Skip to main content An official website of the United States government Here's how you know Here's how you know Official websites use .gov A .gov website belongs to an official government organization in the United States. Secure .gov websites use HTTPS A lock ( Lock Locked padlock icon ) or https:// means you've safely connected to the .gov website. Share sensitive information only on official, secure websites. Search Log in Dashboard Publications Account settings Log out Search… Search NCBI Primary site navigation Search Logged in as: Dashboard Publications Account settings Log in Search PMC Full-Text Archive Search in PMC Journal List User Guide PERMALINK Copy As a library, NLM provides access to scientific literature. Inclusion in an NLM database does not imply endorsement of, or agreement with, the contents by NLM or the National Institutes of Health. Learn more: PMC Disclaimer | PMC Copyright Notice Int Dent J . 2026 Mar 31;76(3):109480. doi: 10.1016/j.identj.2026.109480 Search in PMC Search in PubMed View in NLM Catalog Add to search Aspartate Aminotransferase/Alanine Aminotransferase Ratio Predicts the Post-surgical Prognosis of Patients with Oral Squamous Cell Carcinoma Baochang He Baochang He a Stomatological Center, Department of Oral and Maxillofacial Surgery, The First Affiliated Hospital, Fujian Medical University, Fuzhou, China b Department of Epidemiology and Health Statistics, Fujian Provincial Key Laboratory of Environment Factors and Cancer, School of Public Health, Fujian Medical University, Fuzhou, China Find articles by Baochang He a, b, # , Liling Shen Liling Shen b Department of Epidemiology and Health Statistics, Fujian Provincial Key Laboratory of Environment Factors and Cancer, School of Public Health, Fujian Medical University, Fuzhou, China Find articles by Liling Shen b, # , Yulan Lin Yulan Lin b Department of Epidemiology and Health Statistics, Fujian Provincial Key Laboratory of Environment Factors and Cancer, School of Public Health, Fujian Medical University, Fuzhou, China Find articles by Yulan Lin b , Fa Chen Fa Chen b Department of Epidemiology and Health Statistics, Fujian Provincial Key Laboratory of Environment Factors and Cancer, School of Public Health, Fujian Medical University, Fuzhou, China Find articles by Fa Chen b , Fengqiong Liu Fengqiong Liu b Department of Epidemiology and Health Statistics, Fujian Provincial Key Laboratory of Environment Factors and Cancer, School of Public Health, Fujian Medical University, Fuzhou, China Find articles by Fengqiong Liu b , Yu Qiu Yu Qiu a Stomatological Center, Department of Oral and Maxillofacial Surgery, The First Affiliated Hospital, Fujian Medical University, Fuzhou, China Find articles by Yu Qiu a , Bin Shi Bin Shi a Stomatological Center, Department of Oral and Maxillofacial Surgery, The First Affiliated Hospital, Fujian Medical University, Fuzhou, China Find articles by Bin Shi a , Lisong Lin Lisong Lin a Stomatological Center, Department of Oral and Maxillofacial Surgery, The First Affiliated Hospital, Fujian Medical University, Fuzhou, China Find articles by Lisong Lin a , Jing Wang Jing Wang c Laboratory Center, The Major Subject of Environment and Health of Fujian Key Universities, School of Public Health, Fujian Medical University, Fuzhou, China Find articles by Jing Wang c , Xiaodan Bao Xiaodan Bao d School of Health Management, Fujian Medical University, Fuzhou, China Find articles by Xiaodan Bao d, ⁎ Author information Article notes Copyright and License information a Stomatological Center, Department of Oral and Maxillofacial Surgery, The First Affiliated Hospital, Fujian Medical University, Fuzhou, China b Department of Epidemiology and Health Statistics, Fujian Provincial Key Laboratory of Environment Factors and Cancer, School of Public Health, Fujian Medical University, Fuzhou, China c Laboratory Center, The Major Subject of Environment and Health of Fujian Key Universities, School of Public Health, Fujian Medical University, Fuzhou, China d School of Health Management, Fujian Medical University, Fuzhou, China ⁎ Corresponding author . School of Health Management, Fujian Medical University, 1 Xueyuan Road, Fuzhou 350108, China. [email protected] # These authors have contributed equally to this work. Collection date 2026 Jun. © 2026 The Authors This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). PMC Copyright notice PMCID: PMC13068816 PMID: 41921448 Abstract Objectives Survival rates in patients with oral squamous cell carcinoma (OSCC) have failed to improve greatly over time, and novel biomarkers are urgently needed. Therefore, the aim was to evaluate the predictive ability of the pre-treatment serum aspartate aminotransferase/alanine aminotransferase (AST/ALT) ratio (SLR) for the prognosis of patients with OSCC. Materials and methods This cohort study based on a prospectively maintained database included 552 patients who were diagnosed with primary OSCC and underwent surgery between August 2011 and January 2022. The pre-treatment SLR was determined in each patient by dividing the AST levels by the ALT concentrations. Propensity score matching (PSM) analysis was performed to adjust for potential confounders. COX proportional hazards models, as well as generalized boosted regression model (GBM) and random survival forest (RSF) algorithms were used to investigate the predictive value of the SLR in determining the prognosis of patients with OSCC. Results X-tile software was used to determine the SLR to stratify patients into low-SLR (SLR < 1.50, n = 445) and high-SLR (SLR ≥ 1.50, n = 107). After PSM, all covariates were determined to be well-balanced between the two groups in the matched cohort (all standardized mean differences (SMDs) < 0.1). The high-SLR group exhibited significantly poorer overall survival (OS) compared with that of the low-SLR group based on the multivariate Cox regression modeling in both the original (hazard ratio (HR) = 1.55; 95% confidence interval (CI): 1.02-2.36) and matched (HR = 1.81; 95% CI: 1.13-2.91) cohorts. Furthermore, comparative model analysis demonstrated a moderate improvement in predictive performance when SLR was included in the models ( P = .017 in the original cohort; P = .008 in the matched cohort). Conclusions In conclusion, the pre-treatment SLR serves as an independent prognostic factor for OS in surgical OSCC patients and enhances the performance of prognostic models, indicating its potential as a clinical tool for outcome prediction. Key words: Aspartate aminotransferase, Alanine aminotransferase, SLR, Oral squamous cell carcinoma, Prognosis Introduction Oral cancer represents a common malignant neoplasm in the head and neck region. According to the Global Cancer Observatory (GLOBOCAN) 2022 data estimates, approximately 389,485 new cases were diagnosed, accompanied by 188,230 cancer-related deaths worldwide in 2022. 1 Histologically, oral squamous cell carcinoma (OSCC) is the predominant type of oral cancer. The 5-year survival rate of OSCC has been reported to be approximately 64% to 68%, 2 , 3 and there have been no significant improvements in survival outcomes among patients with OSCC in the past few decades. This persistent plateau underscores the critical need for identifying clinically actionable prognostic biomarkers to achieve precise risk stratification and personalized selection of patients most likely to benefit from specific therapies. Recent cancer research has focused on the characterization of novel biomarkers. 4 As blood-based parameters are easily measurable, various preoperative blood indicators have been used as prognostic predictors in patients with oral cancers. 5 , 6 , 7 , 8 Routine pretreatment evaluations in patients with cancer include the quantification of enzyme concentration to assess liver function, such as aspartate aminotransferase (AST) and alanine aminotransferase (ALT) levels. The AST/ALT ratio (SLR) was described by De Ritis et al 9 in 1957 as a means of distinguishing between various causes of liver disease. Other researchers subsequently confirmed that both aminotransaminases play an important role in regulating cellular metabolism and cancer cell renewal. 10 Numerous studies have shown that a higher SLR may be associated with a poorer prognosis in various cancers, such as renal cell carcinoma, 11 upper urinary tract urothelial carcinoma, 12 hepatocellular carcinoma 13 and prostate cancer. 14 A retrospective study concluded that the SLR could serve as a prognostic factor for overall survival (OS) in patients with head and neck cancers, 15 and another study conducted in Europe found that a high SLR was significantly associated with poorer survival outcomes among patients with oral and oropharyngeal squamous cell carcinomas. 16 However, few studies have investigated the prognostic significance of the SLR in patients with OSCC, and its role has not been verified in the Chinese OSCC population. Therefore, this study aimed to identify a reliable serum prognostic biomarker for OSCC patients undergoing surgery, specifically evaluating the pre-treatment SLR in a cohort from southeastern China. Using propensity score matching (PSM) to control for confounders, we assessed the prognostic significance of the SLR through Cox regression and machine learning (ML) methods, including generalized boosted regression models (GBMs) and random survival forests (RSFs). Methods Sample size justification The minimum sample size was determined using the Cox Regression module in PASS software (Version 08.0.3; NCSS, LLC, Kaysville, UT). The calculation accounted for the following parameters: (1) hazard ratio (HR) = 1.70 (literature-derived range: 1.36-3.09 15 , 16 , 17 , 18 , 19 ); (2) overall event rate ( P ) = 40% (institutional historical data); (3) coefficient of determination ( R ²) = 0.30 (based on our pilot data); (4) standard deviation ( S ) = 0.66 (literature reported 16 ); (5) type I error rate ( α ) = 0.05; (6) statistical power (1- β ) = 90%. The initial estimate required 306 participants. After inflating the sample size by 10% to account for anticipated attrition, the final target was 337 participants. Study design and study population A total of 552 patients who were diagnosed with primary OSCC at the First Affiliated Hospital of Fujian Medical University (Fuzhou City, Fujian Province, China) between August 2011 and January 2022. The inclusion criteria were consistent with those of previous studies, 20 , 21 which can be briefly summarized as: age ≥ 25 years, histologically confirmed primary OSCC, and having undergone surgical treatment. Patients were excluded based on the following criteria: (1) the presence of second primary malignancy; (2) a follow-up time of less than 6 months; (3) pathological diagnosis was unclear or not OSCC; and (4) did not undergo surgical treatment. Data collection and definition Data acquisition followed established methodologies from prior research 20 , 21 : baseline demographic characteristics were systematically collected using interviewer-administered structured questionnaires. Concurrently, data pertaining to clinical features and laboratory indicators were obtained from the patients’ medical records. The cancer stage was determined based on the tumor, node, metastasis (TNM) Classification of Malignant Tumors (AJCC 8th edition), and tumor differentiation was defined according to the World Health Organization (WHO) guidelines. Body mass index (BMI) categorization was based on the following classification standards for Chinese adults: underweight (<18.5 kg/m 2 ), normal weight (18.5-23.9 kg/m 2 ), and overweight/obese (≥ 24 kg/m 2 ). The data for our indicators were obtained from standardized preoperative evaluation results following patient admission, including complete blood count and biochemical tests (comprising AST and ALT). The pre-treatment SLR was calculated by dividing the AST concentration by the ALT levels. Follow-up Follow-up data collection methodology aligned with prior research protocols 20 , 21 : structured telephone interviews were systematically conducted by uniformly trained investigators to ascertain patient outcomes. Survival status was cross-verified against the institutional medical record system. Standardized follow-up assessments were performed at 6-month intervals, with primary endpoints including post-discharge survival outcomes (causes of mortality, date of death, etc.). The follow-up deadline was February, 2025. OS was estimated from the date of the OSCC diagnosis until the date of death or the end of the study, whichever occurred first. Statistical analysis The optimal SLR cutoff value was calculated using X-tile software (Yale University, New Haven, Connecticut). Standardized mean differences (SMDs) were calculated to determine whether statistically significant differences in baseline characteristics existed before and after PSM. The proportional hazards assumption was tested using the Schoenfeld residuals. To identify factors that were independently associated with OS, a multivariate Cox regression model was constructed based on the minimal Akaike information criterion (AIC) from the stepwise regression procedure. Univariate and multivariate Cox regression analyses were performed to evaluate the prognostic significance of factors in predicting OS. Kaplan-Meier curves were constructed to calculate the survival rate, and the log-rank test was used to compare the survival rates between the SLR groups. To better assess the relationships between the pre-treatment SLR and the prognosis of patients with OSCC who underwent surgical intervention, PSM analysis (R package: ‘Matchlt’) was performed to adjust for potential confounders in the different SLR groups. The outcomes of the multivariate logistic regression model in which all relevant confounding variables were included as covariates (including age, sex, place of residence, education, BMI, tobacco use, alcohol consumption, hypertension, diabetes, cancer site, tumor differentiation, TNM stage, postoperative chemotherapy, and postoperative radiation) were calculated alongside the propensity scores of each patient. For PSM, the nearest-neighbor matching algorithm for 2:1 matching and exact matching for two variables (sex and tobacco use) were used to match individuals with different SLR levels. In total, 214 and 107 patients were allocated to the high- and low-SLR groups, respectively. For the GBM- and RSF-based versions, the relative importance (RIMP) or relative influence (RINF) of the pre-treatment SLR in predicting prognosis was also assessed. Conventional receiver operating characteristic (ROC) curves (R package: ‘pROC’) and time-dependent ROC curves (R package: ‘timeROC’) were used to assess and compare the area under the curve (AUC) values when predicting OS and 1-, 3- and 5-year survival based on those three models in both the training and testing sets. Missing data were imputed using the multiple imputations by chain equations (MICE, maxit = 10 and m = 5) method ( Supplement Table 1 ). An SMD < 0.10 was indicative of a small intergroup difference and good equilibrium. Statistical significance was defined as a two-tailed P value of < .05. All statistical analyses were performed using R software (version 4.3.0). Results Patients’ characteristics This cohort study based on a prospectively maintained database enrolled and analyzed data from 552 surgically resected OSCC patients ( Table 1 ). The flow diagram illustrate the patient recruitment process, exclusion criteria, and final sample size allocation across the study groups ( Supplement Figure 1 ). The cohort demonstrated a median age of 61 years (interquartile range (IQR): 52-69 years) with male predominance (62.0%, 342/552) and a substantial proportion (45.7%, 252/552) of stage IV disease. Low proportions of patients had hypertension (26.8%) or diabetes (12.7%). In terms of treatment received at baseline, 165 (29.9%) patients had received chemotherapy and 282 (51.1%) patients had received radiotherapy. Table 1. Baseline characteristics in original cohort and matched cohort. Variables Original cohort Matched cohort Low-SLR High-SLR SMD Low-SLR High-SLR SMD n (%) n (%) n (%) n (%) All 445 (80.6) 107 (19.4) 214 (66.7) 107 (33.3) Age (years) 0.295 0.087 <60 218 (49.0) 37 (34.6) 83 (38.8) 37 (34.6) ≥60 227 (51.0) 70 (65.4) 131 (61.2) 70 (65.4) Gender 0.364 <0.001 Male 291 (65.4) 51 (47.7) 102 (47.7) 51 (47.7) Female 154 (34.6) 56 (52.3) 112 (52.3) 56 (52.3) Residence 0.016 0.010 City 278 (62.5) 66 (61.7) 133 (62.1) 66 (61.7) Rural 167 (37.5) 41 (38.3) 81 (37.9) 41 (38.3) Education 0.141 0.066 Junior high school and below 318 (71.5) 83 (77.6) 160 (74.8) 83 (77.6) High school and above 127 (28.5) 24 (22.4) 54 (25.2) 24 (22.4) BMI (kg/m 2 ) 0.536 0.090 18.5-23.9 265 (59.6) 82 (76.6) 171 (79.9) 82 (76.6) <18.5 41 (9.2) 14 (13.1) 26 (12.1) 14 (13.1) ≥ 24 139 (31.2) 11 (10.3) 17 (7.9) 11 (10.3) Tobacco use 0.251 <0.001 Never 228 (51.2) 68 (63.6) 136 (63.6) 68 (63.6) Ever 217 (48.8) 39 (36.4) 78 (36.4) 39 (36.4) Alcohol consumption 0.184 0.010 Never 282 (63.4) 77 (72.0) 153 (71.5) 77 (72.0) Ever 163 (36.6) 30 (28.0) 61 (28.5) 30 (28.0) Hypertension 0.210 0.035 No 318 (71.5) 86 (80.4) 169 (79.0) 86 (80.4) Yes 127 (28.5) 21 (19.6) 45 (21.0) 21 (19.6) Diabetes 0.252 0.088 No 382 (85.8) 100 (93.5) 195 (91.1) 100 (93.5) Yes 63 (14.2) 7 (6.5) 19 (8.9) 7 (6.5) Cancer site 0.023 0.028 Oral tongue 213 (47.9) 50 (46.7) 103 (48.1) 50 (46.7) Buccal mucosa and others 232 (52.1) 57 (53.3) 111 (51.9) 57 (53.3) Tumor differentiation 0.066 0.009 Well 235 (52.8) 53 (49.5) 107 (50.0) 53 (49.5) Poor and moderate 210 (47.2) 54 (50.5) 107 (50.0) 54 (50.5) TNM Stage 0.020 0.056 I/ II/ III 241 (54.2) 59 (55.1) 112 (52.3) 59 (55.1) IV 204 (45.8) 48 (44.9) 102 (47.7) 48 (44.9) Chemotherapy 0.125 0.030 No 317 (71.2) 70 (65.4) 143 (66.8) 70 (65.4) Yes 128 (28.8) 37 (34.6) 71 (33.2) 37 (34.6) Radiotherapy 0.078 0.056 No 221 (49.7) 49 (45.8) 104 (48.6) 49 (45.8) Yes 224 (50.3) 58 (54.2) 110 (51.4) 58 (54.2) Open in a new tab BMI, body mass index; SLR, aspartate aminotransferase/alanine aminotransferase (AST/ALT) ratio. The median follow-up time was 52.77 (IQR: 24.84-82.19) months, and the median SLR was 1.08 (IQR: 0.86-1.40). According to the optimal cutoff value of the pre-treatment SLR (SLR = 1.50, Supplement Figure 2 ), all patients were assigned to either the low-SLR group (SLR < 1.50, n = 445) or the high-SLR group (SLR ≥ 1.50, n = 107). After PSM, there were no intergroup differences for any of the covariates in the matched cohort (all SMD < 0.1; Table 1 ). Kaplan-Meier analysis for the SLR and survival outcomes The Kaplan-Meier analysis yielded 1-year, 3-year, and 5-year OS rates of 93.6%, 82.0%, and 78.4%, respectively ( Supplement Figure 3 ). OS curves were plotted for both SLR groups based on the Kaplan-Meier analysis ( Fig. 1 , Fig. 2 ). Briefly, a higher SLR was significantly associated with poorer OS in both the original and matched cohorts (all P < .05). Fig. 1. Open in a new tab Kaplan–Meier curves of OS according to the SLR in the original cohorts. Fig. 2. Open in a new tab Kaplan–Meier curves of OS according to the SLR in the PSM cohorts. Association between SLR and OS of OSCC patients All variables and full model met the assumption of equal proportional hazards (all P > .05, Supplement Table 2 ). The results of the univariate and multivariate Cox regression analysis in the original and matched cohorts are shown in the Table 2 and Figure 3 . The multivariate Cox regression revealed that patients in the high-SLR group exhibited significantly worse OS compared with that of the patients in the low-SLR group in both cohorts (original cohort: HR = 1.55, 95% CI: 1.02-2.36, P = .040; matched cohort: HR = 1.81, 95% CI: 1.13-2.91, P = .014). Furthermore, the C-index of the models that incorporated the SLR for the prediction of OS was modestly improved in both the original and matched cohorts ( P = .017 and P = .008, respectively; Table 3 ). The Area Under the Curve (AUC) for SLR was 0.558, which was comparable to that of PLR (Platelet-to-Lymphocyte Ratio, AUC = 0.546) and NLR (Neutrophil-to-Lymphocyte Ratio, AUC = 0.614), the results were showed in Supplement Figure 4 . Table 2. Cox regression analyses of OS of operated patients with OSCC in original cohort and matched cohort. Variables Original cohort Matched cohort Univariate cox Multivariate cox Univariate cox Multivariate cox HR (95% CI) P HR (95% CI) P HR (95% CI) P HR (95% CI) P Age (years) <60 Ref. Ref. Ref. Ref. ≥60 1.57 (1.07-2.30) .022 1.64 (1.11-2.44) .013 1.78 (1.04-3.03) .034 1.94 (1.13-3.35) .016 Gender Male Ref. Ref. Female 1.02 (0.70-1.48) .953 0.86 (0.54-1.36) .514 Residence City Ref. Ref. Rural 0.95 (0.65-1.39) .796 0.78 (0.48-1.27) .321 Education Junior high school and below Ref. Ref. High school and above 0.83 (0.54-1.27) .385 0.66 (0.36-1.21) .178 BMI (kg/m 2 ) 18.5-23.9 Ref. Ref. Ref. Ref. <18.5 2.75 (1.73-4.39) <.001 2.98 (1.84-4.82) <.001 2.73 (1.57-4.75) <.001 2.94 (1.67-5.20) <.001 ≥ 24 0.72 (0.45-1.17) .183 0.72 (0.44-1.18) .192 0.99 (0.42-2.31) .981 0.72 (0.30-1.71) .455 Tobacco use Never Ref. Ref. Ever 1.04 (0.72-1.51) .820 1.16 (0.72-1.87) .535 Alcohol consumption Never Ref. Ref. Ever 1.16 (0.79-1.69) .449 1.08 (0.65-1.80) .753 Hypertension No Ref. Ref. Yes 1.11 (0.74-1.66) .614 1.24 (0.72-2.14) .437 Diabetes No Ref. Ref. Ref. Ref. Yes 1.66 (1.03-2.67) .036 1.59 (0.98-2.56) .060 1.60 (0.80-3.22) .187 1.33 (0.65-2.71) .433 Cancer site Oral tongue Ref. Ref. Buccal mucosa and others 1.10 (0.76-1.59) .604 1.17 (0.74-1.87) .504 Tumor differentiation Well Ref. Ref. Ref. Ref. Poor and moderate 1.31 (0.91-1.89) .150 1.41 (0.96-2.06) .077 1.35 (0.85-2.15) .209 1.52 (0.94-2.45) .088 TNM stage I/ II/ III Ref. Ref. Ref. Ref. IV 2.84 (1.93-4.20) <.001 2.91 (1.97-4.31) <.001 2.22 (1.38-3.58) .001 2.27 (1.40-3.68) <.001 Chemotherapy No Ref. Ref. Yes 1.27 (0.86-1.87) .228 1.27 (0.79-2.04) .333 Radiotherapy No Ref. Ref. Yes 1.15 (0.80-1.67) .445 0.86 (0.54-1.37) .534 SLR <1.50 Ref. Ref. Ref. Ref. ≥1.50 1.79 (1.19-2.69) .005 1.55 (1.02-2.36) .040 1.80 (1.13-2.87) .014 1.81 (1.13-2.91) .014 Open in a new tab BMI, body mass index; SLR, aspartate aminotransferase/alanine aminotransferase (AST/ALT) ratio. Fig. 3. Open in a new tab A forest plot is shown of the hazard ratios for OS in both cohorts. In the multivariate model, adjustments were made for the following variables: age, BMI, diabetes, tumor differentiation, and TNM stage. Table 3. The C-Index of models in original cohort and matched cohort. Variables Original cohort Matched cohort C-index (95% CI) P C-index (95% CI) P Model 1 0.535 (0.504-0.566) 0.549 (0.508-0.590) Model 2 0.553 (0.522-0.584) 0.569 (0.530-0.608) Model 2 vs Model 1 .017 .008 Open in a new tab SLR, aspartate aminotransferase/alanine aminotransferase (AST/ALT) ratio. Model 1: included age, BMI, diabetes status, differentiation grade, and TNM stage. Model 2: incorporated SLR in addition to all variables in Model 1. In this study, depth of invasion (DOI) was documented for only 271 patients. Within this subgroup, DOI was included as an additional covariate in the multivariate Cox regression analysis. When adjusted for a very strong factor like DOI, the independent predictive power of SLR diminishes and becomes statistically marginal in this smaller subset (HR = 2.03, 95% CI: 0.96-4.33, P = .065; Supplement Table 3 ). Additionally, only 317 pathology reports contained perineural invasion, lymphatic invasion, and vascular invasion in our study. The sensitivity analysis revealed no statistically significant association was observed between SLR and OS in this cohort of 317 OSCC patients (HR= 1.36, 95% CI: 0.72-2.55, Supplement Table 3 ), which may be attributable to the limited sample size. Assessment of the SLR based on ML in the original cohort No statistically significant differences were observed in the distribution of the variables and OS between the training and testing sets ( P > .05, as showed in Supplement Table 4 and Supplement Figure 5 ). In the GBM procedure, each predictor was randomly arranged, and correlation reductions in the predictive performance of the model were calculated to obtain a ranking graph of the RINF of the predictors ( Supplement Figure 6 , Supplement Table 5 ). Among the factors evaluated, the pre-treatment SLR ranked fifth (RINF = 6.05). For the RSF-based model, the log-rank splitting rule was used to evaluate the RIMP of each predictor ( Supplement Figure 7 , Supplement Table 6 ). Among the factors assessed, the pre-treatment SLR ranked sixth (RIMP = 0.0587). The performance of GBM, RSF and CPH in predicting survival in original cohort The ROC curves were used to compare the performance of the GBM, RSF, and CPH models ( Supplement Table 6 ). The AUC of the GBM model was significantly higher than those of the RSF and CPH models (AUC = 0.804, 0.724, and 0.761, separately), which further demonstrated the reliability of the results of GBM analysis of the importance of SLR. The time-dependent ROC curves were also compared between models ( Supplement Table 7 , Supplement Figure 8 ), compared with CPH model, GBM model had higher accuracy in predicting 1-, 3-, and 5- year OS of postoperative patients with OSCC (all P < .05). These results further support the reliability of the analysis demonstrating the importance of the SLR in predicting survival outcomes based on the GBM. Discussions To date, few studies have explored the effect of the SLR on survival in patients with OSCC, especially in China. Therefore, this study evaluated survival outcomes among 552 patients with OSCC who underwent surgery in southeastern China. The results demonstrated that the pre-treatment SLR can serve as an independent prognostic predictor of OS in patients with OSCC. Although recent advances have been made in identifying factors capable of predicting cancer prognosis in terms of genetic, epigenetic, and common molecular alterations, there are still no standardized, inexpensive, and time-saving tests available that can be applied in daily clinical practice. Blood-based parameters such as AST and ALT levels are routinely quantified to assess liver function, and samples are relatively easy to collect. The results of the present study confirmed that a higher SLR was significantly associated with poorer OS in patients with OSCC who underwent surgery, which is consistent with the findings of a retrospective study involving patients with head and neck cancers 15 and those of a European cohort study that investigated outcomes in oral and oropharyngeal cancer cases. 16 Recent studies have suggested that the SLR is significantly associated with the prognosis of other cancers; for example, Bezan et al 11 reported that the preoperative SLR was a predictor of poor prognosis in a cohort of patients with non-metastatic renal cell carcinoma. A study conducted by Nishikawa et al 12 also suggested that the SLR was an important predictor of recurrence-free survival in patients with upper urothelial carcinoma after nephrectomy. In addition, a study involving Japanese patients with prostate cancer reported that OS was significantly worse in the high-SLR group than in the low-SLR group. 14 The precise pathophysiological mechanism underlying the relationship between the SLR and survival in patients with OSCC remains unclear; however, one study confirmed that these two aminotransaminases (AST and ALT) play an important role in the regulation of cellular metabolism and cancer cell renewal. 10 In terms of individual functions, ALT participates in the ‘glucose–alanine cycle’ and significantly contributes to the regeneration of glucose consumed by muscle tissues, and AST is involved in the relocation of nicotinamide adenine dinucleotide (NADH) within mitochondria, which is essential for aerobic glycolysis. These two aminotransferase reactions are particularly important in muscles, hepatocytes, and other cells with high metabolic activity. 22 Based on these mechanisms, high tumor progression activity may increase tumor metabolism, thereby leading to elevated AST/ALT ratios. This results of this study demonstrated that incorporating the SLR into models significantly improved their predictive performance; this was consistent with the results reported by Li et al, 19 who showed that the C-index of a nomogram based on the SLR, age, and clinical stage of the patient was significantly higher than that of a nomogram based on the clinical stage alone. Furthermore, the present study showed that the SLR ranked fifth and sixth in terms of the RINF and RIMP in the GBM and RSF models, respectively, providing further evidence that the SLR has a significant impact on the prognosis of patients with OSCC who have undergone surgery. ML algorithms have been shown to exhibit satisfactory to excellent accuracy in predicting oral cancer treatment responses, prognosis, lymph node metastasis, and malignant transformation. 23 However, to our knowledge, no previous study has investigated the accuracy of the SLR in predicting postoperative OS in patients with OSCC using ML. Therefore, in this study, two ML methods, GBM and RSF, were used to predict the prognosis of patients with OSCC; the results demonstrated that the GBM outperformed the CPH model, with significantly higher predictive performance. Some studies have also found that ML algorithms may significantly improve the accuracy of models for predicting the prognosis of patients with OSCC compared to that of CPH models. For example, Tseng et al 24 found that the elastic net penalized Cox proportional hazards model was more accurate in stratifying the survival risk of patients with advanced oral cancer. And Kim et al 25 found that convolutional neural network (CNN)-based histological image analyses can accelerate the discovery of better risk models for OC progression. This study has several strengths. First, multiple imputations were used to scientifically compensate for missing data, which reduced the selection bias. Second, PSM was conducted to ensure there were no intergroup differences in baseline characteristics. Finally, the influence of the SLR in predicting the prognosis of patients with OSCC was evaluated for the first time using ML-based approaches, including the GBM and RSF models; this improved the accuracy of the prognostic prediction modeling. Regarding the translational value of this study, SLR is derived from inexpensive, readily available routine parameters, ensuring favorable cost-effectiveness and accessibility. Furthermore, it significantly enhances the accuracy of prognostic assessment in OSCC patients, thereby guiding more personalized and precise clinical decision-making while ultimately improving survival outcomes. However, this study also had some limitations. First, the study was conducted at a single center, its external validity can be compromised. Second, the SLR can change over time and was not monitored dynamically. Further research is required to validate these findings. Third, PSM also has recognized limitation: (1) the inability to control for unmeasured confounding; (2) PSM-induced exclusion of unmatched subjects poses potential limitations to the external validity of the results; (3) matching reduced the effective sample size, potentially limiting the precision of results estimates. Fourth, the predictive accuracy of our models remains limited, which restricts its direct clinical applicability. Additionally, the unavailability of DOI data in a substantial subset is another major limitation, as it may affect the robustness of establishing SLR’s independent prognostic value. Finally, given that machine learning reliability depends on thousands if not millions of point data, our restricted sample size ( n = 552) inherently curtails the predictive models’ stability and extrapolation capacity. Moreover, ML may be affected by the so-called ‘black box effect’: researchers and clinicians typically know the inputs and the results, but it is hard to understand what is going on inside of the algorithm. The relative absence of transparency continues to be a limitation and potentially hindering the confidence of results. Conclusion In conclusion, a high pre-treatment SLR is associated with poor OS in patients with OSCC undergoing surgery. Incorporating the pre-treatment SLR into predictive models improves their prognostic accuracy, which can be conducive in guiding clinical decision-making when generating personalized and scientific treatment plans to improve survival rates. Informed consent statement Informed consent was obtained from all subjects involved in the study. Author contributions Xiaodan Bao: Writing–original draft; investigation; methodology. Baochang He: Conceptualization; data curation; project administration. Liling Shen: Writing–original draft; investigation. Yulan Lin: Data curation; formal analysis. Fa Chen: Data curation; formal analysis. Fengqiong Liu: Data curation; formal analysis; project administration; validation. Yu Qiu: Supervision; validation; writing–review and editing. Bin Shi: Supervision; validation; writing–review and editing. Lisong Lin: Supervision; writing–review and editing. Jing Wang: Supervision; validation; writing–review and editing. Ethics statement All procedures involving human participants were performed in accordance with the ethical standards of the Helsinki Declaration and the ethical standards of the Institutional Review Board (IRB) of Fujian Medical University (2011053). Written informed consent was obtained from all participants. Conflict of interest The authors declare no competing interests. Acknowledgments Funding This study was supported by the Mutual Expert Research Project of the First Affiliated Hospital of Fujian Medical University (Grant/Award Number: YJRCHP- 2023HBC). Data availability The analysis dataset is available from the corresponding author on reasonable request. Footnotes Supplementary material associated with this article can be found in the online version at doi:10.1016/j.identj.2026.109480 . Appendix. Supplementary materials mmc1.docx (898.3KB, docx) References 1. Bray F., Laversanne M., Sung H., et al. Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin. 2024;74(3):229–263. doi: 10.3322/caac.21834. [ DOI ] [ PubMed ] [ Google Scholar ] 2. Graillon N., Iocca O., Carey R.M., et al. What has the National Cancer Database taught us about oral cavity squamous cell carcinoma? Int J Oral Maxillofac Surg. 2022;51(1):10–17. doi: 10.1016/j.ijom.2021.03.015. [ DOI ] [ PubMed ] [ Google Scholar ] 3. Bao X., Liu F., Lin J., et al. Nutritional assessment and prognosis of oral cancer patients: a large-scale prospective study. BMC Cancer. 2020;20(1):146. doi: 10.1186/s12885-020-6604-2. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 4. Wang H.C., Chan L.P., Wu C.C., et al. Silencing DNA polymerase beta induces aneuploidy as a biomarker of poor prognosis in oral squamous cell cancer. Int J Mol Sci. 2021;22(5) doi: 10.3390/ijms22052402. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 5. Wang Y.T., Fang K.H., Hsu C.M., et al. Retrospective study on the potential of albumin/globulin ratio as a prognostic biomarker for oral cavity cancer patients. Eur Arch Otorhinolaryngol. 2021;278(1):227–238. doi: 10.1007/s00405-020-06145-x. [ DOI ] [ PubMed ] [ Google Scholar ] 6. Chen Q., Fan Y., Li Y., et al. A novel nutritional risk score and prognosis of oral cancer patients: a prospective study. Oral Dis. 2022;28(1):108–115. doi: 10.1111/odi.13733. [ DOI ] [ PubMed ] [ Google Scholar ] 7. Zhu Z., Zhou J., Xiong X., et al. Prognostic value of serum liver enzymes in oral and oropharynx squamous cell carcinomas. J Oral Pathol Med. 2019;48(1):36–42. doi: 10.1111/jop.12803. [ DOI ] [ PubMed ] [ Google Scholar ] 8. Wang N., Lin Y., Chen F., et al. Utility of gamma-glutamyl transpeptidase to lymphocyte count ratio in predicting prognosis of patients with oral cancer: a prospective cohort study in Southeastern China. Head Neck. 2023;45(5):1172–1183. doi: 10.1002/hed.27331. [ DOI ] [ PubMed ] [ Google Scholar ] 9. De Ritis F., Coltorti M., Giusti G. An enzymic test for the diagnosis of viral hepatitis; the transaminase serum activities. Clin Chim Acta. 1957;2(1):70–74. doi: 10.1016/0009-8981(57)90027-x. [ DOI ] [ PubMed ] [ Google Scholar ] 10. Koppenol W.H., Bounds P.L., Dang CV. Otto Warburg’s contributions to current concepts of cancer metabolism. Nat Rev Cancer. 2011;11(5):325–337. doi: 10.1038/nrc3038. [ DOI ] [ PubMed ] [ Google Scholar ] 11. Bezan A., Mrsic E., Krieger D., et al. The preoperative AST/ALT (De Ritis) ratio represents a poor prognostic factor in a cohort of patients with nonmetastatic renal cell carcinoma. J Urol. 2015;194(1):30–35. doi: 10.1016/j.juro.2015.01.083. [ DOI ] [ PubMed ] [ Google Scholar ] 12. Nishikawa M., Miyake H., Fujisawa M. De Ritis (aspartate transaminase/alanine transaminase) ratio as a significant predictor of recurrence-free survival in patients with upper urinary tract urothelial carcinoma following nephroureterectomy. Urol Oncol. 2016;34(9) doi: 10.1016/j.urolonc.2016.04.001. 417 e419-417 e415. [ DOI ] [ PubMed ] [ Google Scholar ] 13. Liu C., Jia B.S., Zou B.W., et al. Neutrophil-to-lymphocyte and aspartate-to-alanine aminotransferase ratios predict hepatocellular carcinoma prognosis after transarterial embolization. Medicine (Baltimore) 2017;96(45):e8512. doi: 10.1097/MD.0000000000008512. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 14. Miyake H., Matsushita Y., Watanabe H., et al. Significance of De Ritis (Aspartate transaminase/Alanine transaminase) ratio as a significant prognostic but not predictive biomarker in Japanese patients with metastatic castration-resistant prostate cancer treated with cabazitaxel. Anticancer Res. 2018;38(7):4179–4185. doi: 10.21873/anticanres.12711. [ DOI ] [ PubMed ] [ Google Scholar ] 15. Takenaka Y., Takemoto N., Yasui T., et al. Transaminase activity predicts survival in patients with head and neck cancer. PLoS One. 2016;11(10) doi: 10.1371/journal.pone.0164057. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 16. Knittelfelder O., Delago D., Jakse G., et al. The AST/ALT (De Ritis) ratio predicts survival in patients with oral and oropharyngeal cancer. Diagnostics (Basel) 2020;10(11) doi: 10.3390/diagnostics10110973. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 17. Lee H., Choi Y.H., Sung H.H., et al. De Ritis ratio (AST/ALT) as a significant prognostic factor in patients with upper tract urothelial cancer treated with surgery. Clin Genitourin Cancer. 2017;15(3):e379–e385. doi: 10.1016/j.clgc.2016.08.023. [ DOI ] [ PubMed ] [ Google Scholar ] 18. Ghahari M., Salari A., Ghafoori Yazdi M., et al. Association between preoperative De Ritis (AST/ALT) ratio and oncological outcomes following radical cystectomy in patients with urothelial bladder cancer. Clin Genitourin Cancer. 2022;20(2):e89–e93. doi: 10.1016/j.clgc.2021.10.007. [ DOI ] [ PubMed ] [ Google Scholar ] 19. Li L., Zeng Q., Xue N., et al. A nomogram based on aspartate aminotransferase/alanine aminotransferase (AST/ALT) ratio to predict prognosis after surgery in gastric cancer patients. Cancer Control. 2020;27(1) doi: 10.1177/1073274820954458. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 20. Wang N., Lin Y., Shen L., et al. Prognostic value of pretreatment lymphocyte percentage in oral cancer: a prospective cohort study. Oral Dis. 2024;30(4):2176–2187. doi: 10.1111/odi.14658. [ DOI ] [ PubMed ] [ Google Scholar ] 21. Shen L., Ji Y., Chen F., et al. An excessive weight loss percentage over the two years before treatment is an independent prognostic factor for operated patients with advanced oral squamous cell carcinoma. Int J Oral Maxillofac Surg. 2025;54(3):208–216. doi: 10.1016/j.ijom.2024.08.042. [ DOI ] [ PubMed ] [ Google Scholar ] 22. Botros M., Sikaris KA. The de Ritis ratio: the test of time. Clin Biochem Rev. 2013;34(3):117–130. [ PMC free article ] [ PubMed ] [ Google Scholar ] 23. Adeoye J., Tan J.Y., Choi S.W., et al. Prediction models applying machine learning to oral cavity cancer outcomes: a systematic review. Int J Med Inform. 2021 doi: 10.1016/j.ijmedinf.2021.104557. [ DOI ] [ PubMed ] [ Google Scholar ] 24. Tseng Y.J., Wang H.Y., Lin T.W., et al. Development of a machine learning model for survival risk stratification of patients with advanced oral cancer. JAMA Netw Open. 2020;3(8) doi: 10.1001/jamanetworkopen.2020.11768. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 25. Kim D.W., Lee S., Kwon S., et al. Deep learning-based survival prediction of oral cancer patients. Sci Rep. 2019;9(1):6994. doi: 10.1038/s41598-019-43372-7. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Associated Data This section collects any data citations, data availability statements, or supplementary materials included in this article. Supplementary Materials mmc1.docx (898.3KB, docx) Data Availability Statement The analysis dataset is available from the corresponding author on reasonable request. Articles from International Dental Journal are provided here courtesy of Elsevier ACTIONS View on publisher site PDF (794.8 KB) Cite Collections Permalink PERMALINK Copy RESOURCES Similar articles Cited by other articles Links to NCBI Databases Cite Copy Download .nbib .nbib Format: AMA APA MLA NLM Add to Collections Create a new collection Add to an existing collection Name your collection * Choose a collection Unable to load your collection due to an error Please try again Add Cancel Follow NCBI NCBI on X (formerly known as Twitter) NCBI on Facebook NCBI on LinkedIn NCBI on GitHub NCBI RSS feed Connect with NLM NLM on X (formerly known as Twitter) NLM on Facebook NLM on YouTube National Library of Medicine 8600 Rockville Pike Bethesda, MD 20894 Web Policies FOIA HHS Vulnerability Disclosure Help Accessibility Careers NLM NIH HHS USA.gov Back to Top