Patients undergoing surgical resection for oral tongue SCC at the Royal Melbourne Hospital (RMH) or the Peter MacCallum Cancer Centre (PMCC) between 2006-2017 were eligible for inclusion.   Multisite ethics approval was obtained for this study (HREC/17/PMCC/220). Eligible patients were identified from two separate institutional databases, one of which was prospectively collected, and was retrospectively analysed.  Patients were excluded if they were aged less than 18 years or had a prior history of head and neck cancer (HNC).  PMCC provides the large majority of adjuvant treatment for patients treated surgically at both centres, and only cases where adjuvant treatment was delivered at PMCC were included. This was to ensure treatment was delivered at a high volume radiation oncology HNC centre using the subsequently mentioned predefined treatment policy with an established radiotherapy quality assurance program. All cases were discussed pre- and post-operatively at a multi-disciplinary tumour meeting. Pre-operative staging generally included computed topography (CT) and selected patients underwent magnetic resonance imaging (MRI) or (18F)-fluorodeoxyglucose positron emission tomography/CT (FDG PET/CT) imaging. 

Surgical excision of the primary was undertaken to obtain clear margins.  If feasible, re-resection was performed for positive margins.  Neck dissection was  performed where there was clinical involvement of the neck or electively in the cN0 neck where the DOI was assessed as being ≥4mm, the role of dissection for DOI between 2mm and 4mm was resolved within the MDM on an individual case basis. The policy of both institutions was to prophylactically dissect the contralateral neck if the primary tumour reached or crossed the midline, or if there was clinical involvement of the contralateral neck.

Postoperative radiotherapy (PORT) was indicated in suitably fit patients in the setting of adverse pathological features, including T3/4 tumour, close/positive margin, lymphovascular space invasion (LVSI), perineural invasion (PNI), DOI≥7mm and any nodal disease (>N1).  Contralateral neck PORT was delivered if the neck was pathologically involved or if it was dissected but pathologically negative, but to a lower dose.  Elective irradiation (ENI) of the contralateral neck was dictated by a predefined institutional protocol, which included tumour extension to within 1cm of the midline or 2cm to the tongue tip or if there was extensive base of tongue involvement. Distance to midline or the tip was determined from clinical and radiological findings.  Standard dose fractionation schedules were 66Gy in 33 fractions for positive margins or extracapsular extension (ECE) and 60Gy in 30 fractions for all other patients. Dissected but pathologically uninvolved neck levels were treated to an equivalent dose of 54Gy.  Over the study period we transitioned from 3D conformal radiotherapy to intensity modulated radiotherapy (IMRT) in the adjuvant setting. In suitably fit patients, concomitant chemotherapy was recommended in patients with a positive margin or ECE.   Unless contraindicated, weekly cisplatin was the recommended chemotherapy regimen. Patients were followed up at least 3 monthly for the first 2 years and at least 6 monthly for the next 3 years.

Patient medical records were accessed for demographic and tumour factors and primary, adjuvant and salvage treatment details.  Patients were staged in accordance with the American Joint Committee on Cancer (AJCC) 7^th edition. Time to event outcomes included freedom from locoregional failure (FFLRF), freedom from contralateral neck failure (CNF), distant-metastasis free survival (DMFS), disease-specific survival (DSS) and overall survival (OS). Local, regional or distant failure was defined as radiological or pathological recurrence. Time to event outcomes were measured from the date of surgery to the date of the event. Other causes of death were censoring events for DSS and patients with unknown cause of death were not included in the DSS estimates. Any CNF was recorded, even if not isolated.

Statistical Analysis

Baseline characteristics were summarised using descriptive statistics. The Kaplan-Meier method was used to estimate the time-to-event curves. Estimates and associated 95% confidence intervals at key time points were reported. No imputation of missing values was performed. All statistical analyses were performed in R version 3.6.1.