Cytologic Features Useful for Distinguishing Small Cell From Non Small Cell Carcinoma in Bronchial Brush and Wash Specimens

Transcription

1 Anatomic Pathology / SMALL CELL VS NON SMALL CELL CARCINOMA Cytologic Features Useful for Distinguishing Small Cell From Non Small Cell Carcinoma in Bronchial Brush and Wash Specimens Charles D. Sturgis, MD, 1 Diana L. Nassar, SCT(ASCP), 1 Joyce A. D Antonio, PhD, 2 and Stephen S. Raab, MD 1 Key Words: Bronchial brush; Bronchial wash; Small cell carcinoma; Non small cell carcinoma; Cytology/cytopathology; Regression analysis Abstract One blinded observer (C.D.S.) retrospectively reviewed 76 previously diagnosed and biopsyconfirmed malignant bronchial brush and wash specimens, 46 non small cell and 30 small cell carcinomas, obtained from 55 patients. Each case was scored for the presence or absence of 36 standard criteria (architectural, cytoplasmic, and nuclear). Logistic regression analysis was used to determine which criteria were most useful for separating small cell from non small cell lesions. Although no single criterion displayed 100% sensitivity and specificity for small cell cancer, univariate statistical analysis indicated that 3 individual criteria (nuclear molding, finely granular or salt and pepper chromatin, and scant delicate cytoplasm) were more than 90% sensitive and specific in cases of small cell carcinoma. The presence of nuclear molding alone provided the best fit for the logistic regression model. When nuclear molding was present, the odds of a small cell diagnosis increased more than 300-fold. Nuclear molding, finely granular or salt and pepper chromatin, and scant, delicate cytoplasm are the 3 most sensitive and specific cytomorphologic features traditionally used to separate small cell from non small cell carcinoma. Nuclear molding alone represents the most significant cytomorphologic feature for distinguishing between these malignant lesions. Lung cancer is the leading cause of cancer death for men and women in the United States, accounting for more cancerrelated mortality since 1930 than all other cancer deaths combined. 1 At the present time, most major medical centers throughout the world use cytologic diagnostic techniques in the evaluation of suspected primary pulmonary malignant neoplasms. Because of simplicity of acquisition, sputum specimens continue to be the most frequently examined specimens worldwide. 2 When the results of sputum studies are negative and clinical suspicion persists, or when lesions are located in the pulmonary periphery, fiberoptic bronchoscopy may be used to obtain diagnostic cellular material. The literature indicates that brushing cytology is accurate for the diagnosis of lung cancer and that cytologic specimens obtained by bronchoscopy generally demonstrate better preserved cellular samples than does analysis of sputum. 3-6 Bronchogenic neoplasms historically have been divided into several different categories based on morphologic features, cell type, and clinical outcome patterns. The first step in the cytologic categorization of lung carcinomas is identification of the features used to distinguish small cell cancer (SCCa) from non small cell cancer (NSCCa). Several large studies have queried the relative distribution of lung carcinoma types in the population. These reviews indicate that approximately one fifth of all primary lung carcinomas are of the small cell undifferentiated type (range, 16%-26%) This separation of malignant epithelial lesions into 2 basic subsets is essential because the clinical management of lung cancer is based largely on stage at diagnosis and on the tumor type as established by the pathologist, with cases of NSCCa considered as 1 treatment group and cases of SCCa representing another. 11 The disseminated nature of most cases of SCCa, displaying a high frequency of metastases at the time American Society of Clinical Pathologists Am J Clin Pathol 2000;114:

2 Sturgis et al / SMALL CELL VS NON SMALL CELL CARCINOMA of diagnosis combined with a high sensitivity to cytostatic agents, has led to the use of chemotherapy as the treatment of choice in all stages. 11 (In some centers, stage I SCCa may be treated with surgery alone.) On the other hand, NSCCa is among the most chemoresistant of all solid tumors. Because of this dichotomy in management, accurate separation of these 2 basic tumor groups is essential. The separation of cases of SCCa from those of NSCCa may be difficult in the clinical setting, with previous reviews indicating cytologic-histologic concordance rates for small cell lesions as low as 71%. 12 Although many features that are useful for the identification of and separation of these malignant pulmonary neoplasms are reported in the literature, rigorous statistical methods have not been used to establish the individual and collective import of these criteria. In the present study, 76 biopsy-proven malignant bronchial brush and wash specimens were reviewed and evaluated for the presence or absence of 36 standardly used cytomorphologic criteria to assess which criteria are most important for separating small cell from non small cell carcinomas of the lung. Materials and Methods By retrospective review, 105 bronchoscopically acquired pulmonary cytology specimens were retrieved from the files of the cytology laboratory at Allegheny General Hospital, Pittsburgh, PA. Each of the cases had been diagnosed during the interval from January 1994 through December 1998, and each was coded as positive for carcinoma. All of the slides in each of these cases were reviewed for the presence or absence of 36 cytomorphologic criteria Table 1, all of which were deemed useful for the identification of malignancy and for the separation of SCCa from the various types of NSCCa by standard cytology texts. 2,13,14 One blinded observer (C.D.S.) reviewed each case and recorded an overall diagnosis in addition to identifying the presence or absence of the 36 criteria. (The number of slides reviewed per case ranged from 2 to 16 [mean, 4 slides per case]. All specimens were alcohol fixed and Papanicolaou stained.) After the data collection phase was concluded, the reviewer s overall diagnosis was compared with the original diagnosis, and the medical record was reviewed to ascertain whether biopsy or other tissue confirmation existed for each case. Cases in which the reviewer s diagnosis differed from the original diagnosis, tissue confirmation had not been obtained, or the original diagnosis had been made on very scant material were excluded from the study set to ensure accuracy in the assignment of criteria to SCCa and NSCCa categories. The final set of biopsy-proven and diagnostically concurrent bronchial brush and wash specimens included 76 cases, 46 cases of NSCCa and 30 cases of SCCa, obtained from 55 adult patients, 29 men and 26 women, ranging in age from 43 to 83 years (mean age, 66 years). Multiple cytologic specimens were studied from the tumors of a subset of the patients. This subset represented the persons from whom more than 1 bronchoscopic specimen was used, ie, a bronchial brush specimen and a bronchial wash specimen from the same patient. (Corresponding histologic samples from these 76 cases were studied by one of us [C.D.S.] to ensure accuracy and 100% correlation between the historically reported histologic diagnoses and the cytologic interpretations.) Although the 36 cytomorphologic criteria listed in Table 1 have been described previously and are used in standard texts, 2,13,14 certain of them need further description. Cases were marked as positive for high nuclear/cytoplasmic ratio when the tumor cells displayed an increased nuclear size relative to the amount of cellular cytoplasm as seen in normal internal control bronchial epithelial cells. Scant cytoplasm was defined as a thin, lymphocyte-like, rim of cytosol. The attribute of nuclear molding was assigned to cases in Table 1 Cytomorphologic Criteria Small Cell Carcinoma Non Small Cell Carcinoma (Squamous) Non Small Cell Carcinoma (Adenocarcinoma) High nuclear/cytoplasmic ratios Orangeophilic cytoplasm* Indistinct cell borders Scant delicate cytoplasm* Keratinized cytoplasm Multiple nucleoli Basophilic cytoplasm* Intercellular bridges Macronucleoli* Nuclear molding* Pearl formation Intracytoplasmic mucin Crush artifact* Cell-in-cell arrangement Foamy cytoplasm* Nuclear breakdown or apoptotic bodies Hyperkeratosis Powdery to open chromatin Diathesis or necrosis Spindle cells Bubbly or vesicular chromatin Granular salt and pepper chromatin* Hyperchromatic smooth nuclei* Coarsely granular chromatin Inconspicuous nucleoli* Ink dot or pyknotic chromatin 3-dimensional groups Single-file arrangements Malignant single cells Irregular nuclear membranes Cells 1.5 times the size of lymphocytes* Distinct cell borders Nuclear overlap in groups Gland formation Papillary structures Tumor giant cells * Criteria used in regression analysis. 198 Am J Clin Pathol 2000;114: American Society of Clinical Pathologists

3 Anatomic Pathology / ORIGINAL ARTICLE which tumor cells closely abutted adjacent cells with nuclear membranes altering to conform to and contour against one another as seen in Image 1 and Image 2. Crush artifact and apoptotic bodies were defined as the smearing of fragile nuclear contents and subcellular fragments of granular, degenerated, nuclear debris, respectively. Salt and pepper chromatin pattern was attributed to a case when the cells displayed a uniformly dispersed and finely granular nuclear substance as seen in Image 1. Single-file arrangement was identified when 4 or more cells were present in cohesive uniform strings presenting in single-file display, often trapped in mucous strands. Keratinized cytoplasm was defined as firm, eosinophilic to glassy cytoplasm with perinuclear intracytoplasmic refractile ringing. Pearl formation, seen classically in squamoid lesions, was attributed to the cases with roughly spherical tumor fragments in which peripheral, somewhat spindled cells encircled larger, more polygonal cells resembling a ball or oyster pearl. Hyperkeratosis was defined as the presence of anucleate squamous cells. Cases marked as positive for spindle cells contained elongated slender cells with tapering cytoplasmic tips in which the cytoplasmic cell length was 4 or more times the length of the nucleus. The categories of distinct vs indistinct cell borders pertained to cohesive fragments of malignant cells and whether or not cells within those fragments were readily separated into individual units or the cytoplasm of the cells formed more of a syncytium. Intracytoplasmic mucin was attributed only to the cases with distinct targetoid intracytoplasmic inclusions of homogenous material. (Mucin stains were not used.) The cases marked as having powdery to open chromatin displayed peripheral nuclear chromatin condensation with only scant nuclear material filling the majority of the nuclear space, giving a washed out appearance. Gland formation was identified only when truly intact acinar architecture with central lumen formation could be appreciated in tissue fragments. Intact tips of papillae and 3- dimensional fragments with fibrovascular cores qualified a case as having papillary structures, and tumor giant cells were defined as large cells (4 or more times the size of the majority of tumor cells in any given case) with nuclear multilobation or multinucleation as seen in Image 3. Univariate analysis was performed for each of the 36 criteria with calculation of sensitivity, specificity, and the Fisher exact 2-sided test P value. Logistic regression analysis was used to model the relationship between the most sensitive and specific individual criteria and the separation of cases of SCCa from NSCCa. The regression was first conducted by examining a combination of nuclear, cytoplasmic, and architectural variables, and then it was repeated looking separately at the criteria divided into architectural, cytoplasmic, and nuclear subsets. The technique of logistic regression analysis as it applies to cytopathologic reviews has been described previously. 15,16 To briefly summarize, variables to be entered into the logistic regression are identified by using univariate analyses of association. These analyses are performed between the dependent variable (diagnosis of SCCa) and the presence or absence of each independent variable (each cytomorphologic criterion). The Image 1 A 3-dimensional fragment of small cell carcinoma acquired by fiberoptic bronchoscopy demonstrating the classic features of high nuclear/cytoplasmic ratios, scant delicate cytoplasm, nuclear molding, finely granular salt and pepper chromatin, and cells 1.5 times the size of lymphocytes (Papanicolaou, 400). American Society of Clinical Pathologists Image 2 A bronchial brush smear with cells of small cell carcinoma demonstrating high nuclear/cytoplasmic ratios, scant delicate cytoplasm, nuclear molding, crush artifact, scattered apoptotic bodies, granular salt and pepper chromatin, and inconspicuous nucleoli (Papanicolaou, 400). Am J Clin Pathol 2000;114:

4 Sturgis et al / SMALL CELL VS NON SMALL CELL CARCINOMA Image 3 A 3-dimensional fragment of non small cell carcinoma (adenocarcinoma) from a bronchial brush specimen demonstrating multiple nucleoli, macronucleoli, relatively abundant foamy cytoplasm, powdery to open chromatin, irregular nuclear membranes, nuclear overlap, and tumor giant cells (Papanicolaou, 400). independent variables marked with asterisks in Table 1 were included in the actual analytic model. Some criteria were excluded from the regression analysis because the logistic model could not handle as many separate features as were assessed given the sample size. In logistic regression, it is assumed that the relationship between the dependent and independent variables can be explained by a log distribution on an S-shaped curve that ranges from 0 to 1. All statistical analyses were conducted using the Statistical Package for the Social Sciences (SPSS, Chicago, IL). classically associated with adenocarcinoma demonstrated 100% specificity for the diagnosis of NSCCa, including multiple nucleoli, macronucleoli, and foamy cytoplasm (significant P values), and vesicular chromatin, intracytoplasmic mucin, gland formation, and papillary structures (nonsignificant P values). Certain features that classically have been associated with given types of malignant neoplasm did not correlate well with information provided in the standard texts. 2,13,14 For example, the presence of 3-dimensional groups has long been associated with the adenocarcinoma subset of NSCCa. Our data demonstrated only a 60% specificity for NSCCa with this criterion, and more than one fourth (27%) of the SCCa cases were marked positive for this criterion. The feature of ink dot or pyknotic chromatin that has been attributed to cells of the squamous subset of NSCCa was found in 50% of the cases of SCCa. Logistic regression analysis demonstrated that the criterion of nuclear molding was the most effective criterion for separating SCCa from NSCCa. The addition of other criteria resulted in diminished significance. When nuclear molding was present in a malignant brush or wash specimen, the odds of a diagnosis of SCCa increased more 300-fold (P =.000 [model chi-square, 70.8]). Multiple regression analyses were conducted with specific subsets of features used as independent variables. When architectural features alone and cytoplasmic features alone were considered, no individual feature or combination of features could be identified as important for separating SCCa from NSCCa. When nuclear features alone were used in the logistic regression, the combined criteria of salt and pepper chromatin and nuclear molding proved important for separating the 2 types of carcinoma (model chi-square, 86.4). Results Of the 36 cytomorphologic criteria evaluated, 21 were found to be individually significant for separating cases diagnosed as SCCa from those diagnosed as NSCCa (P <.05) with 9 classic SCCa features and 12 classic NSCCa features proving significant by univariate analysis as listed in Table 2. Three SCCa criteria had sensitivities and specificities greater than 90%, namely, scant delicate cytoplasm, nuclear molding, and salt and pepper chromatin. Five of the criteria traditionally associated with NSCCa were 100% specific for squamous carcinoma. Features with significant P values were keratinized cytoplasm, spindled tumor cells, and hyperkeratosis. Two of the criteria that were not significant, intercellular bridges and pearl formation, also were 100% specific for squamous carcinoma. Seven of the 14 criteria Discussion Although limited-stage SCCa of the lung may be managed with surgical intervention, the early disseminated nature of this disease process combined with a high sensitivity to cytostatic agents has led to the use of chemotherapy as the first-line treatment of choice, regardless of surgical intervention, for almost all patients diagnosed with this type of bronchogenic malignant neoplasm. The most commonly used chemotherapeutic regimens include combinations of platinum analogues and podophyllotoxins. 11 Before patients are committed to these aggressive therapies, representative malignant cellular tissue must be obtained for accurate diagnoses to be given. When sputum studies fail, fiberoptic bronchoscopy is often the next attempted intervention. Bronchial brushings and washings that contain diagnostic cellular material are optimal specimens because they lack much of 200 Am J Clin Pathol 2000;114: American Society of Clinical Pathologists

5 Anatomic Pathology / ORIGINAL ARTICLE Table 2 Sensitivity, Specificity, and Statistical Significance of Cytologic Criteria for Predicting Diagnoses of Small Cell Carcinoma Cytomorphologic Criteria Sensitivity (%) Specificity (%) P Value Small cell carcinoma Scant delicate cytoplasm <.000 Nuclear molding <.000 Salt and pepper chromatin <.000 Cells 1.5 times the size of lymphocytes <.000 Basophilic cytoplasm <.000 Crush artifact <.000 Nuclear breakdown or apoptotic bodies <.000 Single file arrangements Inconspicuous nucleoli High nuclear/cytoplasmic ratios Diathesis or necrosis Squamous carcinoma Spindled tumor cells <.000 Keratinized cytoplasm <.000 Hyperchromatic smooth nuclei Orangeophilic cytoplasm Hyperkeratosis Distinct cell borders Cell-in-cell arrangement Ink dot or pyknotic chromatin Pearl formation Malignant single cells Intercellular bridges Adenocarcinoma Macronucleoli <.000 Powdery to open chromatin <.000 Foamy cytoplasm <.000 Multiple nucleoli <.000 Irregular nuclear membranes Tumor giant cells Coarsely granular chromatin Bubbly or vesicular chromatin Gland formation Nuclear overlap in groups Papillary structures Indistinct cell borders dimensional groups Intracytoplasmic mucin the cellular distortion and degeneration seen in sputum preparations. 14 Because decisions to treat with chemotherapy vs other definitive treatment strategies may be based on cytologic diagnoses alone, it is important for practicing pathologists to have well-established and proven diagnostic criteria for separating cases of SCCa from cases of NSCCa. In the present study, the presence or absence of 36 standard cytomorphologic criteria was assessed in 76 cases of previously diagnosed and biopsy-proven malignant bronchial cytology specimens to establish the relative import of each criterion by rigorous statistical methods. The data (Table 2) indicate that many of the criteria set forth in standard cytopathology texts 2,13,14 are significant for separating the small cell subset from the group of bronchogenic malignant neoplasms as a whole. The features of scant delicate cytoplasm, nuclear molding, salt and pepper chromatin, cells 1.5 times the size of lymphocytes, basophilic cytoplasm, crush artifact, nuclear breakdown or apoptotic bodies, single-file arrangements, and inconspicuous nucleoli all were significant (P <.05). Logistic regression analysis using a combination of architectural, cytoplasmic, and nuclear features as listed in Table 1 indicates that the presence of nuclear molding (P <.000; residual chi-square, 70.8) is the best criterion for separating cases of SCCa from cases of NSCCa. The addition of other criteria to the logistic model weakens the significance and decreases the fit to the logistic curve. When nuclear molding is present in a bronchial cytology specimen that contains diagnostic cellular material from a bronchogenic carcinoma, the odds of diagnosing the lesions correctly as a SCCa increase more than 300-fold. In the present study, the logistic model was modified in many ways, sometimes excluding certain criteria and other times including them to assure the best fit. When the regression was conducted for architectural and cytoplasmic features alone, no increased significance of any criterion was American Society of Clinical Pathologists Am J Clin Pathol 2000;114:

6 Sturgis et al / SMALL CELL VS NON SMALL CELL CARCINOMA demonstrated. When nuclear features alone were studied, the criteria of nuclear molding and salt and pepper chromatin combined were identified as important for separating SCCa from NSCCa (model chi-square, 86.4). Regardless of the criteria selected and the independent variables used, the best fit to the regression model always included the criterion of nuclear molding. It is interesting that the criterion of nuclear molding alone provides the best chi-square value and fit to the logistic curve by regression analysis. By inspection of the statistical data, one might expect that addition of other variables that were highly sensitive and specific by univariate analysis (such as salt and pepper chromatin) might increase the accuracy of separating cases of SCCa from cases of NSCCa. One explanation for this not being the case is that these criteria are so closely related in the SCCa group that their addition does not allow increased separation. While some criteria may be useful independently, their combined benefit is no better than the presence of nuclear molding alone. One limitation of the present study is the limited sample size. A sacrifice in study design was made when 29 of the original 105 cases were excluded to ensure accuracy in the assignment of criteria, leaving only 76 cases for the study. (The larger original study set would have allowed for the inclusion of more independent variables in the regression models.) Another important limitation is that the present study does not take other lesions into account that may be important differential diagnoses for SCCa, such as atypical carcinoid tumors, lymphomas, and large cell neuroendocrine carcinomas. Including criteria traditionally assigned to these diagnostic categories would round out the study nicely, and accumulation of such data is a nice idea for future investigation. We made a rigorous assessment of the time-honored criteria used in bronchial cytology diagnosis. Certain features that classically have been associated with given types of malignant neoplasm did not correlate well with historically reported associations to given tumor types. For instance, 3-dimensional groups with nuclear overlapping are classically reported to occur in the adenocarcinoma subset of the NSCCa group; however, 3-dimensional tumor fragments were noted in the majority of SCCa cases (73%), and nuclear overlap also was appreciated in more than half of the SCCa cases (53%). Historically contrary criteria such as these still may be diagnostically useful as adjunctive information, but methodical reviews such as ours with statistical backing will help to direct diagnosticians to the most accurate diagnoses. References 1. Friedberg JS, Kaiser LR. Epidemiology of lung cancer. Semin Thorac Cardiovasc Surg. 1997;9: Johnston WW, Elson CE. Respiratory tract. In: Bibbo M, ed. Comprehensive Cytopathology. 2nd ed. Philadelphia, PA: Saunders; 1997: Bedrossian CW, Rybka DL. Bronchial brushing during fiberoptic bronchoscopy for the cytodiagnosis of lung cancer: comparison with sputum and bronchial washings. Acta Cytol. 1976;20: Pedersen U, Balle VH, Greisen O. Diagnostic value of brush biopsy in suspected bronchial carcinoma with the use of the flexible fibre bronchoscope. Clin Otolaryngol. 1981;6: Jay SJ, Wehr K, Nicholson DP, et al. Diagnostic sensitivity and specificity of pulmonary cytology: comparison of techniques used in conjunction with flexible fiber optic bronchoscopy. Acta Cytol. 1980;24: Chopra SK, Genovesi MG, Simmons DH, et al. Fiberoptic bronchoscopy in the diagnosis of lung cancer comparison of pre- and post-bronchoscopy sputa, washings, brushings and biopsies. Acta Cytol. 1977;21: Melamed MR, Flehinger BJ, Zaman MB, et al. Screening for early lung cancer: results of the Memorial Sloan-Kettering study in New York. Chest. 1984;86: Flehinger BJ, Melamed MR, Zaman MB, et al. Early lung cancer detection: results of the initial (prevalence) radiologic and cytologic screening in the Memorial Sloan- Kettering study. Am Rev Respir Dis. 1984;130: Woolner LB, Fontana RS, Sanderson DR, et al. Mayo Lung Project: evaluation of lung cancer screening through December Mayo Clin Proc. 1981;56: Flehinger BJ, Kimmel M, Polyak T, et al. Screening for lung cancer: the Mayo Lung Project revisited. Cancer. 1993;72: Hansen HH, Pappot H. Primary malignant tumours of the lung and pleura. In: Cavalli F, Hansen HH, Kaye SB, eds. Textbook of Medical Oncology. London, England: Martin Dunitz; 1997: Raab SS, Silverman J. Clinical utility of cytologic typing of lung tumors. Diagn Cytopathol. 1994;10: DeMay RM. Respiratory cytology. In: The Art & Science of Cytopathology. Chicago, IL: ASCP Press; 1996: Koss LG. Tumors of the lung: conventional cytology and aspiration biopsy. In: Diagnostic Cytology and Its Histopathologic Bases. 4th ed. Philadelphia, PA: Lippincott; 1992: Raab SS, Isacson C, Layfield LJ, et al. Atypical glandular cells of undetermined significance: cytologic criteria to separate clinically significant from benign lesions. Am J Clin Pathol. 1995;104: Raab SS, Lenel JC, Cohen MB. Low grade transitional cell carcinoma of the bladder: cytologic diagnosis by key features as identified by logistic regression analysis. Cancer. 1994;74: From the 1 Department of Pathology, Allegheny General Hospital, Pittsburgh, PA, and the 2 Rangos School of Health Sciences, Duquesne University, Pittsburgh. Address reprint requests to Dr Sturgis: Allegheny General Hospital Dept of Pathology, 320 E North Ave, Pittsburgh, PA Am J Clin Pathol 2000;114: American Society of Clinical Pathologists