We sought to compare baseline characteristics of randomized participants with nonrandomized participants meeting the same enrollment criteria and to determine whether combining the 2 cohorts in a comprehensive cohort design would be useful for analysis.
Premenopausal women with symptomatic uterine fibroids seeking interventional therapy at 3 US academic medical centers were randomized (1:1) in 2 strata based on calculated uterine volume (<700 and ≥700 cc3) to undergo embolization or focused ultrasound surgery. Women who met the same inclusion criteria but declined randomization were offered enrollment in a parallel cohort. Both cohorts were followed up for a maximum of 36 months after treatment. The measures addressed in this report were baseline demographics, symptoms, fibroid and uterine characteristics, and scores on validated quality-of-life measures.
Of 723 women screened, 57 were randomized and 49 underwent treatment (27 with focused ultrasound and 22 with embolization). Seven of the 8 women randomized but not treated were assigned to embolization. Of 34 women in the parallel cohort, 16 elected focused ultrasound and 18 elected embolization. Compared with nonrandomized participants, randomized participants had higher mean body mass index (28.7 vs 25.3 kg/m2; P = .01) and were more likely to be gravid (77% vs 47%; P = .003) and smokers (42% vs 12%; P = .003). Age, race, uterine volume, number of fibroids, and baseline validated measures of general and disease-specific quality of life, pain, depression, and sexual function did not differ between the groups. When we performed a comprehensive cohort analysis and analyzed by treatment arm, the only baseline difference observed was a higher median McGill Pain Score among women undergoing focused ultrasound (10.5 vs 6; P = .03); a similar but nonsignificant trend was seen in visual analog scale scores for pain (median, 39.0 vs 24.0; P = .06).
Using a comprehensive cohort analysis of study data could result in additional power and greater generalizability if results are adjusted for baseline differences.