This investigation was designed as a retrospective and comparative cohort study of prospectively collected data. It was conducted following approval from the Institutional Review Boards of The First Affiliated Hospital of Zhengzhou University and in strict accordance with the ethical principles outlined in the Declaration of Helsinki. Written consent was obtained from every patient before enrollment.

The study cohort was identified by screening the electronic medical records of all consecutive adult patients (aged 18 years or older) with a histological confirmation of locally advanced OSCC staged as III-IVB according to the AJCC 8^th edition. To be eligible for screening, patients must have been diagnosed between January 1, 2018, and December 31, 2023, and their pre-treatment imaging (CT, MRI, or PET/CT) had to document that the primary tumor was involving or directly abutting the mandibular cortex.

Inclusion in the final analysis was contingent upon meeting all of the following criteria:

Patients were excluded from the study if any of the following applied:

Based on the definitive surgical procedure performed, patients were allocated into two distinct cohorts for comparative analysis. The Mandibular Preservation (MP) Cohort consisted of patients for whom intraoperative findings, following neoadjuvant therapy, permitted a deviation from the pre-operatively planned mandibulectomy; preservation was strictly defined as the absence of a segmental mandibulectomy and encompassed procedures such as periosteal stripping, marginal resection, or cortical shaving of the mandible. In contrast, the Mandibular Sacrificing (MS) Cohort comprised patients who ultimately underwent a segmental or a large composite marginal mandibulectomy as initially planned, despite the documented rCR; this involved the full-thickness removal of a mandibular segment, creating a discontinuity defect that necessitated complex reconstruction, typically with an osseous or osteocutaneous free flap. The final intraoperative decision to pursue a preservation or sacrificing approach was made by the same attending surgeon (Q.Y.), informed by a combination of factors including the degree of tumor detachment from the bone, the quality of the underlying mandible, and the results of frozen section analysis of the soft tissue margins adjacent to the mandible.

A comprehensive and standardized data extraction protocol was employed to collect detailed information from the electronic health records of all eligible patients. The collected variables were systematically categorized into several key domains. Baseline and demographic characteristics included age, sex, BMI, comorbidities summarized by the Charlson Comorbidity Index (CCI), smoking pack-year history, alcohol use, primary tumor site and subsite, and clinical TNM stage according to the AJCC 8th edition. Neoadjuvant and surgical treatment details encompassed the specific NAT regimen, the number of cycles administered, the occurrence and grade of any immune-related adverse events (irAEs), and specific surgical procedure details such as the type of mandibular resection, total operation time, estimated intraoperative blood loss, and the method of reconstruction. Furthermore, data were collected on the pathologic assessment, which involved the final surgical margin status, the pathologic tumor stage (ypTNM), and the degree of pathologic response in the primary tumor specimen as evaluated by the College of American Pathologists tumor regression grading system, ranging from Complete Response to Poor or No Response. The presence or absence of perineural invasion and lymphovascular invasion was also documented from the final pathology reports.

The study evaluated two primary outcome domains: surgical complication and QoL. Surgical complication was rigorously assessed by the incidence of major postoperative complications occurring within 90 days of the definitive procedure, with severity graded according to the Clavien-Dindo (CD) classification system; a major complication was prospectively defined as CD grade III or higher, indicating a requirement for surgical, endoscopic, or radiological intervention. Specific complications tracked included unplanned return to the operating room, free flap failure, the development of an orocutaneous or pharyngocutaneous fistula, wound dehiscence or major infection requiring surgical intervention, significant postoperative hemorrhage, unplanned intensive care unit (ICU) admission, and mortality. The second primary outcome, health-related quality of life, was measured using the validated European Organization for Research and Treatment of Cancer (EORTC) Quality of Life Questionnaires, specifically the core EORTC QLQ-C30 (version 3.0) and the head and neck-specific EORTC QLQ-H&N35 modules. The questionnaire was required to complete preoperatively, 3 months, 6 months, and 12 months after operation.

Secondary outcomes focused on functional recovery and short-term oncologic efficacy. Functional outcomes were evaluated by measuring the duration of the initial postoperative hospital stay and by determining patient dependency on supportive measures, specifically the presence of a gastrostomy tube and/or tracheostomy at the time of hospital discharge and at the 3-month follow-up visit. Furthermore, short-term oncologic efficacy was measured by calculating local recurrence-free survival (LRFS), defined as the time from surgery to the first recurrence at the primary site; regional recurrence-free survival (RRFS), defined as the time to recurrence in the cervical lymph nodes; and disease-free survival (DFS), defined as the time to any recurrence (local, regional, or distant) or death from any cause, with all three metrics being reported at 3 years postoperatively.

All statistical analyses were performed using IBM SPSS Statistics (version 28.0) and R software, with a two-sided p-value of < 0.05 deemed significant. Continuous variables were presented as mean ± standard deviation or median with interquartile range based on their distribution, while categorical variables were summarized as frequencies and percentages. Initial comparisons of baseline demographic, clinical, and treatment-related characteristics between the MP and MS cohorts were conducted using independent samples t-tests, Mann-Whitney U tests, or Chi-square/Fisher’s exact tests, as appropriate, to identify potential confounding factors for adjustment.

The primary outcomes were analyzed with rigorous methodologies. For surgical complication, the incidence of major complications (Clavien-Dindo grade ≥ III) was compared using the Chi-square test, followed by a multivariable logistic regression model to calculate adjusted odds ratios (aOR). For the longitudinal QoL data, a linear mixed-effects model was employed to analyze the EORTC QLQ-C30 and QLQ-H&N35 scores collected preoperatively and at 3, 6, and 12 months postoperatively. This model included fixed effects for group, time, and their interaction, with a random intercept for each patient, allowing for a robust comparison of QoL trajectories between cohorts over time while handling within-patient correlation and missing data.

Secondary outcomes were compared using Chi-square or Mann-Whitney U tests. LRFS, RRFS, and DFS were estimated using Kaplan-Meier curves and compared with the log-rank test. The association of cohort allocation with survival outcomes was further quantified using univariate and multivariable Cox proportional hazards models to calculate hazard ratios (HR), adjusting for pathologic stage and margin status. A post-hoc power analysis was conducted to inform the interpretability of the findings given the final cohort sizes.

In total, 78 patients were enrolled. The MP (n=42) and MS (n=36) cohorts were generally comparable in several key areas. No statistically significant differences were observed in demographic factors such as age, sex, and BMI, or in lifestyle factors including smoking history and alcohol use. Comorbidity burden and details of neoadjuvant treatment, including the number of cycles and the incidence of irAEs, were also similar between the two groups. All patients received the same neoadjuvant immunochemotherapy regimen (ICI + cisplatin-based chemotherapy). Accordingly, cortical bone erosion on pre-treatment imaging was more common in the MS cohort (33.3%) than in the MP cohort (16.7%), although this difference did not reach statistical significance (p=0.089). The MS cohort presented with more advanced disease, featuring significantly higher Clinical T Stage (p=0.008), with 50.0% having T4a tumors compared to 23.8% in the MP cohort, and a more advanced Overall Stage Group (p=0.003), including a greater proportion of Stage IVB disease (25.0% vs. 9.5%). Clinical N Stage was the only tumor characteristic that did not differ significantly between the cohorts. All patients achieved a complete pathologic response and a clear margin. (Supplementary Table 1).

The rate of major complications (Clavien-Dindo grade III or higher) was significantly greater in the MS group (19.4%, 7/36) compared to the MP group (2.4%, 1/42), with a p-value of 0.013. While the overall incidence of specific major complications such as unplanned return to the operating room, orocutaneous fistula, free flap failure, significant postoperative hemorrhage, wound dehiscence/major infection, and unplanned ICU admission was low, all individual instances of these events, except for one case of hemorrhage, occurred exclusively in the MS cohort. However, the differences in the rates of these specific complications did not reach statistical significance individually (Table 1).

Univariate and multivariable analyses were performed to identify factors associated with the occurrence of major complications. In the univariate analysis, the MS cohort was associated with a significantly higher odds of major complications compared to the MP cohort (OR 3.76, 95% CI 1.42-9.97, p=0.008). A clinical T stage of T4a (vs. T2-T3; OR 2.67, 95% CI 1.05-6.78, p=0.039) were also significant risk factors. Other demographic, lifestyle, tumor, and treatment factors, including age, comorbidities, smoking, and neoadjuvant treatment details, were not significantly associated. When these factors were adjusted for in the multivariable analysis, the MS cohort remained the sole independent and statistically significant predictor of major complications (Adjusted OR 3.85, 95% CI 1.42-10.45, p=0.008). The associations for tumor site and T stage were attenuated and were no longer statistically significant in the adjusted model, indicating that the increased risk they presented in univariate analysis may be confounded by their relationship with the surgical approach (Table 2).

In EORTC QLQ-C30 module (Figure 1; Supplementary Table 2), while baseline scores were comparable, the MP cohort demonstrated a superior and more rapid improvement across most domains, particularly in Global Quality of Life, Physical Functioning, and Role Functioning, with these differences becoming statistically significant (p < 0.05) by the 6- and 12-month follow-ups. Concurrently, the MP cohort reported significantly lower severity of key symptoms, including pain, fatigue, insomnia, and appetite loss, at these later time points. This pattern indicates that patients who underwent mandibular preservation experienced a substantially lower symptom burden and a faster return to normal daily activities and overall well-being compared to those who required segmental mandibulectomy. These findings were further reinforced by the head and neck-specific EORTC QLQ-H&N35 module (Figure 2; Supplementary Table 3). The MS cohort consistently reported significantly worse outcomes (p < 0.05) related to swallowing, speech, social eating, and social contact throughout the follow-up period. Notably, domains directly impacted by mandibular integrity, such as problems with teeth and opening the mouth, were also markedly more severe in the MS group. Furthermore, the greater reliance on supportive measures in the MS cohort was reflected in significantly higher scores for pain medication use, nutritional supplements, and feeding tube dependence at the 3-month mark, with these differences persisting, though narrowing, at one year.

Significant differences of postoperative functional outcomes were observed between the MP and MS cohorts. The median postoperative hospital stay was markedly shorter for the MP cohort (9 days, IQR 7-12) compared to the MS cohort (16 days, IQR 13-22), a difference that was highly statistically significant (p<0.001). At the time of hospital discharge, functional dependence was substantially more common in the MS group, with a significantly higher proportion of patients dependent on a gastrostomy tube (63.9% vs. 28.6%, p=0.002) and a tracheostomy (55.6% vs. 19.0%, p<0.001). While functional status improved in both groups by the 3-month follow-up, a significant disparity in gastrostomy tube dependence persisted, with 11.1% of MS patients remaining dependent compared to 0% in the MP group (p=0.037). The difference in tracheostomy dependence at 3 months was no longer statistically significant, with only one patient (5.6%) in the MS cohort still dependent. These findings indicate that while mandibular sacrifice may be a necessary oncologic procedure, it is associated with a significantly longer initial recovery period and a higher burden of functional impairment, particularly with regard to swallowing, which persists for a notable subset of patients at three months postoperatively (Table 3).

No significant detriment was associated with the mandibular preservation approach for LRFS (Figure 3). The univariate Cox regression revealed a non-significant HR of 0.51 (95% CI 0.05–4.98; p=0.562) for the MS cohort compared to the MP cohort, indicating no increased risk of local recurrence with preservation. This finding persisted after multivariable adjustment for potential confounders, with an adjusted HR of 0.48 (95% CI 0.05–4.82; p=0.534). This suggests that achieving a rCR at the primary site following neoadjuvant therapy may permit mandibular preservation without compromising local tumor control (Table 4).

For RRFS, the surgical cohort was not a significant predictor of outcome (Figure 3). The MS cohort showed a non-significant trend toward a higher hazard of regional recurrence in both univariate (HR 1.80, 95% CI 0.53–6.15; p=0.350) and multivariable analyses (Adjusted HR 1.92, 95% CI 0.55–6.76; p=0.305). No other factors, including clinical N stage, or tumor site were significantly associated with RRFS in the models. This implies that the choice of mandibular resection did not independently influence the risk of recurrence in the cervical lymph nodes. (Supplementary Table 4).

The composite endpoint of DFS was also comparable between the two cohorts (Figure 3). Interestingly, the point estimates for the MS cohort hazard ratio were below 1.0 in both univariate (HR 0.35, 95% CI 0.04–3.18) and multivariable models (Adjusted HR 0.33, 95% CI 0.03–3.12), though these results were highly non-significant (p=0.350 and p=0.332, respectively). Over a median follow-up of 3 years, this analysis provides preliminary evidence that an organ-preservation strategy, when applied to select patients with an excellent response to neoadjuvant therapy, can achieve short-to-medium-term disease control rates that are not inferior to those of traditional, more radical surgery. (Supplementary Table 5).

To address potential confounding from the uneven distribution of baseline T stage, subgroup analyses stratified by clinical T category (T2/3 vs. T4a) were performed. Within each stratum, multivariable Cox proportional hazards models were fitted with the same adjustment variables used in the primary cohort analysis. This ensured methodological consistency across all survival analyses. Within the T2/3 subgroup (n=50), there was no significant difference in LRFS survival between the MS and MP cohorts (aHR 0.42, 95% CI 0.04-4.85; p=0.489). Similarly, in the T4a subgroup (n=28), the MS cohort did not show an increased risk of LRFS compared to the MP cohort (aHR 0.55, 95% CI 0.0-6.45; p=0.632). The results were consistent for RRFS and DFS across both T stage strata (Figure 4).