Abstract
In T1 gastric cancer (GC), lymph nodes metastasis (LNM) is considered as a significant prognostic predictor and closely associated with following therapeutic approaches as well as distant metastasis (DM). This study aimed to not only seek risk factors of LNM and DM but also unpack the prognosis in T1 GC patients. We performed a retrospective study enrolling 5547 patients in T1 GC from the Surveillance, Epidemiology, and End Results (SEER) database. Univariate and multivariate logistic regression models were produced to recognize independent risk factors of LNM and DM. Cox regression analyses were performed to identify important prognostic factors of overall survival (OS). Cancer-specific cumulative incidence was plotted by cumulative incidence function. Three nomograms of LNM, DM and OS were established and validated by receiver operating characteristic (ROC) and calibration curves to evaluate discrimination and accuracy. Decision curve analysis (DCA), clinical impact curves (CIC) and subgroups based on risk scores were constructed to measure nomograms clinical utility. The area under the curve (AUC) of LNM nomogram and DM nomogram were 0.735 and 0.896, respectively. OS nomogram was constructed and the corresponding C-index was 0.797. In conclusion, our user-friendly nomograms, which aimed to predict LNM, DM and OS in T1 gastric cancer patients, have shown high efficiency of discrimination and accuracy. These useful and visual tools may have advantageous clinical utility to identify high-risk T1 gastric patients and help clinicians to draw up an individual therapeutic strategy.
Keywords: T1 gastric cancer, lymph nodes metastasis, distant metastasis, prognosis, nomogram, SEER
Introduction
In the world, gastric cancer (GC) ranks sixth in cancer incidence but second in mortality [1]. In 2020, there are 27,600 new diagnosed cases and 11,010 dead cases in USA [2]. T1 GC is defined as GC limited to the mucosa or submucosa according to the 8th AJCC TNM stage system. The natural history of GC is nonhomogeneous and crucial to determine disease prognosis [3]. T1 GC progresses slowly [4] but lymph nodes metastasis (LNM) occurs to approximately 20% of T1 GC patients [5]. As we understood, the occurrence of LNM or distant metastasis (DM) usually connected with the poor prognosis. Therefore, when thinking of treatment options and timing, we need to take the probability of LMM and DM into consider.
According to different stage of GC, clinicians choose different therapeutic strategy, such as endoscopic therapy, gastrectomy, and systemic treatment [6,7]. Several studies have shown endoscopic therapy is an effective mode with similar advantageous results compared to traditional surgical resection [8-10]. In addition, endoscopic therapy like endoscopic mucosal resection (EMR) and endoscopic submucosal dissection (ESD) [11] is associated with shorter hospitalization, fewer complications and better quality of life [12-14]. However, endoscopic therapy has its indication [15,16], for T1N1-3M0 GCs, this technique might not solve the problem of LNM. And for T1N0-3M1 GCs, systemic treatment might be better according to Japanese gastric cancer treatment guidelines 2018 (5th edition) [17].
Given the heterogeneity of T1 GCs and its totally different treatment, we decide to develop predictive models based on the Surveillance, Epidemiology, and End Results (SEER) program to help clinicians make appropriate therapeutic strategy. Our study aims were to (1) calculate independent risk factors of LNM and DM among T1 GCs, then plot relevant nomograms; (2) proceed survival analyses and construct an OS prognostic nomogram in T1 GCs; (3) assess clinical effects through subgroups.
Materials and methods
Data sources and population selection
In present study, the data onto patients were extracted from the SEER program which was a public-use database and afforded relative authoritative information about cancer relevant records covering approximately 35% population of the United State, so no ethical approval was required for our study and the quality of data sources was guaranteed. A total of 85,128 cases, diagnosed as T1 gastric cancer from 2004 to 2016, were obtained from the database. The exclusion criteria are as follows: (1) Age at diagnosis < 18; (2) Absence or ambiguity of important clinicopathological variables, such as race, grade, primary site, regional nodes examined, tumor size; (3) Incomplete survival information; (4) Patients with gastric cancer in T0, T2-4, TX, NX or MX. Given the existing evidence that patients with distant metastasis were regarded as terminal stage and the situation of lymph nodes wasn’t the decisive factor of treatment, we divided the aggregate cases into two study groups. Group N (n = 4747) was the T1N0-3M0 gastric cancer population to predict risk factors of LNM while group M (n = 5547) was the total T1 gastric cancer population to predict DM. The flow chart of case inclusion and exclusion is represented in Figure 1.
Figure 1.
Research flowchart.
Variables and outcomes
Accordingly, 16 clinicopathological variables were obtained from SEER database, including year of diagnosis, age at diagnosis, race, sex, grade, histology, primary site, tumor size, regional nodes examined, AJCC 7th N stage, survival status, marital status, radiation, chemotherapy, surgery and follow-up time. Overall survival (OS) was explained as span from the diagnosis date to the last time of follow-up or the date of death for any reason. Cancer-specific survival (CSS) was explained as span from diagnosis date until death only because of gastric cancer. For demographic variables, we registered year of diagnosis (2004-2008, 2009-2012, 2013-2016), race (white, black, others), age at diagnosis (18-55, 56-65, 66-75, 76+ years old) and marital status (married, others). Tumor covariates enrolled grade (well differentiated, moderately differentiated, poorly differentiated, undifferentiated), histology was reclassified based on the International Classification of Diseases for Oncology, 3rd Edition (ICD-O-3) Hist/behav, malignant (adenocarcinoma: 8140, 8144, 8210, 8211, 8260, 8261, 8263; signet ring cell carcinoma: 8490; others), primary site was defined by the International Classification of Diseases for Oncology (ICD-O-2) codes (cardia: C16.0-cardia; non-cardia: C16.1-fundus, C16.2-body, C16.3-antrum, C16.4-pylorus, C16.5-lesser curvature, C16.6-greater curvature, C16.7-overlapping lesion), tumor size (1-9 mm, 10-19 mm, 20-29 mm, 30+ mm), regional nodes examined (≤12, >12), AJCC 7th N stage (N0, non-N0: N1/2/3). Treatment related characteristics included radiation (Yes, No/Unknown), chemotherapy (Yes, No/Unknown) and surgery (Yes, No/Unknown). Other covariates included survival status (alive, dead of cancer, dead of other cause) and follow-up time.
Construction of nomograms and evaluation
We produced univariate and multivariate logistic regression models [18] to recognize independent risk factors of lymph nodes metastasis in group N and distant metastasis in group M. Cox regression analyses were performed to identify important prognostic factors of OS and CSS, respectively. Odd ratio (OR) and hazard ratio (HR) were used to weigh the impact of each insignificant factor on LNM, DM and OS, respectively. Sub-distribution hazard ration (SHR) was applied to measure the effect of each prognostic variable on CSS. Overall survival curves were plotted by Kaplan-Meier method and cancer-specific cumulative incidence was plotted by cumulative incidence function. Then, based on the results of multivariate binary logistic regression models, two nomograms were conducted to predict the risk factors of LNM and DM in T1 gastric cancer patients. Meanwhile, according to Cox proportional hazard model, one prognostic nomogram was established to calculate the OS probability of gastric cancer patients. These nomograms [19] were validated by ROC and calibration curves to evaluate the discrimination and accuracy. The value of C-index is as same as that of AUC in logistic regression model. The maximum C-index is 1.0, indicating a perfect differentiation ability, while 0.5 represents a random chance of the nomogram. Decision curve analysis (DCA), as a tool to assess the clinical application value of the nomogram [20], was constructed in this study to evaluate the net benefits. Furthermore, we performed clinical impact curves to reveal the value of nomogram models more intuitively. Based on the risk scores, all cases were divided into low-, medium-, and high-risk subgroups to measure the utility of nomograms.
Statistical analysis
We operated all statistical analyses by R software (version 4.0.0) and GraphPad Prism (version 8.4.3). Using R packages and functions, curves (nomogram, Kaplan-Meier, ROC, calibration, DCA, and clinical impact curve) were plotted. A two-sided P-value < .05 was accounted statistically significant.
Results
Clinical characteristics of patients with T1 gastric cancer
A total of 5547 eligible T1 gastric cancer patients were enrolled from SEER program, retrospectively. Demographic and clinical characteristics of patients are shown in Table 1. In the present study, we divided the total 5547 into two groups, group N (T1N0-3M0, n = 4747) and group M (T1N0-3M0-1, n = 5547). The incidence of lymph node metastasis was 17.86% in group N and the occurrence of distant metastasis was 14.42% in group M. The medium follow-up time was 38 months in group N (ranging from 13 to 76 months) and 30 months in group M (ranging from 9 to 69 months), respectively.
Table 1.
Clinicopathological variables of patients with T1 gastric cancer
| Clinicopathological variables | Nt (%) | Ne (%) | Nne (%) | P | Mt (%) | Me (%) | Mne (%) | P |
|---|---|---|---|---|---|---|---|---|
| N = 4747 | N = 848 | N = 3899 | N = 5547 | N = 800 | N = 4747 | |||
| Year of diagnosis | .086 | .088 | ||||||
| 2004-2008 | 1635 (34.4) | 317 (37.4) | 1318 (33.8) | 1883 (33.9) | 248 (31.0) | 1635 (34.4) | ||
| 2009-2012 | 1562 (32.9) | 277 (32.7) | 1285 (33.0) | 1853 (33.4) | 291 (36.4) | 1562 (32.9) | ||
| 2013-2016 | 1550 (32.7) | 254 (30.0) | 1296 (33.2) | 1811 (32.6) | 261 (32.6) | 1550 (32.7) | ||
| Age at diagnosis | < .001 | < .001 | ||||||
| 18-55 | 732 (15.4) | 161 (19.0) | 571 (14.6) | 950 (17.1) | 218 (27.3) | 732 (15.4) | ||
| 56-65 | 1011 (21.3) | 190 (22.4) | 821 (21.1) | 1220 (22.0) | 209 (26.1) | 1011 (21.3) | ||
| 66-75 | 1411 (29.7) | 271 (32.0) | 1140 (29.2) | 1612 (29.1) | 201 (25.1) | 1411 (29.7) | ||
| 76+ | 1593 (33.6) | 226 (26.7) | 1367 (35.1) | 1765 (31.8) | 172 (21.5) | 1593 (33.6) | ||
| Race | .797 | < .001 | ||||||
| White | 3099 (65.3) | 562 (66.3) | 2537 (65.1) | 3696 (66.6) | 597 (74.6) | 3099 (65.3) | ||
| Black | 568 (12.0) | 98 (11.6) | 470 (12.1) | 664 (12.0) | 96 (12.0) | 568 (12.0) | ||
| Others | 1080 (22.8) | 188 (22.2) | 892 (22.9) | 1187 (21.4) | 107 (13.4) | 1080 (22.8) | ||
| Sex | .005 | < .001 | ||||||
| Female | 1830 (38.6) | 290 (34.2) | 1540 (39.5) | 2076 (37.4) | 246 (30.8) | 1830 (38.6) | ||
| Male | 2917 (61.4) | 558 (65.8) | 2359 (60.5) | 3471 (62.6) | 554 (69.2) | 2917 (61.4) | ||
| Grade | < .001 | < .001 | ||||||
| Well differentiated | 766 (16.1) | 47 (5.5) | 719 (18.4) | 793 (14.3) | 27 (3.4) | 766 (16.1) | ||
| Moderately differentiated | 1774 (37.4) | 283 (33.4) | 1491 (38.2) | 1994 (35.9) | 220 (27.5) | 1774 (37.4) | ||
| Poorly differentiated | 2113 (44.5) | 500 (59.0) | 1613 (41.4) | 2646 (47.7) | 533 (66.6) | 2113 (44.5) | ||
| Undifferentiated | 94 (2.0) | 18 (2.1) | 76 (1.9) | 114 (2.1) | 20 (2.5) | 94 (2.0) | ||
| Histology | .023 | .060 | ||||||
| Adenocarcinoma | 3440 (72.5) | 583 (68.8) | 2857 (73.3) | 4046 (72.9) | 606 (75.8) | 3440 (72.5) | ||
| Signet ring cell carcinoma | 699 (14.7) | 146 (17.2) | 553 (14.2) | 792 (14.3) | 93 (11.6) | 699 (14.7) | ||
| Others | 608 (12.8) | 119 (14.0) | 489 (12.5) | 709 (12.8) | 101 (12.6) | 608 (12.8) | ||
| Primary site | .022 | < .001 | ||||||
| Cardia | 1497 (31.5) | 296 (34.9) | 1201 (30.8) | 1926 (34.7) | 429 (53.6) | 1497 (31.5) | ||
| Non-cardia | 3250 (68.5) | 552 (65.1) | 2698 (69.2) | 3621 (65.3) | 371 (46.4) | 3250 (68.5) | ||
| Tumor size (mm) | < .001 | < .001 | ||||||
| 1-9 | 890 (18.7) | 32 (3.8) | 858 (22.0) | 897 (16.2) | 7 (0.9) | 890 (18.7) | ||
| 10-19 | 1264 (26.6) | 169 (19.9) | 1095 (28.1) | 1321 (23.8) | 57 (7.1) | 1264 (26.6) | ||
| 20-29 | 949 (20.0) | 182 (21.5) | 767 (19.7) | 1040 (18.7) | 91 (11.4) | 949 (20.0) | ||
| 30+ | 1644 (34.6) | 465 (54.8) | 1179 (30.2) | 2289 (41.3) | 645 (80.6) | 1644 (34.6) | ||
| Regional nodes examined | < .001 | < .001 | ||||||
| ≤12 | 2761 (58.2) | 357 (42.1) | 2404 (61.7) | 3530 (63.6) | 2761 (58.2) | 769 (96.1) | ||
| >12 | 1986 (41.8) | 491 (57.9) | 1495 (38.3) | 2017 (36.4) | 1986 (41.8) | 31 (3.9) | ||
| N stage | NA | < .001 | ||||||
| N0 | NA | NA | NA | 4309 (77.7) | 410 (51.2) | 3899 (82.1) | ||
| Non-N0 (N1/2/3) | NA | NA | NA | 1238 (22.3) | 390 (48.8) | 848 (17.9) | ||
| Survival status | < .001 | < .001 | ||||||
| Alive | 2929 (61.7) | 448 (52.8) | 2481 (63.6) | 3013 (54.3) | 84 (10.5) | 2929 (61.7) | ||
| Dead of cancer | 1065 (22.4) | 293 (34.6) | 772 (19.8) | 1744 (31.4) | 679 (84.9) | 1065 (22.4) | ||
| Dead of other cause | 753 (15.9) | 107 (12.6) | 646 (16.6) | 790 (14.2) | 37 (4.6) | 753 (15.9) | ||
| Marital status | .019 | .205 | ||||||
| Married | 2845 (59.9) | 539 (63.6) | 2306 (59.1) | 3344 (60.3) | 499 (62.4) | 2845 (59.9) | ||
| Others | 1902 (40.1) | 309 (36.4) | 1593 (40.9) | 2203 (39.7) | 301 (37.6) | 1902 (40.1) | ||
| Radiation | < .001 | < .001 | ||||||
| Yes | 709 (14.9) | 398 (46.9) | 311 (8.0) | 903 (16.3) | 194 (24.2) | 709 (14.9) | ||
| None/Unknown | 4038 (85.1) | 450 (53.1) | 3588 (92.0) | 4644 (83.7) | 606 (75.8) | 4038 (85.1) | ||
| Chemotherapy | < .001 | < .001 | ||||||
| Yes | 968 (20.4) | 512 (60.4) | 456 (11.7) | 1522 (27.4) | 554 (69.2) | 968 (20.4) | ||
| None/Unknown | 3779 (79.6) | 336 (39.6) | 3443 (88.3) | 4025 (72.6) | 246 (30.8) | 3779 (79.6) | ||
| Surgery | .554 | < .001 | ||||||
| Yes | 4020 (84.7) | 712 (84.0) | 3308 (84.8) | 4092 (73.8) | 72 (9.0) | 4020 (84.7) | ||
| None/Unknown | 727 (15.3) | 136 (16.0) | 591 (15.2) | 1455 (26.2) | 728 (91.0) | 727 (15.3) | ||
| Follow-up time (months) | 38 (13-76) | 30 (12-70) | 40 (13-77) | .004 | 30 (9-69) | 7 (3-13) | 38 (13-76) | < .001 |
Note: Nt (%), total number of the cohort N; Ne (%), number of LNM events; Nne (%), number of non-LNM events; Mt (%), total number of the cohort M; Me (%), number of DM events; Mne (%), number of non-DM events. Abbreviation: LNM, lymph node metastasis; DM, distant metastasis.
Risk factors of LNM and its nomogram
According to univariable and multivariable logistic regression models, we finally included five significant risk factors of LNM as follows: age at diagnosis, grade, primary site, tumor size and regional nodes examined (Table 2). As for age, the oldest patients had lower risk of LNM (OR = 0.59, 95% CI = 0.46-0.76, P < .001). Compared to well differentiated T1 gastric cancer, moderately differentiated (OR = 2.00, 95% CI = 1.45-2.82, P < .001), poorly differentiated (OR = 2.97, 95% CI = 2.15-4.17, P < .001) and undifferentiated (OR = 2.18, 95% CI = 1.15-4.01, P = .014) patients all had higher risk of lymph nodes metastasis. Patients with non-cardia gastric cancer had statistical lower risk of lymph nodes metastasis than those with cardia (OR = 0.75, 95% CI = 0.63-0.90, P = .002) gastric cancer. In terms of tumor size, an increasing LNM risk was detected in lager size patients, especially for size over 30 mm (OR = 9.79, 95% CI = 6.83-14.54, P < .001).
Table 2.
Logistic regression for the associated risk factors of LNM in T1N0-3M0 gastric cancer
| Clinicopathological variables | Univariate analysis | Multivariate analysis | ||
|---|---|---|---|---|
|
|
|
|||
| OR (95% CI) | P | OR (95% CI) | P | |
| Age at diagnosis | ||||
| 18-55 | Reference | Reference | ||
| 56-65 | 0.82 (0.65-1.04) | .100 | 0.78 (0.61-1.01) | .056 |
| 66-75 | 0.84 (0.68-1.05) | .127 | 0.83 (0.66-1.06) | .134 |
| 76+ | 0.59 (0.47-0.73) | < .001 | 0.59 (0.46-0.76) | < .001 |
| Race | ||||
| White | Reference | |||
| Black | 0.94 (0.74-1.19) | .615 | ||
| Others | 0.95 (0.79-1.14) | .592 | ||
| Sex | ||||
| Female | Reference | Reference | ||
| Male | 1.26 (1.08-1.47) | .004 | 1.15 (0.96-1.37) | .127 |
| Grade | ||||
| Well differentiated | Reference | Reference | ||
| Moderately differentiated | 2.90 (2.13-4.05) | < .001 | 2.00 (1.45-2.82) | < .001 |
| Poorly differentiated | 4.74 (3.51-6.55) | < .001 | 2.97 (2.15-4.17) | < .001 |
| Undifferentiated | 3.62 (1.96-6.46) | < .001 | 2.18 (1.15-4.01) | .014 |
| Histology | ||||
| Adenocarcinoma | Reference | Reference | ||
| Signet ring cell carcinoma | 1.29 (1.05-1.58) | .013 | 0.98 (0.77-1.24) | .865 |
| Others | 1.19 (0.95-1.48) | .115 | 1.26 (0.98-1.61) | .067 |
| Primary site | ||||
| Cardia | Reference | Reference | ||
| Non-cardia | 0.83 (0.71-0.97) | .020 | 0.75 (0.63-0.90) | .002 |
| Tumor size (mm) | ||||
| 1-9 | Reference | Reference | ||
| 10-19 | 4.14 (2.85-6.20) | < .001 | 3.54 (2.42-5.34) | < .001 |
| 20-29 | 6.36 (4.38-9.54) | < .001 | 5.44 (3.72-8.22) | < .001 |
| 30+ | 10.57 (7.43-15.58) | < .001 | 9.79 (6.83-14.54) | < .001 |
| Regional nodes examined | ||||
| ≤12 | Reference | Reference | ||
| >12 | 2.21 (1.90-2.57) | < .001 | 2.11 (1.79-2.48) | < .001 |
| Marital status | ||||
| Married | Reference | Reference | ||
| Others | 0.83 (0.71-0.97) | .017 | 0.96 (0.81-1.13) | .612 |
Abbreviations: LNM, lymph node metastasis; 95% CI, 95% confidence intervals; OR, odd ratio.
A nomogram was established to show the risk factors of LNM intuitively (Figure 2A). In addition, precise score of each factor in the nomogram was displayed in Table 4. We could see that the tumor size took up the maximum proportion. ROC (Figure 2B) demonstrated a great discrimination of nomogram with AUC of 0.735 (95% CI = 0.717-0.752) and the calibration curve illustrated an effectively accuracy (Figure 2C). Furthermore, decision curve analysis and clinical impact curve showed that threshold probabilities of 0-0.5 were the best benefit by nomogram in group N (Figure 2D, 2E).
Figure 2.
Nomogram (A), receiver operating characteristic curve (B), the calibration curve (C), decision curve analysis (D), and the clinical impact curve (E) for forecasting LNM in T1N0-3M0 gastric cancer patients.
Table 4.
Nomogram score of significant factors in T1 gastric cancer
| Clinicopathological variables | Nomogram score | ||
|---|---|---|---|
|
| |||
| LNM | DM | OS | |
| Age at diagnosis | |||
| 18-55 | 24 | 42 | 0 |
| 56-65 | 14 | 24 | 6 |
| 66-75 | 16 | 17 | 25 |
| 76+ | 0 | 0 | 62 |
| Race | |||
| White | 16 | ||
| Black | 24 | ||
| Others | 0 | ||
| Sex | |||
| Female | 0 | ||
| Male | 12 | ||
| Grade | |||
| Well differentiated | 0 | 0 | 0 |
| Moderately differentiated | 30 | 18 | 17 |
| Poorly differentiated | 47 | 38 | 27 |
| Undifferentiated | 35 | 26 | 28 |
| Histology | |||
| Adenocarcinoma | 20 | ||
| Signet ring cell carcinoma | 0 | ||
| Others | 19 | ||
| Primary site | |||
| Cardia | 13 | 16 | 17 |
| Non-cardia | 0 | 0 | 0 |
| Tumor size (mm) | |||
| 1-9 | 0 | 0 | 0 |
| 10-19 | 55 | 48 | 13 |
| 20-29 | 74 | 64 | 24 |
| 30+ | 100 | 100 | 37 |
| Regional nodes examined | |||
| ≤12 | 0 | 91 | 26 |
| >12 | 33 | 0 | 0 |
| Marital status | |||
| Married | 0 | ||
| Others | 12 | ||
| Chemotherapy | |||
| Yes | 0 | ||
| None/Unknown | 26 | ||
| Surgery | |||
| Yes | 0 | ||
| None/Unknown | 100 | ||
| N stage | |||
| N0 | 0 | 0 | |
| N1 | 33 | 19 | |
| N2 | 33 | 46 | |
| N3 | 33 | 42 | |
| M stage | |||
| M0 | 0 | ||
| M1 | 63 | ||
Abbreviations: LNM, Lymph nodes metastasis; DM, Distant metastasis; OS, Overall survival.
Predictors of DM and its nomogram
Univariate and multivariate logistic regression for the presence of DM showed that seven variables - including age at diagnosis, grade, histology, primary site, tumor size, regional nodes examined, N stage - were related to DM (Table 3). The older patients were at lower risk to suffer DM (aged at 56-65, OR = 0.55, 95% CI = 0.42-0.73, P < .001; aged at 66-75, OR = 0.43, 95% CI = 0.33-0.57, P < .001; aged at 76+, OR = 0.23, 95% CI = 0.18-0.30, P < .001). The DM always happened when the grade was poorly differentiated (OR = 3.75, 95% CI = 2.46-5.93, P < .001), undifferentiated (OR = 2.49, 95% CI = 1.20-5.15, P = .014), or moderately differentiated (OR = 1.87, 95% CI = 1.21-2.98, P = .006). Compared with patients who didn’t have LNM, those suffered from LNM (N1/2/3, OR = 3.10, 95% CI = 2.54-3.78, P < .001) were at a higher risk of distant metastasis. Furthermore, we analyzed the critical risk factors associated with the specific site of distant metastasis. Since the definite metastasis site was only recorded from 2010 and onward, patients from 2010 to 2016 were enrolled. We screened 1875 T1 GC patients with a clear record of distant metastasis site from 5547 patients. Among them, there were 21 cases of bone metastasis, 5 cases of brain metastasis, 31 cases of lung metastasis, and 98 cases of liver metastasis. According to the theory of Events Per Variable (EPV) [21,22], we can only analyze the risk factors of liver metastasis. Associated risk factors to liver metastasis were calculated with univariate and multivariate logistic regression (Supplementary Table 1), which revealed that grade, histology, tumor size, regional nodes examined, N stage were all independent risk factors. Larger tumors were more prone to liver metastases (size >30 mm, OR = 21.26, 95% CI = 4.55-379.25, P = .003). Compared with non-adenocarcinoma, patients with adenocarcinoma were more likely to develop liver metastases (signet ring cell carcinoma, OR = 0.11, 95% CI = 0.03-0.30, P < .001; other type of histology, OR = 0.40, 95% CI = 0.16-0.90, P = .038). The risk factors of liver metastasis were almost the same as those of distant metastasis, which showed the credibility of our study.
Table 3.
Logistic regression for the associated risk factors of DM in T1 gastric cancer
| Clinicopathological variables | Univariate analysis | Multivariate analysis | ||
|---|---|---|---|---|
|
|
|
|||
| OR (95% CI) | P | OR (95% CI) | P | |
| Age at diagnosis | ||||
| 18-55 | Reference | Reference | ||
| 56-65 | 0.69 (0.56-0.86) | .001 | 0.55 (0.42-0.73) | < .001 |
| 66-75 | 0.48 (0.39-0.59) | < .001 | 0.43 (0.33-0.57) | < .001 |
| 76+ | 0.36 (0.29-0.45) | < .001 | 0.23 (0.18-0.30) | < .001 |
| Race | ||||
| White | Reference | Reference | ||
| Black | 0.88 (0.69-1.10) | .272 | 0.83 (0.62-1.10) | .203 |
| Others | 0.51 (0.41-0.64) | < .001 | 0.85 (0.64-1.11) | .234 |
| Sex | ||||
| Female | Reference | Reference | ||
| Male | 1.41 (1.20-1.66) | < .001 | 0.89 (0.73-1.10) | .281 |
| Grade | ||||
| Well differentiated | Reference | Reference | ||
| Moderately differentiated | 3.52 (2.38-5.41) | < .001 | 1.87 (1.21-2.98) | .006 |
| Poorly differentiated | 7.16 (4.92-10.88) | < .001 | 3.75 (2.46-5.93) | < .001 |
| Undifferentiated | 6.04 (3.23-11.15) | < .001 | 2.49 (1.20-5.15) | .014 |
| Histology | ||||
| Adenocarcinoma | Reference | Reference | ||
| Signet ring cell carcinoma | 0.76 (0.60-0.95) | .018 | 0.50 (0.37-0.67) | < .001 |
| Others | 0.94 (0.75-1.18) | .613 | 0.98 (0.72-1.31) | .879 |
| Primary site | ||||
| Cardia | Reference | Reference | ||
| Non-cardia | 0.40 (0.34-0.46) | < .001 | 0.59 (0.48-0.72) | < .001 |
| Tumor size (mm) | ||||
| 1-9 | Reference | Reference | ||
| 10-19 | 5.73 (2.79-13.85) | < .001 | 5.26 (2.52-12.86) | < .001 |
| 20-29 | 12.19 (6.05-29.11) | < .001 | 9.46 (4.59-22.91) | < .001 |
| 30+ | 49.88 (25.52-116.86) | < .001 | 32.66 (16.37-77.51) | < .001 |
| Regional nodes examined | ||||
| ≤12 | Reference | Reference | ||
| >12 | 0.06 (0.04-0.08) | < .001 | 0.04 (0.03-0.06) | < .001 |
| Marital status | ||||
| Married | Reference | |||
| Others | 0.90 (0.77-1.05) | .192 | ||
| N stage | ||||
| N0 | Reference | Reference | ||
| Non-N0 (N1/2/3) | 4.37 (3.74-5.12) | < .001 | 3.10 (2.54-3.78) | < .001 |
Abbreviations: DM, distant metastasis; 95% CI, 95% confidence intervals; OR, odd ratio.
On the basis of independent risk factors, a nomogram was produced in M group to predict DM (Figure 3A). Each factor possessed a specific score in Table 4. Tumor size also accounted for the largest portion, followed by regional nodes examined, age at diagnosis, grade, N stage, histology, primary site. ROC analysis (Figure 3B) was conducted in order to ensure that the nomogram had brilliant efficacy. The nomogram for the prediction of DM had a nearly perfect AUC which was 0.896 (95% CI = 0.885-0.907). The calibration curve did not elucidate obvious deviation from reference line, which indicated an advantageous predictive accuracy (Figure 3C). Moreover, according to DCA and CIC, we discovered that the most beneficial threshold probabilities for forecasting DM in group M were 0-0.8 (Figure 3D, 3E). And we also constructed the corresponding Nomogram (Supplementary Figure 1A), receiver operating characteristic curve (Supplementary Figure 1B), the calibration curve (Supplementary Figure 1C), decision curve analysis (Supplementary Figure 1D), and the clinical impact curve (Supplementary Figure 1E) for forecasting liver metastasis in T1NXM0-1 gastric cancer patients. The AUC (0.896, 95% CI = 0.885-0.907) was also excellent.
Figure 3.
Nomogram (A), receiver operating characteristic curve (B), the calibration curve (C), decision curve analysis (D), and the clinical impact curve (E) for forecasting DM in T1NXM0-1 gastric cancer patients.
Survival analyses on LNM and DM
We used the Kaplan-Meier and Gray method to calculate the influence of LNM and DM on the survival. Overall survival was significantly connected with LNM (HR = 1.38, 95% CI = 1.23-1.55, P < .0001) and DM (HR = 6.32, 95% CI = 5.70-7.00, P < .0001) according to Kaplan-Meier curves (Figure 4A, 4B). In terms of cancer-specific death, LNM (SHR = 1.86, 95% CI = 1.63-2.12, P < .0001) and DM (SHR = 8.24, 95% CI = 7.44-9.13, P < .0001) were also associated with CSS (cancer-specific survival) using Gray method (Figure 4C, 4D).
Figure 4.
Impact of lymph nodes metastasis on overall survival (A) and cancer-specific survival (C) in T1 gastric cancer. Effect of distant metastasis on overall survival (B) and cancer-specific survival (D) in T1 gastric cancer.
Construction of a prognostic nomogram in T1 gastric cancer
Based on univariate and multivariate Cox proportional hazard regression, we revealed prognostic factors of overall survival. The prognostic factors consisted of age at diagnosis, race, sex, grade, primary site, tumor size, regional nodes examined, marital status, chemotherapy, surgery, N stage and M stage. We made a forest plot to display the results of multivariate Cox proportional hazard model more intuitively (Figure 5). Patients who were 66-75 years old (HR = 1.42, 95% CI = 1.24-1.63, P < .001) or 76+ years old (HR = 2.36, 95% CI = 2.07-2.69, P < .001) would have higher risk of death compared to those aged at 18-55. The probability of death increased in patients who suffered LNM (N1, HR = 1.31, 95% CI = 1.18-1.45, P < .001; N2, HR = 1.91, 95% CI = 1.45-2.52, P < .001; N3, HR = 1.78, 95% CI = 1.15-2.76, P = .01) or DM (HR = 2.41, 95% CI = 2.14-2.71, P < .001).
Figure 5.
Forest plot depicting the significance of multivariate Cox proportional hazard regression prognostic factors of OS in T1 gastric cancer.
Then, the study further researched independent risk factors of gastric cancer-specific death (GCSD) in T1 patients by performing competing risk model. Ten variables were included: age at diagnosis, race, grade, primary site, tumor size, regional nodes examined, chemotherapy, surgery, N stage and M stage (Table 5). In matters of age, the older patients (aged more than 76 years old, SHR = 1.40, 95% CI = 1.21-1.63, P < .001) had higher risk of GCSD when compared to those younger. The worsen grade had a gradual rising risk of GCSD (moderately differentiated, SHR = 1.30, 95% CI = 1.05-1.60, P = .014; poorly differentiated, SHR = 1.59, 95% CI = 1.29-1.96, P < .001; undifferentiated, SHR = 1.62, 95% CI = 1.12-2.33, P = .010) when well differentiated was regarded as reference. Consistently, the probability of GCSD increased in patients who suffered LNM (N1, SHR = 1.37, 95% CI = 1.20-1.55, P < .001; N2, SHR = 2.17, 95% CI = 1.62-2.90, P < .001; N3, SHR = 1.84, 95% CI = 1.22-2.77, P = .004) or DM (SHR = 2.24, 95% CI = 1.93-2.62, P < .001).
Table 5.
Competing risk regression analysis for risk factors of CSS in T1 gastric cancer
| Clinicopathological variables | Univariate analysis | Multivariate analysis | ||
|---|---|---|---|---|
|
|
|
|||
| SHR (95% CI) | P | SHR (95% CI) | P | |
| Age at diagnosis | ||||
| 18-55 | Reference | Reference | ||
| 56-65 | 0.89 (0.77-1.04) | .150 | 0.97 (0.84-1.12) | .700 |
| 66-75 | 0.79 (0.69-0.92) | .002 | 0.99 (0.85-1.15) | .900 |
| 76+ | 1.22 (1.06-1.40) | .004 | 1.40 (1.21-1.63) | < .001 |
| Race | ||||
| White | Reference | Reference | ||
| Black | 0.90 (0.77-1.04) | .150 | 0.97 (0.82-1.14) | .670 |
| Others | 0.57 (0.50-0.65) | < .001 | 0.86 (0.75-0.98) | .023 |
| Sex | ||||
| Female | Reference | Reference | ||
| Male | 1.30 (1.18-1.44) | < .001 | 1.09 (0.98-1.21) | .110 |
| Grade | ||||
| Well differentiated | Reference | Reference | ||
| Moderately differentiated | 2.14 (1.75-2.62) | < .001 | 1.30 (1.05-1.60) | .014 |
| Poorly differentiated | 3.06 (2.52-3.73) | < .001 | 1.59 (1.29-1.96) | < .001 |
| Undifferentiated | 3.31 (2.35-4.66) | < .001 | 1.62 (1.12-2.33) | .010 |
| Histology | ||||
| Adenocarcinoma | Reference | Reference | ||
| Signet ring cell carcinoma | 0.88 (0.77-1.02) | .084 | 1.08 (0.93-1.25) | .320 |
| Others | 0.85 (0.73-0.99) | .041 | 1.00 (0.85-1.18) | .980 |
| Primary site | ||||
| Cardia | Reference | Reference | ||
| Non-cardia | 0.52 (0.47-0.57) | < .001 | 0.84 (0.75-0.94) | .003 |
| Tumor size (mm) | ||||
| 1-9 | Reference | Reference | ||
| 10-19 | 1.71 (1.34-2.18) | < .001 | 1.42 (1.10-1.82) | .007 |
| 20-29 | 2.87 (2.27-3.64) | < .001 | 1.89 (1.48-2.43) | < .001 |
| 30+ | 6.34 (5.10-7.87) | < .001 | 2.22 (1.75-2.82) | < .001 |
| Regional nodes examined | ||||
| ≤12 | Reference | Reference | ||
| >12 | 0.27 (0.24-0.31) | < .001 | 0.71 (0.61-0.83) | < .001 |
| Marital status | ||||
| Married | Reference | |||
| Others | 1.09 (0.99-1.20) | .069 | ||
| Radiation | ||||
| Yes | Reference | Reference | ||
| None/Unknown | 0.52 (0.46-0.57) | < .001 | 1.02 (0.89-1.16) | .800 |
| Chemotherapy | ||||
| Yes | Reference | Reference | ||
| None/Unknown | 0.35 (0.32-0.38) | < .001 | 1.18 (1.02-1.36) | .025 |
| Surgery | ||||
| Yes | Reference | Reference | ||
| None/Unknown | 9.38 (8.48-10.40) | < .001 | 4.26 (3.65-4.97) | < .001 |
| N stage | ||||
| N0 | Reference | Reference | ||
| N1 | 2.25 (2.03-2.49) | < .001 | 1.37 (1.20-1.55) | < .001 |
| N2 | 2.31 (1.75-3.04) | < .001 | 2.17 (1.62-2.90) | < .001 |
| N3 | 3.88 (2.69-5.58) | < .001 | 1.84 (1.22-2.77) | .004 |
| M stage | ||||
| M0 | Reference | Reference | ||
| M1 | 8.24 (7.44-9.13) | < .001 | 2.24 (1.93-2.62) | < .001 |
Abbreviations: CSS, cancer-specific survival; 95% CI, 95% confidence intervals; SHR, sub-distribution hazard ration.
Based on the COX regression model, the 3-, 5- and 10-year OS prognostic nomogram has shown in Figure 6A. We can calculate the T1 GC patients’ probabilities of 3-, 5-, and 10-year OS by summing up each factor’s score. The C-index was 0.797 and the calibration curves illustrated brilliant agreement between the prediction and practical situation (Figure 6B-D). Additionally, the DCA curves showed that the optimal threshold probabilities were 0-0.6, 0-0.7, 0-0.8 for forecasting the 3-, 5-, 10-year OS (Figure 6E-G), respectively.
Figure 6.
Nomogram (A), the calibration curve (B-D) and decision curve analysis (E-G) for forecasting overall survival in T1 gastric cancer patients.
Clinical effects
Finally, we used the interquartile range of the nomogram score to reclassify group N into three subgroups (low-, medium-, and high-risk). The score was 0-100, 101-160, 161+ for each subgroup. Occurrence of LNM amid three subgroups was statistically different (P < .001) (Figure 7A). In a similar way to reclassify group M, the incidence of DM among subgroups demonstrated the statistical significance (P < .001) (Figure 7B). And the probability of OS also showed a significantly difference among these subgroups (P < .001) (Figure 7C).
Figure 7.
Stacked bar charts of clinical effect on LNM nomogram (A) and DM nomogram (B). Kaplan-Meier curve of clinical effect on OS nomogram (C).
Discussion
Because of various clinical features and prognoses, patients with T1 gastric cancer always need distinct treatments. In terms of T1 GCs without LNM, patients just need to accept endoscopy resection to remove the lesions. It can acquire similar beneficial prognosis compared with radical surgery but decrease the occurrence of adverse effects. By contrast, T1 GCs with LNM, making up 20.5% roughly [5], ought to accept radical surgical resection and undergo adjuvant chemotherapy. Unfortunately, T1 GCs with DM lose the opportunity of surgical therapy and have to experience systemic therapy, such as palliative chemotherapy, molecular targeted anti-tumor drugs and immunotherapy [6,17]. In a word, it is important to forecast the probability of LNM and DM in T1 GCs. Certainly, it is significant to calculate the probability of overall survival based on clinical features in T1 GCs, too.
We can see increasing related studies recently, but there are still some disadvantages and limitations. First of all, although earlier researches [23,24] established models through regression methods, all of them didn’t possess the capacity of prediction since rendering models were difficult to exert in real world conveniently. We used the nomograms to display the prediction of lymph node metastasis, distant metastasis and overall survival intuitively and make further validation and clinical decision. Secondly, in the researches including nomograms of T1 GC population [25-27], they were still lack of completeness and were not able to comprehensively reflect the prognosis of T1 GCs in one dataset, which only study the probability of LNM or DM, ignoring the survival outcomes of this population. Thirdly, the inclusion criteria seem to be confused. For instance, several studies used the whole data of T1 GCs directly [28,29]. Nonetheless, they all had predictive results of lymph node metastasis, it is lack of rigor, because only nonmetastatic GC patients have a clinical value in forecasting lymph node metastasis.
Base on these, we reclassified the included cases into N group (T1N0-3M0 GCs for LNM) and M group (T1N0-3M0-1 GCs for DM and OS). Not only the risk factors of LNM and DM but also the prognostic predictors of OS for T1 GCs were all constructed, with the parallel nomograms plotted too. These nomograms were further examined by several methods such as ROC, calibration curve, DCA and CIC. LNM nomogram involves five factors: age at diagnosis, grade, primary site, tumor size and reginal nodes examined while DM nomogram involves seven factors: age at diagnosis, histology, grade, primary site, tumor size, reginal nodes examined and N stage. 3-, 5- and 10-year OS nomogram includes twelve predictors: age at diagnosis, race, sex, grade, primary site, tumor size, reginal nodes examined, marital status, chemotherapy, surgery, N stage and M stage. Furthermore, we analyzed the critical risk factors associated to the specific site of distant metastasis. Due to data limitations, we can only analyze liver metastasis. Nomogram of liver metastasis involves five factors: histology, grade, tumor size, reginal nodes examined and N stage, which was almost consistent with the risk factors of DM.
All these nomograms illustrated great discrimination and accuracy. AUC of the LNM and DM nomograms were 0.735 and 0.896, respectively. And C-index of OS nomogram was 0.797. Good clinical utility could be seen in the corresponding proper threshold probability. Moreover, we forged low-, medium- and high-risk subgroups according to interquartile scores of nomograms along with plotted stacked bar charts and K-M survival curve to show the ability of nomograms discrimination intuitively.
In term of the LNM nomogram, we discovered that tumor size and grade took up the largest portion of nomogram scores, which was consistent with prior studies [26,29]. The larger tumor size was, the more probability of occurrence of LNM was. Compared with well differentiated gastric cancer, the probability of LNM in the poorly differentiated and undifferentiated cancer were 2.97 and 2.18 (both P < .05). Unlike Jiang et al’s argument that younger age increased risk of LNM [30], we didn’t find a significant relationship between age and lymph node metastasis. In term of DM nomogram, tumor size accounted for the largest percentage of nomogram scores. We already have seen this parameter confirmed as a significant prognostic factor in solid cancers [31-33]. Interestingly, we found that there was an inverse association of distant metastasis with age in T1 GC population. Above all, the relationship between age and LNM or DM will need a further study, which indicates our next efforts. For OS nomogram, the largest proportion in nomogram were surgery and age at diagnosis. It is not surprising that for T1 GCs who did not have surgery, the risk for death would increase. Consistent with our study, previous studies [23,34] reported age as one independent prognostic predictor affecting the prognosis of GC. The OS for elderly patients was shorter when compared with younger patients.
In present study, we screened 5,547 eligible cases from SEER database and analyzed the cases relying on suitable statistical methods. What’s more, we concluded these convincing inferences. Our study yet has some shortcomings - this retrospective analysis is lack of the specific site of distant metastasis and some critical risk factors such as the infection of HP, the detection of key molecules like Her-2. Additionally, all of these models were not validated by external cohort, which remained continuously improved on the basis of future’s application. We hope this article can be an inspiration and provide reference for clinical trials to obtain more clinical cohort data from large centers to verify our models. So as to promote the progress of individualized treatment.
Conclusion
In conclusion, we established corresponding nomograms of LNM, DM and OS of T1 gastric cancer population depending on independent risk factors through logistic and COX regression analyses. Furthermore, all these nomograms were verified to acquire reliable ability of discrimination and accuracy. Besides, they also had brilliant profit in clinical utility. We hope that our results can assist doctors in making individual clinical decisions for T1 GC patients and promote the process of individualized therapy.
Acknowledgements
This work was funded by the National project of clinical collaboration of traditional Chinese and western medicine for the treatment of major diseases (Bin Lv, Gastric Cancer, http://yzs.satcm.gov.cn), Top ten thousand talents program of Zhejiang Province (Shanming Ruan, no. 2019-97, http://www.zjzzgz.gov.cn/), Zhejiang Provincial Program for the Cultivation of the Young and Middle-Aged Academic Leaders in Colleges and Universities (Shanming Ruan, no. 2017-248, http://www.zjedu.gov.cn/), Zhejiang Provincial Project for the key discipline of Traditional Chinese Medicine (Yong Guo, no. 2017-XK-A09, http://www.zjwjw.gov.cn/). We thank the Surveillance, Epidemiology, and End Results database for the tremendous contribution. We also thank Konstantinos for the help in language polishing.
Disclosure of conflict of interest
None.
Supporting Information
References
- 1.Bray F, Ferlay J, Soerjomataram I, Siegel RL, Torre LA, Jemal A. Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin. 2018;68:394–424. doi: 10.3322/caac.21492. [DOI] [PubMed] [Google Scholar]
- 2.Siegel RL, Miller KD, Jemal A. Cancer statistics, 2020. CA Cancer J Clin. 2020;70:7–30. doi: 10.3322/caac.21590. [DOI] [PubMed] [Google Scholar]
- 3.Oh SY, Lee JH, Lee HJ, Kim TH, Huh YJ, Ahn HS, Suh YS, Kong SH, Kim GH, Ahn SJ, Kim SH, Choi Y, Yang HK. Natural history of gastric cancer: observational study of gastric cancer patients not treated during follow-up. Ann Surg Oncol. 2019;26:2905–2911. doi: 10.1245/s10434-019-07455-z. [DOI] [PubMed] [Google Scholar]
- 4.Everett SM, Axon AT. Early gastric cancer: disease or pseudo-disease? Lancet. 1998;351:1350–1352. doi: 10.1016/s0140-6736(98)04365-7. [DOI] [PubMed] [Google Scholar]
- 5.Hanada Y, Choi AY, Hwang JH, Draganov PV, Khanna L, Sethi A, Bartel MJ, Goel N, Abe S, De Latour RA, Park K, Melis M, Newman E, Hatzaras I, Reddy SS, Farma JM, Liu X, Schlachterman A, Kresak J, Trapp G, Ansari N, Schrope B, Lee JY, Dhall D, Lo S, Jamil LH, Burch M, Gaddam S, Gong Y, Del Portillo A, Tomizawa Y, Truong CD, Brewer Gutierrez OI, Montgomery E, Johnston FM, Duncan M, Canto M, Ahuja N, Lennon AM, Ngamruengphong S. Low frequency of lymph node metastases in patients in the united states with early-stage gastric cancers that fulfill japanese endoscopic resection criteria. Clin Gastroenterol Hepatol. 2019;17:1763–1769. doi: 10.1016/j.cgh.2018.11.031. [DOI] [PubMed] [Google Scholar]
- 6.Smyth EC, Nilsson M, Grabsch HI, van Grieken NC, Lordick F. Gastric cancer. Lancet. 2020;396:635–648. doi: 10.1016/S0140-6736(20)31288-5. [DOI] [PubMed] [Google Scholar]
- 7.Draganov PV, Wang AY, Othman MO, Fukami N. AGA institute clinical practice update: endoscopic submucosal dissection in the United States. Clin Gastroenterol Hepatol. 2019;17:16–25. e11. doi: 10.1016/j.cgh.2018.07.041. [DOI] [PubMed] [Google Scholar]
- 8.Choi IJ, Lee JH, Kim YI, Kim CG, Cho SJ, Lee JY, Ryu KW, Nam BH, Kook MC, Kim YW. Long-term outcome comparison of endoscopic resection and surgery in early gastric cancer meeting the absolute indication for endoscopic resection. Gastrointest Endosc. 2015;81:333–341. e331. doi: 10.1016/j.gie.2014.07.047. [DOI] [PubMed] [Google Scholar]
- 9.Pyo JH, Lee H, Min BH, Lee JH, Choi MG, Lee JH, Sohn TS, Bae JM, Kim KM, Ahn JH, Carriere KC, Kim JJ, Kim S. Long-term outcome of endoscopic resection vs. surgery for early gastric cancer: a non-inferiority-matched cohort study. Am J Gastroenterol. 2016;111:240–249. doi: 10.1038/ajg.2015.427. [DOI] [PubMed] [Google Scholar]
- 10.Pourmousavi MK, Wang R, Kerdsirichairat T, Kamal A, Akshintala VS, Hajiyeva G, Lopimpisuth C, Hanada Y, Kumbhari V, Singh VK, Khashab MA, Brewer OG, Shin EJ, Canto MI, Lennon AM, Ngamruengphong S. Comparable cancer-specific mortality of patients with early gastric cancer treated with endoscopic therapy vs. surgical resection. Clin Gastroenterol Hepatol. 2020;18:2824–2832. e2821. doi: 10.1016/j.cgh.2020.04.085. [DOI] [PubMed] [Google Scholar]
- 11.Ono H, Yao K, Fujishiro M, Oda I, Nimura S, Yahagi N, Iishi H, Oka M, Ajioka Y, Ichinose M, Matsui T. Guidelines for endoscopic submucosal dissection and endoscopic mucosal resection for early gastric cancer. Dig Endosc. 2016;28:3–15. doi: 10.1111/den.12518. [DOI] [PubMed] [Google Scholar]
- 12.Hahn KY, Park CH, Lee YK, Chung H, Park JC, Shin SK, Lee YC, Kim HI, Cheong JH, Hyung WJ, Noh SH, Lee SK. Comparative study between endoscopic submucosal dissection and surgery in patients with early gastric cancer. Surg Endosc. 2018;32:73–86. doi: 10.1007/s00464-017-5640-8. [DOI] [PubMed] [Google Scholar]
- 13.Jeon HK, Kim GH, Lee BE, Park DY, Song GA, Kim DH, Jeon TY. Long-term outcome of endoscopic submucosal dissection is comparable to that of surgery for early gastric cancer: a propensity-matched analysis. Gastric Cancer. 2018;21:133–143. doi: 10.1007/s10120-017-0719-4. [DOI] [PubMed] [Google Scholar]
- 14.Wang S, Zhang Z, Liu M, Li S, Jiang C. Endoscopic resection compared with gastrectomy to treat early gastric cancer: a systematic review and meta-analysis. PLoS One. 2015;10:e0144774. doi: 10.1371/journal.pone.0144774. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Gotoda T, Yanagisawa A, Sasako M, Ono H, Nakanishi Y, Shimoda T, Kato Y. Incidence of lymph node metastasis from early gastric cancer: estimation with a large number of cases at two large centers. Gastric Cancer. 2000;3:219–225. doi: 10.1007/pl00011720. [DOI] [PubMed] [Google Scholar]
- 16.Guo A, Du C, Tian S, Sun L, Guo M, Lu L, Peng L. Long-term outcomes of endoscopic submucosal dissection versus surgery for treating early gastric cancer of undifferentiated-type. Medicine (Baltimore) 2020;99:e20501. doi: 10.1097/MD.0000000000020501. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Japanese Gastric Cancer Association. Japanese gastric cancer treatment guidelines 2018 (5th edition) Gastric Cancer. 2021;24:1–21. doi: 10.1007/s10120-020-01042-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Tolles J, Meurer WJ. Logistic regression: relating patient characteristics to outcomes. JAMA. 2016;316:533–534. doi: 10.1001/jama.2016.7653. [DOI] [PubMed] [Google Scholar]
- 19.Balachandran VP, Gonen M, Smith JJ, DeMatteo RP. Nomograms in oncology: more than meets the eye. Lancet Oncol. 2015;16:e173–180. doi: 10.1016/S1470-2045(14)71116-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Vickers AJ, Elkin EB. Decision curve analysis: a novel method for evaluating prediction models. Med Decis Making. 2006;26:565–74. doi: 10.1177/0272989X06295361. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Peduzzi P, Concato J, Kemper E, Holford TR, Feinstein AR. A simulation study of the number of events per variable in logistic regression analysis. J Clin Epidemiol. 1996;49:1373–1379. doi: 10.1016/s0895-4356(96)00236-3. [DOI] [PubMed] [Google Scholar]
- 22.Wynants L, Bouwmeester W, Moons KG, Moerbeek M, Timmerman D, Van Huffel S, Van Calster B, Vergouwe Y. A simulation study of sample size demonstrated the importance of the number of events per variable to develop prediction models in clustered data. J Clin Epidemiol. 2015;68:1406–1414. doi: 10.1016/j.jclinepi.2015.02.002. [DOI] [PubMed] [Google Scholar]
- 23.Han J, Tu J, Tang C, Ma X, Huang C. Clinicopathological characteristics and prognosis of cT1N0M1 gastric cancer: a population-based study. Dis Markers. 2019;2019:5902091. doi: 10.1155/2019/5902091. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Sun Z, Liu H, Yu J, Huang W, Han Z, Lin T, Chen H, Zhao M, Hu Y, Jiang Y, Li G. Frequency and prognosis of pulmonary metastases in newly diagnosed gastric cancer. Front Oncol. 2019;9:671. doi: 10.3389/fonc.2019.00671. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Guo CG, Zhao DB, Liu Q, Zhou ZX, Zhao P, Wang GQ, Cai JQ. A nomogram to predict lymph node metastasis in patients with early gastric cancer. Oncotarget. 2017;8:12203–12210. doi: 10.18632/oncotarget.14660. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Zhao LY, Yin Y, Li X, Zhu CJ, Wang YG, Chen XL, Zhang WH, Chen XZ, Yang K, Liu K, Zhang B, Chen ZX, Chen JP, Zhou ZG, Hu JK. A nomogram composed of clinicopathologic features and preoperative serum tumor markers to predict lymph node metastasis in early gastric cancer patients. Oncotarget. 2016;7:59630–59639. doi: 10.18632/oncotarget.10732. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Zheng Z, Zhang Y, Zhang L, Li Z, Wu X, Liu Y, Bu Z, Ji J. A nomogram for predicting the likelihood of lymph node metastasis in early gastric patients. BMC Cancer. 2016;16:92. doi: 10.1186/s12885-016-2132-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Mu J, Jia Z, Yao W, Song J, Cao X, Jiang J, Wang Q. Predicting lymph node metastasis in early gastric cancer patients: development and validation of a model. Future Oncol. 2019;15:3609–3617. doi: 10.2217/fon-2019-0377. [DOI] [PubMed] [Google Scholar]
- 29.Yin XY, Pang T, Liu Y, Cui HT, Luo TH, Lu ZM, Xue XC, Fang GE. Development and validation of a nomogram for preoperative prediction of lymph node metastasis in early gastric cancer. World J Surg Oncol. 2020;18:2. doi: 10.1186/s12957-019-1778-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Jiang Y, Huang W, Xie J, Han Z, Chen C, Xi S, Sun Z, Hu Y, Zhao L, Yu J, Li T, Zhou Z, Cai S, Li G. Young age increases risk for lymph node positivity in gastric cancer: a Chinese multi-institutional database and US SEER database study. J Cancer. 2020;11:678–685. doi: 10.7150/jca.37531. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Tran B, Roshan D, Abraham E, Wang L, Garibotto N, Wykes J, Campbell P, Ebrahimi A. The prognostic impact of tumor size in papillary thyroid carcinoma is modified by age. Thyroid. 2018;28:991–996. doi: 10.1089/thy.2017.0607. [DOI] [PubMed] [Google Scholar]
- 32.Wang HM, Huang CM, Zheng CH, Li P, Xie JW, Wang JB, Lin JX, Lu J. Tumor size as a prognostic factor in patients with advanced gastric cancer in the lower third of the stomach. World J Gastroenterol. 2012;18:5470–5475. doi: 10.3748/wjg.v18.i38.5470. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Wang X, Wan F, Pan J, Yu GZ, Chen Y, Wang JJ. Tumor size: a non-neglectable independent prognostic factor for gastric cancer. J Surg Oncol. 2008;97:236–240. doi: 10.1002/jso.20951. [DOI] [PubMed] [Google Scholar]
- 34.Yagi Y, Seshimo A, Kameoka S. Prognostic factors in stage IV gastric cancer: univariate and multivariate analyses. Gastric Cancer. 2000;3:71–80. doi: 10.1007/pl00011699. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.