| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
J Thorac Cardiovasc Surg 2002;124:113-122
© 2002 The American Association for Thoracic Surgery
General Thoracic Surgery (GTS) |
From the Departments of Thoracic and Cardiovascular Surgery,a Biostatistics and Epidemiology,b Hematology and Medical Oncology,c Neurological Surgery,d and Radiation Oncology,e The Cleveland Clinic Foundation, Cleveland, Ohio.
Read at the Eighty-first Annual Meeting of The American Association for Thoracic Surgery, San Diego, Calif, May 6-9, 2001.
Received for publication April 27, 2001. Revisions requested July 10, 2001; revisions received Sept 28, 2001. Accepted for publication Nov 9, 2001. Address for reprints: Thomas W. Rice, MD, The Cleveland Clinic Foundation, 9500 Euclid Ave, Desk F25, Cleveland, OH 44195 (E-mail: ricet{at}ccf.org).
 |
Abstract |
|---|
 Top Abstract IntroductionPatients and methodsResultsDiscussionAppendix AAppendix BAppendix: DiscussionReferences |
|---|
|
Introduction |
|---|
|
TopAbstractIntroductionPatients and methodsResultsDiscussionAppendix AAppendix BAppendix: DiscussionReferences |
|---|
1,2 Neoadjuvant therapy for stage III NSCLC has produced a unique cohort of patients who have survived to have metachronous brain metastases develop. 3 Optimal management in these cases is uncertain, 4 a difficulty that provided the impetus for this study. The purposes of this study were to identify management strategies that maximize survival of patients with stage III NSCLC and metachronous brain metastases and to determine whether any apparent improved survival was due to treatment or simply patient selection. |
Patients and methods |
|---|
|
TopAbstractIntroductionPatients and methodsResultsDiscussionAppendix AAppendix BAppendix: DiscussionReferences |
|---|
Cross-sectional systematic follow-up was done in February 2001. Eighty-three patients were dead. Mean follow-up was 36 months (range 2.5 to 96 months) for 8 living patients. Time zero for time-related analysis was measured from diagnosis of metachronous brain metastases. All-cause mortality was the study end point.
Data analysis
Identification of optimal management
The goal of assessing management was identification of strategies that maximized survival. Nonparametric estimates of survival after development of brain metastases were obtained using the Kaplan-Meier method. Parametric estimates of survival were obtained by hazard function analysis. 5 Survival estimates were compared between subgroups using the log-rank test.
Both multivariable Cox proportional hazards analysis and parametric hazard function analysis of all variables listed in Appendix Table A were used to identify risk factors for death after development of brain metastases. A stepwise selection procedure was used, with a P = .05 criterion for retention of variables in the final model. This procedure was verified by bootstrap bagging with 1000 samples. 6 Variables identified in at least 50% of the bootstrap analyses were considered reliable. Continuous and ordinal variables were assessed univariably by decile risk analysis to determine whether their relationship was well calibrated to model outcome or whether transformations of scale were required.
Treatment benefit or patient selection?
Because there was no treatment protocol, patient selection itself could lead to spurious conclusions that certain therapies improved survival. At the same time, the resulting variance in management permitted separation of treatment effect from that of patient selection by means of propensity score matching. 7 The propensity model included 10 variables: gender, histologic type, stage at treatment of NSCLC, lung resection (vs none) for NSCLC, disease-free interval, number of brain metastases, local recurrence of NSCLC, presence of distant metastases other than in the brain, Eastern Cooperative Oncology Group (ECOG) performance status at diagnosis of brain metastases, and a recursive partitioning analysis (RPA) score. (RPA score ranges from 1 to 3 and is composed of age, performance status, recurrence of NSCLC, and presence of other distant metastases. 8) All these variables were entered into a multivariable logistic regression model of metastasectomy or stereotactic radiosurgery versus palliation. For each patient, the model was used to calculate the probability of undergoing metastasectomy or stereotactic radiosurgery. This probability is the propensity score.
The propensity score was used in two ways to separate treatment from selection effects. First, we used it to compare survival benefit of metastasectomy or stereotactic radiosurgery within propensity-matched groups. For this, patients were sorted according to the propensity score and divided into two equal-sized groups. We then determined the effect of metastasectomy or stereotactic radiosurgery on survival within each of the groups using the Kaplan-Meier method and log-rank test. To assess similarity within each group, t tests and contingency tables were used to compare patient factors, tumor factors, and treatment of NSCLC between patients who underwent metastasectomy or stereotactic radiosurgery and those who did not.
Second, the propensity score was used to confirm the results of the multivariable analysis that metastasectomy or stereotactic radiotherapy was beneficial. 9 For this, the propensity score was incorporated into the multivariable survival model of optimal management, and its adjustment of treatment effect magnitude and statistical significance was observed.
Presentation
Hazard and odds ratios are accompanied by 95% confidence intervals (CIs). Depiction of the ideal candidate for metastasectomy or stereotactic radiosurgery and the ideal candidate for palliation of brain metastases used the hazard function analysis. Specific values were entered into the multivariable equation, the equation was solved, and results were presented graphically with 68% CIs equivalent to 1 SE.
|
Results |
|---|
|
TopAbstractIntroductionPatients and methodsResultsDiscussionAppendix AAppendix BAppendix: DiscussionReferences |
|---|
Characterization and treatment of brain metastases
Forty-two patients (46%) had a single brain metastasis, 15 (16%) had 2 metastases, 16 (18%) had 3, and 18 (20%) had 4 or more. Sixty-two patients (68%) had unilateral brain metastases. Treatment of brain metastases was whole-brain radiation therapy in 43 patients (47%), metastasectomy alone or in combination with whole-brain radiation therapy in 23 patients (25%), and stereotactic radiosurgery with or without whole-brain radiation therapy in 24 patients (26%). One patient received no treatment.
The impact of stage III NSCLC and its treatment on metachronous brain metastases is described in Appendix A.
Optimal management
Risk-unadjusted survivals at 3, 6, 12, and 24 months after diagnosis of metachronous brain metastases were 70%, 45%, 22%, and 10%, respectively, with a median survival of 5.2 months (Figure 1). Risk factors for death are listed in Table 1. Good survival was predicted by stage IIIA, resection of NSCLC, younger age at diagnosis of brain metastases, good performance status, no other distant metastases, and metastasectomy or stereotactic radiosurgery (Figure 2 and Appendix Figure A).
|
|
|
, A, demonstrates the clinically important survival benefit of metastasectomy or stereotactic radiosurgery relative to palliation for metachronous brain metastases in a patient with characteristics predictive of good survival. Figure 3, B, demonstrates the statistically significant but clinically unimportant survival benefit of metastasectomy or stereotactic radiosurgery relative to palliation for metachronous brain metastases in a patient with characteristics predictive of poor survival.
|
|
|
|
|
Discussion |
|---|
|
TopAbstractIntroductionPatients and methodsResultsDiscussionAppendix AAppendix BAppendix: DiscussionReferences |
|---|
10 Survival and quality of life for patients with solitary brain metastases are superior when metastasectomy is included in the treatment. 11 However, complete resection of the primary lung cancer and control of the primary and other distant extracranial metastatic sites determine outcome after brain metastasectomy (Appendices A and B).
Solitary brain metastasis has been a selection factor for metastasectomy; however, many patients have more than one brain metastasis. 10,12 This and other selection factors have produced a literature of treatment of brain metastases in highly selected patients. Many readers attribute good survival to a given treatment; however, it may actually be due to patient selection. Thus, although it is intuitively apparent that some patients will benefit from aggressive treatment of their brain metastases, it remains unclear which patients should be treated and how. Optimal treatment of patients with NSCLC brain metastases cannot be generalized, because they are heterogeneous and treatment must be individually selected.
One aspect of heterogeneity is stage of primary NSCLC. Neoadjuvant treatment of stage III NSCLC produces a unique cohort of patients who have the primary NSCLC controlled and survive to have brain metastases develop. 3 Another aspect is the method of treatment of the brain metastases. Although no phase III trials comparing metastasectomy with stereotactic radiosurgery have been completed, our report and others have demonstrated similar survivals between these modalities. 13-15 Only technical considerations (large metastasis and neurologic deterioration from mass effect) limit the use of stereotactic radiosurgery.
In this report, both metastasectomy and stereotactic radiosurgery produced a survival advantage relative to whole-brain radiation or symptomatic therapy (palliation). For the patient with characteristics predictive of poor survival, this advantage was statistically significant but clinically unimportant. Therefore, in such patients quality of life becomes the goal of therapy. Stereotactic radiosurgery spares the patient the trauma of metastasectomy and may provide the best palliation. Whole-brain radiation therapy should be reserved for patients who are not candidates for stereotactic radiosurgery.
For the patient with characteristics predictive of good survival—younger age (Appendix Figure A), previous resection of NSCLC, stage IIIA, good performance status at diagnosis of brain metastasis, and no other distant metastases—survival was better if cure rather than palliation was the goal. The "good-risk" patient is denied long-term survival if treated with palliative intent using only whole-brain radiation or symptomatic therapy. Because of equivalent survivals with metastasectomy and stereotactic radiosurgery in these patients and the lack of any phase III trials, there is uncertainty—even among ourselves—about which of these two modalities should be offered to which patient.
Treatment benefit or treatment selection?
Lack of a management protocol produced variations in treatment and thus provided a spectrum of patient-therapy combinations. A patient with characteristics predictive of good survival was more likely to be selected for metastasectomy or stereotactic radiosurgery, whereas a patient with characteristics predictive of poor survival was more likely to be selected for whole-brain radiotherapy or symptomatic therapy. Propensity score methodology was developed to address nonrandomized comparisons in which therapy has not been applied either randomly or systematically. 7,9
Use of the propensity score in this study allowed patient selection to be separated from treatment effect. We used propensity scores in two ways in the analysis. First, we found in propensity-matched patients a uniform benefit of treating metachronous brain metastases with metastasectomy or stereotactic radiosurgery. Second, we introduced the propensity score into the multivariable analysis of optimal treatment and confirmed the benefit of metastasectomy or stereotactic radiosurgery. Both analyses demonstrated a clinically significant benefit only for the good-risk patient. 9
Limitations
This was a small, single-institution clinical experience. The study required that all patients survive to develop brain metastases, making interim death a selection factor. Patients were selected for treatment by clinical criteria; treatment was administered in a nonrandom fashion. To limit the influence of nonrandomization, we used propensity scores as described previously. Despite our use of 10 variables to develop the score and produce well-matched groups, the propensity score did not match for unrecorded variables that may have influenced selection.
Conclusion
Younger patients with resected stage IIIA NSCLC who have isolated metachronous brain metastases and good performance status do best when treated with metastasectomy or stereotactic radiosurgery. Survival benefit is a brain treatment effect and not the result of selecting the best candidates for aggressive therapy.
|
Appendix A |
|---|
|
TopAbstractIntroductionPatients and methodsResultsDiscussionAppendix AAppendix BAppendix: DiscussionReferences |
|---|
Stage III non–small cell lung cancer
Time interval between treatment of NSCLC and diagnosis of brain metastases was highly variable. According to multivariable analysis, longer disease-free intervals were seen in men (hazard ratio 0.55, 95% CI 0.35-0.86, P = .009), for squamous cell carcinoma versus large cell undifferentiated carcinoma (hazard ratio 0.47, 95% CI 0.24-0.91, P = .02), for adenocarcinoma versus large cell carcinoma (hazard ratio 0.58, 95% CI 0.33-1.03, P = .06), and in patients who had undergone pneumonectomy (hazard ratio 0.61, 95% CI 0.37-1.0, P = .05).
Patients who had undergone pneumonectomy had fewer brain metastases (n = 18/22 [82%] with 1 or 2 brain metastases) than did those who had undergone segmentectomy or lobectomy (n = 20/37 [54%] with 1 or 2 brain metastases, P = .03). Unilateral brain metastases were present in 21 patients who had undergone pneumonectomy (95%) and in only 24 who had undergone segmentectomy or lobectomy (65%, P = .008). Bilateral brain metastases were more common among patients who did not undergo resection of NSCLC (n = 15/32, 47%) than among patients who had resection (n = 14/59, 24%, P = .02). However, the numbers of brain metastases in these two groups were similar (P = .6). Other distant metastases were more frequent at the time of diagnosis of brain metastases among patients who had undergone resection of NSCLC (n = 12/59, 20%, vs n = 2/32, 6%, P = .08).
|
|
Sites of active tumor found at death differed with treatment of NSCLC. Among 32 patients without resection, sites of cancer were locoregional (lung) in 25 cases (78%), brain in 21 (66%), and other distant metastases in 9 (28%). In contrast, among 32 patients with segmentectomy or lobectomy, sites of cancer were locoregional in 14 cases (44%), brain in 18 (56%), and other distant metastases in 7 (22%). Among 19 patients with pneumonectomy, sites of cancer were locoregional in 7 cases (37%), brain in 8 (42%), and other distant metastases in 8 (42%).
|
|
Appendix B |
|---|
|
TopAbstractIntroductionPatients and methodsResultsDiscussionAppendix AAppendix BAppendix: DiscussionReferences |
|---|
Treatment also varied with disease-free interval. Disease-free intervals were longest for patients who had metastasectomy, shorter for those receiving stereotactic radiosurgery, and shortest for those receiving palliation.
Treatment varied with the number of brain metastases. Two or fewer brain metastases were seen in 87% of patients who had metastasectomy, in 75% who had stereotactic radiosurgery, and in 43% who had palliation. Unilateral brain metastases were more likely to be treated with metastasectomy (87%) or stereotactic radiosurgery (79%) than with palliation (52%). NSCLC was controlled in 92% of patients receiving stereotactic radiosurgery, in 65% undergoing metastasectomy, and in 48% receiving palliation. Other distant metastases were present in 4% of patients receiving stereotactic radiosurgery and in 9% undergoing metastasectomy, but in 25% receiving palliation.
|
Appendix: Discussion |
|---|
|
TopAbstractIntroductionPatients and methodsResultsDiscussionAppendix AAppendix BAppendix: DiscussionReferences |
|---|
I would like to pose a few questions. First, please tell us more about the long-term survivors. Were they all young patients with resected stage IIIA disease and single brain metastases, or are there some older ones in the group? Also, perhaps we should use functional status rather than age in the treatment algorithm, because even though posttreatment survival in your series was a little worse among elderly patients, there is no other effective therapy.
Second, do you currently ever treat patients with stage IIIB lung cancer with aggressive brain resection since the analysis of your results? If so, why?
Third, what is your denominator for this study, first in terms of all patients treated at your institution with stage III lung cancer and second in terms of all patients with stage III cancer who had distant metastases develop? How many of those are just in the brain?
Fourth, are isolated brain metastases the most common first site of failure in your institution for stage III lung cancer? We and others have noted that, with a rate greater than 50% in our series. If that is the case, should we consider prophylactic whole-brain radiation for patients who undergo resection of stage IIIA lung cancer?
Finally, how do you currently treat a patient with a T1 N2 M1 NSCLC whose only synchronous metastatic focus is an isolated 2 cm brain metastasis?
Dr Moazami. Dr Bueno, thank you for your comments. I will try to address your questions in the order that you have addressed them.
You asked about the long-term survivors in this study. We had a total of 91 patients, 8 of whom were alive at the end of our follow-up period, which was completed in February 2001. Mean follow-up of the patients was 36 months. Their mean age was 48 years (range 36-61 years). As you suggested, the patients in the survival group actually correlated well with the variables that we identified for best survival; that is, they were generally younger patients with good performance status and controlled stage IIIA disease in whom we had not detected any distant metastases other than in the brain.
Your second question was whether we would aggressively treat the brain lesion in a patient with stage IIIB given the results of our data. I think that in general the results of all retrospective studies such as this provide us with some guidelines, but obviously treatment needs to be considered on a patient-to-patient basis. For a patient with stage IIIB NSCLC that has been aggressively treated with good locoregional control, we believe that it is reasonable to proceed with aggressive treatment of an isolated brain lesion with either resection or perhaps stereotactic radiosurgery. Future studies should clarify whether stereotactic radiosurgery or metastasectomy is the better treatment. Again, one of the purposes of this extensive analysis was to show that, regardless of other factors, metastasectomy or stereotactic radiosurgery does confer a survival benefit on these patients.
You asked about the denominator of the study. We did an extensive review of all the medical records at the Cleveland Clinic during this 10-year period to try to select all the patients diagnosed with stage III NSCLC. The total number of patients was 891. Of those, we selected only the patients who had metachronous brain metastases develop, which we considered as 3 months from the time of diagnosis of lung cancer. This gave us the numerator, which was 91 patients.
You asked whether we believe that the brain is commonly the isolated site of failure and whether prophylactic whole-brain radiation therapy should be given to this patient population. We similarly found that in our study group, 46% of these patients had an isolated brain metastasis. Whether we should prophylactically radiate the brain has not been clarified in the literature. Most of the studies that I am aware of have not shown any benefit of prophylactic whole-brain radiation in any of the populations, with the possible exception of those with small cell cancer.
Finally, you asked about a patient with a stage IIIA disease, T1 N2 M1, who has only a synchronous brain metastasis. Obviously, the study was not meant to address synchronous brain metastases. However, we believe that once an aggressive approach to locoregional control is established, it is also reasonable to proceed with an aggressive treatment regimen for the brain. How to proceed in treating a patient like this depends on the variables we have discussed, including performance status.
Dr Robert J. Ginsberg (Toronto, Ontario, Canada). You had 8 long-term survivors. How many had 2 or more brain metastases?
Dr Moazami. Of the 8 survivors, 5 had 1 brain metastasis, 1 had 2 metastases, and 2 had 3 metastases.
Dr Ginsberg. Why was it not one of your conclusions that you should limit this aggressive therapy to single metastases?
Dr Moazami. Well, there was a sharp cutoff in survival when we divided the patients according to 1 to 2 brain metastases versus 3 or more. In our conclusion we stated that the best candidates are the ones that need to be aggressively treated. In this best profile that we constructed, one factor was the presence of 1 to 2 metastases.
Dr Ginsberg. But you did not differentiate between 1 versus 2?
Dr Moazami. There was no difference in survival between patients with 1 or 2 brain metastases. Predicted Kaplan-Meier median survival was 6.3 months for 1 brain metastasis and 7.3 months for 2 brain metastases.
Dr Ginsberg. Were any patients with 2 metastases treated with surgery?
Dr Moazami. Of 15 patients with 2 brain metastases, 7 had palliative whole-brain radiotherapy, 4 had stereotactic radiotherapy, and 4 had resection with whole-brain radiotherapy.
Dr Douglas E. Wood (Seattle, Wash). How do you feel that you corrected for selection bias in the second half of your analysis, that is, that the factors you identified, in vanishingly smaller numbers, represent a treatment effect as opposed to a selection effect of patients? And do you believe that the success in the aggressively treated patients was due to systemic tumor control or improved palliation within the brain?
Dr Moazami. The issue of selection bias was a concern. We tried to address it as best we could by using the statistical method of propensity matching. This allowed us to assign a propensity score on the basis of the 10 variables and match the patients. It was our best statistical tool to determine from this retrospective data analysis whether the treatment was effective. Prospective, randomized studies are another way to address such questions.
The data suggest that the patients who were aggressively treated had control both locoregionally and in the brain. Most failures were at distant sites.
|
References |
|---|
|
TopAbstractIntroductionPatients and methodsResultsDiscussionAppendix AAppendix BAppendix: DiscussionReferences |
|---|
This article has been cited by other articles:
|
|
 |
|
  A. Modi, H. A. Vohra, and D. F. Weeden Does surgery for primary non-small cell lung cancer and cerebral metastasis have any impact on survival? Interactive CardioVascular and Thoracic Surgery, April 1, 2009; 8(4): 467 - 473. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 T. M. De Pas, F. de Braud, G. Catalano, C. Putzu, G. Veronesi, F. Leo, P. G. Solli, D. Brambilla, G. Paganelli, and L. Spaggiari Oligometastatic Non-Small Cell Lung Cancer: A Multidisciplinary Approach in the Positron Emission Tomographic Scan Era Ann. Thorac. Surg., January 1, 2007; 83(1): 231 - 234. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. Furak, I. Trojan, T. Szoke, L. Agocs, A. Csekeo, J. Kas, E. Svastics, J. Eller, and L. Tiszlavicz Lung Cancer and Its Operable Brain Metastasis: Survival Rate and Staging Problems Ann. Thorac. Surg., January 1, 2005; 79(1): 241 - 247. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 D. G. Pfister, D. H. Johnson, C. G. Azzoli, W. Sause, T. J. Smith, S. Baker Jr, J. Olak, D. Stover, J. R. Strawn, A. T. Turrisi, et al. American Society of Clinical Oncology Treatment of Unresectable Non-Small-Cell Lung Cancer Guideline: Update 2003 J. Clin. Oncol., January 15, 2004; 22(2): 330 - 353. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| ANN THORAC SURG | ASIAN CARDIOVASC THORAC ANN | EUR J CARDIOTHORAC SURG |
| J THORAC CARDIOVASC SURG | ICVTS | ALL CTSNet JOURNALS |
