A list of consecutive patients who were diagnosed with OCSCC and received RT between 2002 and 2018 was extracted from an institutional cancer registry; a total of 486 patients were identified. The inclusion criteria were as follows: pathologically confirmed SCC of oral cavity, resection of primary tumor with/without neck node dissection, and received postoperative RT using three-dimensional conformal RT (3D-CRT) or IMRT. Exclusion criteria were as follows: PORT followed by salvage resection after recurrence of disease (n = 145); pathologically not a SCC, such as adenoid cystic carcinoma or sarcoma (n = 41); palliative treatment due to distant metastasis (n = 21); and OCSCC with double primary lung cancer (n = 4). After all exclusions, the data of 275 patients were analyzed.

The procedures followed in this study were in accordance with the Helsinki Declaration of 1975, as revised in 2000. This study was approved by our Institutional Review Board (IRB # 4-2019-0401).

Pretreatment evaluation included a complete history, physical examination and laboratory studies including a complete blood cell count and serum chemistry profile. Patients underwent imaging studies such as computed tomography (CT), magnetic resonance imaging for primary tumor and neck node involvement evaluation, and positron emission tomography (PET) for systemic evaluation.

The surgical techniques included resection either by open approaches or by trans-oral robotic surgery using da Vinci Robot (Intuitive Surgical Inc., Sunnyvale, CA). Neck dissection was performed on the involved side or both sides of neck in order to examine the regional lymph node involvement.

RT was delivered using megavoltage photons (≥6 MV). Using 3D-CRT, a cone-down technique was used. Using IMRT, the simultaneous integrated boost technique was used in all patients. The high-risk clinical target volume (CTV)1 encompassed the primary tumor bed (based on preoperative imaging, physical examination, and operative findings) and extranodal extension (ENE) or microscopically involved surgical margin lesions (RM+). The intermediate-risk CTV2 encompassed the pathologically positive hemi neck; this frequently required coverage of nodal levels I, IIa-b, III, and IV for most cases. The low-risk CTV3 usually encompassed the prophylactically treated neck with a low risk of harboring microscopic disease (e.g., the uninvolved low or contralateral neck). For the planning target volume (PTV), a 2–5 mm margin was applied to the CTV. The intended total dose for PTV1 was 60–66 Gy in 2.0 Gy per fraction. If ENE or RM+ were present, the region was treated with 64–66 Gy. The intended total doses for PTV2 and PTV3 were 60 Gy and 45-50 Gy, respectively. The target volume was delineated on simulation CT fused with PET and other images. Helical tomotherapy (HT), an image-guided IMRT system using megavoltage CT (MVCT) that provides precise delivery, was used in IMRT. HT was demonstrated to have better target volume dose conformity and homogeneity than other IMRT  (8). The daily MVCT images were fused with the original treatment planning based on soft tissue and bony structures at each fraction. The position was corrected manually to align target volume after automatic registration  (9).

Concurrent chemotherapy was added to RT in patients with high risk OCSCC. High risk was defined by ENE, RM+, perineural invasion (PNI), lymphovascular invasion (LVI), and/or multiple nodes involvement. ENE was defined as extension of cancer cells through the lymph node capsule. The pathological margins were classified as negative (>5 mm), close (≤5 mm), and positive (presence of cancer cells microscopically [RM+]). Patients received concurrent chemotherapy as follows: cisplatin was administered as a weekly dose of 25–40 mg/m² or a triweekly dose of 100 mg/m² from the first day of RT.

Each patient was examined by a dental team for pre-radiotherapy dental care, which was completed before the initiation of PORT. In addition to clinical examination, radiographic examination was performed to determine the periodontal status and the presence of periapical inflammation and other dental diseases. Each patient was examined at least once a week to monitor treatment-related toxicities. Treatment-related toxicities were graded according to the Common Toxicity Criteria for Adverse Events version 5.0.

Patterns of first failure were defined as loco-regional failure or distant metastases. The date of failure was the date of tissue confirmation or imaging study showing evidence of failure. Local failure was defined as failure occurring within the same site of the primary tumor, regional failure if occurring within the regional lymph nodes, and distant failure if occurring outside of the local and regional areas.

Statistical analyses were conducted using IBM SPSS version 25.0 (IBM Corp., Armonk, NY). The differences in characteristics and toxicities were compared using chi-square tests, and the Kaplan–Meier method was used to calculate the OS, progression-free survival (PFS), loco-regional failure-free survival and distant metastasis-free survival; differences between the curves were analyzed using the log-rank test. Cox proportional hazards models were used to assess the association of variables with the survival and hazard ratio (HR) and confidence interval (CI). Statistical significance was defined as p < 0.05. Factors showing p < 0.10 in the univariate analyses were included in the multivariate analyses.

The patient and pathologic characteristics are summarized in Table 1 . The median age of patients was 58 years (range, 18–86) and the male-to-female ratio was 6:4. The most common primary site was the oral tongue (54%, n = 148), followed by the buccal mucosa (13%, n = 37), retromolar trigone (12%, n = 32), and alveolar ridge (12%, n = 32). The most common pathologic T and N status were T4 (36%, n = 98) and N3 (30%, n = 82). A total of 119 patients (43%) had stage IVA disease according to American Joint Committee On Cancer (AJCC) 8^th edition and 82 patients (30%) had stage IVB disease. A total of 103 patients (37%) had ENE and 72 patients (26%) had RM+.

The characteristics that differed by treatment group included pathologic N status, AJCC stage, ENE, resection margin status, RT dose and RT modality. More patients in the CCRT group had N3 disease and AJCC stage IVB disease (46.6% vs. 10.2%, p<0.001), ENE (56.1% vs. 15.7%, p<0.001), LVI (28.4% vs. 13.4%, p=0.033). The mean RT dose was higher in the CCRT group (62.2 Gy vs. 60.9 Gy, p=0.002) and fewer patients received IMRT (7.4% vs. 19.7%, p=0.004). The other characteristics were well balanced between treatment groups.

Ipsilateral neck dissection was performed for 268 patients (98%), of whom 129 (47%) received modified radical neck dissection (mRND), 121 (44%) received supra-omohyoid neck dissection (SOND), and 18 (7%) received selective neck dissection (SND). Contralateral neck dissection was performed for 87 patients (32%), of whom 10 (4%) received mRND, 58 (21%) received SOND, and 19 (7%) received SND.

A total of 36 patients (13%) received 3D-CRT and 239 patients (87%) received IMRT. The median RT dose was 63 Gy (range, 50–67.5), median high risk CTV fractional dose was 2.1 Gy (range, 1.8–2.5) and median fraction number was 30 (range, 20-36). The median treatment day of RT was 42 days (range, 32–70) and total treatment time from surgery to finish of RT was 82 days (range, 63–177). A total of 148 patients (54%) received concurrent chemotherapy, three patients received induction chemotherapy consisting of cisplatin and gimeracil, also known as TS-1, before surgery, and one patient received maintenance gimeracil chemotherapy after RT. The median cumulative dose of cisplatin was 200 mg/m² (range, 40–300 mg/m²). Most patients (66%, n = 97) completed chemotherapy without interruption. A total of 17 patients (12%) discontinued concurrent chemotherapy because of grade 3 poor oral intake (n = 7), grade 3 fatigue (n = 5), and grade 3 neutropenia (n = 5). Furthermore, the chemotherapy dose was reduced in 15 (10%) patients because of grade 3 fatigue (n = 8), and grade 3 neutropenia (n = 7).

With a median follow-up of 40 months (range, 5–203), the 5-year OS and PFS rates were 65% and 61%, respectively ( Figure 1 ). Median OS was not reached, and median PFS was 140 months (95% CI, 67.6–212.4). According to primary site, the 5-year OS were follows: oral tongue (69%), buccal mucosa (64%), retromolar trigone (55%), gingiva and alveolar ridge (56%), floor of mouth (76%), hard palate (42%).

The prognostic factors associated with OS are summarized in Table 2 . Univariate analysis revealed that pathologic T and N status, AJCC stage, ENE, and PNI were significant prognostic factors associated with OS. In the multivariate analysis, pathologic T status and PNI were associated with poorer OS. The prognostic factors associated with PFS are summarized in Table 3 . Univariate analysis revealed that pathologic T and N status, AJCC stage, ENE, and PNI were significant prognostic factors associated with PFS. In the multivariate analysis, pathologic T status and PNI were associated with poorer PFS.

There were no significant differences in the 5-year OS (64% vs. 65%, p = 0.974) and PFS (62% vs. 60%, p=0.846) between treatment groups ( Figure 2 ). No significant difference was observed in the 5-year loco-regional failure-free survival (79% vs. 77%, p = 0.599) and distant metastasis-free survival (78% vs. 81%, p = 0.475).

In the subgroup analysis according to high risk (ENE, RM+, PNI, LVI, and multiple node), the concurrent use of chemotherapy showed significantly improved OS in patients with ENE (HR 0.39, 95% CI 0.21–0.43, p = 0.003, Figure 3 ), while there was no advantage in OS in patients with RM+, PNI, LVI, and multiple node.

In total, 103 and 72 patients had ENE and RM+, respectively; 22 patients had both ENE and RM+ and 153 patients had either ENE or RM+. Concurrent chemotherapy significantly improved the 5-year OS in patients with ENE (56% vs. 32%, p = 0.002); however, it did not improve the 5-year OS in patients with RM+ (64% vs. 64%, p = 0.899). Moreover, concurrent chemotherapy was not beneficial in patients with either ENE or RM+ (p = 0.116), but it was beneficial in terms of survival in patients with both ENE and RM+ (p < 0.001). The mean RT dose for patients with ENE was statistically higher than that for patients without ENE (62.7 Gy vs. 60.9 Gy, p < 0.001). Similarly, the mean RT dose for patients with RM+ was statistically higher than that for patients without RM+ (62.6 Gy vs. 61.2 Gy, p < 0.001).

The most common first pattern of failure in patients with ENE was distant failure (26%), followed by local (16%) and regional failures (15%). In patients with RM+, the rates of local (13%), regional (15%), and distant (13%) failures were similar. The overall rates of distant failure in patients with ENE, RM+, and negative resection margins were 32%, 21%, and 16%, respectively.

Treatment failure occurred in 48 and 59 patients in the RT alone and CCRT groups, respectively. The patterns of first failure are summarized in Figure 4 . A total of 37 patients (13%) had local failures, 33 patients (12%) had regional failures, and 37 patients (13%) had distant failures ( Figure 4A ). The most common patterns of failure in CCRT was distant failure (15%) followed by regional (14%) and local (11%) ( Figure 4B ), and the most common patterns of failure in RT alone was local failure (17%), followed by distant (12%) and regional failures (9%) ( Figure 4C ).

A total of 105 patients (38.2%) had grade 3 mucositis, and three patients had grade 3 skin reaction. With respect to late toxicities, 10 patients (3.6%) suffered from osteoradionecrosis (ORN) of the mandible and five patients had orocutaneous fistula. Among the 10 patients with ORN, five patients had a primary tumor near the mandible (one with retromolar trigone and four with gingiva), five had pathologic T4a disease and eight received more than 60 Gy of RT. Among the five patients with orocutaneous fistula, all patients had pathologic T4a disease and received more than 60 Gy, and three had RM+.