Demographic Data and Clinicopathological Characteristics
This study extracted 13606 patients diagnosed with ASCC from SEER database and excluded 1715 patients with incomplete TNM stage info. A total of 11891 patients diagnosed with ASCC between 2004 and 2015 were included in our study finally. Clinicopathologic data for different 8th AJCC clinical stages are presented in Table 2. There was a higher proportion of men among SEER patients diagnosed with ASCCs. Meanwhile, the main proportion in the SEER cohort was represented by the patients of a White race. Furthermore, for patients with most of the clinical stages among the ASCC patient cohort, no surgical resection was performed (59.1–83.5%), except for those with stage I (32.9%). Except for I stage ASCC patients, the number of ASCC patients who underwent radiation therapy and chemotherapy treatment was not significantly different between the 8th AJCC clinical stages of ASCC patients.
Disease-specific survival and Overall Survival for different T, N and M Classifications
The 5-year DSS rates was 86.1% for T1, 83.6% for T2, 67.6% for T3, and 62.3% for T4 tumours and the 5-year OS rates was 80.0% for T1, 65.2% for T2, 47.1% for T3 and 42.9% for T4 for according to the AJCC 8th edition staging system. When stratified by the AJCC 7th nodal classification system, DSS did not differ significantly (λ = 1.429, p = 0.232) between N1 and N2 nodal status in patients with ASCC but OS differed significantly (λ = 4.775, p = 0.029). When evaluated using the AJCC 8th nodal classification system (N0 vs. N1), there were significant differences in DSS (λ = 466.352, p < 0.001) and OS (λ = 273.702, p < 0.001) between the stages (Figure 1).
Furthermore, the comparison between the 5-year DSS rates in patients with TxN+M0 stages defined using 7th and 8th editions showed that there was no significant difference in DSS between N1, N2 and N3 among ASCC patients with T1 and T3 according to the AJCC 7th, similar with OS (Supplementary 1). In the AJCC 8th M classification, DSS (λ = 1219.892, p < 0.001) and OS (λ = 1091.599, p < 0.001) also differed significantly.
AJCC stage groupings
According to the AJCC 8th edition staging system, the 5-year DSS rates were 94.2% for stage I, 86.4% for stage IIA, 74.9% for stage IIB, 79.4% for stage IIIA, 72.2% for stage IIIB, 64.8% for stage IIIC, and 35.4% for stage IV ASCC patients and the 5-year OS rates were 81.7% for stage I, 68.1% for stage IIA, 53.4% for stage IIB, 65.1% for stage IIIA, 50.7% for stage IIIB, 48.5% for stage IIIC, and 21.2% for stage IV ASCC patients (Figure 2a). The new AJCC subclassification of stage II ASCCs into A and B showed a favourable discrimination in DSS (λ = 80.833, p < 0.001) and OS (λ = 82.121, p < 0.001) between IIA and IIB (Figure 2b). Furthermore, 8th AJCC subclassification of stage III ASCC patients also showed a better differentiation strategy of OS (λ = 41.451, p < 0.001) than that of the 7th AJCC (λ = 1.071, p = 0.301) for IIIA vs. IIIB (Figure 2c).
Modification of the 8th AJCC clinical staging system
A few deficiencies were found when we verified the efficacy of discrimination in the 8th AJCC clinical staging system, especially between IIB and IIIA stages. KM analysis showed that the DSS (λ = 10.153, p = 0.002) and OS (λ = 47.652, p < 0.001) of IIIA ASCC patient showed much better than IIB patients with significant differences. Further, there was no significant difference in OS between IIIB and IIIC (λ = 0.037, p = 0.848) (Figure 3a). To identify the impact of deficiencies on the discriminatory value of the system, we calculated univariable Cox proportional hazards for DSS and OS in each of the 8th clinical stages and actual TNM stages among the ASCC patients (Supplementary 2) and made heatmaps (Figure 3b).
Furthermore, we analysed the hazards for actual TNM stage ASCC patients and modified the AJCC 8th clinical staging system (Table 3). The novel AJCC IIB stage was modified as T1N1M0, the IIIA stage was modified as T3N0M0 and T2N1M0, the IIIB stage was modified as T4N0M0 and T3N1M0 and the IIIC stage was modified as T4N1M0, while the corresponding 8th IIB stage was defined as T3N0M0, IIIA stage as T1-2N1M0, IIIB stage as T4N0M0 and IIIC stage as T3-4N1M0.
Validation of the modified 8th AJCC staging system
The prognostic utility of the 8th AJCC clinical staging system was enhanced after our modifications. ASCC patients experienced worse prognosis as the AJCC stage increased, except that the differences in prognostic outcomes were not significant between IIB and IIIA stage ASCC patients. After the modification, IIIA patients showed logically worse DSS (λ = 4.952, p = 0.022) and OS (λ = 14.210, p < 0.001) than IIB patients with significant difference. Moreover, the discrimination of OS between IIIB and IIIC patients has been promoted although with no significant difference (λ = 1.994, p = 0.120) (Figure 3c).
The C-indices of the different staging systems had been promoted after modifications. For DSS of ASCC patients in the SEER cohort, the C-index of the 8th staging system increased from 0.697 (95%CI: 0.686–0.709) to 0.711 (95%CI: 0.699–0.722) compared with that of the 7th Staging System for all stages of ASCC. For II and III stages ASCC patients, the C-index was also increased from 0.585 (95%CI: 0.571–0.601) to 0.612 (95%CI: 0.597–0.626). The C-index increased significantly after our modification, from 0.612 (95%CI: 0.597–0.626) to 0.636 (95%CI: 0.611–0.645) for II and III stage ASCC patients, and from 0.711 (95%CI: 0.699–0.722) to 0.721 (95%CI: 0.703–0.729) for the entire cohort (the OSs promotion showed in Supplementary 3).