Smoking Status and Recurrence in Oral Tongue Squamous Cell Carcinoma: Evidence for Convergent Clinical Behavior Authors: Jaafar A Hadi, Guojun Li, Zohra N Nizami, Jennifer Wang, Bridget Reeves, Ratna Veeramachaneni, Dianna Roberts, Lorena Isabel Gomez, Humam Kadara, Mena Mansour, Jeffrey N Myers Published: 2026 ABSTRACT ABSTRACTObjectivesOral tongue squamous cell carcinomas (SCCOT) arising in never‐smokers and smokers have similar genomic profiles, suggesting tobacco exposure may not be the dominant driver of tumor biology at this site. We investigated whether smoking is related to recurrence risk in SCCOT.MethodsWe analyzed disease‐free survival (DFS) in two SCCOT cohorts: a discovery cohort of 86 patients and a validation cohort of 410 patients, both treated and followed at MD Anderson Cancer Center with available smoking histories and demographic, clinical, and pathologic information. The initial validation cohort included 424 patients; however, 14 patients overlapping with the discovery cohort were excluded to preserve cohort independence. The risk of recurrence was assessed using univariate and multivariable Cox regression models adjusted for prognostic factors. Smoking was evaluated as ever‐smoker versus never‐smoker and as high versus low cumulative exposure (> 15 vs. ≤ 15 pack‐years).ResultsIn the discovery cohort, no statistically significant differences in DFS were observed by smoking status (log‐rank p = 0.227) or cumulative exposure (log‐rank p = 0.460). Additionally, neither smoking status (aHR: 0.96; 95% CI: 0.46–1.20) nor exposure (aHR: 0.99; 95% CI: 0.48–2.03) was associated with recurrence. These findings were consistent in the validation cohort.ConclusionThese findings suggest smoking history was not independently associated with recurrence risk in SCCOT. Larger prospective studies are required to validate these findings. Further studies integrating genomic and transcriptomic profiling are needed to define high‐risk SCCOT biology and identify biomarkers for personalized treatment.ImpactBy demonstrating smoking status does not influence recurrence risk, our findings provide clinical support for prior genomic evidence of similarity between SCCOT in smokers and never‐smokers, arguing against the use of smoking history to define prognostic subgroups.Level of Evidence2.Keywords: disease‐free survival, oral tongue squamous cell carcinoma, recurrence risk, smoking Introduction Oral tongue squamous cell carcinoma (SCCOT) is increasingly affecting patients who lack traditional carcinogenic exposures, especially younger adults who are never‐smokers or light smokers [1]. This epidemiologic pattern has raised the hypothesis that SCCOT in never‐smokers may be biologically distinct from SCCOT in smokers. However, genomic profiling studies have not supported this proposed biological divergence. For example, Pickering et al. demonstrated that SCCOT tumors arising in smokers and never‐smokers exhibited similar mutational burdens, driver‐gene alterations, and copy‐number landscapes [1]. Similarly, a previous integrative genomic analysis by the same group showed that oral squamous cell carcinomas including SCCOT do not segregate into tobacco‐driven versus non‐tobacco‐driven molecular subclasses [2]. Additional sequencing studies demonstrated that regardless of smoking exposure, SCCOT rarely harbors site‐specific mutations [3]. Furthermore, it has been demonstrated that early‐onset and familial SCCOT cases appear to develop through carcinogenic pathways unrelated to tobacco exposure [4, 5]. Despite these genomic findings, it is not known whether clinical outcomes reflect the molecular similarities between smokers and non‐smokers with SCCOT. This knowledge gap may reflect limitations of prior studies evaluating clinical outcomes by smoking status, including small cohort sizes, incomplete quantification of smoking exposure, inclusion of mixed oral cavity subsites, and lack of independent validation. A solid and rigorous evaluation of recurrence risk that incorporates both smoking status and smoking exposure examined in independent SCCOT cohorts has not been performed. We sought to determine whether smoking history influences recurrence risk in SCCOT and whether clinical outcomes mirror the reported genomic similarity between tobacco‐driven and non‐tobacco‐driven SCCOT [1, 2]. Methods Study Population Two independent cohorts of patients with SCCOT treated at The University of Texas MD Anderson Cancer Center were analyzed. The discovery cohort comprised 86 patients with primary SCCOT identified as a site‐specific subset of a larger prospective biospecimen‐collection protocol for previously untreated head and neck squamous cell carcinoma initiated at MD Anderson Cancer Center in 2008. SCCOT was defined as squamous cell carcinoma arising from the oral tongue (ICD‐O C02.x), excluding tumors involving the base of the tongue (C01), floor of mouth (C04), or other oral cavity subsites. Eligible patients were required to have histologically confirmed SCCOT and available longitudinal follow‐up for recurrence outcomes. Smoking and clinicopathologic variables were analyzed using all available data for each variable of interest. All patients in the discovery cohort were treatment‐naive at enrollment and had detailed clinicopathologic data, smoking history, and longitudinal follow‐up available for analysis. This study was conducted in accordance with the Declaration of Helsinki and was approved by the Institutional Review Board of The University of Texas MD Anderson Cancer Center under a prospective biospecimen‐collection and registry protocol (LAB08‐0848). Written informed consent was obtained from all living participants prior to enrollment, including consent for tissue acquisition, clinical data collection, and longitudinal follow‐up. The validation cohort initially included 424 patients with SCCOT treated at MD Anderson Cancer Center between 1995 and 2013 and available follow‐up for recurrence outcomes. Fourteen patients overlapping with the discovery cohort were identified and excluded to preserve cohort independence, leaving 410 patients in the final validation analysis. Patients were included in analyses when the required variables for that specific analysis were available. Because this study involved retrospective analysis of existing clinical data, patient allocation to exposure groups was not randomized. Smoking Exposure Definitions Smoking status was classified as never‐smoker or ever‐smoker. Never‐smokers were defined as individuals with no documented history of cigarette use, and ever‐smokers were defined as individuals with any history of cigarette exposure, regardless of duration or intensity. Cumulative smoking exposure was quantified using pack‐years, calculated as the average number of cigarette packs smoked per day multiplied by the number of years of smoking. Patients were categorized as having ≤ 15 or > 15 pack‐years of smoking. The threshold of 15 pack‐years is commonly used in SCCOT studies to distinguish minimal from clinically meaningful lifetime tobacco exposure [6, 7]. DFS was defined as time from initial treatment to recurrence or death. Outcome Definition Disease‐free survival (DFS) was defined as the time from initial treatment to the first documented recurrence or death from any cause, whichever occurred first, consistent with conventional DFS methodology. Because death was incorporated directly into the DFS endpoint rather than treated as a censoring event, a separate competing‐risk analysis was not performed. Patients without recurrence or death were censored at the date of last follow‐up. Statistical Analysis Kaplan–Meier curves and log‐rank tests were used to estimate DFS across smoking groups, and differences between groups were assessed using the log‐rank test. Cox proportional hazards regression models were used to evaluate univariate and multivariable associations between smoking variables and DFS. Multivariable models, included age, sex, race, TNM stage, and treatment modality, which are established prognostic factors in SCCOT. Given the bimodal epidemiology of SCCOT, characterized by a distinct early‐onset peak associated with familial and genetically predisposed cases and a later‐onset peak consistent with sporadic disease [5], age‐stratified analyses were conducted (≤ 50 vs. > 50 years) to evaluate whether age modified the relationships between smoking history and recurrence risk. In the discovery cohort, for patients aged ≤ 50 years, multivariable modeling included age alone for analysis of smoking status, whereas analyses of smoking exposure were adjusted for age and race due to limited event counts. For patients aged > 50 years, full multivariable models including all covariates were used. All statistical analyses were performed using IBM SPSS Statistics for Windows, Version 29.0.1 (IBM Corp., Armonk, NY, USA). Two‐sided p‐values < 0.05 were considered statistically significant. The proportional hazards assumption for Cox regression models was evaluated using graphical inspection of log–log survival plots, which demonstrated no major violations of proportionality. Formal time‐varying coefficient or Schoenfeld residual analyses were not performed. Sex as a Biological Variable Both male and female patients were included in the study cohorts. Sex was incorporated as a covariate in multivariable regression models to account for potential biological differences in SCCOT outcomes. Missing Data and Attrition The discovery cohort initially included 86 patients, and the validation cohort initially included 424 patients. Fourteen overlapping patients identified between the cohorts were excluded from the validation cohort to preserve cohort independence, leaving 410 patients in the final validation cohort. Patients were not universally excluded for missing covariate data. Instead, analyses were performed using all available data for each variable and outcome of interest. In the discovery cohort, cumulative smoking exposure (pack‐years) was unavailable for 4 patients, and alcohol history was unavailable for 2 patients. In the validation cohort, treatment information was unavailable for 21 patients, pack‐year smoking history was unavailable for 30 patients, and alcohol history was unavailable for 31 patients. Accordingly, sample sizes varied slightly across subgroup and multivariable analyses depending on variable availability. No additional attrition occurred after cohort assembly beyond exclusion of overlapping patients and patients lacking required data for a given analysis. Results Baseline Characteristics Baseline demographic and clinicopathologic characteristics of the discovery and validation cohorts are summarized in Table 1. The two cohorts differed substantially in several baseline characteristics: the discovery cohort was older than the validation cohort (mean age 65.7 vs. 55.3 years), had a higher proportion of advanced disease (Stage III–IV: 62.8% vs. 43.4%), received adjuvant therapy more frequently (surgery alone: 32.6% vs. 54.1%), and had a higher recurrence rate (54.7% vs. 21.5%). Median follow‐up was also longer in the discovery cohort (5.0 years; IQR: 1.2–7.7 years) than in the validation cohort (3.4 years; IQR: 1.5–5.6 years), which may have contributed to the higher observed recurrence rate in the discovery cohort. Smoking variables also differed modestly between cohorts, with the validation cohort including a higher proportion of ever‐smokers (50.0% vs. 60.7%) and a higher proportion of patients with > 15 pack‐years (24.4% vs. 42.4%). TABLE 1Baseline characteristics of the discovery and validation cohorts. a VariableDiscovery cohort (n = 86)Validation cohort (n = 410)Age, mean ± SD, years65.7 ± 12.655.3 ± 14.2SexMale52 (60.5)243 (59.3)Female34 (39.5)167 (40.7)RaceNon‐Hispanic White70 (81.4)351 (85.6)Other16 (18.6)59 (14.4)Smoking statusNever‐smoker43 (50.0)161 (39.3)Ever‐smoker43 (50.0)249 (60.7)Pack‐years b ≤ 1560 (69.8)205 (50.0)> 1521 (24.4)175 (42.4)Stage (TNM) c I–II31 (36.0)232 (56.6)III–IV54 (62.8)178 (43.4)TreatmentSurgery d 28 (32.6)222 (54.1)Surgery + RT36 (41.9)94 (22.9)Surgery + CRT21 (24.4)83 (17.8)AlcoholYes38 (44.2)225 (54.9)No46 (53.5)154 (37.6)RecurrenceYes47 (54.7)88 (21.5)No39 (45.3)322 (78.5)Median follow‐up (IQR), years5.0 (1.2–7.7)3.4 (1.5–5.6)Abbreviations: CRT, chemoradiation; IQR, interquartile range; RT, radiation therapy. a Values in the table are number of patients (percentage) unless indicated otherwise. b In the discovery cohort, pack‐year smoking history was unavailable for five patients. In the validation cohort, pack‐year smoking history was unavailable for 30 patients. c In the discovery cohort, TNM stage data were missing for one patient. d In the discovery cohort, treatment information was unavailable for one patient. In the validation cohort, treatment information was unavailable for 21 patients. Baseline characteristics of the discovery and validation cohorts. a Abbreviations: CRT, chemoradiation; IQR, interquartile range; RT, radiation therapy. Values in the table are number of patients (percentage) unless indicated otherwise. In the discovery cohort, pack‐year smoking history was unavailable for five patients. In the validation cohort, pack‐year smoking history was unavailable for 30 patients. In the discovery cohort, TNM stage data were missing for one patient. In the discovery cohort, treatment information was unavailable for one patient. In the validation cohort, treatment information was unavailable for 21 patients. As shown in Table 2, TNM stage and recurrence rates did not differ significantly by smoking status in either cohort. However, baseline differences were observed between the cohorts after smoking‐related stratification. In the discovery cohort, smoking status differed significantly by race (p = 0.011), and pack‐year category differed significantly by sex (p = 0.040). In the validation cohort, ever‐smokers were significantly older (p = 0.001), more frequently male (p = 0.001), consumed more alcohol (p = 0.001), and received adjuvant therapy more frequently (p = 0.033) compared with never‐smokers. Importantly, these differences (cohort‐level and smoking‐stratified) were accounted for in the multivariable models by adjusting for age, sex, race, treatment, smoking status, alcohol, and pack‐years. TABLE 2Distribution of patient characteristics by smoking status and pack‐year category in the discovery and validation cohorts. a , b VariableDiscovery cohort (n = 86)Validation cohort (n = 410)Smoking statusPack‐yearsSmoking statusPack‐yearsEver‐smokers/never‐smokers p ≤ 15/> 15 p Ever‐smokers/never‐smokersp≤ 15/> 15 p Age, mean ± SD, years66.9 ± 11.4/64.5 ± 13.70.3964.7 ± 13.3/68.9 ± 11.30.257.3 ± 13.0/52.3 ± 15.60.00152.8 ± 16.2/57.5 ± 11.00.001Sex0.270.040.0010.001Male29 (55.8)/23 (44.2)33 (66.0)/17 (34.0)171 (70.4)/72 (29.6)99 (40.7)/121 (49.8)Female14 (41.2)/20 (58.8)27 (87.1)/4 (12.9)78 (46.7)/89 (53.3)106 (63.5)/53 (31.7)Race0.0110.5030.1130.217Non‐Hispanic White30 (42.9)/40 (57.1)51 (76.1)/16 (23.9)219 (62.4)/132 (37.6)160 (48.1)/154 (43.9)Other13 (81.3)/3 (18.8)9 (64.3)/5 (35.7)30 (50.8)/29 (49.2)36 (61.0)/20 (33.9)Stage (TNM)0.3690.4360.4750.524I–II13 (41.9)/18 (58.1)24 (80.0)/6 (20.0)137 (59.1)/95 (40.9)120 (51.7)/93 (40.1)III–IV29 (53.7)/25 (46.3)36 (70.6)/15 (29.4)112 (62.9)/66 (37.1)85 (47.8)/81 (45.5)Treatment0.9610.7160.0330.013Surgery14 (50.0)/14 (50.0)21 (77.8)/6 (22.2)120 (54.1)/102 (45.9)126 (56.8)/78 (34.7)Surgery + RT19 (52.8)/17 (47.2)23 (67.6)/11 (32.4)66 (70.2)/28 (29.8)37 (39.4)/50 (53.2)Surgery + CRT10 (47.6)/11 (52.4)15 (78.9)/4 (21.1)48 (65.8)/25 (34.2)34 (46.6)/37 (50.7)Alcohol0.1240.1370.0010.001Yes23 (60.5)/15 (39.5)25 (65.8)/13 (34.2)175 (77.8)/50 (22.2)78 (34.7)/147 (65.3)No19 (41.3)/27 (58.7)34 (81)/8 (19)43 (27.9)/111 (72.1)127 (82.5)/27 (17.5)Recurrence0.1930.4480.5410.582Yes27 (57.4)/20 (42.6)30 (69.8)/13 (30.2)56 (63.6)/32 (36.4)43 (48.9)/36 (40.9)No16 (41.0)/23 (59.0)30 (78.9)/8 (21.1)193 (59.9)/129 (40.1)162 (50.3)/138 (42.9)Abbreviations: CRT, chemoradiation; RT, radiation therapy. a Values in table are number of patients (percentage) unless otherwise indicated. b Statistically significant differences were observed between groups for race (p = 0.011) in the discovery cohort, and for age (p = 0.001), sex (p = 0.001), and treatment (p = 0.039) in the validation cohort. These differences were accounted for through multivariable adjustment in Cox regression models. Distribution of patient characteristics by smoking status and pack‐year category in the discovery and validation cohorts. a , b Abbreviations: CRT, chemoradiation; RT, radiation therapy. Values in table are number of patients (percentage) unless otherwise indicated. Statistically significant differences were observed between groups for race (p = 0.011) in the discovery cohort, and for age (p = 0.001), sex (p = 0.001), and treatment (p = 0.039) in the validation cohort. These differences were accounted for through multivariable adjustment in Cox regression models. Univariate Analyses Kaplan–Meier curves for DFS by smoking status are shown in Figure 1, and corresponding numerical results are shown in Table 3. In the discovery cohort, Kaplan–Meier curves for never‐smokers and ever‐smokers were nearly superimposable (log‐rank p = 0.227; Figure 1A). A similar pattern of largely overlapping survival curves was observed in the validation cohort (log‐rank p = 0.475; Figure 1B). In both cohorts, the recurrence curves mirrored each other in terms of shape, timing, and slope, indicating that smokers did not experience either earlier or more frequent recurrences. Disease‐free survival (DFS) according to smoking status in the discovery and validation cohorts. Kaplan–Meier curves showing DFS among patients with oral tongue squamous cell carcinoma stratified by smoking status. (A) Discovery cohort (n = 86) comparing never‐smokers (n = 43) and ever‐smokers (n = 43). (B) Validation cohort (n = 424) comparing never‐smokers (n = 169) and ever‐smokers (n = 255). Differences between groups were assessed using the log‐rank test. TABLE 3Univariate and multivariable analysis of the effect of smoking status and pack‐year category on disease‐free survival in the discovery and validation cohorts.Discovery cohort (n = 86)Validation cohort (n = 410)Univariate HR (95% CI)Multivariable HR (95% CI) a Univariate HR (95% CI)Multivariable HR (95% CI) a Smoking statusNever‐smoker1111Ever‐smoker1.42 (0.80–2.54)0.96 (0.46–1.20) b 1.19 (0.77–1.84)1.11 (0.66–1.85)Pack‐years≤ 151111> 151.28 (0.67–2.46)0.99 (0.48–2.03)1.00 (0.64–1.56)0.99 (0.60–1.64) a The multivariate hazard ratio was adjusted for age, sex, race, stage, alcohol, and treatment. b Of 43 ever‐smokers, 42 were included in this analysis. Univariate and multivariable analysis of the effect of smoking status and pack‐year category on disease‐free survival in the discovery and validation cohorts. The multivariate hazard ratio was adjusted for age, sex, race, stage, alcohol, and treatment. Of 43 ever‐smokers, 42 were included in this analysis. Kaplan–Meier curves for DFS by cumulative smoking exposure are shown in Figure 2, and corresponding numerical results are shown in Table 3. In the discovery cohort, DFS did not differ between patients with ≤ 15 pack‐years and those with > 15 pack‐years (log‐rank p = 0.460; Figure 2A). This pattern persisted in the validation cohort (log‐rank p = 0.879; Figure 2B). Disease‐free survival (DFS) according to cumulative smoking exposure in the discovery and validation cohorts. Kaplan–Meier curves illustrating DFS stratified by cumulative smoking exposure measured in pack‐years. (A) Discovery cohort comparing ≤ 15 pack‐years (n = 60) versus > 15 pack‐years (n = 21). (B) Validation cohort comparing ≤ 15 pack‐years (n = 213) versus > 15 pack‐years (n = 177). Differences between groups were assessed using the log‐rank test. Multivariable Cox Regression Adjusted hazard ratios (HRs) for smoking status are provided in Table 3. In the discovery cohort, smoking status was not associated with recurrence after adjustment for age, sex, race, TNM stage, alcohol, and treatment (adjusted HR: 0.96; 95% CI: 0.46–1.20). Similarly, in the validation cohort, smoking status was not associated with recurrence risk (adjusted HR: 1.11; 95% CI: 0.66–1.85). Adjusted HRs for cumulative smoking exposure are also provided in Table 3. In the discovery cohort, cumulative smoking exposure was not associated with recurrence after adjustment for age, sex, race, TNM stage, alcohol, and treatment (adjusted HR: 0.99; 95% CI: 0.48–2.03). Similarly, in the validation cohort, cumulative smoking exposure was not associated with recurrence risk (adjusted HR: 0.99; 95% CI: 0.60–1.64). Molecular correlates were not available in both cohorts; therefore, the observed lack of association between smoking and recurrence should be interpreted in the context of prior genomic studies demonstrating similar driver mutation spectra between smokers and never‐smokers. Age‐Stratified Analyses DFS curves for patients ≤ 50 years are shown in Figure 3, and the corresponding HRs are presented in Table 4. In the discovery cohort, DFS differed between never‐smokers and ever‐smokers (log‐rank p = 0.017; Figure 3A), but the analysis was underpowered due to the very small number of events (6 never‐smokers and 5 ever‐smokers), resulting in wide confidence intervals on the HR (univariate HR: 10.13; 95% CI: 1.06–96.76; Table 4). Multivariable analysis for smoking status, adjusted only for age due to limited sample size, showed no significant association with recurrence (HR: 16.33; 95% CI: 0.42–642.06; Table 4). In the discovery cohort, DFS also did not differ significantly by pack‐year category (log‐rank p = 0.055; Figure 3C), and pack‐year category was not associated with recurrence in univariate or multivariate analysis (Table 4). Disease‐free survival (DFS) among patients ≤ 50 years of age stratified by smoking status and smoking exposure. Kaplan–Meier analyses evaluating DFS among younger patients (≤ 50 years). (A) Discovery cohort: DFS by smoking status (never‐smokers vs. ever‐smokers). (B) Validation cohort: DFS by smoking status. (C) Discovery cohort: DFS by cumulative smoking exposure (≤ 15 vs. > 15 pack‐years). (D) Validation cohort: DFS by cumulative smoking exposure. Differences between groups were assessed using the log‐rank test. TABLE 4Univariate and multivariable analysis of disease‐free survival by age group (≤ 50 years vs. > 50 years) in the Discovery and Validation Cohorts.Discovery cohort (n = 86)Validation cohort (n = 410)UnivariateMultivariate a UnivariateMultivariate a N HR (95% CI) N aHR (95% CI) N HR (95% CI) N aHR (95% CI)≤ 50 years b Smoking statusNever6161821821Ever510.13 (1.06–96.76)516.33 (0.42–642.06)711.02 (0.53–1.95)711.06 (0.51–2.19)Pack‐years≤ 15919110211021> 1525.63 (0.78–40.84)22.14 (0.19–23.73)460.73 (0.34–1.56)460.73 (0.31–1.72)> 50 yearsSmoking statusNever371371791701Ever381.09 (0.60–2.01)381.10 (0.60–2.03)1781.52 (0.80–2.90)1781.12 (0.52–2.41)Pack‐years≤ 1551151110311031> 15191.03 (0.51–2.08)191.07 (0.51–2.23)1281.38 (0.740–2.55)1281.03 (0.50–2.14) a The multivariable analysis was adjusted for age, sex, race, smoking status, smoking exposure, alcohol, and treatment. b In the discovery cohort (patients ≤ 50 years), multivariable models adjusted for age when evaluating smoking status and for age and race when evaluating smoking exposure. Univariate and multivariable analysis of disease‐free survival by age group (≤ 50 years vs. > 50 years) in the Discovery and Validation Cohorts. The multivariable analysis was adjusted for age, sex, race, smoking status, smoking exposure, alcohol, and treatment. In the discovery cohort (patients ≤ 50 years), multivariable models adjusted for age when evaluating smoking status and for age and race when evaluating smoking exposure. Among patients ≤ 50 years, in the validation cohort, where the sample size was larger, DFS did not differ by smoking status (log‐rank p = 0.901; Figure 3B) or pack‐years (log‐rank p = 0.444; Figure 3D). Similarly, neither smoking status nor pack‐year category showed a significant association with recurrence in univariate or multivariable Cox models (Table 4). Among patients > 50 years, in both the discovery and validation cohorts, DFS curves for smoking status showed near‐complete overlap, as did DFS curves for pack‐year categories (Figure 4). Lastly, neither smoking status nor pack‐year category showed a significant association with recurrence in univariate or multivariable Cox models (Table 4). Disease‐free survival (DFS) among patients > 50 years of age stratified by smoking status and smoking exposure. Kaplan–Meier curves evaluating DFS among patients older than 50 years. (A) Discovery cohort: DFS by smoking status. (B) Validation cohort: DFS by smoking status. (C) Discovery cohort: DFS by smoking exposure (≤ 15 vs. > 15 pack‐years). (D) Validation cohort: DFS by smoking exposure (≤ 15 vs. > 15 pack‐years). Differences between groups were assessed using the log‐rank test. Discussion In this study, we analyzed two independent SCCOT cohorts totaling 496 patients and found no association between smoking history and recurrence risk, regardless of whether smoking history was evaluated categorically (ever‐smoker vs. never‐smoker) or quantitatively (≤ 15 vs. > 15 pack‐years). Kaplan–Meier curves, univariate Cox models, and fully adjusted multivariable Cox models consistently demonstrated null effects of smoking on DFS. Moreover, age‐stratified analyses confirmed that neither smoking status nor cumulative smoking exposure altered recurrence patterns in younger or older patients. These consistent results across multiple analytic frameworks strongly support the conclusion that smoking history was not independently associated with recurrence risk in SCCOT. Our findings are consistent with prior genomic studies demonstrating that SCCOT tumors do not diverge molecularly between smokers and non‐smokers. Pickering et al. showed that SCCOT tumors arising in smokers and never‐smokers are genomically indistinguishable, sharing similar mutation burdens, driver‐gene alterations, and copy‐number landscapes [1, 2]. Additional sequencing studies likewise failed to identify distinct tobacco‐driven subclasses [3], demonstrated convergence of oncogenic alterations across age groups [4], and suggested similar tumorigenic pathways in early‐onset and familial SCCOT [5]. Thus, the absence of clinical divergence seen in our cohorts is compatible with the reported molecular convergence of SCCOT across smoking exposure groups. Also, the age‐stratified analyses provided overall consistent evidence. In the validation cohort, where statistical power was greater, neither smoking status nor cumulative exposure was associated with DFS in either younger or older adults. In the discovery cohort, a nominally significant difference in DFS was observed among patients ≤ 50 years (p = 0.017), with a point estimate suggesting higher hazard in ever‐smokers; however, this was based on only 11 events and was not replicated in the validation cohort. Taken together, these findings support the broader conclusion that smoking‐associated and non‐smoking‐associated SCCOT likely share similar clinical behavior, though the observation in younger patients in the discovery cohort warrants attention in larger future studies. Our findings also align with prior clinical reports noting no prognostic advantage for never‐smokers in SCCOT. Vered et al. demonstrated that recurrence patterns relate to histopathologic and local invasion variables rather than smoking history [8]. Similarly, several studies have shown that recurrence is linked to depth of invasion, perineural invasion, and tumor stage, not exposure variables [9, 10, 11]. These data further support our conclusions that recurrence determinants in SCCOT are likely not tied to tobacco. Although this study does not directly examine tumor microenvironmental biology, the existing literature provides a potential mechanistic framework for our finding that recurrence in SCCOT was not strongly associated with smoking exposure. Previous studies in head and neck squamous cell carcinoma have linked recurrence and progression to mechanisms that occur independently of tobacco‐related mutational burden, including cancer‐associated fibroblast‐mediated stromal remodeling, immunosuppressive microenvironmental dynamics, and tumor microenvironment‐driven invasive phenotype [12, 13, 14]. These findings raise the speculation that tumor‐extrinsic drivers, particularly tumor microenvironment biology, rather than tobacco‐associated genomic alterations alone, may contribute substantially to recurrence in SCCOT. This hypothesis warrants further investigation in studies integrating molecular and microenvironmental profiling with clinical outcome data. A major strength of this study is the use of two independent cohorts, including a large validation cohort, each with rigorous documentation of smoking history and long‐term follow‐up. The consistency of null results across all analytic approaches reduces the likelihood that our findings reflect statistical underpowering or misclassification. This study has several limitations. Although the multivariable models adjusted for key demographic and clinical factors, several established predictors of SCCOT recurrence were not accounted for, including depth of invasion, perineural and lymphovascular invasion, margin status, extranodal extension, detailed alcohol exposure patterns, and molecular or tumor microenvironment features, raising the possibility of residual confounding. The two cohorts also had baseline differences in age, stage distribution, treatment intensity, and crude recurrence rate, which may reflect differences in follow‐up (as discussed in the Results), referral patterns, or treatment era. Additionally, some subgroup analyses were limited to a small number of events. Moreover, although death was incorporated into the DFS endpoint, separate competing‐risk analyses were not performed and may provide additional insight in future studies, particularly in older patient populations. Lastly, the ≤ 15 pack‐year group included never‐smokers, and peri‐ or posttreatment smoking behavior was not available, limiting the assessment of dose–response relationships and the impact of smoking cessation or continued exposure after diagnosis. Nonetheless, the striking concordance across models and cohorts strongly supports the central conclusion. Conclusion Smoking history, whether measured categorically (ever vs. never) or quantitatively (≤ 15 vs. > 15 pack‐years), was not independently associated with recurrence risk in SCCOT in two independent cohorts. Our findings are consistent with prior genomic evidence of molecular convergence between tobacco‐driven versus non‐tobacco‐driven SCCOT, extending that evidence into clinical outcomes, though larger prospective studies are required to validate these findings. Further, these findings suggest that future prognostic studies for SCCOT might benefit from incorporating microenvironmental and molecular features alongside clinical covariates, rather than relying on smoking history for primary stratification. Funding The authors have nothing to report. Disclosure Generative AI tools were used for language editing and grammar refinement. All scientific content, statistical analyses, interpretation, and final manuscript decisions were reviewed and verified by the authors. Conflicts of Interest The authors declare no conflicts of interest. Acknowledgments The authors have nothing to report. Data Availability Statement The data generated and analyzed in this study are not publicly available due to patient privacy and institutional review board restrictions. De‐identified data may be made available from the corresponding author upon reasonable request and with appropriate institutional approvals. References Associated Data Data Availability Statement The data generated and analyzed in this study are not publicly available due to patient privacy and institutional review board restrictions. De‐identified data may be made available from the corresponding author upon reasonable request and with appropriate institutional approvals.