Lung Cancer 28 (2000) 21 – 27 www.elsevier.nl/locate/lungcan
Temporal trends in survival after surgical resection of localized non-small cell lung cancer Jeffrey B. Rubins a,*, Stephen L. Ewing b, Suzanne Leroy c, Edward W. Humphrey d, Vicki Morrison c a
Pulmonary Di6ision (111N), Department of Medicine, Minneapolis Veterans Affairs Medical Center, Uni6ersity of Minnesota, 1 Veterans Dri6e, Minneapolis, MN 55417, USA b Department of Pathology, Minneapolis Veterans Affairs Medical Center, Uni6ersity of Minnesota, Minneapolis, MN 55417, USA c Oncology Di6ision, Department of Medicine, Minneapolis Veterans Affairs Medical Center, Uni6ersity of Minnesota, Minneapolis, MN 55417, USA d Department of Surgery, Minneapolis Veterans Affairs Medical Center, Uni6ersity of Minnesota, Minneapolis, MN 55417, USA Received 16 August 1999; received in revised form 3 November 1999; accepted 9 November 1999
Abstract To test whether modern preoperative staging modalities and perioperative care improve survival after resection of localized non-small cell lung cancer (NSCLC), we retrospectively reviewed outcomes of 454 patients with NSCLC resected from 1947 through 1969 (designated pre-1970 cases), and 540 patients with cancers resected from 1981 through 1994 (designated post-1980 cases). Mean ages, histological subtypes, surgical stages, and types of surgical procedures differed significantly between the two groups. Compared with pre-1970 cases, post-1980 cases were older, had more adenocarcinoma and less squamous cell carcinoma, and had lesser proportions of advanced stage disease. Postoperative (day 30) mortality was significantly higher for resections of localized (stages 1 and 2) NSCLC prior to 1970. For patients surviving at least 30 days after surgery, subsequent survival after resection of localized NSCLC differed minimally between pre-1970 and post-1980 groups. We conclude that perioperative mortality after resection of localized NSCLC improved, but subsequent postoperative survival for these patients did not significantly improve over the 45-year period studied. © 2000 Elsevier Science Ireland Ltd. All rights reserved. Keywords: Lung neoplasms; Cancer; Lobectomy; Survival; Computed tomography; Mediastinoscopy
1. Introduction
* Corresponding author. Tel.: +1-612-725-2000, ext. 4400; fax: +1-612-727-5634. E-mail address:
[email protected] (J.B. Rubins)
Surgical resection remains the only potentially curative treatment for non-small cell lung cancer (NSCLC). By definition, resection with curative intent presupposes that all of the malignant tumor can be completely resected. Accordingly, preoper-
0169-5002/00/$ - see front matter © 2000 Elsevier Science Ireland Ltd. All rights reserved. PII: S 0 1 6 9 - 5 0 0 2 ( 9 9 ) 0 0 1 1 6 - 6
22
J.B. Rubins et al. / Lung Cancer 28 (2000) 21–27
ative evidence of advanced lymphatic or extrapulmonary spread of cancer precludes curative resection. Prior to modern radiologic and surgical techniques, extended resections and complete mediastinal node dissections were routinely performed in efforts to completely resect lung cancer. However, the advent of improved preoperative staging modalities, particularly computed tomography and mediastinoscopy in the mid-1970s, allowed a more accurate estimate of the local extent of tumor and identified patients with advanced nodal disease or metastatic cancer before surgery. Presumably the improved ability to identify preoperatively patients with localized NSCLC, as well as the advances in intensive care support for managing patients postoperatively, have translated into substantially improved outcomes for patients undergoing curative resection of NSCLC. We tested this assumption by comparing outcomes after surgical resection of localized NSCLC at a single institution before the availability of computed tomography and mediastinoscopy (1947–1969) and after their routine use in preoperative staging (1981 – 1994).
2. Methods
2.1. Identification of cases The clinical records of all patients admitted to the Minneapolis Veterans Affairs Medical Center between 1947 and 1969 with a diagnosis of carcinoma of the lung were extracted previously as described [5]. Of the 1707 admissions, the 454 cases of NSCLC that underwent resection and had adequate histology and staging data were designated for this study as the pre-1970 cases. For the comparison group, all patients who underwent thoracotomies for pulmonary lesions between January 1981 and June 1994 at the Minneapolis Veterans Affairs Medical Center were identified by review of thoracic surgery case logs as previously described [7]. The 1994 cut-off date was chosen to allow at least 5 years of followup for all patients. Patient demographic data and pathology reports were obtained from the hospital computer database and the pathology records.
Between 1981 and 1994, 540 of 786 patients underwent resection of NSCLC with curative intent at the Minneapolis VA and were designated post1980 surgery cases for this study. All tumors were staged by the TNM system of the American Joint Committee based upon pathological data reported from examination of the resected specimens and clinical staging data, where applicable [6]. Date last seen alive or date of death was obtained by scanning the hospital computer database for admissions, outpatient visits, and dates of death. For patients who could not be clearly identified as alive (outpatient or inpatient visit within the past 6 months) or dead from the hospital computer records, the Veterans Affairs Beneficiary Identification and Records Locator System database was used to ascertain dates of death.
2.2. Analysis Associations between categorical variables were analyzed with x 2 analysis and continuous variables with unpaired Student’s t-test. Survival curves were calculated by the Kaplan–Meier method, and compared using log-rank tests. Calculations of proportions of patients surviving 2 and 5 years after surgery were performed using life-tables analysis. Statistical comparison of ageadjusted survival was calculated by entering age and time of surgery (dichotomized as before 1970 or after 1980) using Cox regression analysis. Calculations were performed using SPSS® for Windows (release 6.1, Chicago, IL). All P values are two-tailed.
3. Results Patients undergoing resection for NSCLC prior to 1970 (between 1947 and 1969) were slightly but significantly younger than those resected after 1980 (between 1981 and 1994) (61.1 99.5 vs. 65.69 7.4, PB 0.001). Both groups of cases were greater than 99% male, and both groups had similar proportions who were actively smoking at least 1 pack of cigarettes per day at the time of surgery, or had smoked within the preceding
J.B. Rubins et al. / Lung Cancer 28 (2000) 21–27 Table 1 Histological subtypes among resected non-small cell lung cancer in pre-1970 and post-1980 cases Number of cases (percent of total)
Age, mean 9 SD* Male gender Active or recent tobacco use Histology** Adenocarcinoma Squamous cell carcinoma Adenosquamous carcinoma Large cell carcinoma
Pre-1970 (n =454)
Post-1980 (n=540)
61.19 9.5 99% 65%
65.697.4 99% 62%
81 (17.8) 335 (73.8)
231 (42.8) 273 (50.6)
3 (0.7)
1 (0.2)
35 (7.7)
33 (6.1)
* PB0.001 by t-test. ** PB0.0001 by x 2 analysis.
2 years (65% for pre-1970 and 62% for post-1980 groups). The distribution of histological subtypes among the resected NSCLC cases in the pre-1970 group was significantly different (PB0.0001) than the post-1980 cases (Table 1). Squamous cell carcinoma was the most common histological subtype in both groups. However, a significantly greater proportion of pre-1970 cases had squamous cell carcinoma compared with post1980 cases. Conversely, adenocarcinoma constituted over three times as many post-1980 cases compared with pre-1970 cases. Table 2 Surgical stage among resected non-small cell lung cancer in pre-1970 and post-1980 cases Stage*
1 2 3A 3B 4
Number of cases (percent of total) Pre-1970 (n = 454)
post-1980 (n= 540)
196 87 134 34 3
340 100 65 20 15
(43.2) (19.2) (29.5) (7.5) (0.7)
* PB0.0001 by x 2 analysis.
(63.0) (18.5) (12.0) (3.7) (2.8)
23
The distribution of surgical stages also differed significantly (PB 0.0001) between the pre-1970 and post-1980 cases. A greater proportion of cases undergoing resection prior to availability of CT and mediastinoscopy was advanced stage disease (stages 3A, 3B, or 4) at the time of surgery, whereas the majority of cases undergoing resection after 1980 were stage 1 (Table 2). Proportions of cases that were stage 2, determined by the presence of malignant cells in resected hilar nodes, and thus less accessible to preoperative staging by CT or mediastinoscopy, were similar in the two groups. Because both squamous cell carcinoma and more advanced disease were more common in the pre-1970 group, we speculated that squamous cell carcinoma might be associated with more advanced tumor stage at presentation because of its tendency to develop more proximally in the respiratory tract and thus remain radiographically occult longer. However, there was no significant correlation between histological subtype and surgical stage in either the pre-1970 or post-1980 groups. Specifically, for pre-1970 cases, 73.8% of stages 1 and 2 NSCLC was squamous cell carcinoma and 18.7% was adenocarcinoma (compared with 73.8% and 17.8%, respectively, of total pre-1970 cases). For post-1980 cases, 50% of stages 1 and 2 NSCLC was squamous cell carcinoma and 42.9% was adenocarcinoma, compared with 50.6% and 42.8% of total post-1980 cases. The types of surgical procedures performed for resection of NSCLC also differed significantly (PB 0.0001) between the pre-1970 and post-1980 groups. Mediastinoscopy, available only for the post-1980 group, was performed in 21.6% of cases which went on to resection with curative intent. Prior to 1970, 43% of resections were lobectomies, and 54% were pneumonectomies. After 1980, 65.7% of resections were lobectomies, and only 14% were pneumonectomies. In addition, the number of resections of less than a full lobe (wedge resections or segmentectomies) increased strikingly after the availability of CT scanning, from 14 of 454 resections (3.1%) before 1970 to 102 of 540 resections (18.8%) after 1980. The surgical stage of NSCLC disease resected using less complete procedures also changed between
J.B. Rubins et al. / Lung Cancer 28 (2000) 21–27
24
Table 3 Survival analysis of patients alive 30 days after resection of localized NSCLC Pre-1970
Post-1980
Stages 1 and 2 NSCLC Mean (months) Median (months) Two-year actuarial (%) Five-year actuarial (%)
68.3 38.8 51.6 32.1
67.8 44.5 53.8 32.3
Stage 1 NSCLC Mean (months) Median (months) Two-year actuarial (%) Five-year actuarial (%)
72.6 42.0 54.8 33.9
75.0 48.3 58.4 36.7
Lobectomy cases Mean (months) Median (months)
71.0 44.5
73.6 46.7
the two study periods. Only 50% of wedge resections and segmentectomies were performed for stages 1 and 2 disease prior to 1970, whereas 89.2% of those were performed for stages 1 and 2 disease after 1980. Analysis of total survival after surgery for stages 1 and 2 NSCLC revealed that postsurgical
survival was significantly (PB 0.001) better for cases resected after 1980 compared with those resected prior to 1970. Mean and median survivals for post-1980 cases were 66.2 and 42.5 months, compared with 59.4 and 27.9 months for pre-1970 cases. However, inspection of the survival curves revealed that much of the survival difference in the two groups occurred during the first month after surgery. Further analysis showed that a significantly greater proportion of pre-1970 cases died within 30 days after surgery, compared with post-1980 cases (13.1% vs. 3.9%, PB 0.0001). Prior to 1970, 35 of 245 (14.3%) pneumonectomy cases died within 30 days after surgery, compared with only 2 of 71 (2.8%) cases resected after 1980 (P=0.008). Similarly, 28 of 195 (14.4%) lobectomy cases prior to 1970 died within 30 days after surgery, compared with only 15 of 332 (4.5%) cases after 1980 (P=0.0001). When the survival analysis was restricted to stages 1 and 2 NSCLC cases alive at least 30 days after surgery, survival was slightly but not significantly better for the post-1980 group compared with the pre-1970 group (Table 3, Fig. 1). Mean survival differed by approximately 1 month, and
Fig. 1. Kaplan –Meier analysis of survival for patients alive 30 days after resection of stages 1 and 2 NSCLC.
J.B. Rubins et al. / Lung Cancer 28 (2000) 21–27
median survival by approximately 6 months. Two- and 5-year survival after resection differed by less than 5% between groups. Further restricting the analysis to only stage 1 NSCLC yielded similar results (Table 3). Because of the statistically significant differences in patient age between groups, survival was further analyzed by entering both age and time of surgery (dichotomized as pre-1970 and post-1980) into a Cox regression model. Age-adjusted survival of patients alive 30 days after resection of localized NSCLC tended towards a significantly better survival for post1980 cases, compared with pre-1970 cases (P = 0.051), although the magnitude of the difference was small. Because of differences in the distribution of surgical procedures performed before 1970 and after 1980, survival was also compared for patients undergoing lobectomy for localized NSCLC who survived 30 days after resection (Table 3). Again, mean and median survival differed minimally between groups, with a trend towards statistical significance for age-adjusted survival (P = 0.08). Additional analysis of survival by type of surgical procedure revealed that overall survival after wedge resection or segmentectomy was not statistically different from that after lobectomy in the resections performed after 1980. Survival was actually slightly better by approximately 10% for wedge/segmentectomy cases than for lobectomy cases during the first 2 postoperative years (even when discounting postoperative mortality), but then declined to approximately 10% below that of lobectomy cases from 2 to 5 years after surgery.
4. Discussion Our retrospective study of surgical resection of NSCLC at a single institution over a 45-year period was designed to test the assumption that survival after resection of NSCLC has substantially improved in the modern era. Significant advances in preoperative staging modalities, particularly computed tomography and mediastinoscopy, and in perioperative intensive care, might have been expected to translate into signifi-
25
cantly improved outcomes of surgical treatment of NSCLC. We did find that postoperative 30 day mortality was significantly less after surgical resections, both pneumonectomies and lobectomies, performed after 1980 compared with those performed before 1970. Thus, our data suggest that modern surgical techniques and postoperative care have had a major impact on short-term outcomes after resection of localized NSCLC. However, subsequent survival for patients alive 30 days after resection of stages 1 and 2 NSCLC improved minimally in the post-1980 cases compared with the pre-1970 era. The difference in calculated mean survival was less than 2 months, and actuarial survival estimates at 5 years were essentially identical in both pre-1970 and post1980 groups. The retrospective design of this study limits the interpretation of these findings. The actual impact of chest CT and mediastinoscopy on preoperative staging cannot be evaluated in a retrospective study. Also, rigorous definitions for assigning cause of death could not be established; consequently, only total survival rather than diseasefree survival can be measured. However, total survival is often considered a more reliable outcome, because assignment of cause of death may be erroneous. In addition, the fact that most of the mortality occurred within the first 2 years after surgery in both groups makes this measure clinically relevant. Another potential limitation of a retrospective study is the absence of established definitions of medical comorbidities and functional status prior to surgery; thus, it was not possible in this study to reliably compare survival in the two groups stratitified by severity of underlying illnesses. However, because of the sizable number of patients in both groups and the fact that most were evaluated and resected under the supervision of the same surgeon (E.W.H.) during the period of this study, it seems unlikely that these factors differed enough between groups to offset real differences in survival. Because of the significant differences in the types of surgical procedures performed in the pre-1970 and post-1980 cases, we considered that true differences in survival after resection of localized NSCLC might be obscured by differences in survival related to sur-
26
J.B. Rubins et al. / Lung Cancer 28 (2000) 21–27
gical procedure. For instance, wedge resections and segmentectomies might be performed selectively in patients whose underlying conditions would prohibit more complete resections, and these same medical comorbidities might shorten their postoperative survival. However, restricting the survival analysis to only cases undergoing lobectomy, the most common surgical procedure in both groups and a procedure requiring reasonably robust preoperative functional status, still did not reveal significant differences in survival between the two groups. Assuming then that patients undergoing resection before 1970 and after 1980 had comparable performance status and that the substantial mortality within 5 years of intended curative resection was related to NSCLC, our data suggest that our ability to identify truly localized NSCLC that would be cured by resection has not improved markedly over 5 decades, despite dramatic innovations in preoperative staging modalities. Chest and body CT scanning, which have become routine in the evaluation of the local and distant extent of lung cancer over the past 20 years, identify patients who require nodal staging by mediastinoscopy. The impact of the availability of chest CT and mediastinoscopy after 1980 was suggested in our study by the greater proportion of stage 1 cases among resected NSCLC. Although not directly testable in a retrospective study, the shift towards more stage 1 disease among cases undergoing resection after 1980 is likely the result of the prethoracotomy identification and exclusion of patients with advanced local or nodal disease using modern staging modalities. In support of this interpretation, the proportion of stage 2 NSCLC, which cannot be reliably identified preoperatively by chest CT and mediastinoscopy, was not different in the pre-1970 and post-1980 groups. However, CT imaging and mediastinoscopy likely lack sufficient sensitivity to reliably detect the extent of clinically-occult spread of lung cancer [1]. Therefore, a substantial proportion of patients who seem to have localized NSCLC by preoperative CT imaging, mediastinoscopy, and even intraoperative staging succumb to recurrent (or possibly a second) lung cancer within a few years after intended curative resection.
Our results did confirm two ongoing trends in treatment of NSCLC that have been noted previously. We found an increased proportion of adenocarcinoma, and a decrease in squamous cell carcinomas, in resections performed after 1980 compared with those performed prior to 1970, in agreement with previous reports of the Veterans Administration Lung Group [2] and national survey of the Commission on Cancer of the American College of Surgeons [4]. The reasons for this shift in histological subtypes have not been clear from previous reports. Adenocarcinoma is reported to be more prevalent among younger populations with lung cancer, but in the present study, our post-1980 population with a higher prevalence of adenocarcinoma was older than the pre-1970 group. Conceivably, the increased prevalence of adenocarcinoma among resected NSCLC may reflect an increased detection rate for incidental peripheral lung lesions, typically adenocarcinomas, due to the increased use of routine chest radiography after 1970. We also confirmed a substantially increased use of wedge or segmental resection in 1980s compared with earlier decades [4]. Presumably, this increase reflects the increased use of chest CT in the modern era to define more patients as candidates for a more restricted resection. Again, our retrospective analysis did not allow us to confirm our clinical suspicion that more limited procedures were performed selectively on patients unable to tolerate a more complete resection. Of note, overall survival after wedge resection/segmentectomy was not significantly different than that after lobectomy. Initial mortality was actually lower for limited resection than for lobectomy, but mortality from 2 to 5 years after limited resection was higher, presumably from recurrent disease [3].
5. Conclusion Perioperative mortality after resection of localized NSCLC has improved in the modern era compared with surgery prior to 1970, but overall mortality for patients surviving surgery has not improved dramatically over nearly 50 years. The continued substantial recurrence rate and mortal-
J.B. Rubins et al. / Lung Cancer 28 (2000) 21–27
ity within a few years after intended curative resection of NSCLC suggests the need to incorporate cellular and genetic markers of tumor biology in the staging of NSCLC to more accurately distinguish patients with truly localized disease who might be cured by surgery from those with microadvanced disease, who would require new systemic treatment modalities.
Acknowledgements We thank Mary Macauley for her invaluable assistance in maintaining the Thoracic Surgery database, Patti Baribeau for assistance with the Minneapolis VA Tumor Registry, and Drs Dennis Niewoehner and Hanna Rubins for their critical reviews of this manuscript. Supported by the Department of Veterans Affairs Research Service.
.
27
References [1] Colice GL. Chest CT for known or suspected lung cancer. Chest 1994;106:1538 – 50. [2] Cox JD, Yesner RA. Adenocarcinoma of the lung: recent results from the Veterans Administration Lung Group. Am Rev Resp Dis 1979;120:1025– 9. [3] Ginsberg RJ, Rubinstein LV. Randomized trial of lobectomy versus limited resection for T1 N0 non-small cell lung cancer: Lung Cancer Study Group. Ann Thorac Surg 1995;60:615 – 22. [4] Humphrey EW, Smart CR, Winchester DP, et al. National survey of the pattern of care for carcinoma of the lung. J Thorac Cardiovasc Surg 1990;100:837 – 43. [5] Mayer JE Jr, Ewing SL, Ophoven JJ, Sumner HW, Humphrey EW. Influence of histologic type on survival after curative resection for undifferentiated lung cancer. J Thorac Cardiovasc Surg 1982;84:641 – 8. [6] Mountain CF. A new international staging system for lung cancer. Chest 1986;89:225S – 33S. [7] Rubins JB, Rubins HB. Temporal trends in the prevalence of malignancy in resected solitary pulmonary lesions. Chest 1996;109:100 – 3.