General results
A total of 3864 patients undergoing curative gastrectomy with nodal dissection were eligible for this study between February 2011 and February 2016.The 5-year survival rate (YSR) of 3864 GC patients was 54.7%, with a median overall survival (OS) of 63.03 months (range: 1.13-110.33 months) (Figure S2).
2872 (5-YSR: 55.1%) patients were male and 992 (5-YSR: 53.6%) were female, with a median age of 63 years (range: 19–90 years). 2570 (5-YSR: 52.8%) patients were aged 60 years or older while 1294 (5-YSR: 58.6%) were less than 60 years old. Regarding the tumor location, there were 1860 (5-YSR: 53.5%), 606 (5-YSR: 52.8%), 1104 (5-YSR: 62.5%), and 294 (5-YSR: 36.7%) tumors in the lower third, middle third, upper third, and > 2/3 stomach groups, respectively.
In view of the type of gastrectomy, there were 2628 (5-YSR: 49.5%), 1048 (5-YSR: 65.6%), and 188 (5-YSR: 66.0%) patients in the total gastrectomy (TG), distal gastrectomy (DG), proximal gastrectomy (PG) groups, respectively. Considering perineural invasion, there were 390 (5-YSR: 43.1%), and 3474 (5-YSR: 56.0%) patients in the positive, and negative groups, respectively. With respect to vascular invasion, there were 642 (5-YSR: 39.6%), and 3222 (5-YSR: 57.1%) patients in the positive, and negative groups, respectively.
In addition, of 3864 patients, 608 (5-YSR: 89.5%) were staged as having pT1 tumors, 326 (5-YSR: 77.9%) as pT2, 254 (5-YSR: 61.4%) as pT3, 2124 (5-YSR: 45.3%) as pT4a, and 552 (5-YSR: 35.9%) as pT4b. Patients were divided into five groups based on N stage (N0, N1, N2, N3a, and N3b), there were 1476 (5-YSR: 80.1%), 640 (5-YSR: 58.6%), 796 (5-YSR: 41.7%), 753 (5-YSR: 27.5%) and 200 (5-YSR: 9.0%) patients in each group respectively. Also, patients were divided into three groups based on TNM stage (I, II, III), there were 784 (5-YSR: 90.1%), 834 (5-YSR: 70.9%), and 2246 (5-YSR: 36.3%) patients in each group respectively.
The most appropriate cut-off value for MLNR was 0.25 (X2 = 866.433, P < 0.001) based on the "minimum p-value" approach by the X-tile software (Fig. 1). Of 3864 patients, 2548 (5-YSR: 69.5%) were classified into a low-MLNR (L-MLNR) group, and 1316 (5-YSR: 26.1%) were classified into a high-MLNR (H-MLNR) group. The most appropriate cut-off value for tumor size was 4.5 cm (X2 = 454.250, P < 0.001) (Figure S1). Patients were divided into two groups based on tumor size (< 4.5 cm and ≥ 4.5 cm), and there were 1872 (5-YSR: 71.8%), and 1948 (5-YSR: 38.4%) patients in each group respectively. The data of the included 3864 patients were analyzed, with the clinicopathological characteristics of the patients shown in Table 1.
Table 1
Clinicopathological characteristics and survival analyses of GC patients (N = 3864)
Characteristics | n | 5-YSR (%) | X2 | Univariate P value | HR (95% CI) | Multivariate P value |
Gender | | | 0.014 | 0.906 | | 0.699 |
Male | 2872 | 55.1 | | | | |
Female | 992 | 53.6 | | | | |
Age at surgery (year) | | | 15.521 | < 0.001 | | < 0.001 |
< 60 | 1294 | 58.6 | | | Reference | |
≥ 60 | 2570 | 52.8 | | | 1.242 (1.123–1.375) | |
Tumor location | | | 77.078 | < 0.001 | | 0.447 |
Upper third | 1860 | 53.5 | | | | |
Middle third | 606 | 52.8 | | | | |
Lower third | 1104 | 62.5 | | | | |
> 2/3 stomach | 294 | 36.7 | | | | |
Tumor size (cm)* | | | 457.908 | < 0.001 | | < 0.001 |
<4.5 | 1872 | 71.8 | | | Reference | |
≥4.5 | 1948 | 38.4 | | | 1.599 (1.437–1.781) | |
Type of gastrectomy | | | 95.569 | < 0.001 | | < 0.001 |
TG | 2628 | 49.5 | | | Reference | Reference |
DG | 1048 | 65.6 | | | 0.745 (0.662–0.838) | < 0.001 |
PG | 188 | 66.0 | | | 0.827 (0.646–1.060) | 0.133 |
T stage | | | 488.722 | < 0.001 | | 0.302 |
T1 | 608 | 89.5 | | | | |
T2 | 326 | 77.9 | | | | |
T3 | 254 | 61.4 | | | | |
T4a | 2124 | 45.3 | | | | |
T4b | 552 | 35.9 | | | | |
N stage | | | 1105.510 | < 0.001 | | < 0.001 |
N0 | 1476 | 80.1 | | | 0.375 (0.271–0.518) | < 0.001 |
N1 | 640 | 58.6 | | | 0.408 (0.311–0.535) | < 0.001 |
N2 | 796 | 41.7 | | | 0.497 (0.406–0.609) | < 0.001 |
N3a | 752 | 27.5 | | | 0.566 (0.476–0.673) | < 0.001 |
N3b | 200 | 9.0 | | | Reference | Reference |
ELN | | | 1.268 | 0.260 | | < 0.001 |
<16 | 1238 | 56.9 | | | Reference | |
≥16 | 2626 | 53.7 | | | 0.807 (0.726–0.897) | |
Perineural invasion | | | 6.271 | 0.012 | | 0.023 |
Positive | 390 | 43.1 | | | Reference | |
Negative | 3474 | 56.0 | | | 1.187 (1.024–1.376) | |
Vascular invasion | | | 100.000 | < 0.001 | | 0.005 |
Positive | 642 | 39.6 | | | Reference | |
Negative | 3222 | 57.7 | | | 0.841 (0.746–0.949) | |
TNM stage | | | 788.209 | < 0.001 | | < 0.001 |
I | 784 | 90.1 | | | Reference | Reference |
II | 834 | 70.9 | | | 2.237 (1.753–2.855) | < 0.001 |
III | 2246 | 36.3 | | | 3.642 (2.714–4.888) | < 0.001 |
MLNR | | | 883.883 | < 0.001 | | < 0.001 |
< 0.25 | 2548 | 69.5 | | | Reference | |
≥ 0.25 | 1316 | 26.1 | | | 1.526 (1.279–1.821) | |
GC, gastric cancer; TG, total gastrectomy; DG, distal gastrectomy; PG, proximal gastrectomy; ELN, examined lymph node; 5-YSR,5-year survival rate; HR, hazard ratio; 95% CI, 95% confidence interval;* ,median of tumor diameter: 4.5 (0.3 − 27.0) cm;MLNR: metastatic lymph |
node ratio; The eighth edition of TNM classification for GC was adopted for postoperatively pathological stages of all included patients |
Survival analysis
The median follow-up time was 62.7 (range: 1.4 − 110.2) months. Clinicopathological factors were analyzed to verify prognostic significance in terms of OS in patients. Univariable analysis demonstrated that age at surgery (P < 0.001), tumor location (P < 0.001), tumor size (P < 0.001), type of gastrectomy (P < 0.001), T stage (P < 0.001), N stage (P < 0.001), perineural invasion (P = 0.012), vascular invasion (P < 0.001), TNM stage (P < 0.001), and MLNR (P < 0.001) were significant predictors of OS (Fig. 2,3).
Significant differences were observed among the low and high-MLNR groups (P < 0.001); the 5-YSRs were 69.5%, and 26.1%, respectively. Multivariable analyses demonstrated that age at surgery (P < 0.001), tumor size (P < 0.001), type of gastrectomy (P < 0.001), N stage (P < 0.001), ELN (P < 0.001), vascular invasion (P = 0.005), TNM stage (P < 0.001), and MLNR (P < 0.001) were significant independent predictors of OS (Table 1).
Relationship between clinicopathological characteristics and MLNR of GC patients
Of 3864 GC patients, the Chi-square test analysis demonstrated that tumor location (P < 0.001), tumor size (P < 0.001), type of gastrectomy (P < 0.001), T stage (P < 0.001), N stage (P < 0.001), perineural invasion (P < 0.001), vascular invasion (P < 0.001), and TNM stage (P < 0.001) were significantly related to the MLNR.
Multivariable analyses demonstrated that age at surgery (P = 0.030, BIC: 876), tumor location (P = 0.005, BIC: 868), N stage (P < 0.001, BIC: 1916) and ELN (P < 0.001, BIC: 1023) were significant independent factors of MLNR. In addition, the multinomial logistic regression model showed the smallest BIC values for age at surgery and tumor location, indicating the most intensive relationship with MLNR (Table 2). The linear correlation between the ELN count and the MLNR was illustrated in Supplementary Fig. 3(R2 = 0.0008, P = 0.0778).
Table 2
Relationship between clinicopathological characteristics and Metastatic Lymph Node Ratio (n = 3864)
Characteristics | Metastatic Lymph Node Ratio | | Univariable analysis | | Multivariable analysis |
| < 0.25 | ≥ 0.25 | | X2 | P value | | OR (95%CI) | P value | BIC value |
Gender | | | | 0.357 | 0.550 | | | 0.316 | |
Male | 1906 | 966 | | | | | | | |
Female | 648 | 344 | | | | | | | |
Age at surgery (year) | | | | 3.319 | 0.068 | | | 0.030 | 876 |
< 60 | 830 | 464 | | | | | Reference | | |
≥ 60 | 1724 | 846 | | | | | 0.702 (0.511–0.966) | | |
Tumor location | | | | 46.407 | < 0.001 | | | 0.005 | 868 |
Upper third | 1308 | 552 | | | | | Reference | Reference | |
Middle third | 358 | 248 | | | | | 1.785 (1.184–2.693) | 0.006 | |
Lower third | 728 | 376 | | | | | 0.915 (0.639–1.309) | 0.627 | |
> 2/3 stomach | 160 | 134 | | | | | 0.579 (0.293–1.147) | 0.117 | |
Tumor size (cm) | | | | 276.692 | < 0.001 | | | 0.402 | |
<4.5 | 1482 | 390 | | | | | | | |
≥4.5 | 1046 | 902 | | | | | | | |
Type of gastrectomy | | | | 17.698 | < 0.001 | | | 0.060 | |
TG | 1690 | 938 | | | | | | | |
DG | 718 | 330 | | | | | | | |
PG | 146 | 42 | | | | | | | |
T stage | | | | 462.895 | < 0.001 | | | 0.334 | |
T1 | 590 | 18 | | | | | | | |
T2 | 278 | 48 | | | | | | | |
T3 | 174 | 80 | | | | | | | |
T4a | 1262 | 862 | | | | | | | |
T4b | 250 | 302 | | | | | | | |
N stage | | | | 2888.807 | < 0.001 | | | < 0.001 | 1916 |
N0 | 1476 | 0 | | | | | | | |
N1 | 628 | 12 | | | | | | | |
N2 | 434 | 362 | | | | | | | |
N3a | 10 | 742 | | | | | | | |
N3b | 0 | 200 | | | | | | | |
ELN | | | | 2.041 | 0.153 | | | < 0.001 | 1023 |
༜16 | 836 | 402 | | | | | Reference | | |
≥16 | 1712 | 914 | | | | | 0.129 (0.091–0.182) | | |
Perineural invasion | | | | 81.004 | < 0.001 | | | 0.719 | |
Positive | 178 | 212 | | | | | | | |
Negative | 2376 | 1098 | | | | | | | |
Vascular invasion | | | | 198.583 | < 0.001 | | | 0.914 | |
Positive | 270 | 372 | | | | | | | |
Negative | 2284 | 938 | | | | | | | |
TNM stage | | | | 1240.679 | < 0.001 | | | 0.471 | |
I | 784 | 0 | | | | | | | |
II | 794 | 40 | | | | | | | |
III | 970 | 1276 | | | | | | | |
BIC: Bayesian information criterion; TG, total gastrectomy; DG, distal gastrectomy; PG, proximal gastrectomy; ELN, examined lymph node; OR, odds ratio; 95% CI, 95% confidence interval; The depth of primary tumor invasion (T stage), classification of regional metastasis lymph nodes (N stage) and TNM stage were based on the 8th edition TNM staging system. |
3.4 Effects of other clinicopathological characteristics on MLNR for predicting prognosis of GC patients
Stratum analysis was used to assess the impact of the clinicopathological characteristics on the efficiency of prognostic prediction of the MLNR for GC patients.
We validated four clinicopathological factors (age at surgery, tumor location, N stage and ELN) that were statistically significant for MLNR in Table 3.
Table 3
Effects of clinicopathological characteristics on Metastatic Lymph Node Ratio for predicting prognosis of GC patients
Characteristics | MLNR | No. of patients (%) | 5-YSR (%) | X2 | P value |
Age at surgery (year) | | | | | |
< 60 | < 0.25 | 830 | 616 (74.2) | 300.799 | < 0.001 |
| ≥ 0.25 | 464 | 142 (30.6) | | |
≥ 60 | < 0.25 | 1724 | 1154 (66.9) | 586.187 | < 0.001 |
| ≥ 0.25 | 846 | 202 (23.9) | | |
Tumor location | | | | | |
Upper third | < 0.25 | 1308 | 858 (65.6) | 359.092 | < 0.001 |
| ≥ 0.25 | 552 | 138 (25.0) | | |
Middle third | < 0.25 | 358 | 246 (68.7) | 123.769 | < 0.001 |
| ≥ 0.25 | 248 | 74 (29.8) | | |
Lower third | < 0.25 | 728 | 572 (78.6) | 295.891 | < 0.001 |
| ≥ 0.25 | 376 | 118 (31.4) | | |
> 2/3 stomach | < 0.25 | 160 | 94 (58.8) | 82.958 | < 0.001 |
| ≥ 0.25 | 134 | 14 (10.4) | | |
N stage | | | | | |
N0 | < 0.25 | 1476 | 1182 (80.1) | - | - |
| ≥ 0.25 | 0 | | | |
N1 | < 0.25 | 628 | 372 (59.2) | 6.712 | 0.010 |
| ≥ 0.25 | 12 | 4 (33.3) | | |
N2 | < 0.25 | 434 | 206 (47.5) | 17.955 | < 0.001 |
| ≥ 0.25 | 362 | 126 (34.8) | | |
N3a | < 0.25 | 10 | 10 (100.0) | 13.552 | < 0.001 |
| ≥ 0.25 | 742 | 196 (26.4) | | |
N3b | < 0.25 | 0 | | - | - |
| ≥ 0.25 | 200 | 18 (9.0) | | |
ELN | | | | | |
<16 | < 0.25 | 836 | 592 (70.8) | 282.830 | < 0.001 |
| ≥ 0.25 | 402 | 112 (27.9) | | |
≥16 | < 0.25 | 1712 | 1178 (68.8) | 599.797 | < 0.001 |
| ≥ 0.25 | 914 | 232 (25.4) | | |
MLNR: metastatic lymph node ratio; ELN, examined lymph node; 5-YSR,5-year survival rate |
Table 4
MLNR predicting prognosis of GC patients (univariate Cox regression)
Characteristics | MLNR < 0.25 | | MLNR ≥ 0.25 |
HR (95% CI) | P | | HR (95% CI) | P |
Gender | Reference | Reference | | 0.269 (0.245–0.295) | < 0.001 |
Age at surgery | Reference | Reference | | 0.266 (0.242–0.292) | < 0.001 |
Tumor location | Reference | Reference | | 0.271 (0.247–0.297) | < 0.001 |
Tumor size | Reference | Reference | | 0.335 (0.304–0.369) | < 0.001 |
Type of gastrectomy | Reference | Reference | | 0.274 (0.249-0.300) | < 0.001 |
T stage | Reference | Reference | | 0.362 (0.328–0.398) | < 0.001 |
N stage | Reference | Reference | | 0.625 (0.528–0.740) | < 0.001 |
ELN | Reference | Reference | | 0.269 (0.245–0.295) | < 0.001 |
Perineural invasion | Reference | Reference | | 0.270 (0.245–0.296) | < 0.001 |
Vascular invasion | Reference | Reference | | 0.278 (0.253–0.306) | < 0.001 |
TNM stage | Reference | Reference | | 0.478 (0.431–0.531) | < 0.001 |
GC, gastric cancer; ELN, examined lymph node; HR, hazard ratio; 95% CI, 95% confidence interval; MLNR: metastatic lymph node ratio; The depth of primary tumor invasion (T stage), classification of regional metastasis lymph nodes (N stage) and TNM stage were based on the 8th edition TNM staging system. |
Firstly, we have evaluated the influence of MLNR on prognosis in both group 1 (age < 60) and group 2 (age ≥ 60). In group 1, there were 1294 patients, who were divided into L-MLNR and H-MLNR (P < 0.001) (Fig. 4A); the 5-YSRs were 74.2%, and 30.6%, respectively. In group 2, there were 2626 patients, who were divided into MLNR < 0.25 and MLNR ≥ 0.25 (P < 0.001) (Fig. 4B); the 5-YSRs were 66.9%, and 23.9%, respectively.
Secondly, we have evaluated the influence of MLNR on prognosis in group 1 (Upper third), group 2 (Middle third), group 3 (Lower third) and group 4 (> 2/3 stomach). In group 1, there were 1860 patients, who were divided into MLNR < 0.25 and MLNR ≥ 0.25 (P < 0.001) (Fig. 4E); the 5-YSRs were 65.6%, and 25.0%, respectively. In group 2, there were 606 patients, who were divided into MLNR < 0.25 and MLNR ≥ 0.25 (P < 0.001) (Fig. 4F); the 5-YSRs were 68.7%, and 29.8%, respectively. In group 3, there were 1104 patients, who were divided into MLNR < 0.25 and MLNR ≥ 0.25 (P < 0.001) (Fig. 4G); the 5-YSRs were 78.6%, and 31.4%, respectively. In group 4, there were 294 patients, who were divided into MLNR < 0.25 and MLNR ≥ 0.25 (P < 0.001) (Fig. 4H); the 5-YSRs were 58.8%, and 10.4%, respectively.
Thirdly, we have evaluated the influence of MLNR on prognosis in group 1 (N0), group 2 (N1), group 3 (N2), group 4 (N3a), and group 5 (N3b). In group 1, there were 1476 patients, who were MLNR < 0.25; the 5-YSRs were 80.1%. In group 2, there were 640 patients, who were divided into MLNR < 0.25 and MLNR ≥ 0.25 (P < 0.001) (Fig. 5B); the 5-YSRs were 59.2%, and 33.3%, respectively. In group 3, there were 796 patients, who were divided into MLNR < 0.25 and MLNR ≥ 0.25 (P < 0.001) (Fig. 5C); the 5-YSRs were 47.5%, and 34.8%, respectively. In group 4, there were 752 patients, who were divided into MLNR < 0.25 and MLNR ≥ 0.25 (P < 0.001) (Fig. 5D); the 5-YSRs were 100.0%, and 26.4%, respectively. In group 5, there were 200 patients, who were MLNR ≥ 0.25; the 5-YSRs were 9.0%.
Finally, we have evaluated the influence of MLNR on prognosis in group 1 (ELN < 16) and group 2 (ELN ≥ 16). In group 1, there were 1238 patients, who were divided into MLNR < 0.25 and MLNR ≥ 0.25 (P < 0.001) (Fig. 4C); the 5-YSRs were 70.8%, and 27.9%, respectively. In group 2, there were 2570 patients, who were divided into MLNR < 0.25 and MLNR ≥ 0.25 (P < 0.001) (Fig. 4D); the 5-YSRs were 68.8%, and 25.4%, respectively. The results showed that MLNRs were statistically significant at all levels of clinicopathological factors.
To make the above results more convincing, we used the univariate COX proportional hazards analysis to assess the impact of MLNR on the prognosis of GC patients. We discovered that MLNR had a significant impact on prognosis in distinguishing substages of patients with clinicopathological features, such as gender [P < 0.001, 0.269 (0.245–0.295)], age at surgery [P < 0.001, 0.266 (0.242–0.292)], tumor location [P < 0.001, 0.271 (0.247–0.297)], tumor size [P < 0.001, 0.335 (0.304–0.369) ], type of gastrectomy [P < 0.001, 0.274 (0.249-0.300)], T stage [P < 0.001, 0.362 (0.328–0.398)], N stage [P < 0.001, 0.625 (0.528–0.740)], ELN [P < 0.001, 0.269 (0.245–0.295)], perineural invasion [P < 0.001, 0.270 (0.245–0.296)], vascular invasion [P < 0.001, 0.278 (0.253–0.306)], and TNM stage [P < 0.001, 0.478 (0.431–0.531)].