Methods	GOG orchestrated multicenter phase 3 randomised controlled trial
Participants	A total of 206 women with any stage of previously untreated locally confined mixed mesenchymal mullerian malignant tumour (primary carcinosarcomas of the uterus) were randomised to whole abdominal irradiation (WAI) arm (n = 105) or chemotherapy (n = 101) with Cispltain, Ifosfamide and mesna All women had a hysterectomy, bilateral salpingo-oopherectomy and optimal resection of all gross intra-abdominal/pelvic disease, including macroscopically involved pelvic and para-aortic nodes, leaving no residual disease any larger than 1 cm before randomisation Twenty-five cases were excluded from the analysis based on review of histology and one was excluded because the GOG committee felt there was inappropriate residual disease at randomisation
Interventions	Chemotherapy consisted of intravenous (IV) cisplatin (20mg/m2/day × 4 days) followed by a 1 hour infusion of ifosfamide (1.5g/m2/day IV × 4 days) with mesna (120 mg/m2 IV bolus over 15 min on day one, followed by 1.5 g/m2/day IV continuous infusion × 4 days beginning with day one) every 3 weeks for three cycles within 8 weeks following initial surgery Whole abdominal irradiation involved 30 Gray to the whole abdomen at 1 Gray per fraction, two fractions per day, and 5 days each week with a minimum of 6 hour between morning and afternoon fractions (hyperfractionation). Due to slow patient accrual, in August 1996, the dose fractionation schedule was changed to once-daily fractions of 1.5 Gray for 5 days each week to the same total dose to the whole abdomen of 30 Gray. Early recruits were given a pelvic boost of hyperfractionated 20 Gray. The fractionation schedule was also changed in August 1996 to once-daily fractions of 1.8 Gray for 5 days each week to a total dose of 19.8 Gray (cumulative true pelvic dose of 49.8 Gray)
Outcomes	Overall survival was defined as the time from randomisation to death from any cause or, for living patients, the date of last contact. The recurrence-free interval was assessed from the date of randomisation to the date when clinically evident disease was observed
Results (OS & PFS)	The estimated crude probability of recurring within 5 years was 58% (WAI) and 52% (CIM). Adjusting for stage and age, the recurrence rate was 21% lower for CIM patients than for WAI patients (relative hazard (RH) = 0.789, 95% confidence interval (CI): (0.530 to 1.176), P = 0.245, 2-tail test). The estimated death rate was 29% lower among the CIM group (RH = 0.712, 95% CI: 0.484 to 1.048, P = 0.085, two-tail test) The estimated crude probability of surviving at least five years following diagnosis was approximately 35% for those randomised to WAI versus 45% for those randomised to CIM. After adjusting for stage and age at diagnosis, the estimated death rate was 29% lower for CIM patients than WAI patients (RH = 0.712, 95% CI: 0.484 to 1.048). A similar result was obtained when all women (regardless of eligibility) were included in the analysis of survival for an intention-to-treat analysis (RH = 0.727, 95% CI: 0.503 to 1.050)
Results (site of recurrence)	There have been 112 recurrences reported from 206 women. These include 60 in the radiotherapy group and 52 in the chemotherapy group. There were slightly more vaginal recurrences in the CIM group and more abdominal recurrences in the WAI group, but these differences were not statistically significant The sites of first recurrence for women allocated radiotherapy included the vagina (4/105; 3.8%), pelvis (14/105; 13.3%), abdomen (29/105; 27.6%) and distant sites 13/105; 25.7%). The corresponding sites of first recurrence for women allocated chemotherapy was the vagina (10/101; 9.9%), pelvis (14/101; 13.8%), abdomen (19/101; 18.8%) and other distant sites (10/101; 23.8%)
Toxicity	Grade 3 or 4 acute anaemia (11 versus 1, P < 0.01) and neuropathy (central and peripheral) (9 versus 0, P < 0.01) occurred more often among those treated with CIM chemotherapy than radiotherapy. Regarding grade 2, 3 or 4 late effects, gastrointestinal events occurred more often among those treated with radiotherapy (WAI) than chemotherapy (10 versus 0, P < 0.001) Two women in the radiotherapy cohort died as a direct consequence of radiation hepatitis, while one patient in the chemotherapy arm died of an acute systemic blood infection which originated at the port site and was complicated by neutropenia
Notes
Risk of bias
Bias	Authors’ judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Randomisation was orchestrated by the central GOG office block randomisation based
Allocation concealment (selection bias)	Low risk	The authors confirm that treatment assignments remained concealed from randomisation code generation to registration
Blinding (performance bias and detection bias) All outcomes	High risk	It was not possible to blind the observers or participants to treatment allocation in this type of trial
Incomplete outcome data (attrition bias) All outcomes	Low risk	Primary outcome data were available on all 206 women and data on adverse events was available from 197/206 (96%) cases. A total of 232 women were randomised, but 26 were found to be ineligible and excluded (appropriately before analysis
Selective reporting (reporting bias)	Low risk	There is no suggestion of selective reporting and all important survival events have been reported
Other bias	Low risk	We have been unable to find any suggestion from any author that the study was biased
Main problems with the trial as perceived by this analysis	Low risk	The protocol for delivering radiotherapy changed midway through the study but this is unlikely to have any impact on the conclusions that can be derived from this study