- Email: [email protected]
Risk Factors for Recurrence After Lung Cancer Resection as Estimated Using the Survival Tree Method
CHEST
Original Research LUNG CANCER
Risk Factors for Recurrence After Lung Cancer Resection as Estimated Using the Survival Tree Method Shigeki Sawada, MD, PhD; Natsumi Yamashita, MD, PhD; Hiroshi Suehisa, MD, PhD; and Motohiro Yamashita, MD, PhD
Background: Patients with lung cancer often present with recurrence, even after resection. The identification of risk factors for recurrence after resection is useful. Methods: Among 1,338 patients with lung cancer who underwent a complete resection, 277 developed recurrences post surgery. Data regarding the TNM factors, histologic subtype, and presence/absence of vessel invasion were analyzed retrospectively using the survival tree method to identify groups with a high risk of recurrence after resection. Results: The results revealed that the T factor, the N factor, and lymphatic (ly) and blood (v) vessel invasion were related to the risk of recurrence, and six combinations of these factors were identified using the survival tree method: group A: v 5 0, T ⱕ 1b, ly 5 0; group B: v 5 0, T ⱕ 1b, ly ⱖ 1; group C: v 5 0, T ⱖ 2a; group D: v ⱖ 1, N ⱕ 1, T ⱕ 2b; group E: v ⱖ 1, N ⱕ 1, T ⱖ 3; and group F: v ⱖ 1, N ⱖ 2. The six groups were then further classified into three groups: a low-risk group (group A), a moderate-risk group (groups B, C, and D), and a high-risk group (groups E and F). The 5-year recurrence-free survival rate was approximately 98% for the low-risk group, 75% for the moderate-risk group, and 30% for the high-risk group. Conclusions: Combining the T, N, v, and ly factors allowed the precise identification of a group with a high risk of recurrence after resection. CHEST 2013; 144(4):1238–1244 Abbreviations: G 5 tumor differentiation grade; ly 5 lymphatic vessel invasion; NSCLC 5 non-small cell lung cancer; p 5 pathologic; RFS 5 recurrence-free survival; v 5 blood vessel invasion
staging system for non-small cell lung canThecerTNM (NSCLC) is an internationally accepted system
used to determine the disease stage.1-3 This staging system is a measure of the extent of disease and is used to guide management and predict patient prognosis.4 Surgical resection is performed in patients with stage I and stage II disease as well as in some patients with
Manuscript received December 19, 2012; revision accepted April 19, 2013. Affiliations: From the Department of Thoracic Surgery (Drs Sawada, Suehisa, and M. Yamashita), and the Division of Clinical Biostatistics (Dr N. Yamashita), National Hospital Organization Shikoku Cancer Center, Matsuyama, Japan. Funding/Support: The authors have reported to CHEST that no funding was received for this study. Correspondence to: Shigeki Sawada, MD, PhD, Department of Thoracic Surgery, Shikoku Cancer Center, 160 Minamiumemotocho Kou, Matsuyama-shi, Ehime, 791-0280, Japan; e-mail: ssawada@ shikoku-cc.go.jp © 2013 American College of Chest Physicians. Reproduction of this article is prohibited without written permission from the American College of Chest Physicians. See online for more details. DOI: 10.1378/chest.12-3034
stage IIIA disease. Examinations, including physical examinations, chest radiography, and so forth, are performed periodically after resection, in accordance with a generally recognized postoperative follow-up or surveillance protocol. The purpose of postoperative follow-up is to detect any recurrences early after resection, so that adequate treatment can be offered in an attempt to improve the survival duration and quality of life. For doctors in charge of such postoperative follow-up, it would be useful to have an understanding of the factors associated with a high risk of recurrence and of the interval until the development of recurrence after resection. The TNM stage is well known to be associated with prognosis. Other factors, such as the presence/absence of vessel invasion and the tumor differentiation grade (G), have also been reported to be associated with prognosis.5-7 In this study, to identify a population with a high risk of recurrence after resection for NSCLC, we evaluated the T factor, N factor, extent of vessel invasion, and so on as part of
1238
Downloaded From: http://journal.publications.chestnet.org/ by David Kinnison on 10/03/2013
Original Research
a risk estimate for recurrence using the survival tree method. Materials and Methods This retrospective study based on a review of medical records was conducted with the approval of our hospital’s internal review board (SCC2012-225). From 1998 to 2009, 1,338 patients with NSCLC underwent complete resection (R0) with a segmentectomy or more extended pulmonary resection, followed by systemic mediastinal lymph node dissection; all the patients were followed up at our institution after resection. The seventh edition of the TNM staging system was applied in this study, and all 1,338 patients were restaged according to the seventh edition.8 Information about the 1,338 patients, including the tumor histology, pathologic (p) stage, history of adjuvant chemotherapy, and the survival and recurrence data were carefully reviewed. Regarding the histology, we collected information not only on the histologic subtypes but also on the extent of lymphatic vessel invasion (ly), the extent of blood vessel invasion (v), and the G. The specimens were stained with hematoxylin-eosin and elastica van Gieson and were examined under a 3 200 magnification. The ly factor was semiquantitatively scored into four grades in accordance with the following criteria: ly0, no lymphatic vessel involvement; ly1, involvement of one lymphatic site per low-power field; ly2, involvement of two or three lymphatic sites per lowpower field; and ly3, involvement of four or more lymphatic sites per low-power field. The v factor was similarly scored. The G factor was classified into four grades of severity according to the guidelines of the American Joint Commission on Cancer, as follows: G1, well differentiated; G2, moderately differentiated; G3, poorly differentiated; G4, undifferentiated.9 Survival data and recurrence data were obtained for all 1,338 patients in June 2011. Kaplan-Meier curves for recurrencefree survival (RFS) and the estimated hazard ratios were constructed according to the TNM stage. A definition of RFS in this study was a count of recurrence only as an event. Deaths were regarded as censored cases when recurrence was not known to have occurred. To identify a population with a high risk of recurrence after resection, the survival tree method was used in this study. The algorithm for the survival tree method is as follows.10 The survival tree model involves two components: (1) the examination of all allowable splits on each predictor variable and (2) the selection and execution (to create left and right daughter nodes) of the optimal splits. At each step of the analysis, all the possible splits into disjoint subsets are examined by a computer algorithm based on the log-rank test. The best split is then selected so that the split produces two subgroups of patients with the largest difference in Kaplan-Meier survival curves. The same procedure is applied recursively to increase the number of splits until each group contains only a few subjects. In this study, T factor, N factor, ly factor, v factor, G factor, histologic subtype, and history of adjuvant chemotherapy were evaluated as variables using the survival tree method according to the frequency of recurrence after resection, and the patients were classified into subgroups. Kaplan-Meier curves for RFS were constructed according to the subgroups identified by the survival tree method. The estimated hazard ratios for recurrence classified according to these subgroups were also calculated. The survival tree analysis was performed using the rpart package in the R statistical software, and the hazard ratios for recurrence were estimated using the muhaz package in the R statistical software (http://www.r-project.org/). Other statistical analyses, including the Kaplan-Meier method, were performed using Stata, Ver. 11.2 (StataCorp LP). P values , .05 were considered to denote statistical significance.
Preoperative Workup and Postoperative Follow-up Protocol All the patients who were considered to be candidates for surgical resection underwent a CT scan examination of the thorax and upper abdomen, abdominal ultrasonography, bone scintigraphy, and enhanced MRI of the brain. PET/CT scan was initiated in April 2006 at our institute. All the patients underwent PET/CT scan and an enhanced MRI examination of the brain as part of the preoperative workup thereafter. During the first 2 years after surgery, a follow-up examination, including history taking and a physical examination, was performed every 3 to 4 months. During the next 3 years, the same follow-up procedures were repeated every 5 to 6 months. A CT scan examination of the thorax and upper abdomen was performed 12 months after the operation and was repeated annually. Chest radiography was performed 3 months and 6 months after the operation and was then repeated once a year between the CT scan examinations. The follow-up procedures, including the imaging examinations, were reduced to a minimum after the passage of 5 years after the operation but were nonetheless continued for up to 7 to 10 years. It was important to distinguish metachronous lung cancer from recurrence when a new pulmonary malignancy was detected after resection. Lou et al11 reported their protocol for distinguishing recurrence and metachronous lung cancer after resection. Our process was similar to the method of Lou et al.11 When a new pulmonary malignancy was detected during the follow-up surveillance, restaging was performed just as was done during the preoperative workup, which included CT scan, PET/CT scan, MRI, ultrasonography, and bone scintigraphy. A diagnosis of recurrence was based on the image findings in most cases. In some cases in which it was difficult to discriminate between recurrence and metachronous lung cancer, histologic confirmation was performed, and a diagnosis of metachronous lung cancer was decided according to the criteria of Martini and Melamed12 as follows: (1) the histologic results differed from those of the index tumor; (2) the same histologic results as the index tumor were obtained, but diagnosis was made 2 years after the primary tumor; or (3) the same histologic results as the index tumor were obtained, the diagnosis was made within 2 years of the primary tumor, but the tumors were located in different lobes or segments, with no positive intervening lymph nodes and no evidence of metastasis. When a new pulmonary malignancy fulfilled any one of these criteria, it was considered to be metachronous lung cancer. In addition, a new pulmonary malignancy was diagnosed as metachronous lung cancer when it was accompanied by a ground-glass opacity on high-resolution CT scan, regardless of its location and the timing of its appearance. A new pulmonary malignancy that met these criteria was considered to be metachronous lung cancer and was excluded from this study.
Results The characteristics of the 1,338 patients are listed in Table 1. The median age was 66 years (range, 27-88 years), and 742 patients were men. Histologic examination revealed adenocarcinoma in 1,046 patients, squamous cell carcinoma in 226 patients, and others in 66 patients; the p stage was p-stage IA in 663 patients, IB in 305, IIA in 119, IIB in 73, IIIA in 140, and IIIB in four patients. No ly (ly0) was observed in 851 patients (65.7%), whereas 287 patients (22.2%) were classified as ly1, 134 patients (10.3%) were classified as ly2, and 22 patients (1.7%) were classified as ly3. Similarly, no v (v0) was observed in 731 patients (56.6%),
journal.publications.chestnet.org
Downloaded From: http://journal.publications.chestnet.org/ by David Kinnison on 10/03/2013
CHEST / 144 / 4 / OCTOBER 2013
1239
Table 1—Patient Characteristics Characteristic Age, median (range), y Sex Male Female Histologic subtype Adenocarcinoma Squamous cell carcinoma Other p Stage IA IB IIA IIB IIIA IIIB pT 1a 1b 2a 2b 3 4 Unknown pN 0 1 2 3 Unknown ly 0 1 2 3 Unknown v 0 1 2 3 Unknown G Well differentiated (G1) Moderately differentiated (G2) Poorly differentiated (G3) Undifferentiated (G4) Unknown Adjuvant chemotherapy No Yes Recurrence No Yes Dead or alive Alive Dead
Value 66 (27-88) 742 (55.5) 596 (44.5) 1,046 (78.2) 226 (16.9) 66 (4.9) 663 (50.8) 305 (23.4) 119 (9.1) 73 (5.6) 140 (10.7) 4 (0.3) 497 (38.1) 243 (18.7) 411 (31.5) 42 (3.2) 105 (8.1) 5 (0.4) 35 (2.6) 1,087 (81.2) 126 (9.5) 119 (8.9) 1 (0.1) 5 (3.7) 851 (65.7) 287 (22.2) 134 (10.3) 22 (1.7) 44 (3.3) 731 (56.6) 334 (25.9) 188 (14.6) 37 (2.9) 48 (3.6) 440 (35.0) 521 (41.4) 280 (22.2) 18 (1.4) 79 (5.9)
the G factor, 440 patients (35.0%) were classified as G1, 521 patients (41.4%) were classified as G2, 280 patients (22.2%) were classified as G3, and 18 patients (1.4%) were classified as G4. Preoperative chemoradiotherapy was performed in 29 patients, and postoperative adjuvant chemotherapy was performed in 327 patients. The median follow-up period was 54.2 months (range, 1-159 months). Recurrence was observed in 277 patients (26.5%) during the study period, and 296 patients (22.1%) had died as of June 2011. The 1-, 3-, and 5-year RFS for the all patients were 77.1%, 22.6%, and 7.9%, respectively, and most recurrences occurred within 5 years of resection. The RFS curves classified according to the TNM stage are shown in Figure 1. The 5-year RFS rates for stage IA, IB, IIA, IIB, and IIIA were 90.8%, 68.9%, 54.3%, 43.5%, and 22.0%, respectively, and the RFS was significantly different among the TNM stage (P , .001). The estimated hazard ratios for recurrence classified according to the TNM stage are shown in Figure 2. The estimated hazard ratios for recurrence increased and reached a peak at between 6 months and 2 years after resection, gradually decreasing thereafter. The hazard ratios for recurrence in stage IIB-IIIB disease seemed to reach their peak earlier than those in stage IA-IIA disease. The RFS was analyzed according to the grade of the ly factor, and the results indicated that the grade of the ly factor was significantly related to the RFS (P , .001). The RFS of patients with ly0 was better than those with ly1, ly2, and ly3, and the RFS became worse as the grade of the ly advanced. Similarly, the RFS was analyzed according to the grade of the v factor and the G factor, and the results indicated that the RFS also became worse as the grade of the v factor and the G factor advanced. Regarding the histologic subtype, patients with adenocarcinoma had a better RFS than the patients with other histologic subtypes (P , .001). To identify a population with a high risk of recurrence after resection, the T factor, N factor, ly factor,
1,011 (75.6) 327 (24.4) 984 (73.5) 277 (26.5) 1,042 (77.9) 296 (22.1)
Data are presented as No. (%) unless otherwise noted. G 5 tumor differentiation grade; ly 5 lymphatic vessel invasion; p 5 pathologic; v 5 blood vessel invasion.
whereas 334 patients (25.9%) were classified as v1, 188 patients (14.6%) were classified as v2, and 37 patients (2.9%) were classified as v3. Regarding
Figure 1. Recurrence-free survival curve classified according to the TNM stage.
1240
Downloaded From: http://journal.publications.chestnet.org/ by David Kinnison on 10/03/2013
Original Research
Figure 2. Estimated hazard rates for recurrence classified according to the TNM stage.
v factor, G factor, histologic subtype, and history of adjuvant chemotherapy were analyzed using the survival tree method. The T, N, v, and ly factors were selected as variables in the split steps, but the G factor, histologic subtype, and history of adjuvant chemotherapy were not selected in this study (Fig 3). The first split was produced by the v factor, and the patients were separated according to v 5 0 vs v ⱖ 1. The second split of the v 5 0 branch was produced by the T factor, and the patients were separated according to T ⱕ 1b vs T ⱖ 2a. The V 5 0, T ⱕ b group was separated with ly 5 0 vs ly ⱖ 1 at the third sprit. Similarly, in the v ⱖ 1 branch, the second split was produced by the N factor, and the patients were separated according to N ⱕ 1 vs N ⱖ 2. The third split of the v ⱖ 1, N ⱕ 1 group was produced by the T factor, and the patients were separated according to T ⱕ 2b vs T ⱖ 3. Finally, the 1,338 patients were classified into six groups; group A: v 5 0, T ⱕ 1b, ly 5 0; group B: v 5 0, T ⱕ 1b, ly ⱖ 1; group C: v 5 0, T ⱖ 2a; group D: v ⱖ 1, N ⱕ 1, T ⱕ 2b; group E: v ⱖ 1, N ⱕ 1, T ⱖ 3; and group F: v ⱖ 1, N ⱖ 2. The Kaplan-Meier curves for the RFS classified according to these six groups are shown in Figure 4.
Figure 3. Survival tree. ly 5 lymphatic vessel invasion; p 5 pathologic; Pt. No. 5 patient number; v 5 blood vessel invasion.
The curves for groups B, C, and D were identical; similarly, those for groups E and F were also almost identical. The six groups were further classified into three risk groups, as follows: a low-risk group (group A), a moderate-risk group (groups B, C, and D), and a high-risk group (groups E and F) (Figs 3, 4). The 5-year RFS rates were approximately 98% for the low-risk group, 75% for the moderate-risk group, and 30% for the high-risk group. The estimated hazard ratios for recurrence classified according to these three groups are shown in Figure 5. The estimated hazard ratios for recurrence increased and reached their peak at between 1 and 2 years after resection, then gradually decreased thereafter. The hazard ratio for recurrence in the high-risk group seemed to reach its peak earlier than that in the low-risk group. Discussion The TNM staging system is based on the characteristics of the primary tumor, the status of metastasis to the lymph nodes, and the presence/absence of distant metastasis. It is used worldwide, because it is simple and is correlated with patient prognosis. In patients with lung cancer, other factors, such as ly, v, and G factors, and more recently, the presence/absence of epidermal growth factor receptor mutation and the expressions of excision repair cross-complementation group 1 and ribonucleotide reductase M1 polypeptide, have been reported to be correlated with prognosis.5-7,13-16 However, testing for epidermal growth factor receptor mutation and the expressions of excision repair cross-complementation group 1and ribonucleotide reductase M1 polypeptide are not routinely performed in clinical practice. In this study, we focused on the T, N, ly, v, and G factors, all of which are easily evaluable in typical clinical practice, in an attempt to identify the risk factors for recurrence after resection in patients with lung cancer. The results of this study suggest that the combining of T, N, ly, and v factors in the risk estimation allowed the precise identification of a population with a high risk of recurrence.
Figure 4. Recurrence-free survival curves classified according to the six groups identified using the survival tree method.
journal.publications.chestnet.org
Downloaded From: http://journal.publications.chestnet.org/ by David Kinnison on 10/03/2013
CHEST / 144 / 4 / OCTOBER 2013
1241
Figure 5. Estimated hazard rates for recurrence classified according to the low-, moderate-, and high-risk groups.
A precise identification of the high-risk population for recurrence after resection might contribute to optimizing the postoperative follow-up surveillance. For instance, postoperative follow-up examinations might be unnecessary in the low-risk group (group A: v 5 0, T ⱕ 1b, ly 5 0), since recurrence is expected to be rare in this group. The Cox proportional hazards analysis is commonly used for multivariate analyses in many medical studies. The primary assumption in a Cox proportional hazards analysis is that the hazard functions of all data groups are proportional to one another. The survival tree model makes no assumption of proportional hazards; through a precise and complex program, it segregates groups by sequential partitioning based on statistical parameters related to the number of events (recurrences, deaths), allowing the progressive segregation of smaller groups with statistically significantly differences.17 The survival tree model, therefore, can identify meaningful prognostic subsets in a study population, which usually do not emerge from a routine proportional hazards analysis.10,18,19 At each step of the analysis in the survival tree model, all the possible splits into disjoint subsets are examined by a computer algorithm based on the log-rank test. The best split is then selected so that the split produces two subgroups of patients with the greatest differences in their Kaplan-Meier survival curves. Therefore, the variable that is used for the first split is considered to be the most significant factor. The first separation was created based on the presence/absence of v (v 5 0 vs v ⱖ 1) in this study. This outcome suggested that the presence/absence of v might be the most critical predictive factor of recurrence after resection. Interestingly, the variable in the third step, which separated group A and group B, was the presence/absence of ly (ly 5 0 vs ly ⱖ 1). Although the grades of the v factor and the ly factor were significantly related to the RFS, the presence/absence of vessel invasion might have
more influence than the grade for the prediction of recurrence. Another factor considered to be associated with recurrence might be visceral pleural invasion. Several investigators have reported that visceral pleural invasion is related to survival,20,21 but this factor was not evaluated in this study. Including the visceral pleural invasion factor among the ly and v factors study might have led to different results. Preoperative chemoradiotherapy was performed in 29 patients, and postoperative chemotherapy was performed in 327 patients; these patients were analyzed together in this study. Especially in the cases with preoperative chemoradiotherapy, the ly factor and v factor might have been influenced by the treatment. However, only 29 patients underwent preoperative chemoradiotherapy, and this number was so small that the effect of preoperative chemoradiotherapy was unlikely to have had a significant influence on the results. The presence/absence of adjuvant chemotherapy was included and analyzed using the survival tree method to evaluate the influence on recurrence but was not selected as a variable. Ideally, the patients with preoperative chemoradiotherapy and adjuvant chemotherapy should have been analyzed separately, or a subgroup analysis might have been needed. Metachronous lung cancer was diagnosed according to the criteria described in the Materials and Methods section, and patients with metachronous lung cancer were excluded from the present study. Of the 1,338 patients who were evaluated in this study, 53 patients (4.0%) developed metachronous lung cancer during the study period. The time intervals from the resection of the first lung cancer until the development of the second lung cancer ranged from 8 to 99 months. The hazard rate for the development the metachronous lung cancer over time is shown in Figure 6. The hazard rate increased gradually and reached a plateau at approximately 2 years, then decreased approximately 8 years after the first lung cancer. The possible reasons for the reduction in hazard rate might be related to patient age and
Figure 6. Estimated hazard rates for the development of metachronous lung cancer.
1242
Downloaded From: http://journal.publications.chestnet.org/ by David Kinnison on 10/03/2013
Original Research
the follow-up period of the study. The median age of the 53 patients with metachronous lung cancer was 69 years, and the patients would have been 77 years old after the passage of 8 years. Such an age might be too old for the development of second lung cancer. The longest follow-up period in this study was approximately 12 years. The number of patients who developed metachronous lung cancer would have likely increased if the follow-up period had been longer. The hazard rate for recurrence increased after resection and reached a peak between 1 and 2 years, gradually decreasing thereafter, as shown in Figures 1 and 5. In theory, the shape of the recurrence risk curve should be a single-peaked smooth curve; however, our results revealed curves with multiple peaks. One reason for this result might have been the study size. There were only 277 patients with recurrence in this study, which might not be a sufficiently large sample size to allow a smooth single-peaked curve to be obtained. However, Demicheli et al22 analyzed the recurrence dynamics after resection in a larger study and reported a similar multipeak recurrence risk curve. They explained this result using the tumor dormancy theory, which proposes that disseminated tumor cells interact with the microenvironment and show growth arrest if the environment is not permissive; then, once the environment becomes favorable again, the cells start to show uncontrolled proliferation and form metastases.23-26 This theory has been previously discussed in detail in relation to breast cancer, but it is still not clearly understood.27,28 Another possibility explaining the multipeak pattern might be related to our postoperative follow-up surveillance. As described in the Materials and Methods section, chest CT scan examinations were routinely repeated every year after resection in all patients. This procedure possibly detected asymptomatic recurrences, explaining the small peaks in the recurrence risk curves observed at around 2 and 3 years after resection. There is insufficient evidence, however, to arrive at a definitive conclusion, and as far as this study was concerned, we considered the multipeak pattern of the recurrence risk curve to be attributable to our postoperative follow-up surveillance protocol. Conclusions Risk factors for recurrence after resection were analyzed using the survival tree model in patients who had undergone a complete resection for NSCLC. The patients were classified into three groups (low, moderate, and high risk) according to the risk of recurrence by the combining the T, N, ly, and v factors. The results suggested that combining the T, N, ly, and v factors might enable an accurate estimation of the risk of recurrence after resection.
Acknowledgments Author contributions: Dr Sawada takes responsibility for the integrity of the data and the accuracy of the data analysis. Dr Sawada: contributed to the design, overseeing, and conduct of the study and to the writing and editing of the manuscript. Dr N. Yamashita: had full access to the data and contributed to the statistical analysis and drafting of the manuscript. Dr Suehisa: contributed to the study conception and design, acquisition of data, analysis and interpretation of data, drafting of the manuscript, and revision of the manuscript. Dr M. Yamashita: contributed to the study conception and design, acquisition of data, analysis and interpretation of data, drafting of the manuscript, and revision of the manuscript. Financial/nonfinancial disclosures: The authors have reported to CHEST that no potential conflicts of interest exist with any companies/organizations whose products or services may be discussed in this article. Other contributions: We thank Norihiro Teramoto, MD, PhD; Rieko Nishimura, MD, PhD; and Hiroyuki Takahata, MD, PhD of the Department of Pathology and Laboratory Medicine for their contribution to the pathological examinations. Without their invaluable assistance, this study would not have been possible.
References 1. Asamura H, Goya T, Koshiishi Y, et al; Japanese Joint Committee of Lung Cancer Registry. A Japanese Lung Cancer Registry study: prognosis of 13,010 resected lung cancers. J Thorac Oncol. 2008;3(1):46-52. 2. Goldstraw P, Crowley J, Chansky K, et al; International Association for the Study of Lung Cancer International Staging Committee; Participating Institutions. The IASLC Lung Cancer Staging Project: proposals for the revision of the TNM stage groupings in the forthcoming (seventh) edition of the TNM Classification of malignant tumours. J Thorac Oncol. 2007;2(8):706-714. 3. Groome PA, Bolejack V, Crowley JJ, et al; IASLC International Staging Committee; Cancer Research and Biostatistics; Observers to the Committee; Participating Institutions. The IASLC Lung Cancer Staging Project: validation of the proposals for revision of the T, N, and M descriptors and consequent stage groupings in the forthcoming (seventh) edition of the TNM classification of malignant tumours. J Thorac Oncol. 2007;2(8):694-705. 4. Silvestri GA, Tanoue LT, Margolis ML, Barker J, Detterbeck F; American College of Chest Physicians. The noninvasive staging of non-small cell lung cancer: the guidelines. Chest. 2003;123(suppl 1):147S-156S. 5. Macchiarini P, Fontanini G, Hardin MJ, et al. Blood vessel invasion by tumor cells predicts recurrence in completely resected T1 N0 M0 non-small-cell lung cancer. J Thorac Cardiovasc Surg. 1993;106(1):80-89. 6. Ito M, Moriya T, Ishida T, et al. Significance of pathological evaluation for lymphatic vessel invasion in invasive breast cancer. Breast Cancer. 2007;14(4):381-387. 7. Hashizume S, Nagayasu T, Hayashi T, et al. Accuracy and prognostic impact of a vessel invasion grading system for stage IA non-small cell lung cancer. Lung Cancer. 2009;65(3): 363-370. 8. Sobin LH, Gospodarowicz MK, Wittekind C. International Union Against Cancer. TNM Classification of Malignant Tumours. 7th ed. Chichester, West Sussex, England: WileyBlackwell; 2010. 9. Cancer AJCo. AJCC Cancer Staging Manual. 6th ed. New York, NY: Springer; 2002. 10. Biegon A, Hanau M, Greenberger V, Segal M. Aging and brain cholinergic muscarinic receptor subtypes: an autoradiographic study in the rat. Neurobiol Aging. 1989;10(4):305-310.
journal.publications.chestnet.org
Downloaded From: http://journal.publications.chestnet.org/ by David Kinnison on 10/03/2013
CHEST / 144 / 4 / OCTOBER 2013
1243
11. Lou F, Huang J, Sima CS, Dycoco J, Rusch V, Bach PB. Patterns of recurrence and second primary lung cancer in early-stage lung cancer survivors followed with routine computed tomography surveillance. J Thorac Cardiovasc Surg. 2013;145(1): 75-81. 12. Martini N, Melamed MR. Multiple primary lung cancers. J Thorac Cardiovasc Surg. 1975;70(4):606-612. 13. Hirsch FR, Varella-Garcia M, Bunn PA Jr, et al. Epidermal growth factor receptor in non-small-cell lung carcinomas: correlation between gene copy number and protein expression and impact on prognosis. J Clin Oncol. 2003;21(20):3798-3807. 14. Selvaggi G, Novello S, Torri V, et al. Epidermal growth factor receptor overexpression correlates with a poor prognosis in completely resected non-small-cell lung cancer. Ann Oncol. 2004;15(1):28-32. 15. Zheng Z, Chen T, Li X, Haura E, Sharma A, Bepler G. DNA synthesis and repair genes RRM1 and ERCC1 in lung cancer. N Engl J Med. 2007;356(8):800-808. 16. Lipford EH III, Eggleston JC, Lillemoe KD, Sears DL, Moore GW, Baker RR. Prognostic factors in surgically resected limited-stage, nonsmall cell carcinoma of the lung. Am J Surg Pathol. 1984;8(5):357-365. 17. Sevin BU, Lu Y, Bloch DA, Nadji M, Koechli OR, Averette HE. Surgically defined prognostic parameters in patients with early cervical carcinoma. A multivariate survival tree analysis. Cancer. 1996;78(7):1438-1446. 18. Altman RD, Medsger TA Jr, Bloch DA, Michel BA. Predictors of survival in systemic sclerosis (scleroderma). Arthritis Rheum. 1991;34(4):403-413. 19. Piette JD, Intrator O, Zierler S, Mor V, Stein MD. An exploratory analysis of survival with AIDS using a nonparametric tree-structured approach. Epidemiology. 1992;3(4):310-318.
20. Kudo Y, Saji H, Shimada Y, et al. Impact of visceral pleural invasion on the survival of patients with non-small cell lung cancer. Lung Cancer. 2012;78(2):153-160. 21. Hung JJ, Jeng WJ, Hsu WH, Chou TY, Lin SF, Wu YC. Prognostic significance of the extent of visceral pleural invasion in completely resected node-negative non-small cell lung cancer. Chest. 2012;142(1):141-150. 22. Demicheli R, Fornili M, Ambrogi F, et al. Recurrence dynamics for non-small-cell lung cancer: effect of surgery on the development of metastases. J Thorac Oncol. 2012;7(4): 723-730. 23. Páez D, Labonte MJ, Bohanes P, et al. Cancer dormancy: a model of early dissemination and late cancer recurrence. Clin Cancer Res. 2012;18(3):645-653. 24. Aguirre-Ghiso JA. Models, mechanisms and clinical evidence for cancer dormancy. Nat Rev Cancer. 2007;7(11): 834-846. 25. Braun S, Vogl FD, Naume B, et al. A pooled analysis of bone marrow micrometastasis in breast cancer. N Engl J Med. 2005;353(8):793-802. 26. Ranganathan AC, Adam AP, Aguirre-Ghiso JA. Opposing roles of mitogenic and stress signaling pathways in the induction of cancer dormancy. Cell Cycle. 2006;5(16):1799-1807. 27. Demicheli R, Miceli R, Moliterni A, et al. Breast cancer recurrence dynamics following adjuvant CMF is consistent with tumor dormancy and mastectomy-driven acceleration of the metastatic process. Ann Oncol. 2005;16(9):1449-1457. 28. Demicheli R, Abbattista A, Miceli R, Valagussa P, Bonadonna G. Time distribution of the recurrence risk for breast cancer patients undergoing mastectomy: further support about the concept of tumor dormancy. Breast Cancer Res Treat. 1996; 41(2):177-185.
1244
Downloaded From: http://journal.publications.chestnet.org/ by David Kinnison on 10/03/2013
Original Research
Original Research LUNG CANCER
Risk Factors for Recurrence After Lung Cancer Resection as Estimated Using the Survival Tree Method Shigeki Sawada, MD, PhD; Natsumi Yamashita, MD, PhD; Hiroshi Suehisa, MD, PhD; and Motohiro Yamashita, MD, PhD
Background: Patients with lung cancer often present with recurrence, even after resection. The identification of risk factors for recurrence after resection is useful. Methods: Among 1,338 patients with lung cancer who underwent a complete resection, 277 developed recurrences post surgery. Data regarding the TNM factors, histologic subtype, and presence/absence of vessel invasion were analyzed retrospectively using the survival tree method to identify groups with a high risk of recurrence after resection. Results: The results revealed that the T factor, the N factor, and lymphatic (ly) and blood (v) vessel invasion were related to the risk of recurrence, and six combinations of these factors were identified using the survival tree method: group A: v 5 0, T ⱕ 1b, ly 5 0; group B: v 5 0, T ⱕ 1b, ly ⱖ 1; group C: v 5 0, T ⱖ 2a; group D: v ⱖ 1, N ⱕ 1, T ⱕ 2b; group E: v ⱖ 1, N ⱕ 1, T ⱖ 3; and group F: v ⱖ 1, N ⱖ 2. The six groups were then further classified into three groups: a low-risk group (group A), a moderate-risk group (groups B, C, and D), and a high-risk group (groups E and F). The 5-year recurrence-free survival rate was approximately 98% for the low-risk group, 75% for the moderate-risk group, and 30% for the high-risk group. Conclusions: Combining the T, N, v, and ly factors allowed the precise identification of a group with a high risk of recurrence after resection. CHEST 2013; 144(4):1238–1244 Abbreviations: G 5 tumor differentiation grade; ly 5 lymphatic vessel invasion; NSCLC 5 non-small cell lung cancer; p 5 pathologic; RFS 5 recurrence-free survival; v 5 blood vessel invasion
staging system for non-small cell lung canThecerTNM (NSCLC) is an internationally accepted system
used to determine the disease stage.1-3 This staging system is a measure of the extent of disease and is used to guide management and predict patient prognosis.4 Surgical resection is performed in patients with stage I and stage II disease as well as in some patients with
Manuscript received December 19, 2012; revision accepted April 19, 2013. Affiliations: From the Department of Thoracic Surgery (Drs Sawada, Suehisa, and M. Yamashita), and the Division of Clinical Biostatistics (Dr N. Yamashita), National Hospital Organization Shikoku Cancer Center, Matsuyama, Japan. Funding/Support: The authors have reported to CHEST that no funding was received for this study. Correspondence to: Shigeki Sawada, MD, PhD, Department of Thoracic Surgery, Shikoku Cancer Center, 160 Minamiumemotocho Kou, Matsuyama-shi, Ehime, 791-0280, Japan; e-mail: ssawada@ shikoku-cc.go.jp © 2013 American College of Chest Physicians. Reproduction of this article is prohibited without written permission from the American College of Chest Physicians. See online for more details. DOI: 10.1378/chest.12-3034
stage IIIA disease. Examinations, including physical examinations, chest radiography, and so forth, are performed periodically after resection, in accordance with a generally recognized postoperative follow-up or surveillance protocol. The purpose of postoperative follow-up is to detect any recurrences early after resection, so that adequate treatment can be offered in an attempt to improve the survival duration and quality of life. For doctors in charge of such postoperative follow-up, it would be useful to have an understanding of the factors associated with a high risk of recurrence and of the interval until the development of recurrence after resection. The TNM stage is well known to be associated with prognosis. Other factors, such as the presence/absence of vessel invasion and the tumor differentiation grade (G), have also been reported to be associated with prognosis.5-7 In this study, to identify a population with a high risk of recurrence after resection for NSCLC, we evaluated the T factor, N factor, extent of vessel invasion, and so on as part of
1238
Downloaded From: http://journal.publications.chestnet.org/ by David Kinnison on 10/03/2013
Original Research
a risk estimate for recurrence using the survival tree method. Materials and Methods This retrospective study based on a review of medical records was conducted with the approval of our hospital’s internal review board (SCC2012-225). From 1998 to 2009, 1,338 patients with NSCLC underwent complete resection (R0) with a segmentectomy or more extended pulmonary resection, followed by systemic mediastinal lymph node dissection; all the patients were followed up at our institution after resection. The seventh edition of the TNM staging system was applied in this study, and all 1,338 patients were restaged according to the seventh edition.8 Information about the 1,338 patients, including the tumor histology, pathologic (p) stage, history of adjuvant chemotherapy, and the survival and recurrence data were carefully reviewed. Regarding the histology, we collected information not only on the histologic subtypes but also on the extent of lymphatic vessel invasion (ly), the extent of blood vessel invasion (v), and the G. The specimens were stained with hematoxylin-eosin and elastica van Gieson and were examined under a 3 200 magnification. The ly factor was semiquantitatively scored into four grades in accordance with the following criteria: ly0, no lymphatic vessel involvement; ly1, involvement of one lymphatic site per low-power field; ly2, involvement of two or three lymphatic sites per lowpower field; and ly3, involvement of four or more lymphatic sites per low-power field. The v factor was similarly scored. The G factor was classified into four grades of severity according to the guidelines of the American Joint Commission on Cancer, as follows: G1, well differentiated; G2, moderately differentiated; G3, poorly differentiated; G4, undifferentiated.9 Survival data and recurrence data were obtained for all 1,338 patients in June 2011. Kaplan-Meier curves for recurrencefree survival (RFS) and the estimated hazard ratios were constructed according to the TNM stage. A definition of RFS in this study was a count of recurrence only as an event. Deaths were regarded as censored cases when recurrence was not known to have occurred. To identify a population with a high risk of recurrence after resection, the survival tree method was used in this study. The algorithm for the survival tree method is as follows.10 The survival tree model involves two components: (1) the examination of all allowable splits on each predictor variable and (2) the selection and execution (to create left and right daughter nodes) of the optimal splits. At each step of the analysis, all the possible splits into disjoint subsets are examined by a computer algorithm based on the log-rank test. The best split is then selected so that the split produces two subgroups of patients with the largest difference in Kaplan-Meier survival curves. The same procedure is applied recursively to increase the number of splits until each group contains only a few subjects. In this study, T factor, N factor, ly factor, v factor, G factor, histologic subtype, and history of adjuvant chemotherapy were evaluated as variables using the survival tree method according to the frequency of recurrence after resection, and the patients were classified into subgroups. Kaplan-Meier curves for RFS were constructed according to the subgroups identified by the survival tree method. The estimated hazard ratios for recurrence classified according to these subgroups were also calculated. The survival tree analysis was performed using the rpart package in the R statistical software, and the hazard ratios for recurrence were estimated using the muhaz package in the R statistical software (http://www.r-project.org/). Other statistical analyses, including the Kaplan-Meier method, were performed using Stata, Ver. 11.2 (StataCorp LP). P values , .05 were considered to denote statistical significance.
Preoperative Workup and Postoperative Follow-up Protocol All the patients who were considered to be candidates for surgical resection underwent a CT scan examination of the thorax and upper abdomen, abdominal ultrasonography, bone scintigraphy, and enhanced MRI of the brain. PET/CT scan was initiated in April 2006 at our institute. All the patients underwent PET/CT scan and an enhanced MRI examination of the brain as part of the preoperative workup thereafter. During the first 2 years after surgery, a follow-up examination, including history taking and a physical examination, was performed every 3 to 4 months. During the next 3 years, the same follow-up procedures were repeated every 5 to 6 months. A CT scan examination of the thorax and upper abdomen was performed 12 months after the operation and was repeated annually. Chest radiography was performed 3 months and 6 months after the operation and was then repeated once a year between the CT scan examinations. The follow-up procedures, including the imaging examinations, were reduced to a minimum after the passage of 5 years after the operation but were nonetheless continued for up to 7 to 10 years. It was important to distinguish metachronous lung cancer from recurrence when a new pulmonary malignancy was detected after resection. Lou et al11 reported their protocol for distinguishing recurrence and metachronous lung cancer after resection. Our process was similar to the method of Lou et al.11 When a new pulmonary malignancy was detected during the follow-up surveillance, restaging was performed just as was done during the preoperative workup, which included CT scan, PET/CT scan, MRI, ultrasonography, and bone scintigraphy. A diagnosis of recurrence was based on the image findings in most cases. In some cases in which it was difficult to discriminate between recurrence and metachronous lung cancer, histologic confirmation was performed, and a diagnosis of metachronous lung cancer was decided according to the criteria of Martini and Melamed12 as follows: (1) the histologic results differed from those of the index tumor; (2) the same histologic results as the index tumor were obtained, but diagnosis was made 2 years after the primary tumor; or (3) the same histologic results as the index tumor were obtained, the diagnosis was made within 2 years of the primary tumor, but the tumors were located in different lobes or segments, with no positive intervening lymph nodes and no evidence of metastasis. When a new pulmonary malignancy fulfilled any one of these criteria, it was considered to be metachronous lung cancer. In addition, a new pulmonary malignancy was diagnosed as metachronous lung cancer when it was accompanied by a ground-glass opacity on high-resolution CT scan, regardless of its location and the timing of its appearance. A new pulmonary malignancy that met these criteria was considered to be metachronous lung cancer and was excluded from this study.
Results The characteristics of the 1,338 patients are listed in Table 1. The median age was 66 years (range, 27-88 years), and 742 patients were men. Histologic examination revealed adenocarcinoma in 1,046 patients, squamous cell carcinoma in 226 patients, and others in 66 patients; the p stage was p-stage IA in 663 patients, IB in 305, IIA in 119, IIB in 73, IIIA in 140, and IIIB in four patients. No ly (ly0) was observed in 851 patients (65.7%), whereas 287 patients (22.2%) were classified as ly1, 134 patients (10.3%) were classified as ly2, and 22 patients (1.7%) were classified as ly3. Similarly, no v (v0) was observed in 731 patients (56.6%),
journal.publications.chestnet.org
Downloaded From: http://journal.publications.chestnet.org/ by David Kinnison on 10/03/2013
CHEST / 144 / 4 / OCTOBER 2013
1239
Table 1—Patient Characteristics Characteristic Age, median (range), y Sex Male Female Histologic subtype Adenocarcinoma Squamous cell carcinoma Other p Stage IA IB IIA IIB IIIA IIIB pT 1a 1b 2a 2b 3 4 Unknown pN 0 1 2 3 Unknown ly 0 1 2 3 Unknown v 0 1 2 3 Unknown G Well differentiated (G1) Moderately differentiated (G2) Poorly differentiated (G3) Undifferentiated (G4) Unknown Adjuvant chemotherapy No Yes Recurrence No Yes Dead or alive Alive Dead
Value 66 (27-88) 742 (55.5) 596 (44.5) 1,046 (78.2) 226 (16.9) 66 (4.9) 663 (50.8) 305 (23.4) 119 (9.1) 73 (5.6) 140 (10.7) 4 (0.3) 497 (38.1) 243 (18.7) 411 (31.5) 42 (3.2) 105 (8.1) 5 (0.4) 35 (2.6) 1,087 (81.2) 126 (9.5) 119 (8.9) 1 (0.1) 5 (3.7) 851 (65.7) 287 (22.2) 134 (10.3) 22 (1.7) 44 (3.3) 731 (56.6) 334 (25.9) 188 (14.6) 37 (2.9) 48 (3.6) 440 (35.0) 521 (41.4) 280 (22.2) 18 (1.4) 79 (5.9)
the G factor, 440 patients (35.0%) were classified as G1, 521 patients (41.4%) were classified as G2, 280 patients (22.2%) were classified as G3, and 18 patients (1.4%) were classified as G4. Preoperative chemoradiotherapy was performed in 29 patients, and postoperative adjuvant chemotherapy was performed in 327 patients. The median follow-up period was 54.2 months (range, 1-159 months). Recurrence was observed in 277 patients (26.5%) during the study period, and 296 patients (22.1%) had died as of June 2011. The 1-, 3-, and 5-year RFS for the all patients were 77.1%, 22.6%, and 7.9%, respectively, and most recurrences occurred within 5 years of resection. The RFS curves classified according to the TNM stage are shown in Figure 1. The 5-year RFS rates for stage IA, IB, IIA, IIB, and IIIA were 90.8%, 68.9%, 54.3%, 43.5%, and 22.0%, respectively, and the RFS was significantly different among the TNM stage (P , .001). The estimated hazard ratios for recurrence classified according to the TNM stage are shown in Figure 2. The estimated hazard ratios for recurrence increased and reached a peak at between 6 months and 2 years after resection, gradually decreasing thereafter. The hazard ratios for recurrence in stage IIB-IIIB disease seemed to reach their peak earlier than those in stage IA-IIA disease. The RFS was analyzed according to the grade of the ly factor, and the results indicated that the grade of the ly factor was significantly related to the RFS (P , .001). The RFS of patients with ly0 was better than those with ly1, ly2, and ly3, and the RFS became worse as the grade of the ly advanced. Similarly, the RFS was analyzed according to the grade of the v factor and the G factor, and the results indicated that the RFS also became worse as the grade of the v factor and the G factor advanced. Regarding the histologic subtype, patients with adenocarcinoma had a better RFS than the patients with other histologic subtypes (P , .001). To identify a population with a high risk of recurrence after resection, the T factor, N factor, ly factor,
1,011 (75.6) 327 (24.4) 984 (73.5) 277 (26.5) 1,042 (77.9) 296 (22.1)
Data are presented as No. (%) unless otherwise noted. G 5 tumor differentiation grade; ly 5 lymphatic vessel invasion; p 5 pathologic; v 5 blood vessel invasion.
whereas 334 patients (25.9%) were classified as v1, 188 patients (14.6%) were classified as v2, and 37 patients (2.9%) were classified as v3. Regarding
Figure 1. Recurrence-free survival curve classified according to the TNM stage.
1240
Downloaded From: http://journal.publications.chestnet.org/ by David Kinnison on 10/03/2013
Original Research
Figure 2. Estimated hazard rates for recurrence classified according to the TNM stage.
v factor, G factor, histologic subtype, and history of adjuvant chemotherapy were analyzed using the survival tree method. The T, N, v, and ly factors were selected as variables in the split steps, but the G factor, histologic subtype, and history of adjuvant chemotherapy were not selected in this study (Fig 3). The first split was produced by the v factor, and the patients were separated according to v 5 0 vs v ⱖ 1. The second split of the v 5 0 branch was produced by the T factor, and the patients were separated according to T ⱕ 1b vs T ⱖ 2a. The V 5 0, T ⱕ b group was separated with ly 5 0 vs ly ⱖ 1 at the third sprit. Similarly, in the v ⱖ 1 branch, the second split was produced by the N factor, and the patients were separated according to N ⱕ 1 vs N ⱖ 2. The third split of the v ⱖ 1, N ⱕ 1 group was produced by the T factor, and the patients were separated according to T ⱕ 2b vs T ⱖ 3. Finally, the 1,338 patients were classified into six groups; group A: v 5 0, T ⱕ 1b, ly 5 0; group B: v 5 0, T ⱕ 1b, ly ⱖ 1; group C: v 5 0, T ⱖ 2a; group D: v ⱖ 1, N ⱕ 1, T ⱕ 2b; group E: v ⱖ 1, N ⱕ 1, T ⱖ 3; and group F: v ⱖ 1, N ⱖ 2. The Kaplan-Meier curves for the RFS classified according to these six groups are shown in Figure 4.
Figure 3. Survival tree. ly 5 lymphatic vessel invasion; p 5 pathologic; Pt. No. 5 patient number; v 5 blood vessel invasion.
The curves for groups B, C, and D were identical; similarly, those for groups E and F were also almost identical. The six groups were further classified into three risk groups, as follows: a low-risk group (group A), a moderate-risk group (groups B, C, and D), and a high-risk group (groups E and F) (Figs 3, 4). The 5-year RFS rates were approximately 98% for the low-risk group, 75% for the moderate-risk group, and 30% for the high-risk group. The estimated hazard ratios for recurrence classified according to these three groups are shown in Figure 5. The estimated hazard ratios for recurrence increased and reached their peak at between 1 and 2 years after resection, then gradually decreased thereafter. The hazard ratio for recurrence in the high-risk group seemed to reach its peak earlier than that in the low-risk group. Discussion The TNM staging system is based on the characteristics of the primary tumor, the status of metastasis to the lymph nodes, and the presence/absence of distant metastasis. It is used worldwide, because it is simple and is correlated with patient prognosis. In patients with lung cancer, other factors, such as ly, v, and G factors, and more recently, the presence/absence of epidermal growth factor receptor mutation and the expressions of excision repair cross-complementation group 1 and ribonucleotide reductase M1 polypeptide, have been reported to be correlated with prognosis.5-7,13-16 However, testing for epidermal growth factor receptor mutation and the expressions of excision repair cross-complementation group 1and ribonucleotide reductase M1 polypeptide are not routinely performed in clinical practice. In this study, we focused on the T, N, ly, v, and G factors, all of which are easily evaluable in typical clinical practice, in an attempt to identify the risk factors for recurrence after resection in patients with lung cancer. The results of this study suggest that the combining of T, N, ly, and v factors in the risk estimation allowed the precise identification of a population with a high risk of recurrence.
Figure 4. Recurrence-free survival curves classified according to the six groups identified using the survival tree method.
journal.publications.chestnet.org
Downloaded From: http://journal.publications.chestnet.org/ by David Kinnison on 10/03/2013
CHEST / 144 / 4 / OCTOBER 2013
1241
Figure 5. Estimated hazard rates for recurrence classified according to the low-, moderate-, and high-risk groups.
A precise identification of the high-risk population for recurrence after resection might contribute to optimizing the postoperative follow-up surveillance. For instance, postoperative follow-up examinations might be unnecessary in the low-risk group (group A: v 5 0, T ⱕ 1b, ly 5 0), since recurrence is expected to be rare in this group. The Cox proportional hazards analysis is commonly used for multivariate analyses in many medical studies. The primary assumption in a Cox proportional hazards analysis is that the hazard functions of all data groups are proportional to one another. The survival tree model makes no assumption of proportional hazards; through a precise and complex program, it segregates groups by sequential partitioning based on statistical parameters related to the number of events (recurrences, deaths), allowing the progressive segregation of smaller groups with statistically significantly differences.17 The survival tree model, therefore, can identify meaningful prognostic subsets in a study population, which usually do not emerge from a routine proportional hazards analysis.10,18,19 At each step of the analysis in the survival tree model, all the possible splits into disjoint subsets are examined by a computer algorithm based on the log-rank test. The best split is then selected so that the split produces two subgroups of patients with the greatest differences in their Kaplan-Meier survival curves. Therefore, the variable that is used for the first split is considered to be the most significant factor. The first separation was created based on the presence/absence of v (v 5 0 vs v ⱖ 1) in this study. This outcome suggested that the presence/absence of v might be the most critical predictive factor of recurrence after resection. Interestingly, the variable in the third step, which separated group A and group B, was the presence/absence of ly (ly 5 0 vs ly ⱖ 1). Although the grades of the v factor and the ly factor were significantly related to the RFS, the presence/absence of vessel invasion might have
more influence than the grade for the prediction of recurrence. Another factor considered to be associated with recurrence might be visceral pleural invasion. Several investigators have reported that visceral pleural invasion is related to survival,20,21 but this factor was not evaluated in this study. Including the visceral pleural invasion factor among the ly and v factors study might have led to different results. Preoperative chemoradiotherapy was performed in 29 patients, and postoperative chemotherapy was performed in 327 patients; these patients were analyzed together in this study. Especially in the cases with preoperative chemoradiotherapy, the ly factor and v factor might have been influenced by the treatment. However, only 29 patients underwent preoperative chemoradiotherapy, and this number was so small that the effect of preoperative chemoradiotherapy was unlikely to have had a significant influence on the results. The presence/absence of adjuvant chemotherapy was included and analyzed using the survival tree method to evaluate the influence on recurrence but was not selected as a variable. Ideally, the patients with preoperative chemoradiotherapy and adjuvant chemotherapy should have been analyzed separately, or a subgroup analysis might have been needed. Metachronous lung cancer was diagnosed according to the criteria described in the Materials and Methods section, and patients with metachronous lung cancer were excluded from the present study. Of the 1,338 patients who were evaluated in this study, 53 patients (4.0%) developed metachronous lung cancer during the study period. The time intervals from the resection of the first lung cancer until the development of the second lung cancer ranged from 8 to 99 months. The hazard rate for the development the metachronous lung cancer over time is shown in Figure 6. The hazard rate increased gradually and reached a plateau at approximately 2 years, then decreased approximately 8 years after the first lung cancer. The possible reasons for the reduction in hazard rate might be related to patient age and
Figure 6. Estimated hazard rates for the development of metachronous lung cancer.
1242
Downloaded From: http://journal.publications.chestnet.org/ by David Kinnison on 10/03/2013
Original Research
the follow-up period of the study. The median age of the 53 patients with metachronous lung cancer was 69 years, and the patients would have been 77 years old after the passage of 8 years. Such an age might be too old for the development of second lung cancer. The longest follow-up period in this study was approximately 12 years. The number of patients who developed metachronous lung cancer would have likely increased if the follow-up period had been longer. The hazard rate for recurrence increased after resection and reached a peak between 1 and 2 years, gradually decreasing thereafter, as shown in Figures 1 and 5. In theory, the shape of the recurrence risk curve should be a single-peaked smooth curve; however, our results revealed curves with multiple peaks. One reason for this result might have been the study size. There were only 277 patients with recurrence in this study, which might not be a sufficiently large sample size to allow a smooth single-peaked curve to be obtained. However, Demicheli et al22 analyzed the recurrence dynamics after resection in a larger study and reported a similar multipeak recurrence risk curve. They explained this result using the tumor dormancy theory, which proposes that disseminated tumor cells interact with the microenvironment and show growth arrest if the environment is not permissive; then, once the environment becomes favorable again, the cells start to show uncontrolled proliferation and form metastases.23-26 This theory has been previously discussed in detail in relation to breast cancer, but it is still not clearly understood.27,28 Another possibility explaining the multipeak pattern might be related to our postoperative follow-up surveillance. As described in the Materials and Methods section, chest CT scan examinations were routinely repeated every year after resection in all patients. This procedure possibly detected asymptomatic recurrences, explaining the small peaks in the recurrence risk curves observed at around 2 and 3 years after resection. There is insufficient evidence, however, to arrive at a definitive conclusion, and as far as this study was concerned, we considered the multipeak pattern of the recurrence risk curve to be attributable to our postoperative follow-up surveillance protocol. Conclusions Risk factors for recurrence after resection were analyzed using the survival tree model in patients who had undergone a complete resection for NSCLC. The patients were classified into three groups (low, moderate, and high risk) according to the risk of recurrence by the combining the T, N, ly, and v factors. The results suggested that combining the T, N, ly, and v factors might enable an accurate estimation of the risk of recurrence after resection.
Acknowledgments Author contributions: Dr Sawada takes responsibility for the integrity of the data and the accuracy of the data analysis. Dr Sawada: contributed to the design, overseeing, and conduct of the study and to the writing and editing of the manuscript. Dr N. Yamashita: had full access to the data and contributed to the statistical analysis and drafting of the manuscript. Dr Suehisa: contributed to the study conception and design, acquisition of data, analysis and interpretation of data, drafting of the manuscript, and revision of the manuscript. Dr M. Yamashita: contributed to the study conception and design, acquisition of data, analysis and interpretation of data, drafting of the manuscript, and revision of the manuscript. Financial/nonfinancial disclosures: The authors have reported to CHEST that no potential conflicts of interest exist with any companies/organizations whose products or services may be discussed in this article. Other contributions: We thank Norihiro Teramoto, MD, PhD; Rieko Nishimura, MD, PhD; and Hiroyuki Takahata, MD, PhD of the Department of Pathology and Laboratory Medicine for their contribution to the pathological examinations. Without their invaluable assistance, this study would not have been possible.
References 1. Asamura H, Goya T, Koshiishi Y, et al; Japanese Joint Committee of Lung Cancer Registry. A Japanese Lung Cancer Registry study: prognosis of 13,010 resected lung cancers. J Thorac Oncol. 2008;3(1):46-52. 2. Goldstraw P, Crowley J, Chansky K, et al; International Association for the Study of Lung Cancer International Staging Committee; Participating Institutions. The IASLC Lung Cancer Staging Project: proposals for the revision of the TNM stage groupings in the forthcoming (seventh) edition of the TNM Classification of malignant tumours. J Thorac Oncol. 2007;2(8):706-714. 3. Groome PA, Bolejack V, Crowley JJ, et al; IASLC International Staging Committee; Cancer Research and Biostatistics; Observers to the Committee; Participating Institutions. The IASLC Lung Cancer Staging Project: validation of the proposals for revision of the T, N, and M descriptors and consequent stage groupings in the forthcoming (seventh) edition of the TNM classification of malignant tumours. J Thorac Oncol. 2007;2(8):694-705. 4. Silvestri GA, Tanoue LT, Margolis ML, Barker J, Detterbeck F; American College of Chest Physicians. The noninvasive staging of non-small cell lung cancer: the guidelines. Chest. 2003;123(suppl 1):147S-156S. 5. Macchiarini P, Fontanini G, Hardin MJ, et al. Blood vessel invasion by tumor cells predicts recurrence in completely resected T1 N0 M0 non-small-cell lung cancer. J Thorac Cardiovasc Surg. 1993;106(1):80-89. 6. Ito M, Moriya T, Ishida T, et al. Significance of pathological evaluation for lymphatic vessel invasion in invasive breast cancer. Breast Cancer. 2007;14(4):381-387. 7. Hashizume S, Nagayasu T, Hayashi T, et al. Accuracy and prognostic impact of a vessel invasion grading system for stage IA non-small cell lung cancer. Lung Cancer. 2009;65(3): 363-370. 8. Sobin LH, Gospodarowicz MK, Wittekind C. International Union Against Cancer. TNM Classification of Malignant Tumours. 7th ed. Chichester, West Sussex, England: WileyBlackwell; 2010. 9. Cancer AJCo. AJCC Cancer Staging Manual. 6th ed. New York, NY: Springer; 2002. 10. Biegon A, Hanau M, Greenberger V, Segal M. Aging and brain cholinergic muscarinic receptor subtypes: an autoradiographic study in the rat. Neurobiol Aging. 1989;10(4):305-310.
journal.publications.chestnet.org
Downloaded From: http://journal.publications.chestnet.org/ by David Kinnison on 10/03/2013
CHEST / 144 / 4 / OCTOBER 2013
1243
11. Lou F, Huang J, Sima CS, Dycoco J, Rusch V, Bach PB. Patterns of recurrence and second primary lung cancer in early-stage lung cancer survivors followed with routine computed tomography surveillance. J Thorac Cardiovasc Surg. 2013;145(1): 75-81. 12. Martini N, Melamed MR. Multiple primary lung cancers. J Thorac Cardiovasc Surg. 1975;70(4):606-612. 13. Hirsch FR, Varella-Garcia M, Bunn PA Jr, et al. Epidermal growth factor receptor in non-small-cell lung carcinomas: correlation between gene copy number and protein expression and impact on prognosis. J Clin Oncol. 2003;21(20):3798-3807. 14. Selvaggi G, Novello S, Torri V, et al. Epidermal growth factor receptor overexpression correlates with a poor prognosis in completely resected non-small-cell lung cancer. Ann Oncol. 2004;15(1):28-32. 15. Zheng Z, Chen T, Li X, Haura E, Sharma A, Bepler G. DNA synthesis and repair genes RRM1 and ERCC1 in lung cancer. N Engl J Med. 2007;356(8):800-808. 16. Lipford EH III, Eggleston JC, Lillemoe KD, Sears DL, Moore GW, Baker RR. Prognostic factors in surgically resected limited-stage, nonsmall cell carcinoma of the lung. Am J Surg Pathol. 1984;8(5):357-365. 17. Sevin BU, Lu Y, Bloch DA, Nadji M, Koechli OR, Averette HE. Surgically defined prognostic parameters in patients with early cervical carcinoma. A multivariate survival tree analysis. Cancer. 1996;78(7):1438-1446. 18. Altman RD, Medsger TA Jr, Bloch DA, Michel BA. Predictors of survival in systemic sclerosis (scleroderma). Arthritis Rheum. 1991;34(4):403-413. 19. Piette JD, Intrator O, Zierler S, Mor V, Stein MD. An exploratory analysis of survival with AIDS using a nonparametric tree-structured approach. Epidemiology. 1992;3(4):310-318.
20. Kudo Y, Saji H, Shimada Y, et al. Impact of visceral pleural invasion on the survival of patients with non-small cell lung cancer. Lung Cancer. 2012;78(2):153-160. 21. Hung JJ, Jeng WJ, Hsu WH, Chou TY, Lin SF, Wu YC. Prognostic significance of the extent of visceral pleural invasion in completely resected node-negative non-small cell lung cancer. Chest. 2012;142(1):141-150. 22. Demicheli R, Fornili M, Ambrogi F, et al. Recurrence dynamics for non-small-cell lung cancer: effect of surgery on the development of metastases. J Thorac Oncol. 2012;7(4): 723-730. 23. Páez D, Labonte MJ, Bohanes P, et al. Cancer dormancy: a model of early dissemination and late cancer recurrence. Clin Cancer Res. 2012;18(3):645-653. 24. Aguirre-Ghiso JA. Models, mechanisms and clinical evidence for cancer dormancy. Nat Rev Cancer. 2007;7(11): 834-846. 25. Braun S, Vogl FD, Naume B, et al. A pooled analysis of bone marrow micrometastasis in breast cancer. N Engl J Med. 2005;353(8):793-802. 26. Ranganathan AC, Adam AP, Aguirre-Ghiso JA. Opposing roles of mitogenic and stress signaling pathways in the induction of cancer dormancy. Cell Cycle. 2006;5(16):1799-1807. 27. Demicheli R, Miceli R, Moliterni A, et al. Breast cancer recurrence dynamics following adjuvant CMF is consistent with tumor dormancy and mastectomy-driven acceleration of the metastatic process. Ann Oncol. 2005;16(9):1449-1457. 28. Demicheli R, Abbattista A, Miceli R, Valagussa P, Bonadonna G. Time distribution of the recurrence risk for breast cancer patients undergoing mastectomy: further support about the concept of tumor dormancy. Breast Cancer Res Treat. 1996; 41(2):177-185.
1244
Downloaded From: http://journal.publications.chestnet.org/ by David Kinnison on 10/03/2013
Original Research