenge to resolve discrepancies (FNA vs. CNB). Therefore, our FN rate is cal1066
Diagnostic Cytopathology, Vol. 42, No 12
culated from a combined diagnosis, with the worst diagnosis of either the FNA or CNB being the final diagnosis. Our FN rate is in the low range in comparison to recent studies in which suspicious pulmonary lesions are sampled with the assistance of CT-guidance. A limitation to comparing interlaboratory study data is the inconsistent or unclear definition of a “FN” TFNA specimen. Two studies, both performed on similar cohorts of patients in which CT fluoroscopy was used to attain specimens from pulmonary lesions, used a similar follow-up gold standard (histological, clinical, or both) and achieved similar FN rates of 4.1%6 and 4.8%.7 A study by Tsukada et al.9 of CT-guided FNA’s showed a higher FN rate (15.2%); however, the investigators required radiologic confirmation of an increase in lesion size or proven metastasis for follow-up—a more strict definition than in our study and others. One study8 had an FN rate of 6.4%, or almost twice as high as our study, but had a less clear definition of follow-up: “proof of malignancy or obvious clinical outcome.” We considered a diagnosis of “suspicious for malignancy” to be a malignant diagnosis; a high number of our interpretive error cases (8; 73%) received this diagnosis. In our previous study,11 we determined that cases with a
Diagnostic Cytopathology DOI 10.1002/dc
FALSE-NEGATIVE PULMONARY FNA SPECIMENS
suspicious diagnosis had a high likelihood of being malignant on follow-up (a high-positive predictive value), thus we applied the same standard in the present study. Conversely, we also determined that an atypical diagnosis was likely to be negative for malignancy on patient follow-up (a high-negative predictive value), so we were comfortable in considering them to be negative in the present study. Certainly, in other institutions, the same reasoning may not apply. Previous studies have been mixed in regard to how these cases are handled. Some investigators have clearly excluded these cases from analysis altogether,4,12 others have considered them as malignant cases because of a high-positive predictive value.13 Yet, interestingly, in another study they were considered to be negative for malignancy despite having a highpositive predictive value of 87.5%.2 But by far, investigators seem to be more inclined to be unclear as to whether any suspicious cases existed or where they fit into their statistical calculations.3,6,7,9,14–16 This area clearly warrants further investigation with more clear and consistent reporting. To our knowledge, few investigators have attempted to systematically and in a case-control fashion, determine the true source of an FN pulmonary FNA diagnosis. The Memorial Sloan-Kettering Cancer Center performed a study on 45 FN cases obtained from 176 fluoroscopically or CT-guided pulmonary lesion samples.1 Both a cytotechnologist and a pathologist rescreened and reclassified these cases independently and without prior knowledge of the outcome. They found a total FN rate of 46.7% (21/ 45) in that cohort and 95.2% (20 of the 21) were determined to be sampling errors (calculations made from published raw data). Other studies seem to corroborate the assertion that sampling errors seem to be much more common than interpretive errors.2,3,12,14 It was not clear whether negative and positive controls were included to help blind the study authors to case status. Carefully matching negative and positive controls to the FN cases in our study, we believed was the best way to blind the screeners to case status and determine whether features present in FN cases are unique. In our study, we found that the lesion sizes of interpretive error cases (2.88 cm) were no different than the overall average (2.89 cm). However, these cases were more likely to abut the pleura than the overall cohort and the controls, again bringing up the possible difficulty in accessing these lesions. But, we are limited by having 11 interpretive errors making it impossible to draw any far reaching conclusions based on these low numbers. In general, these cases were paucicellular with only rare tumor cells (less than 100 cells) present oftentimes with extensive necrosis. Diagnoses on these cases can vary significantly depending on an individual pathologists experi-
ence and confidence in handling such limited material. Low-grade malignancies as well as unusual metastatic presentations mimicking benign components can also cause an increased chance of misinterpretation. Sampling errors tended to be slightly larger radiographically than the overall average lesion size and also tended to be closer to and more frequently abut the pleura possibly causing them to be difficult to sample. Also, as expected, pathologically there were very few abnormal cells on these slides with the majority of cases containing less than 25 cells. They were more likely to have an abundance (over 50 cells) of pulmonary macrophages (9/ 30) than interpretive error cases (1/11), negative controls (2/15), and positive controls (2/15). The significance of this finding is unclear. Presence of inflammation also did not seem to play a factor in our sampling error or interpretive error cases. Recently, investigators have explored the use of rapid on-site evaluation and the area shows promise.17–19 This could have potentially helped some of the FN cases where cellularity was a confounding factor. Currently, at our institution we regularly perform adequacy assessments on our TFNAs, however, even when informed of a nondiagnostic or negative aspirate, our clinicians rarely rebiopsy patients due to a number of contributing factors such as difficult to access lesions, lesions too small to sample, low suspicion of malignancy or satisfaction with quality of tissue biopsy. Having a second pathologist review pulmonary FNA’s could potentially reduce lab error rates. This would significantly increase workload if all negative or nondiagnostic lung FNAs were reviewed by a second pathologist (over 300 specimens of the 1045 study set representing almost one third of lung FNAs). It is not known if the small number of interpretive errors (11 of 300) would be picked up in a nonstudy clinical practice setting. This has been investigated more in FNAs of the thyroid and with little published in regard to lung FNAs. Some practices might consider a second review, but it is unknown if such a review would be of benefit, considering the fact that paucity of material played such a large role in these errors, and the large number of second reviews that would be required (all negative and nondiagnostic cases). A study by Montaudon et al.8 in which the investigators evaluated the 5-year performance and complication rate of their CT-guided lung biopsy practice, revealed a 6.4% (39/ 605) FN rate. On the basis of multivariate analysis, they determined that lesion size equal to or smaller than 10 mm in diameter was the only variable to reach statistical significance. They found that both lesion distance from the pleura and pleura contact were not statistically significant variables. Other reasons for their FNs included abundant necrosis and the existence of a benign abscess or pseudoinflammatory tumor coexisting with a malignant Diagnostic Cytopathology, Vol. 42, No 12
1067
Diagnostic Cytopathology DOI 10.1002/dc
MINOT ET AL.
tumor in the same lesion. In a smaller, 3-year study in which biopsies were collected under CT-guidance, a FN rate of 15.2% was found, with the discovery that lesion size, too, had a significant role in the FN rate.9 We found in the present study that of the 11 interpretive errors, 8 were from either primary or metastatic adenocarcinoma and the other 3 were from primary squamous cell carcinoma, metastatic renal cell carcinoma, and lymphoma. A diagnosis of adenocarcinoma can sometimes be difficult to determine, particularly when necrosis is abundant, which was present in 45% of our interpretive error cases. Reported mimics of adenocarcinoma include reactive type II pneumocytes or mesothelial cells,10 bronchial metaplasia as well as atypical adenomatous hyperplasia.20 It has also been reported that squamous cell cancers can sometimes lead to a FN diagnosis due to the presence of necrosis and sometimes severe, acute inflammation.21 Mimics of squamous cell carcinoma can sometimes lead to false-positive diagnoses, making these cancers particularly challenging.20,22,23 A limitation to our study is its retrospective design. Information gleaned from reviews of patient demographic, clinical, and diagnostic records can sometimes be incomplete leading to erroneous assumptions. Furthermore, our reviewing pathologists were limited to the material at hand. It is possible that if they were allowed to order additional ancillary studies (i.e., immunohistochemical stains) the diagnoses or even diagnostic categories could have changed for some of the cases. In summary, we conclude that in our pulmonary FNA practice and with our relatively few errors, inadequate tumor sampling seems to be a larger issue than errors related to incorrect interpretations. The use of CT has led to the biopsying of smaller lesions, many of which prove to be benign. However, in an early primary malignancy or metastasis in which the malignant lesions are very small, adequate material can be difficult to collect particularly when the tumors abut the pleura. Diagnostically, it seems necrosis can be a factor leading to incorrect diagnoses. As radiological methods continue to advance, clinicians need to be cognizant of the anatomic limitations to the collection of biologic material as well as the presence of extracellular material that can hinder pathologic diagnoses.
References
3. Winning AJ, McIvor J, Seed WA, Husain OA, Metaxas N. Interpretation of negative results in fine needle aspiration of discrete pulmonary lesions. Thorax 1986;41:875–879. 4. Calhoun P, Feldman PS, Armstrong P, et al. The clinical outcome of needle aspirations of the lung when cancer is not diagnosed. Ann Thorac Surg 1986;41:592–596. 5. Johnston WW. Percutaneous fine needle aspiration biopsy of the lung. A study of 1,015 patients. Acta Cytol 1984;28:218–224. 6. Hiraki T, Mimura H, Gobara H, et al. CT fluoroscopy-guided biopsy of 1,000 pulmonary lesions performed with 20-gauge coaxial cutting needles: Diagnostic yield and risk factors for diagnostic failure. Chest 2009;136:1612–1617. 7. Matsui Y, Hiraki T, Mimura H, et al. Role of computed tomography fluoroscopy-guided cutting needle biopsy of lung lesions after transbronchial examination resulting in negative diagnosis. Clin Lung Cancer 2011;12:51–55. 8. Montaudon M, Latrabe V, Pariente A, Corneloup O, Begueret H, Laurent F. Factors influencing accuracy of CT-guided percutaneous biopsies of pulmonary lesions. Eur Radiol 2004;14:1234–1240. 9. Tsukada H, Satou T, Iwashima A, Souma T. Diagnostic accuracy of CT-guided automated needle biopsy of lung nodules. AJR Am J Roentgenol 2000;175:239–243. 10. Saad RS, Silverman JF. Respiratory cytology: Differential diagnosis and pitfalls. Diagn Cytopathol 2009;38:297–307. 11. Minot DM, Jaben E, Aubry MC, et al. Evolution of transthoracic fine needle aspiration and core needle biopsy practice: A comparison of two time periods, 1996–1998 and 2003–2005. Diagn Cytopathol 2012;40:876–881. 12. Afify A, Davila RM. Pulmonary fine needle aspiration biopsy. Assessing the negative diagnosis. Acta Cytol 1999;43:601–604. 13. Khouri NF, Stitik FP, Erozan YS, et al. Transthoracic needle aspiration biopsy of benign and malignant lung lesions. AJR Am J Roentgenol 1985;144:281–288. 14. Cagle PT, Kovach M, Ramzy I. Causes of false results in transthoracic fine needle lung aspirates. Acta Cytol 1993;37:16–20. 15. Gobien RP, Valicenti JF, Paris BS, Daniell C. Thin-needle aspiration biopsy: Methods of increasing the accuracy of a negative prediction. Radiology 1982;145:603–605. 16. Priola AM, Priola SM, Cataldi A, et al. Diagnostic accuracy and complication rate of CT-guided fine needle aspiration biopsy of lung lesions: A study based on the experience of the cytopathologist. Acta Radiol 2010;51:527–533. 17. Collins BT, Chen AC, Wang JF, Bernadt CT, Sanati S. Improved laboratory resource utilization and patient care with the use of rapid on-site evaluation for endobronchial ultrasound fine-needle aspiration biopsy. Cancer Cytopathol 2013;121:544–551. 18. Ecka RS, Sharma M. Rapid on-site evaluation of EUS-FNA by cytopathologist: An experience of a tertiary hospital. Diagn Cytopathol 2013;41:1075–1080. 19. Mondoni M, Carlucci P, Di Marco F, et al. Rapid on-site evaluation improves needle aspiration sensitivity in the diagnosis of central lung cancers: A randomized trial. Respiration 2013; 86:52–58. 20. Butnor KJ. Avoiding underdiagnosis, overdiagnosis, and misdiagnosis of lung carcinoma. Arch Pathol Lab Med 2008;132:1118–1132.
1. Zakowski MF, Gatscha RM, Zaman MB. Negative predictive value of pulmonary fine needle aspiration cytology. Acta Cytol 1992;36: 283–286.
21. Silverman JF. Inflammatory and neoplastic processes of the lung: Differential diagnosis and pitfalls in FNA biopsies. Diagn Cytopathol 1995;13:448–462.
2. Caya JG, Clowry LJ, Wollenberg NJ, Tieu TM. Transthoracic fineneedle aspiration cytology. Analysis of 82 patients with detailed verification criteria and evaluation of false-negative cases. Am J Clin Pathol 1984;82:100–103.
22. Idowu MO, Powers CN. Lung cancer cytology: Potential pitfalls and mimics—A review. Int J Clin Exp Pathol 2010;3:367–385.
1068
Diagnostic Cytopathology, Vol. 42, No 12
23. Crapanzano JP, Zakowski MF. Diagnostic dilemmas in pulmonary cytology. Cancer 2001;93:364–375.
