Tumor Biol. DOI 10.1007/s13277-014-1848-6
REVIEW
Evaluation of the seventh AJCC TNM staging system for gastric cancer: a meta-analysis of cohort studies Jizhun Zhang & Yangbing Zhou & Kewei Jiang & Zhanlong Shen & Yingjiang Ye & Shan Wang
Received: 6 September 2013 / Accepted: 23 September 2013 # International Society of Oncology and BioMarkers (ISOBM) 2014
Abstract The AJCC seventh edition TNM classification for gastric cancer was released in 2010 and included major revision. Large-volume gastric cancer centers have evaluated the prognostic significance of the new system and obtained paradoxical results. The authors performed a meta-analysis of these studies to evaluate the new classification. Fifteen eligible studies with 38,972 patients were included in the analysis. Hazard ratios (HRs) and associated 95 % confidence intervals were extracted from identified studies. The primary outcome was overall survival. The HRs for the seventh edition T classification and N classification were found to increase steadily and reasonably. The cumulative survival rates of the seventh edition subgroups of T classifications demonstrated obvious differences; meanwhile, the differences between subgroups of N classifications including N3a and N3b categories were also significant. The 5year survival rates according to the seventh edition TNM staging system were 94.71 % (stage IA), 88.72 % (stage IB), 80.45 % (stage IIA), 67.24 % (stage IIB), 53.68 % (stage IIIA), 37.56 % (stage IIIB), and 21.26 % (stage IIIC), respectively. The results of this study indicate that the seventh edition of the TNM classification was considered valid, although further evaluation was needed for N3a and N3b categories.
Keywords Gastric cancer . TNM classification . Prognosis . AJCC
Introduction Gastric cancer is the fourth most common cancer and the second leading cause of cancer-related death worldwide [1]. Nowadays, the American Joint Committee on Cancer (AJCC) TNM classification has become the principal method for prognosis assessment in gastric cancer. In 2010, the AJCC seventh edition for gastric cancer was released based on recent medical evidence. This new classification made revisions in depth of invasion (T category), number of metastatic lymph nodes (N category), and the final staging groupings (TNM category). According to the seventh TNM classification, the sixth edition T2 classification was subclassified into T2 (muscularis propria) and T 3 (subserosa) and T 1 was subclassified into T1a (mucosa) and T1b (submucosa). The new N stage was categorized to N0 (no regional lymph node (LN) metastasis), N1 (1–2 involved regional LNs), N2 (3–6 involved regional LNs), N3a (7–15 involved regional LNs), and N3b (>15 involved regional LNs). M classification was based on the presence of distant metastasis [2, 3]. A few of cancer centers have investigated the effect of the new TNM classification both in Asian and Western populations and obtained paradoxical results [4–18]. According to Kwon SJ, the seventh edition even failed to resolve any noticeable matters [19]. With the aim of evaluating the efficacy of the new system, we conducted a meta-analysis of studies from 20 large gastric cancer centers in 5 countries.
Electronic supplementary material The online version of this article (doi:10.1007/s13277-014-1848-6) contains supplementary material, which is available to authorized users. J. Zhang : K. Jiang : Z. Shen : Y. Ye (*) : S. Wang (*) Department of Gastroenterological Surgery, People’s Hospital, Peking University, Beijing 100044, People’s Republic of China e-mail: [email protected] e-mail: [email protected] Y. Zhou Department of General Surgery, Affiliated Hospital of Qingdao, University Medical College, Qingdao, China
Methods Search strategy and selection criteria PubMed, Embase, Web of Science, and CBM databases were searched by using combinations of the following keywords: “gastric cancer,” gastric carcinoma,” “carcinoma of stomach,” “stomach neoplasms,” “TNM classification,” “TNM stage,”
Tumor Biol.
“TNM system,” “prognostic,” and “prognosis” (last update on March 1, 2013). The referenced literatures in the identified reports were also retrieved for other potentially important researches. All researches were carefully estimated to identify repeated data. Criteria used to define duplicate data included study period, hospital, treatment information, and any additional inclusion criteria. Two reviewers independently searched all the identified abstracts and retrieved full articles for detailed information. No language restrictions were applied. To be included in the meta-analysis of all-cause mortality, eligibility criteria were as follows: overall survival curves generated from individual patient data, hazard ratios (HRs) could be calculated from the outcomes, and more than 15 LNs examined in most cases. Data extraction and quality assessment Standardized data collection forms were used for information collection. Data retrieved from the reports included publication details and study characteristics, such as sample size, median follow-up, and outcome measures. Principal authors of the studies were contacted for missing information and updated outcome if needed. To provide valid assessments of answers to prognostic questions, evaluation of the quality of studies was carried out strictly in the meta-analysis [20]. Statistical analysis The primary outcome was overall survival, defined as the time after surgery until any cause death. Patients who had not experienced the event by the time of their last followup were censored. The HRs and their variances were directly extracted from the articles or estimated indirectly from the number of events and the survival curves [21]. Absolute differences in survival rates were calculated from survival curves in identified articles at annual intervals. Median follow-up was directly extracted from the studies or calculated according to the inverted Kaplan– Meier technique [22]. Quantity I [2] was calculated to indicate the approximate proportion of total variability (0 to 100 %) in point estimates that can be attributed to systematic differences across studies rather than to chance [23]. Heterogeneity was explored by using sensitivity analyses. To be consistent, if there was significant heterogeneity in some groups (P >0.05, I 2 >50 %), HRs would be calculated with a random effects model; otherwise, primary analyses would be done with a fixed effects model [24]. All analyses were conducted with STATA 11 (College Station, TX). All P values were two sided and the cutoff for statistical significance was 0.05.
Results Characteristics of the studies The steps of the selection diagram were presented in Supplementary Fig. 1. Twenty-four studies were included for detailed evaluation. Of those, four studies did not provide sufficient information to calculate HRs, even after contacting the authors [25–28]. Two studies were excluded due to double evaluation [29, 30]. Longer follow-up was warranted in another study (median follow-up after surgery was only 14 months) [31]. In other two studies, most patients in their series did not have more than 15 lymph nodes harvested [32, 33]. A total of 15 studies were eligible for analysis. Two studies were conducted in Germany [5, 10], one study in Italy [8], one study in Netherlands [17], six in Korea [6, 9, 12, 15, 16, 18], and five in China [4, 7, 11, 13, 14]. By using strict inclusion, we aimed to guarantee the quality of the included studies. The main characteristics of the patients and studies included were listed in Table 1. Survival discrimination by T stage Eleven studies were available for T stage assessment. Overall survival results of T parameter are based on 26,571 patients with a median follow-up of 76 months [4–8, 10–13, 15, 16]. The HRs of each seventh edition T classification were 2.37 in T2, 3.94 in T3, 7.17 in T4a, and 17.13 in T4b compared with T1 (Table 2). The 5-year survival rates of each seventh edition T classification were as follows: T1, 92.40 %; T2, 74.83 %; T3, 56.06 %; T4a, 37.40 %; and T4b, 16.19 % (Fig. 1). Survival discrimination by N stage Thirteen studies were qualified for N stage assessment. Overall survival results of N parameter were based on 27,359 patients with a median follow-up of 78 months [4–8, 10–16, 18]. The HRs of each seventh edition N classification were 2.53 in N1, 3.95 in N2, 6.61 in N3a , and 11.03 in N3b compared with N1 (Table 2). The 5-year survival rates of each seventh edition N classification were as follows: N0, 87.04 %; N1, 67.72 %; N2, 49.06 %; N3a, 32.14 %; and N3b, 16.60 % (Fig. 2). Survival discrimination by TNM staging group Eleven studies were qualified for final TNM stage assessment. Overall survival results of TNM parameter are based on 36,588 patients with a median follow-up of 61 months [4–10, 12, 13, 15, 17]. The HRs of each seventh edition TNM classification were 1.70 in IB, 2.60 in IIA, 3.70 in IIB, 5.31 in IIIA, 7.45 in IIIB, and 11.19 in IIIC compared with IA (Fig. 3). The 5-year survival rates according to the seventh edition TNM staging
Tumor Biol. Table 1 Study characteristics Study by first author
Year
Country
No. of patients
Median followup (months)
Ahn [15] Yoon [6] Chae [18] Jung [12]
2010 2011 2011 2010
Korea Korea Korea Korea
9,998 1,799 295 2,916
86.7 48 57.2 72
Kim [9] Ha [16] Viktoria [10] Reim [5] Marrelli [8] Wang [11] Sun [7] Deng [14] Fang [13] Zhang [4] Dikken [17]
2011 2009 2011 2012 2011 2010 2012 2010 2011 2013 2012
Korea Korea Germany Germany Italy China China China China China Netherlands
10,060 1,633 554 1,767 2,090 1,503 1,998 456 1,380 964 1,559
61 50 16 77 87 50 36 54 62.2 24.5 98
Fig. 1 Survival curves for gastric cancer patients following curative resection according to stage subgroup of the seventh AJCC T stage
system were 94.71 % (stage IA), 88.72 % (stage IB), 80.45 % (stage IIA), 67.24 % (stage IIB), 53.68 % (stage IIIA), 37.56 % (stage IIIB), and 21.26 % (stage IIIC) (Fig. 4).
Discussion Ideal cancer staging should not only provide an indication of prognosis and a framework for treatment decisions but should Table 2 Hazard ratios and 95 % confidence intervals (95 %CI) for survival by stage (vs. stageT1, N0, and IA) Seventh No. of patients HRs
95 %CI
χ2
P value I 2 (%)
T1 T2 T3 T4a T4b N0 N1 N2 N3a N3b Ia
8,862 3,356 6,023 7,385 945 12,826 3,696 4,068 4,061 2,515 12,011
1 2.37 3.94 7.17 17.13 1 2.53 3.95 6.61 11.03 1
– 2.12–2.64 3.37–4.60 6.19–8.31 12.84–22.87 – 2.23–2.86 3.19–4.89 5.54–7.89 8.90–13.68 –
– 8.94 21.16 21.26 30.96 – 35.49 113.4 67.80 82.05 –
– 0.54 0.02 0.02
REVIEW
Evaluation of the seventh AJCC TNM staging system for gastric cancer: a meta-analysis of cohort studies Jizhun Zhang & Yangbing Zhou & Kewei Jiang & Zhanlong Shen & Yingjiang Ye & Shan Wang
Received: 6 September 2013 / Accepted: 23 September 2013 # International Society of Oncology and BioMarkers (ISOBM) 2014
Abstract The AJCC seventh edition TNM classification for gastric cancer was released in 2010 and included major revision. Large-volume gastric cancer centers have evaluated the prognostic significance of the new system and obtained paradoxical results. The authors performed a meta-analysis of these studies to evaluate the new classification. Fifteen eligible studies with 38,972 patients were included in the analysis. Hazard ratios (HRs) and associated 95 % confidence intervals were extracted from identified studies. The primary outcome was overall survival. The HRs for the seventh edition T classification and N classification were found to increase steadily and reasonably. The cumulative survival rates of the seventh edition subgroups of T classifications demonstrated obvious differences; meanwhile, the differences between subgroups of N classifications including N3a and N3b categories were also significant. The 5year survival rates according to the seventh edition TNM staging system were 94.71 % (stage IA), 88.72 % (stage IB), 80.45 % (stage IIA), 67.24 % (stage IIB), 53.68 % (stage IIIA), 37.56 % (stage IIIB), and 21.26 % (stage IIIC), respectively. The results of this study indicate that the seventh edition of the TNM classification was considered valid, although further evaluation was needed for N3a and N3b categories.
Keywords Gastric cancer . TNM classification . Prognosis . AJCC
Introduction Gastric cancer is the fourth most common cancer and the second leading cause of cancer-related death worldwide [1]. Nowadays, the American Joint Committee on Cancer (AJCC) TNM classification has become the principal method for prognosis assessment in gastric cancer. In 2010, the AJCC seventh edition for gastric cancer was released based on recent medical evidence. This new classification made revisions in depth of invasion (T category), number of metastatic lymph nodes (N category), and the final staging groupings (TNM category). According to the seventh TNM classification, the sixth edition T2 classification was subclassified into T2 (muscularis propria) and T 3 (subserosa) and T 1 was subclassified into T1a (mucosa) and T1b (submucosa). The new N stage was categorized to N0 (no regional lymph node (LN) metastasis), N1 (1–2 involved regional LNs), N2 (3–6 involved regional LNs), N3a (7–15 involved regional LNs), and N3b (>15 involved regional LNs). M classification was based on the presence of distant metastasis [2, 3]. A few of cancer centers have investigated the effect of the new TNM classification both in Asian and Western populations and obtained paradoxical results [4–18]. According to Kwon SJ, the seventh edition even failed to resolve any noticeable matters [19]. With the aim of evaluating the efficacy of the new system, we conducted a meta-analysis of studies from 20 large gastric cancer centers in 5 countries.
Electronic supplementary material The online version of this article (doi:10.1007/s13277-014-1848-6) contains supplementary material, which is available to authorized users. J. Zhang : K. Jiang : Z. Shen : Y. Ye (*) : S. Wang (*) Department of Gastroenterological Surgery, People’s Hospital, Peking University, Beijing 100044, People’s Republic of China e-mail: [email protected] e-mail: [email protected] Y. Zhou Department of General Surgery, Affiliated Hospital of Qingdao, University Medical College, Qingdao, China
Methods Search strategy and selection criteria PubMed, Embase, Web of Science, and CBM databases were searched by using combinations of the following keywords: “gastric cancer,” gastric carcinoma,” “carcinoma of stomach,” “stomach neoplasms,” “TNM classification,” “TNM stage,”
Tumor Biol.
“TNM system,” “prognostic,” and “prognosis” (last update on March 1, 2013). The referenced literatures in the identified reports were also retrieved for other potentially important researches. All researches were carefully estimated to identify repeated data. Criteria used to define duplicate data included study period, hospital, treatment information, and any additional inclusion criteria. Two reviewers independently searched all the identified abstracts and retrieved full articles for detailed information. No language restrictions were applied. To be included in the meta-analysis of all-cause mortality, eligibility criteria were as follows: overall survival curves generated from individual patient data, hazard ratios (HRs) could be calculated from the outcomes, and more than 15 LNs examined in most cases. Data extraction and quality assessment Standardized data collection forms were used for information collection. Data retrieved from the reports included publication details and study characteristics, such as sample size, median follow-up, and outcome measures. Principal authors of the studies were contacted for missing information and updated outcome if needed. To provide valid assessments of answers to prognostic questions, evaluation of the quality of studies was carried out strictly in the meta-analysis [20]. Statistical analysis The primary outcome was overall survival, defined as the time after surgery until any cause death. Patients who had not experienced the event by the time of their last followup were censored. The HRs and their variances were directly extracted from the articles or estimated indirectly from the number of events and the survival curves [21]. Absolute differences in survival rates were calculated from survival curves in identified articles at annual intervals. Median follow-up was directly extracted from the studies or calculated according to the inverted Kaplan– Meier technique [22]. Quantity I [2] was calculated to indicate the approximate proportion of total variability (0 to 100 %) in point estimates that can be attributed to systematic differences across studies rather than to chance [23]. Heterogeneity was explored by using sensitivity analyses. To be consistent, if there was significant heterogeneity in some groups (P >0.05, I 2 >50 %), HRs would be calculated with a random effects model; otherwise, primary analyses would be done with a fixed effects model [24]. All analyses were conducted with STATA 11 (College Station, TX). All P values were two sided and the cutoff for statistical significance was 0.05.
Results Characteristics of the studies The steps of the selection diagram were presented in Supplementary Fig. 1. Twenty-four studies were included for detailed evaluation. Of those, four studies did not provide sufficient information to calculate HRs, even after contacting the authors [25–28]. Two studies were excluded due to double evaluation [29, 30]. Longer follow-up was warranted in another study (median follow-up after surgery was only 14 months) [31]. In other two studies, most patients in their series did not have more than 15 lymph nodes harvested [32, 33]. A total of 15 studies were eligible for analysis. Two studies were conducted in Germany [5, 10], one study in Italy [8], one study in Netherlands [17], six in Korea [6, 9, 12, 15, 16, 18], and five in China [4, 7, 11, 13, 14]. By using strict inclusion, we aimed to guarantee the quality of the included studies. The main characteristics of the patients and studies included were listed in Table 1. Survival discrimination by T stage Eleven studies were available for T stage assessment. Overall survival results of T parameter are based on 26,571 patients with a median follow-up of 76 months [4–8, 10–13, 15, 16]. The HRs of each seventh edition T classification were 2.37 in T2, 3.94 in T3, 7.17 in T4a, and 17.13 in T4b compared with T1 (Table 2). The 5-year survival rates of each seventh edition T classification were as follows: T1, 92.40 %; T2, 74.83 %; T3, 56.06 %; T4a, 37.40 %; and T4b, 16.19 % (Fig. 1). Survival discrimination by N stage Thirteen studies were qualified for N stage assessment. Overall survival results of N parameter were based on 27,359 patients with a median follow-up of 78 months [4–8, 10–16, 18]. The HRs of each seventh edition N classification were 2.53 in N1, 3.95 in N2, 6.61 in N3a , and 11.03 in N3b compared with N1 (Table 2). The 5-year survival rates of each seventh edition N classification were as follows: N0, 87.04 %; N1, 67.72 %; N2, 49.06 %; N3a, 32.14 %; and N3b, 16.60 % (Fig. 2). Survival discrimination by TNM staging group Eleven studies were qualified for final TNM stage assessment. Overall survival results of TNM parameter are based on 36,588 patients with a median follow-up of 61 months [4–10, 12, 13, 15, 17]. The HRs of each seventh edition TNM classification were 1.70 in IB, 2.60 in IIA, 3.70 in IIB, 5.31 in IIIA, 7.45 in IIIB, and 11.19 in IIIC compared with IA (Fig. 3). The 5-year survival rates according to the seventh edition TNM staging
Tumor Biol. Table 1 Study characteristics Study by first author
Year
Country
No. of patients
Median followup (months)
Ahn [15] Yoon [6] Chae [18] Jung [12]
2010 2011 2011 2010
Korea Korea Korea Korea
9,998 1,799 295 2,916
86.7 48 57.2 72
Kim [9] Ha [16] Viktoria [10] Reim [5] Marrelli [8] Wang [11] Sun [7] Deng [14] Fang [13] Zhang [4] Dikken [17]
2011 2009 2011 2012 2011 2010 2012 2010 2011 2013 2012
Korea Korea Germany Germany Italy China China China China China Netherlands
10,060 1,633 554 1,767 2,090 1,503 1,998 456 1,380 964 1,559
61 50 16 77 87 50 36 54 62.2 24.5 98
Fig. 1 Survival curves for gastric cancer patients following curative resection according to stage subgroup of the seventh AJCC T stage
system were 94.71 % (stage IA), 88.72 % (stage IB), 80.45 % (stage IIA), 67.24 % (stage IIB), 53.68 % (stage IIIA), 37.56 % (stage IIIB), and 21.26 % (stage IIIC) (Fig. 4).
Discussion Ideal cancer staging should not only provide an indication of prognosis and a framework for treatment decisions but should Table 2 Hazard ratios and 95 % confidence intervals (95 %CI) for survival by stage (vs. stageT1, N0, and IA) Seventh No. of patients HRs
95 %CI
χ2
P value I 2 (%)
T1 T2 T3 T4a T4b N0 N1 N2 N3a N3b Ia
8,862 3,356 6,023 7,385 945 12,826 3,696 4,068 4,061 2,515 12,011
1 2.37 3.94 7.17 17.13 1 2.53 3.95 6.61 11.03 1
– 2.12–2.64 3.37–4.60 6.19–8.31 12.84–22.87 – 2.23–2.86 3.19–4.89 5.54–7.89 8.90–13.68 –
– 8.94 21.16 21.26 30.96 – 35.49 113.4 67.80 82.05 –
– 0.54 0.02 0.02
