From January 2010 to April 2020, a total of 504 patients with pathologically diagnosed HNACC were treated at our center. The following patients were excluded from this study: 155 patients who did not undergo radical surgery combined with postoperative IMRT, 82 patients with recurrence or metastasis, 56 patients with incomplete data or lost follow-up, 3 patients with concomitant other malignancies, and 2 patients who received docetaxel chemotherapy. Finally, 206 patients were included in the study, and all patients were restaged according to the 8th edition of the American Joint Committee on Cancer criteria. The entire cohort included 147 patients who received PORT (71.4%) and 59 patients who received POCRT (28.6%). Within the entire cohort, 15.5% (n = 32) of patients underwent neck dissection. All patients underwent IMRT to a median dose of 67.5 Gy (range 60–74 Gy). Patients undergoing PORT alone were treated to a median dose of 68 Gy (range 60–74 Gy), and those undergoing POCRT were treated to a median dose of 67 Gy (range 60–70 Gy). The skull base was treated in 64.6% (n = 133) of patients (PORT group = 89, 60.5%; POCRT group = 44, 74.6%), and the neck was irradiated in 62.6% (n = 129) of patients (PORT group = 95, 64.6%; POCRT group = 34, 57.6%). The POCRT group consisted of 30 patients (50.8%) who received cisplatin chemotherapy, 12 patients (20.3%) who received nedaplatin chemotherapy, 12 patients (20.3%) who received lobaplatin chemotherapy, and 5 patients (8.5%) who received oxaliplatin chemotherapy. Among the POCRT group, 69.5% (n = 41) underwent a tri-weekly regimen of chemotherapy, while the remaining 30.5% (n = 18) received weekly chemotherapy. The use of POCRT was associated with the primary location (P = 0.048*) and pathological T stage (P = 0.011*). Propensity score matching was used to match 51 pairs of patients who received PORT or POCRT, and patient characteristics were balanced across all covariates. The clinical characteristics of the two cohorts are summarized in Table 1.

After a median follow-up of 73.5 months (range, 15–227 months), 21 (10.2%) of the 206 patients experienced local–regional failure, 70 (34.0%) developed distant metastasis, and 47 (22.8%) died (42 from cancer, 5 from non-cancer-related diseases or accidents). The 3-, 5-, and 10-yr LRC for the cohort were 92.0%, 90.6%, and 86.9%, respectively. The 3-, 5-, and 10-yr DMFS were 76.1%, 68.5%, and 56.7%, respectively. The 3-, 5-, and 10-yr DFS were 73.2%, 65.0%, and 54.8%, respectively. The 3-, 5-, and 10-yr OS were 91.7%, 85.3%, and 67.0%, respectively.

The most frequent pattern of failure was distant metastasis, which occurred in a median time of 30 months (range, 4–116 months). The lung was the most commonly affected site of distant metastasis (80%), followed by bone (19%), liver (17%), and brain (11%). The median time to local–regional failure was 22 months (range, 1–98 months), with 90.9% of these failures occurring in high-dose areas. Table 2 provides a detailed summary of the characteristics of patients who experienced local–regional failure.

Within the entire cohort, a comparison between the POCRT group and patients undergoing PORT indicated superior locoregional control (LRC) after accounting for competing risk events (P = 0.048*, Gray’s test, Fig. 1A). In the POCRT group, the 3-year, 5-year, and 10-year locoregional failure cumulative incidence rates were 1.7%, 1.7%, and 4.3% respectively, while in the PORT group, they were 9.5%, 11.2%, and 15.2%. However, no statistically significant differences were observed between the two groups in terms of DMFS, DFS, and OS (all P > 0.05, Log-rank test, Fig. 1B–D). To minimize inherent selection bias in the retrospective cohort, propensity score matching was used to balance the PORT and POCRT groups. Within the matched cohort, the POCRT group continued to demonstrate superior LRC compared to the PORT group (P = 0.022*, Gray's test, Fig. 2A). No significant differences were observed between the POCRT and PORT groups in terms of DMFS, DFS, and OS in the matched cohort (all P > 0.05, Log-rank test, Fig. 2B–D).

LRFCI, DMFS, DFS, and OS rates in patients with HNACC treated with PORT or POCRT before propensity score matching. LRFCI locoregional failure cumulative incidence, DMFS distant metastasis free survival, DFS disease free survival, OS overall survival, PORT postoperative radiotherapy, POCRT postoperative chemoradiotherapy. P values were calculated by Gray's test or log-rank test, *Statistically significant difference (P value < 0.05)

LRFCI, DMFS, DFS, and OS rates in patients with HNACC treated with PORT or POCRT after propensity score matching. Abbreviations: LRFCI locoregional failure cumulative incidence, DMFS distant metastasis free survival, DFS disease free survival, OS overall survival, PORT postoperative radiotherapy, POCRT postoperative chemoradiotherapy. P values were calculated by Gray's test or log-rank test, *Statistically significant difference (P value < 0.05)

Multivariate analysis, which included adjuvant concurrent chemotherapy, clinical-pathological factors and skull base/neck RT, identified adjuvant concurrent chemotherapy as an independent prognostic factor for LRC (Competing risks regression, HR = 0.144, 95% CI 0.026–0.802, P = 0.027*, Table 3). Independent prognostic factors for LRC also included pathological T stage, pathological N status, PNI, LVI, resection status and neck RT. PNI, LVI and histologic solid component were independent prognostic factors for DMFS, and pathological N status, PNI, LVI, histologic solid component, and resection status were independent prognostic factors for DFS, and pathological N status, PNI, LVI and histologic solid component were independent prognostic factors for OS (Cox proportional hazards regression, all P < 0.05*, Table 3). Adjuvant concurrent chemotherapy was not an independent prognostic factor for DMFS, DFS, or OS.

The acute toxicities during radiotherapy were evaluated and listed in Table 4, with comparison between the PORT and POCRT groups. Six patients experienced unscheduled treatment interruptions during radiotherapy, with five patients interrupting radiotherapy due to grade 4 toxicity reactions and one patient interrupting radiotherapy due to nasal bleeding. All patients resumed treatment within one week after the interruption and completed the full course of treatment. Overall, compared with the PORT group, the POCRT group had a higher incidence of unscheduled radiotherapy interruptions, but the difference was not significant (P = 0.057). In addition, the POCRT group had higher incidences of upper gastrointestinal toxicity and hematologic toxicities, including leukopenia, neutropenia, and anemia (all P < 0.05*). No grade 4 hematologic toxicity or treatment-related deaths were observed.