Stage migration and prognostic refinement of oral squamous cell carcinoma under the AJCC 8th edition staging system: a retrospective cohort study with additional evaluation using the SEER database Authors: Jianing Cui, Qianrui Liu, Hongshi Cai, Jianfeng Liang, Hongyu Li, Yue Zhu, Meng Xu, Jinsong Hou Published: 2026 ABSTRACT ObjectiveThe eighth edition of the American Joint Committee on Cancer (AJCC) staging system incorporated pathologic depth of invasion (pDOI) and pathologic extranodal extension (pENE) for prognostic stratification in oral squamous cell carcinoma (OSCC). However, the extent of stage migration and the consistency of survival discrimination across populations remained controversial. This study aimed to evaluate stage migration associated with the AJCC 8th edition, assess its prognostic performance relative to the AJCC 7th edition, and explore potential refinements to the pT classification.MethodsPatient data from the Hospital of Stomatology, Sun Yat-sen University (2014–2021), and the SEER database (2018–2019) were used. Overall survival (OS) and disease-specific survival (DSS) were estimated using the Kaplan–Meier method to compare the prognostic outcomes of the AJCC 7th and 8th editions. An exploratory refinement of the AJCC 8th pT classification was then developed using the SYSU-OSCC cohort. Multivariable Cox regression was used to assess the prognostic value of different cutoff thresholds, and a refined pT classification was proposed. Model performance was evaluated according to the established Groome criteria.ResultsIn the SYSU-OSCC cohort, upstaging was observed in 31.3% of patients for pT classification, 10.9% for pN classification, and 31.4% for overall pathologic stage. Compared with the AJCC 7th edition, the AJCC 8th showed clearer prognostic separation between pT3 and pT4a in the SYSU-OSCC cohort (5-year OS rates: 69.2% vs. 51.7%, p = 0.012; 5-year DSS rates: 76.6% vs. 62.0%, p = 0.014) and improved survival discrimination, particularly at intermediate and advanced-stage disease. However, prognostic overlap persisted between pT1 and pT2 categories (5-year OS rates: 82.4% vs. 76.9%, p = 0.14; 5-year DSS rates: 87.0% vs. 83.9%, p = 0.45). In the SEER-OSCC cohort, however, risk separation between pT3 and pT4a remained limited (3-year OS rates: 58.1% vs. 54.3%, p = 0.16; 3-year DSS rates: 64.5% vs. 63.4%, p = 0.41). Based on the SYSU-OSCC cohort, tumors with pDOI > 5 mm and ≤ 10 mm and a tumor size ≤ 2 cm were proposed for reclassification into the pT1 category. In the SYSU-OSCC training and internal validation cohorts, the exploratory refined classification showed improved prognostic performance compared with the original AJCC 8th edition.ConclusionThe AJCC 8th edition improved prognostic stratification in OSCC, particularly for intermediate and advanced-stage disease. However, residual prognostic overlap remained between specific pT categories. The exploratory refinement of selected pT thresholds may modestly improve prognostic accuracy. However, rigorous validation in larger independent external cohorts is required before its clinical applicability can be determined.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1186/s12903-026-08922-2. Introduction Oral squamous cell carcinoma (OSCC) is the most common malignancy of the oral cavity and remains a major global health concern [1]. Despite advances in surgical techniques and adjuvant therapies, survival outcomes for patients with OSCC have improved only modestly over recent decades, highlighting the need for accurate risk stratification to guide treatment planning and prognostic assessment [2]. The 8th edition of the American Joint Committee on Cancer (AJCC) Cancer Staging Manual introduced pathologic depth of invasion (pDOI) and pathologic extranodal extension (pENE) as key parameters for the pathologic tumor (pT) and pathologic nodal (pN) classifications of OSCC, aiming to improve risk stratification and prognostic accuracy. The initial AJCC 8th edition criteria for oral cavity cancer were released in 2017 and were subsequently modified during the implementation period. In an interim revision implemented on January 1, 2018, pDOI > 20 mm was introduced as a criterion for pT4a classification [3, 4]. This criterion was later revised in the final update released on June 25, 2018, in which tumors larger than 4 cm with pDOI > 10 mm were classified as pT4a. In the present study, restaging was performed according to the final AJCC 8th edition criteria released on June 25, 2018. The National Comprehensive Cancer Network (NCCN) Clinical Practice Guidelines in Oncology are widely used in clinical practice and provide detailed recommendations for diagnosis and treatment according to tumor type and stage. The 2018 NCCN guidelines for oral cavity cancers reflected the changes in the initial AJCC 8th edition released in 2017 [5]. In the 2019 updates and subsequent versions of the NCCN guidelines, further adjustments to stratification recommendations for selected tumor characteristics were made based on the final AJCC 8th edition staging criteria released on June 25, 2018 [6, 7]. These changes suggest, to some extent, that further refinement of the staging system remains clinically relevant. Although the stage redistribution after the implementation of AJCC 8th has been reported in several studies [8–12], the magnitude of redistribution and its impact on survival discrimination have varied considerably across studies and populations. Some reports have suggested improved prognostic stratification with clearer stage separation [8, 11–13], whereas others have reported no significant improvement [14, 15], raising questions about the generalizability of the revised criteria. Therefore, a retrospective analysis was conducted using two data sources. Using data from a high-incidence OSCC region in China, the extent of stage migration associated with pDOI and pENE was evaluated. The prognostic performance of the AJCC 8th edition was compared with that of the AJCC 7th edition in the SYSU-OSCC cohort, and an external cohort from the Surveillance, Epidemiology, and End Results (SEER) database was used for an additional evaluation of the AJCC 8th edition. The strengths and limitations of the current staging system were evaluated, and potential directions for refinement were explored. Materials and methods Data sources and quality assessment Two retrospective data sources were included in this study. The SYSU-OSCC cohort included 937 patients who were diagnosed with OSCC at the Hospital of Stomatology, Sun Yat-sen University, between January 2014 and March 2021, with follow-up data available through March 2026. For the exploratory model development and internal validation of the refined pT classification, the SYSU-OSCC cohort was randomly divided at a 1:1 ratio into a training cohort and an internal validation cohort. A fixed random seed was used to ensure reproducibility of the random allocation. This study was conducted in accordance with the Declaration of Helsinki and was approved by the Medical Research Ethics Committee of the Hospital of Stomatology, Sun Yat-sen University (approval number: KQEC-2024–122-01). Because of the retrospective design and use of anonymized data, the requirement for informed consent was waived. The SEER-OSCC cohort was derived from the SEER database and included patients diagnosed between January 2018 and December 2019, with follow-up data available through December 2022. Patient selection and variable definitions for the SEER-OSCC cohort were based on SEER coding rules and data documentation. This retrospective study was conducted and reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) Statement for cohort studies. A completed STROBE checklist has been provided [see Supplementary Table 1]. Patient selection Patients with histopathologically confirmed OSCC before surgery were eligible for inclusion. Histologic codes included 8052/3, 8070–8076/3, 8078/3, 8083/3, and 8084/3. Primary tumor subsites included the oral tongue (C02.0–C02.3), upper and lower gingiva (C03.0–C03.9), floor of mouth (C04.0–C04.9), hard palate (C05.0), and other oral cavity sites, including the buccal mucosa and retromolar trigone (C06.0–C06.9). Ambiguous or overlapping sites that could potentially originate from the oropharynx (e.g., C02.8–C02.9, C05.8–C05.9) were excluded. Distant metastasis status in the SYSU-OSCC cohort was assessed using pretreatment imaging, including chest CT and abdominal imaging. All included patients had no evidence of distant metastasis and were classified as M0. Patients were excluded if they had received induction chemotherapy, preoperative radiotherapy or chemoradiotherapy, or palliative treatment. Cases were also excluded if key demographic data, tumor stage, grade, histological type, surgical treatment information, or cause of death were missing or unspecified. Patients who were lost to follow-up before the study end date without documented events were also excluded. Treatment and follow-up All patients in the SYSU-OSCC cohort underwent curative-intent resection of the primary tumor, with adequate margins pursued according to standard oncologic principles. Neck management, including elective or therapeutic neck dissection, was performed according to clinical stage and institutional protocols. Postoperative radiotherapy or chemoradiotherapy was recommended according to established risk factors, including positive margins, nodal metastasis, advanced pT category, and pENE. Follow-up was conducted via structured telephone interviews and/or scheduled outpatient visits at our institution. Follow-up intervals were scheduled as follows: every 3 months during the first 2 postoperative years, every 4–6 months during years 3–5, and every 6–12 months thereafter. At each outpatient visit, a comprehensive head and neck examination was performed by specialized oral oncology clinicians. Any clinical suspicion of local recurrence, regional nodal recurrence, or distant metastasis prompted diagnostic imaging, including contrast-enhanced CT, MRI, and/or PET/CT, as clinically indicated. Active follow-up for the SYSU-OSCC cohort ended on March 31, 2026. For the SEER-OSCC cohort, follow-up data were available through December 2022. TNM classification All patients in the SYSU-OSCC cohort were initially staged according to the 7th edition of the AJCC staging system and later reviewed and restaged according to the final 8th edition criteria published on June 25, 2018. The pDOI and pENE were assessed directly from histological slides. For the pT classification, primary tumor maximum diameter and invasion of adjacent structures, including the mandibular or maxillary medullary bone, maxillary sinus, and facial skin, were evaluated. Extrinsic tongue muscle invasion was not used as a criterion for pT4a classification in accordance with the final AJCC 8th edition criteria. Patients in the SEER-OSCC cohort were staged according to the final AJCC 8th edition staging system published on June 25, 2018. Patients without neck dissection were classified as pNx and were excluded from pN-specific survival analyses and analyses requiring pathologic stage grouping. Accordingly, pN classification and pathologic stage migration were evaluated only among patients with available pathologic nodal information. Statistical analysis Statistical analyses were performed using SPSS version 26.0 (IBM Corp., Armonk, NY, USA) and R version 4.2.0 (R Project for Statistical Computing). Overall survival (OS) was defined as the time from diagnosis to death from any cause or last follow-up. Disease-specific survival (DSS) was defined as the time from diagnosis to death from oral cancer, with deaths from other causes censored. Median follow-up was estimated using the reverse Kaplan–Meier method. Continuous variables were compared using the Mann–Whitney U test, and categorical variables were compared using Pearson’s chi-squared test or Fisher’s exact test, as appropriate. Baseline characteristics were compared between the training and internal validation cohorts, and no statistically significant differences were observed [see Supplementary Table 2]. OS and DSS were evaluated separately in the SYSU-OSCC and SEER-OSCC cohorts. In the SYSU-OSCC cohort, 5-year OS and DSS were estimated because long-term follow-up data were available. For the SEER-OSCC cohort, 3-year OS and DSS were reported because follow-up data were available only through December 2022. Survival estimates were used to assess prognostic stratification within each cohort; no direct comparison of survival rates between the SYSU-OSCC and SEER-OSCC cohorts was performed. Survival curves were generated using the Kaplan–Meier method, and between-group differences were assessed using the log-rank test. Univariable Cox regression analyses were first performed to assess associations between clinicopathological variables and survival outcomes. Variables were selected for multivariable Cox proportional hazards regression based on clinical relevance, prior evidence, potential confounding effects, and univariable analysis results, rather than solely on statistical significance. Multicollinearity among candidate variables was assessed using variance inflation factors (VIFs), and variables with substantial collinearity were not entered simultaneously into the same model. The proportional hazards assumption was evaluated using Schoenfeld residuals. Additional survival analyses and visualizations were performed using the survival, survminer, and ggplot2 packages in R. Harrell’s concordance index (C-index) was used to evaluate the discriminative ability of different staging systems for survival outcomes. The 95% confidence intervals (CIs) for the C-index were calculated using the corresponding standard errors. Model performance was compared across staging systems according to the established Groome criteria: hazard consistency, hazard discrimination, sample size balance, and outcome prediction [16]. Comparisons were performed using the AutoRPA framework through the online analytical tool developed by Ren et al. [17] (available at http://rpa.renlab.org/). Staging systems were ranked according to an overall summary score, with lower scores indicating better performance. The robustness of the evaluation was assessed using 200 bootstrap resampling iterations. A two-sided p-value < 0.05 was considered statistically significant. Results Baseline characteristics and prognosis based on AJCC 8th A total of 937 patients were included in the SYSU-OSCC cohort (median age, 54 years), with a median follow-up of 78 months (range, 1.0–144.0 months). Neck dissection was performed in 69.3% of patients (649/937), and pathologic nodal metastasis was identified in 36.5%. Postoperative radiotherapy and chemotherapy were administered to 17.1% and 9.6% of patients, respectively. The SEER-OSCC cohort comprised 2589 patients from the SEER database, with a median follow-up of 41 months (range, 2.0–59.0 months). Neck dissection was performed in 97.5% of patients (2525/2589), and pathologic nodal metastasis was identified in 43.0%. Postoperative radiotherapy and chemotherapy were administered to 55.9% and 23.3% of patients, respectively. Baseline clinicopathological characteristics are summarized in Table 1. During follow-up, 293 deaths from any cause and 183 oral cancer-specific deaths were recorded in the SYSU-OSCC cohort. In the SEER-OSCC cohort, 935 deaths from any cause and 666 oral cancer-specific deaths were recorded.Table 1Baseline characteristicsCharacteristicSYSU-OSCC cohortSEER-OSCC cohort(n = 2589)Training cohort(n = 490)Validation cohort(n = 447)Age, median (IQR), y55 (46–64)54 (45–63.5)64 (56–73)Sex, No. (%) Male309 (63.0)261 (58.4)1547 (59.8) Female181 (36.9)186 (41.6)1042 (40.2)Subsite, No. (%) Anterior tongue285 (58.2)276 (61.7)1175 (45.4) Gum/RMT71 (14.5)53 (11.9)614 (23.7) Buccal mucosa98 (20.0)74 (16.6)262 (10.1) Hard palate11 (2.2)13 (2.9)41 (1.6) Floor of mouth25 (5.1)31 (6.9)374 (14.4) NOS--123 (4.8)AJCC 7th pT, No. (%) pT1207 (42.2)192 (42.9)- pT2216 (44.1)199 (44.5)- pT334 (6.9)23 (5.1)- pT4a26 (5.3)27 (6.0)- pT4b7 (1.4)6 (1.3)-AJCC 7th pN, No. (%) pN0208 (42.4)204 (45.6)- pN157 (11.6)45 (10.1)- pN274 (15.1)61 (13.6)-AJCC 8th pT, No. (%) pT1140 (28.6)133 (29.7)478 (18.5) pT2214 (43.7)193 (43.7)812 (31.4) pT383 (16.9)76 (15.3)393 (15.2) pT4a46 (9.4)39 (10.0)850 (32.8) pT4b7 (1.4)6 (1.3)56 (2.2)AJCC 8th pN, No. (%) pN0208 (42.4)204 (45.6)1438 (55.5) pN145 (9.2)43 (9.6)278 (10.7) pN255 (11.2)37 (8.3)420 (16.2) pN331 (6.3)26 (5.8)405 (15.6)pDOI, No. (%) ≤ 5 mm205 (41.8)187 (41.8)743 (28.7) > 5 and ≤ 10 mm151 (30.8)146 (32.7)551 (21.3) > 10 mm134 (27.3)114 (25.5)571 (22.1)pENE, No. (%) pENE-296 (60.4)282 (63.1)2047 (80.8) pENE + 43 (8.8)28 (6.3)498 (19.2)Neck dissection, No. (%)339 (69.2)310 (69.4)2525 (97.5)Postoperative Radiotherapy, No. (%)91 (18.6)70 (15.7)1448 (55.9)Postoperative Chemotherapy, No. (%)47 (9.6)43 (9.6)603 (23.3)Percentages (%) indicate the proportion of patients within the corresponding group or category. pN classification and pENE status were summarized among patients with available pathologic nodal informationRMT Retromolar trigone, NOS Not otherwise specified, pN Pathologic nodal, pT Pathologic tumorIQR, interquartile range Table 1Baseline characteristicsCharacteristicSYSU-OSCC cohortSEER-OSCC cohort(n = 2589)Training cohort(n = 490)Validation cohort(n = 447)Age, median (IQR), y55 (46–64)54 (45–63.5)64 (56–73)Sex, No. (%) Male309 (63.0)261 (58.4)1547 (59.8) Female181 (36.9)186 (41.6)1042 (40.2)Subsite, No. (%) Anterior tongue285 (58.2)276 (61.7)1175 (45.4) Gum/RMT71 (14.5)53 (11.9)614 (23.7) Buccal mucosa98 (20.0)74 (16.6)262 (10.1) Hard palate11 (2.2)13 (2.9)41 (1.6) Floor of mouth25 (5.1)31 (6.9)374 (14.4) NOS--123 (4.8)AJCC 7th pT, No. (%) pT1207 (42.2)192 (42.9)- pT2216 (44.1)199 (44.5)- pT334 (6.9)23 (5.1)- pT4a26 (5.3)27 (6.0)- pT4b7 (1.4)6 (1.3)-AJCC 7th pN, No. (%) pN0208 (42.4)204 (45.6)- pN157 (11.6)45 (10.1)- pN274 (15.1)61 (13.6)-AJCC 8th pT, No. (%) pT1140 (28.6)133 (29.7)478 (18.5) pT2214 (43.7)193 (43.7)812 (31.4) pT383 (16.9)76 (15.3)393 (15.2) pT4a46 (9.4)39 (10.0)850 (32.8) pT4b7 (1.4)6 (1.3)56 (2.2)AJCC 8th pN, No. (%) pN0208 (42.4)204 (45.6)1438 (55.5) pN145 (9.2)43 (9.6)278 (10.7) pN255 (11.2)37 (8.3)420 (16.2) pN331 (6.3)26 (5.8)405 (15.6)pDOI, No. (%) ≤ 5 mm205 (41.8)187 (41.8)743 (28.7) > 5 and ≤ 10 mm151 (30.8)146 (32.7)551 (21.3) > 10 mm134 (27.3)114 (25.5)571 (22.1)pENE, No. (%) pENE-296 (60.4)282 (63.1)2047 (80.8) pENE + 43 (8.8)28 (6.3)498 (19.2)Neck dissection, No. (%)339 (69.2)310 (69.4)2525 (97.5)Postoperative Radiotherapy, No. (%)91 (18.6)70 (15.7)1448 (55.9)Postoperative Chemotherapy, No. (%)47 (9.6)43 (9.6)603 (23.3)Percentages (%) indicate the proportion of patients within the corresponding group or category. pN classification and pENE status were summarized among patients with available pathologic nodal informationRMT Retromolar trigone, NOS Not otherwise specified, pN Pathologic nodal, pT Pathologic tumorIQR, interquartile range Baseline characteristics Percentages (%) indicate the proportion of patients within the corresponding group or category. pN classification and pENE status were summarized among patients with available pathologic nodal information RMT Retromolar trigone, NOS Not otherwise specified, pN Pathologic nodal, pT Pathologic tumor IQR, interquartile range Patients in the SYSU-OSCC cohort were staged according to both AJCC 7th and AJCC 8th editions, whereas those in the SEER-OSCC cohort were staged only according to AJCC 8th edition. Stage migration trends from AJCC 7th to AJCC 8th After restaging from the AJCC 7th edition to the final AJCC 8th edition, the pT category was upstaged in 31.3% of patients, downstaged in 0.1%, and remained unchanged in 68.6% (Table 2). The pN category remained unchanged in 89.1% of patients and was upstaged in 10.9% (Table 3). Overall pathologic stage was upstaged in 31.4% of patients and remained unchanged in 68.6% (Table 4).Table 2Upstaging and downstaging trends for the pT categoryAJCC 8th, No. (%)AJCC 7thpT1pT2pT3pT4apT4bTotal, NopT1273 (68.4)126 (31.6)399pT2281 (67.5)134 (32.5)416pT324 (42.1)33 (57.9)57pT4a1 (1.9)52 (98.1)53pT4b13 (100)13Total, No2734071598513937Percentages (%) indicate the proportion of patients within the corresponding categorypT Pathologic tumorTable 3Upstaging trends for the pN categoryAJCC 8th, No. (%)AJCC 7thpN0pN1pN2apN2bpN2cpN3bTotal, NopN0412 (100)412pN188 (86.3)14 (13.7)102pN2a2 (50.0)2 (50.0)4pN2b68 (59.1)47 (40.9)115pN2c8 (50.0)8 (50.0)16Total, No412881668857649Percentages (%) indicate the proportion of patients within the corresponding categorypN Pathologic nodalTable 4Upstaging trends for the overall pathologic stageAJCC 8th, No. (%)AJCC 7thpStageIpStageIIpStageIIIpStageIVApStageIVBTotal, NopStageI88 (59.5)60 (40.5)148pStageII153 (69.9)66 (30.1)219pStageIII96 (80.7)23 (19.3)119pStageIVA95 (63.3)55 (36.7)150pStageIVB13 (100)13Total, No8821316211868649Percentages (%) indicate the proportion of patients within the corresponding categorypStage, pathologic stage Table 2Upstaging and downstaging trends for the pT categoryAJCC 8th, No. (%)AJCC 7thpT1pT2pT3pT4apT4bTotal, NopT1273 (68.4)126 (31.6)399pT2281 (67.5)134 (32.5)416pT324 (42.1)33 (57.9)57pT4a1 (1.9)52 (98.1)53pT4b13 (100)13Total, No2734071598513937Percentages (%) indicate the proportion of patients within the corresponding categorypT Pathologic tumor Upstaging and downstaging trends for the pT category Percentages (%) indicate the proportion of patients within the corresponding category pT Pathologic tumor Table 3Upstaging trends for the pN categoryAJCC 8th, No. (%)AJCC 7thpN0pN1pN2apN2bpN2cpN3bTotal, NopN0412 (100)412pN188 (86.3)14 (13.7)102pN2a2 (50.0)2 (50.0)4pN2b68 (59.1)47 (40.9)115pN2c8 (50.0)8 (50.0)16Total, No412881668857649Percentages (%) indicate the proportion of patients within the corresponding categorypN Pathologic nodal Upstaging trends for the pN category Percentages (%) indicate the proportion of patients within the corresponding category pN Pathologic nodal Table 4Upstaging trends for the overall pathologic stageAJCC 8th, No. (%)AJCC 7thpStageIpStageIIpStageIIIpStageIVApStageIVBTotal, NopStageI88 (59.5)60 (40.5)148pStageII153 (69.9)66 (30.1)219pStageIII96 (80.7)23 (19.3)119pStageIVA95 (63.3)55 (36.7)150pStageIVB13 (100)13Total, No8821316211868649Percentages (%) indicate the proportion of patients within the corresponding categorypStage, pathologic stage Upstaging trends for the overall pathologic stage Percentages (%) indicate the proportion of patients within the corresponding category pStage, pathologic stage Stage migration and survival changes We analyzed 5-year OS and DSS among patients whose pT category changed after incorporation of pDOI in the AJCC 8th edition (Tables 5 and 6). Compared with the corresponding AJCC 7th edition categories, the final AJCC 8th edition pT1–pT3 categories showed higher stage-specific survival estimates, except for a slight decrease in 5-year DSS in the pT1 category. These changes were interpreted as survival redistribution after restaging rather than as changes in survival within the same patients. Conversely, the final AJCC 8th edition pT4a category showed lower 5-year OS and DSS estimates, likely because high-risk tumors originally classified as pT3 were reclassified as pT4a, resulting in risk concentration within the pT4a category.Table 5Stage migration and 5-year OS by pT classificationAJCC 7thStage migration (AJCC 8th)AJCC 8thΔOS (% change)pT1: 82.2%pT1: 82.4%pT1: 82.4% + 0.2% (0.2%)pT2: 81.7%pT2: 72.8%pT2: 74.8%pT2: 76.9% + 4.1% (5.6%)pT3: 68.7%pT3: 52.5%pT3: 74.8%pT3: 69.2% + 16.7% (31.8%)pT4a: 36.4%pT4a: 60.4%pT3: 0.0%pT4a: 51.7%−8.7% (14.4%)pT4a: 61.5%pT4b: 38.5%NonepT4b: 38.5%-Percentages (%) for OS and DSS indicate calculated 5-year survival estimatesTable 6Stage migration and 5-year DSS by pT classificationAJCC 7thStage migration (AJCC 8th)AJCC 8thΔDSS (% change)pT1: 87.7%pT1: 87.0%pT1: 87.0%−0.7% (0.8%)pT2: 89.3%pT2: 80.0%pT2: 81.5%pT2: 83.9% + 3.9% (4.9%)pT3: 76.8%pT3: 63.1%pT3: 78.7%pT3: 76.6% + 13.5% (21.4%)pT4a: 51.1%pT4a: 67.7%pT3: 0.0%pT4a: 62.0%−5.7% (8.4%)pT4a: 69.1%pT4b: 44.9%NonepT4b: 44.9%-Percentages (%) for OS and DSS indicate calculated 5-year survival estimates Table 5Stage migration and 5-year OS by pT classificationAJCC 7thStage migration (AJCC 8th)AJCC 8thΔOS (% change)pT1: 82.2%pT1: 82.4%pT1: 82.4% + 0.2% (0.2%)pT2: 81.7%pT2: 72.8%pT2: 74.8%pT2: 76.9% + 4.1% (5.6%)pT3: 68.7%pT3: 52.5%pT3: 74.8%pT3: 69.2% + 16.7% (31.8%)pT4a: 36.4%pT4a: 60.4%pT3: 0.0%pT4a: 51.7%−8.7% (14.4%)pT4a: 61.5%pT4b: 38.5%NonepT4b: 38.5%-Percentages (%) for OS and DSS indicate calculated 5-year survival estimates Stage migration and 5-year OS by pT classification Percentages (%) for OS and DSS indicate calculated 5-year survival estimates Table 6Stage migration and 5-year DSS by pT classificationAJCC 7thStage migration (AJCC 8th)AJCC 8thΔDSS (% change)pT1: 87.7%pT1: 87.0%pT1: 87.0%−0.7% (0.8%)pT2: 89.3%pT2: 80.0%pT2: 81.5%pT2: 83.9% + 3.9% (4.9%)pT3: 76.8%pT3: 63.1%pT3: 78.7%pT3: 76.6% + 13.5% (21.4%)pT4a: 51.1%pT4a: 67.7%pT3: 0.0%pT4a: 62.0%−5.7% (8.4%)pT4a: 69.1%pT4b: 44.9%NonepT4b: 44.9%-Percentages (%) for OS and DSS indicate calculated 5-year survival estimates Stage migration and 5-year DSS by pT classification Percentages (%) for OS and DSS indicate calculated 5-year survival estimates Similarly, substantial changes were observed across pN categories (Tables 7 and 8). Compared with early-stage categories, changes in stage-specific survival estimates were more pronounced in advanced-stage categories. Because of the small number of pN2a patients in the 7th edition and the reclassification of some original pN1 patients after the incorporation of the pENE, the AJCC 8th pN2a subgroup was largely reconstituted. Meanwhile, high-risk patients were more clearly concentrated in the pN3b category under the AJCC 8th edition.Table 7Stage migration and 5-year OS by pN classificationAJCC 7thStage migration (AJCC 8th)AJCC 8thΔOS (% change)pN0: 83.5%pN0: 83.5%pN0: 83.5%-pN1: 63.7%pN1: 64.8%pN1: 64.8% + 1.1% (1.7%)pN2a: 57.1%pN2a: 100%pN2a: 100%pN2a: 62.5%−37.5% (37.5%)pN3b: 100%pN2b: 47.8%pN2b: 52.9%pN2b: 52.9% + 5.1% (10.7%)pN3b: 40.3%pN2c: 43.8%pN2c: 50%pN2c: 50% + 6.2% (14.2%)pN3b: 37.5%pN3NonepN3b: 42.0%-Percentages (%) for OS and DSS indicate calculated 5-year survival estimatesTable 8Stage migration and 5-year DSS by pN classificationAJCC 7thStage migration (AJCC 8th)AJCC 8thΔDSS (% change)pN0: 89.3%pN0: 89.3%pN0: 89.3%-pN1: 71.2%pN1: 72.5%pN1: 72.5% + 1.3% (1.8%)pN2a: 62.9%pN2a: 100%pN2a: 100%pN2a: 67.7%−32.3% (32.3%)pN3b: 100%pN2b: 56.5%pN2b: 62.2%pN2b: 62.2% + 6.3% (10.1%)pN3b: 48.4%pN2c: 46.7%pN2c: 50%pN2c: 50% + 3.3% (7.1%)pN3b: 42.9%pN3NonepN3b: 49.7%-Percentages (%) for OS and DSS indicate calculated 5-year survival estimates Table 7Stage migration and 5-year OS by pN classificationAJCC 7thStage migration (AJCC 8th)AJCC 8thΔOS (% change)pN0: 83.5%pN0: 83.5%pN0: 83.5%-pN1: 63.7%pN1: 64.8%pN1: 64.8% + 1.1% (1.7%)pN2a: 57.1%pN2a: 100%pN2a: 100%pN2a: 62.5%−37.5% (37.5%)pN3b: 100%pN2b: 47.8%pN2b: 52.9%pN2b: 52.9% + 5.1% (10.7%)pN3b: 40.3%pN2c: 43.8%pN2c: 50%pN2c: 50% + 6.2% (14.2%)pN3b: 37.5%pN3NonepN3b: 42.0%-Percentages (%) for OS and DSS indicate calculated 5-year survival estimates Stage migration and 5-year OS by pN classification Percentages (%) for OS and DSS indicate calculated 5-year survival estimates Table 8Stage migration and 5-year DSS by pN classificationAJCC 7thStage migration (AJCC 8th)AJCC 8thΔDSS (% change)pN0: 89.3%pN0: 89.3%pN0: 89.3%-pN1: 71.2%pN1: 72.5%pN1: 72.5% + 1.3% (1.8%)pN2a: 62.9%pN2a: 100%pN2a: 100%pN2a: 67.7%−32.3% (32.3%)pN3b: 100%pN2b: 56.5%pN2b: 62.2%pN2b: 62.2% + 6.3% (10.1%)pN3b: 48.4%pN2c: 46.7%pN2c: 50%pN2c: 50% + 3.3% (7.1%)pN3b: 42.9%pN3NonepN3b: 49.7%-Percentages (%) for OS and DSS indicate calculated 5-year survival estimates Stage migration and 5-year DSS by pN classification Percentages (%) for OS and DSS indicate calculated 5-year survival estimates At the overall pathologic stage level, stage-specific survival estimates remained stable or increased across stages after restaging (Tables 9 and 10). Early stages showed minimal change in survival estimates, indicating the relative stability of early disease classification.Table 9Stage migration and 5-year OS by overall pathologic stageAJCC 7thStage migration (AJCC 8th)AJCC 8thΔOS (% change)I: 89.1%I: 89.7%I: 89.7% + 0.6% (0.8%)II: 88.3%II: 82.3%II: 81.8%II: 83.6% + 1.3% (1.6%)III: 83.3%III: 63.8%III: 67.7%III: 74.1% + 10.3% (16.1%)IVA: 47.8%IVA: 54.7%IVA: 61.1%IVA: 58.5% + 3.8% (6.9%)IVB: 43.6%IVB: 38.5%IVB: 38.5%IVB: 42.6% + 4.1% (10.6%)Percentages (%) for OS and DSS indicate calculated 5-year survival estimatesTable 10Stage migration and 5-year DSS by overall pathologic stageAJCC 7thStage migration (AJCC 8th)AJCC 8thΔDSS (% change)I: 93.8%I: 91.8%I: 91.8%−2.0% (2.1%)II: 96.5%II: 88.5%II: 88.8%II: 91.0% + 2.5% (2.8%)III: 87.6%III: 71.7%III: 75.9%III: 80.8% + 9.1% (12.7%)IVA: 54.7%IVA: 62.4%IVA: 68.7%IVA: 65.9% + 3.5% (5.6%)IVB: 51.6%IVB: 44.9%IVB: 44.9%IVB: 50.3% + 5.4% (12.0%)Percentages (%) for OS and DSS indicate calculated 5-year survival estimates Table 9Stage migration and 5-year OS by overall pathologic stageAJCC 7thStage migration (AJCC 8th)AJCC 8thΔOS (% change)I: 89.1%I: 89.7%I: 89.7% + 0.6% (0.8%)II: 88.3%II: 82.3%II: 81.8%II: 83.6% + 1.3% (1.6%)III: 83.3%III: 63.8%III: 67.7%III: 74.1% + 10.3% (16.1%)IVA: 47.8%IVA: 54.7%IVA: 61.1%IVA: 58.5% + 3.8% (6.9%)IVB: 43.6%IVB: 38.5%IVB: 38.5%IVB: 42.6% + 4.1% (10.6%)Percentages (%) for OS and DSS indicate calculated 5-year survival estimates Stage migration and 5-year OS by overall pathologic stage Percentages (%) for OS and DSS indicate calculated 5-year survival estimates Table 10Stage migration and 5-year DSS by overall pathologic stageAJCC 7thStage migration (AJCC 8th)AJCC 8thΔDSS (% change)I: 93.8%I: 91.8%I: 91.8%−2.0% (2.1%)II: 96.5%II: 88.5%II: 88.8%II: 91.0% + 2.5% (2.8%)III: 87.6%III: 71.7%III: 75.9%III: 80.8% + 9.1% (12.7%)IVA: 54.7%IVA: 62.4%IVA: 68.7%IVA: 65.9% + 3.5% (5.6%)IVB: 51.6%IVB: 44.9%IVB: 44.9%IVB: 50.3% + 5.4% (12.0%)Percentages (%) for OS and DSS indicate calculated 5-year survival estimates Stage migration and 5-year DSS by overall pathologic stage Percentages (%) for OS and DSS indicate calculated 5-year survival estimates Evaluation of AJCC 7th and AJCC 8th Figure 1A–C show the Kaplan–Meier survival curves for the AJCC 7th edition in the SYSU-OSCC cohort. The AJCC 7th edition showed limited discrimination between pT3 and pT4a for both OS and DSS (5-year OS rates: 52.5% vs. 60.4%, p = 0.48; 5-year DSS rates: 63.1% vs. 67.7%, p = 0.97). In addition, survival overlap was observed between pathologic stage I and stage II (5-year OS rates: 89.1% vs. 82.3%, p = 0.67; 5-year DSS rates: 93.8% vs. 88.5%, p = 0.27).Fig. 1Kaplan–Meier curves for AJCC 7th in the SYSU-OSCC cohort (n = 937). OS and DSS of AJCC 7th pT category (A), pN category (B), pStage category (C) in the SYSU-OSCC cohort Kaplan–Meier curves for AJCC 7th in the SYSU-OSCC cohort (n = 937). OS and DSS of AJCC 7th pT category (A), pN category (B), pStage category (C) in the SYSU-OSCC cohort After restaging according to the final AJCC 8th edition, the pT4a category showed significantly poorer survival than the pT3 category in the SYSU-OSCC cohort (5-year OS rates: 69.2% vs. 51.7%, p = 0.012; 5-year DSS rates: 76.6% vs. 62.0%, p = 0.014). The 8th edition showed improved discrimination among intermediate and advanced pT categories, including pT3, pT4a, and pT4b. However, prognostic overlap persisted between pT1 and pT2 (5-year OS rates: 82.4% vs. 76.9%, p = 0.14; 5-year DSS rates: 87.0% vs. 83.9%, p = 0.45), and between pathologic stage I and stage II (5-year OS rates: 89.7% vs. 83.6%, p = 0.55; 5-year DSS rates: 91.8% vs. 91.0%, p = 0.98) (Fig. 2A–C).Fig. 2Kaplan–Meier curves for AJCC 8th in the SYSU-OSCC cohort (n = 937). A–C OS and DSS of AJCC 8th pT category (A), pN category (B), pStage category (C) in the SYSU-OSCC cohort. D Radar chart for hazard consistency, hazard discrimination, sample size balance, and outcome prediction (the smaller value, the better performance) Kaplan–Meier curves for AJCC 8th in the SYSU-OSCC cohort (n = 937). A–C OS and DSS of AJCC 8th pT category (A), pN category (B), pStage category (C) in the SYSU-OSCC cohort. D Radar chart for hazard consistency, hazard discrimination, sample size balance, and outcome prediction (the smaller value, the better performance) Staging performance was evaluated according to the established Groome criteria using the online tool developed by Ren et al. [17]. The AJCC 8th edition showed better overall performance than the AJCC 7th edition, particularly in hazard consistency and sample size balance, with lower scores indicating better performance (Fig. 2D, also displayed in Table 11).Table 11Comparison of the performance between the AJCC 7th edition and 8th edition under four evaluation criteria in SYSU-OSCC cohort, related to Fig. 2DEvaluation criteriaSYSU-OSCC cohortAJCC 7thAJCC 8thHazard Consistency4.50 ± 0.794.10 ± 0.84 Normalized score0.68 ± 0.460.31 ± 0.46 Normalized rank1.681.31Hazard Discrimination0.53 ± 0.120.56 ± 0.17 Normalized score0.55 ± 0.500.45 ± 0.50 Normalized rank1.551.45Balance0.39 ± 0.040.31 ± 0.05 Normalized score0.92 ± 0.270.08 ± 0.27 Normalized rank1.921.08Outcome prediction score0.29 ± 0.0190.29 ± 0.01 Normalized score0.59 ± 0.490.40 ± 0.49 Normalized rank1.591.40Overall score2.38 ± 0.901.62 ± 0.90 Normalized overall rank1.491.18Times in rank 1102163 Table 11Comparison of the performance between the AJCC 7th edition and 8th edition under four evaluation criteria in SYSU-OSCC cohort, related to Fig. 2DEvaluation criteriaSYSU-OSCC cohortAJCC 7thAJCC 8thHazard Consistency4.50 ± 0.794.10 ± 0.84 Normalized score0.68 ± 0.460.31 ± 0.46 Normalized rank1.681.31Hazard Discrimination0.53 ± 0.120.56 ± 0.17 Normalized score0.55 ± 0.500.45 ± 0.50 Normalized rank1.551.45Balance0.39 ± 0.040.31 ± 0.05 Normalized score0.92 ± 0.270.08 ± 0.27 Normalized rank1.921.08Outcome prediction score0.29 ± 0.0190.29 ± 0.01 Normalized score0.59 ± 0.490.40 ± 0.49 Normalized rank1.591.40Overall score2.38 ± 0.901.62 ± 0.90 Normalized overall rank1.491.18Times in rank 1102163 Comparison of the performance between the AJCC 7th edition and 8th edition under four evaluation criteria in SYSU-OSCC cohort, related to Fig. 2D In the SEER-OSCC cohort, the AJCC 8th edition showed limited separation between the pT3 and pT4a categories for both OS and DSS (3-year OS rates: 58.1% vs. 54.3%, p = 0.16; 3-year DSS rates: 64.5% vs. 63.4%, p = 0.41) (Fig. 3A–C). Notably, no prognostic overlap was observed between the pT1 and pT2 categories in the SEER-OSCC cohort. This pattern differed from that observed in the SYSU-OSCC cohort.Fig. 3Kaplan–Meier curves for AJCC 8th in the SEER-OSCC cohort (n = 2589). A–C OS and DSS of AJCC 8th pT category (A), pN category (B), pStage category (C) in the SEER-OSCC cohort Kaplan–Meier curves for AJCC 8th in the SEER-OSCC cohort (n = 2589). A–C OS and DSS of AJCC 8th pT category (A), pN category (B), pStage category (C) in the SEER-OSCC cohort Harrell’s C-index was additionally calculated to provide a supplementary assessment of prognostic discrimination in both cohorts, and the detailed results are presented [see Supplementary Table 3]. Refinement of the T classification To improve the prognostic separation between AJCC 8th pT1 and pT2 categories in the SYSU-OSCC cohort, subsets of pT2 tumors with favorable features were evaluated for possible reclassification as pT1. The SYSU-OSCC cohort was randomly divided into a training cohort and an internal validation cohort using a 1:1 ratio for model development and validation. In the training cohort, after adjusting for age, sex and treatment modality, multivariable Cox regression analyses showed that tumors with a pDOI > 5 mm and ≤ 10 mm and a tumor size ≤ 2 cm had OS and DSS risks comparable to those of pT1 tumors (Table 12).Table 12Multivariable analysis of refined pT categories in the SYSU-OSCC cohortRisk stratificationNumberAHROS (95% CI)p-valueAHRDSS (95% CI)p-valueTraining cohort pDOI ≤ 5 mm with Tumor size ≤ 2 cm140ReferenceReferenceReferenceReference pDOI ≤ 5 mm with Tumor size > 2 and ≤ 4 cm601.36 (0.77, 2.42)0.281.44 (0.69, 3.02)0.33 pDOI > 5 and ≤ 10 mm with Tumor size ≤ 2 cm521.13 (0.59, 2.14)0.721.14 (0.50, 2.60)0.75 pDOI > 5 and ≤ 10 mm with Tumor size > 2 and ≤ 4 cm871.45 (0.87, 2.41)0.151.06 (0.52, 2.19)0.86 pDOI > 10 mm with Tumor size ≤ 2 cm141.55 (0.61, 3.97)0.361.51 (0.45, 5.10)0.50Validation cohort pDOI ≤ 5 mm with Tumor size ≤ 2 cm133ReferenceReferenceReferenceReference pDOI ≤ 5 mm with Tumor size > 2 and ≤ 4 cm531.30 (0.85, 1.99)0.231.33 (0.77, 2.32)0.30 pDOI > 5 and ≤ 10 mm with Tumor size ≤ 2 cm491.16 (0.73, 1.83)0.530.81 (0.41, 1.59)0.54 pDOI > 5 and ≤ 10 mm with Tumor size > 2 and ≤ 4 cm811.55 (1.07, 2.25)0.021.60 (0.99, 2.57)0.05 pDOI > 10 mm with Tumor size ≤ 2 cm101.53 (0.73, 3.20)0.250.94 (0.29, 3.05)0.92Age, sex (male vs. female), radiation (No vs. Yes) were included in the multivariable model. Multivariable model was stratified on pN category (pN3 vs. pN2 vs. pN1 vs. pN0) because of nonproportional hazards Table 12Multivariable analysis of refined pT categories in the SYSU-OSCC cohortRisk stratificationNumberAHROS (95% CI)p-valueAHRDSS (95% CI)p-valueTraining cohort pDOI ≤ 5 mm with Tumor size ≤ 2 cm140ReferenceReferenceReferenceReference pDOI ≤ 5 mm with Tumor size > 2 and ≤ 4 cm601.36 (0.77, 2.42)0.281.44 (0.69, 3.02)0.33 pDOI > 5 and ≤ 10 mm with Tumor size ≤ 2 cm521.13 (0.59, 2.14)0.721.14 (0.50, 2.60)0.75 pDOI > 5 and ≤ 10 mm with Tumor size > 2 and ≤ 4 cm871.45 (0.87, 2.41)0.151.06 (0.52, 2.19)0.86 pDOI > 10 mm with Tumor size ≤ 2 cm141.55 (0.61, 3.97)0.361.51 (0.45, 5.10)0.50Validation cohort pDOI ≤ 5 mm with Tumor size ≤ 2 cm133ReferenceReferenceReferenceReference pDOI ≤ 5 mm with Tumor size > 2 and ≤ 4 cm531.30 (0.85, 1.99)0.231.33 (0.77, 2.32)0.30 pDOI > 5 and ≤ 10 mm with Tumor size ≤ 2 cm491.16 (0.73, 1.83)0.530.81 (0.41, 1.59)0.54 pDOI > 5 and ≤ 10 mm with Tumor size > 2 and ≤ 4 cm811.55 (1.07, 2.25)0.021.60 (0.99, 2.57)0.05 pDOI > 10 mm with Tumor size ≤ 2 cm101.53 (0.73, 3.20)0.250.94 (0.29, 3.05)0.92Age, sex (male vs. female), radiation (No vs. Yes) were included in the multivariable model. Multivariable model was stratified on pN category (pN3 vs. pN2 vs. pN1 vs. pN0) because of nonproportional hazards Multivariable analysis of refined pT categories in the SYSU-OSCC cohort Age, sex (male vs. female), radiation (No vs. Yes) were included in the multivariable model. Multivariable model was stratified on pN category (pN3 vs. pN2 vs. pN1 vs. pN0) because of nonproportional hazards Based on these findings, tumors with a pDOI > 5 mm and ≤ 10 mm and a tumor size ≤ 2 cm were proposed for reclassification into the pT1 category. This proposed reclassification was further supported by the results from the internal validation cohort (Table 12). The proposed pT classification was then evaluated in both the training and internal validation cohorts (Table 13).Table 13Multivariable analysis for proposed pT classification in the SYSU-OSCC cohortNumberAHROS (95% CI)p-valueAHRDSS (95% CI)p-valueTraining cohort Proposed pT1192ReferenceReferenceReferenceReference Proposed pT21611.38 (0.93, 2.06)0.111.20 (0.70, 2.06)0.50 pT3841.94 (1.22, 3.07)0.0042.06 (1.15, 3.69)0.015 pT4a463.77 (2.27, 6.26) < 0.0013.71 (1.96, 6.98) < 0.001 pT4b76.26 (2.18, 18.02) < 0.0016.40 (1.86, 22.03)0.003Validation cohort Proposed pT1182ReferenceReferenceReferenceReference Proposed pT21441.39 (0.91, 2.13)0.131.52 (1.03, 2.24)0.004 pT3761.99 (1.22, 3.24)0.0062.42 (1.58, 3.71) < 0.001 pT4a393.50 (2.04, 5.98) < 0.0014.75 (3.02, 7.47) < 0.001 pT4b69.29 (3.12, 27.66) < 0.00111.08 (4.83, 25.39) < 0.001Age, sex (male vs. female), radiation (No vs. Yes) were included in the multivariable model. Multivariable model was stratified on pN category (pN3 vs. pN2 vs. pN1 vs. pN0) because of nonproportional hazards Table 13Multivariable analysis for proposed pT classification in the SYSU-OSCC cohortNumberAHROS (95% CI)p-valueAHRDSS (95% CI)p-valueTraining cohort Proposed pT1192ReferenceReferenceReferenceReference Proposed pT21611.38 (0.93, 2.06)0.111.20 (0.70, 2.06)0.50 pT3841.94 (1.22, 3.07)0.0042.06 (1.15, 3.69)0.015 pT4a463.77 (2.27, 6.26) < 0.0013.71 (1.96, 6.98) < 0.001 pT4b76.26 (2.18, 18.02) < 0.0016.40 (1.86, 22.03)0.003Validation cohort Proposed pT1182ReferenceReferenceReferenceReference Proposed pT21441.39 (0.91, 2.13)0.131.52 (1.03, 2.24)0.004 pT3761.99 (1.22, 3.24)0.0062.42 (1.58, 3.71) < 0.001 pT4a393.50 (2.04, 5.98) < 0.0014.75 (3.02, 7.47) < 0.001 pT4b69.29 (3.12, 27.66) < 0.00111.08 (4.83, 25.39) < 0.001Age, sex (male vs. female), radiation (No vs. Yes) were included in the multivariable model. Multivariable model was stratified on pN category (pN3 vs. pN2 vs. pN1 vs. pN0) because of nonproportional hazards Multivariable analysis for proposed pT classification in the SYSU-OSCC cohort Age, sex (male vs. female), radiation (No vs. Yes) were included in the multivariable model. Multivariable model was stratified on pN category (pN3 vs. pN2 vs. pN1 vs. pN0) because of nonproportional hazards We evaluated the performance of the two versions of staging systems for patients in the training and internal validation cohorts. Fig. 4A–D showed the Kaplan–Meier curves for the proposed pT classification in the two cohorts. The pT2 category showed significantly poorer survival than the pT1 category. The overall pathologic stage based on the proposed pT classification showed a similar pattern, with clearer prognostic stratification. Survival curve overlap was reduced compared with the original AJCC 8th edition. Model performance was then evaluated according to the Groome criteria. The proposed classification showed better performance than the original AJCC 8th edition, particularly in hazard consistency, hazard discrimination, outcome prediction, and sample size balance (Fig. 4E–F; also displayed in Table 14). Lower scores indicated better performance. Harrell’s C-index was additionally calculated in both cohorts and the results are presented [see Supplementary Table 4].Fig. 4Kaplan–Meier curves for proposed classification and comparison of overall stage models in the SYSU-OSCC training (n = 490) and validation (n = 447) cohorts. A–B OS and DSS of proposed pT category (A), pStage category (B) in the training cohort. C–D OS and DSS for proposed pT category (C), pStage category (D) in the validation cohort. E–F Radar chart for hazard consistency, hazard discrimination, sample size balance, and outcome prediction in training cohort (E), and validation cohort (F) (lower values indicate better performance)Table 14Comparison of the performance between the exploratory proposed edition and current 8th edition under four evaluation criteria in the SYSU-OSCC training and internal validation cohorts, related to Figs. 4E–4FEvaluation criteriaTraining cohortValidation cohortProposed editionAJCC 8thProposed editionAJCC 8thHazard Consistency4.43 ± 0.874.80 ± 0.905.14 ± 1.015.29 ± 1.05 Normalized score0.37 ± 0.480.63 ± 0.480.40 ± 0.500.56 ± 0.50 Normalized rank1.371.631.401.56Hazard Discrimination0.64 ± 0.230.61 ± 0.210.62 ± 0.170.61 ± 0.22 Normalized score0.46 ± 0.500.54 ± 0.500.50 ± 0.500.50 ± 0.50 Normalized rank1.471.541.51.5Balance0.22 ± 0.030.33 ± 0.050.28 ± 0.050.41 ± 0.05 Normalized score0.04 ± 0.200.96 ± 0.200.03 ± 0.170.97 ± 0.17 Normalized rank1.041.961.031.97Outcome prediction score0.29 ± 0.020.29 ± 0.020.29 ± 0.020.29 ± 0.02 Normalized score0.47 ± 0.500.53 ± 0.500.46 ± 0.500.55 ± 0.50 Normalized rank1.471.531.461.55Overall score1.67 ± 0.782.48 ± 0.881.55 ± 0.932.46 ± 0.93 Normalized overall rank1.121.491.171.51Times in rank 117510316598 Kaplan–Meier curves for proposed classification and comparison of overall stage models in the SYSU-OSCC training (n = 490) and validation (n = 447) cohorts. A–B OS and DSS of proposed pT category (A), pStage category (B) in the training cohort. C–D OS and DSS for proposed pT category (C), pStage category (D) in the validation cohort. E–F Radar chart for hazard consistency, hazard discrimination, sample size balance, and outcome prediction in training cohort (E), and validation cohort (F) (lower values indicate better performance) Table 14Comparison of the performance between the exploratory proposed edition and current 8th edition under four evaluation criteria in the SYSU-OSCC training and internal validation cohorts, related to Figs. 4E–4FEvaluation criteriaTraining cohortValidation cohortProposed editionAJCC 8thProposed editionAJCC 8thHazard Consistency4.43 ± 0.874.80 ± 0.905.14 ± 1.015.29 ± 1.05 Normalized score0.37 ± 0.480.63 ± 0.480.40 ± 0.500.56 ± 0.50 Normalized rank1.371.631.401.56Hazard Discrimination0.64 ± 0.230.61 ± 0.210.62 ± 0.170.61 ± 0.22 Normalized score0.46 ± 0.500.54 ± 0.500.50 ± 0.500.50 ± 0.50 Normalized rank1.471.541.51.5Balance0.22 ± 0.030.33 ± 0.050.28 ± 0.050.41 ± 0.05 Normalized score0.04 ± 0.200.96 ± 0.200.03 ± 0.170.97 ± 0.17 Normalized rank1.041.961.031.97Outcome prediction score0.29 ± 0.020.29 ± 0.020.29 ± 0.020.29 ± 0.02 Normalized score0.47 ± 0.500.53 ± 0.500.46 ± 0.500.55 ± 0.50 Normalized rank1.471.531.461.55Overall score1.67 ± 0.782.48 ± 0.881.55 ± 0.932.46 ± 0.93 Normalized overall rank1.121.491.171.51Times in rank 117510316598 Comparison of the performance between the exploratory proposed edition and current 8th edition under four evaluation criteria in the SYSU-OSCC training and internal validation cohorts, related to Figs. 4E–4F Discussion In this study, we systematically evaluated the impact of incorporating pDOI and pENE on stage migration and prognostic stratification in OSCC under the AJCC 8th staging system. Incorporation of pDOI led to an upstaging of pT in approximately one-third of patients, whereas pENE was associated with an upstaging of pN in about 10% of patients. The observed proportion of pENE-driven upstaging in our cohort was notably lower than the 29.2% reported by Matos et al. [18] and the 22.9% reported by Mattavelli et al. [19], likely reflecting differences in patient populations and pathological assessment practices. In the SYSU-OSCC cohort, the AJCC 8th pT classification demonstrated greater improvement in survival outcomes for patients classified as pT2 or pT3, but survival declined in the pT4a category. These results indicated that a subset of tumors with greater depth of invasion, more aggressive biological behavior, and poorer prognosis had not been adequately identified. These tumors were subsequently reclassified as pT4a in the AJCC 8th edition, resulting in a more concentrated high-risk population within this category. This adjustment addressed the previously observed insufficient separation between pT3 and pT4a in the AJCC 7th edition. However, it should be noted that changes in stage-specific survival following restaging may partially reflect patient redistribution across stage categories rather than a true improvement in individual prognostic prediction. Before classifying tumors with pDOI > 10 mm and tumor size > 4 cm as pT4a, results from the SYSU-OSCC cohort indicated substantial overlap between the survival curves of the pT3 and pT4 categories. Several studies have reported consistent findings, indicating that no significant prognostic distinction existed between pT3 and pT4 categories after pDOI incorporation in the AJCC 8th edition [14, 15, 20]. Additionally, a meta-analysis indicated that the risks associated with T3 and T4 are highly comparable [21]. Notably, a study by Yusuke Yokota et al. [22], which incorporated clinical DOI and ENE, reported conclusions similar to those observed in the SYSU-OSCC cohort of the present study. Furthermore, no overlap was observed between pT1 and pT2 categories in the SEER-OSCC cohort. These discrepancies may be attributable to differences in database sources, cohort size, and potential inaccuracies associated with retrospective restaging and measurement variability [15]. Shrinkage of pathological sections following excision and formalin fixation may also contribute [23]. In the present study, all cases were uniformly restaged based on pDOI measured directly from histological slides, which may explain the significantly improved discrimination observed among intermediate and advanced pT categories. Another major modification introduced was the incorporation of pENE. ENE has been consistently demonstrated to serve as an important independent prognostic factor in oral cancer [24, 25]. In the SYSU-OSCC cohort of the present study, pathologic extranodal extension was identified in 7.5% of all patients and 29.9% of patients with nodal metastasis. Patients upstaged to pN2 or pN3 due to pENE exhibited significantly worse 5-year OS than non-upstaged counterparts, with survival risks comparable to those of the traditional high-risk group. Several recent publications have suggested that positive lymph node counts are associated with poor prognosis and may guide the implementation of treatment intensification [26]. Meanwhile, several modified nodal staging systems have been proposed to incorporate the number of positive metastatic lymph nodes, offering innovative refinement to existing classification systems [20, 27–29]. The results of these studies may provide additional insights for further refinement of the TNM staging system. Notably, survival analysis revealed pronounced numerical fluctuations within the pN2a category. This instability was largely attributable to the extremely small number of patients classified as pN2a under the AJCC 7th edition, followed by restructuring after incorporation of pENE in the 8th edition. This redistribution inevitably affected the stability of survival estimates and underscored the necessity of interpreting subgroup analyses while considering sample size and confidence intervals. At the pathologic stage level, AJCC 8th demonstrated a clearer survival gradient among patients with stage III–IV disease compared with AJCC 7th, particularly regarding prognostic distinction between stages IVA and IVB, which appeared more clinically rational. These results are consistent with those reported by Singhavi et al. [30]. Collectively, the AJCC 8th edition staging system represents a substantial advancement in risk re-stratification for intermediate and advanced OSCC. However, certain limitations remain. In the SYSU-OSCC cohort, considerable prognostic overlap remained between the pT1 and pT2 categories, which was also observed in earlier pathologic stages. To address this issue, tumors with a pDOI > 5 mm and ≤ 10 mm and a tumor size ≤ 2 cm were proposed to be reclassified into the pT1 category. Performance evaluation suggested that this exploratory modification improved staging performance within the training and internal validation cohorts, including hazard consistency within stages, hazard discrimination between stages, outcome prediction, and sample size balance. Overall, incorporation of pDOI and pENE into the AJCC 8th staging system enabled more precise identification of high-risk OSCC patients, demonstrating superior risk concentration and clinical utility. Our findings also suggested that targeted adjustments at specific staging thresholds could further enhance prognostic accuracy. However, due to the relatively small sample size of SYSU-OSCC cohort, assessment of certain classifications (e.g., pN2a) was not sufficiently comprehensive. Moreover, as the information required to accurately reconstruct AJCC 7th staging was incomplete in the SEER database, the SEER-OSCC cohort could not be staged according to the AJCC 7th edition. Therefore, this study was unable to externally confirm whether the AJCC 8th edition outperformed the AJCC 7th edition. Importantly, the proposed refinement was exploratory and has not yet been externally validated in larger, independent populations. Nevertheless, substantial variability in evaluation results persists across data sources, implying that when interpreting the relevant research results and developing risk stratification strategies, heterogeneity across different populations and clinical contexts should be carefully considered. These limitations underscore the need for larger and more diverse cohorts to improve the robustness and generalizability of the findings. Conclusion The AJCC 8th staging system improved risk stratification for OSCC through the incorporation of pDOI and pENE, particularly in patients with intermediate- and advanced-stage disease. However, residual prognostic overlap remained between selected pT categories. The proposed pT refinement suggests a trend toward better prognostic stratification in the SYSU-OSCC cohorts, but the observed improvement remains limited. Due to the absence of robust external validation, this refinement cannot currently be utilized to guide clinical implementation or individualized treatment decisions. Rigorous validation in larger, independent external cohorts remains an essential prerequisite. Supplementary Information Supplementary Material 1. Supplementary Material 1.