Fu 2000.

Methods	Location of trial: US, Canada. Number of centres: > 40. Funding: National Cancer Institute (grants CA21661, CA 37422, CA 32115, CA 06294). Trial ID: RTOG 9003.
Participants	Inclusion criteria: patients aged at least 18 years, with Karnofsky performance status ≥ 60%, no prior treatment, with Stage II‐IV disease M0 squamous cell carcinoma of oral cavity, oropharynx or supraglottic larynx or Stage II‐IV cancer of base of tongue or hypopharynx. T1‐T4 included. Exclusion criteria: prior or synchronous malignancy. Recruitment period: September 1991 to August 1997. OC: 110/1073 (10%). OP: 649/1073 (60%). OC+OP: 759/1073 (71%). Number randomised: 1113. Number analysed: 1073.
Interventions	Hyperfractionated radiotherapy versus hyperfractionated/accelerated/split course radiotherapy versus accelerated/boost radiotherapy versus conventional radiotherapy Hyperfractionated (n = 263): 1.2 Gy per fraction, 2 fractions per day, interfraction interval 6 hours, 5 times per week to total dose of 81.6 Gy in 35 fractions over 7 weeks. Hyperfractionated/accelerated/split (n = 274): 1.6 Gy per fraction, 2 fractions per day, interfraction interval 6 hours, 5 times per week to 38.4 Gy then 2 weeks rest then resume as for Phase 1 with further 28.8 Gy for total of 67.2 Gy in 42 fractions over 6 weeks. Accelerated/boost (n = 268): 1.8 Gy per fraction, daily, interfraction interval 6 hours, 5 times per week together with a 1.5 Gy boost field for last 12 treatment days to a total of 72 Gy in 42 fractions over 6 weeks. Conventional (n = 268): 2 Gy per fraction, 5 fractions per week to a total of 70 Gy in 35 fractions over 7 weeks.
Outcomes	Primary: locoregional control at 2 years. Secondary: overall survival, disease free survival, acute and late toxicity. Duration of follow‐up: median follow‐up was 23 months for all analysable patients and 41.2 for surviving patients.
Notes	Sample size calculation. Study was designed to detect an increase in locoregional control from 40% to 55% with a type 1 error of 0.05 and power of 80%. Sample size was increased by 20% to allow for patients being found to be ineligible, lost to follow‐up or dying without locoregional failure within 2 years. Sample planned 1080 participants. Data from Kaplan‐Meier graphs (numbers at risk presented).
Risk of bias
Bias	Authors' judgement	Support for judgement
Adequate sequence generation?	Low risk	Stratified by Karnofsky performance status (90‐100 versus 60‐80), N stage (N+ versus N‐) and primary site. Randomisation was according to the scheme of Zelen, used to achieve balance in treatment assignment among the institutions.
Allocation concealment?	Low risk	Patients were enrolled by means of a telephone call to RTOG headquarters.
Blinding ‐ Outcome Assessors	High risk	Not mentioned.
Incomplete outcome data addressed?	Unclear risk	1113 patients randomised, 28 were found to be ineligible, 5 refused protocol treatment or died before treatment started, 7 had inadequate data, total 40/1113 = 4% excluded. Exclusions not described by treatment group; appears likely that more excluded from hyperfractionated group (A) and fewer from split accelerated group (B).
Free of selective reporting?	Low risk	Important outcomes reported ‐ locoregional control, overall survival and disease free survival.
Free of other bias?	Unclear risk	Appears to be some difference between groups at baseline.