The roles of adjuvant chemotherapy and thoracic irradiation

Transcription

1 Factors Predicting Patterns of Recurrence After Resection of N1 Non-Small Cell Lung Carcinoma Timothy E. Sawyer, MD, James A. Bonner, MD, Perry M. Gould, MD, Robert L. Foote, MD, Claude Deschamps, MD, Carla M. Lange, BS, and Hongzhe Li, PhD Divisions of Radiation Oncology, Thoracic and Cardiovascular Surgery, and the Section of Biostatistics, Mayo Clinic and Mayo Foundation, Rochester, Minnesota Background. Although irradiation and chemotherapy are unproved adjuvant treatments for completely resected N1 non small cell lung carcinoma, previous studies may have been diluted by the inclusion of low-risk patients. Risk factors in this situation, however, are not yet well defined. Methods. One hundred seven consecutive patients with complete resection of N1 disease who received no other therapy were studied to identify factors independently predicting the risk of freedom from local recurrence (FFLR), freedom from distant metastasis (FFDM), and overall survival (OS). Results. Twelve factors were assessed for a potential prognostic relationship with FFLR, FFDM, and OS. Regression analyses revealed that the factors independently associated with an improved outcome were positive bronchoscopic findings (FFLR, p 0.005), a greater number of dissected N1 nodes (FFDM, p 0.02), and a lesser T stage (OS, p 0.01). Classification and regression tree analyses were then used to separate the patients into risk groups. Conclusions. Although these results require corroboration in further studies, they may aid the design of trials examining therapies used to decrease rates of local recurrence or distant metastasis. (Ann Thorac Surg 1999;68:1171 6) 1999 by The Society of Thoracic Surgeons The roles of adjuvant chemotherapy and thoracic irradiation after surgical resection of non small cell lung carcinoma (NSCLC) with N1 involvement, as defined by the American Joint Committee on Cancer (AJCC) [1], are unclear. Although a recent meta-analysis suggested that adjuvant chemotherapy in this situation leads to improved survival [2], patients with N1 disease were not analyzed independently, and most prospective studies have concluded that both postsurgical chemotherapy and radiation therapy confer no overall survival benefit. It is possible that adjuvant chemotherapy or radiation therapy (or both) truly does not affect the survival outcomes of any patients with N1 disease. However, many of the studies examining this issue have considered patients in the AJCC N1 category as a single group. It is therefore also possible that any potential benefit of adjuvant therapy has been obscured by a diluting effect from inclusion of patients with N1 disease who are at low risk for distant metastasis, local recurrence, and death. It is further possible that examining the effect of adjuvant therapy specifically in patients with N1 involvement who are at particularly high risk for recurrence might uncover a previously obscured benefit. Several prognostic factors in resected NSCLC have been identified [3 8]. To date, most studies have used survival as the end-point. The current analysis attempts to define specific tumor and treatment related prognostic Accepted for publication April 3, Address reprint requests to Dr Bonner, Mayo Clinic, 200 First Street SW, Rochester, MN factors associated with a high risk for distant metastasis, local recurrence, and death (all as separate end-points) and therefore to suggest which patients are most likely to benefit from the administration of adjuvant chemotherapy or radiation therapy, or both. Material and Methods The medical records of all patients undergoing thoracotomy for lung cancer at the Mayo Clinic from January 1, 1987, to December 31, 1990, were reviewed. Patients with NSCLC who underwent resection of all gross tumors and who had AJCC N1 involvement of lymph nodes were included in this analysis. Specifically, patients with squamous cell carcinoma, adenocarcinoma, adenosquamous carcinoma, and large cell carcinoma were included, whereas those with bronchoalveolar carcinoma, carcinoid tumors, lymphoma, small cell carcinoma, and sarcoma were excluded. Other features that led to elimination from this analysis were AJCC N2 or N3 disease, proven metastatic (M1) disease before or at the time of surgical resection, administration of either preoperative or postoperative chemotherapy or radiation therapy, and gross residual disease after surgery. Pretreatment evaluation generally included history and physical examination, complete blood cell count, and serum chemistry evaluation (to include alkaline phosphatase, aspartate transaminase, calcium, and electrolytes), chest radiography, and chest and upper abdominal computed tomography scans. Other tests, such as bronchoscopy, head 1999 by The Society of Thoracic Surgeons /99/$20.00 Published by Elsevier Science Inc PII S (99)

2 1172 SAWYER ET AL Ann Thorac Surg PREDICTING RECURRENCE IN LUNG CARCINOMA 1999;68: computed tomography, and bone scan, were performed as clinically indicated. Specifically, results of bronchoscopy were considered positive only in patients with visible endobronchial disease. In general, bone scans and head computed tomography scans were performed only when symptoms suggested that results of these tests might be positive. The overwhelming majority of patients did not undergo either test. During the period of this study, the thoracic surgeons at the Mayo Clinic generally took a similar approach to thoracotomy and lymphadenectomy for patients with lung cancer. Except when pulmonary reserve was questionable in patients with very poor lung function, wedge resections and segmentectomies were not performed; lobectomy was done whenever bilobectomy or pneumonectomy was not anatomically necessary. In general, complete mediastinal lymph node dissection was performed, with complete removal of the high paratracheal, mid-low paratracheal, subcarinal, paraesophageal, and inferior pulmonary ligament lymph nodes. Aortic and subaortic lymph nodes were dissected for left-sided tumors. N1 nodes also were dissected completely, as appropriate for the primary surgical procedure performed. After the operative procedure and pathologic examination, patients were assigned a surgical stage in accordance with the AJCC. Lymph nodes were also assigned to one of 14 lymph node stations adapted from Naruke and associates [9]. N1 nodes corresponded to Naruke stations 10 through 12, with station 10 representing hilar nodes, station 11 representing interlobar nodes, and station 12 representing lobar nodes. Stations 13 and 14 represented segmental and subsegmental nodes, respectively. During the period of this study, however, stations 13 and 14 were not differentiated from each other; nodes from these stations were grouped together under the term intrapulmonary peribronchial. In the present analysis, all intrapulmonary nodes were therefore assigned to station 13. Although surgical intervention was quite uniform for the patients in this analysis, there was no uniform policy at the Mayo Clinic during the period of this study regarding indications for postoperative therapy. Given the lack of a proven survival benefit, the great majority of patients with N1 disease who had surgical resection during the period of this study received no adjuvant therapy. However, adjuvant irradiation was occasionally recommended and administered; a total of 13 patients with N1 involvement received radiation therapy. As stated above, patients who received radiation therapy were excluded from this analysis. No patients in the N1 group were given chemotherapy during the period of this study. End points in this analysis were freedom from local recurrence (FFLR), freedom from distant metastasis (FFDM), and survival. Local recurrences were defined as recurrences at the bronchial margin, in the original involved lobe or lobes, or in N1, N2, or N3 lymph nodes. Recurrences beyond these sites were considered distant metastatic lesions. The characteristic balance of patient, tumor, and treatment between the patients in this analysis and the patients excluded purely on the basis of having received neoadjuvant or adjuvant therapy was evaluated with Fisher s exact test [10]. Times to local recurrence, distant metastasis, and death were measured from the date of surgical resection. Patients were scored as having distant metastasis regardless of whether they had already experienced local recurrence, and vice versa. All FFLR, FFDM, and survival distributions were estimated by the actuarial method of Kaplan and Meier [11]. Comparisons between actuarial curves were made with the log rank test [12]. Multivariate analyses were performed by the methods of Cox [13]. For determination of FFLR, FFDM, and survival risk groups, classification and regression tree analyses were performed by the method of Breiman and associates [14]. Results The median follow-up for all patients in this analysis, including those who died, was 2.3 years. For the 24 patients still alive who had not had local recurrence or distant metastasis, minimum follow-up was 2.4 years (maximum, 8.1 years). The median age of all 107 patients in this analysis was 67.0 years, and the range was 40 to 84 years. Treatment and tumor related characteristics for all patients are shown in Table 1. Initially a comparison was made between the patients in this analysis, none of whom had received neoadjuvant or adjuvant therapy, and the patients excluded from analysis solely on the basis of having received additional primary therapy. Fisher s exact test performed for multiple potential prognostic factors revealed that there were no significant imbalances between the groups, with one exception. Patients not receiving adjuvant therapy (the patients included in the present analysis) were more likely than those excluded from the analysis to have involvement of nodal station 10 or 11 (p 0.05) (Table 1). Next, FFLR, FFDM, and survival rates were determined for the entire group of 107 patients who underwent complete resection and were found to have involvement of N1 lymph nodes. The Kaplan-Meier curves are shown in Figure 1. The 5-year actuarial FFLR, FFDM, and survival rates were 62%, 53%, and 32%, respectively. Conversely, the respective 5-year actuarial rates of local recurrence, distant metastasis, and death were 38%, 47%, and 68%. The absolute numbers of patients experiencing local recurrence, distant metastasis, and death were 25, 41, and 75, respectively. Multiple tumor and treatment related factors were analyzed for potential prognostic significance. These factors are shown in Table 2. Univariate analyses were performed to determine associations between these covariates and FFLR, FFDM, and survival. For two factors, pathologic margin status and tumor grade, univariate analyses were not performed. There were too few patients with grade 1 or 2 tumors (2) and positive margins (1) to allow for meaningful comparisons. Two factors had statistically significant associations with FFLR. Although patients with negative preoperative bronchoscopic findings had an FFLR rate very similar to that of patients who did not have a preoperative bronchoscopic procedure, both of these groups had a lower FFLR rate than patients with positive bronchoscopic

3 Ann Thorac Surg SAWYER ET AL 1999;68: PREDICTING RECURRENCE IN LUNG CARCINOMA 1173 Table 1. Patient, Tumor, and Treatment Characteristics in Patients With Complete Resection of N1 Non-Small Cell Lung Carcinoma Characteristic Number of Patients Surgery Alone Surgery With Irradiation a p Value b Preoperative bronchoscopy Positive results Negative results or not 59 5 performed Surgical procedure Segmentectomy or wedge resection Lobectomy, bilobectomy, or pneumonectomy No. of involved N1 nodes No. of N1 nodes dissected Tumor histologic type Squamous Nonsquamous 48 4 T classification T1 or T T3 or T Maximal tumor dimension, cm Nodal stations involved c 12 or 13 (or both) only or 11 (or both) with or without 12 or a Administration of radiation therapy led to exclusion from analysis. b Fisher s exact test. c As described by Naruke et al [9]. findings (log rank, p 0.03). Additionally, patients with nonsquamous carcinomas had a lower FFLR rate than patients with squamous tumors (log rank, p 0.04). Table 2. Potential Prognostic Factors Considered in Univariate Cox, Regression, and Classification and Regression Tree Analyses in 107 Patients With Complete Resection of N1 Non-Small Cell Lung Carcinoma Factor Comparison Groups Number a Preoperative Positive 48 bronchoscopy Negative 25 Not performed 34 Surgical procedure Segmentectomy or 9 wedge resection Lobectomy 45 Bilobectomy or 52 pneumonectomy Number of involved N1 nodes 3 21 Location of involved Station 12 or 13 (or 23 N1 nodes b both) only Station 10 or 11 (or 4 both) only 10 or 11 (or both) and or 13 (or both) Number of N1 nodes dissected Number of N2 nodes dissected Number of lung 1 95 lobes involved 1 12 Tumor grade 1 or 2 2 3or4 105 Tumor histology Squamous 48 Nonsquamous 59 AJCC tumor T1 31 classification T2 55 T3 or T4 21 Pathologic tumor 0 3 cm 53 size; in greatest 3cm 53 dimension Pathologic margin Negative 106 status Positive 1 a The numbers for each factor do not add up to 107. Occasionally, particular data could not be retrospectively ascertained. b Stations as described by Naruke et al [9]. AJCC American Joint Committee on Cancer. Fig 1. Kaplan-Meier freedom from local recurrence, freedom from distant metastasis, and survival curves for 107 patients with complete resection of N1 non small cell lung carcinoma. No patient received adjuvant or neoadjuvant therapy. Only one factor had a statistically significant association with FFDM. Patients with ten or fewer N1 nodes dissected and evaluated had a lower FFDM rate than those with more than ten nodes removed (log rank, p 0.04). Similarly, only one factor had a significant association with survival. Patients with T3 or T4 tumors had a lower survival rate than those with either T1 or T2 tumors (log rank, p 0.02; the Kaplan-Meier curves for T1 and T2 tumors were nearly superimposed and crossed 4 times). Subsequently, factors with independent prognostic significance for FFLR, FFDM, and survival were sought. Separate Cox regression analyses were performed for each end-point. Factors independently associated with an improved outcome were positive preoperative bronchoscopic result (for FFLR, p 0.005), more than 10 N1

4 1174 SAWYER ET AL Ann Thorac Surg PREDICTING RECURRENCE IN LUNG CARCINOMA 1999;68: Table 3. Freedom From Local Recurrence, Freedom From Distant Metastasis, and Overall Survival Risk Groups for 107 Patients With Complete Resection of N1 Non-Small Cell Lung Carcinoma a Risk FFLR FFDM OS Low Positive bronchoscopy 10 N1 nodes dissected and T1 or T2 Intermediate Negative bronchoscopy (or not No patients performed) and not meeting high-risk criteria High Negative bronchoscopy (or not 10 N1 nodes dissected performed) and tumor 3cm and/or T3 or T4 and 10 N1 nodes dissected T1 or T2 and positive bronchoscopy T1 or T2 and negative bronchoscopy (or not performed) T3 or T4 a Using a classification and regression tree analysis [14]. FFDM freedom from distal metastasis; FFLR freedom from local recurrence; OS overall survival. nodes dissected (for FFDM, p 0.02), and a primary tumor classification of T1 or T2 (for survival, p 0.01). Classification and regression tree analyses were used to divide patients into low, intermediate, and high-risk groups for FFLR, FFDM, and survival. The following factors were considered: status of the preoperative bronchoscopy, number of N1 nodes dissected, number of positive N1 nodes, location of positive N1 nodes, tumor histology, tumor size, and AJCC primary tumor classification. The number of lobes involved, the type of operation performed, and the number of N2 nodes dissected were not considered, because they completely lacked statistical significance on univariate analysis. Tumor grade and pathologic margin status also were not considered, because, as noted previously, not enough patients had positive margins or low grade. As mentioned above, the Kaplan-Meier survival curves for patients with T1 and T2 tumors overlapped, as did the FFLR curves for patients with negative bronchoscopic findings and those without bronchoscopy. Therefore, in the regression tree analyses, patients with T1 and T2 tumors were considered as one group, as were patients with negative bronchoscopic results and those without bronchoscopy. As expected, the factors with independent prognostic significance were those most useful for dividing patients into risk groups. Table 3 shows the results of the regression tree analyses for FFLR, FFDM, and survival. Of note, the analysis for FFDM proved useful in separating patients into only two risk groups. In Figure 2, the risk groups for each end-point are shown in graphic form. Next, the efficacy of the regression tree analyses in separating patients into risk groups was determined. First, Kaplan-Meier FFLR curves were generated to compare the outcomes for patients at low, intermediate, and high risk for local recurrence. The actuarial 5-year FFLR rates were 83% (n 48), 50% (n 50), and 24% (n 9), respectively (log rank, p 0.003) (Fig 3, top). A similar risk group comparison was performed with FFDM as the end-point. The respective actuarial 5-year FFDM rates were 66% (n 50) and 40% (n 57) for the low and high-risk groups (log rank, p 0.01) (Fig 3, middle). Finally, a similar comparison was performed for patients at low, intermediate, and high risk for a lower rate of survival. The actuarial 5-year survival rates were 44% (n 37), 28% (n 49), and 19% (n 21), respectively (log rank, p 0.01) (Fig 3, bottom). Fig 2. Classification and regression tree analyses were used to separate patients into risk groups according to (top) freedom from local recurrence, (middle) freedom from distant metastasis, and (bottom) survival. AJCC American Joint Committee on Cancer.

5 Ann Thorac Surg SAWYER ET AL 1999;68: PREDICTING RECURRENCE IN LUNG CARCINOMA 1175 Fig 3. Freedom from local recurrence (top), freedom from distant metastasis (middle), and overall survival (bottom) rates for patients at low, intermediate, and high risk for local recurrence were well separated, supporting the efficacy of classification and regression tree analyses in determining risk groups. Comment A primary advantage of the present analysis compared with previous analyses of patients with AJCC N1 involvement is the uniform treatment; as emphasized above, no patients in the present study received chemotherapy or radiation therapy as part of their primary treatment. Therefore, the recurrence patterns in these patients can be considered pure, because they were uninfluenced by nonsurgical treatment. A potential concern is that this group of 107 patients with completely resected N1 disease does not accurately represent all patients with complete resection of N1 disease, because patients receiving additional therapies are excluded. The potential bias exists that patients with more advanced N1 disease received adjuvant therapy, so that the patients in this series are those with early N1 disease. However, during the period of this study, radiation oncologists, medical oncologists, and surgeons at the authors institution did not routinely recommend adjuvant therapy for any patients with completely resected N1 disease, because a demonstrated survival benefit was lacking. As further evidence that it is reasonable to assume that this is a representative group of patients with N1 involvement, the actuarial 5-year survival rate (32%) is not dissimilar to the 5-year survival rates for patients with completely resected AJCC stage II or N1 disease seen at other institutions [3 6, 8, 15]. Finally, the application of Fisher s exact test revealed no significant prognostic factor imbalances, with one exception. Patients in the present analysis (those receiving surgical therapy only) were more likely to have involvement of nodal station 10 or 11 (Table 1). However, this factor was of no prognostic significance in this analysis. In addition, this imbalance must be interpreted with caution; only 13 patients were excluded because of having received adjuvant therapy. The results of this analysis suggest that multiple prognostic factors are important in predicting outcome. In particular, they lend strong support to the current AJCC staging system [1], because AJCC T classification was useful in defining groups at risk for FFDM and was an independent predictor of overall survival. At least one previous analysis suggested the importance of T classification in predicting distant metastasis in a similar group of patients [7]. The importance of T classification in predicting survival in patients with N1 also has been previously demonstrated [3, 4]. Conversely, at least one series has suggested that T classification does not have prognostic significance in this group of patients [6]. In addition, in an analysis by Martini and associates [5], size was an important prognostic factor but T classification was not, refuting the results of the present and other studies by suggesting that tumor size may be a more useful predictor of outcome than T classification. The importance of the extent of N1 nodal dissection was an unexpected but interesting finding. Possibly, removal of N1 lymph nodes is therapeutic as well as diagnostic and prognostic. However, it is also possible that increasing the number of lymph nodes removed simply increases staging accuracy, thereby increasing the likelihood that a given patient truly has only N1 disease. In contradiction to this hypothesis, however, the extent of N2 lymph node dissection was not of prognostic significance in the present study. Indirect support for a therapeutic benefit conferred by increasing the extent of lymph node dissection can be seen elsewhere in the lung cancer literature. Retrospective studies suggested, for example, that radical, systematic mediastinal lymph node dissection (as opposed to lymph node sampling) is associated with improved outcome in patients with resected N2 disease [16 18]. This finding was refuted in a later randomized, prospective study of 182 patients with N2 disease [19]. However, further analysis of the same 182 patients showed that

6 1176 SAWYER ET AL Ann Thorac Surg PREDICTING RECURRENCE IN LUNG CARCINOMA 1999;68: radical mediastinal lymph node dissection was associated with an increased ability to assess the number of N2 levels involved, suggesting that more disease is removed when the lymph node dissection is more nearly complete [20]. Certainly, if future studies confirm that the extent of nodal dissection correlates with prognosis, randomized trials would require that patients undergo complete dissection. It would no longer be appropriate to study patients who are high risk by virtue of having a few nodes removed. At present, however, the extent of dissection remains unproved as a prognostic factor. An interesting study design would use the extent of nodal dissection as a stratification factor. This would serve two purposes, allowing further analysis of the prognostic relevance of this factor and allowing the study of the effect of postoperative therapy, both in patients who have had and in those who have not had complete dissections. In the present analysis, the lower local recurrence and higher survival rates associated with positive preoperative bronchoscopic findings also were intriguing. Little work has been done investigating the status of preoperative bronchoscopy as a prognostic factor. It is possible that in the present study, the use of bronchoscopy also led to earlier identification of lung cancer, at a point when the disease was less extensive. However, cancer is more likely to be identified at the time of bronchoscopy in patients with more central disease [21]. Thus, another hypothesis is that when N1 lymph nodes are involved by more peripheral tumors, the lymphatic pathways through which lung cancer cells have traveled are longer, are less anatomically predictable, and are less likely to be completely removed at the time of surgical resection. Further work is indicated to assess the prognostic significance of preoperative bronchoscopy and the significance of central rather than peripheral location of the primary tumor. One potential limitation of this analysis is that patients without bronchoscopy were grouped together (in the regression tree analysis) with those who had negative bronchoscopic findings. As stated above, these two subgroups of patients were found to have very similar outcomes. Our presumption is that patients not undergoing bronchoscopy before surgery were those who were at low risk for a positive result. The overwhelming probability is that this presumption is safe; however, it cannot be retrospectively proven. The results of the present study refute data suggesting that hilar lymph node involvement (stations 10 and 11, as defined by Naruke and associates [9]) is associated with a less favorable outcome than involvement of only intrapulmonary or lobar lymph nodes (stations 12 and 13) [4, 6]. In the present analysis, the location of N1 involvement was not associated with the rate of FFLR, FFDM, or survival. The results of the present study suggest that the patients most likely to benefit from treatment that effectively reduces the rate of local recurrence are those with large peripheral (bronchoscopy-negative) tumors who have had fewer N1 nodes dissected. Those most likely to benefit from treatment that effectively reduces the rate of distant metastasis are those with T3 or T4 lesions or those with fewer N1 nodes dissected. In addition, although not supported by the present analysis, patients with other than squamous histologic type may also be at high risk for distant metastasis after surgical resection. The results of this study may be useful in the design of future trials evaluating the role of adjuvant therapy in patients with pathologically staged NSCLC and AJCC N1 disease. They should be corroborated in further studies and modified as appropriate. References 1. American Joint Committee on Cancer. Manual for staging of cancer. 4th ed. Philadelphia: JB Lippincott, 1992: Non-small Cell Lung Cancer Collaborative Group. Chemotherapy in non-small cell lung cancer: a meta-analysis using updated data on individual patients from 52 randomized clinical trials. Br Med J 1995;311: Mountain CF. A new international staging system for lung cancer. Chest 1986;89 Suppl:225S 33S. 4. Naruke T, Goya T, Tsuchiya R, Suemasu K. Prognosis and survival in resected lung carcinoma based on the new international staging system. J Thorac Cardiovasc Surg 1988;96: Martini N, Burt ME, Bains MS, McCormack PM, Rusch VW, Ginsberg RJ. Survival after resection of stage II non-small cell lung cancer. Ann Thorac Surg 1992;54: Yano T, Yokoyama H, Inoue T, Asoh H, Tayama K, Ichinose Y. Surgical results and prognostic factors of pathologic N1 disease in non-small cell carcinoma of the lung. Significance of N1 level: lobar or hilar nodes. J Thorac Cardiovasc Surg 1994;107: Ramacciato G, Paolini A, Volpino P, et al. Modality of failure following resection of stage I and stage II non-small cell lung cancer. Int Surg 1995;80: Komaki R, Fujii T, Perkins P, et al. Apoptosis and mitosis as prognostic factors in pathologically staged N1 non-small cell lung cancer. Int J Radiat Oncol Biol Phys 1996;36: Naruke T, Suemasu K, Ishikawa S. Lymph node mapping and curability at various levels of metastasis in resected lung cancer. J Thorac Cardiovasc Surg 1978;76: Kleinbaum DG, Kupper LL, Morgenstern H. Epidemiologic research: principles and quantitative methods. Belmont, CA: Lifetime Learning Publications, 1982: Kaplan EL, Meier P. Nonparametric estimation from incomplete observations. J Am Stat Assoc 1958;53: Mantel N. Evaluation of survival data and two new rank order statistics arising in its consideration. Cancer Chemother Rep 1966;50: Cox DR. Regression models and lifetables (with discussion). J R Stat Soc (B) 1972;34: Breiman L, Friedman JH, Olshen RA, Stone CJ. Classification and regression trees. Belmont: Wadsworth International Group, Holmes EC. Treatment of stage II lung cancer (T1N1 and T2N1). Surg Clin North Am 1987;67: Martini N, Flehinger BJ. The role of surgery in N2 lung cancer. Surg Clin North Am 1987;67: Naruke T, Goya T, Tsuchiya R, Suemasu K. The importance of surgery to non-small cell carcinoma of lung with mediastinal lymph node metastasis. Ann Thorac Surg 1988;46: Deschamps C, Shaw E, Jett J, et al. Significance of complete versus incomplete nodal staging and sampling in stages II/IIIA nonsmall cell lung cancer (NSCLC) [Abstract]. Proc Am Soc Clin Oncol 1994;13: Izbicki JR, Thetter O, Habekost M, et al. Radical systematic mediastinal lymphadenectomy in nonsmall cell lung cancer: a randomized controlled trial. Br J Surg 1994;81: Izbicki JR, Passlick B, Karg O, et al. Impact of radical systematic mediastinal lymphadenectomy on tumor staging in lung cancer. Ann Thorac Surg 1995;59: Radke JR, Conway WA, Eyler WR, Kvale PA. Diagnostic accuracy in peripheral lung lesions. Factors predicting success with flexible fiberoptic bronchoscopy. Chest 1979;76: