Our study was conducted in full accordance with ethical principles, including the World Medical Association’s Declaration of Helsinki (2002 version), with IRB approval (CMUSH-IRB-KJ-PJ-2024-29). We performed a single-institution, retrospective cohort study. From January 2016 to December 2021, patients treated in Beijing Stomatological Hospital who satisfied the following inclusion criteria were included. The process for selecting patients for model development is presented in Figure 1. The inclusion criteria were as follows: 1) patients pathologically diagnosed as squamous cell carcinoma. 2) With a tumor located on the tongue, lower gingiva, upper gingiva, buccal mucosa, floor of the mouth, or hard palate. 3) No preoperative chemotherapy, radiotherapy or other radical treatments before the operation; Relevant routine laboratory indicators of peripheral blood should be examined within 7 days before the operation; Patients were excluded if the follow-up time was less than 1 year, and complete clinical and laboratory data were not completed.

The dataset includes patients’ baseline information (age, gender, 8^th AJCC TNM stage, tobacco and alcohol use), Preoperative routine hematological parameters within 7 days (absolute neutrophil count, absolute monocyte count, absolute lymphocyte count, platelet count, albumin, hemoglobin). and follow-up data.

All the indicators mentioned in our study were evaluated and collected as follows:

(1) Neutrophil-to-Lymphocyte Ratio (NLR) = Absolute neutrophil count/Absolute lymphocyte count

Patients were followed for at least 1 year after the operation, and phone call records or visits to the clinic were used for the verification of disease and survival status.

Normally distributed continuous data are reported as the means ± standard deviation (SD). Qualitative data are described by frequency and percentage. the outcome assessment parameters included overall survival (OS) and disease-free survival (DFS). DFS was defined as the time from the first day after treatment to the time of disease progression (including local/regional recurrence, second primary or distant metastasis) or death due to any cause. OS was defined as the period from the end of treatment until death caused by any reason. A receiver operating characteristic (ROC) curve was used to determine the optimal cutoff values for the indicators related to preoperative routine hematological parameters. The Kaplan-Meier method was used to compare the outcomes of the different risk groups Statistical significance was determined with the log-rank test, to explore the timeliness of each indicator, time-dependent receiver operating characteristic curves were constructed using the “survival ROC” package in R software, with calculation of the area under the curve (AUC) for postoperative survival at 1–5 years. The relevant forest plots were drawn using the “forest plot” package in R software. The analyses were performed with SPSS 22.0 (IBM, Chicago, IL, United States), and a two-sided P-value <0.05 was considered significant.

According to the inclusion criteria, we identified a total of 624 patients with OSCC who underwent primary surgery with surgically negative margins. Among these, 354 patients were men and 270 were women, 1:1.31. The most common tumor site was the tongue (39.7%), followed by the gingiva (26.4%). The proportions of smokers and drinkers were 41.7% and 34.5% respectively. From the clinical staging perspective, the cohort included 59 cases of Tis (9.5%), 98 Stage I cases (15.7%), 150 Stage II cases (24.0%), 68 Stage III cases (10.9%), and 249 Stage IV cases (39.9%), while the pathological T staging comprised 62 Tis cases (9.9%), 87 T1 cases (13.9%), 174 T2 cases (27.9%), 94 T3 cases (15.1%), and 207 T4 cases (33.2%).

For the indicators related to preoperative routine hematological parameters, the mean NLR was 2.09 ± 0.06, and for the other indicators, the data were as follows: mean PLR: 131.85 ± 2.03, mean LMR:4.52 ± 0.07, mean SII: 470.0 ± 14.74, mean PNI: 45.61 ± 0.18, mean ALI:54.60 ± 1.13, mean SIM:0.95 ± 0.06, and H-index: 2.19 ± 0.29. Further, the mentioned continuous variables were converted to binary classifications by cutoff values as follows: NLR (2.3), PLR (132.7), LMR (3.0), SII (661.6), PNI (47.3), ALI (40.0), SIM (1.6), and H-index (2.7). Previous studies have demonstrated that inflammatory states are characterized by elevated absolute neutrophil count, absolute monocyte count, and platelet count, coupled with reduced serum albumin and absolute lymphocyte count (Wu et al., 2022; Yang et al., 2019; Sakai et al., 2023; Cao and Kagan, 2023; Wei et al., 2021; Lin et al., 2022; Gaudioso et al., 2021; Murad et al., 2022; Valero et al., 2020; Diao et al., 2018; Park et al., 2018). Consequently, based on established formulas for these parameters, patients with elevated NLR, PLR, SII, SIM, and H-index, or reduced LMR, PNI, and ALI, were categorized as having a high-inflammatory trend. The remaining cases were classified as having a low-inflammatory trend.

Through February 2023, the end of the follow-up, the median follow-up time was 35months (IQR: 23-52). During the follow-up period, a total of 116 death events were observed (all-cause mortality rate 18.6%), of which 199 patients (31.9%) developed local recurrence or lymph node metastasis. The baseline characteristics of the patients are summarized in Table 1.

To evaluate the impact of preoperative inflammatory status on clinical outcomes, we performed Kaplan-Meier survival analysis comparing postoperative OS and DFS between patients with high- and low-inflammatory trend as mentioned above. The survival curves showed significantly worse OS and DFS in patients with high-inflammatory trends in each inflammatory-related indicator (log-rank test, P < 0.005 for both endpoints), as illustrated in Figure 2. This result suggests that the preoperative systemic inflammatory status may be an independent risk factor affecting the survival prognosis of patients with OSCC.

Supported by the above evidence, we further analyzed the prognostic predictive ability of each indicator item by item based on Cox univariate or multivariate analysis. Before applying the COX univariate and multivariate regression analysis, the proportional hazards of each index were verified (Supplementary Figure S1). Univariate Cox regression analysis of postoperative overall survival (OS) identified significant associations with the following variables: clinical stage (III vs 0, HR = 8.523, 95% CI: 1.080–67.277,P = 0.042), and inflammation-related indicators including NLR (high vs. low: HR = 2.599, 95% CI: 1.776-3.686, P < 0.001), PLR (high vs. low: HR = 1.986, 95% CI: 1.374-2.869, P < 0.001), LMR (high vs. low: HR = 0.345, 95% CI: 0.232-0.513, P < 0.001), SII (high vs. low: HR = 2.593, 95% CI: 1.759-3.822, P < 0.001), PNI (high vs. low: HR = 0.388, 95% CI: 0.242-0.623, P < 0.001), ALI (high vs. low: HR = 0.300, 95% CI: 0.207–0.434, P < 0.001), SIM (high vs. low: HR = 3.083, 95% CI: 2.034–4.672, P < 0.001), and H-index (high vs low: HR = 3.275, 95% CI: 2.246–4.776, P < 0.001). After adjusting for confounding factors, we performed Multivariate Cox regression analysis and confirmed that 2 inflammation-related indicators served as independent prognostic factors for OS in OSCC patients (Table 2). Specifically, elevated levels of PNI (high vs low: HR = 0.583, 95% CI: 0.348–0.977, P = 0.041), and ALI (high vs low: HR = 0.502, 95% CI: 0.282–0.892, P = 0.019) demonstrated significant correlations with OS, indicating that elevated levels of these markers may correlate with patient survival (Table 2).

Similarly, Univariate Cox regression analysis of postoperative DFS risk factors demonstrated significant associations with sex, age, smoking and drinking history, clinical stage, and all inflammation-related markers (Supplementary Table S1). Multivariate Cox regression analysis after adjusting for confounding factors confirmed that only clinical stage served as an independent prognostic factor for DFS in OSCC patients (Supplementary Table S1). This result indicated that all inflammation-related indicators only influence the overall survival status of patients. In addition, time-dependent ROC revealed that among the eight included indicators, the ALI demonstrated the optimal predictive performance, achieving a 1-year postoperative AUC of 0.65, though no statistically significant differences were observed between the indicator groups. Notably, seven preoperative markers—NLR, LMR, SII, ALI, PLR, SIM, and H-index—exhibited a significant decline in AUC values at the 2-year postoperative interval compared to the 1-year results (Figure 3A), suggesting that these preoperative indicators may only hold predictive value for short-term prognosis after treatment.

In order to further clarify the prognostic value of the above indicators and minimize the influence of potential confounding factors, we conducted stratified subgroup analyses using multivariate Cox proportional hazards models to evaluate age-specific (<60 vs. ≥60 years) and gender-specific prognostic associations. The indicators related to preoperative routine hematological parameters demonstrated differential predictive efficacy across subgroups, with PNI (HR = 0.256, 95% CI = 0.125–0.522, P < 0.001), ALI (HR = 0.372, 95% CI = 0.211–0.658, P = 0.001), along with clinical stage showing independent associations with postoperative OS in younger patients (<60 years, n = 257). Similarly, PNI also correlated with DFS in younger patients. In contrast, none exhibited significant prognostic value for OS or DFS in elderly patients (≥60 years, n = 367), indicating age-dependent prognostic relevance of preoperative inflammation-related indicators in OSCC (Table 3; Supplementary Tables S2-S4).

Similarly, in the male subgroup (n = 354), multivariate-adjusted analysis demonstrated that 2 indicators independently influenced postoperative OS, including PNI: HR = 0.371, 95% CI = 0.175- 0.789, P = 0.010, ALI: HR = 0.468, 95% CI = 0.223–0.980, P = 0.044. However, only the clinical stages serve as independent prognostic factors for DFS. Conversely, in female patients (n = 270), none of these biomarkers showed independent prognostic significance for either OS or DFS (P > 0.05) (Table 4; Supplementary Tables S5-S7).

Building upon these findings, the aforementioned indicators demonstrated significant prognostic predictive value exclusively in patients under 60 years and the younger male subgroups. To evaluate specific populations, we further constructed time-dependent ROC curves and systematically calculated AUC values at critical postoperative intervals (1-year, 3-year, and 5-year). Analytical results revealed that compared to the overall population, all indicators exhibited enhanced predictive efficacy in this targeted subgroup despite maintaining analogous prognostic trends. Notably, ALI demonstrated superior predictive performance among all parameters, achieving an AUC of 0.74 during the first postoperative year, significantly outperforming other indicators (Figure 3B). Furthermore, we evaluated the differences in prognostic effects of the 8 indicators among patients of different ages and genders through interaction analysis. Based on the Cox proportional hazards regression model, patients were divided into different subgroups by age (≥60 years vs. <60 years) and gender. The HR and 95%CI of each indicator in each subgroup were calculated, and the interaction P-values were used to assess the statistical differences of the 8 indicators across different stages and subgroups. Forest plots visualizing the results indicated that ALI and PNI, like previously mentioned, also demonstrated statistically significant effects in patients under 60 years old and in male patients (Figure 4). This discovery suggests ALI and PNI may serve as clinically valuable biomarkers for prognostic evaluation in male OSCC patients under 60 years.