The patient data in this study were extracted from the Surveillance, Epidemiology, and End Results (SEER) database between 2010 and 2018. The SEER database collects information on cancer statistics, treatment, and survival, covering approximately 48% of the population in the United States [16]. Patients with the following criteria were included: (1) pathologically diagnosed with invasive BC; (2) White American or Chinese individuals with BC; (3) having detailed information on age, TNM stage, ER status, PR status, HER2 status, histology subtype, tumor grade, surgery, radiotherapy, and chemotherapy administration. Male patients, patients with contralateral BC, and those diagnosed with a distant metastatic stage were excluded from this study.

We selected the variables including age (<50, 50-70, and >70 years), race and ethnicity (White American and Chinese), grade (well-, moderately-, poorly differentiated, and undifferentiated), histological subtype (infiltrating duct carcinoma, lobular carcinoma, mixed, and other carcinomas), molecular subtype (luminal A, luminal B, HER2-enriched, and triple-negative BC), T stage (T0-T4), N stage (N0-N3), and the seventh and eighth AJCC staging (IA-IIIC). The races and ethnicities of White American (code 01) and Chinese (code 04) were chosen for analysis using “Race/ethnicity” codes in the SEER database. The end point of this study was CSS, which was calculated as the time from BC diagnosis to the occurrence of BC-related death.

This study was approved by the ethics committee of the First Affiliated Hospital of Xiamen University (Xiamen) and West China Hospital, Sichuan University (Chengdu; 2021GGB027). Informed consent is not required because the data were extracted from the SEER database after obtaining permission from the administrator. In addition, the privacy of the participants was well protected through anonymization and deidentification of their information.

The Pearson chi-square test was used to compare the differences in baseline characteristics and stage migration changes between groups of Chinese and White American individuals with BC. The receiver operating characteristic (ROC) curve was used to identify the discriminating ability of the AS and PPS. The Kaplan–Meier method was used to plot the survival curves, and the log-rank test was used to compare the differences. Multivariate Cox proportional hazards models were used to calculate the independent risk predictors of CSS. Sensitivity analyses were used to investigate the effect of race on CSS after stratification by different AJCC substages. SPSS (version 22.0; IBM Corp) was used for analyzing all the data. A P value less than .05 was defined as the threshold for statistical significance.

In total, 376,818 individuals with BC were identified in this study. Of them, 369,522 (98.1%) were White American and 7296 (1.9%) were Chinese. The proportions of patients aged <50, 50-70, and ≥70 years were 18.3% (n=68,833), 50.9% (n=191,831), and 30.8% (n=116,154), respectively. The majority of the patients had infiltrating ductal carcinoma (n=294,359, 78.1%), a low-grade tumor (n=279,429, 74.2%), luminal A subtype BC (n=288,545, 76.6%), T1-2 stage BC (n=347,931, 92.3%), and N0-1 stage BC (n=351,666, 93.3%).

Regarding the distribution of baseline characteristics between White American and Chinese individuals, the latter were more likely to be younger (P<.001) and have a lower-grade tumor (P<.001), infiltrating duct carcinoma (P<.001), luminal B or HER2-enriched subtype (P<.001), and a lower-stage tumor (P<.001) than White American participants. Among all participants, stage IA, IB, IIA, IIB, IIIA, IIIB, and IIIC disease accounted for 53.3% (n=200,926), 2.3% (n=8,569), 23.0% (n=86,651), 11.1% (n=41,837), 5.9% (n=22,237), 2.0% (n=7649), and 2.4% (n=8949) of cases in the seventh AJCC AS and for 66.1% (n=249,173), 15.5% (n=58,405), 7.8% (n=29,379), 3.2% (n=12,116), 3.9% (n=14,800), 1.9% (n=7175), and 1.5% (n=5770) of cases in the eighth AJCC PPS, respectively.

With regard to the treatments, 96.1% (n=362,189) of the participants received surgical intervention, 52.7% (n=198,399) of them received radiotherapy, and 36.7% (n=139,160) of them received chemotherapy. White American participants were more likely to undergo breast-conserving surgery, while more Chinese participants were treated with mastectomy (P<.001). In addition, Chinese participants were more prone to receiving chemotherapy (P<.001), while White American patients were more likely to be treated with radiotherapy (P<.001). Detailed information on the study population is presented in Table 1.

According to the seventh AJCC AS, 197,231 (53.4%), 8422 (2.3%), 84,765 (22.9%), 40,962 (11.1%), 21,827 (5.9%), 7523 (2.0%), and 8792 (2.4%) White American participants versus 3695 (50.6%), 147 (2.0%), 1886 (25.8%), 875 (12.0%), 410 (5.6%), 126 (1.7%), and 157 (2.2%) Chinese participants had stage IA, IB, IIA, IIB, IIIA, IIIB, and IIIC disease, respectively (P<.001). According to the eighth AJCC PPS, 244,477 (66.2%), 57,236 (15.5%), 28,722 (7.8%), 11,843 (3.2%), 14,538 (3.9%), 7023 (1.9%), and 5683 (1.5%) White American participants versus 4696 (64.4%), 1169 (16.0%), 657 (9.0%), 273 (3.7%), 262 (3.6%), 152 (2.1%), and 87 (1.2%) Chinese participants had stage IA, IB, IIA, IIB, IIIA, IIIB, and IIIC disease, respectively (P<.001; Table 1). A total of 149,452 (39.7%) participants migrated from the seventh AJCC AS to the eighth AJCC PPS (n=22,516, 6.0% upstaged and n=126,936, 33.7% downstaged). Among the upstaged participants, 22,127 (6.0%) were White American and 389 (6.3%) were Chinese, while among the downstaged participants, 124,368 (33.6%) were White American and 2586 (35.2%) were Chinese (P=.004). Furthermore, the disease stages of 223,027 (60.4%) White American and 4339 (59.5%) Chinese participants remained unchanged. There was a significant difference in stage migration (upstaging, downstaging, and unchanging stage) between White American and Chinese participants (P=.004). The frequencies of stage discrepancies between White American and Chinese participants are shown in Table 2.

With a median follow-up duration of 44 months (range 0-107 months), 42,522 deaths and 17,807 breast cancer–related deaths occurred. The Kaplan–Meier curves showed that 5-year overall survival and CSS for the entire group were 87.4% and 95.9%, respectively. The seventh AJCC AS (P<.001; Figure 1A) and the eighth AJCC PPS (P<.001; Figure 1B) could significantly predict the survival outcome of BC. Multivariate Cox proportional hazards analysis revealed that the AS and PPS both had significant prognostic predicting value in the study population (Table 3). Kaplan–Meier curves (Figure 1) and ROC curve (area under the curve [AUC] 0.769 vs 0.753, P<.001; Figure 2A) indicated that the eighth AJCC PPS had better discriminating ability than the seventh AJCC AS.

In White American participants, the AJCC AS (P<.001; Figure 3A) and the AJCC PPS (P<.001; Figure 3B) could also significantly predict survival and prognosis consistent with stages, and multivariate analysis showed that the 2 staging systems were significant prognostic predictors of CSS (Table 4). In addition, the eighth AJCC PPS had better discriminating ability than the seventh AJCC AS in White American participants (AUC 0.769 vs 0.753, P<.001; Figure 2B). Similar results were obtained for Chinese participants, in that the seventh AJCC AS (P<.001; Figure 4A) and the eighth AJCC PPS (P<.001; Figure 4B) both had significant prognostic values (Table 4), and the eighth AJCC PPS still showed better discriminating ability (AUC 0.790 vs 0.776, P<.001; Figure 2C) in this population.

According to the seventh AJCC AS, Chinese women had a better 5-year CSS in stage IA (98.6% vs 99.1%, P=.01), stage IIA (95.8% vs 97.7%, P<.001), and stage IIIB (78.0% vs 86.5%, P=.01) disease than White American women, while no significant 5-year CSS was observed in those with stage IB (98.1% vs 99.3%, P=.37), IIB (93.1% vs 93.8%, P=.69), IIIA (88.9% vs 90.7%, P=.92), and IIIC (78.4% vs 79.6%, P=.93) disease (Figure 5). Multivariate Cox proportional hazards analysis revealed that race was an independent predictor in stage IA (hazard ratio [HR] 0.679, 95% CI 0.491-0.939, P=.02), IIA (HR 0.683, 95% CI 0.524-0.892, P=.005), and IIIB (HR 0.616, 95% CI 0.392-0.969, P=.04) disease but not in stage IB (P=.56), IIB (P=.94), IIIA (P=.82), and IIIC (P=.70) disease (Table 5).

When further stratified by the eighth AJCC PPS, Chinese women had a better 5-year CSS in stage IA (98.7% vs 99.2%, P<.001), stage IIA (92.0% vs 94.8%, P=.049), and IIIA (85.5% vs 90.8%, P=.02) disease than White American women, and the 5-year CSS was not significant in those with stage IB (95.6% vs 96.3%, P=.44), IIB (89.4% vs 91.2%, P=.42), IIIB (78.2% vs 78.2%, P=.89), and IIIC (64.5% vs 73.5%, P=.42) disease between White American and Chinese women (Figure 6). Cox multivariate analysis revealed that Chinese women had a better CSS in stage IA (HR 0.673, 95% CI 0.476-0.853, P=.002) and IIIA (HR 0.689, 95% CI 0.478-0.994, P=.046) disease than White American women, but race was not a prognostic factor in those with stage IB (P=.85), IIA (P=.27), IIB (P=.90), IIIB (P=.63), and IIIC (P=.25) disease (Table 5).

This study aimed to evaluate stage-specific survival in BC between Chinese and White American women in accordance with the eighth AJCC PPS. Our results show that the eighth AJCC PPS had a similar discriminating ability between White American and Chinese participants with BC compared with the seventh AJCC AS. Our study provides additional data on the use of new PPS in different races based on current real-world practices.

Advances in molecular biomarkers (ER, PR, and HER2) and their close relationship with treatment responses and prognosis rendered the traditional AS unable to meet the trend of individualized treatment [6,7,17]. The eighth AJCC PPS, which integrates the aforementioned biomarkers and grade, facilitates more precise prognosis prediction than the seventh AJCC AS [8-10]. However, the small sample size and treatment heterogeneity in their studies limited the application of the eighth AJCC PPS in BC. In our real-world study with a large sample size (n=376,818), the eighth AJCC PPS revealed better prognostic accuracy than the seventh AJCC AS (P<.001) and performed well with discriminating ability consistent with disease stages. Therefore, the new AJCC staging system could better predict prognosis and guide the treatment of BC.

Our study shows that 149,452 (39.7%) individuals with BC migrated from the AJCC AS to AJCC PPS, which was similar to the rates observed in previous studies (20.7%-52.8%) [10,18-22]. The downstaging rate was significantly higher than upstaging rate (33.7% vs 6.0%) in this study, and the results are consistent with those of previous studies (downstaging: 15.2%-42.1%; upstaging: 5.5%-41.0%) [10,18-22]. Change in stage leads to diverse therapeutic decisions. The new AJCC staging enabled 126,936 (33.7%) participants to be downstaged, and these patients might be exempt from the therapies, such as chemotherapy and radiotherapy, which could ensure efficacy and reduce the treatment burden of patients [23]. In our previous studies, we found that the new AJCC staging can accurately guide individualized treatment of patients with BC in clinical decision-making. Patients who were downstaged from the eighth AJCC PPS can safely avoid adjuvant chemotherapy or radiotherapy [24-26]. Therefore, the 8th AJCC staging better reflects the trend of personalized treatment. In addition, among the patients with stage changes in this study, White American participants had a higher upstaging rate than Chinese participants (6.0% vs 5.3%), while the latter had a higher downstaging rate than the former (35.2% vs 33.6%). The Will Rogers phenomenon might explain the differences in “stage migration” in individuals with cancer who are of different races. Differences in culture, education, and diet lead to differences in migration rates between different ethnic groups [27-29]. In addition, socioeconomic status might be another critical factor affecting stage distribution and survival outcomes in different races. A study by Kantor et al [30] included 259,852 individuals with BC who are of different races and reported that non-Hispanic Black individuals and those of lower socioeconomic status had a lower disease-specific survival, even in all substages of the PPS [30].

The initial model for establishing the new AJCC PPS in BC was based on 305,519 patients from National Cancer Database between 2010 and 2012 [11]. However, the majority of the participants were White American and Hispanic American [11]. Therefore, the applicability of the eighth edition of the AJCC staging in Asian individuals, especially in Chinese individuals, remains unclear. Several retrospective studies explored the value of new AJCC staging in Chinese individuals with BC. However, only partial subgroups of BC, such as T1-2N1 and TNBC were included [13-15,31]. The cohort study conducted by He et al [15] recruited patients with TNBC from Sun Yat-sen University Cancer Center (n=611) and the SEER database (n=31,941) to examine the prognostic value of the eighth AJCC PPS in comparison with the seventh AJCC AS. However, no significant discriminatory ability was observed between the 2 staging systems in Chinese individuals with BC in this study and patients from the SEER database [15]. The opposite result was obtained in another study conducted by Yang et al [31], which included 1556 Chinese individuals with BC and compared the prognostic value of the 2 staging systems. They found that the new AJCC PPS had better accuracy of prognosis prediction than AS in Chinese individuals with BC [28]. However, their sample size was relatively small, especially in the eighth AJCC PPS of stage IIB (n=83) and IIIC (n=22). Therefore, their result might not accurately reflect the value of new staging [31]. In addition, most of the studies assessing the effect of the eighth AJCC PPS in the Chinese population lacked a comparison with the standard population, and their applicability may not be adequate [13-15,31]. In our study, we used a much larger sample to evaluate the new AJCC staging, and we observed a better discriminating value than that of the AJCC AS regardless of race. Therefore, our study better verified the applicability of the new staging in Chinese individuals with BC.

In a previous SEER study, Lim et al [4] reported that Chinese women with BC in the United States have better CSS than White American women, and the largest survival differences between Chinese and White American women were observed for stage I and node-negative cancers [4]. In this study, using the new AJCC staging, we found that Chinese women had superior CSS among those with stage IA and stage IIIA disease compared to White American women. The main reasons for this difference are not clear. The differences in treatment compliance and inherent genetic predisposition may lead to differences in survival between the 2 ethnic groups [32-35].

There are several limitations to be acknowledged in this study. First, we extracted the patient data from the SEER database, and selection biases inherently existing in retrospective studies should not be disregarded. Second, although the sample size of the group of Chinese individuals was much larger than that in previous studies, the number of individuals in some substages, such as stage IIIC (n=87), was still small. Therefore, the value of new AJCC staging in Chinese individuals with BC should be further explored. Third, details of treatment were not collected in the SEER database, including radiotherapy (technique, target volume, and radiation dose), chemotherapy regimens, endocrine therapy (regimen and duration), and targeted therapy, which may potentially affect the final analysis. Even with these limitations, our study reflects real-world practices and extends the applicability of the new staging.

In conclusion, our study suggests that the eighth AJCC PPS has a similar discriminating ability in White American and Chinese individuals with BC than the AJCC AS. Therefore, the new staging is also applicable to Chinese individuals with BC. Further studies are needed to explore the value of the PPS in Chinese individuals with BC.