Fibroid interventions: reducing symptoms today and tomorrow: extending generalizability by using a comprehensive cohort design with a randomized controlled trial.

Uterine fibroids (leiomyomas) are a common and burdensome disease of reproductive-aged women, yet quality evidence to inform treatment decisions is currently lacking ( 1 ). The few randomized controlled trials (RCTs) comparing fibroid therapies have largely been performed outside the United States ( 2 – 10 ). US-based RCTs studying therapies for the related problem of heavy menstrual bleeding have often experienced recruitment challenges ( 11 – 14 ). The Fibroid Interventions: Reducing Symptoms Today and Tomorrow (FIRSTT) study is an RCT ( NCT00995878 , clinicaltrials.gov ) funded by the National Institutes of Health to compare 2 minimally invasive therapies approved by the US Food and Drug Administration: uterine artery embolization (UAE) and magnetic resonance imaging (MRI)-guided focused ultrasound surgery (MRgFUS). Because of slow recruitment for the RCT, women meeting the FIRSTT inclusion criteria who declined randomization were offered enrollment in an abbreviated study protocol. Prior studies have demonstrated that such a comprehensive cohort design (CCD) can be useful because outcomes do not appear to differ when both groups receive the same treatments ( 15 , 16 ). The current report summarizes the baseline data for this trial and tests the hypothesis that contemporaneously recruited women meeting the same enrollment criteria who decline participation can be used in a CCD for fibroid treatment trials. Subsequent reports will use these data to compare safety, efficacy, economics, and ovarian reserve after treatment with UAE or MRgFUS.

Evaluating the comparative effectiveness of interventional therapies for treatment of uterine fibroids is an important area of research, and an RCT provides the highest level of evidence. However, RCTs for interventional fibroid therapies have been difficult to perform in United States populations; the FIRSTT study further demonstrates this challenge. Recruitment for the FIRSTT study also faced the challenge that despite receiving US Food and Drug Administration approval in 2004, many insurers continue to classify MRgFUS as an investigational therapy and deny coverage for treatment. Thus, the conundrum exists that insurers await high-quality evidence of efficacy for new treatments, while RCTs remain difficult to conduct because of lack of insurance coverage. Our study findings provide evidence that a CCD analysis would be appropriate for FIRSTT trial data. This type of design has been advocated in the past to increase the generalizability of RCT results and has been used in high-profile research studies, including the Women’s Health Initiative ( 15 , 16 , 28 , 29 ). In our study, the enrollment criteria that defined the RCT and PC1 groups resulted in similar participant characteristics, which suggests that the criteria, rather than the act of randomization, was the key variable. In addition, women not meeting these criteria who enrolled in the PC2 cohort were different in many important ways. The FIRSTT trial also demonstrates that having financial coverage for treatment in place substantially increases enrollment. The fact that all but 1 of the participants who chose not to undergo treatment were assigned to UAE suggests that women enrolling in the RCT were seeking access to MRgFUS, and the study funding of treatment appears to have been an enticement to enroll. However, given the overall similarity between treatment arms and the higher levels of baseline pain reported in the MRgFUS group, it does not appear that this coverage induced recruitment of women who reported fewer symptoms at baseline. Interestingly, women meeting the same enrollment criteria but deciding to choose their treatment selected UAE and MRgFUS in nearly equal numbers. This suggests that the critical distinction for fibroid studies could be the definition of the population based on enrollment criteria rather than the act of randomization. Going forward, assessing FIRSTT trial outcomes of only participants undergoing treatment appears to be more appropriate given the substantial number of withdrawals that differentially affected the UAE arm of the RCT. Although the statistical difference in pain scores is attenuated, adjustments for baseline pain scores should be made in future analyses. Other differences between arms also may affect outcomes. For example, women undergoing UAE were more likely to have only 1 or 2 fibroids of 3 cm or larger; this may be important because MRgFUS targets individual fibroids and UAE is a global uterine therapy ( 1 ). A clear limitation of the study, however, is that important unmeasured confounders may exist among women in the PC1 cohort that differ between arms and may affect outcomes. If there are additional confounders, this effect may be magnified because the sample size is relatively small. Another key limitation of this study was the failure to meet our enrollment targets. More specifically, our goal had been to recruit more women of African descent given the increased severity of disease in this population and amid evidence that these women have different patterns of disease and prefer uterine-sparing therapies ( 30 , 31 ). Other than this limitation, our cohort did reflect the population of women with symptomatic fibroids and significant impairment of quality of life, which is consistent with other fibroid studies as measured by standardized instruments. Women in the RCT also reflect the US population in terms of smoking status, with an approximate rate of 18% ( 32 ). Our finding that women in the RCT and PC1 groups were similar in several key features, including symptoms, objective measures of fibroid burden, and most baseline validated measures, was reassuring given the broad categories of inclusion. Conducting a CCD study with adjustment for baseline factors may prove to be a useful model for future studies of interventional fibroid therapies.

Across study sites, 723 women were prescreened and 568 (79%) were excluded. The remaining 155 women consented to study participation and were seen for an initial study visit ( Figure ). Of these, 57 women (37%) made up the RCT group. Uterine size was less than 700 cc 3 in 41 (72%) of the RCT participants. Among the 57 women, 49 (86%) underwent study treatment: 22 UAE and 27 MRgFUS ( Figure ). Of the 8 women randomized but not treated, 7 were assigned to the UAE group. A protocol violation occurred in 1 of these 7 participants: she was randomized to UAE but then was subsequently assigned to MRgFUS at a second site before the dual randomization was discovered. None of these 8 participants underwent either study treatment. A majority of the 27 women undergoing MRgFUS had treatment funded either by industry (8; 30%) or institutional (7; 26%) funding. Thirty-eight women (25%) met all enrollment criteria but declined participation; 34 of these women participated in the PC1 observational cohort: 18 underwent UAE and 16 underwent MRgFUS ( Figure ). Of the women who had consented, 60 (39%) did not meet the inclusion criteria for the RCT. Having an MRI finding that was an exclusion criterion was the most common reason for exclusion. Thirty-eight of the 60 women excluded were at the site participating in PC2, and 28 (74%) consented to be included in PC2 ( Figure ). Complete demographic and clinical information for the RCT cohort is shown in Table 1 . Women enrolled in the RCT were typical of women with uterine fibroids: mean (SD) age was 44.3 (4.7) years, body mass index was 28.7 (5.5) kg/m 2 , and 24% had experienced infertility. The RCT participants also tended to reflect the racial diversity of the United States, except for considerable underrepresentation of Hispanic women (5%). The majority of RCT participants were current alcohol users (82%) and regularly exercised (75%), whereas 18% were current smokers and 42% reported having more than 100 cigarettes in their lifetime. Median (IQR) uterine size was 563 (402–693) cc 3 ; 44% had 1 fibroid and 15% had 4 or more fibroids at least 3 cm in greatest diameter ( Table 1 ). Most women had multiple fibroid symptoms. Bulk symptoms were the most common dominant symptom reported ( Table 1 ). Medical comorbid conditions were relatively common; the most frequent conditions among 55 respondents were ovarian cysts (15%), depression (11%), fibrocystic breast disease (11%), hypertension (11%), endometriosis (9%), and hypothyroidism (9%). Other gynecologic conditions were rare and included adenomyosis (4%), endometrial polyps (2%), and uterine prolapse (2%). Among 56 participants responding, most had some type of prior surgery (79%) and reported taking regular medication (68%), but none had diabetes. Validated measures confirmed that women in the RCT had substantial fibroid symptoms, with a mean UFS-QOL SSS of 53.8 (19.5), and impaired quality of life, with a UFS-QOL total health-related quality of life subscale score of 50.8 (18.8) ( Table 1 ). General health status was also impaired compared with population means, as captured by the Short Form-36 score ( Table 1 ); the most impairment was seen on the energy/fatigue subscore (37.3 [23.7]). Pain as measured by the MPS (median [IQR], 9.0 [4.0–17.0]) and VAS (median [IQR], 35.5 [7.5–60.5]) was consistent with mild pain and was comparable to baseline pain in the previous uterine fibroid study using MPS ( 27 ). The CES-D mean score (19.5 [7.2]) suggested mild depression, and the mean FSFI score (19.5 [10.7]) suggested lower sexual function than in the general population. Women enrolled in PC1 were very similar to their RCT counterparts ( Table 1 ). They did, however, have a significantly lower mean body mass index ( P =.01), were less likely to have had a pregnancy ( P =.003), and were less likely to be current ( P =.048) or ever-smokers ( P =.003) ( Table 1 ). RCT and PC1 participants had no significant differences in fibroid characteristics, self-reported fibroid symptoms, median uterine volumes, or scores on any validated instruments ( Table 1 ). Women who enrolled in PC2 ( Supplemental Table 2 ) were different from women who met enrollment criteria for the other protocols ( Table 1 ). Compared with RCT and PC1 participants, PC2 participants had larger uteri (median, 670 vs 584 cc 3 ), were younger at the time of fibroid diagnosis (35.2 [9.3] vs 40.3 [6.9] years), and had a greater proportion of women with 4 or more measured fibroids (30% vs 14%). PC2 participants were also more likely to report infertility (50% vs 23%) but had similar validated measures ( Supplemental Table 2 , Table 1 ). Within each cohort (RCT and PC1), there were few differences between the UAE and MRgFUS arms ( Tables 2 and 3 ). When the 2 cohorts were combined, there were no significant differences in patient or fibroid characteristics between treatment arms ( Table 2 ). However, the median baseline MPS score was significantly higher in the MRgFUS arm (10.5 vs 6.0 [UAE]; P =.03); a similar but nonsignificant trend was seen in VAS scores (median, 39.0 vs 24.0; P =.06) ( Table 3 ). No correlation was seen between smoking and pain scores (data not shown). Because of the differential loss of participants between randomization and treatment in the RCT, an analysis of treated participants was also performed ( Supplemental Tables 3 and 4 ). Again, no significant differences were found between the MRgFUS and UAE arms in patient or fibroid characteristics ( Supplemental Table 3 ). Although not statistically significant, the same trend was seen in pain scores as in the intention-to-treat analysis, with patients in the MRgFUS arm reporting higher pain scores than patients in the UAE arm on the MPS (median, 10.0 vs 7.0; P =.08) and the VAS (median, 38.0 vs 25.0; P =.10) ( Supplemental Table 4 ).

The overview of the design of the FIRSTT study has been previously reported ( 17 ). Since the initial report, the University of California, San Francisco was added in June 2013 to the initial sites: Mayo Clinic, Rochester, Minnesota, and Duke University, Durham, North Carolina. The institutional review board of each site approved the same study protocol. Briefly, UAE and MRgFUS were performed according to the clinical standard of care, and treatment costs were designed to be paid by the participant’s health insurance. Insurance approval for both procedures was confirmed before randomization. Initially, funding was obtained for a 6-month study (RC1HD063312). Study visits occurred at baseline, day of study treatment, and 6 months after treatment. Telephone follow-up and review of study diaries took place 2, 4, and 6 weeks after treatment. Protocol modifications are outlined in Supplemental Table 1 . Full funding (R01HD060503) allowed for visits at 12, 24, and 36 months after treatment, which included MRI at 24 and 36 months, measurement of ovarian reserve at baseline, 12, 24, and 36 months, and economic analysis of treatments. Enrollment had been substantially limited because of lack of insurance coverage for MRgFUS, and industry funding was obtained to support treatment if insurance coverage was denied (InSightec, Tirat Carmel, Israel). Thus, women randomized to MRgFUS were allowed to proceed with treatment while insurance appeals took place. This backup payment method was closed at 1 site as of November 1, 2013, and the institution committed to cover these costs. Women in both treatment arms were responsible for copayments and deductibles. The primary efficacy end point for the 36-month study was additional intervention for symptomatic fibroids during the study period. Change in the symptom severity score (SSS) subscale of the Uterine Fibroid Symptom and Quality of Life instrument (UFS-QOL) was the key secondary efficacy end point ( 18 ). Safety was assessed by examining adverse events. Other specific aims of the study included assessing fibroid regression with MRI, assessing ovarian reserve by examining antimüllerian hormone levels, and conducting an economic analysis of treatment. Full inclusion and exclusion criteria at the study outset have been reported ( 15 ). In brief, participants were premenopausal women with symptomatic fibroids and uteri less than 20 gestational weeks in size who were not actively trying for pregnancy. All women underwent MRI with gadolinium contrast during screening. Uterine volume was calculated using the formula for the volume of a prolate ellipsoid, and the number of fibroids with maximal diameter of 3 cm or larger was recorded. Changes in enrollment criteria over time are outlined in Supplemental Table 1 . None of the previously treated participants would have been excluded by these protocol changes. At 2 of the 3 sites, a standard telephone screening instrument was used for prescreening, which sequentially identified exclusion criteria, including perimenopausal or postmenopausal status, women actively seeking pregnancy, prior fibroid interventional therapy, current use of a gonadotropin-releasing hormone analog, and medical contraindications to either study treatment. A study gynecologist subsequently screened each participant to exclude other causes of symptoms and to assess contraindications to either therapy, and MRI results were reviewed by a study radiologist for imaging enrollment criteria. Multiple modes of recruitment were used and were extended over time ( Supplemental Table 1 ). Two observational cohorts were established in 2011 for women who did not enroll in the RCT. Parallel cohort 1 (PC1) consisted of women who met the RCT enrollment criteria but declined participation. PC1 participants completed all study instruments and underwent assessment of ovarian reserve but did not have follow-up MRIs and received a smaller stipend. Parallel cohort 2 (PC2) included all women undergoing fibroid therapy. The protocol for PC2 involved obtaining limited baseline data, assessing ovarian reserve, and collecting economic costs; participants received a minimal stipend. Two sites participated in PC1, and 1 participated in PC2. Baseline questionnaires were administered at the participant’s initial visit; we attempted to schedule women in the early proliferative phase. The collected validated measures included the UFS-QOL ( 18 ), Short Form-36 (RAND Corporation) ( 19 ), the Center for Epidemiologic Studies Depression Scale (CES-D) ( 20 ), Female Sexual Function Index (FSFI) ( 21 ), McGill Pain Score (MPS) ( 22 ), and Visual Analog Scale for Pain (VAS). Detailed information on study instruments has been reported previously ( 17 ). The randomization scheme was stratified by study site and calculated uterine volume: smaller than 700 cc 3 and 700 cc 3 or larger. The randomization was performed using a web-based application created and supported by the Division of Biomedical Statistics and Informatics at Mayo Clinic, which used a dynamic allocation approach based on the Pocock-Simon method to achieve balance within each stratum ( 23 ). After randomization, treatment was typically scheduled within 10 days to minimize the chance of subject loss. Neither participants nor clinical investigators were blinded to study assignment. UAE and MRgFUS were performed using standardized protocols, and all participants received the same discharge instructions and prescriptions. The Data Safety Monitoring Board (DSMB) consisted of the study statistician (A.L.W.), Dr. Bradley Van Voorhis, a nationally recognized gynecologist and fibroid expert at the University of Iowa, and Dr. James Spies, a nationally recognized interventional radiologist at Georgetown University with expertise in UAE. The National Institutes of Health project officer was an ex officio member of the DSMB. An intention-to-treat analysis was conducted. The initial sample size calculations were conducted to detect differences between the 2 treatment arms for a) the need for additional interventional therapy for symptomatic fibroids over the course of the follow-up period, and b) the mean (SD) reduction (compared with baseline) in UFS-QOL SSS. Without published data on MRgFUS outcomes at 36 months at that time, calculations were based on published outcomes at 24 months: a) 20% and 37.5% of patients needing additional therapy after UAE and MRgFUS, respectively, and b) mean (SD) SSS decreases of 40.1 (25.2) and 28.1 (23.6) from baseline scores for UAE and MRgFUS, respectively ( 24 – 26 ). The study was designed to recruit 99 women per treatment arm, which provided statistical power of 78% and 93%, respectively, to detect the anticipated differences in outcomes a and b. These calculations were based on a 2-sided χ 2 test and t test with a type I error of 0.05. There were no a priori stopping rules. An interim analysis was conducted by the study statistician in February 2014 to assess the results and to determine whether study enrollment should be terminated early given the slow enrollment. Although too few participants had reached the 36-month point to assess the primary end point, statistically significant differences were observed between treatments for SSS over 24 months. The results were presented by blinded group assignment and reviewed by study investigators and the DSMB. The decision was made to end study enrollment as of August 1, 2014, to allow all participants to have at least 12 months of follow-up within the study. In performing the interim analysis, missed follow-up visits were identified at 1 site, and enrollment was closed at that site on March 18, 2014. Demographic and baseline characteristics were summarized using standard descriptive statistics: frequency (percentage) for categorical variables and mean (SD) or median (interquartile range [IQR]) for continuous variables. Comparisons between groups (RCT vs PC1 and MRgFUS vs UAE) were evaluated using the χ 2 test or Fisher’s exact test for categorical variables and the 2-sample t test or Wilcoxon rank sum test for continuous variables. All calculated P values were 2-sided, and P values less than .05 were considered statistically significant. Analyses were performed using SAS software version 9.3 (SAS Institute, Inc).