- Email: [email protected]
A nomogram to predict brain metastasis as the first relapse in curatively resected non-small cell lung cancer patients
Accepted Manuscript Title: A Nomogram To Predict Brain Metastasis As the First Relapse in Curatively Resected Non-Small Cell Lung Cancer Patients Author: Young-Woong Won Jungnam Joo Tak Yun Geon-Kook Lee Ji-Youn Han Heung Tae Kim Jin Soo Lee Moon Soo Kim Jong Mog Lee Hyun-Sung Lee Jae Ill Zo Sohee Kim PII: DOI: Reference:
S0169-5002(15)00094-X http://dx.doi.org/doi:10.1016/j.lungcan.2015.02.006 LUNG 4791
To appear in:
Lung Cancer
Received date: Revised date: Accepted date:
20-10-2014 28-1-2015 8-2-2015
Please cite this article as: Won Y-W, Joo J, Yun T, Lee G-K, Han J-Y, Kim HT, Lee JS, Kim MS, Lee JM, Lee H-S, Zo JI, Kim S, A Nomogram To Predict Brain Metastasis As the First Relapse in Curatively Resected Non-Small Cell Lung Cancer Patients, Lung Cancer (2015), http://dx.doi.org/10.1016/j.lungcan.2015.02.006 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Original article Title: A Nomogram To Predict Brain Metastasis As the First Relapse in Curatively Resected NonSmall Cell Lung Cancer Patients Young-Woong Won, M.D. a,b*, Jungnam Joo, Ph.D. e*, Tak Yun, M.D.c, Geon-Kook Lee, M.D. c, Ji-Youn Han, M.D. c, Heung Tae Kim, M.D. c, Jin Soo Lee, M.D. c, Moon Soo Kim, M.D. c, Jong Mog Lee, M.D. c, Hyun-
ip t
Sung Lee, M.D. c, Jae Ill Zo, M.D. c,d, Sohee Kim, M.S.e *Y. –W Won and J Joo contributed equally to this work.
Center for Clinical Trials, National Cancer Center Hospital, National Cancer Center, Goyang, South Korea :
323 Ilsan-ro, Ilsandong-gu, Goyang-si, Gyeonggi-go, Republic of Korea
Division of Hematology and Oncology, Department of Internal Medicine, Hanyang University College of
us
b
cr
a
Medicine, Seoul, South Korea :153 Gyeongchun-ro, Guri-si, Gyeonggi-do, Republic of Korea Center for Lung Cancer, Research Institute and Hospital, National Cancer Center, Goyang, South Korea :
an
c
323 Ilsan-ro, Ilsandong-gu, Goyang-si, Gyeonggi-go, Republic of Korea d
Department of Thoracic and Cardiovascular Surgery, Samsung Medical Center, Sungkyunkwan University
e
M
School of Medicine, Seoul, South Korea : 81 Irwon-Ro, Gangnam-gu, Seoul, Republic of Korea Biometric Research Branch, Research Institute, National Cancer Center, Goyang, South Korea : 323 Ilsan-
Tak Yun, M.D.
te
Corresponding author:
d
ro, Ilsandong-gu, Goyang-si, Gyeonggi-go, Republic of Korea
Ac ce p
Center for Lung Cancer, Research Institute and Hospital, National Cancer Center Address :323 Ilsan-ro, Ilsandong-gu, Goyang-si, Gyeonggi-go, 410-769, Republic of Korea Email: [email protected] Office: +82-31-920-1621 Fax: +82-31-920-2798
1
Page 1 of 20
Abstract Objectives: Development of brain metastasis results in a significant reduction in overall survival. However, there is no an effective tool to predict brain metastasis in non-small cell lung cancer (NSCLC) patients. We conducted this study to develop a feasible nomogram that can predict metastasis to the brain as the first relapse site in patients with curatively resected NSCLC.
ip t
Material and Methods: A retrospective review of NSCLC patients who had received curative surgery at
cr
National Cancer Center (Goyang, South Korea) between 2001 and 2008 was performed. We chose metastasis to the brain as the first relapse site after curative surgery as the primary endpoint of the study. A
us
nomogram was modeled using logistic regression.
Results: Among 1,218 patients, brain metastasis as the first relapse developed in 87 patients (7.14%) during
an
the median follow-up of 43.6 months. Occurrence rates of brain metastasis were higher in patients with adenocarcinoma or those with a high pT and pN stage. Younger age appeared to be associated with brain
M
metastasis, but this result was not statistically significant. The final prediction model included histology, smoking status, pT stage, and the interaction between adenocarcinoma and pN stage. The model showed
d
fairly good discriminatory ability with a C-statistic of 69.3% and 69.8% for predicting brain metastasis
te
within 2 years and 5 years, respectively. Internal validation using 2,000 bootstrap samples resulted in C-
Ac ce p
statistics of 67.0% and 67.4% which still indicated good discriminatory performances. Conclusion: The nomogram presented here provides the individual risk estimate of developing metastasis to the brain as the first relapse site in patients with NSCLC who have undergone curative surgery. Surveillance programs or preventive treatment strategies for brain metastasis could be established based on this nomogram.
Keywords: Nomogram; Brain Metastasis; Non-small Cell Lung Cancer; Relapse
2
Page 2 of 20
1. Introduction Lung cancer is the major cause of death from cancer worldwide [1]. Among lung cancers, non-small cell lung cancer (NSCLC) is the most common histology [2]. In general, for patients with early-stage NSCLC, surgery provides the best chance for cure. Some patients with advanced disease can also be treated curatively with surgery. Although prognosis of patients with resectable NSCLC has improved, recurrence of
ip t
disease is still a critical problem. Unfortunately, it is not clear what the optimal interval and modalities for
cr
detecting recurrent disease are. Additionally, it is not clear that early detection of a recurrence without symptoms would improve prognosis. Despite the paucity of scientific evidence, however, various
us
surveillance strategies to monitor recurrence after surgery for NSCLC have been proposed and accepted. However, there are no strategies to detect an asymptomatic brain relapse [3].
an
A nomogram is a graphical predictive tool that uses a statistical regression model to measure the impact of various clinicopathological parameters on the likelihood of an event in question [4]. It helps
M
clinicians to estimate an individual patient’s risk of recurrence and prognosis, to individualize treatment, and to select suitable patients for randomized controlled trials. We conducted this study to develop a feasible
d
nomogram that can predict metastasis to the brain as the first relapse site in patients with curatively resected
te
NSCLC. This nomogram was developed with clinical and pathologic variables using a multivariable model
Ac ce p
to predict the probability of occurrence of brain metastasis after surgery.
2. Materials and Methods
2.1. Study patients and clinical data collection We performed a retrospective review of NSCLC patients who had received curative surgery at National Cancer Center (Goyang, South Korea) from 2001 to 2008. Among patients who had histologically confirmed NSCLC, those who were treated with curative surgery for NSCLC, had complete resection without residual disease after surgery, and had no history of prior other malignancy were eligible for the study. A follow-up time of at least 24 months was a criterion for inclusion unless brain metastasis was observed earlier than 24 months. Patients with distant metastasis prior to surgery were excluded. In our 3
Page 3 of 20
institution, there was no regular surveillance program for detecting brain metastases. Therefore, radiologic examinations for brain metastases were performed when a brain metastasis was suspected by the physician. Clinicopathological data were retrieved from electronic medical records. TNM stage was classified according to the UICC/AJCC 6th TNM staging system [5]. The course of treatment and disease were also reviewed. This study was approved by the Institutional Review Board of the National Cancer Center
cr
ip t
(Goyang, South Korea [NCCNCS-11-446]).
2.2. Definition of terms
us
Brain metastasis includes metastatic disease to brain parenchyma or leptomeningeal seeding. Patients with brain metastasis only were described as ‘isolated brain metastasis’ patients. Patients with
an
simultaneously documented brain metastasis and other systemic metastatic site(s) were described as ‘simultaneous brain metastasis’ patients. Patients with local relapse or systemic metastasis except brain
M
metastasis followed by brain metastasis were described as ‘subsequent brain metastasis’ patients. Metastasis to the brain as the first relapse site was the focus of this study, and included isolated and simultaneous brain
d
metastasis. Therefore, in this paper, ‘brain metastasis’ indicates a metastasis to the brain as the first relapse
Ac ce p
and subsequent brain metastasis.
te
site. In contrast, ‘overall brain metastasis’ includes isolated brain metastasis, simultaneous brain metastasis,
2.3. Primary endpoint
Clinical, laboratory, and radiologic examinations are recommended every 6-12 months for 2 years according to the National Comprehensive Cancer Network (NCCN) guidelines for follow-up of NSCLC. In this study, the primary endpoint of interest was time from the surgery to brain metastasis as a first relapse site after curative surgery in NSCLC patient who received curative surgery. Time to brain metastasis was defined from the date of surgery to the date of brain metastasis as a first relapse site. Patients without relapse were censored at the time of last follow-up. Patients with either local relapse or systemic metastasis except brain metastasis followed by brain metastasis were censored at the time of first relapse. 4
Page 4 of 20
2.4. Statistical analyses Continuous variables were summarized as medians and ranges, and categorical variables as proportions. Wilcoxon rank sum test and chi-squared test were used to test differences in distributions of continuous and categorical variables between groups, respectively. Impact of potential risk factors on the
ip t
primary endpoint was evaluated by univariable and multivariable Cox proportional hazard model. Factors
cr
considered in the model included age, gender, histology, smoking status, pT stage, and pN stage. pT stage was categorized as 1 vs. 2-4 and pN stage as 0-1 vs. 2-3 based on the results from the univariable model. For
us
multivariable analysis, a variable selection method with selection criteria of 0.1 was considered, and interactions among factors significant at 0.1 were also included in the model. The goodness of fit of the
an
model was assessed using the Hosmer-Lemeshow (HL) type chi-square test developed for survival model. The discrimination ability of the multivariable Cox model for predicting 2-year and 5-year probabilities of
M
developing brain metastasis as a first relapse site was assessed using the C-statistic, a concordance measure analogous to area under the receiver operating characteristic (ROC) curve of the logistic regression[6-9].The
d
performance of the model was validated using 2,000 bootstrap re-samplings. Biases of the performance
te
measures were first estimated using the bootstrap samples, and then corrected with the apparent measures to
Ac ce p
produce bias-corrected values [4]. Finally, after checking the acceptability of the bias corrected performance measure, a nomogram was built based on this model for predicting 2-year and 5-year brain metastasis-free survival which can be easily generalized to predict that of other time point. All statistical analyses were performed using SAS version 9.1 (Cary, NC) and R statistical software, version 2.12.
3. Results
3.1. Patient characteristics and occurrence of brain metastasis after curative surgery We identified a total of 1,218 patients who were treated with curative surgery between February 2001 and December 2008, and who were followed-up for at least 24 months unless event occurred. The patients’ demographic and clinical characteristics are summarized in Table 1. Median age at the time of 5
Page 5 of 20
surgery was 62 years (range, 24 to 87 years). Median follow-up duration was 43.6 months (interquartile range [IQR], 28.3 to 63.2 months). Overall brain metastasis, including subsequent brain metastasis, occurred in 104 patients (8.5%). Metastasis to the brain as the first relapse site occurred in 87 patients (7.1%). Patients’clinicopathological factors according to the status of metastasis to the brain as the first relapse site are also summarized in Table 1. Patients with adenocarcinoma histology showed a high occurrence of brain
ip t
metastasis (adenocarcinoma [ADC] vs. non-ADC, 10.8% [60/553] vs. 4.06% [27/665], odds ratio [OR] 2.67,
cr
P<0.001). pN stage was also associated with the development of brain metastases (pN0-1 vs. pN2-3, 5.2% [49/943] vs.13.8% [38/275], OR 2.66, P <0.001). Occurrence rate of brain metastasis was also significantly
us
higher in patients with higher pT stage (pT1 vs. pT2-4, 4.5% [14/312] vs. 8.1% [73/906], OR 1.80, P =0.035). Brain metastasis (as the first relapse)-free survival according to clinicopathological factors is
an
indicated in Figure 1.
M
3.2. Prediction model and nomogram for brain metastasis as the first relapse within 2 years In the univariable model, adenocarcinoma histology, pT2-4 stage, and pN 2-3 stage were significant
d
predictors of brain metastasis as the first relapse, with p-values less than 0.1 (Table 2). Final multivariable
te
prediction model after backward variable selection with interaction included histology (ADC vs. non-ADC),
Ac ce p
pT stage (1 vs. 2-4), pN stage (0-1 vs. 2-3), smoking status (ever-smoker vs. never-smoker), and the interaction between ADC and pN stage. Time dependent ROC curves of this prediction model are shown in Figure 2 for predicting 2 and 5 years probabilities of developing metastasis to the brain as the first relapse site; the AUC was 69.3% (95% confidence interval [CI], 62.6%-76.0%), and 69.8% (95% CI, 63.6%-75.9%) for 2 and 5 years, respectively, which implied that the model had fairly good discrimination ability. HL-type chi-square statistic (P-value) for 2 and 5 years were 1.42 (P -value =0.841) and 0.4 (P-value = 0.982), which demonstrated that the model was well calibrated. Internal validation using 2,000 bootstrap resamples revealed that the model still had acceptable performance with a bias-corrected AUC of 67.0% and 67.4% , respectively. Using this model, we developed a nomogram that predicts the probability of developing metastasis to the brain as the first relapse site within 2 and 5 years after surgery; this nomogram is shown in 6
Page 6 of 20
Figure 3.
4. Discussion In the present study, we developed a nomogram that uses readily available clinicopathological information to predict the probability of developing brain metastasis after curative surgery in NSCLC
ip t
patients. Since the prediction model used time to brain metastasis as a primary endpoint, the nomogram can
cr
be developed for any time point of interest, and we chose 2 years and 5 years in this study.
Clinicopathological factors included in the nomogram are histology, pN stage, pT stage, and smoking status.
us
The model had fairly good discrimination ability with a C-statistic of 69.3%, and 69.8% for prediction of 2 years and 5 years. Internal validation using 2,000 bootstrap samples resulted in a C-statistic of 67.0% and
an
67.4%, indicating that the discrimination performance of the model was still good. Calibration of the model evaluated with the HL chi-square test supported the goodness of fit of the nomogram to the data. To our
M
knowledge, the present nomogram is the first that has been developed specifically to predict the probability of developing brain metastasis in NSCLC patients who have undergone curative surgery.
d
In previous studies, the occurrence rates of brain metastasis in NSCLC patients who received
te
curative surgery were 10% to 38% [10-12]. However, in our study, brain metastasis occurred in only 8.5% of
Ac ce p
patients. This lower occurrence rate was also reflected in the tumor stage results (stage I/II, 5.1%; stage III, 12.3%). It is unclear why the incidence of brain metastasis in our study was lower than that reported in previous studies, as the follow-up duration of our study was not shorter than that of previous studies. Possible explanations could be high proportions of stage I (50.2%) disease and squamous histology (46.1%) patients than were included in the previous studies. Similar to our study results, only 6.8% of 1,532 patients who underwent complete resection, mostly for stage I and II NSCLC, had documented first recurrences involving the brain [13]. Hubbs et al. reported diagnosis of overall brain metastasis in 6.2% of patients among 975 stage I/II NSCLC patients after surgery. The occurrence rate of isolated brain metastasis was 2.05%. Median follow-up duration was 33 months. The median time from surgery to development of brain metastasis was 10 months [14]. In contrast, Carolan et al. reported that overall brain metastasis occurred in 7
Page 7 of 20
34.9% of patients with stage III NSCLC after combined modality therapy. Metastasis to the brain as the first relapse site was observed in 18.1% of patients. The median time between diagnosis and first failure (brain metastasis) was 11.7 months [15]. Risk factors for development of brain metastasis identified in our study include adenocarcinoma histology, smoking, high pT stage, and high pN stage. Smoking was not a significant risk factor in the
ip t
univariable analysis; however it became significant when other risk factors were adjusted in the
cr
multivariable analysis. Non-squamous histology (especially, adenocarcinoma) has consistently been reported to be a risk factor for brain metastasis. Furthermore, higher T and N stage and younger age (<60 years) have
us
also been shown to be associated with an increased risk of developing brain metastasis [13, 14, 16-19]. However, age failed to reach statistical significance in our data set. In previously published nomograms, the
an
range of variables considered is usually determined based on data availability and clinical evidence, rather than solely on statistical significance [20]. However, since the inclusion of this variable did not alter the
M
development and the performance of the prediction model, we excluded age in the final model. Based on these risk factors, we chose to use parameters that were objective and easily measurable to facilitate accurate
d
prediction of brain metastasis by clinicians.
te
The prognosis of NSCLC patients without metastasis has improved substantially in the past decade.
Ac ce p
Optimized loco-regional control and adjuvant chemotherapy have contributed to improved survival [21, 22]. However, these advances do not seem to alter brain relapse rates [23-25]. Despite treatment of brain metastasis by radiation therapy, surgery, or medical therapy, prognosis of NSCLC patients with brain metastasis is poor, with a median survival of 3-6 months [26-29]. Prophylactic cranial irradiation (PCI) has been discussed as a strategy to reduce the risk of brain metastases. A number of trials have shown that PCI is effective at reducing brain metastases and improving survival in small cell lung cancer (SCLC) patients [30, 31]. However, although PCI in patients with locally-advanced NSCLC significantly decreases the risk of brain metastasis, there is no survival benefit of PCI [32, 33]. Additionally, whole brain radiation therapy has the potential risk of neurocognitive complications [34, 35]. By identifying the subgroup of patients at high risk for developing brain metastasis, the nomogram we developed can potentially increase the benefits and 8
Page 8 of 20
decrease the risks of PCI. NCCN guidelines for follow-up of NSCLC include history taking, physical examination, and chest computed tomography (CT) every 6-12 months for 2 years, then annually. In contrast, diagnostic tests for brain, such as brain CT or magnetic resonance imaging (MRI), are performed only when there is a clinical suspicion of brain metastasis. Yokoi et al. performed frequent follow-up brain CT in patients following
ip t
surgical resection for NSCLC. They noted a longer median survival in the asymptomatic group than the
cr
symptomatic group [36]. Nishikawa et al. evaluated the optimal brain MRI follow-up interval for NSCLC patients to detect radiosurgically manageable brain metastases. They suggested that early detection of brain
us
metastases by biannual MRI follow-up may decrease the need for invasive treatment of NSCLC patients [37]. Additionally, a randomized trial showed improved survival for patients with surgically treated brain
an
metastases than those with palliatively-treated brain metastases [29]. Therefore, close observation with regular brain radiologic studies could lead to early detection of brain metastases before the patient becomes
M
symptomatic, and prognosis of brain metastasis may be improved by early detection and treatment. Our nomogram could be used to establish an optimal screening strategy for brain relapse.
d
Our study had some limitations. First, we included patients from a single center. In contrast, models
te
derived from multicenter or population-based cohorts are more likely to be generalizable. However, these
Ac ce p
latter types of studies may also be hampered by the lack of consistent availability of detailed data elements that may improve prognostic accuracy. Second, we validated the nomogram using bootstrap resampling. Although bootstrapping is a sample reuse method that prevents over-interpretation of current data, this method does not ensure external applicability [4]. Third, we classified TNM stage according to the UICC/AJCC 6th TNM staging system. Since 2009, however, the UICC/AJCC 7thTNM staging system has been used [38]. In our study, T stages were divided into T1 and T2-4. Fortunately, there was no change in definition of the T1 stage between the two versions of the UICC/AJCC TNM staging system. Definition of the N stage also remained the same, therefore this is unlikely to have affected our results. In conclusion, the nomogram presented here integrates clinicopathological variables to provide an individual risk estimate of developing metastasis to the brain as the first relapse site in patients with NSCLC 9
Page 9 of 20
who have undergone curative surgery. Surveillance programs or preventive treatment strategies for brain metastasis could potentially be established using this nomogram, and it could aid in identification of patients with similar prognoses.
Acknowledgement: This work was supported by a grant from the National Cancer Center, South Korea
ip t
(grant number. 1210060).
Ac ce p
te
d
M
an
us
cr
Disclosure: The authors have no conflicts of interest to declare.
10
Page 10 of 20
References 1.
Siegel R, Ward E, Brawley O, Jemal A. Cancer statistics, 2011. CA: A Cancer Journal for Clinicians 2011.
2.
Ferlay J, Parkin D, Curado M, Bray F, Edwards B, Shin H. Cancer incidence in five continents, volumes I to IX: IARC CancerBase no. 9 [Internet]. Lyon (France): International Agency for Research on Cancer; 2010. In, 2011. Calman L, Beaver K, Hind D, Lorigan P, Roberts C, Lloyd-Jones M. Survival Benefits from Follow-Up of
ip t
3.
Patients with Lung Cancer: A Systematic Review and Meta-Analysis. Journal of Thoracic Oncology 2011;6: 1993.
Iasonos A, Schrag D, Raj GV, Panageas KS. How to build and interpret a nomogram for cancer prognosis. Journal of Clinical Oncology 2008;26: 1364-1370.
5.
Goldstraw P, Crowley JJ. The International Association for the Study of Lung Cancer international staging
us
project on lung cancer. Journal of Thoracic Oncology 2006;1: 281. 6.
cr
4.
Harrell F, Lee KL, Mark DB. Tutorial in biostatistics multivariable prognostic models: issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors. Statistics in medicine
7.
an
1996;15: 361-387.
Harrell FE, Califf RM, Pryor DB, Lee KL, Rosati RA. Evaluating the yield of medical tests. Jama 1982;247: 2543-2546.
Pencina MJ, D'Agostino RB. Overall C as a measure of discrimination in survival analysis: model specific
M
8.
population value and confidence interval estimation. Statistics in medicine 2004;23: 2109-2123. 9.
Royston P, Altman DG. External validation of a Cox prognostic model: principles and methods. BMC Douillard JY, Rosell R, De Lena M, Carpagnano F, Ramlau R, Gonzáles-Larriba JL, Grodzki T, Pereira JR, Le
te
10.
d
medical research methodology 2013;13: 33.
Groumellec A, Lorusso V. Adjuvant vinorelbine plus cisplatin versus observation in patients with completely resected stage IB–IIIA non-small-cell lung cancer (Adjuvant Navelbine International Trialist 11.
Ac ce p
Association [ANITA]): a randomised controlled trial. The lancet oncology 2006;7: 719-727. Gilbert S, Reid KR, Lam MY, Petsikas D. Who should follow up lung cancer patients after operation? The Annals of thoracic surgery 2000;69: 1696-1700. 12.
Yano T, Yokoyama H, Inoue T, Asoh H, Tayama K, Takai E, Ichinose Y. The first site of recurrence after complete resection in non-small-cell carcinoma of the lung. Comparison between pN0 disease and pN2 disease. The Journal of thoracic and cardiovascular surgery 1994;108: 680.
13.
Figlin RA, Piantadosi S, Feld R. Intracranial recurrence of carcinoma after complete surgical resection of stage I, II, and III non-small-cell lung cancer. N Engl J Med 1988;318: 1300-1305.
14.
Hubbs JL, Boyd JA, Hollis D, Chino JP, Saynak M, Kelsey CR. Factors associated with the development of brain metastases. Cancer 2010;116: 5038-5046.
15.
Carolan H, Sun AY, Bezjak A, Yi QL, Payne D, Kane G, Waldron J, Leighl N, Feld R, Burkes R. Does the incidence and outcome of brain metastases in locally advanced non-small cell lung cancer justify prophylactic cranial irradiation or early detection? Lung Cancer 2005;49: 109-115.
16.
Bajard A, Westeel V, Dubiez A, Jacoulet P, Pernet D, Dalphin J, Depierre A. Multivariate analysis of factors predictive of brain metastases in localised non-small cell lung carcinoma. Lung Cancer 2004;45: 317-323. 11
Page 11 of 20
17.
Ceresoli GL, Reni M, Chiesa G, Carretta A, Schipani S, Passoni P, Bolognesi A, Zannini P, Villa E. Brain metastases in locally advanced nonsmall cell lung carcinoma after multimodality treatment: risk factors analysis. Cancer 2002;95: 605.
18.
Feld R, Rubinstein LV, Weisenberger TH. Sites of recurrence in resected stage I non-small-cell lung cancer: a guide for future studies. Journal of Clinical Oncology 1984;2: 1352-1358.
19.
Martini N, Bains MS, Burt ME, Zakowski MF, McCormack P, Rusch VW, Ginsberg RJ. Incidence of local recurrence and second primary tumors in resected stage I lung cancer. The Journal of thoracic and
20.
ip t
cardiovascular surgery 1995;109: 120-129. Wong SL, Kattan MW, McMasters KM, Coit DG. A Nomogram That Predicts the Presence of Sentinel Node Metastasis in Melanoma With Better Discrimination Than the American Joint Committee on CancerStaging 21.
cr
System. Annals of surgical oncology 2005;12: 282-288.
Arriagada R, Auperin A, Burdett S, Higgins J, Johnson D, Le Chevalier T, Le Pechoux C, Parmar M, Pignon J, Souhami R. Adjuvant chemotherapy, with or without postoperative radiotherapy, in operable non-small-
22.
us
cell lung cancer: two meta-analyses of individual patient data. Lancet 2010;375: 1267.
Aupérin A, Le Péchoux C, Rolland E, Curran WJ, Furuse K, Fournel P, Belderbos J, Clamon G, Ulutin HC, Paulus R. Meta-analysis of concomitant versus sequential radiochemotherapy in locally advanced non–
23.
an
small-cell lung cancer. Journal of Clinical Oncology 2010;28: 2181-2190.
Andre F, Grunenwald D, Pujol JL, Girard P, Dujon A, Brouchet L, Brichon PY, Westeel V, Le Chevalier T. Patterns of relapse of N2 nonsmall-cell lung carcinoma patients treated with preoperative chemotherapy:
24.
M
should prophylactic cranial irradiation be reconsidered? Cancer 2001;91: 2394-2400. Gaspar LE, Chansky K, Albain KS, Vallieres E, Rusch V, Crowley JJ, Livingston RB, Gandara DR. Time from treatment to subsequent diagnosis of brain metastases in stage III non-small-cell lung cancer: a 25.
d
retrospective review by the Southwest Oncology Group. J Clin Oncol 2005;23: 2955-2961. Strauss GM, Herndon JE, Sherman DD, Mathisen DJ, Carey RW, Choi NC, Rege VB, Modeas C, Green MR.
te
Neoadjuvant chemotherapy and radiotherapy followed by surgery in stage IIIA non-small-cell carcinoma of the lung: report of a Cancer and Leukemia Group B phase II study. J Clin Oncol 1992;10: 1237-1244. Alexander III E, Moriarty TM, Davis RB, Wen PY, Fine HA, Black PM, Kooy HM, Loeffler JS. Stereotactic
Ac ce p
26.
radiosurgery for the definitive, noninvasive treatment of brain metastases. Journal of the National Cancer Institute 1995;87: 34-40. 27.
Dimitropoulos C, Hillas G, Nikolakopoulou S, Kostara I, Sagris K, Vlastos F, Alchanatis M. Prophylactic cranial irradiation in non-small cell lung cancer patients: who might be the candidates? Cancer management and research 2011;3: 287.
28.
Jacot W, Quantin X, Boher J, Andre F, Moreau L, Gainet M, Depierre A, Quoix E, Le Chevalier T, Pujol J. Brain metastases at the time of presentation of non-small cell lung cancer: a multi-centric AERIO analysis of prognostic factors. British journal of cancer 2001;84: 903.
29.
Patchell RA, Tibbs PA, Walsh JW, Dempsey RJ, Maruyama Y, Kryscio RJ, Markesbery WR, Macdonald JS, Young B. A randomized trial of surgery in the treatment of single metastases to the brain. New England Journal of Medicine 1990;322: 494-500.
30.
Aupérin A, Arriagada R, Pignon JP, Le Péchoux C, Gregor A, Stephens RJ, Kristjansen PEG, Johnson BE, Ueoka H, Wagner H. Prophylactic cranial irradiation for patients with small-cell lung cancer in complete remission. New England Journal of Medicine 1999;341: 476-484. 12
Page 12 of 20
31.
Meert AP, Paesmans M, Berghmans T, Martin B, Mascaux C, Vallot F, Verdebout JM, Lafitte JJ, Sculier JP. Prophylactic cranial irradiation in small cell lung cancer: a systematic review of the literature with metaanalysis. BMC cancer 2001;1: 5.
32.
Gore EM, Bae K, Wong SJ, Sun A, Bonner JA, Schild SE, Gaspar LE, Bogart JA, Werner-Wasik M, Choy H. Phase III comparison of prophylactic cranial irradiation versus observation in patients with locally advanced non–small-cell lung cancer: Primary analysis of Radiation Therapy Oncology Group Study RTOG 0214. Journal of Clinical Oncology 2011;29: 272-278. Lester JF, MacBeth FR, Coles B. Prophylactic cranial irradiation for preventing brain metastases in patients
ip t
33.
undergoing radical treatment for non–small-cell lung cancer: A Cochrane Review. International Journal of Radiation Oncology* Biology* Physics 2005;63: 690-694.
Aoyama H, Tago M, Kato N, Toyoda T, Kenjyo M, Hirota S, Shioura H, Inomata T, Kunieda E, Hayakawa K,
cr
34.
Nakagawa K, Kobashi G, Shirato H. Neurocognitive function of patients with brain metastasis who received either whole brain radiotherapy plus stereotactic radiosurgery or radiosurgery alone. Int J Radiat Oncol 35.
us
Biol Phys 2007;68: 1388-1395.
Meyers CA, Smith JA, Bezjak A, Mehta MP, Liebmann J, Illidge T, Kunkler I, Caudrelier JM, Eisenberg PD, Meerwaldt J, Siemers R, Carrie C, Gaspar LE, Curran W, Phan SC, Miller RA, Renschler MF. Neurocognitive
an
function and progression in patients with brain metastases treated with whole-brain radiation and motexafin gadolinium: results of a randomized phase III trial. J Clin Oncol 2004;22: 157-165. 36.
Yokoi K, Miyazawa N, Arai T. Brain metastasis in resected lung cancer: value of intensive follow-up with
37.
M
computed tomography. Ann Thorac Surg 1996;61: 546-550; discussion 551. Nishikawa T, Ueba T, Kawashima M, Kajiwara M, Iwata R, Kato M, Miyamatsu N, Yamashita K. Early detection of metachronous brain metastases by biannual brain MRI follow-up may provide patients with 770-774.
te
Goldstraw P, Crowley J, Chansky K, Giroux DJ, Groome PA, Rami-Porta R, Postmus PE, Rusch V, Sobin L. 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. Journal of Thoracic
Ac ce p
38.
d
non-small cell lung cancer with more opportunities to have radiosurgery. Clin Neurol Neurosurg 2010;112:
Oncology 2007;2: 706.
13
Page 13 of 20
Figure legend Figure 1. Brain metastasis-free survival according to clinicopathological factors. Figure 2. ROC curves (A) and calibration plots (B,C) for nomogram predicting brain metastasis as the first relapse within 2 and 5 years after curative resection
ip t
Figure 3. Nomogram predicting brain metastasis as the first relapse within 2 and 5 years after curative
Ac ce p
te
d
M
an
us
cr
resection.
14
Page 14 of 20
Tables Table 1. Patients’ demographic and clinical characteristics Total (N=1218)
Characteristics
Pathologic stage
0.169$ 0.462
500
(44.2)
42
(48.3)
> 60 years
676
(55.5)
631
(55.8)
45
(51.7)
Male
905
(74.3)
842
(74.4)
63
(72.4)
Female
313
(25.7)
289
(25.6)
24
(27.6)
Adenocarcinoma
553
(45.4)
493
(43.6)
60
(69.0) <0.001a
Squamous cell carcinoma
562
(46.1)
545
(48.2)
17
(19.5)
Others
103
(8.5)
93
(8.2)
10
(11.5)
BAC
32
(2.6)
31
(2.7)
1
(1.2)
Sarcomatoid carcinoma
27
(2.2)
23
(2.1)
4
(4.6)
Large cell NEC
36
(3.0)
31
(2.7)
5
(5.7)
8
(0.7)
0
(0.0)
361
(31.9)
28
(32.2)
770
(68.1)
59
(67.8)
ip t
(44.5)
0.676
cr
us
Undifferentiated or Pleomorphic carcinoma Never-smoker
8
(0.7)
389
(31.9)
Ever-smoker
829
(68.1)
IA
222
(18.2)
213
(18.8)
9
IB
389
(31.9)
373
(33.0)
16
(18.4)
IIA
42
(3.5)
40
(3.5)
2
(2.3)
IIB
214
(17.6)
197
(17.4)
17
(19.5)
0.959
(10.4) <0.001b
297
(24.4)
262
(23.2)
35
(40.2)
54
(4.4)
46
(4.1)
8
(9.2)
312
(25.6)
298
(26.3)
14
(16.1)
727
(59.7)
672
(59.4)
55
(63.2)
3
131
(10.8)
118
(10.5)
13
(14.9)
4
48
(3.9)
43
(3.8)
5
(5.8)
0
686
(56.3)
657
(58.1)
29
(33.3) <0.001d
1
257
(21.1)
237
(21.0)
20
(23.0)
2
268
(22.0)
233
(20.6)
35
(40.2)
3
7
(0.6)
4
(0.3)
3
(3.5)
798
(65.5)
749
(66.2)
49
(56.3)
420
(34.5)
382
(33.8)
38
(43.7)
1
Ac ce p
2
No
Yes $ Wilcoxon rank-sum test, Other tests are chi-square test. a
(%)
542
IIIB
Adjuvant Chemotherapy
N
61 (32-74)
≤ 60 years
IIIA
pT stage
(%)
Event (N=87) P-value
an
Smoking status
N
62 (24-87)
M
Histology
(%)
62 (24-87)
d
Gender
pN stage
N
median (min-max)
te
Age
No Event (N=1131)
0.035c
0.061
Histology: ADC vs. Non-ADC, bPathologic stage: I-II vs. III, cpT stage T1 vs. T2-4, dpN stage: N0-1 vs. N2-3
Abbreviations: BAC, bronchoalveolar carcinoma; NEC, neuroendocrine carcinoma, ADC, adenocarcinoma
15
Page 15 of 20
Table 2. Univariable and multivariable Cox regression analyses estimating the risk factors of time to brain metastasis as a first relapse site after curative resection
2.26 1.69 1.87 1.18
Yes vs. No
3.46
0.007 0.032 0.033 0.720
1.25 1.05 1.05 0.48
4.08 2.73 3.34 2.93
0.019
1.22
9.80
Ac ce p
te
d
M
an
Abbreviations: ADC, adenocarcinoma; CI, confidence interval
95% CI
ip t
Female vs. Male ADC vs. non-ADC Ever vs. Never T2-4 vs. T1 N2-3 vs. N0-1 Yes vs. No
P-value
cr
Age (year) Gender Histology Smoking status pT stage pN stage Adjuvant Chemotherapy Interaction between ADC and N2-3
Multivariable analysis Hazard ratio
us
Univariable analysis Hazard P-value 95% CI ratio 0.98 0.156 0.96 1.01 1.10 0.692 0.69 1.76 2.85 <0.001 1.81 4.49 1.02 0.944 0.65 1.59 2.01 0.017 1.14 3.56 3.19 <0.001 2.09 4.88 1.59 0.034 1.04 2.43
Clinicopathological factors
16
Page 16 of 20
Brain metastasis was higher in patients with adenocarcinoma histology. After surgery, high pT and pN stage were also risk factors for brain metastasis.
Ac ce p
te
d
M
an
us
cr
ip t
The nomogram provides the individual risk estimate of brain metastasis.
17
Page 17 of 20
Ac
ce
pt
ed
M
an
us
cr
i
Figure1
Page 18 of 20
Ac
ce
pt
ed
M
an
us
cr
i
Figure2
Page 19 of 20
Ac
ce
pt
ed
M
an
us
cr
i
Figure3
Page 20 of 20
S0169-5002(15)00094-X http://dx.doi.org/doi:10.1016/j.lungcan.2015.02.006 LUNG 4791
To appear in:
Lung Cancer
Received date: Revised date: Accepted date:
20-10-2014 28-1-2015 8-2-2015
Please cite this article as: Won Y-W, Joo J, Yun T, Lee G-K, Han J-Y, Kim HT, Lee JS, Kim MS, Lee JM, Lee H-S, Zo JI, Kim S, A Nomogram To Predict Brain Metastasis As the First Relapse in Curatively Resected Non-Small Cell Lung Cancer Patients, Lung Cancer (2015), http://dx.doi.org/10.1016/j.lungcan.2015.02.006 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Original article Title: A Nomogram To Predict Brain Metastasis As the First Relapse in Curatively Resected NonSmall Cell Lung Cancer Patients Young-Woong Won, M.D. a,b*, Jungnam Joo, Ph.D. e*, Tak Yun, M.D.c, Geon-Kook Lee, M.D. c, Ji-Youn Han, M.D. c, Heung Tae Kim, M.D. c, Jin Soo Lee, M.D. c, Moon Soo Kim, M.D. c, Jong Mog Lee, M.D. c, Hyun-
ip t
Sung Lee, M.D. c, Jae Ill Zo, M.D. c,d, Sohee Kim, M.S.e *Y. –W Won and J Joo contributed equally to this work.
Center for Clinical Trials, National Cancer Center Hospital, National Cancer Center, Goyang, South Korea :
323 Ilsan-ro, Ilsandong-gu, Goyang-si, Gyeonggi-go, Republic of Korea
Division of Hematology and Oncology, Department of Internal Medicine, Hanyang University College of
us
b
cr
a
Medicine, Seoul, South Korea :153 Gyeongchun-ro, Guri-si, Gyeonggi-do, Republic of Korea Center for Lung Cancer, Research Institute and Hospital, National Cancer Center, Goyang, South Korea :
an
c
323 Ilsan-ro, Ilsandong-gu, Goyang-si, Gyeonggi-go, Republic of Korea d
Department of Thoracic and Cardiovascular Surgery, Samsung Medical Center, Sungkyunkwan University
e
M
School of Medicine, Seoul, South Korea : 81 Irwon-Ro, Gangnam-gu, Seoul, Republic of Korea Biometric Research Branch, Research Institute, National Cancer Center, Goyang, South Korea : 323 Ilsan-
Tak Yun, M.D.
te
Corresponding author:
d
ro, Ilsandong-gu, Goyang-si, Gyeonggi-go, Republic of Korea
Ac ce p
Center for Lung Cancer, Research Institute and Hospital, National Cancer Center Address :323 Ilsan-ro, Ilsandong-gu, Goyang-si, Gyeonggi-go, 410-769, Republic of Korea Email: [email protected] Office: +82-31-920-1621 Fax: +82-31-920-2798
1
Page 1 of 20
Abstract Objectives: Development of brain metastasis results in a significant reduction in overall survival. However, there is no an effective tool to predict brain metastasis in non-small cell lung cancer (NSCLC) patients. We conducted this study to develop a feasible nomogram that can predict metastasis to the brain as the first relapse site in patients with curatively resected NSCLC.
ip t
Material and Methods: A retrospective review of NSCLC patients who had received curative surgery at
cr
National Cancer Center (Goyang, South Korea) between 2001 and 2008 was performed. We chose metastasis to the brain as the first relapse site after curative surgery as the primary endpoint of the study. A
us
nomogram was modeled using logistic regression.
Results: Among 1,218 patients, brain metastasis as the first relapse developed in 87 patients (7.14%) during
an
the median follow-up of 43.6 months. Occurrence rates of brain metastasis were higher in patients with adenocarcinoma or those with a high pT and pN stage. Younger age appeared to be associated with brain
M
metastasis, but this result was not statistically significant. The final prediction model included histology, smoking status, pT stage, and the interaction between adenocarcinoma and pN stage. The model showed
d
fairly good discriminatory ability with a C-statistic of 69.3% and 69.8% for predicting brain metastasis
te
within 2 years and 5 years, respectively. Internal validation using 2,000 bootstrap samples resulted in C-
Ac ce p
statistics of 67.0% and 67.4% which still indicated good discriminatory performances. Conclusion: The nomogram presented here provides the individual risk estimate of developing metastasis to the brain as the first relapse site in patients with NSCLC who have undergone curative surgery. Surveillance programs or preventive treatment strategies for brain metastasis could be established based on this nomogram.
Keywords: Nomogram; Brain Metastasis; Non-small Cell Lung Cancer; Relapse
2
Page 2 of 20
1. Introduction Lung cancer is the major cause of death from cancer worldwide [1]. Among lung cancers, non-small cell lung cancer (NSCLC) is the most common histology [2]. In general, for patients with early-stage NSCLC, surgery provides the best chance for cure. Some patients with advanced disease can also be treated curatively with surgery. Although prognosis of patients with resectable NSCLC has improved, recurrence of
ip t
disease is still a critical problem. Unfortunately, it is not clear what the optimal interval and modalities for
cr
detecting recurrent disease are. Additionally, it is not clear that early detection of a recurrence without symptoms would improve prognosis. Despite the paucity of scientific evidence, however, various
us
surveillance strategies to monitor recurrence after surgery for NSCLC have been proposed and accepted. However, there are no strategies to detect an asymptomatic brain relapse [3].
an
A nomogram is a graphical predictive tool that uses a statistical regression model to measure the impact of various clinicopathological parameters on the likelihood of an event in question [4]. It helps
M
clinicians to estimate an individual patient’s risk of recurrence and prognosis, to individualize treatment, and to select suitable patients for randomized controlled trials. We conducted this study to develop a feasible
d
nomogram that can predict metastasis to the brain as the first relapse site in patients with curatively resected
te
NSCLC. This nomogram was developed with clinical and pathologic variables using a multivariable model
Ac ce p
to predict the probability of occurrence of brain metastasis after surgery.
2. Materials and Methods
2.1. Study patients and clinical data collection We performed a retrospective review of NSCLC patients who had received curative surgery at National Cancer Center (Goyang, South Korea) from 2001 to 2008. Among patients who had histologically confirmed NSCLC, those who were treated with curative surgery for NSCLC, had complete resection without residual disease after surgery, and had no history of prior other malignancy were eligible for the study. A follow-up time of at least 24 months was a criterion for inclusion unless brain metastasis was observed earlier than 24 months. Patients with distant metastasis prior to surgery were excluded. In our 3
Page 3 of 20
institution, there was no regular surveillance program for detecting brain metastases. Therefore, radiologic examinations for brain metastases were performed when a brain metastasis was suspected by the physician. Clinicopathological data were retrieved from electronic medical records. TNM stage was classified according to the UICC/AJCC 6th TNM staging system [5]. The course of treatment and disease were also reviewed. This study was approved by the Institutional Review Board of the National Cancer Center
cr
ip t
(Goyang, South Korea [NCCNCS-11-446]).
2.2. Definition of terms
us
Brain metastasis includes metastatic disease to brain parenchyma or leptomeningeal seeding. Patients with brain metastasis only were described as ‘isolated brain metastasis’ patients. Patients with
an
simultaneously documented brain metastasis and other systemic metastatic site(s) were described as ‘simultaneous brain metastasis’ patients. Patients with local relapse or systemic metastasis except brain
M
metastasis followed by brain metastasis were described as ‘subsequent brain metastasis’ patients. Metastasis to the brain as the first relapse site was the focus of this study, and included isolated and simultaneous brain
d
metastasis. Therefore, in this paper, ‘brain metastasis’ indicates a metastasis to the brain as the first relapse
Ac ce p
and subsequent brain metastasis.
te
site. In contrast, ‘overall brain metastasis’ includes isolated brain metastasis, simultaneous brain metastasis,
2.3. Primary endpoint
Clinical, laboratory, and radiologic examinations are recommended every 6-12 months for 2 years according to the National Comprehensive Cancer Network (NCCN) guidelines for follow-up of NSCLC. In this study, the primary endpoint of interest was time from the surgery to brain metastasis as a first relapse site after curative surgery in NSCLC patient who received curative surgery. Time to brain metastasis was defined from the date of surgery to the date of brain metastasis as a first relapse site. Patients without relapse were censored at the time of last follow-up. Patients with either local relapse or systemic metastasis except brain metastasis followed by brain metastasis were censored at the time of first relapse. 4
Page 4 of 20
2.4. Statistical analyses Continuous variables were summarized as medians and ranges, and categorical variables as proportions. Wilcoxon rank sum test and chi-squared test were used to test differences in distributions of continuous and categorical variables between groups, respectively. Impact of potential risk factors on the
ip t
primary endpoint was evaluated by univariable and multivariable Cox proportional hazard model. Factors
cr
considered in the model included age, gender, histology, smoking status, pT stage, and pN stage. pT stage was categorized as 1 vs. 2-4 and pN stage as 0-1 vs. 2-3 based on the results from the univariable model. For
us
multivariable analysis, a variable selection method with selection criteria of 0.1 was considered, and interactions among factors significant at 0.1 were also included in the model. The goodness of fit of the
an
model was assessed using the Hosmer-Lemeshow (HL) type chi-square test developed for survival model. The discrimination ability of the multivariable Cox model for predicting 2-year and 5-year probabilities of
M
developing brain metastasis as a first relapse site was assessed using the C-statistic, a concordance measure analogous to area under the receiver operating characteristic (ROC) curve of the logistic regression[6-9].The
d
performance of the model was validated using 2,000 bootstrap re-samplings. Biases of the performance
te
measures were first estimated using the bootstrap samples, and then corrected with the apparent measures to
Ac ce p
produce bias-corrected values [4]. Finally, after checking the acceptability of the bias corrected performance measure, a nomogram was built based on this model for predicting 2-year and 5-year brain metastasis-free survival which can be easily generalized to predict that of other time point. All statistical analyses were performed using SAS version 9.1 (Cary, NC) and R statistical software, version 2.12.
3. Results
3.1. Patient characteristics and occurrence of brain metastasis after curative surgery We identified a total of 1,218 patients who were treated with curative surgery between February 2001 and December 2008, and who were followed-up for at least 24 months unless event occurred. The patients’ demographic and clinical characteristics are summarized in Table 1. Median age at the time of 5
Page 5 of 20
surgery was 62 years (range, 24 to 87 years). Median follow-up duration was 43.6 months (interquartile range [IQR], 28.3 to 63.2 months). Overall brain metastasis, including subsequent brain metastasis, occurred in 104 patients (8.5%). Metastasis to the brain as the first relapse site occurred in 87 patients (7.1%). Patients’clinicopathological factors according to the status of metastasis to the brain as the first relapse site are also summarized in Table 1. Patients with adenocarcinoma histology showed a high occurrence of brain
ip t
metastasis (adenocarcinoma [ADC] vs. non-ADC, 10.8% [60/553] vs. 4.06% [27/665], odds ratio [OR] 2.67,
cr
P<0.001). pN stage was also associated with the development of brain metastases (pN0-1 vs. pN2-3, 5.2% [49/943] vs.13.8% [38/275], OR 2.66, P <0.001). Occurrence rate of brain metastasis was also significantly
us
higher in patients with higher pT stage (pT1 vs. pT2-4, 4.5% [14/312] vs. 8.1% [73/906], OR 1.80, P =0.035). Brain metastasis (as the first relapse)-free survival according to clinicopathological factors is
an
indicated in Figure 1.
M
3.2. Prediction model and nomogram for brain metastasis as the first relapse within 2 years In the univariable model, adenocarcinoma histology, pT2-4 stage, and pN 2-3 stage were significant
d
predictors of brain metastasis as the first relapse, with p-values less than 0.1 (Table 2). Final multivariable
te
prediction model after backward variable selection with interaction included histology (ADC vs. non-ADC),
Ac ce p
pT stage (1 vs. 2-4), pN stage (0-1 vs. 2-3), smoking status (ever-smoker vs. never-smoker), and the interaction between ADC and pN stage. Time dependent ROC curves of this prediction model are shown in Figure 2 for predicting 2 and 5 years probabilities of developing metastasis to the brain as the first relapse site; the AUC was 69.3% (95% confidence interval [CI], 62.6%-76.0%), and 69.8% (95% CI, 63.6%-75.9%) for 2 and 5 years, respectively, which implied that the model had fairly good discrimination ability. HL-type chi-square statistic (P-value) for 2 and 5 years were 1.42 (P -value =0.841) and 0.4 (P-value = 0.982), which demonstrated that the model was well calibrated. Internal validation using 2,000 bootstrap resamples revealed that the model still had acceptable performance with a bias-corrected AUC of 67.0% and 67.4% , respectively. Using this model, we developed a nomogram that predicts the probability of developing metastasis to the brain as the first relapse site within 2 and 5 years after surgery; this nomogram is shown in 6
Page 6 of 20
Figure 3.
4. Discussion In the present study, we developed a nomogram that uses readily available clinicopathological information to predict the probability of developing brain metastasis after curative surgery in NSCLC
ip t
patients. Since the prediction model used time to brain metastasis as a primary endpoint, the nomogram can
cr
be developed for any time point of interest, and we chose 2 years and 5 years in this study.
Clinicopathological factors included in the nomogram are histology, pN stage, pT stage, and smoking status.
us
The model had fairly good discrimination ability with a C-statistic of 69.3%, and 69.8% for prediction of 2 years and 5 years. Internal validation using 2,000 bootstrap samples resulted in a C-statistic of 67.0% and
an
67.4%, indicating that the discrimination performance of the model was still good. Calibration of the model evaluated with the HL chi-square test supported the goodness of fit of the nomogram to the data. To our
M
knowledge, the present nomogram is the first that has been developed specifically to predict the probability of developing brain metastasis in NSCLC patients who have undergone curative surgery.
d
In previous studies, the occurrence rates of brain metastasis in NSCLC patients who received
te
curative surgery were 10% to 38% [10-12]. However, in our study, brain metastasis occurred in only 8.5% of
Ac ce p
patients. This lower occurrence rate was also reflected in the tumor stage results (stage I/II, 5.1%; stage III, 12.3%). It is unclear why the incidence of brain metastasis in our study was lower than that reported in previous studies, as the follow-up duration of our study was not shorter than that of previous studies. Possible explanations could be high proportions of stage I (50.2%) disease and squamous histology (46.1%) patients than were included in the previous studies. Similar to our study results, only 6.8% of 1,532 patients who underwent complete resection, mostly for stage I and II NSCLC, had documented first recurrences involving the brain [13]. Hubbs et al. reported diagnosis of overall brain metastasis in 6.2% of patients among 975 stage I/II NSCLC patients after surgery. The occurrence rate of isolated brain metastasis was 2.05%. Median follow-up duration was 33 months. The median time from surgery to development of brain metastasis was 10 months [14]. In contrast, Carolan et al. reported that overall brain metastasis occurred in 7
Page 7 of 20
34.9% of patients with stage III NSCLC after combined modality therapy. Metastasis to the brain as the first relapse site was observed in 18.1% of patients. The median time between diagnosis and first failure (brain metastasis) was 11.7 months [15]. Risk factors for development of brain metastasis identified in our study include adenocarcinoma histology, smoking, high pT stage, and high pN stage. Smoking was not a significant risk factor in the
ip t
univariable analysis; however it became significant when other risk factors were adjusted in the
cr
multivariable analysis. Non-squamous histology (especially, adenocarcinoma) has consistently been reported to be a risk factor for brain metastasis. Furthermore, higher T and N stage and younger age (<60 years) have
us
also been shown to be associated with an increased risk of developing brain metastasis [13, 14, 16-19]. However, age failed to reach statistical significance in our data set. In previously published nomograms, the
an
range of variables considered is usually determined based on data availability and clinical evidence, rather than solely on statistical significance [20]. However, since the inclusion of this variable did not alter the
M
development and the performance of the prediction model, we excluded age in the final model. Based on these risk factors, we chose to use parameters that were objective and easily measurable to facilitate accurate
d
prediction of brain metastasis by clinicians.
te
The prognosis of NSCLC patients without metastasis has improved substantially in the past decade.
Ac ce p
Optimized loco-regional control and adjuvant chemotherapy have contributed to improved survival [21, 22]. However, these advances do not seem to alter brain relapse rates [23-25]. Despite treatment of brain metastasis by radiation therapy, surgery, or medical therapy, prognosis of NSCLC patients with brain metastasis is poor, with a median survival of 3-6 months [26-29]. Prophylactic cranial irradiation (PCI) has been discussed as a strategy to reduce the risk of brain metastases. A number of trials have shown that PCI is effective at reducing brain metastases and improving survival in small cell lung cancer (SCLC) patients [30, 31]. However, although PCI in patients with locally-advanced NSCLC significantly decreases the risk of brain metastasis, there is no survival benefit of PCI [32, 33]. Additionally, whole brain radiation therapy has the potential risk of neurocognitive complications [34, 35]. By identifying the subgroup of patients at high risk for developing brain metastasis, the nomogram we developed can potentially increase the benefits and 8
Page 8 of 20
decrease the risks of PCI. NCCN guidelines for follow-up of NSCLC include history taking, physical examination, and chest computed tomography (CT) every 6-12 months for 2 years, then annually. In contrast, diagnostic tests for brain, such as brain CT or magnetic resonance imaging (MRI), are performed only when there is a clinical suspicion of brain metastasis. Yokoi et al. performed frequent follow-up brain CT in patients following
ip t
surgical resection for NSCLC. They noted a longer median survival in the asymptomatic group than the
cr
symptomatic group [36]. Nishikawa et al. evaluated the optimal brain MRI follow-up interval for NSCLC patients to detect radiosurgically manageable brain metastases. They suggested that early detection of brain
us
metastases by biannual MRI follow-up may decrease the need for invasive treatment of NSCLC patients [37]. Additionally, a randomized trial showed improved survival for patients with surgically treated brain
an
metastases than those with palliatively-treated brain metastases [29]. Therefore, close observation with regular brain radiologic studies could lead to early detection of brain metastases before the patient becomes
M
symptomatic, and prognosis of brain metastasis may be improved by early detection and treatment. Our nomogram could be used to establish an optimal screening strategy for brain relapse.
d
Our study had some limitations. First, we included patients from a single center. In contrast, models
te
derived from multicenter or population-based cohorts are more likely to be generalizable. However, these
Ac ce p
latter types of studies may also be hampered by the lack of consistent availability of detailed data elements that may improve prognostic accuracy. Second, we validated the nomogram using bootstrap resampling. Although bootstrapping is a sample reuse method that prevents over-interpretation of current data, this method does not ensure external applicability [4]. Third, we classified TNM stage according to the UICC/AJCC 6th TNM staging system. Since 2009, however, the UICC/AJCC 7thTNM staging system has been used [38]. In our study, T stages were divided into T1 and T2-4. Fortunately, there was no change in definition of the T1 stage between the two versions of the UICC/AJCC TNM staging system. Definition of the N stage also remained the same, therefore this is unlikely to have affected our results. In conclusion, the nomogram presented here integrates clinicopathological variables to provide an individual risk estimate of developing metastasis to the brain as the first relapse site in patients with NSCLC 9
Page 9 of 20
who have undergone curative surgery. Surveillance programs or preventive treatment strategies for brain metastasis could potentially be established using this nomogram, and it could aid in identification of patients with similar prognoses.
Acknowledgement: This work was supported by a grant from the National Cancer Center, South Korea
ip t
(grant number. 1210060).
Ac ce p
te
d
M
an
us
cr
Disclosure: The authors have no conflicts of interest to declare.
10
Page 10 of 20
References 1.
Siegel R, Ward E, Brawley O, Jemal A. Cancer statistics, 2011. CA: A Cancer Journal for Clinicians 2011.
2.
Ferlay J, Parkin D, Curado M, Bray F, Edwards B, Shin H. Cancer incidence in five continents, volumes I to IX: IARC CancerBase no. 9 [Internet]. Lyon (France): International Agency for Research on Cancer; 2010. In, 2011. Calman L, Beaver K, Hind D, Lorigan P, Roberts C, Lloyd-Jones M. Survival Benefits from Follow-Up of
ip t
3.
Patients with Lung Cancer: A Systematic Review and Meta-Analysis. Journal of Thoracic Oncology 2011;6: 1993.
Iasonos A, Schrag D, Raj GV, Panageas KS. How to build and interpret a nomogram for cancer prognosis. Journal of Clinical Oncology 2008;26: 1364-1370.
5.
Goldstraw P, Crowley JJ. The International Association for the Study of Lung Cancer international staging
us
project on lung cancer. Journal of Thoracic Oncology 2006;1: 281. 6.
cr
4.
Harrell F, Lee KL, Mark DB. Tutorial in biostatistics multivariable prognostic models: issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors. Statistics in medicine
7.
an
1996;15: 361-387.
Harrell FE, Califf RM, Pryor DB, Lee KL, Rosati RA. Evaluating the yield of medical tests. Jama 1982;247: 2543-2546.
Pencina MJ, D'Agostino RB. Overall C as a measure of discrimination in survival analysis: model specific
M
8.
population value and confidence interval estimation. Statistics in medicine 2004;23: 2109-2123. 9.
Royston P, Altman DG. External validation of a Cox prognostic model: principles and methods. BMC Douillard JY, Rosell R, De Lena M, Carpagnano F, Ramlau R, Gonzáles-Larriba JL, Grodzki T, Pereira JR, Le
te
10.
d
medical research methodology 2013;13: 33.
Groumellec A, Lorusso V. Adjuvant vinorelbine plus cisplatin versus observation in patients with completely resected stage IB–IIIA non-small-cell lung cancer (Adjuvant Navelbine International Trialist 11.
Ac ce p
Association [ANITA]): a randomised controlled trial. The lancet oncology 2006;7: 719-727. Gilbert S, Reid KR, Lam MY, Petsikas D. Who should follow up lung cancer patients after operation? The Annals of thoracic surgery 2000;69: 1696-1700. 12.
Yano T, Yokoyama H, Inoue T, Asoh H, Tayama K, Takai E, Ichinose Y. The first site of recurrence after complete resection in non-small-cell carcinoma of the lung. Comparison between pN0 disease and pN2 disease. The Journal of thoracic and cardiovascular surgery 1994;108: 680.
13.
Figlin RA, Piantadosi S, Feld R. Intracranial recurrence of carcinoma after complete surgical resection of stage I, II, and III non-small-cell lung cancer. N Engl J Med 1988;318: 1300-1305.
14.
Hubbs JL, Boyd JA, Hollis D, Chino JP, Saynak M, Kelsey CR. Factors associated with the development of brain metastases. Cancer 2010;116: 5038-5046.
15.
Carolan H, Sun AY, Bezjak A, Yi QL, Payne D, Kane G, Waldron J, Leighl N, Feld R, Burkes R. Does the incidence and outcome of brain metastases in locally advanced non-small cell lung cancer justify prophylactic cranial irradiation or early detection? Lung Cancer 2005;49: 109-115.
16.
Bajard A, Westeel V, Dubiez A, Jacoulet P, Pernet D, Dalphin J, Depierre A. Multivariate analysis of factors predictive of brain metastases in localised non-small cell lung carcinoma. Lung Cancer 2004;45: 317-323. 11
Page 11 of 20
17.
Ceresoli GL, Reni M, Chiesa G, Carretta A, Schipani S, Passoni P, Bolognesi A, Zannini P, Villa E. Brain metastases in locally advanced nonsmall cell lung carcinoma after multimodality treatment: risk factors analysis. Cancer 2002;95: 605.
18.
Feld R, Rubinstein LV, Weisenberger TH. Sites of recurrence in resected stage I non-small-cell lung cancer: a guide for future studies. Journal of Clinical Oncology 1984;2: 1352-1358.
19.
Martini N, Bains MS, Burt ME, Zakowski MF, McCormack P, Rusch VW, Ginsberg RJ. Incidence of local recurrence and second primary tumors in resected stage I lung cancer. The Journal of thoracic and
20.
ip t
cardiovascular surgery 1995;109: 120-129. Wong SL, Kattan MW, McMasters KM, Coit DG. A Nomogram That Predicts the Presence of Sentinel Node Metastasis in Melanoma With Better Discrimination Than the American Joint Committee on CancerStaging 21.
cr
System. Annals of surgical oncology 2005;12: 282-288.
Arriagada R, Auperin A, Burdett S, Higgins J, Johnson D, Le Chevalier T, Le Pechoux C, Parmar M, Pignon J, Souhami R. Adjuvant chemotherapy, with or without postoperative radiotherapy, in operable non-small-
22.
us
cell lung cancer: two meta-analyses of individual patient data. Lancet 2010;375: 1267.
Aupérin A, Le Péchoux C, Rolland E, Curran WJ, Furuse K, Fournel P, Belderbos J, Clamon G, Ulutin HC, Paulus R. Meta-analysis of concomitant versus sequential radiochemotherapy in locally advanced non–
23.
an
small-cell lung cancer. Journal of Clinical Oncology 2010;28: 2181-2190.
Andre F, Grunenwald D, Pujol JL, Girard P, Dujon A, Brouchet L, Brichon PY, Westeel V, Le Chevalier T. Patterns of relapse of N2 nonsmall-cell lung carcinoma patients treated with preoperative chemotherapy:
24.
M
should prophylactic cranial irradiation be reconsidered? Cancer 2001;91: 2394-2400. Gaspar LE, Chansky K, Albain KS, Vallieres E, Rusch V, Crowley JJ, Livingston RB, Gandara DR. Time from treatment to subsequent diagnosis of brain metastases in stage III non-small-cell lung cancer: a 25.
d
retrospective review by the Southwest Oncology Group. J Clin Oncol 2005;23: 2955-2961. Strauss GM, Herndon JE, Sherman DD, Mathisen DJ, Carey RW, Choi NC, Rege VB, Modeas C, Green MR.
te
Neoadjuvant chemotherapy and radiotherapy followed by surgery in stage IIIA non-small-cell carcinoma of the lung: report of a Cancer and Leukemia Group B phase II study. J Clin Oncol 1992;10: 1237-1244. Alexander III E, Moriarty TM, Davis RB, Wen PY, Fine HA, Black PM, Kooy HM, Loeffler JS. Stereotactic
Ac ce p
26.
radiosurgery for the definitive, noninvasive treatment of brain metastases. Journal of the National Cancer Institute 1995;87: 34-40. 27.
Dimitropoulos C, Hillas G, Nikolakopoulou S, Kostara I, Sagris K, Vlastos F, Alchanatis M. Prophylactic cranial irradiation in non-small cell lung cancer patients: who might be the candidates? Cancer management and research 2011;3: 287.
28.
Jacot W, Quantin X, Boher J, Andre F, Moreau L, Gainet M, Depierre A, Quoix E, Le Chevalier T, Pujol J. Brain metastases at the time of presentation of non-small cell lung cancer: a multi-centric AERIO analysis of prognostic factors. British journal of cancer 2001;84: 903.
29.
Patchell RA, Tibbs PA, Walsh JW, Dempsey RJ, Maruyama Y, Kryscio RJ, Markesbery WR, Macdonald JS, Young B. A randomized trial of surgery in the treatment of single metastases to the brain. New England Journal of Medicine 1990;322: 494-500.
30.
Aupérin A, Arriagada R, Pignon JP, Le Péchoux C, Gregor A, Stephens RJ, Kristjansen PEG, Johnson BE, Ueoka H, Wagner H. Prophylactic cranial irradiation for patients with small-cell lung cancer in complete remission. New England Journal of Medicine 1999;341: 476-484. 12
Page 12 of 20
31.
Meert AP, Paesmans M, Berghmans T, Martin B, Mascaux C, Vallot F, Verdebout JM, Lafitte JJ, Sculier JP. Prophylactic cranial irradiation in small cell lung cancer: a systematic review of the literature with metaanalysis. BMC cancer 2001;1: 5.
32.
Gore EM, Bae K, Wong SJ, Sun A, Bonner JA, Schild SE, Gaspar LE, Bogart JA, Werner-Wasik M, Choy H. Phase III comparison of prophylactic cranial irradiation versus observation in patients with locally advanced non–small-cell lung cancer: Primary analysis of Radiation Therapy Oncology Group Study RTOG 0214. Journal of Clinical Oncology 2011;29: 272-278. Lester JF, MacBeth FR, Coles B. Prophylactic cranial irradiation for preventing brain metastases in patients
ip t
33.
undergoing radical treatment for non–small-cell lung cancer: A Cochrane Review. International Journal of Radiation Oncology* Biology* Physics 2005;63: 690-694.
Aoyama H, Tago M, Kato N, Toyoda T, Kenjyo M, Hirota S, Shioura H, Inomata T, Kunieda E, Hayakawa K,
cr
34.
Nakagawa K, Kobashi G, Shirato H. Neurocognitive function of patients with brain metastasis who received either whole brain radiotherapy plus stereotactic radiosurgery or radiosurgery alone. Int J Radiat Oncol 35.
us
Biol Phys 2007;68: 1388-1395.
Meyers CA, Smith JA, Bezjak A, Mehta MP, Liebmann J, Illidge T, Kunkler I, Caudrelier JM, Eisenberg PD, Meerwaldt J, Siemers R, Carrie C, Gaspar LE, Curran W, Phan SC, Miller RA, Renschler MF. Neurocognitive
an
function and progression in patients with brain metastases treated with whole-brain radiation and motexafin gadolinium: results of a randomized phase III trial. J Clin Oncol 2004;22: 157-165. 36.
Yokoi K, Miyazawa N, Arai T. Brain metastasis in resected lung cancer: value of intensive follow-up with
37.
M
computed tomography. Ann Thorac Surg 1996;61: 546-550; discussion 551. Nishikawa T, Ueba T, Kawashima M, Kajiwara M, Iwata R, Kato M, Miyamatsu N, Yamashita K. Early detection of metachronous brain metastases by biannual brain MRI follow-up may provide patients with 770-774.
te
Goldstraw P, Crowley J, Chansky K, Giroux DJ, Groome PA, Rami-Porta R, Postmus PE, Rusch V, Sobin L. 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. Journal of Thoracic
Ac ce p
38.
d
non-small cell lung cancer with more opportunities to have radiosurgery. Clin Neurol Neurosurg 2010;112:
Oncology 2007;2: 706.
13
Page 13 of 20
Figure legend Figure 1. Brain metastasis-free survival according to clinicopathological factors. Figure 2. ROC curves (A) and calibration plots (B,C) for nomogram predicting brain metastasis as the first relapse within 2 and 5 years after curative resection
ip t
Figure 3. Nomogram predicting brain metastasis as the first relapse within 2 and 5 years after curative
Ac ce p
te
d
M
an
us
cr
resection.
14
Page 14 of 20
Tables Table 1. Patients’ demographic and clinical characteristics Total (N=1218)
Characteristics
Pathologic stage
0.169$ 0.462
500
(44.2)
42
(48.3)
> 60 years
676
(55.5)
631
(55.8)
45
(51.7)
Male
905
(74.3)
842
(74.4)
63
(72.4)
Female
313
(25.7)
289
(25.6)
24
(27.6)
Adenocarcinoma
553
(45.4)
493
(43.6)
60
(69.0) <0.001a
Squamous cell carcinoma
562
(46.1)
545
(48.2)
17
(19.5)
Others
103
(8.5)
93
(8.2)
10
(11.5)
BAC
32
(2.6)
31
(2.7)
1
(1.2)
Sarcomatoid carcinoma
27
(2.2)
23
(2.1)
4
(4.6)
Large cell NEC
36
(3.0)
31
(2.7)
5
(5.7)
8
(0.7)
0
(0.0)
361
(31.9)
28
(32.2)
770
(68.1)
59
(67.8)
ip t
(44.5)
0.676
cr
us
Undifferentiated or Pleomorphic carcinoma Never-smoker
8
(0.7)
389
(31.9)
Ever-smoker
829
(68.1)
IA
222
(18.2)
213
(18.8)
9
IB
389
(31.9)
373
(33.0)
16
(18.4)
IIA
42
(3.5)
40
(3.5)
2
(2.3)
IIB
214
(17.6)
197
(17.4)
17
(19.5)
0.959
(10.4) <0.001b
297
(24.4)
262
(23.2)
35
(40.2)
54
(4.4)
46
(4.1)
8
(9.2)
312
(25.6)
298
(26.3)
14
(16.1)
727
(59.7)
672
(59.4)
55
(63.2)
3
131
(10.8)
118
(10.5)
13
(14.9)
4
48
(3.9)
43
(3.8)
5
(5.8)
0
686
(56.3)
657
(58.1)
29
(33.3) <0.001d
1
257
(21.1)
237
(21.0)
20
(23.0)
2
268
(22.0)
233
(20.6)
35
(40.2)
3
7
(0.6)
4
(0.3)
3
(3.5)
798
(65.5)
749
(66.2)
49
(56.3)
420
(34.5)
382
(33.8)
38
(43.7)
1
Ac ce p
2
No
Yes $ Wilcoxon rank-sum test, Other tests are chi-square test. a
(%)
542
IIIB
Adjuvant Chemotherapy
N
61 (32-74)
≤ 60 years
IIIA
pT stage
(%)
Event (N=87) P-value
an
Smoking status
N
62 (24-87)
M
Histology
(%)
62 (24-87)
d
Gender
pN stage
N
median (min-max)
te
Age
No Event (N=1131)
0.035c
0.061
Histology: ADC vs. Non-ADC, bPathologic stage: I-II vs. III, cpT stage T1 vs. T2-4, dpN stage: N0-1 vs. N2-3
Abbreviations: BAC, bronchoalveolar carcinoma; NEC, neuroendocrine carcinoma, ADC, adenocarcinoma
15
Page 15 of 20
Table 2. Univariable and multivariable Cox regression analyses estimating the risk factors of time to brain metastasis as a first relapse site after curative resection
2.26 1.69 1.87 1.18
Yes vs. No
3.46
0.007 0.032 0.033 0.720
1.25 1.05 1.05 0.48
4.08 2.73 3.34 2.93
0.019
1.22
9.80
Ac ce p
te
d
M
an
Abbreviations: ADC, adenocarcinoma; CI, confidence interval
95% CI
ip t
Female vs. Male ADC vs. non-ADC Ever vs. Never T2-4 vs. T1 N2-3 vs. N0-1 Yes vs. No
P-value
cr
Age (year) Gender Histology Smoking status pT stage pN stage Adjuvant Chemotherapy Interaction between ADC and N2-3
Multivariable analysis Hazard ratio
us
Univariable analysis Hazard P-value 95% CI ratio 0.98 0.156 0.96 1.01 1.10 0.692 0.69 1.76 2.85 <0.001 1.81 4.49 1.02 0.944 0.65 1.59 2.01 0.017 1.14 3.56 3.19 <0.001 2.09 4.88 1.59 0.034 1.04 2.43
Clinicopathological factors
16
Page 16 of 20
Brain metastasis was higher in patients with adenocarcinoma histology. After surgery, high pT and pN stage were also risk factors for brain metastasis.
Ac ce p
te
d
M
an
us
cr
ip t
The nomogram provides the individual risk estimate of brain metastasis.
17
Page 17 of 20
Ac
ce
pt
ed
M
an
us
cr
i
Figure1
Page 18 of 20
Ac
ce
pt
ed
M
an
us
cr
i
Figure2
Page 19 of 20
Ac
ce
pt
ed
M
an
us
cr
i
Figure3
Page 20 of 20