THE RANDOMIZED CONTROLLED TRIAL (RCT), INTROduced into clinical research more than 50 years ago, has assumed increasing importance in evaluating medical interventions. Most studies comparing randomized and nonrandomized trials addressing the same question have found that RCTs generally yield less favorable assessments of treatment efficacy than the nonrandomized trials. Although this conclusion is not quite unanimous, there has been increasing recognition of the importance of random assignment in medical research and the RCT now occupies a prominent position in the hierarchy of evidence on which to base clinical judgments. In addition to random assignment, allocation concealment and other features of experimental design are also critical for proper conduct of an RCT. Attention to these essential aspects improves study quality and, presumably, the trust clinicians can place in the findings. High-quality RCTs are challenging to perform. Under the best circumstances, recruitment can be difficult, attrition constantly threatens sample size adequacy, breaks in study design can negate an entire study, costs are inevitably high, and long periods of time are needed for completion. Conducting a rigorous RCT becomes even more difficult when surgical therapies are evaluated, particularly when surgery is compared with nonsurgical treatment. Masking usually is not possible and placebo controls, such as sham procedures, are theoretically possible yet ethically questionable. Furthermore, technical skills are important for determining surgical outcome but inevitably differ from surgeon to surgeon. Moreover, when surgery is involved, recruitment can be complicated by the unique personal and emotional characteristics of the patient-physician relationship. Presumably, these are some of the reasons that surgical treatments are, in general, less likely than pharmacologic therapy to be supported by high-quality evidence from RCTs. Nevertheless, in view of their high cost and treatment-associated risk, it is all the more important that operative therapies are supported by the best possible evidence. Two articles in this issue of THE JOURNAL describe RCTs designed to assess the efficacy of hysterectomy in women with excessive uterine bleeding unrelated to pregnancy or malignancy. Hysterectomy is an appropriate intervention for investigation, given its frequency as the most common major, nonobstetric operation and that it usually is elective (ie, not typically a life-saving measure). The 2 studies differed with respect to the controls. One study used “expanded medical therapy” with sex steroid hormones, prostaglandin synthetase inhibitors, or both. The other study used intrauterine administration of a continuous-release progestational hormone. Both trials were designed and conducted carefully and analyzed rigorously, and both focused on subjective patient responses as principal outcomes. The study by Kuppermann and colleagues, conducted at 5 university-based centers across the United States, enrolled women who were dissatisfied after previous treatment with medroxyprogesterone acetate, an orally effective, synthetic progestational hormone. These women were randomly assigned to undergo hysterectomy or receive any of a variety of oral or injectable hormone regimens with or without prostaglandin synthetase inhibitors. The investigators encountered serious difficulties in recruiting study participants. This is understandable because the women being recruited did not achieve symptom relief with one form of hormonal therapy and were being asked to participate in a study in which they had a 50% chance of being assigned to receive a similar treatment. Because of these recruitment problems, the enrollment goal was lowered, such that ultimately, 63 women were randomly assigned: 31 to hysterectomy and 32 to medical therapy (of whom 17 crossed over to have hysterectomies by the end of the 2-year study). Despite the potential problem with statistical power, intention-to-treat analysis revealed statistically significant advantages for hysterectomy 6 months after randomization in nearly all of the outcome measures assessed. These differences mostly disappeared by 2 years, reflecting the large proportion of women assigned to receive medical treatment who opted for hysterectomy while the study was progressing. The authors concluded that among women with excessive uter-