Introduction

Surgery, including total laryngectomy, and concurrent chemoradiotherapy (CCRT) are standard treatment options for locally advanced laryngeal squamous cell carcinoma (LASCC) [1,2,3]. Cisplatin-based CCRT or induction chemotherapy followed by radiotherapy/CCRT is commonly used for laryngeal preservation; however, patients who are unsuitable for CCRT due to advanced age, renal dysfunction, or poor performance status often undergo total laryngectomy [3,4,5]. Postoperative CCRT is recommended for patients with adverse pathological features such as extranodal extension (ENE) or positive surgical margins [6,7,8]. Postoperative radiotherapy alone is indicated for patients with other adverse pathological features, including pT4 primary tumors, pT3 supraglottic tumors, pN2 or pN3 nodal disease, perineural invasion, vascular invasion, lymphatic invasion, and subglottic extension [9, 10].

Perioperative pembrolizumab, in addition to surgery and adjuvant therapy, has been recognized as the standard of care for stage III and IVA LASCC with a PD-L1 combined positive score (CPS) of ≥ 1 [11]. However, patients with stage III–IVA LASCC are a clinically heterogeneous population, and the extent to which different patient subgroups may benefit from perioperative treatment intensification remains uncertain. This issue is particularly relevant in real-world surgical practice because some patients undergo surgery precisely because advanced age, comorbidities, renal dysfunction, or other background factors limit the feasibility of definitive CCRT [12,13,14,15,16,17]. Such factors may also affect the delivery of postoperative therapy. Therefore, understanding outcomes after surgery under real-world conditions and identifying clinical and pathological factors associated with recurrence and mortality may help characterize patient subgroups with greater unmet needs and inform future studies of perioperative treatment intensification.

The aim of this study was to clarify the oncological outcomes of patients with stage III and IVA LASCC who underwent surgery as initial treatment in a real-world setting and to identify clinical and pathological factors associated with survival outcomes. These findings were intended to provide prognostic information that may help inform the selection of patients for future studies evaluating perioperative treatment intensification. To achieve this objective, data from a multicenter, retrospective, 10-year cohort involving all 13 university hospitals and cancer centers in the northern region of Japan, serving a population of approximately 14 million people, were used to evaluate the associations of pretreatment factors and postoperative pathological features with survival outcomes.

Patients and methods

Patients

This multicenter, retrospective study was conducted from January 2014 to December 2023 at 13 participating hospitals, including all university hospitals and cancer centers in northern Japan. A total of 667 patients were newly diagnosed with stage III or IV LASCC during this period. Of these, 264 patients who received non-surgical treatment were excluded. Of the remaining 403 patients who underwent curative surgery, 11 patients with stage IVB disease were excluded, and 11 patients with missing key clinical or outcome data were also excluded, resulting in a final cohort of 381 patients. The patient characteristics of those who underwent surgery (n = 381) and those who received chemoradiotherapy (n = 206) are compared in Supplemental Table 1.

Table 1 Patient and treatment characteristics (N = 381)
Full size table

Ethics

This study was conducted in accordance with the Declaration of Helsinki. The study protocol was approved by the institutional review board (2023-1-1093), which waived the requirement for informed consent because all data were anonymized prior to analysis.

Study design

Information on age, sex, primary tumor site, clinical and pathological T and N status, and stage based on the UICC 8th edition, as well as smoking and drinking habits, type of surgery (total laryngectomy or others), pathological features including ENE and positive surgical margins, adjuvant therapy, time to recurrence, site of recurrence, and prognosis, was obtained from medical records. Adjuvant therapy was determined according to institutional policies and tailored to each patient’s clinical condition. Disease-free survival (DFS) was defined as the time from initial surgery to cancer recurrence or death from any cause. Overall survival (OS) was defined as the time from initial surgery to death from any cause. Disease-specific survival (DSS) was defined as the time from initial surgery to death due to the primary disease. Treatment outcomes were analyzed according to clinical stage (Fig. 1). Subgroup analyses were conducted according to age (< 72 versus ≥ 72 years), primary tumor site, and N status in patients with stage III disease (Table 2, Fig. 2), and according to age (< 72 versus ≥ 72 years), primary tumor site, T status (T1-3 versus T4), and N status (N0-1 versus N2) in patients with stage IVA disease (Fig. 3, Supplemental Fig. 1). The age cutoff of 72 years was based on the median age of the study cohort. Sensitivity analyses were performed by treating age as a continuous variable in multivariate Cox regression models. Additional analyses were performed according to pathological recurrence risk stratification and adjuvant treatment modality (Fig. 3). The implementation rates of adjuvant therapy were also compared between patients aged < 72 and ≥ 72 years within each pathological recurrence risk group (Table 5). Pathological recurrence risk stratification was defined as follows: high risk, ENE or positive surgical margins; intermediate risk, pT4 disease, pT3 supraglottic tumors, pN2–3 disease, and subglottic extension; and low risk, absence of all adverse pathological features. Adjuvant treatment modality was categorized as CCRT, radiotherapy alone, or no adjuvant treatment.

Fig. 1
Fig. 1
Full size image

Kaplan–Meier survival curves for stage III and stage IVA disease. A Disease-free survival (DFS), B overall survival (OS), and C disease-specific survival (DSS). Patients with stage III disease show significantly better DFS and DSS than those with stage IVA disease, whereas OS does not differ significantly between the two groups. Three-year survival rates and log-rank P values are shown

Table 2 Univariate and multivariate Cox regression analyses for survival outcomes in patients with stage III disease
Full size table
Fig. 2
Fig. 2
Full size image

Kaplan–Meier survival curves according to age (< 72 years versus ≥72 years) in patients with stage III disease. A Disease-free survival (DFS), B overall survival (OS), and C disease-specific survival (DSS). Younger patients (< 72 years) show good survival outcomes. Three-year survival rates and log-rank P values are shown

Fig. 3
Fig. 3
Full size image

Kaplan–Meier survival curves according to pathological risk group and adjuvant treatment modality. A–C Disease-free survival (DFS), overall survival (OS), and disease-specific survival (DSS) according to pathological risk group. Patients were classified into three risk groups: high-risk (extranodal extension or positive surgical margins), intermediate-risk (pT4 disease, pT3 supraglottic tumor, pN2–3 disease, or subglottic extension), and low-risk (no adverse pathological features). Significant differences in DFS, OS, and DSS are observed among the three risk groups. D–F Survival outcomes according to adjuvant treatment modality in the high-risk group. Patients were categorized into three treatment groups: adjuvant concurrent chemoradiotherapy (CCRT), adjuvant radiotherapy (RT) alone, and no adjuvant therapy. Although no significant differences are observed among treatment modalities, patients treated with adjuvant CCRT tend to have better survival outcomes. Three-year survival rates and log-rank P values are shown

Statistical analysis

The Kaplan–Meier method was used to estimate DFS, OS, and DSS. Survival outcomes were compared using the log-rank test. When appropriate, post hoc pairwise comparisons of multiple groups were performed with Bonferroni adjustment. Univariate and multivariate Cox regression analyses were performed to identify risk factors for DFS, OS, and DSS. Variables for multivariate Cox regression models were selected based on clinical relevance and potential prognostic importance, rather than on univariate P values. For stage III analysis, age, primary tumor site, and N status were included as covariates, whereas T status was additionally included for stage IVA analysis. The proportional hazards assumption was assessed using Schoenfeld residuals, and no evidence of violation was observed in any of the models (p > 0.05). Restricted cubic spline analyses with four knots were performed to assess the linearity assumption for age in the multivariable Cox models, with age 72 years used as the reference value. There was no missing data for the analyzed variables. For comparisons of patient characteristics between the surgery and chemoradiotherapy groups, the Student’s t-test was used for age, the chi-square test or Fisher’s exact test was used for sex, primary tumor site, and clinical stage, and the Mann–Whitney U test was used for clinical T and N status. All statistical analyses were performed using R for Mac (version 3.5.0; R Foundation for Statistical Computing, Vienna, Austria).

Results

A total of 381 patients with stage III or IVA LASCC who underwent curative-intent surgery as initial treatment at the 13 participating hospitals between January 2014 and December 2023 were included in the analysis. The patient and treatment characteristics are summarized in Table 1. Of the 381 patients, 202 (53.0%) had glottic cancer, 147 (38.6%) had supraglottic cancer, and 32 (8.4%) had subglottic cancer. Stage III disease was observed in 190 patients (49.9%), whereas 191 patients (50.1%) had stage IVA disease. Compared with the patients who received chemoradiotherapy (n = 206), patients who underwent surgery were older and had a higher proportion of glottic cancer, more advanced clinical T status, and a higher proportion of clinical stage IVA disease (Supplemental Table 1).

The Kaplan–Meier curves for DFS, OS, and DSS according to clinical stage (stage III versus stage IVA) are shown in Fig. 1. The 3 year DFS, OS, and DSS rates were 64.2, 85.2, and 90.0% for stage III disease, and 49.5, 75.4, and 78.9% for stage IVA disease, respectively. Significant differences in DFS (p = 0.022) and DSS (p = 0.004) were observed between stage III and stage IVA disease, whereas no significant difference was observed in OS (p = 0.18).

In patients with stage III disease, univariate and multivariate Cox regression analyses showed that age ≥ 72 years was independently associated with poorer DFS, OS, and DSS. Sensitivity analyses treating age as a continuous variable showed that each 1 year increase in age was independently associated with poorer DFS, OS, and DSS (Table 2). Restricted cubic spline analyses showed no significant evidence of non-linearity for DFS (P = 0.603), OS (P = 0.304), or DSS (P = 0.678), supporting the linearity assumption for age in the multivariable Cox models for stage III disease (Supplemental Fig. 2). The Kaplan–Meier curves according to age are shown in Fig. 2. The 3 year DFS, OS, and DSS rates were 72.1, 89.8, and 94.7% for patients aged < 72 years and 57.8, 74.8, and 85.4% for patients aged ≥ 72 years, respectively.

In patients with stage IVA disease, univariate and multivariate Cox regression analyses showed no independent prognostic factors for DFS. Age ≥ 72 years was independently associated with poorer OS and DSS, T4 disease was independently associated with poorer DSS, and N2 disease was independently associated with poorer OS, and DSS. Sensitivity analyses treating age as a continuous variable showed that each 1-year increase in age was independently associated with poorer OS, but not DFS or DSS (Table 3). The Kaplan–Meier curves according to age, T status, and N status are shown in Supplemental Fig. 1.

Table 3 Univariate and multivariate Cox regression analyses for survival outcomes in patients with stage IVA disease
Full size table

Patients’ clinicopathological characteristics are summarized in Table 4. Eighty-one (21.3%) patients had adverse pathological features, such as ENE or positive surgical margins; 188 (49.3%) had other risk features, including pT4 disease, pT3 supraglottic tumor, pN2–3 disease, or subglottic extension; and 112 (29.4%) had no adverse features. Overall, 24 (6.3%) patients received adjuvant CCRT, 43 (11.3%) received adjuvant radiotherapy alone, and 314 (82.4%) received no adjuvant therapy. In the pathological high-risk group (ENE or positive surgical margins), 18 (22.2%) patients received adjuvant CCRT, 28 (34.6%) received radiotherapy alone, and 35 (43.2%) received no adjuvant therapy (Fig. 3D–F). The corresponding numbers in the pathological intermediate-risk group (pT4 disease, pT3 supraglottic tumor, pN2–3 disease, or subglottic extension) were 6 (3.1%), 15 (8.0%), and 167 (88.8%), respectively. No patients in the low-risk group received adjuvant therapy. Among patients with high-risk pathological features (ENE or positive surgical margins), the proportion receiving adjuvant CCRT was significantly lower in patients aged ≥ 72 years than in those aged < 72 years (10.5% vs 32.6%, p = 0.003). Among patients with other risk features, however, there was no significant difference in the proportion receiving adjuvant CCRT or radiotherapy alone between patients age ≥ 72 years and those aged < 72 years (8.6% vs 13.7%, p = 0.36; Table 5).

Table 4 Clinicopathological characteristics (N = 381)
Full size table
Table 5 Comparison of adjuvant therapy implementation rates by age, stratified by recurrence risk group
Full size table

During a median follow-up period of 40 months, 137 (36.0%) patients experienced recurrence at a median of 8 months after surgery. Recurrence patterns included local recurrence in 25 (6.6%) patients, nodal recurrence in 63 (16.5%), and distant metastasis in 63 (16.5%, Table 1).

The Kaplan–Meier curves according to recurrence risk stratification based on pathological factors are shown in Fig. 3. Significant differences in DFS, OS, and DSS were observed among the three risk groups (all p < 0.001). Patients with high-risk pathological features, such as ENE or positive surgical margins, had significantly worse DFS, OS, and DSS than those in the intermediate- (DFS, p < 0.001; OS, p = 0.010; DSS, p < 0.001) and low-risk groups (DFS, p < 0.001; OS, p = 0.005; DSS, p < 0.001). After stratification by recurrence risk, no significant differences in DFS, OS, or DSS were observed among adjuvant treatment modalities (DFS, p = 0.79; OS, p = 0.28; DSS, p = 0.63); therefore, no post hoc pairwise comparisons were performed. Kaplan–Meier curves showed numerically better DFS, OS, and DSS in high-risk patients treated with adjuvant CCRT than in those receiving adjuvant radiotherapy alone or no adjuvant therapy (Fig. 3D–F). In the intermediate-risk group, no clear survival benefit was observed with adjuvant CCRT; however, the number of patients receiving CCRT was limited (n = 6, data not shown).

Discussion

In this multicenter, retrospective study, oncological outcomes of patients with stage III or IVA LASCC who underwent surgery as the initial treatment with curative intent were evaluated. Patients with stage III disease had significantly better DFS and DSS than those with stage IVA disease. Notably, the 3 year DSS rate in patients with stage III disease reached 90.0%. These findings support the role of upfront surgery as a curative treatment option for selected patients with stage III LASCC. In contrast, the poorer outcomes in stage IVA disease suggest the need for more effective multimodal treatment strategies.

In the subgroup analysis of stage III disease, age was the only factor significantly associated with survival outcomes on Cox regression analysis. Patients aged <72 years had better survival outcomes than those aged ≥ 72 years, with a 3-year DSS rate of 94.7%. This association was also observed when age was analyzed as a continuous variable, with each 1-year increase in age independently associated with poorer DFS, OS, and DSS. In the stage IVA subgroup, no prognostic factor showed consistent associations across survival endpoints. When age was analyzed as a continuous variable, each 1-year increase in age was independently associated with poorer OS, but not DFS or DSS. Older age, T4 disease, and N2 disease were associated with worse DSS.

These findings provide prognostic information that may help characterize patients with different unmet clinical needs. Given the favorable outcomes observed in younger patients with stage III disease after surgery-based treatment, the absolute room for further improvement in survival may be relatively limited in this subgroup, although whether these patients derive less benefit from perioperative pembrolizumab cannot be determined from the present study. In contrast, the poorer outcomes observed in stage IVA disease suggest a greater unmet need for more effective multimodal treatment strategies. However, because no patients in this cohort received perioperative pembrolizumab and PD-L1 CPS was not evaluated, the present study cannot determine which patients are more likely to derive a benefit from perioperative pembrolizumab. Prospective studies incorporating predictive biomarkers and treatment-response data are needed to identify patients who may benefit most from perioperative immunotherapy.

Of the 381 patients, 81 (21.3%) had adverse pathological features such as ENE or positive surgical margins, and 188 (49.3%) had other risk features. Despite the relatively high proportion of patients with adverse pathological risk features, adjuvant therapy was administered infrequently, with only 24 (6.3%) receiving CCRT and 43 (11.3%) receiving radiotherapy alone, whereas 314 (82.4%) did not receive any adjuvant therapy. Although no statistically significant differences in survival outcomes were observed according to adjuvant treatment modality, Kaplan–Meier curves showed numerically better DFS, OS, and DSS among patients with high-risk pathological features who received adjuvant CCRT. However, these findings should be interpreted with caution because of the potential for confounding by indication. Patients selected for adjuvant CCRT may have differed from those receiving radiotherapy alone or no adjuvant therapy in terms of age, performance status, renal function, comorbidities, and other clinical factors. In addition, the small number of patients receiving CCRT precluded meaningful adjusted analyses. Therefore, the observed differences cannot be interpreted as evidence of a treatment benefit from adjuvant CCRT in this observational study. These results suggest that a substantial proportion of patients with adverse pathological features did not receive guideline-recommended adjuvant therapy in routine clinical practice. Notably, the proportion of patients receiving guideline-recommended adjuvant therapy was significantly lower among those age ≥ 72 years than among those aged < 72 years, in patients with high-risk pathological features, whereas no significant difference was observed in those with other risk features. This lower use of postoperative adjuvant therapy may have contributed to the poorer DFS, OS, and DSS observed in older patients, particularly those with high-risk pathological features. Although the reasons for treatment omission were unavailable in the present study, advanced age, impaired renal function, comorbidities, or previous radiotherapy may have contributed. This highlights an important unmet clinical need and supports further investigation of perioperative treatment strategies, including pembrolizumab-based approaches, particularly in patients at higher risk of poor outcomes or those who may be unable to receive standard postoperative therapy. However, the present study does not provide evidence that these patients would derive greater benefit from pembrolizumab.

The present study had several limitations. First, because this was a multicenter, retrospective study, treatment decisions were made at each institution according to the clinical conditions of individual patients. As a result, the reason for omitting adjuvant CCRT or adjuvant radiotherapy was not available for analysis. Second, this study did not assess several adverse pathological features, including perineural invasion, vascular invasion, and lymphatic invasion. Therefore, some patients classified as low-risk may have harbored intermediate-risk pathological features, potentially resulting in misclassification. Third, because this study covered a 10-year period, treatment for recurrent disease was not uniform throughout the study. Advances in systemic therapy, including the introduction of immunotherapy and molecular targeted therapies, may have affected post-recurrence management and consequently affected survival outcomes [18,19,20,21]. Fourth, patients who underwent surgery were significantly older than those who received definitive chemoradiotherapy, suggesting that the surgical cohort may have been subjected to treatment selection bias. The relatively advanced age of the surgical cohort may have influenced the use of adjuvant CCRT among patients with high-risk pathological features, and the potential influence of comorbidities and other factors associated with treatment selection cannot be excluded. In addition, comorbidity, performance status, renal function, and some detailed pathological factors were not available for inclusion in the multivariate models; therefore, residual confounding cannot be excluded when interpreting age as an independent prognostic factor.

Despite these limitations, this study provides real-world prognostic data that may inform risk stratification and future studies of perioperative treatment strategies for patients with stage III and IVA LASCC. The findings may help identify patient populations with greater unmet clinical needs; however, whether these subgroups derive greater benefit from pembrolizumab-based treatment cannot be determined from the present study. Prospective studies are needed to evaluate the predictive factors and treatment effects of perioperative immunotherapy in resectable LASCC.

Conclusion

In this multicenter, real-world study, younger patients with stage III LASCC achieved good outcomes after surgery-based treatment. In contrast, patients with stage IVA disease and adverse pathological features had worse survival outcomes. Despite the substantial proportion of patients with high-risk pathological features, postoperative adjuvant therapy was infrequently administered in routine clinical practice. These findings suggest that the unmet clinical need for treatment intensification may differ among patients with stage III and IVA LASCC and provide prognostic information for future studies of risk-adapted perioperative treatment strategies. However, the present study cannot determine which patients derive greater benefit from pembrolizumab-based perioperative therapy.