MSG 1994.

Methods	Methods of randomisation and allocation: computer‐generated schedule with packaged coded medication Multicentre (99 sites), double‐blind with parallel group design, placebo controlled Number of women randomised: 1724 Number of withdrawals: 255 after 1 year of follow‐up; 339 after 2 years of follow‐up Source of funding: Wyeth‐Ayerst Research
Participants	Countries: US and Europe Healthy women aged 45 to 65 years with an intact uterus were recruited from 99 sites Inclusion criteria: last natural menstrual cycle at least 12 months before the baseline screening; FSH higher than the lower limit for postmenopausal women for the given laboratory; no use of estrogen‐ or progestogen‐containing medication for at least 2 weeks before the pre‐study screening Exclusion criteria: any contraindication for estrogen or progestogen use, or if they had used any estrogen‐containing medication within 3 months of entry; major medical illness, liver, kidney or diabetes; hypertension, systolic blood pressure > 160 mmHg or diastolic pressure > 90 mmHg; abnormal cervical cytology or endometrial hyperplasia at baseline biopsy
Interventions	Continuous vs sequential (1) 0.625 mg per day CEE + placebo (2) 0.625 mg per day CEE + 2.5 mg per day MPA continuous (3) 0.625 mg per day CEE + 5 mg per day MPA continuous (4) 0.625 mg per day CEE + 5 mg per day MPA last 14 days of the cycle (days 15 to 28), + placebo (days 1 to 14) (5) 0.625 mg per day CEE + 10 mg per day MPA last 14 days of the cycle (days 15 to 28), + placebo (days 1 to 14) Duration: 1 year (13 cycles)
Outcomes	Frequency of hyperplasia, carcinoma, or both (confirmed by endometrial biopsy) at 6 and 12 months (at the end of cycles 6 and 13) Frequency of irregular bleeding or spotting (number of cycles)
Power calculation described	Yes, "With a hypothesized incidence of endometrial hyperplasia of 7.55 in the Unopposed group and 25 in the CEE/MPA groups it was calculated that 215 patients per treatment group would provide 80% power to detect at least one statistically significant difference at the 0.0125 level (includes Bonferroni adjustment for multiple comparisons)" (see Woodruff 1994 p. 1214 under MSG 1994)
Analysis by Intention to Treat	No
Notes	If hyperplasia was confirmed, the patient was withdrawn from the study and given appropriate treatment
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Methods of randomisation and allocation: computer‐generated schedule with packaged coded medication
Allocation concealment (selection bias)	Low risk	Randomisation conducted centrally
Blinding (performance bias and detection bias) All outcomes	Low risk	‐
Incomplete outcome data (attrition bias) All outcomes	Low risk	Endometrial biopsies after 1 year of treatment on 1385 of the 1724 women randomised (80%)
Other bias	Low risk	5 parallel groups well balanced at baseline