The clinical characteristics of the patients from the SEER database and our center are displayed in Tables 1 and 2, respectively. Eight thousand hundred and sixty patients in the SEER cohort and 277 patients in our center cohort were excluded for lack of complete information on tumor, node, metastasis (TNM) stage, pathology, and survival information. The flow chart shows the study’s detailed inclusion and exclusion process in Figure S1. A total of 2943 NPC patients aged between 20 and 80 years from the SEER database were found eligible to be included in this study; among them, 935 and 2008 patients were diagnosed with EONPC and LONPC, respectively. Asian or Pacific Islanders represented 52.1% of the EONPC group, while they represented only 37.2% of the LONPC group (p < 0.001). In the EONPC group, the median age was 42 years (IQR 36–46 years), while it was 61 years (IQR 55–67 years) in the LONPC group. While the SEER cohort had more men than women, the EONPC group had more female patients (Fig. 1a) (p < 0.001). Among the eligible 833 NPC patients in our cancer treatment center, 518 and 315 were < 50 and ≥ 50 years old, respectively. The median ages of the two groups were 41 (IQR 35–46 years) and 56 years (IQR 52–61 years), respectively. Most patients were males, as in the SEER cohort. The EONPC group had a higher female ratio than the LONPC group (Fig. 1b) (p < 0.001).

Distribution of age at diagnosis and gender among individuals with nasopharyngeal cancer. a Patient age and gender distribution in the SEER Database. b Patient age and gender distribution in our center database

The pathology subtypes were different in the cohorts; the EONPC group was more likely to have WHO types II and III than type I in the SEER cohort (p < 0.001, Table 1), while no difference was observed in our center cohort (p = 0.315, Table 2). Regarding the overall TNM stage, both cohorts had a higher rate of lymph node metastasis in terms of the N stage for EONPC. There were no significant differences in the T, M, and tumor stages between the two groups in the SEER cohort. Statistical differences in the M and tumor stages were found between the two groups in our center cohort. Regarding the treatment, more EONPC patients in both cohorts received chemotherapy (p < 0.001), while radiation therapy (p < 0.001) and induction (IC), synchronous (CC), and adjuvant chemotherapies (AC) (p < 0.001) were additionally administered to those in the SEER and our center cohort, respectively. All patients in our center cohort received IMRT, while only a fraction received targeted therapy. Moreover, EONPC patients presented with a larger primary tumor volume (p = 0.049), while smoking and alcohol consumption were not significantly different. We also found that there were more patients with high comorbidity scores (p < 0.001) and EBV-DNA positivity (p < 0.001) in EONPC patients.

There were 43.2% deaths throughout the study period, of which 694 were disease-related. The mean follow-up was 67.8 months. Kaplan–Meier OS and CSS survival curves are shown in Fig. 2a and b. Five-year survival rates for EONPC and LONPC were 43.8% and 31.9%, respectively (p = 0.002). There were significantly worse survival rates in the LONPC group. Compared to LONPC, EONPC showed better CSS. We conducted univariate and multivariate Cox analyses to evaluate the prognostic significance of clinicopathological factors for OS (Table 3). All factors contributed to OS in the univariate analysis; besides the primary site, those significant in our univariate analysis remained significant after multivariate analysis.

Kaplan–Meier survival curves between early-onset and late-onset nasopharyngeal cancer patients from the SEER database. a Overall survival (OS). b Cancer-specific survival (CSS)

During a median follow-up time of 75 months, 87 patients relapsed, 132 metastasized, and 175 (21%) died. Kaplan–Meier curves of OS and CSS for the two groups are shown in Fig. 3a and b. Patients with EONPC had significantly longer OS (81.8% vs. 75.2%) and CSS (74.9% vs. 55.8%) than those with LONPC (p = 0.002). There were no statistically significant differences between the two groups in PFS, LRFS, RRFS, and DMFS (p = 0.234) (Fig. 3c–f). According to univariate analysis, age, sex, pathology, TNM stage, primary tumor volume, Charlson Comorbidity Index, EBV-DNA and smoking were significantly associated with OS (Table 4). Furthermore, multivariate analysis indicated that age, TNM stage, and primary tumor volume, EBV-DNA and smoking remained significantly associated with OS.

Kaplan–Meier survival curves between early-onset and late-onset nasopharyngeal cancer patients from the our center database. a Overall survival (OS). b Cancer-specific survival (CSS). c Progression‐free survival (PFS). d Locoregional relapse free survival (LRFS). e Regional recurrence-free survival (RRFS). f Distant metastasis-free survival (DMFS)

Moreover, we also performed OS subgroup analysis to explore the relationship between age and clinical factors. The subgroup analysis is shown in Fig. 4. Patients were divided into four age-related subgroups (20–29 {the reference group, 30–39, 40–49, and ≥ 50 years). In patients with stages of T3–T4, N2–N3, M0, no smoking history and tumor volume ≧44.2 mL, those ≥ 50 years old showed significant negative prognostic indication for OS. The age group of 40–49 years was also a negative predictive factor for OS among patients with stage T3–T4 cancer and EBV-DNA status. There was no significant difference in OS between the younger reference and other subgroups.

Forest plots of the associations of age with overall survival by subgroup. Association of age with overall survival in subgroup analysis of T, N stage, M stage, smoking history, EBV-DNA status and primary tumor volume

We evaluated acute toxicities and major late adverse events in all treated patients. The details of acute toxicities results are shown in Table 5. Compared with patients with EONPC patients, LONPC patients had significantly increased grade 3–4 hematotoxicity in acute side effects, such as leukopenia (13.9% vs.7.4%, p = 0.006), anemia (7.3% vs. 3.9%, p = 0.006) and thrombocytopenia (4.3% vs. 0.6%, p < 0.001). However, with the exception of nausea, there was no significant difference in non-hematologic toxicity between the two groups. Among the late side effects, dry mouth and deafness or otitis are the most common, especially in the LONPC group which had a significantly higher rate of grade 2 dry mouth and grade 3–4 deafness or otitis side effects (Table 6).