In the United States, hysterectomy is exceedingly common—second only to cesarean delivery among major operations1—and is a definitive treatment for the control of uterine bleeding. Because uterine removal is irreversible and the surgical risks exceed those of most other treatments, patients with heavy menstrual bleeding are likely to undergo an initial trial of medical treatment. Some women continue to have troublesome symptoms, often for years, and eventually come to the point of seeking a definitive solution through elective hysterectomy. Observational studies suggest that the resulting symptomatic relief improves their quality of life,2 but there are no randomized trials comparing the benefits and harms of hysterectomy with those of oral medical treatment.
We conducted the Medicine or Surgery randomized trial to compare treatment alternatives for women with abnormal bleeding that was refractory to cyclic medroxyprogesterone acetate. Specifically, we compared expanded treatment with oral medications versus hysterectomy among women who did not have a strong preference regarding uterine conservation or removal. We recommended a medical regimen of oral contraceptive pills and prostaglandin inhibitors and permitted variations in this regimen according to patient and provider preference. Our report summarizes clinical outcomes of the Medicine or Surgery trial. The primary outcome, change in overall mental health, is the subject of a separate publication.3
MATERIALS AND METHODS
The study design and recruitment methods for the Medicine or Surgery study were detailed in an earlier publication.4 Subjects were recruited from the gynecology clinics at the University of Alabama, Birmingham; University of Tennessee, Memphis; Wayne State University in Detroit; and clinics affiliated with the University of California, San Diego. The study was approved by the institutional review boards of the 4 clinical centers and the University of California, San Francisco, where the coordinating center was located.
We included English-speaking premenopausal women aged 30–50 years with at least 2 months of abnormal uterine bleeding, defined as more than 7 days of flow each month or flow heavy enough to produce anemia (hematocrit less than or equal to 32%). Women aged 45 years or more were included if their follicle-stimulating hormone levels did not exceed 30 mIU/mL and if their endometrial biopsy did not show evidence of hyperplasia or carcinoma. We excluded from participation women with other causes of anemia, who had desire to preserve fertility, or who had evidence of pregnancy, endocrinopathy, or coagulopathy. We also excluded women who had received treatment with long-acting regimens (depot medroxyprogesterone acetate or gonadotrophin-releasing hormone agonist) within 6 months of screening or who had used oral contraceptive pills or intrauterine devices within 3 months of screening and those who had contraindications to study medications, potential problems with subject compliance or follow-up, or were participating in another medication trial. In addition, we excluded women with evidence of pelvic pathology for which hysterectomy or other specific directed therapy was indicated (ie, ultrasound, hysterogram, hysteroscopy, and/or biopsy showing endometrial polyp, submucous leiomyoma, endometrial hyperplasia or carcinoma, or cervical dysplasia or carcinoma).
Between August 1997 and December 2000, we screened 1,556 women and identified 1,013 individuals who were potentially eligible for the Medicine or Surgery trial (Figure 1). Of the 413 women started on a 3- to 5-month trial of 10–20 mg of cyclic medroxyprogesterone acetate 10–14 days per month, 271 (65.6%) completed the trial and 215 (52%) completed a regimen satisfaction survey. Of the 92 who were dissatisfied, 38 (41.3%) agreed to participate in the randomized trial. The other 25 randomly assigned participants had received a previous treatment with medroxyprogesterone acetate and were dissatisfied; they were not required to repeat a medroxyprogesterone acetate trial. Follow-up data were available at 24 months for 29 of 31 subjects randomly assigned to hysterectomy and for 30 of 32 subjects randomly assigned to expanded medical treatment.
We informed eligible patients that they would be randomly assigned to hysterectomy or medical treatment. Each participant signed a consent form approved by the local institutional review board. Eligible participants who consented were assigned with equal probability to 1 of the 2 treatment groups by using a procedure detailed in an earlier publication.4 Randomization was stratified by clinical center and, within strata, treatment was assigned by using randomly permuted blocks of sizes 4, 6, and 8. Sequenced randomization envelopes were opened in front of a witness. The witnessed and signed assignment sheet was then faxed to the coordinating center, where it was irrevocably entered into the treatment assignment database. The Medicine or Surgery trial was designed to closely reflect what occurs in community practice as is typical in effectiveness trials. Hysterectomies were performed by university and private-sector gynecologists; some were performed with residents as surgical assistants. We allowed surgeons to choose the route and technique of hysterectomy using their own standard criteria. We indicated that bilateral oophorectomy should be performed only to treat abnormal findings or at the patient’s specific request. We recommended that patients assigned to medical treatment receive a combined oral contraceptive and a prostaglandin inhibitor, but we permitted variations in the types and combinations of medications. We encouraged participants to stay in their assigned groups but otherwise did not prohibit participants from seeking alternative therapies.
At the randomization visit and every 3 months for 2 years, participants completed questionnaires assessing demographics, gynecologic and reproductive history (at baseline only), health beliefs and attitudes, and symptomatology, including magnitude and impact of bleeding, pelvic pain, pelvic or bladder pressure, low back pain, and urinary incontinence, and quality-of-life measures, including sexual functioning. We created single- and multi-item health belief and attitude scales as described in a previous report.4 The quality-of-life measures and outcomes are the subject of a separate report.3
Before the study, data collectors (study coordinators and research assistants) at the clinical sites attended an initial training session led by the coordinating center in which they received instruction and practice in the procedures and interviewing techniques used in the study. In addition, a detailed operations manual was prepared by the coordinating center that included instructions on conducting the study procedures and subject interviews. Neither clinic personnel nor participants were blinded to the participants’ randomized treatment assignment.
Local research assistants extracted data on hysterectomy characteristics and inpatient complications from operative reports, pathology reports, and discharge summaries using checklists designed for this purpose. Length of hospitalization was calculated as date of discharge minus date of admission (usually the day of surgery). Hospitalizations were ascertained comprehensively from cross-referencing abstracted data from the hysterectomy admission, institutional data with diagnosis-related groups codes on readmissions to hospitals affiliated with each clinical center, and patients’ reports of hospitalization assessed every 3 months during the study. Admissions were independently classified into categories by 1 of the authors, and as “attributed, at least in part, to hysterectomy,” in an independent review by 3 authors who were blinded to treatment group assignment; discrepancies were resolved by discussion and consensus.
We assessed the validity of the randomization by comparing the treatment groups on baseline characteristics using t tests for most continuous outcomes, the Wilcoxon rank-sum test for duration of bleeding, and Fisher exact test for the categorical characteristics. Changes in clinical outcomes were first analyzed according to treatment assignment (intention to treat) without regard to crossing over to hysterectomy. Differences between the medicine and hysterectomy groups were assessed by using a mixed linear model5 for the changes from baseline to 18 and 24 months, treated as repeated outcomes, and adjusted for the baseline scores. The criterion for statistical significance was set at P < .05.
We also conducted an exploratory within-groups analysis in which women who were randomly assigned to expanded medical treatment were grouped according to whether or not they underwent hysterectomy by the end of the 2-year follow-up period. We assessed the statistical significance of improvements from baseline to 18 and 24 months within each of the 3 groups using mixed linear models, again adjusting for baseline score. A Kaplan–Meier curve was used to describe the timing and cumulative incidence of hysterectomy among women assigned to medical treatment. We examined predictors of crossing over using Cox proportional hazards models6 with both fixed baseline and time-dependent covariates.
Sample size requirements were determined by the main study outcome, which was a clinically significant improvement in the Mental Component Summary score of the 36-item Short Form Health Survey (SF-36) of the Medical Outcomes Study.4 Estimating that the standard deviation of the Mental Component Summary score would be 10 points,7 we calculated that a sample size of at least 60 participants would allow us to reject the null hypothesis of no difference in mental component subscore improvements between hysterectomy and medical treatment with 90% power in a 2-sided test with α = .05 if the true effect size was 6.8 units or greater on a 100-point scale.
A data and safety monitoring board monitored the study every 6–12 months for adverse trends in either of the treatment groups. At each meeting, the board deemed the adverse study outcomes to be consistent with those published for hysterectomies in general and allowed the study to continue without modification.
Between August 1997 and December 2000, we randomly assigned 63 women: 31 to hysterectomy and 32 to expanded medical treatment. Most were recruited from the clinical centers in Birmingham (n = 39) and Memphis (n = 21). Most of the participants were employed and had health insurance, and approximately half were African American (Table 1). Multiple pelvic symptoms were common, and there was a high degree of urinary symptomatology. The randomized groups were similar in their baseline characteristics, with a difference that was statistically significant in only 1 of 38 instances. Women randomly assigned to medicine were more likely to report pelvic or bladder pressure than women randomly assigned to hysterectomy (84% versus 55%, P = .01). There were no differences between groups in the presence of leiomyomata at the time of the initial examination or in reports of pelvic or low back pain. Health-related quality-of-life measures were similar and reflected extremely low levels of satisfaction with symptom level and current health. Participants in both groups endorsed health beliefs and attitudes favoring medical treatments as well as uterine removal, as would be predicted from their willingness to be randomly assigned. The median duration of abnormal bleeding symptoms was 3 to 4 years, with the 75th percentile 11 to 12 years. Previous treatments in addition to medroxyprogesterone acetate included a variety of common medical and surgical approaches.
Twenty-nine of the 32 participants randomly assigned to medical treatment received a hormonal treatment: combined oral contraceptive pills containing 30–35 μg of ethinyl estradiol (n = 12), cyclic progestin pills (n = 5), continuous progestin pills or injection (n = 2), or conjugated estrogen with (n = 8) or without (n = 2) progestin. Seventeen of the 32 participants also received a prostaglandin inhibitor starting on day 1 of menses for 5 days: naproxen sodium (n = 15), ibuprofen (n = 1), or rofecoxib (n = 1). By 24 months after randomization, 17 participants randomly assigned to medical treatment had undergone a hysterectomy. The cumulative rate of hysterectomy in this group was 38% by 9 months and reached a plateau at 53% by 18 months after randomization. There were no perioperative complications or late complications requiring readmission after hysterectomy, but 2 patients assigned to medicine were admitted for other treatments. One had hysteroscopy and suction curettage for abnormal uterine bleeding 7 months after randomization and a gastroplasty for obesity 8 months later. Another patient presented with hemorrhagic vaginal bleeding and received a blood transfusion and injection of gonadotropin-releasing hormone agonist 3 weeks after randomization and 4 months before electing to undergo a hysterectomy. Two patients had hospital admissions that were deemed unrelated to their gynecological problems—1 for removal of a colon polyp and 1 for laparoscopic cholecystectomy.
Twenty-eight of the 31 participants randomly assigned to hysterectomy underwent the procedure, with 18 receiving a vaginal hysterectomy, 7 a total abdominal hysterectomy, and 3 a supracervical abdominal hysterectomy. The majority (83%) of procedures occurred within 2 months of randomization and most (86%) required a 1- to 2-day hospitalization. Operative times averaged 115 minutes (standard deviation [SD] 34), and the mean estimated blood loss was 298 mL (SD 304). Uterine specimens weighed an average of 226 grams (SD 286) and showed leiomyomata in 19 cases and adenomyosis in 11 cases. The endometrial pattern was secretory in 11 cases and proliferative in 14 cases. One specimen showed a high-grade cervical lesion (moderate dysplasia) after a preoperative Pap test showed mildly atypical squamous cells.
Perioperative complications occurred in 2 patients randomly assigned to hysterectomy. One patient had a superficial thermal injury (blanching less than 1 mm) to her distal small bowel during lysis of adhesions to the anterior abdominal wall at the time of supracervical hysterectomy. The injury was oversewn, and the patient had a normal postoperative course with normal return of bowel function before discharge on the third postoperative day. Another patient had a postoperative fever. Three patients experienced late complications requiring readmission after hysterectomy. One patient was readmitted on postoperative day 6 after vaginal hysterectomy with hypovolemia requiring laparotomy and salpingo-oophorectomy; she was admitted again on postoperative day 21 with hematemesis and esophagitis. Another patient presented with a new seizure disorder on postoperative day 9 after abdominal hysterectomy. A patient with persistent cyclic bleeding after supracervical hysterectomy was readmitted 15 months later for trachelectomy. Two patients had readmission deemed unrelated to hysterectomy, 1 for extremity cellulitis 9 months after hysterectomy and diverticulitis 10 months later and another for partial thyroidectomy.
Clinical outcomes were compared as changes from baseline values (Table 2). At 6 months after randomization, the hysterectomy group experienced greater improvements than the medicine group in pelvic pain, urinary urgency, sensation of incomplete bladder emptying, and breast pain (all P < .05). In addition, there were trends toward greater improvement in pelvic or bladder pressure and low back pain that were not statistically significant. By 2 years after randomization, most differences in improvement other than bleeding had narrowed, and few were statistically significant. Those randomly assigned to hysterectomy experienced greater improvement in hot flushes and incomplete bladder emptying. Menopausal elevations of follicle-stimulating hormone (more than 30 mIU/mL) occurred in 6 participants in each group (19.4% for hysterectomy versus 18.8% for medical treatment), a difference that was not statistically significant. As expected, participants randomly assigned to medical treatments were more likely to report cyclic vaginal bleeding than those randomly assigned to hysterectomy at both 6 months (87% versus 11%, P < .001) and 24 months (37% versus 7%, P = .006) after randomization (crossovers contributed to these percentages, eg, continued bleeding in the 3 women randomly assigned to hysterectomy who did not receive the operation) despite substantial crossover to hysterectomy in those assigned to medical treatments. We did not find evidence of a difference in relative improvement in symptoms according to duration of symptoms.
Adjustment for site was performed by including site as a 4-level fixed effect in the regression models used to evaluate treatment effects on study outcomes. After adjustment, a difference favoring hysterectomy for 6-month improvement in pelvic/bladder pressure became statistically significant (P = .04) and a year 2 improvement in incomplete bladder emptying favoring hysterectomy was no longer statistically significant. All other intention-to-treat results were unaffected.
Differences in cumulative restricted activities were not statistically significant in the intention-to-treat analyses. Participants randomly assigned to hysterectomy missed a median of 21 days from work or usual activities (interquartile range 12–37), stayed in bed an additional 3 days (interquartile range, 2–8), and cut down on their usual activities an additional 0 days (interquartile range, 0–16.5). Participants randomly assigned to medicine missed a median of 14 days (interquartile range, 1–42), stayed in bed an additional 5 days (interquartile range, 1–9), and cut down on their usual activities an additional 2 days (interquartile range, 0–13).
Substantial symptom improvement occurred by year 2 within all 3 of the as-treated groups (Table 3). Participants who had a hysterectomy, whether by randomized assignment or after electing to discontinue medical treatment, experienced statistically significant improvements in nearly all symptoms assessed with the exception of incontinence, which did not change over time. Participants randomly assigned to medicine who remained on medical treatment had statistically significant improvements in pelvic pain, pelvic or bladder pressure, and stress incontinence symptoms.
Analyses comparing baseline to year 2 symptom changes in as-treated groups showed greater improvements among those who crossed over to hysterectomy than those remaining on medical treatments. The difference in improvement was statistically significant for pelvic pain (P < .01), low back pain (P = .02), urinary frequency (P = .01), urgency (P = .02), and breast pain (P = .01); there was a trend toward greater improvement in pelvic or bladder pressure (P = .08). Seventy-nine percent of women remaining on medical treatment, and none who crossed over to hysterectomy, continued to report cyclic vaginal bleeding at 24 months (P < .01).
Participants randomly assigned to medicine who underwent a hysterectomy reported more cumulative restricted activity days than those who did not crossover. Statistically significant differences were found in cumulative days missed from work or usual activities (median 40 versus 1.5 days, respectively, P < .001) and additional days staying in bed (median 7 versus 1.5 days, respectively, P = .008). The difference in days in which participants cut down on their usual activities was not statistically significant (median 8 versus 0 days, respectively, P = .094).
Baseline sociodemographic, quality-of-life, and clinical variables, as well as health beliefs and attitudes, created a pool of 48 potential predictors of crossover to hysterectomy in participants assigned to medical treatment. The 9 baseline characteristics with univariate P values less than .10 included bleeding characteristics (days per month, subjective amount, menstrual interval), presence or absence of pelvic pain, body mass index, psychological well-being, the Mental Component Summary score of the Medical Outcomes Study Short Form-36, symptom resolution ratings, and satisfaction with current health. Five predictors considered to be measuring conceptually unique domains were selected for inclusion in multivariable analyses based on their relative measurement properties and reproducibility. The Mental Component Summary score was selected over the psychological well-being score, and days of monthly bleeding was selected over other bleeding characteristics.
In an exploratory multivariable analysis that included the 5 selected predictors with univariate P values less than 0.10, crossover to hysterectomy was more likely among women with more days of bleeding per month (relative hazard 1.06 per day of bleeding, 95% confidence interval [CI] 1.03, 1.10; P < .001) and less likely among women with higher body mass index (relative hazard 0.92 per increment, 95% CI 0.86, 0.99; P = .019), higher symptom resolution scores at the end of the prerandomization medroxyprogesterone acetate treatment (relative hazard 0.97 per increment, 95% CI 0.94, 1.00, P = .037), and more favorable mental component subscore scores (relative hazard 0.95 per increment, 95% CI 0.91, 1.00, P = .047). The presence or absence of pelvic pain was not a statistically significant predictor in the multivariable model. Baseline health beliefs and attitudes, including those favoring medicines, uterine conservation, or importance of stopping bleeding, did not meet the inclusion criterion for entry into the model. The presence of uterine leiomyomata was also not found to predict hysterectomy.
Concern about the potential harms of hysterectomy has created enthusiasm for a variety of medical and surgical treatment alternatives. These harms include the irreversible loss of future fertility and the psychological impact of surgery and uterine removal, as well as postsurgical morbidity, very rare mortality,8 and possibly incontinence years later.9 Consequently, hysterectomy is often reserved for abnormal uterine bleeding that is refractory to treatment alternatives of lesser risk.
If an initial trial of medical treatment fails to control a patient’s symptoms, what is the probability that expanded efforts with a different regimen will succeed? The findings of the Medicine or Surgery trial support consideration of hysterectomy after an unsuccessful trial of oral progestins. In addition to cessation of vaginal bleeding, participants randomly assigned to hysterectomy experienced statistically significant improvements in pelvic pain, several urinary symptoms, and breast pain at 6 months compared with those randomly assigned to medical treatment. By 24 months, more than half of the medicine group had undergone a hysterectomy, obscuring most differences in intention-to-treat analyses. Analyses within as-treated groups at 2 years of follow-up revealed statistically significant improvements in several symptoms among women who remained on medications. However, there were improvements of greater magnitude, and improvement in more symptoms, among women who had a hysterectomy, whether by random assignment or after electing to discontinue medical treatment. Participants remaining on medications experienced fewer days of restricted activity than those who crossed over to hysterectomy.
The strengths of the Medicine or Surgery trial include its randomized design. Randomized trials comparing surgical to nonsurgical interventions are difficult to conduct, and crossover to surgical treatments is common.10 Postrandomization choice is a design characteristic of trials comparing medical and surgical strategies for heavy menstrual bleeding.11,12 A recent systematic review of medical versus surgical treatments for heavy menstrual bleeding identified only 5 randomized trials, of which all were conducted in European countries.13 Most of the trials compared the levonorgestrel intrauterine system to endometrial resection or destruction (3 trials) or hysterectomy (1 trial). The randomized comparison of levonorgestrel intrauterine system versus hysterectomy in Finland permitted hysterectomy to be performed by laparoscopic, vaginal, and abdominal routes. By 1 year after randomization, 20% of those randomized to levonorgestrel intrauterine system crossed over to hysterectomy.14 A randomized trial in Scotland compared transcervical resection of the endometrium with 3 cycles of medical treatment selected by the gynecologist from a wide range of choices (progestins, combined oral contraceptives, antifibrinolytics, prostaglandin inhibitors, androgenic steroids, and hormone replacement therapy). Crossover to surgical treatments was high at 2 years (59%) and 4 years (77%) after randomization.15 Although motivated by assuring study feasibility, the lack of standardization in type of hysterectomy and medical regimen in these studies, and in the Medicine or Surgery Trial, has the potential of obscuring differences that might be present if only 1 of the treatments is effective and all women randomly assigned to medicine receive that treatment.
A lack of blinding is unavoidable in trials of medical versus surgical treatments. Participants randomly assigned to hysterectomy did not ingest placebo tablets, and those randomly assigned to medical treatment did not undergo sham surgery. Most of our study outcomes were patient reports of symptoms. A difference in attitudes toward medical treatments or uterine conservation could set the stage for bias in the patient reports. In our prospective cohort of women potentially eligible for the Medicine or Surgery study, participants valuing uterine conservation were more likely to be satisfied with 3–5 months of cyclic medroxyprogesterone acetate treatment.16 However, baseline health beliefs and attitudes of women participating in the Medicine or Surgery trial did not differ by randomized group and did not predict crossover to hysterectomy in exploratory analyses.
The Medicine or Study trial examined multiple outcomes, and we did not account for multiple comparisons by use of the Bonferroni or similar adjustment to significance levels. Instead, we advise caution when interpreting findings of borderline statistical significance.
Several issues influence the generalizability of our findings. Ninety-five percent of our 63 subjects included women from 2 major race/ethnicity groups of diverse socioeconomic status from the southeastern United States. Our effectiveness approach allowed for variations in hysterectomy techniques and medical regimens that are inclusive of common practices. The chief challenge to generalizability comes from the road participants traveled before randomization and their subsequent willingness to be randomly assigned. Eligible women were required to complete a minimum 3-month trial of cyclic medroxyprogesterone acetate, attend a follow-up visit, indicate their dissatisfaction on a written survey, and accept assignment at random to medical alternatives or hysterectomy. Those accepting randomization expressed attitudes and beliefs endorsing both medications and hysterectomy and lacked strong biases precluding 1 treatment or the other. Thus, our findings may not be applicable to women who have not completed a medroxyprogesterone acetate trial or who have strong treatment preferences. In addition, two thirds of the women had leiomyomata, which may diminish the effectiveness of medical treatment, although this was not the case in the medroxyprogesterone acetate trial that preceded the Medicine or Surgery trial.16
The Medicine or Surgery trial contributes to a small-but-growing number of randomized comparisons addressing the optimal indications and timing of hysterectomy versus alternative treatments for noncancerous uterine conditions. The data from such trials are essential for optimally informed decision-making by patients and their health care providers. The clinical outcomes of the Medicine or Surgery trial converge with our quality-of-life and sexual functioning outcomes to form a clear picture. For symptom improvement in patients with abnormal uterine bleeding refractory to medroxyprogesterone acetate, hysterectomy is superior to continued efforts with oral medications at 6 months and in women assigned to medicine who elect to have a hysterectomy by 24 months. However, quality of life and sexual functioning improve to a statistically and clinically significant degree among all women in the 2 years of follow-up, regardless of whether or not they had a hysterectomy.3 We conclude that hysterectomy may be an optimal choice for women who give high priority to resolving bothersome symptoms, including bleeding, pelvic pain and pressure, low back pain, breast pain, and bladder symptoms. That 10% (3 of 28) of women receiving a hysterectomy by randomization required hospitalization up to 15 months later underscores the potential morbidity of the procedure. For women wishing to avoid hysterectomy, continued efforts at medical treatment may bring substantial improvements in pelvic pain, pelvic/bladder pressure and stress incontinence, and with fewer days of restricted activity.
1. Farquhar CM, Steiner CA. Hysterectomy rates in the United States 1990 to 1997. Obstet Gynecol 2002;99:229–34.
2. Carlson KJ, Miller BA, Fowler FJ Jr. The Maine Women’s Health Study: I. Outcomes of hysterectomy. Obstet Gynecol 1994;83:556–65.
3. Kupperman M, Varner RE, Summitt RL, Learman LA, Ireland C, Vittinghoff E, et al. The effect of hysterectomy versus medical treatment on health-related quality of life and sexual functioning. The Medicine or Surgery (Ms) randomized trial. JAMA 2004;291:1447–55.
4. Varner RE, Ireland CC, Summitt RL, Richter HE, Learman LA, Vittinghoff E, et al. Medicine or Surgery (Ms): a randomized clinical trial comparing hysterectomy and medical treatment in premenopausal women with abnormal uterine bleeding. Controlled Clinical Trials 2004;25:104–18.
5. McCulloch CE, Searle SR. Generalized, linear, and mixed models. New York (NY): John Wiley and Sons; 2001.
6. Cox DR, Oakes D. Analysis of survival data. London, UK: Chapman & Hall; 1984.
7. Ware J, Kosinski M, Keller S. SF-36 physical and mental health summary scales: a user’s manual. Boston (MA): The Health Institute, New England Medical Center; 1994.
8. Myers ER, Steege JF. Risk adjustment for complications of hysterectomy: limitations of routinely collected administrative data. Am J Obstet Gynecol 1999;181:567–75.
9. Brown JS, Sawaya G, Thom DH, Grady D. Hysterectomy and urinary incontinence: a systematic review. Lancet 2000;356:535–9.
10. Rolnick SJ, Flores SK, Fowler SE, Derman R, Davidson B. Conducting randomized, controlled trials: experience with the dysfunctional uterine bleeding intervention trial [review]. J Reprod Med 2001;46:1–5, discussion 5–6.
11. Cooper KG, Grant AM, Garratt AM. The impact of using a partially randomized patient preference design when evaluating alternative managements for heavy menstrual bleeding. Br J Obstet Gynecol 1997;104:1367–73.
12. Broder MS, Landow WJ, Goodwin SC, Brook RH, Sherbourne CD, Harris K. An agenda for research into uterine artery embolization: results of an expert panel conference [review]. J Vasc Interverv Radiol 2000;11:509–15.
13. Marjoribanks J, Lethaby A, Farquhar C. Surgery versus medical therapy for heavy menstrual bleeding (Cochrane Review). In: The Cochrane Library, Issue 2, 2003. Oxford: Update Software.
14. Hurskainen R, Teperi J, Rissanen P, Aalto AM, Grenman S, Kivela A, et al. Quality of life and cost-effectiveness of levonorgestrel-releasing intrauterine system versus hysterectomy for treatment of menorrhagia: a randomized trial. Lancet 2001;357:273–7.
15. Cooper KG, Jack SA, Parkin DE, Grant AM. Five-year follow up of women randomized to medical management or transcervical resection of the endometrium for heavy menstrual loss: clinical and quality of life outcomes. BJOG 2001;108:1222–8.
16. Richter HE, Learman LA, Varner RE, Hendrix SL, Summitt RL, Washington AE. Medroxyprogesterone acetate for treatment of abnormal uterine bleeding: factors predicting success. Am J Obstet Gynecol 2003;189:37–42.
Members of the Medicine or Surgery Research Group include the following:
University of Alabama, Birmingham:
R. E. Varner, MD (principal investigator); H. Richter, MD, PhD (co-investigator). Robert L. Holley, MD (co-investigator); Seine Chiang, MD (co-investigator); V. Bannister, RN (clinic coordinator); L. Abdo, RN, CCRC (clinic coordinator); Alison Myrich (research technician).
University of California, San Diego:
E. Barrett-Connor, MD (principal investigator); D. Goodman-Gruen, MD, PhD (co-investigator); A. Fleming, RN (clinic coordinator).
University of Tennessee, Memphis:
R. L. Summitt, Jr, MD (principal investigator); F. W. Ling, MD (co-investigator); Gary H. Lipscomb, MD (co-investigator); Marie Woodruff, RN, CCRC (clinic coordinator); D. Bran, RN CCRC (clinic coordinator); M. F. Braslow (senior research technician).
Wayne State University, Detroit:
S. G. McNeeley, MD (principal investigator); S. Hendrix, DO (co-investigator); D. Kmak, MD (co-investigator); Paul R. Makela, MD (co-investigator); J. Mowery, RN (clinic coordinator); M. McNamee, RN (research nurse).
University of California, San Francisco:
E. Washington, MD, MSc (study chair); S. B. Hulley, MD, MPH (principal investigator); L. Learman, MD, PhD (co-investigator, gynecology); M. Kuppermann, PhD, MPH (co-investigator, quality of life); J. Showstack, PhD (co-investigator, health economics); E. Vittinghoff, PhD (statistician); F. Lin, MS (statistical programmer); C. Ireland, MPH (project director); C. Gehrman (programmer).
Agency for HealthCare Research and Quality:
H. Hubbard, RN, MPH (project officer).
Data and Safety Monitoring Board:
C. Westhoff, MD MSc (chair); J. Wittes, PhD; M. Hlatky, MD (2000–2003); W. Applegate, MD (1997–2000).