Equivocal cytology in lung cancer diagnosis

Improvement of diagnostic accuracy using adjuvant multicolor FISH, DNA-image cytometry, and quantitative promoter hypermethylation analysis

Authors


Abstract

BACKGROUND:

Sometimes, cytological lung cancer diagnosis is challenging because equivocal diagnoses are common. To enhance diagnostic accuracy, fluorescent in situ hybridization (FISH), DNA-image cytometry, and quantitative promoter hypermethylation analysis have been proposed as adjuncts.

METHODS:

Bronchial washings and/or brushings or transbronchial fine-needle aspiration biopsies were prospectively collected from patients who were clinically suspected of having lung carcinoma. After routine cytological diagnosis, 70 consecutive specimens, each cytologically diagnosed as negative, equivocal, or positive for cancer cells, were investigated with adjuvant methods. Suspicious areas on the smears were restained with the LAVysion multicolor FISH probe set (Abbott Molecular, Des Plaines, Illinois) or according to the Feulgen Staining Method for DNA-image cytometry analysis. DNA was extracted from residual liquid material, and frequencies of aberrant methylation of APC, p16INK4A, and RASSF1A gene promoters were determined with quantitative methylation-specific polymerase chain reaction (QMSP) after bisulfite conversion. Clinical and histological follow-up according to a reference standard, defined in advance, were available for 198 of 210 patients.

RESULTS:

In the whole cohort, cytology, FISH, DNA-image cytometry, and QMSP achieved sensitivities of 83.7%, 78%, 79%, and 49.6%, respectively (specificities of 69.8%, 98.2%, 98.2%, and 98.4%, respectively). Subsequent to cytologically equivocal diagnoses, FISH, DNA-image cytometry, and QMSP definitely identified malignancy in 79%, 83%, and 49%, respectively. With QMSP, 4 of 22 cancer patients with cytologically negative diagnoses were correctly identified.

CONCLUSIONS:

Thus, adjuvant FISH or DNA-image cytometry in cytologically equivocal diagnoses improves diagnostic accuracy at comparable rates. Adjuvant QMSP in cytologically negative cases with persistent suspicion of lung cancer would enhance sensitivity. Cancer (Cancer Cytopathol) 2011;. © 2011 American Cancer Society.

Lung cancer is 1 of the most frequent causes of death worldwide. It is estimated that in 2008, 161,840 people in the United States and 342,000 in Europe died from lung cancer.1, 2 More than 77% of the patients in the United States are in the late stages of the disease with metastases to lymph nodes and distant sites at diagnosis, as often early symptoms are missing.3 In recent years, several attempts have been conducted to improve an early or accelerated diagnosis because patients in early lung cancer stages have a better overall prognosis after therapy.4 Diagnosis with physical examination, chest x-ray, spiral computed tomography (CT), and bronchoscopically obtained histological or cytological specimens represent, in most cases, the first attempt to confirm suspected lung cancer. Cytological methods include investigation of sputa, bronchial washings, bronchial brushings, and transbronchial or transthoracic fine-needle aspiration biopsies.5 Because lung cancer is often diagnosed in late stages, operative therapy is not always recommended, and final diagnosis is not seldom solely based on cytology.6 Sometimes benign and malignant lesions cannot be discriminated by morphology with certainty, and reactive changes of bronchial and alveolar epithelium, air-drying artifacts during smear preparation, and poorly preserved specimens sometimes impede or even render a distinct cytological diagnosis impossible.7-10 Accordingly, cytology sometimes leaves an equivocal (2.9% of cases in a single institution; 8.6% from a single hospital during a 9-month period in our institution) or inconclusive result even in the hands of experienced observers.5, 11, 12 To prevent repeated diagnostic efforts and potentially harmful invasive diagnostic procedures, it is essential to enhance diagnostic accuracy in these cases. The desired result is a definite positive or negative diagnosis of malignancy.

Based on the hypothesis that (chromosomal) aneuploidy essentially contributes to tumorigenesis,13-17 numerical chromosomal aberrations can be detected in cancer cells with fluorescent in situ hybridization (FISH). The LAVysion (Abbott Molecular, Des Plaines, Illinois) multicolor FISH probe18 has been previously used for the early detection of lung cancer on cytological specimens.11, 19-26 DNA-image cytometry uses the detection of DNA aneuploidy via an abnormal cellular DNA content after stoichiometric staining of DNA according to the Feulgen staining method.27, 28 Diagnostic application in pulmonary pathology has been reported for identification of prospective malignant lesions (ie, dysplasia) and prediction of prognosis in manifest cancers.29-35. Both, FISH and DNA-image cytometry can be performed on the same specimen subsequent to a cytological diagnosis.11, 18, 21, 30, 32 There is no need for additional material, which means no further stress for the patient.

Promoter hypermethylation is a major mechanism of tumor suppressor gene inactivation in lung cancer and can be used as a biomarker for early detection.36-39 Panels of aberrantly methylated gene promoters, investigated with quantitative methylation-specific real-time polymerase chain reaction (QMSP) can be used as biomarkers for the detection of lung cancer on residual liquid material from regular diagnostic cytology specimen collection.40

The aim of our prospective cohort study was to compare the potential benefit for diagnostic accuracy on pulmonary cytology of LAVysion multicolor FISH, DNA-image cytometry, and a panel of aberrantly methylated tumor suppressor genes with QMSP.40 The study intended to determine a diagnostic algorithm for the application of these methods, especially in cytologically equivocal cases. Each method was correlated with a predefined reference standard. In a second step, all adjuvant methods were directly compared on the same specimen for their diagnostic power.

MATERIALS AND METHODS

Diploidy is defined as a 2-fold chromosomal set. Euploid polyploidy, in this context, means 2n chromosomal sets (including tetraploidy). Aneuploidy means a chromosomal set ≠2n, which is because integrated-value multiples of single chromosomal sets, apart from 2n, do not occur in non-neoplastic tissues.

Patient Selection and Study Design

The study was approved by the local ethics committee. Bronchoscopically obtained diagnostic material on 843 consecutive patients with suspected lung cancer from the Florence Nightingale Hospital in Düsseldorf, Germany, was sent to the Institute of Cytopathology during May 2008 and February 2009. These materials included bronchial washings, bronchial brushings, and transbronchial fine-needle aspiration biopsies of peribronchial lesions and intrapulmonary and mediastinal lymph nodes. A routine cytological diagnosis was made in accordance with accepted diagnostic groups (see below). Only the first cytological specimen of a patient was included in the study. Patients with manifest lung cancer, whose bronchial aspirates were taken for aftercare purposes, were excluded because this condition is known to bear the risk of a false-positive diagnosis for methylation analysis.40 After application of these inclusion and exclusion criteria, 3 groups were built consisting of the first 70 patients with either a cytologically negative, equivocal, or positive diagnosis (Fig. 1). FISH, DNA-image cytometry, and QMSP were applied to each group. The authors had no additional influence on recruitment of these 210 patients. Six and 11 months after cytological diagnosis, the follow-up reference standard was determined by review of patients' charts. The latter interval was chosen to disclose potentially premalignant lesions.32, 41, 42

Figure 1.

Modified STARD diagram illustrates a patient's way through the study.

Cytological Investigation and Selection of Smears for Adjuvant Methods

Immediately after bronchoscopy, bronchial washings were fixed in Saccomanno fixative (50% ethanol, 2% polyethylene glycol 1.500, and 60 mg/L rifampicin). An aliquot of each sample was used for preparation of 4 routine smears. Residual material was stored at 4°C for subsequent investigation. Bronchial brushings or transbronchial fine-needle aspiration biopsies were smeared on 2-10 glass slides and immediately fixed with alcohol-spray (Merckofix; Merck KGaA, Darmstadt, Germany) by the bronchoscopist. For clinical routine cytology, all specimens were Papanicolaou stained and interpreted by experienced cytopathologists. All specimens were diagnosed according to the accepted diagnostic categories as follows: “negative” (no tumor cells), “doubtful” (probability of a malignant tumor approximately 30%), “suspicious” (probability of a malignant tumor approximately 70%), “positive” (tumor cells present), or “not sufficient” (no cells from deeper airways present or severe artifacts). Examples are presented in Figure 2. Afterward, the smear with the highest amount of atypical cells was selected for DNA-image cytometry. Another smear with a lesser amount of atypical cells was selected for FISH. QMSP was performed on residual, not smeared, bronchial-washing material.

Figure 2.

(A) These are normal respiratory epithelial cells with minor degenerative changes. (B) Regenerative epithelial cells are shown with nuclear enlargement, prominent nucleoli, round to oval nuclear outline, and slightly coarse chromatin. (C) Cluster of poorly differentiated squamous cell carcinoma cells shows irregular nuclear outline, coarse chromatin, and nuclear polymorphism. (D) Cancer or regeneration? Cytological diagnosis is hampered by air- drying artifacts in this cluster of regenerative epithelial cells and may lead to an equivocal result. Exclusion of aneuploidy with FISH or DNA-image cytometry helped to resolve this diagnostic problem and supported a certain negative diagnosis. Follow-up showed no sign of malignancy in this case. (Papanicolaou stain; original magnification, ×63; oil immersion objective)

Follow-Up Reference Standard

The reference standard was obtained by review of patients' charts by reviewers who were blinded to FISH, DNA-image cytometry, or promoter methylation analysis. The positive reference standard was defined as either diagnosis of a malignant tumor with histological biopsy and/or resection specimen from the same pulmonary region and in chronological context with the bronchoscopically obtained material for cytology or cytological diagnosis of a malignant tumor with a consistent clinical course (ie, imaging or adequate therapy). The negative reference standard was defined as a benign histological or cytological diagnosis consistent with the overall clinical context or no proof of a malignant lung tumor within 11 months after cytological diagnosis. All histological diagnoses were reviewed by experienced pathologists. In cases of discrepancies in the adjuvant methods and the reference standard, residual Papanicolaou smears were reviewed without changing initial cytological diagnosis.

FISH Analysis

For bronchial specimens, the Vysis LAVysion multicolor FISH probe was used. It consisted of a mixture of 4 directly labeled DNA FISH probes (chromosomal regions 5p15.2 [green signal], 6p11.1-q11 [blue signal], 7p12 [EGFR, red signal], and 8q24.12-24.13 [C-MYC, yellow signal]). FISH analysis was made on 1 of the Papanicolaou-stained slides previously used for cytological diagnosis after labeling areas with suspicious cells with a diamond pen on the backside of a slide and processing as previously described.43

Briefly, the smears were uncovered, rehydrated, and destained, then digested by using pepsin, washed in phosphate-buffered saline, and then fixed in formalin. After dehydration, denaturation at 73°C, hybridization with the FISH probe mix (7 μL LSI/WCP hybridization buffer, 2 μL purified water, and 1 μL LAVysion multicolor probe), and an additional washing step, the smears were counterstained with 4′,6-diamidino-2-phenylindole dihydrochloride (DAPI) (Vectashield DAPI mounting medium; Vector, Burlingame, California), cover-slipped, and sealed with rubber cement.

FISH cases were analyzed by 2 independent observers, each with knowledge of routine cytology diagnoses but blinded to DNA-image cytometry and QMSP. In cases of discrepancy, an additional opinion was obtained from a third observer, and a decision was made by a majority.

Hybridized areas on the slides were screened for atypical cells (nuclear enlargement, irregular shape, patchy DAPI staining) using the DAPI filter. Signals were recorded from these cells. A cell was defined as chromosomally aneuploid with a gain of 2 or more of the 4 probes.18, 21 Tetrasomy or even octasomy, defined as the presence of 4 or 8 signals of 3 or more probes, was not considered abnormal (Fig. 3). A specimen was considered positive for malignancy when 6 or more cells on a slide exhibited chromosomal aneuploidy.18, 21 When this number could not be reached after counting 25 abnormal cells, the number of cells investigated was extended to 60 in a first step, or the whole hybridized area on a slide had to be scanned. In each specimen, normal respiratory epithelial cells or lymphocytes served as internal controls, and hybridization efficiency was evaluated in these cells.

Figure 3.

Examples of FISH images with the LAVysion probe set (chromosomal regions 5p15.2 [green signal], 6p11.1-q11 [blue signal], 7p12 [EGFR, red signal], 8q24.12-24.13 [C-MYC, yellow signal]) demonstrate (A, arrow) disomy, (B) tetrasomy with a 4-4-4-4 pattern, (C) tetrasomy with a 4-4-4-3 pattern, (D) octasomy, and (E) aneusomy. (F) Aneuploid cells are shown with clusters of epidermal growth factor receptor (EGFR) gene amplification. (Original magnification, ×100; planar objective, several planes)

DNA-Image Cytometry

DNA-image cytometry was applied to 1 of the smears used for cytological diagnosis after suspicious cells were labeled on the coverslip. A photocopy of the slides was made to preserve labels after uncovering. The slides were uncovered in xylene and restained according to the method described by Feulgen.27 Measurements of nuclear DNA contents were performed as previously described.28 A computer-based image analysis system was used consisting of a Motic BA400 microscope (Motic, Xiamen, China) with a ×40 objective, a 12-bit color CCD camera with a resolution of 1360×1024 pixels (MoticamPro 285A; Motic, Xiamen, China), and the MotiCyte-DNA-image cytometry software (Motic, Xiamen, China), which provides shading and glare correction. The Conformité Européene (CE) label as a diagnostic device was available for the MotiCyte-DNA instrument. In each case, at least 30 normal respiratory epithelial cells, lymphocytes, or granulocytes were measured as internal reference cells. A minimum of 70 and optimum of 300 chosen nuclei of interest were measured in the previously labeled areas per specimen. The relevant parameter for a positive diagnosis of malignancy was the proof of DNA aneuploidy. Two algorithms for the identification of DNA aneuploidy were used: abnormal position of any DNA stemline and/or occurrence of cells >9 c (c = DNA content). DNA stemline ploidy was defined as the modal value of a DNA stemline in c U. DNA stemline aneuploidy was assumed when the modal value of a stemline was <1.80 c or >2.20 c and <3.60 c or >4.40 c. Single-cell aneuploidy was diagnosed when at least 1 cell per slide had a DNA content >9 c. Rarely, octaploid cells occur in noncancerous epithelium of inflammatory affected lungs.44 Consequently, the threshold for the detection of rare aneuploid cells had to be set at 9 c and not at 5 c. Examples of a diploid, euploid-polyploid, and aneuploid histogram are shown in Figure 4. All technical instruments, all software used, and guidelines for diagnostic interpretation and quality assurance met the standard requirements of the consensus reports of the European Society for Analytical Cellular Pathology.45-48

Figure 4.

DNA-content (x-axis) is plotted against number of cells (y-axis). (A) Diploid DNA pattern is shown with stemline at 2 c. Euploid polyploid DNA pattern has typical stemlines at 2 c, 4 c and approximately 8 c. (B) This euploid polyploid DNA pattern has typical stemlines at 2 c, 4 c and approximately 8 c. Two cells with a DNA-content greater than 9 c led to a false positive diagnosis. (C) This example of an aneuploid DNA pattern has the biggest stemline at 2.7 c and an additional fraction of proliferating cells between 4 c and 6 c.

QMSP

Analysis of gene promoter hypermethylation was conducted after bisulfite treatment of DNA blinded to the cytological diagnosis as follows: the genomic DNA of cells from bronchial washings was isolated by using the Puregene DNA Isolation kit (Gentra Systems, Minneapolis, Minnesota). Fully methylated DNA (CPGenome Universal Methylated DNA; Millipore, Billerica, Massachusetts) served as positive control. One microgram of DNA per sample was modified by sodium bisulfite treatment according to Herman et al.49 Promoter methylation analysis of adenomatous polyposis coli promoter 1A (APC), cyclin-dependent kinase inhibitor-2A (p16INK4A), and RAS association domain family protein 1 (RASSF1A) was made by using a LightCycler (Roche Diagnostics GmbH, Mannheim, Germany) as previously described (Fig. 5).40, 50 Myogenic differentiation antigen (MYOD1) was used as internal reference to control for input DNA. A sample without DNA served as a negative control. Correct size of amplified MYOD1 DNA and average samples of APC, p16INK4A, and RASSF1A were controlled with agarose gel electrophoresis. This indicates a sufficient DNA extraction, sodium bisulfite treatment, and quantitative polymerase chain reaction (PCR). Sample DNA sequencing was in accordance with published gene bank sequences of APC, p16INK4A, and RASSF1A. A specimen was assigned as positive when at least 1 tumor suppressor gene exhibited promoter hypermethylation.

Figure 5.

Quantitative polymerase chain reaction (PCR) runs are shown. Cycle number (x-axis) is plotted against fluorescence intensity (y-axis). Examples are of target genes (A) p16INK4A, (B) RASSF1A, and (C) APC. APC promoter methylation level was calculated as the ratio of the endpoint fluorescence intensity values (APC/MYOD1 [reference] x 100). The cutoff (bar) was 35%. p16INK4A or RASSF1A promoter hypermethylation is shown at crossing points >0. AC indicates adenocarcinoma; SCC, squamous cell carcinoma; SCLC, small-cell carcinoma; NSCLC, nonsmall-cell carcinoma.

Statistical Analysis

The Fisher exact test was used for contingency-table analysis of categorical data (positive or negative for malignant tumor) provided by both reference standard and tests. Sensitivity and specificity, both with 95% confidence intervals, were calculated for cytological diagnosis, FISH, DNA-image cytometry, and QMSP. Cytologically suspicious and equivocal diagnoses were set as positive for statistical evaluation. Two methods each of FISH, DNA-image cytometry, and QMSP were directly correlated by construction of contingency tables using Fleiss kappa statistics. The level of significance was set to P = .05.

RESULTS

Participants

The flow of the patients (210 in total) through the study is shown in a modified Standards for Reporting of Diagnostic Accuracy (STARD) diagram (Fig. 1).51 Follow-up reference standard was met in 68, 64, and 66 patients with positive, uncertain, or negative cytological tumor diagnosis, respectively. Twelve patients were not eligible for evaluation: 5 patients had missing charts, 2 died during the diagnostic procedure, 1 was treated for cancer after a suspicious cytological diagnosis with no histological tumor confirmation, 2 patients' chart reviews revealed known lung cancer, and 2 patients refused further diagnostic confirmation of their highly suspected lung cancers. Refer to Table 1 for detailed characteristics of the patient population. The remaining 198 patients were enrolled in this study, and FISH, DNA-image cytometry, and QMSP were performed.

Table 1. Clinical Characteristics of Patient Population
 Benign Lung Diseasea n=63Primary Lung Cancer n=124Othersb n=23
 No. (%) Median [Range]No. (%) Median [Range]No. (%) Median [Range]
  • SCC indicates squamous cell carcinoma; AC, adenocarcinoma; SCLC, small cell lung cancer; NSCLC, non-SCLC; cSCLC, combined SCLC.

  • a

    aBenign lung disease: acute or chronic bronchitis (35), pneumonia (9), interstitial lung disease (7), scar (3), pulmonary embolism (3), pleuritis (1), hemoptysis (1), atelectasis (1), gastroesophageal reflux (1), hamartoma (1), mediastinal neurinoma (1).

  • b

    bOther cases include patients with metastasizing carcinoma to the lung: poorly differentiated carcinoma of the breast (1), AC of the vulva (1), endometrioid AC of the cervix uteri (1), AC of the stomach (1), colorectal AC (3), or miscellaneous conditions (refer to g).

  • c

    cOnly primary lung carcinoma included. Percentages were calculated for SCLC and NSCLC separately.

  • d

    dThe exact tumour stage could not be determined for 2 patients. One patient died, 1 patient left the hospital prior to complete staging.

  • e

    eDiffusely metastasizing SCLC (1), cervical and mediastinal lymph node metastases of pulmonary AC with undetected primary (1).

  • f

    fTumor classification was based on cytological diagnoses alone in 9 cases.

  • g

    gMetastatic melanoma (1), atypical carcinoid tumour (1), non-Hodgkin lymphoma (2), no reference standard (12).

Age, y65 [26-88]66 [34-88]66 [26-83]
Sex   
 Female30 (48)48 (39)7 (30)
 Male33 (52)76 (61)16 (70)
Smoking status   
 Smoker34 (54)100 (81)8 (35)
 Pack-years40 [6-150]40 [10-150]40 [30-60]
 Never smoker10 (16)10 (8)5 (22)
 No data19 (30)14 (11)10 (43)
Stagec   
 0 1 (1) 
 IA 4 (4) 
 IB 14 (13.5) 
 IIB 5 (5) 
 IIIA 13 (12.5) 
 IIIB 27 (26) 
 IV 37 (36) 
 Not determinedd 2 (2) 
 Limited disease 7 (35) 
 Extensive disease 13 (65) 
Location   
 Central 84 (68) 
 Peripheral only 38 (30.5) 
 Miscellaneouse 2 (1.5) 
Histology/Cytologyf   
 SCC 31 (25)0 (0)
 AC 52 (42)5 (22)
 NSCLC 17 (14)2 (8.5)
 SCLC 20 (16)0 (0)
 cSCLC 3 (2)0 (0)
 Miscellaneousg 1 (1)16 (69.5)

FISH

A newly developed scoring algorithm for evaluation of chromosomal aneuploidy with multicolor FISH was applied to exclude false-positive results caused by euploid polyploidization, for example, in tissue repair during chronic bronchitis. Of 198 specimens, 189 (95.5%) were evaluable, and 9 specimens were excluded because of severe degenerative changes of the bronchial material or insufficient hybridization of the DNA probe. FISH yielded 98.2% (P < .001) specificity in the whole cohort of patients (Fig. 6). After a negative cytological diagnosis, FISH correctly identified malignancy in 3 patients, thus suggesting a false-negative cytology due to screening errors. Careful rescreening of Papanicolaou-stained slides revealed sparse amounts of atypical cells in 2 cases and misinterpretation of tumor cells as benign caused by severe air-drying artifacts in 1 case. All 68 patients with a certain cytological diagnosis of malignant tumor were evaluable with FISH. Malignancy was correctly identified in 67 cases. One small-cell lung carcinoma (SCLC) with well-recognized typical nuclear molding and crowding pattern on DAPI counter-stained slides showed no chromosomal aneuploidy with the 4 FISH probes used. In the group of patients with equivocal cytological diagnoses, 78.6% (33 of 42) of evaluable specimens were positive for FISH in patients with a malignant lung tumor (Table 2). One cytologically doubtful specimen showed chromosomal aneuploidy in a woman aged 72 years, but the clinical follow-up was a chronic bronchitis. Careful rescreening of residual Papanicolaou-stained smears revealed a misinterpretation of euploid-polyploidy as aneuploidy, which was possibly caused by a tight signal constellation. FISH is of special clinical value in the adjuvant application after highly suspicious cytology (Fig. 7). Under this condition, it confirmed a malignant tumor in 87.9% (29 of 33, P = .147) of patients.

Figure 6.

Specificity and sensitivity of cytology, FISH, DNA-image cytometry, and QMSP are shown in the whole cohort of patients (n = 198). Data are presented as percentages. (No./total).

Figure 7.

(Top) Papanicolaou-stained smear of a bronchial washing contains only 1 group of singly lying cells with severe abnormal changes. The sparse number of cells hampered a clear-cut positive diagnosis of a nonsmall-cell carcinoma; cytology revealed a highly suspicious result only. (Bottom) Same group is shown after hybridization with the LAVysion probe set and computer-assisted relocalization. One cell at the upper right corner was stripped off during hybridization procedure. Chromosomal aneuploidy is shown with clusters of EGFR gene amplification (7p12, red signals) in nonsmall-cell carcinoma. (original magnification, ×40; planar objective)

Table 2. Diagnostic Accuracy of FISH, DNA Image Cytometry, and QMSP in Cytologically Equivocal Cases
Specimen TypeaSensitivityaSpecificityaPκb
 No./Total (%) [95% CI]No./Total (%) [95% CI]  
  • FISH indicates fluorescent in situ hybridization; QMSP, quantitative methylation specific polymerase chain reaction (PCR).

  • a

    aAll specimens that were evaluable with either FISH, DNA image cytometry, or QMSP were included for calculation of diagnostic accuracy.

  • b

    bKappa-values were calculated only in cases when both compared methods were evaluable.

  • c

    cComparison of FISH and DNA image cytometry.

  • d

    dComparison of DNA image cytometry and QMSP.

  • e

    eComparison of FISH and QMSP.

FISH33/42 (79) [66-91]13/14 (93) [79-100]<.0010.90c
DNA image cytometry29/35 (83) [70-95]13/13 (100)<.0010.46d
QMSP19/39 (49) [33-64]18/19 (95) [85-100].0010.38e

DNA-Image Cytometry

Of the 198 specimens, 179 (90.4%) were evaluable, and 19 were excluded because of cell degeneration or subliminal amount of suspicious cells. In the whole cohort, the specificity of DNA-image cytometry was 98.2% (P < .001; Fig. 6). Subsequent to a negative cytological diagnosis, 2 cases were identified as false-negative by DNA ploidy. Both cases were also identified by FISH, and histological follow-up revealed a malignant tumor in each case. One patient with negative clinical follow-up was diagnosed with DNA aneuploidy according to the applied criteria for evaluation. A DNA histogram (Fig. 4B) showed a euploid-polyploid pattern and 2 large, but normally configured, ciliated cells with a DNA content greater than 9 c, the latter leading to the diagnosis of DNA aneuploidy. All 67 cytologically positive diagnoses evaluable with DNA-image cytometry were DNA aneuploid. Subsequent to an equivocal cytological diagnosis, DNA-image cytometry was successfully applied to 75% (48 of 64) of specimens. A malignant tumor was correctly predicted in 82.9% (29 of 35) by DNA aneuploidy (Table 2). Particularly, subsequent to a highly suspicious cytological diagnosis, malignancy was confirmed in 92.9% (26 of 28, P = .014) of patients.

QMSP

Of the 198 specimens, 187 (94.4%) were evaluable, including 18 cases in which the relevant cytological diagnoses for study inclusion were made on transbronchial fine-needle aspiration biopsies (16) or bronchial brush biopsies (2). In each case, an additional bronchial washing specimen was evaluable for QMSP: 9 of the same pulmonary region and 9 of different or unknown pulmonary regions evaluated for follow-up reference diagnosis. The latter 9 patients were excluded from evaluation of QMSP results. QMSP achieved 49.6% and 98.4% (P < .001) sensitivity and specificity, respectively, in the whole cohort (Fig. 6). Aberrant promoter methylation was correctly identified in 4 patients with false-negative cytological diagnoses. Follow-up revealed a malignant tumor in each case. One patient's malignant tumor was also detected by FISH. Rescreening of Papanicolaou-stained smears revealed sparse amounts of atypical cells in 2 cases. The cytology of the other 2 specimens remained negative. Aberrant promoter methylation was observed in 60.3% of evaluable cytologically positive specimens. These patients all had malignant lung tumors. After equivocal cytology, application of aberrant promoter methylation detected 48.7% (19 of 39) of tumors (Table 2). One patient with known rectosigmoid carcinoma and suspect lingula patches on chest x-ray showed aberrant methylation of the APC promoter. Postsurgical histological investigation of this region yielded an inflammatory pseudotumor and no evidence of pulmonary malignancy or colorectal metastasis.

Direct Comparison of Adjuvant Methods

In a subset of 147 patients, 59 with negative, 51 with positive, and 37 with equivocal cytology, all additional methods were performed and interpreted on the same specimen, and a follow-up reference standard was available. This group comprised only bronchial washing specimens, which are often false-negative with reported overall sensitivities of 43% for peripheral and 48% for central tumors in a large meta-analysis.5 More powerful sampling methods like bronchial brushings or transbronchial fine-needle aspiration biopsies were excluded because QMSP was tested only on liquid material. Accordingly, diagnostic accuracy of FISH and DNA-image cytometry was slightly lower than in the whole cohort (Fig. 6). Diagnostic accuracy of DNA-image cytometry and FISH was nearly equal (κ = 0.93), whereas sensitivity of QMSP remained lower (Table 3).

Table 3. Direct Comparison of FISH, DNA Image Cytometry, and QMSP for the Detection of Malignant Lung Tumors in Cases Where All Methods Were Evaluable on the Same Specimen (n = 147)
Specimen TypeSensitivitySpecificityPκa
 No./Total (%) [95% CI]No./Total (%) [95% CI]  
  • FISH indicates fluorescent in situ hybridization; QMSP, quantitative methylation specific polymerase chain reaction (PCR).

  • a

    Kappa-values were calculated only in cases when both compared methods were evaluable.

  • b

    Comparison of FISH and DNA image cytometry.

  • c

    Comparison of DNA image cytometry and QMSP.

  • d

    Comparison of FISH and QMSP.

FISH74/98 (76) [66-84]48/49 (98) [94-100]<.0010.93b
DNA image cytometry75/98 (77) [68-85]48/49 (98) [94-100]<.0010.52c
QMSP52/98 (53) [42-62]49/49 (100)<.0010.53d

DISCUSSION

Equivocal diagnoses in pulmonary cytology are problematic for physicians, pathologists, and patients. Furthermore, diagnostic bronchial cytology suffers from high rates of false-negative samples because the tumor cells are often absent.5 In the present study, we compared diagnostic accuracy of FISH, DNA-image cytometry, and quantitative analysis of aberrant promoter methylation. An algorithm is presented, defining the status of each adjuvant method in the diagnostic procedure.

Beyond that, we have developed a new scoring algorithm for FISH analysis in pulmonary cytology with the LAVysion set, taking the physiological phenomenon of euploid polyploidization into account. With the advent of DNA-image cytometry techniques, euploid polyploidization under benign conditions has been recognized in a large variety of human tissues.52 This includes polyploidy associated with functional adaption (ie, heart muscle cells),53 aging, or reactive conditions (ie, pneumonia). It is known from DNA-image cytometry studies that single nuclei of inflamed bronchial epithelial cells contain a DNA content up to 8 c, which corresponds with a 4-fold chromosomal set.44 Euploid polyploidization, which resembles a more than 30-year old diagnostic problem, gains new importance for the detection of aneuploidy by chromosomal FISH because it has to be differentiated from the latter.

The LAVysion set has been designed primarily to detect lung cancer cells.18 This was confirmed by our results, for 67 of 68 specimens with positive cytology were also chromosomally aneuploid with FISH. FISH was aneuploid in 79% of lung cancer patients subsequent to a cytologically equivocal diagnosis. This is in good agreement with Bubendorf and coworkers, who detected chromosomal alterations in 45 cytological equivocal cases with FISH after automated relocation and achieved 79% sensitivity and 100% specificity.11 Taking euploid polyploidization into account and excluding these cells from aneuploid values did not reduce sensitivity (78%) in the whole series of patients compared with published evidence. Bubendorf found enhanced sensitivity (73%) for bronchial brushings compared with cytology alone (49%) but no additional value in lung cancer diagnosis for bronchial washings and transbronchial needle aspirates in a group of 100 lung cancer patients and 71 controls.21 Halling et al performed cytology and FISH on bronchoscopic brushing and washing specimens and achieved 71% sensitivity on brushings and 49% sensitivity on washing specimens.22 Voss found an additional 32% of detected lung cancers by FISH compared with cytology (41% and 20% for central and peripheral tumors, respectively) in bronchoscopically collected cytology specimens in a cohort of 343 patients with indeterminate pulmonary lesions.26 The latter study defined tetrasomy as nonmalignant for study purposes but did not take euploid polyploidization beyond tetraploidy into account. We observed chromosomal octaploid cells in 8 of 56 cytologically negative or equivocal specimens with negative follow-up. In part, marked euploid polyploidization was detectable in 24 of these specimens. False-positive FISH diagnoses can easily be made when octaploid cells are interpreted as aneuploid, and difficulties in signal counting (ie, tightly positioned or elongated signals, degenerative changes) lead to false classification of in-fact tetraploid cells as chromosomally aneuploid.

Nevertheless, 1 patient with chronic bronchitis revealed a false-positive FISH result. Careful re-evaluation showed, in part, very narrow to close contacting signal positions, which hampered signal counting. With follow-up diagnosis in mind, a cytopathologist could have interpreted the suspicious cells as euploid-polyploid.

FISH was aneuploid in 3 of 65 patients after a negative cytological result; all proved to be malignant lung tumors in follow-up. Two of their specimens were aneuploid on DNA-image cytometry, and 1 showed aberrant promoter hypermethylation, too. In a group of 230 patients, Voss and coworkers detected a much higher number of 37 FISH-positive specimens with confirmed lung cancer.26 This could warrant use of FISH as a reflex test on residual material after negative cytology, but a low rate of FISH-positive specimens after negative cytology findings, as observed in our study, makes this approach appear too expensive.

Nuclear DNA content, measured with DNA-image cytometry, increases with degree of cellular atypia up from mildly dysplastic (metaplastic with mild atypia) changes to invasive lung cancer.29, 31 This has been used to facilitate the distinction between atypical squamous cells (precancer) and nonspecific cellular changes by measuring the nuclear DNA content of dysplastic bronchial epithelial cells.30, 32 The morphological criteria of dysplasia and carcinoma were applied for both cytology and DNA-image cytometry according to Saccomanno.54 We detected 82.9% aneuploid specimens with DNA-image cytometry subsequent to an equivocal cytological diagnosis, confirmed as malignant by follow-up. This is in good agreement with Auffermann et al who detected 17 of 19 cancers with DNA-image cytometry subsequent to cytological dysplastic and suspicious cell findings.32 When their specimens exhibit DNA aneuploidy with the criteria applied in this study, our patients routinely receive the diagnosis of a malignant tumor.

DNA-image cytometry revealed DNA aneuploidy in 3 patients subsequent to a negative cytological diagnosis. Follow-up histological diagnosis of a malignant tumor was obtained for 2 patients. One patient was diagnosed with DNA aneuploidy because of 2 cells with a DNA content >9 c.28 Refer to Figure 4B for the DNA histogram, which displays a typical euploid polyploid pattern with the exception of 2 large, but cytologically normally configured, ciliated cells with a DNA content greater than 9 c. These had not reached the 16 c state yet and were obviously octaploid cells in the S-phase of the cell cycle. Because follow-up did not show any sign of malignancy, we prefer to handle single DNA values up to 16 c as not aneuploid in the case of euploid polyploidization as suggested for bronchial epithelial cells of the human lung by measurement of nuclear area and volume and as reported for other organs with inflammation and reactive changes.55-57 In addition, we strongly recommend strict rules for ploidy interpretation as have been summarized by Böcking, to avoid potential pitfalls (ie, viral cytopathic effect).28 As proof that DNA aneuploidy is a reliable marker of malignant lung tumors, all cytologically positive specimens evaluable with DNA-image cytometry revealed stemline aneuploidy or single-cell aneuploidy.

We have recently suggested quantitative detection of aberrant promoter hypermethylation of APC, p16INK4A, and RASSF1A genes as a reflex test on bronchial cytologic specimens in patients who are clinically suspected of having lung cancer but do not display a final cytological or histological diagnosis of malignancy.40 QMSP had 53% overall sensitivity in this study, which is confirmed by 49.6% sensitivity here. The 98.4% specificity is in good agreement with >99% detected in our previous study. Aberrant promoter methylation was detected in 4 out of 22 patients with proven lung carcinoma subsequent to a negative cytology finding and, therefore, had the best performance of the methods compared in this study. One cytologically negative specimen of a peripheral adenocarcinoma was positive on FISH and on DNA-image cytometry, too. Bronchial washings of 1 central small-cell carcinoma, 1 central adenocarcinoma, and 1 peripheral, multifocal adenocarcinoma were exclusively positive with QMSP. Rescreening of residual Papanicolaou-stained slides displayed suspicious cells in 2 cases; the others remained negative. In this study, 1 nontumor patient displayed a false-positive QMSP assay with aberrant methylation of the APC promoter similar to that described by Schmiemann et al.40 The 68-year-old patient observed in our study with a 90% APC promoter methylation compared with MYOD1 had a 3-month history of rectosigmoidal carcinoma with suspected lung metastasis in the lingula. Surgical resection revealed an inflammatory pseudotumor. It has been suggested that aberrant methylation of the APC promoter is associated with aging.58, 59 This was 1 reason why Grote et al introduced a 35% cutoff—compared with the MYOD1 reference gene—into investigation of aberrant APC promoter methylation with QMSP.50 One could speculate whether this cutoff is not high enough in some rare cases.

This is the first study, to our knowledge, that directly compared the diagnostic power of 3 different approaches in addition to cytological investigation of lung cancer. FISH and DNA-image cytometry achieved very similar rates of aneuploidy detection, whereas sensitivity of QMSP was significantly lower. The prediction of lung cancer incidence with aberrant promoter methylation and FISH was reported in a screening approach by Belinsky et al and Varella-Garcia et al in 2 studies on the same subgroup of patients, all enrolled in the Colorado High-Risk Cohort Study from 1993 to 2003. Belinsky reported increasing prevalence of promoter hypermethylation of multiple genes, diagnosed with nested methylation-specific PCR, in sputum samples with decreasing time to lung cancer diagnosis.42 Methylation of 3 or more genes in sputum, collected within 18 months before manifest cancer, predicted lung cancer with 64% sensitivity and specificity. Varella-Garcia reported 76% sensitivity and 88% specificity for a positive FISH assay with the LAVysion probes within 18 months before cancer diagnosis.20

Subsequent to an equivocal cytological result in the current report, the diagnostic accuracy of FISH and DNA-image cytometry were nearly equal (κ = 0.9), but more smears were evaluable with FISH (56 of 64) than with DNA-image cytometry (48 of 64), in most cases, because of a subliminal amount of atypical cells. We suggest performing DNA-image cytometry when there are enough cells (more than approximately 70 cells) in the specimen, because this method is cheaper than FISH ($120 US dollars for DNA cytometry, $595 US dollars for multicolor FISH according to a German medical-fee schedule) and not as time consuming.60 When there is a sparse amount of cells, which often occurs, hampering a clear-cut cytological cancer diagnosis, we recommend chromosomal FISH because only 6 aneuploid cells are needed for a positive diagnosis.

Conclusion

To a great extent, equivocal cytology in lung cancer diagnosis can be overcome by the use of additional methods on the same specimen (ie, slide). Our diagnostic algorithm recommends DNA-image cytometry on 1 of the cytological smears when there are enough atypical cells, otherwise FISH should be performed. In most cases, unequivocal cancer diagnoses are possible. Subsequent to a negative cytology finding, QMSP can be performed as a reflex test on the residual liquid specimen in the case of persisting lung cancer suspicion.

CONFLICT OF INTEREST DISCLOSURES

Professor A. Böcking is receiving grant support from Motic Company, Xiamen, China, to develop instruments for DNA-image cytometry and multimodal cell analysis.

Ancillary