Image-based Multilevel Subdivision of M1 Category in TNM Staging System for Metastatic Nasopharyngeal Carcinoma 1

Transcription

1 This copy is for personal use only. To order printed copies, contact Lujun Shen, MD Wang Li, MD Siyang Wang, MD Guofeng Xie, MD Qi Zeng, MD Chen Chen, MD Feng Shi, MD Ying Zhang, MD Ming Wu, MD Wanhong Shu, MD Changchuan Pan, MD Yunfei Xia, MD Peihong Wu, MD Image-based Multilevel Subdivision of M1 Category in TNM Staging System for Metastatic Nasopharyngeal Carcinoma 1 Purpose: Materials and Methods: To establish an image-based M1 category subdivision system for personalized prognosis prediction and treatment planning in patients with metastatic nasopharyngeal carcinoma (NPC). A total of 1172 patients with metachronous metastasic NPC were retrospectively enrolled (the dataset is from Sun Yat-sen University Cancer Center for derivation, and the combined datasets are from Guangzhou Medical University Cancer Center and the Fifth Affiliated Hospital of Sun Yat-sen University for validation). The Ethics Committee of the three centers approved this study. A general subdivision system of the M1 category was established on the basis of the most influential metastatic features for overall survival (OS). The following multilevel subdivision system for precise subdivision of the M1 category was designed: M [number of locations]-location [number of lesions], with B indicating bone, L indicating the lung, H indicating the liver, and N indicating a node. The correlation of the M1 subdivisions with OS was determined with Cox regression. The best treatment response was assessed with Response Evaluation Criteria in Solid Tumors 1.1 guidelines and modified Response Evaluation Criteria in Solid Tumors criteria. Original Research n Head and Neck Imaging 1 From the Department of Medical Imaging and Interventional Radiology, Sun Yat-sen University Cancer Center, Guangzhou , People s Republic of China (L.S., W.L., Q.Z., F.S., W.S., P.W.); State Key Laboratory of Oncology in South China, Collaborative Innovation Center of Cancer Medicine, Sun Yat-sen University, Guangzhou, People s Republic of China (L.S., C.C., Y.Z., M.W., Y.X., P.W.); Department of Radiation Oncology, Fifth Affiliated Hospital of Sun Yat-sen University, Zhuhai, People s Republic of China (S.W., Q.Z., C.C., Y.X.); Department of Radiation Oncology, Cancer Center of Guangzhou Medical University, Guangzhou, People s Republic of China (G.X.); Department of Medical Oncology, Sichuan Cancer Hospital and Institute, Chengdu,, People s Republic of China (C.P.); and Zhongshan School of Medicine, Sun Yat-sen University, Guangzhou, People s Republic of China (Y.Z., M.W., W.S.). Received July 6, 2015; revision requested September 2; revision received November 21; accepted December 7; final version accepted January 12, Address correspondence to P.W. ( wuph@sysucc.org.cn). P.W. supported by National High Technology Research and Development Program of China (863 Program). L.S. and W.L. contributed equally to this work. q RSNA, 2016 Results: Conclusion: Multivariate analysis in the derivation cohort showed that the number of metastatic lesions (multiple or single), the number of metastatic locations (multiple or single), liver involvement, and bone involvement were independent prognostic factors for OS. In general, subdividing the cohort by the number of metastatic lesions and the number of metastatic locations resulted in three subcategories of differential OS: M1a, a single lesion in a single organ or location; M1b, multiple lesions in a single organ or location; and M1c, metastases in multiple locations (for M1b vs M1a, hazard ratio [HR] = 2.28, 95% confidence interval [CI]: 1.71, 3.05; for M1c vs M1a, HR = 3.65, 95% CI: 2.75, 4.85); these subdivisions were externally validated. The multilevel subdivision system could be further used to discriminate among subgroups of differential OS under the M1b subcategory. Findings from analysis of multilevel subgroups suggested that patients with a single metastatic lesion (M1-B 1, M1- L 1, M1-H 1, M1-N 1 ) or two lesions in the liver only (M1-H 2 ) had high rates of complete response (CR) or complete surgical resection (CSR) and 3-year OS after treatment (CR plus CSR rates >30%, and 3-year OS rates >50%); there were high 3-year OS rates (.50%) in patients with stage M1-B 2, M1-L 2, or M1-H 3 disease but relatively low rates of CR or CSR. Use of the multilevel M1 subdivision system in patients with NPC could facilitate more precise prognosis prediction and better identification of patients who will respond well to treatment than the conventional subdivision strategy. q RSNA, 2016 Online supplemental material is available for this article. Radiology: Volume 280: Number 3 September 2016 n radiology.rsna.org 805

2 Cancer patients with distant metastasis constitute a very heterogeneous group, of which postmetastatic survival ranges from potentially curable to incurable (1,2). With advances in imaging and Advances in Knowledge nn The numbers of metastatic lesions (multiple or single) and locations (multiple or single) were the most influential metastatic feature for predicting overall survival (OS) in patients with metachronous metastasic nasopharyngeal carcinoma (NPC); multiple lesions in a single organ or location versus a single lesion had a hazard ratio (HR) of 2.28 and a 95% confidence interval (CI) of 1.71 and 3.05, and metastases in multiple locations versus a single lesion had a HR of 3.65 and a 95% CI of 2.75 and nn When multiple metastases were present in a single location, the use of a multilevel nomenclature system (M [number of locations]- Location [number of lesions], with B indicating bone, L indicating the lung, H indicating the liver, and N indicating lymph nodes) demonstrated that patients in the M1-B m subgroup had poorer OS than did those in the M1-B 2 (HR = 2.28, 95% CI: 1.33, 3.91) and M1-B 3 (HR = 2.08, 95% CI: 1.24, 3.49) subgroups, and patients in the M1-H m subgroup had poorer OS than did those in the M1-H 2 subgroup (HR = 3.74, 95% CI: 1.02, 13.66). nn Patients with a single metastatic lesion or two metastatic lesions in the liver had a high chance of achieving complete response (CR) or complete surgical resection (CSR) and high 3-year OS rates after treatment (CR plus CSR.30% and 3-year OS rates.50%), making them ideal candidates for aggressive treatment. image-guided treatment, there is a growing need to develop appropriate nomenclature to characterize metastatic disease for better prognosis prediction and treatment planning (3 5). At present, the tumor node metastasis (TNM) classification, which describes the anatomic extent of cancer, is the most well-rounded and widely used classification system (6). However, of the 46 types of malignancy documented in the TNM Cancer Staging Manual (7th edition), only six have developed subcategories under the M1 category that roughly divide the cohort into two or three subgroups and that are not sophisticated enough (7). Nasopharyngeal carcinoma (NPC) is an ideal malignancy worth exploring for a new nomenclature system. Past studies on NPC showed that lung metastasis alone is a favorable prognostic factor compared with bone or liver metastasis alone (8 10). A single metastatic lesion in a single location (an organ or site) was reported to correlate with prolonged survival compared with multiple metastatic lesions in a single location or metastases in multiple locations (10 12). Recent years have seen increased interest in the use of local therapy for metastases, and a series of reports were published that showed that even long-term survival could be achieved in select patients with NPC and oligometastasis with combined systemic and local therapies (13 16). A nomenclature system that could take all the important anatomic factors into consideration is desirable. Besides, because cancer metastasis covers a wide spectrum of conditions, it is attractive to explore the theoretical margin of Implication for Patient Care nn Use of the image-based multilevel M1 subdivision system and the SMART strategy precisely subdividing the M1 category and assessing the response to treatment in patients with metastatic NPC could serve as a powerful tool in selecting patients who would benefit from aggressive and image-guided treatment. metastatic extent between patients who would benefit from aggressive management and those who would not with the designed system (17,18). Currently, in the seventh edition of the TNM classification of NPC, the metastatic stage (M1) is still a catchall classification. Our previous study on synchronously metastatic NPC proposed a trichotomous M1 subdivision system; however, further exploration was impeded by the relatively small sample size and the limited metastatic features of independent prognostic value identified (19). In contrast, patients with metachronous metastasis account for about 65% of all patients with metastatic NPC; normally, their primary tumors are well controlled, and the extent of metastatic disease may best correlate with survival (10). Therefore, in this current study, we set out to establish a new nomenclature system to subdivide the M1 category of NPC in patients with metachronous Published online before print /radiol Content codes: Radiology 2016; 280: Abbreviations: CI = confidence interval CR = complete response CSR = complete surgical resection GMUCC = Guangzhou Medical University Cancer Center HR = hazard ratio KPS = Karnofsky performance score NPC = nasopharyngeal carcinoma OS = overall survival SYSUCC = Sun Yat-sen University Cancer Center SYSU = Fifth Affiliated Hospital of Sun Yat-sen University UICC = Union for International Cancer Control Author contributions: Guarantors of integrity of entire study, L.S., W.L., Q.Z., F.S., M.W., W.S., P.W.; study concepts/study design or data acquisition or data analysis/interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; approval of final version of submitted manuscript, all authors; agrees to ensure any questions related to the work are appropriately resolved, all authors; literature research, L.S., W.L., S.W., G.X., C.C., F.S., Y.Z., W.S., C.P., P.W.; clinical studies, L.S., W.L., S.W., G.X., Q.Z., C.C., F.S., Y.Z., M.W., W.S., Y.X., P.W.; experimental studies, L.S., F.S., W.S., P.W.; statistical analysis, L.S., W.L., S.W., G.X., Q.Z., C.C., F.S., M.W., W.S., C.P., Y.X., P.W.; and manuscript editing, L.S., W.L., S.W., G.X., Q.Z., C.C., F.S., Y.Z., W.S., Y.X., P.W. Conflicts of interest are listed at the end of this article. 806 radiology.rsna.org n Radiology: Volume 280: Number 3 September 2016

3 metastasis and assess its value in prognosis prediction and patient identification. Materials and Methods Patients This multicenter retrospective study consisted of consecutive series of 1242 patients with metachronous metastasic NPC who were treated at Sun Yat-sen University Cancer Center (SYSUCC; n = 963), the Fifth Affiliated Hospital of Sun Yat-sen University (SYSU; n = 107), and Guangzhou Medical University Cancer Center (GMUCC; n = 172) from 2000 to The Hospital Ethics Committee of the three medical centers approved this study. The following inclusion criteria were used: (a) histologically confirmed NPC and (b) a diagnosis of distant metastasis after primary radiation therapy on the basis of histologic and/or imaging findings and subsequent follow-up. The following exclusion criteria were used: (a) lack of clinical data from computed tomography (CT) or magnetic resonance imaging of the head and neck regions, radiography or CT of the chest, ultrasonography or CT of the abdomen, or whole-body bone scanning at the time metastasis was diagnosed (41 patients from SYSUCC, nine patients from SYSU, and 12 patients from GMUCC were excluded) and (b) other malignancies (six from SYSUCC and two from GMUCC were excluded). A total of 1172 patients were included. Of the SYSUCC dataset, 348 patients were included in our previous study, which was our initial attempt at subdividing the M1 category of NPC (10). In the current study, multicentric data were analyzed, and two subdivision systems (general and multilevel) were proposed and compared in their prognostic value specifically for patients with metachronous metastasic NPC. Also included in previous studies that aimed to investigate the possible subdivision systems in NPC with single-organ metastasis were 195 patients with liver metastasis (27) and 136 with bone-only metastasis (16). Ten patients with metachronous liver metastasis (13) and 10 patients with metachronous lung metastasis (15) who did not undergo radiofrequency ablation were included in previous studies to evaluate the safety and efficacy of radiofrequency ablation. M1 Subdivision Workup and Variables The multidisciplinary team established to assess disease and the metastatic features of NPC was similar to that used in our previous study (16,19,20). A general subdivision system that used trichotomy and a multilevel subdivision system of the M1 category were proposed. In establishing the general subdivision system, the dataset from SYSUCC (n = 916) was used for derivation, and the combined datasets from SYSU (n = 98) and GMUCC (n = 158) were used for validation. In assessing the multilevel subdivision system, the dataset from SYSUCC (the derivation cohort) was used. The T and N descriptors of the primary disease of NPC patients were staged according to the seventh edition of the Union for International Cancer Control (UICC) staging system (21). The other variables assessed included sex, age, Karnofsky performance score (KPS), body mass index, disease-free interval, and local recurrence. Treatment, Follow-up, and Response Assessment A total of 1011 patients (86.3%) from the three medical centers underwent palliative chemotherapy as systemic treatment after metastasis (SYS- UCC, 86.8%; SYSU, 81.6%; GMUCC, 86.1%). The primary chemotherapy used was exclusively plastin-based, with cisplatin used in combination with one or two of the following drugs: fluorouracil, cyclophosphamide, vincristine, paclitaxel, and bleomycin for four to six cycles. Treatment was discontinued as a request by patients or a result of unacceptable drug toxicity. Local therapies such as surgery, radiation therapy, interventional embolization, and ablative therapy served as options for patients with limited or localized metastatic lesions. Patients were routinely followed up every two cycles during systemic chemotherapy and every 2 3 months until death. The primary outcome was overall survival (OS), which was defined as the time from diagnosis of distant metastasis to death by any cause. The imaging data of patients admitted to SYSUCC during follow-up was routinely analyzed by radiologists (P.W., Y.X., W.L., and L. S., with 30, 22, 10, and 4 years of experience in radiology, respectively) to determine the best treatment response according to the Response Evaluation Criteria in Solid Tumors (RECIST) guideline, version 1.1 (22). For patients who underwent local-regional therapies, the modified RECIST criteria were applied (23,24). Surgical resection of oligometastasis was recorded, and complete surgical resection (CSR) was defined as the removal of all image-confirmed metastatic lesions, with no microscopic disease seen at resection margins (25). Statistical Analysis Statistical comparisons of categoric data were carried out with the Pearson x 2 test. Rates of OS were estimated by means of the Kaplan-Meier method and compared between the subgroups with the use of the log-rank test. To determine the independent prognostic factors for OS in the derivation cohort, Cox regression model in backward conditioned method was utilized, with the following covariates: age; UICC T stage; UICC N stage; KPS; body mass index; disease-free interval; number of metastatic lesions (ie, multiple or single); number of metastatic locations (ie, multiple or single); liver, bone, and lymph node involvement; and local recurrence. The general subdivision system was established on the basis of the most influential metastatic features for OS in the derivation cohort; acceptance of the system required that a significantly differential OS be observed between each two subcategories in multivariate analysis by external validation. To assess whether a multilevel subdivision system had higher efficacy in prognosis prediction than the Radiology: Volume 280: Number 3 September 2016 n radiology.rsna.org 807

4 general subdivision system, we evaluated the survival between subgroups of multilevel subdivisions under each general subdivisions through a multiple-adjusted Cox model, with covariates including age, UICC N stage, KPS, and local recurrence. To identify the specific patient subgroups that would benefit from aggressive treatment under the multilevel subdivision system, we established the SMART strategy, which stands for subdividing the M1 category precisely and assessing the response of treatment and which took of both survival and treatment response information into consideration by setting cutoff values. A two-tailed P value of less than.05 was considered to indicate a significant difference. Statistical analysis was performed with SPSS version 20.0 software (IBM, Armonk, NY). Results Value of the General Subdivision of the M1 Category by the Number of Metastatic Locations and Lesions The derivation cohort consisted of 916 patients with a median age of 44 years (range, years), and the two validation cohorts (SYSU and GMUCC) consisted of 158 and 98 patients, respectively, with a median age of 43 years (range, years) and 48 years (range, years), respectively. In patients with NPC, the median follow-up after metastasis was 19.8 months for the derivation cohort and 18.2 months (SYSU) and 16.6 months (GMUCC) for the two validation cohorts. The baseline characteristics of the derivation cohort and validation cohorts are summarized in Table 1 and Table E1 (online). Among the patients in the derivation cohort, the most frequently involved metastatic locations were bone (51.0%), lung (49.1%), liver (32.2%), and distant lymph nodes (31.9%); isolated metastasis was most common among patients with bone involvement (45.8%) and rare in those with distant lymph node involvement (4.2%). Multiple metastatic lesions were detected Table 1 Baseline Characteristics of the Derivation Cohort Characteristic Total more frequently than were single lesions for all metastatic organs and locations (Table E2 [online]). The prognostic value of each specific metastatic location and the number of metastatic locations and lesions in the derivation cohort were evaluated to identify the optimal M1 subdividing strategy (Table 2). In the Cox regression model, together with the presence of local recurrence, age, UICC N stage, and KPS, the number of metastatic lesions (ie, multiple or single) and locations (ie, General M1 Subcategory M1a M1b M1c P Value Age (y).571, (50.8) 68 (54.4) 167 (49.0) 230 (51.1) (49.2) 57 (45.6) 174 (51.0) 220 (48.9) Sex.503 Male 754 (82.3) 107 (85.6) 276 (80.9) 371 (82.4) Female 162 (17.7) 18 (14.4) 65 (19.1) 79 (17.6) UICC T stage.418 T1 54 (5.9) 3 (2.4) 25 (7.3) 26 (5.8) T2 283 (30.9) 45 (36.0) 104 (30.5) 134 (29.8) T3 368 (40.2) 49 (39.2) 130 (38.1) 189 (42.0) T4 211 (23.0) 28 (22.4) 82 (24.0) 101 (22.4) UICC N stage.053 N0 159 (17.4) 32 (25.6) 61 (17.9) 66 (14.7) N1 356 (38.9) 46 (36.8) 126 (37.0) 184 (40.9) N2 296 (32.3) 35 (28.0) 121 (35.5) 140 (31.1) N3 105 (11.5) 12 (9.6) 33 (9.7) 60 (13.3) Histologic type.854* WHO I/II 66 (7.2) 9 (7.2) 24 (7.0) 33 (7.3) WHO III 842 (91.9) 114 (91.2) 315 (92.4) 413 (91.8) Unknown 8 (0.9) 2 (1.8) 2 (0.6) 4 (0.9) KPS (91.9) 122 (97.6) 312 (91.5) 408 (90.7),80 74 (8.1) 3 (2.4) 29 (8.5) 42 (9.3) BMI (kg/m 2 ) (75.9) 102 (81.6) 247 (72.4) 346 (76.9), (24.1) 23 (18.4) 94 (27.6) 104 (23.1) Recurrence at primary site.162 No 718 (78.4) 99 (80.7) 256 (75.1) 363 (80.7) Yes 198 (21.6) 26 (19.3) 85 (24.9) 87 (19.3) DFI (m) (49.6) 73 (58.4) 145 (42.5) 236 (52.4), (50.4) 52 (41.6) 196 (57.5) 214 (47.6) Note. Data are numbers of patients, and data in parentheses are percentages. For the total number of patients, n = 916; for M1a, n = 125; for M1b, n = 341; and for M1c, n = 450. BMI = body mass index, DFI = disease-free interval, WHO = World Health Organization. * Fisher exact test was used when more than 20% of cells had an expected frequency of,5. multiple or single) and the presence of liver and bone involvement were identified as independent prognostic factors for OS (Table 3). On the basis of the number of metastatic lesions and locations, we generally subdivided the M1 category into the following three subcategories: M1a, in which a single lesion is confined within a single organ or location; M1b, in which multiple lesions are found within a single organ or location; and M1c, in which multiple metastases are found in multiple locations (Table 1, 808 radiology.rsna.org n Radiology: Volume 280: Number 3 September 2016

5 Table 2 Univariate Analysis according to Metastatic Features in the Derivation Cohort Variable No. of Patients Median OS (mo) P Value Isolated vs Multiple Metastases No. of metastatic locations,.001 Single Multiple No. of metastatic lesions,.001 Single Multiple Location Involved Bone,.001 Yes No Lung.104 Yes No Liver,.001 Yes No Distant LN,.001 Yes No No. of Lesions and Locations Isolated bone metastases,.001 Single Multiple Isolated lung metastases.001 Single Multiple Isolated liver metastases.225 Single Multiple Isolated distant LN metastases.017 Single Multiple Total isolated metastases*,.001 Single Multiple Note. LN = lymph node. * One patient with NPC and a single pleural metastasis was included. Fig 1, A). After adjusting for age, UICC N stage, KPS, and local recurrence, multivariate analysis suggested that the M1 subcategories significantly differed in terms of OS (M1b vs M1a: hazard ration [HR] = 2.28, 95% confidence interval [CI]: 1.71, 3.05; M1c vs M1a: HR = 3.65, 95% CI: 2.75, 4.85; M1c vs M1b: HR = 1.61, 95% CI: 1.35, 1.92). For external validation, there were no significant differences in the distribution of the M1 subcategories between the derivation cohort and the two validation cohorts (P =.196), whereas the validation cohorts showed longer OS times than did the derivation cohort (SYSU and GMUCC: median OS = 29.6 months, 95% CI: 24.2, 35.1; SYSUCC: median OS = 25.9 months, 95% CI: 23.8, 27.9). Nevertheless, the proposed M1 subcategories significantly discriminated among the three prognostic groups in the validation dataset with both univariate and multivariate analysis (Fig 1, B; Table E3 [online]). Multilevel Subdivision versus General Subdivision in Prognosis Prediction Because the number of metastatic lesions, the number of metastatic locations, and involvement of a specific location (eg, the liver and bone) were independent prognostic factors, we designed the following multilevel nomenclature system, which incorporates all these related anatomic factors (Table 4): M [number of locations]- Location (1). For multilocation metastases, a simplified expression was used: M [number of locations]-[number of total lesions] (2). With this nomenclature applied in the derivation cohort, we further divided the patients in the M1a, M1b, and M1c subcategories into four, 10, and six subgroups, respectively (Fig 2). Two patients with an uncommon metastatic location only (denoted as M1-O 1-m ) were not included in the analysis because of the rarity of this occurrence. In the M1b category, multivariate analysis suggested that the M1- B m subgroup had significantly lower OS rates than did the M1-B 2 (HR = 2.28, 95% CI: 1.33, 3.91) and M1- B 3 (HR = 2.08, 95% CI: 1.24, 3.49) subgroups. The M1-H m subgroup had lower OS rates than did the M1-H 2 subgroup (HR = 3.74, 95% CI: 1.02, 13.66). No significant differences in OS were found between subgroups in either the M1a or M1c category. Treatment Response and the SMART Strategy In the derivation cohort, 80 patients (8.7%) achieved a complete response (CR) after treatment, 554 (60.5%) did not achieve a complete response, and 282 (30.8%) were not evaluated. Of the patients who were not evaluated, 33 (11.7%) underwent CSR of the metastases. Univariate analysis showed that, in the entire cohort, patients who achieved a CR during the course of treatment had improved OS compared with those with a noncomplete response (P,.001). Among patients in the M1a subcategory, Radiology: Volume 280: Number 3 September 2016 n radiology.rsna.org 809

6 Table 3 Independent Prognostic Factors from Multivariate Analysis for OS in the Derivation Cohort Factor HR P Value Age ( 44 y vs,44 y) 1.19 (1.01, 1.40).041 UICC stage T (T3 T4 vs T1 T2) 1.18 (0.99, 1.40).065 N (N2 N3 vs N0 N1) 1.23 (1.04, 1.45).015 KPS (,80 vs 80) 1.87 (1.39, 2.50),.001 No. of metastatic lesions (multiple or single) 2.26 (1.69, 3.03),.001 No. of metastatic locations (multiple or single) 1.45 (1.20, 1.76),.001 Liver involvement (yes or no) 1.26 (1.04, 1.53).018 Bone involvement (yes or no) 1.39 (1.18, 1.65),.001 Local recurrence (yes or no) 1.36 (1.12, 1.66).002 Note. Data in parentheses are 95% CIs. The covariates included for Cox regression model were age; sex; KPS; body mass index; UICC T stage; UICC N stage; disease-free interval; number of metastatic lesions; recurrence at the primary site; number of metastatic locations; and bone, liver, lung, and lymph node involvement. CSR was correlated with favorable OS compared with those with a noncomplete response(p,.001) (Fig 1, C and D). Multiple-adjusted analysis showed that, in all M1 subcategories, a CR was independently associated with improved OS compared with a noncomplete response (Table E4 [online]). We defined the target patients with metastatic NPC as those with high rates of long-term survival and who are likely to achieve a CR or CSR through treatment. The cutoff values (CR plus CSR.30% and 3-year OS rates.50%) were set to include the subdivisions characterized by a single metastasis that was conventionally thought to respond well to aggressive treatment. The results showed that, apart from patients with a single metastatic lesion (M1-B 1, M1-L 1, M1-H 1, and M1-N 1 ), those with two metastatic lesions in the liver only (M1-H 2 ) were also ideal candidates. Patients with stage M1-B 2, M1-L 2, and M1-H 3 disease had high 3-year OS rates but relatively low rates of CR or CSR (CR plus CSR.17%, 3-year OS rates.50%) and were potential target candidates. For patients with more than three metastatic lesions in a specific location (M1-B m and M1-H m ) or a metastasis in multiple locations (M2 3, M2- m, M3-m, and Mm-m), the rates of CR, CSR, and 3-year OS were low (CR,,4%; CSR, 0%; 3-year OS rates,, 30%) (Fig 2, Table E5 [online]). Discussion In this study, we focused on analyzing the prognostic impact of the extent of metastatic disease. There were several notable findings. First, the numbers of metastatic lesions (ie, multiple or single) and metastatic locations (ie, multiple or single) were the most influential metastatic features associated with OS in patients with NPC and metachronous metastasis; in general, subdividing the population by these two factors results in distinct categories regarding OS. Second, the multilevel nomenclature system of the M1 category could further discriminate subgroups with different prognostic value under the general subdivisions. Third, patients with a single metastatic lesion (M1-B 1, M1-L 1, M1-H 1, M1-N 1 ) or two metastatic lesions in the liver only (M1-H 2 ) could be ideal candidates for aggressive treatment. The prognostic impact of the number of metastatic lesions (single or multiple) has been extensively studied and deemed important both in patients with synchronous and metachronous metastatic NPC by a series of studies (10,11,19,26), while its correlation with the impact of the number of metastatic locations in patients with metachronous metastatic NPC remains undetermined. Pan et al (10) retrospectively analyzed the data from 640 patients with metachronous metastasic NPC and found that, rather than the number of metastatic locations (single or multiple), the number of metastatic lesions (single or multiple) was an independent prognostic factor for OS. However, their results should be interpreted with caution, as distant lymph node metastasis was not included in the analysis, and the time interval in which patients were enrolled is relatively long, from 1995 to In our study, patients who were admitted between 2000 and 2009 were enrolled. Multivariate analysis in the derivation cohort suggested that both the number of metastatic lesions (multiple or single) and the number of metastatic locations (multiple or single) were independent prognostic factors for OS, indicating that both factors are important for risk stratification. Our study also showed that liver and bone involvement were independent predictors for OS in patients with metastatic NPC, suggesting that the specific metastatic location involved should be considered when subdividing the M1 category. Moreover, a previous study that specifically focused on metastatic NPC in the liver only showed that a cutoff value of 3 to dichotomize the number of liver metastatic lesions (ie, more than three or three of fewer) was better than a cutoff value of 1 (multiple or single) to achieve optimal prognostic impact (27), indicating that the optimal cutoff value could be different for metastases in different locations and that greater precision can still be achieved. With accumulating evidence supporting that the tumor load closely correlates with the prognosis in patients with NPC (28,29), a sophisticated classification system that takes into account the position of every metastatic lesion is desirable. Therefore, we designed a multilevel nomenclature system for precise subdivision of the M1 category that turned out to have higher accuracy for prognosis prediction than the general subdivision strategy for patients with metastatic NPC, especially those with 810 radiology.rsna.org n Radiology: Volume 280: Number 3 September 2016

7 Figure 1 Figure 1: Kaplan-Meier plot shows OS divided, A, into three subsets according to the proposed general M1 subcategories in the derivation cohort, B, by the proposed general M1 subcategories in the validation cohort, C, by treatment response in the derivation cohort, and, D, into four subsets on the basis of treatment effect in the M1a subcategory of the derivation cohort. NE = not evaluated, non-cr = non-complete response, csr = complete surgical response. multiple metastatic lesions in a single organ or location (M1b). This system could serve as a bridge to personalized treatment under the guidance of modern imaging techniques and is recommended for patients with metachronous metastasic NPC. Recent progress in Epstein Barr virus DNA assay and positron emission tomography-ct provides powerful tools for detecting early metastases (30 32). Meanwhile, there is growing evidence that the combination of local therapy and systemic treatment could achieve a synergistic effect for patients with NPC and limited metastatic lesions. Ma et al (33) analyzed the data from 105 patients with NPC and solitary lung metastases and found that surgery or radiation therapy combined with chemotherapy yielded a significantly better local control and survival outcome (3-year OS, 56% 61%) compared with chemotherapy alone (3-year OS, 21%). (16) assessed the data from 58 patients with NPC and single bone metastases and found that the use of combined chemoradiotherapy could significantly improve OS compared with chemotherapy or radiation therapy alone. Encouraging data on the use of image-guided techniques for local-regional treatment have also been reported. Pan et al (13, 15) reported two retrospective series that used radiofrequency ablation to treat hepatic (n = 17) and pulmonary (n = 23) oligometastases in patients with NPC. The reported local control was 87% for both sites, and the median OS was 48 months for patients with liver metastases and 77 months for patients with lung metastases. These findings suggest that select groups of patients are amenable to aggressive therapy. In clinical practice, the target case selection is individualized on the basis of the number of metastatic lesions, the number of metastatic sites or organs, the feasibility and importance of complete eradication of the target lesion, and clinicians experience (34). The first two factors match the multilevel subdivision system we propose, and the third factor depends on the treatment modality used; therefore, we use CR plus CSR rates and 3-year OS rates to describe it. For this reason, we developed the SMART Radiology: Volume 280: Number 3 September 2016 n radiology.rsna.org 811

8 Table 4 Descriptions of Multilevel Nomenclature System for Subdivision of the M1 Category Nomenclature Description Example Mx Undetermined distant metastasis M1-H 1 Single location Single lesion in the liver M1-H 2 Single location Two lesions in the liver M1-H 3 Single location Three lesions in the liver M1-H m Single location More than three lesions in the liver M2-H 1, L 1 Two locations One lesion in liver and one in lung M2-H 1, L 2 Two locations One lesion in liver and two in lung M2-m* Two locations Multiple lesions M3 3* Three locations Three lesions M3-m* Three locations Multiple lesions Mm-m* Multiple locations Multiple lesions Note. The number after the letter M (1 3) stands for the number of involved locations. The subscript number (1 3) stands for the number of metastatic lesions in each location. The letter m stands for multiple and is used when the number of locations or lesions exceeds three. B = bone, H = liver, L = lung, M = metastases, N = lymph node, O = other. * A simplified expression designed to compare prognoses; a complete expression that indicates lesions in each location is recommended for individual cases. strategy, which stands for subdividing the M1 category precisely and assessing the response of treatment. The results showed that patients with an isolated metastatic lesion (M1-B 1, M1-L 1, M1-H 1, and M1-N 1 ) or two metastatic lesions in the liver only (M1-H 2 ) were ideal target candidates; patients with stage M1-B 2, M1-L 2, or M1-H 3 lesions were potential candidates; and patients with more than three metastatic lesions in a specific location (M1-B m and M1-H m ) or a metastasis in multiple locations (M2 3, M2- m, M3-m, and Mm-m) were best suited for palliative treatment. The advantage of using the SMART strategy for NPC is that it provides a powerful tool for interventional radiologists and oncologists in treatment planning by weighting the risk and benefit of treatment, especially when treatment guidelines are not Figure 2 Figure 2: Graph shows multilevel subdivisions of the M1 category, treatment response, and 3-year OS rates. Precise subdivision of the M1 category with a multilevel nomenclature system may be used to identify the subgroups with a favorable prognosis and treatment response within each general M1 subcategory. NE = not evaluated, non-cr = non-complete response, csr = complete surgical response. 812 radiology.rsna.org n Radiology: Volume 280: Number 3 September 2016

9 supported by level I evidence but rather by clinical experience instead (34,35). Our study has several limitations. First, it is a retrospective study. Second, only a small number of patients (128 [10.9%]) had pathologically confirmed metastatic disease, which could be a potential source of bias. Third, in 249 patients (27.2%) in the derivation cohort, the treatment response was categorized as not evaluated (non-csr) due to a lack of imaging data at re-examination, which may lead to underestimation of the actual therapeutic effect. Lastly, the validation of multilevel subdivisions and the SMART strategy, which stands for subdividing the M1 category precisely and assessing the response of treatment, in the validation dataset were not conducted because of the limited sample size. For these reasons, a large-scale multi-institutional prospective study is needed for validation. In conclusion, the imagebased multilevel subdivision system of the M1 category could provide more precise prognosis prediction than the conventional subdivision strategy and facilitate identification of target patients with metastatic NPC who would benefit from aggressive treatment. Disclosures of Conflicts of Interest: L.S. disclosed no relevant relationships. W.L. disclosed no relevant relationships. S.W. disclosed no relevant relationships. G.X. disclosed no relevant relationships. Q.Z. disclosed no relevant relationships. C.C. disclosed no relevant relationships. F.S. disclosed no relevant relationships. Y.Z. disclosed no relevant relationships. M.W. disclosed no relevant relationships. W.S. disclosed no relevant relationships. C.P. disclosed no relevant relationships. Y.X. disclosed no relevant relationships. P.W. disclosed no relevant relationships. References 1. Khanfir A, Frikha M, Ghorbel A, Karray H, Drira MM, Daoud J. Metastatic nasopharyngeal carcinoma: clinical study and therapeutic results of 95 cases [in French]. Cancer Radiother 2006;10(8): Teo PM, Kwan WH, Lee WY, Leung SF, Johnson PJ. Prognosticators determining survival subsequent to distant metastasis from nasopharyngeal carcinoma. Cancer 1996;77(12): Edge SB, Compton CC. The American Joint Committee on Cancer: the 7th edition of the AJCC cancer staging manual and the future of TNM. Ann Surg Oncol 2010;17(6): Wu F, Li W, Huang Z, et al. Survival analysis and recommendation of subdividing M1 stage according to metastatic status of colorectal cancer [in Chinese]. Zhonghua Yi Xue Za Zhi 2014;94(15): Sánchez de Cos Escuín J, Abal Arca J, Melchor Íñiguez R, et al. Tumor, node and metastasis classification of lung cancer- -M1a versus M1b--analysis of M descriptors and other prognostic factors. Lung Cancer 2014;84(2): Goldstraw P. New TNM classification: achievements and hurdles. Transl Lung Cancer Res 2013;2(4): Edge SB, Byrd DR, Compton CC, et al. AJCC cancer staging manual. 7th ed. New York, NY: Springer, Ong YK, Heng DM, Chung B, et al. Design of a prognostic index score for metastatic nasopharyngeal carcinoma. Eur J Cancer 2003;39(11): Hui EP, Leung SF, Au JS, et al. Lung metastasis alone in nasopharyngeal carcinoma: a relatively favorable prognostic group. A study by the Hong Kong Nasopharyngeal Carcinoma Study Group. Cancer 2004;101(2): Pan CC, Lu J, Yu JR, et al. Challenges in the modification of the M1 stage of the TNM staging system for nasopharyngeal carcinoma: A study of 1027 cases and review of the literature. Exp Ther Med 2012;4(2): Tian YM, Zeng L, Wang FH, et al. Prognostic factors in nasopharyngeal carcinoma with synchronous liver metastasis: a retrospective study for the management of treatment. Radiat Oncol 2013;8: Li JX, Huang SM, Wen BX, Lu TX. Prognostic factors on overall survival of newly diagnosed metastatic nasopharyngeal carcinoma. Asian Pac J Cancer Prev 2014;15(7): Pan C, Wu P, Yu J, et al. CT-guided radiofrequency ablation prolonged metastatic survival in patients with liver metastases from nasopharyngeal carcinoma. Int J Hyperthermia 2011;27(6): Jin Y, Cai YC, Cao Y, et al. Radiofrequency ablation combined with systemic chemotherapy in nasopharyngeal carcinoma liver metastases improves response to treatment and survival outcomes. J Surg Oncol 2012;106(3): Pan CC, Wu PH, Yu JR, et al. Comparative survival analysis in patients with pulmonary metastases from nasopharyngeal carcinoma treated with radiofrequency ablation. Eur J Radiol 2012;81(4):e473 e Shen L, Dong J, Li S, et al. M1 stage subdivision and treatment outcome of patients with bone-only metastasis of nasopharyngeal carcinoma. Oncologist 2015;20(3): Hellman S. Karnofsky Memorial Lecture. Natural history of small breast cancers. J Clin Oncol 1994;12(10): Hellman S, Weichselbaum RR. Oligometastases. J Clin Oncol 1995;13(1): Shen LJ, Wang SY, Xie GF, et al. Subdivision of M category for nasopharyngeal carcinoma with synchronous metastasis: time to expand the M categorization system. Chin J Cancer 2015;37: World Health Organization. International histological classification of tumours. 2nd ed. Geneva, Switzerland: World Health Organization, ; Berlin, Germany: Springer-Verlag, 1988 Present. 21. Lee AW, Ng WT, Chan LK, et al. The strength/weakness of the AJCC/UICC staging system (7th edition) for nasopharyngeal cancer and suggestions for future improvement. Oral Oncol 2012;48(10): Eisenhauer EA, Therasse P, Bogaerts J, et al. New response evaluation criteria in solid tumours: revised RECIST guideline (version 1.1). Eur J Cancer 2009;45(2): Jang HJ, Cho JW, Park B, Choi HC, Kim HS, Kim JH. The assessment of tumor response by measuring the single largest lesion per organ in metastatic tumors: a pooled analysis of previously reported data. J Cancer 2015;6(2): Lencioni R, Llovet JM. Modified RECIST (mrecist) assessment for hepatocellular carcinoma. Semin Liver Dis 2010;30(1): Goorin AM, Delorey MJ, Lack EE, et al. Prognostic significance of complete surgical resection of pulmonary metastases in patients with osteogenic sarcoma: analysis of 32 patients. J Clin Oncol 1984;2(5): Khanfir A, Frikha M, Ghorbel A, Drira MM, Daoud J. Prognostic factors in metastatic nasopharyngeal carcinoma. Cancer Radiother 2007;11(8): Pan C, He N, Zhao M, et al. Subdividing the M1 stage of liver metastasis for nasopharyngeal carcinoma to better predict metastatic survival. Med Oncol 2011;28(4): Radiology: Volume 280: Number 3 September 2016 n radiology.rsna.org 813

10 28. Chen MK, Chen TH, Liu JP, Chang CC, Chie WC. Better prediction of prognosis for patients with nasopharyngeal carcinoma using primary tumor volume. Cancer 2004; 100(10): Chen C, Fei Z, Pan J, Bai P, Chen L. Significance of primary tumor volume and T-stage on prognosis in nasopharyngeal carcinoma treated with intensity-modulated radiation therapy. Jpn J Clin Oncol 2011;41(4): Tang LQ, Chen QY, Fan W, et al. Prospective study of tailoring whole-body dualmodality [18F]fluorodeoxyglucose positron emission tomography/computed tomography with plasma Epstein-Barr virus DNA for detecting distant metastasis in endemic nasopharyngeal carcinoma at initial staging. J Clin Oncol 2013;31(23): Wang WY, Lin PJ, Lin JC. Circulating EBV DNA and ¹⁸F-FDG PET scan in the diagnosis and monitoring of nasopharyngeal carcinoma. J Formos Med Assoc 2012;111(2): Young LS, Dawson CW. Epstein-Barr virus and nasopharyngeal carcinoma. Chin J Cancer 2014;33(12): Ma J, Wen ZS, Lin P, Wang X, Xie FY. The results and prognosis of different treatment modalities for solitary metastatic lung tumor from nasopharyngeal carcinoma: a retrospective study of 105 cases. Chin J Cancer 2010;29(9): Chan OS, Ngan RK. Individualized treatment in stage IVC nasopharyngeal carcinoma. Oral Oncol 2014;50(9): Jin Y, Shi YX, Cai XY, et al. Comparison of five cisplatin-based regimens frequently used as the first-line protocols in metastatic nasopharyngeal carcinoma. J Cancer Res Clin Oncol 2012;138(10): radiology.rsna.org n Radiology: Volume 280: Number 3 September 2016