Survival Comparison of Adenosquamous, Squamous Cell, and Adenocarcinoma of the Lung After Lobectomy

Survival Comparison of Adenosquamous, Squamous Cell, and Adenocarcinoma of the Lung After Lobectomy David T. Cooke, MD, Danh V. Nguyen, PhD, Ying Yang, MS, Steven L. Chen, MD, MBA, Cindy Yu, MD, and Royce F. Calhoun, MD Division of Cardiothoracic Surgery, Division of Surgical Oncology, and Department of Pathology, University of California, Davis Medical Center, Sacramento; and Division of Biostatistics, Department of Public Health Sciences, University of California, Davis, Davis, California Background. Primary adenosquamous carcinoma (ASC) of the lung is a rare tumor that may carry a poor prognosis. We examined a national database to see if ASC exhibited distinct clinical behavior from squamous cell (SC) and adenocarcinoma (AC) of the lung. Methods. This is a retrospective study querying the Surveillance, Epidemiology, and End Results database to identify 872 surgical patients diagnosed with ASC, 7888 with SC, and 12,601 with AC of the lung from 1998 to 2002. Analysis characterized clinical variables to determine patterns of presentation and compared survival among the above three histologic groups after lobectomy for stage I and II disease. Results. ASC represented 4.1% of the 21,361 patients examined. ASC tended toward right side (56.9%) laterality and upper lobe (60.0%) location. Compared with AC, patients with ASC and SC were more likely to be male (p < 0.0001), and ASC patients had worse histologic grade (p< 0.0001). Survival after lobectomy for stage I and II disease was significantly reduced in ASC and SC compared with AC (p < 0.0001). ASC had a significantly increased hazard ratio of 1.35 and 1.27 relative to AC and SC, respectively. Other significant negative predictors of survival included tumor grade of III and IV, stage II, age, and black ethnicity. Conclusions. This large review demonstrates that ASC is an uncommon tumor with distinct clinical behavior and worse prognosis than AC and SC. Further insight into the molecular profile of ASC is needed to determine the cause of its biologic aggressiveness. (Ann Thorac Surg 2010;90:943 8) 2010 by The Society of Thoracic Surgeons Primary adenosquamous carcinoma of the lung (ASC) is a mixed histologic tumor, as defined by the World Heath Organization, as showing components of both squamous cell carcinoma (SC) and adenocarcinoma (AC), with each comprising at least 10% of the tumor (Fig 1A) [1]. ASC is a relatively rare tumor, comprising anywhere from 0.3% to 5% of non-small cell lung cancers [2, 3]. Anecdotally, ASC of the lung has worse outcomes compared with single histology AC (Fig 1B) and SC (Fig 1C). These observations have some support based on small, single-institutional studies that suggest that ASC of the lung has a more advanced clinical and pathologic presentation pattern than AC and SC lung tumors [2, 4 5]. In addition, patients with ASC of the lung may exhibit a poorer survival after resection than patients with SC and AC. We sought to test the hypothesis that ASC exhibits distinct clinical behavior from SC and AC of the lung. We Accepted for publication May 10, 2010. Presented at the Poster Session of the Forty-sixth Annual Meeting of The Society of Thoracic Surgeons, Fort Lauderdale, Florida, Jan 25 27, 2010. Address correspondence to Dr Cooke, Division of Cardiothoracic Surgery, University of California, Davis Medical Center, 2221 Stockton Blvd, Ste 2117, Sacramento, CA 95817; e-mail: david.cooke@ucdavis. ucdmc.edu. evaluated this hypothesis by querying patient data from the United States multiinstitution Surveillance, Epidemiology, and End Results (SEER) cancer database. Material and Methods The University of California, Davis Medical Center Institutional Review Board determined that this project qualified for an exemption. Database This is a retrospective cohort analysis using the SEER database. The SEER database is managed by the National Cancer Institute and is a compilation of deidentified data on cancer incidence and survival from cancer registries representing 26% of the United States population [6]. Search Criteria The SEER database was queried for lung cancer patients diagnosed in the years 1998 to 2002. The site and morphology code was Lung and Bronchus. Patients who were evaluated carried the International Classification of Diseases for Oncology-3 SEER site/histology codes 8070/3 Squamous cell carcinoma, not otherwise specified; 2010 by The Society of Thoracic Surgeons 0003-4975/$36.00 Published by Elsevier Inc doi:10.1016/j.athoracsur.2010.05.025

944 COOKE ET AL Ann Thorac Surg SURVIVAL POSTLOBECTOMY FOR ASC, SC, AND AC 2010;90:943 8 Abbreviations and Acronyms AC Adenocarcinoma ASC Adenosquamous cell CI confidence interval EGFR Epidermal growth factor receptor HR hazard ratio SC squamous cell SD standard deviation SEER Surveillance, Epidemiology, and End Results 8140/3 Adenocarcinoma, not otherwise specified; or 8560/3 Adenosquamous carcinoma, and underwent directed surgical intervention. For analysis of early-stage outcomes, study cohorts were limited to patients undergoing lobectomy for earlystage, node-negative disease and who did not receive radiotherapy. Information on neoadjuvant or adjuvant chemotherapy is not available through the SEER database. For the purposes of this study, we defined early stage using the 7th edition of the T N M lung cancer classification as pathologic (p) T1-3 N0 M0 [7]. In our analysis, the lesions in patients with stage I were pt1a- T2b N0 M0, and stage II lesions were pt2b-t3 N0 M0. According to the 7th edition of the T N M lung cancer classification, tumors by size are T1a, 20 mm or less; T1b, 20 to 30 mm; T2a, 30 to 50 mm; T2b, 50 to 70 mm, and T3, exceeding 70 mm. We did not analyze T3 lesions that are characterized by chest wall invasion or multiple nodules in the same lobe, or stage II tumors that were T1a-T2b N1 M0. Survival was based on cause of death Lung and Bronchus or Alive. The last date of follow-up was December 2006, and the median follow-up time was 5 years for AC and SC and 4.9 years for ASC. Statistical Analysis Our analyses consisted of two parts. Patient and tumor descriptive characteristics were collected, including demographic variables (age, ethnicity, and gender) and clinical variables, such as anatomic tumor characteristics (laterality and location), histologic characteristics (size, grade, number of positive nodes), and number of primaries. Patients with 1 or more positive nodes were considered as having positive lymph nodes. Comparison among carcinomas groups were based on analysis of variance for continuous variables and 2 testing for categoric variables. The second part of our analysis focused on survival among early stage AC, SC and ASC. Early-stage patient characteristics and clinical variables for each group were analyzed as described. Survival curves were estimated using the Kaplan-Meier method for each carcinoma type, and the log-rank statistic was used for the comparison of overall differences in survival distributions. Cox regression analysis was used to obtain hazard ratios of death for AC, SC, and ASC, adjusted for age, sex, ethnicity, laterality, tumor location, size, and grade. We checked departure from the proportional hazard assumption using graphic methods based on Schoenfeld residuals vs time as well as testing procedure based inclusion of timedependent covariates. Unadjusted survival analysis (Kaplan-Meier curves) was based on all data (N 6186), and results are presented with 95% confidence intervals (CI). For the adjusted Cox regression model analysis, 237 patients (3.8%) were excluded because of missing covariate data. To examine the sensitivity of our conclusions to this missing data, we used multiple imputation methods for missing data and the results were no difference. Therefore, we presented results without imputation of missing data (N 5949). Analyses were performed using SAS 9.1 software (SAS, Cary, NC). Results were considered statistically significant at p 0.05. Results Clinical and Pathologic Variables of Patients who Underwent Pulmonary Resection for AC, SC and ASC Carcinoma We identified 7888 patients (Tables 1 and 2) with primary SC (36.93%), 12,601 with AC (58.99%), and 872 with ASC Fig 1. (A) In this high-power photomicrograph of (A) an adenosquamous carcinoma of the lung, squamous component is identified at the right lower half of the picture, and glandular adenocarcinoma component is at the left upper half. High-power micrographs show (B) a well-differentiated adenocarcinoma and a (C) moderately differentiated squamous cell carcinoma of the lung. (Hematoxylin and eosin stain; for (A, B, C) magnification 10.)

Ann Thorac Surg COOKE ET AL 2010;90:943 8 SURVIVAL POSTLOBECTOMY FOR ASC, SC, AND AC 945 Table 1. Clinical Characteristics of Primary Squamous Cell, Adenocarcinoma, and Adenosquamous Carcinoma of the Lung a Variable Squamous Cell N b 7888 Adenocarcinoma N b 12,601 Adenosquamous Cell N b 872 Age, y 67.4 09.5 64.6 SD 66.1 10.5 Gender Male 5149 (65.3) 6146 (48.8) 479 (54.9) Female 2739 (30.4) 6455 (51.2) 393 (45.1) Ethnicity White 6516 (82.6) 10292 (81.7) 674 (77.3) Black 741 (9.4) 1077 (8.6) 89 (10.2) Other 615 (7.8) 1220 (9.7) 107 (12.3) Unknown 969 (12.3) 2083 (16.5) 146 (16.7) a All values were p 0.001. b Combined N for all three cohorts does not equal 21,361 due to incomplete data. SD Standard deviation. (4.08%). Patients with SC were older than AC and ASC patients (67.37 9.53 vs 64.61 10.68 and 66.14 10.52, respectively, p 0.0001). SC patients were more likely to be men than AC and ASC patients (65.28% vs. 48.77% and 54.93% respectively, p 0.0001). ASC patients were less likely to be of white ethnicity than SC and AC patients (77.29% vs 82.61% and 81.68% respectively, p 0.0001). Anatomically, SC tumors were significantly larger then AC and ASC tumors (41.03 mm 23.69 vs 33.00 21.37 and 37.68 26.43 respectively, p 0.0001). All three tumors tended to be right-sided lesions, and upper-lobe and middle-lobe predominant. Pathologically, patients with ASC had an incidence of cancer-positive lymph nodes greater than SC and less than AC patients (34.69% vs 30.98% and 36.43%, respectively, p 0.0001) and more frequently had a histologic grade of III and IV compared with SC and AC patients (55.74% vs 50.09% and 42.72%, respectively, p 0.0001). Early-Stage Patient Characteristics We identified 5309 patients who underwent lobectomy for early-stage disease: 3223 patients (60.7%) had AC (stage I, 2958 patients; stage II, 265 patients), 1862 patients (35.1%) had SC (stage I, 1551 patients; stage II, 311 patients), and 224 patients (4.2%) had ASC (stage I, 192 patients; stage II, 32 patients). Tables 3 and 4 summarize the patient characteristics as well as tumor anatomic and histologic characteristics. The data mirrored the results for combined-stages data. Patients with SC were older than patients with AC and ASC (p 0.0001), and fewer ASC patients were of white ethnicity than SC and AC patients (p 0.0001). All three tumors subtypes tended to be upper-lobe and middle-lobe predominant and right-side laterality. Most early-stage AC were histologic grade I and II (57.7%), whereas 52.3% of SC and 54.8% of ASC tumors were histologic grade III and IV (p 0.0001). Survival in Patients Undergoing Lobectomy for Early-Stage (I and II) Disease Overall survival was significantly reduced in both ASC and SC compared with AC (p 0.0001). Kaplan-Meier survival curves for combined stage I and II for the three groups (Fig 2A) indicated that ASC had the least favor- Table 2. Anatomic and Pathologic Characteristics of Primary Squamous Cell, Adenocarcinoma and Adenosquamous Carcinoma of the Lung a Squamous Cell Adenocarcinoma Adenosquamous Cell Characteristic N b 7022 N b 11,084 N b 822 Size, mm 41.0 23.7 33.0 21.4 37.7 26.4 Positive nodes, No. 6162 0.7 1.7 9201 1.0 2.4 720 0.8 2.7 Patients with positive nodes 1909 (31.0) 3352 (36.4) 247 (34.7) Laterality Left 3484 (44.2) 4949 (39.3) 368 (42.2) Right 4282 (54.3) 7339 (58.3) 496 (56.9) Tumor lobe Upper, lingula, middle 4652 (59.0) 8166 (64.8) 562 (64.5) Lower 2397 (30.4) 3130 (24.8) 247 (28.3) Grade I II 2968 (37.6) 51374 (40.7) 240 (27.5) III IV 3951 (50.1) 5384 (42.7) 486 (55.7) Unknown 969 (12.3) 2083 (16.5) 146 (16.7) a All values were p 0.001. b Combined N for all three cohorts does not equal 21,361 due to incomplete data. SD standard deviation.

946 COOKE ET AL Ann Thorac Surg SURVIVAL POSTLOBECTOMY FOR ASC, SC, AND AC 2010;90:943 8 Table 3. Clinical Characteristics of Primary Squamous Cell, Adenocarcinoma, and Adenosquamous Carcinoma of the Lung for Early-Stage Disease Characteristic Squamous Cell N a 1862 Adenocarcinoma N a 3223 Adenosquamous Cell N a 224 p Value Age, y 68.1 9.0 64.8 10.1 66.9 9.8 Gender Male 1098 (59.0) 1439 (44.7) 118 (52.7) Female 764 (41.0) 1784 (55.4) 106 (47.3) Ethnicity 0.0016 White 1560 (83.8) 2668 (82.8) 169 (75.5) Black 155 (8.3) 230 (7.1) 23 (10.3) Other 147 (7.9) 325 (10.1) 32 (14.3) Unknown 969 (12.3) 2083 (16.5) 146 (16.7) a Combined N for all three cohorts does not equal 5309 due to incomplete data. SD Standard deviation. able relative survival. The 5-year survival rates were 71.7% (95% CI, 70.9% to 72.6%) for AC, 66.1% (95% CI, 65.0% to 67.2%) for SC, and 59.4% (95% CI, 55.8% to 62.9%) for ASC (Fig 2A). Kaplan-Meier survival curves for stage I tumors (Fig 2B) also indicates that ASC has the least favorable survival. Five-year survival rates are 73.2% (95% CI, 71.5% to 75.0%) for AC, 69.2% (95% CI, 66.7% to 72.6%) for SC, and 62.0% (95% CI, 54.7% to 69.4%) for ASC (Fig 2B). For stage II, 5-year survival rates are 55.5% (95% CI, 49.1% to 61.9%) for AC, 50.8% (95% CI, 44.0% to 56.7%) for SC, and 41.5% (95% CI, 20.7 to 62.3%) for ASC (Fig 2C). Although ASC stage II tumors showed a trend toward worse survival, there was no statistical difference between the three groups (p 0.11; Fig 2C). The observed survival was improved in the combined early-stage SC relative to ASC, but the difference in this unadjusted survival was not statistically significant (p 0.187), potentially due to important confounders. Therefore, we next examined the relative survival among the three groups adjusted for important confounders. Clinical Variables Associated With Survival or Disease-Specific Death After Lobectomy for Early-Stage Disease Adjusted survival differences among the carcinoma groups, controlling for the covariates of age, sex, ethnicity, laterality, Table 4. Anatomic and Pathologic Characteristics of Primary Squamous Cell, Adenocarcinoma, and Adenosquamous Carcinoma of the Lung for Early-Stage Disease Characteristic Squamous Cell N a 1827 Adenocarcinoma N a 3182 Adenosquamous Cell N a 224 p Value Size, mm 36.2 19.6 29.0 17.1 33.6 22.5 Stage I 1551 (83.3) 2958 (91.8) 192 (85.7) II 311 (16.7) 265 (8.2) 32 (14.3) Laterality 0.0034 Left 793 (42.6) 1219 (37.8) 86 (38.4) Right 1068 (57.4) 2003 (62.2) 138 (61.6) Tumor lobe Upper, lingula, middle 1164 (64.5) 2291 (72.6) 145 (67.1) Lower 640 (35.5) 865 (27.4) 71 (32.9) Grade I II 823 (44.2) 1860 (57.7) 76 (33.9) III IV 974 (52.3) 1230 (38.2) 123 (54.9) Unknown 65 (3.5) 133 (4.1) 25 (11.2) a Combined N for all three cohorts does not equal 5309 due to incomplete data. SD standard deviation.

Ann Thorac Surg COOKE ET AL 2010;90:943 8 SURVIVAL POSTLOBECTOMY FOR ASC, SC, AND AC 947 tumor location, stage (I and II), and grade were evaluated using Cox proportional hazards regression analysis. Adjusted hazard ratios (HR) from the Cox model are provided in Table 5. ASC had significantly increased HR of 1.35 (95% CI, 1.08 to 1.69; p 0.009) relative to AC and to SC (HR, 1.27; 95% CI, 1.01 to 1.59; p 0.044). However, the HR of death between SC and AC was not significantly different after adjustment for the covariates (age, sex, ethnicity, laterality, tumor location, stage I and II, and grade). Other significant negative predictors of survival included tumor histologic grade of III and IV, stage II, age, male gender, and black ethnicity. Tumor lobe location and laterality were not predictive of survival. Comment Primary ASC of the lung is a rare tumor histology that appears to carry a worse prognosis than single histology primary AC and SC of the lung. Prior reports suggest that ASC represents 0.3% to 5% of non-small cell lung cancers [2, 3]. Our results across a broader population seem to confirm these observations, as we found ASC to be 4% of all postresection, all-stage patients with ASC, AC and SC, as well as early-stage patients. Table 5. Clinical Variables Associated with Survival or Disease Specific Death After Lobectomy for Early-Stage Disease Clinical Variable HR (95% CI) p Value Fig 2. Kaplan-Meier survival curves are shown for patients based on (A) combined early-stage, (B) stage I, and (C) stage II adenocarcinoma, squamous cell, and adenosquamous after lobectomy. (CI confidence interval.) Histologic type Adenosquamous vs 1.35 (1.08 1.69) 0.0094 adenocarcinoma Squamous cell vs 1.07 (0.96 1.12) 0.25 adenocarcinoma Adenosquamous vs 1.27 (1.006 1.593) 0.044 squamous cell Sex Male...... Female 0.77 (0.70 0.86).0001 Age 1.03 (1.02 1.03).0001 Ethnicity White...... Black 1.21 (1.01 1.46) 0.037 Other 0.84 (0.70 1.00) 0.06 Laterality Right...... Left 1.01 (0.91 1.12) 0.83 Tumor location Lower lobe...... Upper and middle lobe 0.95 (0.85 1.06) 0.32 Stage 0.95 (1.01 1.02).0001 I II 1.74 (1.52 1.99) Grade I II...... III IV 1.28 (1.16 1.43) Unknown 1.09 (0.84 1.41) 0.52 CI Confidence Interval; HR hazard ratio.

948 COOKE ET AL Ann Thorac Surg SURVIVAL POSTLOBECTOMY FOR ASC, SC, AND AC 2010;90:943 8 In regards to the more aggressive clinical and pathologic characteristics of ASC, Ruffini and colleagues [2] evaluated 1158 European patients who underwent resection for lung cancer. They found 33 patients (2.8%) with ASC. The ASC group presented with worse histologic grade, more advanced stage, and had a 3-year all-stage survival rate of 28% compared with 46% for patients with single histology. Gawrychowski and colleagues [4] examined data for 96 patients undergoing resection for ASC of the lung. The investigators found that the cumulative postoperative survival rates for patients with ASC at 5 and 10 years were 25.4% and 19.2%, compared with 42.5% and 39.1%, respectively, for a contemporaneous cohort of patients with AC. Nakagawa and colleagues [5] reviewed outcomes in 30 patients undergoing resection for ASC. The cumulative survival rate for patients with ASC and pathologic stages IA to IIB in their study population was similar to that of patients with stage IIIA AC or SC. Similar to these studies, our analysis showed that patients with ASC presented with higher-grade tumors (grade III and IV) and an incidence of nodal-positive disease higher than SC, although lower than AC. Anatomic location and laterality for ASC were similar to AC and SC. The average size of ASC tumors was smaller than SC tumors but larger than AC tumors. In addition, our analysis found that ASC tumors carried a poorer prognosis than AC and SC. For combined early-stage disease, patients with ASC exhibited significantly reduced 5-year survival compared with AC and SC, and ASC was a significant risk factor for death compared with SC and AC after controlling for other available variables. Although this study confirms the more aggressive clinical and pathologic characteristics of ASC compared with AC and SC, it does not explain why the tumor histology of ASC is more virulent. In our analysis, fewer ASC patients are of white ethnicity compared with AC and SC patients (p 0.001). This suggests the potential of a genetic or other molecular etiology for the aggressive clinical and pathologic presentation of ASC. Some groups have identified identical epidermal growth factor receptor (EGFR) abnormalities in ASC tumors that have been described in primary lung AC of Asian, female never-smokers [8 10]. Ohtsuka and colleagues [8] found EGFR tyrosine kinase-domain gene mutations in 2 of 4 patients with ASC, and in 0 of 24 patients with SC. Kang and colleagues [9] found identical EGFR mutations in both the AC and SC components of ASC tumors. The frequency of these EGFR mutations was similar to lung tumors in Asian patients with pure AC. In our study, the subset analysis of the clinical variable other ethnicity demonstrated the ASC cohort had a higher percentage of Asian/Pacific Islander patients (6.42%) compared with AC (5.60%) and SC (3.85%) cohorts (data not shown). However, further investigative studies are needed to confirm that EGFR mutations are the etiology of the aggressive clinical and pathologic characteristics of ASC. In addition, multiple social confounders may complicate the use of information from a large national database to attribute ethnic predilection of tumors to genetic factors. As previous groups have reported, potential socioeconomic health care disparities lead to unequal access to surgical therapy. This may mean that the result of ASC patients more likely being nonwhite, and black ethnicity patients having poorer survival, could have less to do with the genetics of the disease process, but because fewer nonwhites are undergoing surgical resection, or if they are, then at a higher stage, therefore confounding the results [11, 12]. This large review evaluated the clinical and pathologic characteristics of ASC. Our study demonstrates that ASC is an uncommon tumor and confirms that ASC exhibits distinct clinical behavior and carries a poorer prognosis than AC and SC. The decreased survival of patients with earlystage ASC compared with single histology AC and SC suggests a role of clinical trials for adjuvant chemotherapy or targeted molecular therapy, or both, for patients with early-stage ASC. This publication was made possible by Grant Number UL1 RR024146 from The National Center for Research Resources. References 1. Travis WD, Colby TV, Corrin B, Shimosato Y, Brambilla E. In collaboration with L.H. Sobin and pathologists from 14 countries. Histological typing of lung and pleural tumours. (International Histological Classification of Tumours), Geneva: World Health Organization; 1999:1 55. 2. Ruffini E, Rena O, Oliaro A, et al. Lung tumors with mixed histologic pattern. Clinico-pathologic characteristics and prognostic significance. Eur J Cardiothorac Surg 2002;22:701 7. 3. de Jong WK, Schaapveld M, Blaauwgeers JL, Groen HJ. Pulmonary tumours in the Netherlands: focus on temporal trends in histology and stage and on rare tumours. Thorax 2008;63: 1096 102. 4. Gawrychowski J, Bruliński K, Malinowski E, Papla B. Prognosis and survival after radical resection of primary adenosquamous lung carcinoma. Eur J Cardiothorac Surg 2005;27:686 92. 5. Nakagawa K, Yasumitu T, Fukuhara K, Shiono H, Kikui M. Poor prognosis after lung resection for patients with adenosquamous carcinoma of the lung. Ann Thorac Surg 2003;75: 1740 4. 6. National Cancer Institute. Surveillance epidemiology and end results. http://seer.cancer.gov/. 7. Goldstraw P, ed. Site-specific explanatory notes for lung and pleural tumours. International Association for the Study of Lung Cancer staging handbook in thoracic oncology. Orange Park, FL: Editorial Rx Press; 2009:53 74. 8. Ohtsuka K, Ohnishi H, Fujiwara M, et al. Abnormalities of epidermal growth factor receptor in lung squamous-cell carcinomas, adenosquamous carcinomas, and large-cell carcinomas: tyrosine kinase domain mutations are not rare in tumors with an adenocarcinoma component. Cancer 2007;109:741 50. 9. Kang SM, Kang HJ, Shin JH, et al. Identical epidermal growth factor receptor mutations in adenocarcinomatous and squamous cell carcinomatous components of adenosquamous carcinoma of the lung. Cancer 2007;109:581 7. 10. Sasaki H, Endo K, Yukiue H, Kobayashi Y, Yano M, Fujii Y. Mutation of epidermal growth factor receptor gene in adenosquamous carcinoma of the lung. Lung Cancer 2007; 55:129 30. 11. Lathan CS, Neville BA, Earle CC. Racial composition of hospitals: effects on surgery for early-stage non-small-cell lung cancer. J Clin Oncol 2008;26:4347 52. 12. Farjah F, Wood DE, Yanez D 3rd, Symons RG, Krishnadasan B, Flum DR. Temporal trends in the management of potentially resectable lung cancer. Ann Thorac Surg 2008;85:1850 5; discussion 1856.