An official website of the United States government
NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.
Evan Myers, MD, Donna A. Messner, PhD, and Priscilla Velentgas, PhD.
Author Information and AffiliationsUterine fibroids are one of the leading causes of morbidity in reproductive-aged women. Because treatment options for symptomatic fibroids involve significant trade-offs, the lack of evidence on treatment effectiveness limits the ability of patients and their physicians to make informed treatment decisions.
The first objective of this study was to compare the durability of symptom relief after uterus-conserving treatments for symptomatic fibroids in terms of incidence of, and time to, new or recurrent symptoms and subsequent procedures. The second objective of this study was to evaluate the effect stakeholder participation had on the research process.
This was a retrospective database study of patients with uterine fibroids. The study used 2 data sources: Quintiles Electronic Medical Record System (Q-EMR) linked to Truven MarketScan claims and the COMparative effectiveness and PAtient Safety and Surveillance (COMPASS) Research Network. The study included patients with uterine fibroids having 1 of the procedures of interest occurring from January 1, 2005, to December 31, 2011, with the first instance being designated as the index procedure. The study then followed patients for a minimum of 2 years in the electronic medical record and claims data to look for new or recurrent symptoms and subsequent procedures. We created a Stakeholder Participation Council (SPC) with 17 diverse stakeholders (eg, patients/patient advocates, physicians, payers) to provide continuous input on the research process and interpretation and dissemination of the study findings.
The study included a total of 8687 patients in Q-EMR and 3547 patients in COMPASS. Due to the relatively small sample size of the COMPASS data, we based the comparative analyses on the Q-EMR data only. Endometrial ablation was the most common index procedure in Q-EMR. Patients who had uterine artery embolization as their index procedure had a lower risk of new or recurrent symptoms after the index procedure and a lower risk of having a subsequent procedure during the study period compared with patients who underwent endometrial ablation. Myomectomy patients had a lower risk of having a subsequent procedure than patients who received endometrial ablation.
This research provides additional evidence about expectations for the durability of symptom relief and time to subsequent procedures among women with uterine fibroids that may be helpful to women and their health care providers when making treatment choices. Overall, the SPC provided critical input that shaped the study design, data analysis, and dissemination plan for key findings.
Uterine leiomyomata, commonly referred to as fibroids, are one of the leading causes of morbidity in reproductive-aged women. These benign tumors of uterine smooth muscle and extracellular matrix have a cumulative incidence (based on imaging) by age 50 of almost 70% for White women and more than 80% for African American women.1
Although most fibroids are asymptomatic, a substantial proportion cause symptoms. The most common symptoms are heavy menstrual bleeding and/or pain, with a peak incidence for symptoms between ages 35 and 45 years.2 Studies using both global and disease-specific measures of health-related quality of life (HRQOL) show substantial impairment in HRQOL in women with fibroids.3,4
Symptomatic fibroids are by far the most common cause of hospitalization for benign gynecologic conditions in the United States.5 Fibroids may also contribute to adverse reproductive outcomes. Fibroids are associated with some cases of infertility,6,7 although this may be due in part to referral bias.8 Fibroids are common in pregnancy, with approximately 10% of women having detectable fibroids in the first trimester of pregnancy.9 They are also associated with such adverse pregnancy outcomes as miscarriage, preterm birth, and an increased risk of cesarean delivery,2,10,11 although, again, there is uncertainty about whether this association is causal.
Approximately 200 000 hysterectomies, 30 000 myomectomies, and thousands of selective uterine artery embolizations and high-intensity focused ultrasound procedures are performed annually in the United States to remove or destroy uterine fibroids, with the annual economic burden of these tumors estimated to be between $5.9 billion and $34.4 billion.12 Symptoms related to fibroids affect patients’ ability to work and result in substantial out-of-pocket costs. Employed women with symptomatic fibroids report significantly decreased work productivity.13 Nonmedical costs directly incurred by patients with uterine bleeding include not only those associated with lost work due to symptoms and medical visits, but also medications and supplies such as sanitary pads, which in one 2008 study averaged more than $300 per year.14
Despite the high burden of disease associated with fibroids, the quality of the evidence on optimal management is poor. A systematic review conducted in 2002 for the Agency for Healthcare Research and Quality (AHRQ) found almost no high-quality evidence for the comparative effectiveness of different treatment options for symptomatic fibroids.2 An update conducted in 2007 found no substantive changes in the state of the evidence and concluded,
The dearth of high-quality evidence supporting the effectiveness of most interventions for uterine fibroids is remarkable, given how common this problem is. The current state of the literature does not permit definitive conclusions about benefit, harm, or relative costs to help guide women's choices. Significant research gaps include well-conducted trials in US populations that directly compare interventions on short- and, especially, long-term outcomes, studies on therapeutics for medical management, and information on treatment decisions for women who desire a pregnancy.15
Treatment options for symptomatic fibroids involve significant trade-offs, including decisions about relative rates of short-term side effects, and, in the case of procedures, complications and recovery time, the likelihood of cure vs recurrent symptoms, and the effect on future childbearing potential. These trade-offs, coupled with the lack of evidence on outcomes of treatment, limit the ability of patients to make informed treatment decisions. There is considerable uncertainty about all of these outcomes for available treatments for fibroids. Additional research, particularly research on the relative effectiveness of available treatments, is needed.
In addition, significant disparities exist in disease burden and short-term treatment outcomes among fibroid patient subgroups. African American women consistently develop symptomatic fibroids at younger ages, have more severe disease (larger uterine size and number of fibroids at presentation), are more likely to have significant anemia compared with White women, and, in some but not all studies, were found to have more severe symptoms.1,2,9,16-18 African American women with uterine fibroids also tend to experience greater degrees of work impairment, presumably due to more severe symptoms.13 Hysterectomy for any indication is more common among reproductive-aged African American women compared with White women,19 as is myomectomy.2 These factors, along with higher prevalence of such comorbid conditions as obesity and hypertension, make surgical treatment more difficult and contribute to higher rates of complications in African American women for both hysterectomy and myomectomy.20-22
The premise that underlies comparative effectiveness research and patient-centered outcomes research is that research guided by patients and other stakeholders will substantively improve the quality and relevance of clinical research, such that it is more informative to patients and other decision makers. This concept is particularly important for uterine fibroids, for which patient preferences strongly influence treatment decisions. Retrospective observational studies represent 1 approach to studying the effectiveness of treatment options for uterine fibroids. However, little is known about the degree to which stakeholder participation will markedly change the study design or analysis plan for these types of studies.
The long-term objective of this research study is to help patients and care providers make informed decisions about which treatment options for uterine fibroids are most likely to result in the outcomes that are of greatest importance to each patient. To help achieve this long-term objective, this study has the following specific aims, taking advantage of unique data resources that combine administrative and electronic medical record (EMR) data:
To compare the durability of symptom relief after uterus-conserving treatments (myomectomy, endometrial ablation, uterine artery embolization, magnetic resonance imaging [MRI]-guided focused ultrasound ablation) for symptomatic fibroids in terms of:
To compare the durability of symptom relief after uterus-conserving treatments for symptomatic fibroids in terms of:
To evaluate the effect of stakeholder participation in the research process in terms of qualitative and semiquantitative evaluation of the process by investigators and stakeholders.
This observational study includes 2 components: (1) a retrospective cohort study of 2 independent data sources to study the effectiveness of treatment options for uterine fibroids; and (2) a stakeholder engagement plan to evaluate the contributions of a stakeholder committee in influencing the analysis and dissemination plans for this study.
In accordance with PCORI's authorizing legislation, all funded research should be assessed for adherence to PCORI's Methodology Standards27 as part of the peer review process. The investigators hereby identify the following Methodology Standards applicable to the research and confirm the research is adherent to these standards: RQ1 through RQ6; PC1 through PC4; IR1 through IR3, IR5 and IR6; MD1 through MD5; DR1 through DR3; and DN1 and DN2.
This was a retrospective cohort study of uterine fibroid patients to compare the durability of symptom relief following 4 uterus-conserving procedures of interest, including myomectomy, endometrial ablation, uterine artery embolization, and MRI-guided focused ultrasound ablation. The study separately analyzed two independent data sources: Quintiles Electronic Medical Records (Q-EMR) linked with Truven MarketScan claims, and the COMparative effectiveness and PAtient Safety and Surveillance (COMPASS) Research Network. Both data sources include geographically and demographically diverse patient populations and are described in detail below. Use of 2 independent sources of data (Q-EMR and COMPASS) was planned to serve as a validation of the findings from each source, with similar results in the 2 data sources (if observed), strengthening the conclusions and suggesting some generalizability of the findings across health systems and data sources.
Q-EMR is a national network of outpatient offices whose providers allow their deidentified patient-level data to be made available for research. The study captured clinical data from more than 525-member institutions and more than 30 000 providers. The study then linked the Q-EMR data to the Truven MarketScan claims data using an algorithm based on specific data elements (gender, month/year of birth, and 3-digit area codes) and health care service dates.23
The COMPASS Research Network is a network of integrated delivery networks (IDNs) organized to support the conduct of research using EMRs and other linked data sources. Data from these IDNs can be extracted and harmonized using a post hoc common data model to better understand characteristics of and differences between patient populations and changes in treatment over time, and to draw comparisons between geographic regions.
The patient population for the retrospective cohort study was uterine fibroid patients who had a record of at least 1 of the procedures of interest occurring during the period January 1, 2005, to December 31, 2011, with the first instance being designated as the index date. The study period spanned January 1, 2004, to December 31, 2013, to allow for 12 months preindex for evaluating baseline demographic and clinical characteristics and 24 months postindex, ensuring a minimum of 2 years’ follow-up. In Q-EMR linked with Truven MarketScan claims data, we considered patients as being continually followed, provided there was no more than a 30-day gap in their enrollment. In COMPASS, patients were required to have at least 1 office visit during the 12-month preindex period. We defined follow-up duration by encounters within prespecified time windows including 24 months postindex procedure, 24 to 36 months postindex procedure, 36 to 48 months postindex procedure, and 48 to 60 months postindex procedure. We captured diagnosis dates, encounter dates, procedure dates, lab dates, and medication start dates to provide detailed information about patient contact during study follow-up (see figure below).
Approximately 8680 patients with a diagnostic code for uterine fibroids and a procedure of interest during the period January 1, 2004, to June 30, 2012, were eligible for analysis from the linked Q-EMR and claims. The final number of patients included in the analysis was 8687. For the purposes of the present analysis, we used 3 COMPASS sites representing approximately 4000 women with a diagnostic code for uterine fibroids and a procedure of interest. The final number of patients included in the COMPASS analysis was 3547. All data were obtained from patients’ EMR and claims data.
Specific inclusion criteria were as follows:
Subjects meeting any of the following exclusion criteria were not included in the study:
We extracted and used patient demographics and relevant medical history available in the linked Q-EMR and claims and COMPASS data to create 2 separate cohorts to describe patient characteristics and durability of symptom relief after procedures of interest. We obtained data from structured fields in the EMR and claims, using ICD9 codes for symptoms and diagnoses and HCPC and ICD9 for procedures. We evaluated potential predictors of outcomes including patient demographics (ie, age, race/ethnicity, region), comorbid medical conditions, and symptoms (ie, bleeding, pain) at the time of the index procedure. Note that coding definitions for the comorbid medical conditions and symptoms, as well as the medical procedures, can be found in Appendix A.
Descriptive measures computed as described include the following:
Age
Race (White, Black, other)
Region (Northeast, Midwest, Southeast, West)
Hypertension
Type 2 diabetes
Disorders of lipid metabolism
Obesity
Coronary artery disease
Heart failure
Myocardial infarction (MI)
Stroke
Cardiac dysrhythmias
Menorrhagia/anemia
Dysmenorrhea
Pelvic pain not related to menstrual cycle
Other vaginal/pelvic symptoms (leucorrhea, vaginal discharge, dyspareunia)
Urinary symptoms (dysuria, urinary retention, hydronephrosis)
Bowel symptoms (constipation, diarrhea)
Other symptoms (leg pain, back pain, shortness of breath)
Myomectomy
Endometrial ablation
Uterine artery embolization
MRI-guided focused ultrasound ablation
Hysterectomy
The main exposure of interest was a uterus-sparing procedure for uterine fibroid treatment as an index procedure including myomectomy, endometrial ablation, uterine artery embolization, or MRI-guided focused ultrasound. We did not include hysterectomy as an index procedure in comparative analyses across procedures as symptoms because uterine fibroids and subsequent procedures for uterine fibroids cannot be experienced/occur posthysterectomy.
We evaluated durability of symptom relief after uterus-conserving treatments (myomectomy, endometrial ablation, and uterine artery embolization) for symptomatic fibroids using several end points described below. New or recurrent symptoms included only those recorded later than 60 days postprocedure, to eliminate those that may have been postoperative. Subsequent procedures also excluded new procedures recorded less than 60 days after the index procedure, to eliminate what may have been duplicate coding of the index procedure or other misleading coding, with the exception of hysterectomy as a subsequent procedure, based on input from the Stakeholder Partnership Council (SPC).
The following are the specific outcomes assessed in this study:
We maintained and analyzed data obtained from linked Q-EMR and claims and COMPASS separately in SAS.
We described demographic and other baseline characteristics of the 2 cohorts as counts and percentages for categorical variables and measures of central tendency (mean, SD, median, interquartile range) for continuous variables.
We used multivariable logistic regression models to (1) compare the incidence of recurrent or new symptoms, and (2) compare the incidence of subsequent procedures following myomectomy, endometrial ablation, and uterine artery embolization, and (3) to calculate the odds ratios (ORs), controlling for potential confounding variables (race, region, index procedure, symptoms at baseline, comorbidities at baseline, and age) as covariates in the multivariable models. We used Cox proportional hazard models to compare time to new or recurrent symptoms and time to subsequent procedures following the myomectomy, endometrial ablation, and uterine artery embolization, and to calculate the hazard ratios (HRs), controlling for potential confounding variables (race, region, index procedure, symptoms at baseline, comorbidities at baseline, and age) as covariates in the multivariable models. We created Kaplan-Meier curves to graphically present time to first new or recurrent symptom or first subsequent procedure.
To ensure equal follow-up time for study patients, we included only 2 years of follow-up after the index procedure in the logistic regression modeling, regardless of the total follow-up time available for individual patients. We included all available follow-up data in the time-to-event analyses, namely the Cox proportional hazards modeling.
For analyses examining the incidence and time to subsequent procedures, we excluded procedures coded as having occurred within 60 days of the index procedure (except for hysterectomy, which was included regardless of when the procedure occurred). We assumed that these procedures were likely related to coding issues because, clinically, we would not expect to see a subsequent procedure (with the exception of hysterectomy) occurring so soon after the index treatment. Similarly, for analyses examining the incidence and time to new or recurrent symptoms, we did not include symptoms coded as having occurred within 60 days of the index procedure, because many of these symptoms immediately following the index procedure were presumed to be due to the procedure itself (eg, postoperative pain and discomfort) and less likely to be reflective of symptomatic uterine fibroids.
We included the following variables (with referent group asterisked) in the multivariate models (ie, logistic and Cox proportional hazard models):
Symptoms groups included
Comorbidity groups included
We assembled an SPC made up of 17 individuals representing stakeholders including patients and consumers, clinicians, insurers and/or payers, federal agencies, and researchers. We invited stakeholders to participate as members of the SPC based on their experience as experts in the clinical management and/or epidemiology of uterine fibroids, their experience as a uterine fibroids patient and/or their previous patient advocacy activities, or their experience as a stakeholder in 1 of the other areas mentioned.
The specific aims of stakeholder engagement for this project were 4-fold:
the outcomes being measured are relevant to decision makers
the patient population includes important subpopulations of interest to decision makers
the comparators and their definitions are of importance to decision makers, and
other factors relevant to the analysis or interpretation of the data are considered
We engaged the SPC in a series of discussions to gather its feedback on this study. We asked the SPC to provide input on the key research questions that should be addressed by the study and the details of the implementation of the research, such as which variables should be included and how they should be defined, which subpopulations would be of interest, how to define meaningful measures that reflect clinical management of uterine fibroids based on EMR and claims data, how to interpret the study results, what findings would be the most meaningful to different stakeholders, and what means would be appropriate to disseminate the results.
We asked stakeholders to provide informed consent before participating in the research activities, given that their participation included a research component of evaluating the effectiveness of stakeholder input and the process of obtaining it. We sent the informed consent form to stakeholders by email before the first face-to-face meeting. Stakeholders had the option to withdraw their consent and discontinue participation in the SPC at any time. However, no stakeholders withdrew consent.
In accordance with these specific aims for stakeholder engagement, the first face-to-face meeting in June 2014 focused on obtaining stakeholder input to revise a draft version of the study protocol. Subsequent Web meetings and teleconferences provided additional opportunities for the stakeholders to review and help the research team interpret interim results. Stakeholders were able to review preliminary output from the analysis of the first available data source 6 months into the project; they were then asked to provide further input into the refinement of the study protocol before we began our analysis of the second data source. At the second face-to-face meeting in July 2015, the stakeholder group assisted with framing messages that are important to patients, providers, and policymakers, based on the study results and planning dissemination strategies, to ensure information was presented in sources and formats accessible to these decision makers.
A final 2-hour teleconference meeting of stakeholders in November 2015 was devoted to an evaluation of stakeholder experience. In addition to presenting the formal evaluation of the impact of stakeholder refinement on the project's protocol and the outcomes, we requested that participants complete an online evaluation form and provide more detailed individual feedback through comment forms, which we discussed at this meeting.
As previously described, the SPC consisted of 17 individuals. Each stakeholder's affiliations and areas of representation are summarized in Appendix B.
We qualitatively evaluated the effect of stakeholder participation in the research for the 4 aims: effect on the protocol, analyses, dissemination strategy, and overall “success” of engagement as evaluated via stakeholder interview. Stakeholder comments on both the protocol and the analyses are summarized in this report, along with stakeholder recommendations for dissemination. Finally, key findings from the interviews with SPC members are also described below.
There were 8687 patients in the Q-EMR cohort and 3547 patients in the COMPASS cohort (Table 1). As Table 1 presents, hysterectomy was the most common index procedure identified in the data (70% in Q-EMR, 78% in COMPASS), followed by endometrial ablation (15% in Q-EMR, 15% in COMPASS), myomectomy (12% in Q-EMR, 7% in COMPASS), uterine artery embolization (3% in Q-EMR, 1% in COMPASS), and MRI-guided ultrasound (<0.1% in Q-EMR, 0 in COMPASS). Because so few patients were receiving MRI-guided ultrasound as the index procedure, we only performed the comparative analyses across the other uterus-conserving treatments.
Index Procedure, Q-EMR and COMPASS Databases.
The distribution of age was similar between Q-EMR and COMPASS, with patients from both data sources having a mean age of 44 years (Tables 2a and 2b). Patients in COMPASS who underwent a myomectomy were on average younger (40 years) than those patients who underwent other procedures (44 and 45 years for endometrial ablation and uterine artery embolization, respectively). These results are consistent with the Q-EMR findings.
Patient Demographic Characteristics, Q-EMR Database.
Patient Demographic Characteristics, COMPASS Database.
The majority of patients from Q-EMR were from the Southeast (58%), followed by the Northeast (17%) and Midwest (15%), with fewer than 10% living in the West (Table 2a). Patients from COMPASS had a different distribution, with the largest proportion of patients living in the Midwest (47%), followed by the West (46%), and fewer than 10% living in the Northeast or Southeast (Table 2b).
The distribution of procedures across geographic regions was similar between the 2 data sets (see Tables 2a and 2b). In both data sets, hysterectomy represented the most common index procedure, particularly in the Midwest region in COMPASS (86%; see Figure 1).
Index Procedure by Region, Q-EMR, and COMPASS Comparison.
Race was missing for 27% of patients in Q-EMR and 51% of patients in COMPASS, as is often observed in EMR and claims data sources. Among those patients with known race, the majority were White (Q-EMR: 67%; COMPASS: 76%), followed by Black or African American (Q-EMR: 28%; COMPASS: 17%; see Tables 2a and 2b). Figure 2 illustrates findings across both data sets with respect to race. White women were more likely than African American women to undergo endometrial ablation in both data sources.
Index Procedure by Race, Q-EMR, and COMPASS Comparison.
After the initial descriptive analysis of the 2 data sets, we performed further analyses using only the Q-EMR data owing to the substantially larger sample size available and on the consensus recommendation of the SPC. All patients were required to have a minimum of 2 years of follow-up data. Among patients in the Q-EMR cohort included in the analysis, the mean and median follow-up time was 53 months and 48 months, respectively (minimum: 24 months; maximum: 111 months). For both study outcomes, the sample size of patients with complete data for race and other covariates for the logistic regression analyses and Cox proportional hazards analyses in Q-EMR was 1869.
Of the 3547 patients in the COMPASS cohort, the majority underwent hysterectomy as their index procedure (n = 2773; 78%). Of the patients with uterus-conserving procedures (n = 774), race was missing for 66%, resulting in a sample size of 262 patients with uterus-conserving index procedures who had complete data for all covariates. Due to this small sample size, we did not perform regression analyses for COMPASS patients.
Of women who had endometrial ablation as their index procedure, 62% sought medical care for 1 or more symptoms of uterine fibroids within 2 years, compared with 57% of women who had myomectomy and 54% who had uterine artery embolization as index procedures. The time to new or recurrent symptoms among women who did experience such symptoms within 2 years was comparable across the different procedures: endometrial ablation (median time to new or recurrent symptoms, 223 days), myomectomy (median, 240 days), and uterine artery embolization (median, 225 days; see Table 3).
Frequency and Time to Recurrent or New Symptoms and New Procedures Following Initial Uterus-Sparing Procedure, Q-EMR Database.
Results of a multivariable logistic regression analysis, presented in Table 4, indicated that African American patients had higher odds of seeking medical care for new or recurrent uterine fibroid symptoms after an index procedure (OR, 1.28; 95% confidence interval [CI]: 1.02, 1.60) than White patients. There were no significant differences in odds of new or recurrent symptoms across the procedure groups (myomectomy vs endometrial ablation, OR, 1.03; 95% CI, 0.82-1.30; uterine artery embolization vs endometrial ablation, OR, 0.90; 95% CI, 0.66-1.24).
Incidence of Recurrent or New Symptoms Within 2 Years After Initial Uterus-Sparing Procedure, Q-EMR (N = 1869).
Patients with any high-risk comorbidity (OR, 1.80; 95% CI, 1.10-2.97), any other non–high-risk comorbidity (OR, 1.53; 95% CI, 1.20-1.95), or both obesity and any other non–high-risk comorbidity (OR, 1.64; 95% CI, 1.17-2.31) had higher incidence of new or recurrent symptoms compared with patients with no symptoms reported after the index procedure (Table 4). Patients with symptoms in 1 symptom group (OR, 1.72; 95% CI, 1.27-2.34) or 2 or 3 symptom groups at study entry (OR, 2.91; 95% CI, 2.13-3.97) also had higher incidence of new or recurrent symptoms when compared with patients with no recorded symptoms for which they sought medical care at study entry (Table 4).
Results of multivariable Cox proportional hazards regression analysis, presented in Table 5 and Figure 3, showed that African American patients had a higher risk of seeking medical care for uterine fibroid symptoms (HR, 1.20; 95% CI, 1.06-1.37) than White patients. There was no statistically significant difference in risk of seeking medical care for new or recurrent symptoms between myomectomy patients (HR, 0.92; 95% CI, 0.81-1.05) or uterine artery embolization patients (HR, 0.88; 95% CI, 0.73-1.07) and endometrial ablation patients.
Time to Recurrent or New Symptoms With Full Study Period Following Initial Uterus-Sparing Procedure, Q-EMR (N = 1869).
Time to New or Recurrent Symptoms After First 60 Days Postindex Procedure, Kaplan-Meier Curve.
Patients with any high-risk comorbidity (HR, 1.41; 95% CI, 1.09-1.82), any other non–high-risk comorbidity (HR, 1.30; 95% CI, 1.14-1.49), or both obesity and any other non–high-risk comorbidity (HR, 1.39; 95% CI, 1.16-1.67) had a higher risk of new or recurrent symptoms compared with patients with no recorded symptoms for which they sought medical care at study entry. Patients with symptoms in 1 baseline group (HR, 1.44; 95% CI, 1.18-1.75) or 2 or 3 symptom groups at study entry (HR, 2.04; 95% CI, 1.68-2.47) were at a higher risk for recurrent or new symptoms compared with patients with no recorded symptoms at study entry. Patients in the Midwest (HR, 0.74; 95% CI, 0.61-0.90) and Southeast (HR, 0.76; 95% CI, 0.66-0.87) had a lower risk of new or recurrent symptoms compared with patients in the Northeast. Those who underwent a uterine artery embolization or myomectomy as the index procedure had lower risk, albeit not significant, than patients who had endometrial ablation as the index procedure (HR, 0.88; 95% CI, 0.73-1.07; and HR, 0.92; 95% CI, 0.81-1.05, respectively).
Of women who had an endometrial ablation as the index procedure, 18% had a subsequent procedure for treatment of uterine fibroid disease within 2 years of the initial procedure, compared with 13% of women who had myomectomy and 11% who had uterine artery embolization as index procedures. When comparing all available follow-up time, the median and mean times to subsequent procedure were shorter among women who had a myomectomy (mean 307 days, median 298 days) compared with women who underwent uterine artery embolization (mean 416 days, median 477 days) and with those who underwent endometrial ablation (mean 326 days, median 307 days; see Table 3).
Results of a multivariable logistic regression analysis indicated that African American and White patients had similar rates of subsequent procedures within 2 years following their index procedure (OR, 1.13; 95% CI, 0.84-1.52; see Table 6). Patients with myomectomy as the index procedure (OR, 0.68; 95% CI, 0.50-0.92), and patients with uterine artery embolization as the index procedure (OR, 0.59; 95% CI, 0.38-0.94) had lower odds of having a subsequent procedure within 2 years of their index procedure than those with endometrial ablation. Patients with symptoms in 1 symptom group (OR, 1.57; 95% CI, 0.98-2.52) or 2 or 3 symptom groups at study entry (OR, 1.56; 95% CI, 0.98-2.51) had higher odds of having a subsequent procedure within 2 years compared with those with no recorded symptoms for which they sought medical care at study entry, although these results were not statistically significant (Table 6).
Incidence of New Procedures Within 2 Years After Initial Uterus-Sparing Procedure, Q-EMR Database (N = 1869).
Results of a multivariable Cox proportional hazards regression analysis indicated that African American patients underwent subsequent procedures during the study period about as often as White patients (HR, 1.12; 95% CI, 0.91-1.37; see Table 7). Patients who underwent either a myomectomy (HR, 0.74; 95% CI, 0.60-0.92) or uterine artery embolization (HR, 0.61; 95% CI, 0.43-0.86) as their index procedure were at reduced risk of having a subsequent procedure during the study period compared with patients who had endometrial ablation (Table 7, Figure 4).
Time to Subsequent Procedure Within Full Study Period Following Initial Uterus-Sparing Procedure, Q-EMR Database (N = 1869).
Time to Subsequent Procedure After First 60 Days Postindex Procedure, Kaplan-Meier Curve.
We conducted several sensitivity analyses to assess the robustness of the analyses to different assumptions.
First, we chose to exclude symptoms and procedures occurring within 60 days of the index procedure based on SPC input that such symptoms or procedures were likely to be indistinguishable from postoperative symptoms or from the index procedure, or erroneous rather than representative of the true recurrence of uterine fibroid symptoms or new procedures; however, the choice of time period was somewhat subjective. We selected alternative time windows of 30 days postindex and no exclusion period and observed similar associations.
Second, we assessed multiple end points related to durability of uterine fibroid treatments, including onset of new or recurrent symptoms and subsequent procedures, both of which we examined over different time frames (2 years of follow-up and the full available follow-up period extending as far as 9 years), using different analytic approaches (unadjusted comparisons of the proportion of patients in each treatment group with the end point of interest, comparison of Kaplan-Meier curves, multivariate logistic regression and Cox regression analyses). We saw similar direction of effect with regard to treatment benefit in each analysis favoring uterine artery embolization compared with endometrial ablation, with myomectomy in most cases appearing to be superior to endometrial ablation, although the magnitude of effect varied across analyses.
After review of the protocol developed by the investigators, the SPC recommended 15 changes, including 1 on duration of follow-up, 3 on inclusion and exclusion criteria, 7 on patient characteristics, 2 on comparators, 1 on outcomes, and 1 on subgroup analyses. Of the 15 recommended changes, 10 were feasible for implementation in the protocol given the proposed study design (see summary in Table 8 and full list of recommendations in Appendix C). The Investigators considered the remaining 5 recommendations but deemed them either not feasible due to limitations with existing data or beyond the scope and focus of the current study.
Number of Suggestions from the Stakeholder Partnership Council Adopted for Protocol.
After review of the preliminary analyses conducted by the investigators, the SPC provided 18 recommendations about the statistical analyses and presentation of the data: 2 general, 6 on treatment patterns, and 10 on symptoms/additional procedures. After discussion, the investigators implemented 8 of those recommendations (see Table 9 and Appendix D). Of those suggestions that were not implemented, they were either not scientifically advisable given the data limitations or sample size, or were alternative ways of displaying the data that will continue to be evaluated during publication of these results but do not require additional analyses.
Number of Suggestions From the Stakeholder Partnership Council Adopted for Analysis Plan.
The SPC led the discussion of the dissemination plan. Regarding dissemination of results to patients, the SPC recommended building relationships with patient advocate networks, trusted and popular radio shows, magazines, and television talk shows to relay study findings and explain how these findings can be used by patients to support their treatment decisions. Regarding dissemination of results to clinicians, the SPC recommended connecting with clinicians at conferences and through peer reviewed journals.
There was general agreement that a sincere and good effort was made to engage and be responsive to the SPC members’ perspectives. Stakeholders were in unanimous agreement about the value of patient advocate involvement, and 1 person remarked on the benefit of engaging multiple patient advocates. Stakeholders also appreciated the diversity of categories represented within the SPC, and the opportunities to work with and learn from one another. Some of the participants expressed that they would have appreciated more opportunity to interact with the core research team. Many of the stakeholders noted the limitation of the retrospective data; they felt that important questions could not be addressed in the context of this study design.
The majority of patients underwent hysterectomy as their index procedures in both Q-EMR and COMPASS, while endometrial ablation was the most common uterus-conserving index procedure. Overall, distribution of index procedures was similar across both data sources.
When we analyzed treatment patterns by race, as recommended by the SPC, we found that overall, African American women underwent myomectomy more frequently than White women. When we considered both race and region—another SPC recommendation—we found the percentage of women undergoing myomectomies and endometrial ablation differed slightly by region, with a higher proportion of White women undergoing endometrial ablation in the Midwest, and more African American women undergoing endometrial ablation in the Southeast than White women (data not shown).
African American women had a 20% greater risk of seeking medical care for new or recurrent uterine fibroid symptoms during the study period than White women. Regarding index procedure, women who underwent myomectomy had an 8% lower risk, though not statistically significant, of seeking medical care for new or recurrent symptoms during the entire study period than those who underwent endometrial ablation, as did women who underwent uterine artery embolization (12% lower, not statistically significant). Factors statistically significantly associated with risk of new or recurrent symptoms postindex procedure included a history of at least 1 comorbidity and having more symptoms recorded in the medical record at study entry. For the most part, logistic regression analyses that examined the likelihood of having any symptoms within 2 years of index procedure and Cox regression analyses that examined the time to seeking medical care for any symptoms during the full study follow-up period yielded similar results.
Women who underwent a myomectomy or uterine artery embolization had 26% and 39% reduction, respectively, in risk of subsequent procedure during the study period compared with those who underwent endometrial ablation. The risk of subsequent procedure within 2 years of the index procedure was not statistically different between African American and White patients (13% higher risk among African American patients, although, again, not statistically significant). Having more uterine fibroid symptoms recorded in the medical record preprocedure was associated with a 57% increase in odds of having a subsequent procedure within 2 years and a 22% increase during the full study period compared with women without any preprocedure symptoms; however, this association was not statistically significant.
We designed and analyzed the study in a manner to minimize the likelihood that bias would be an alternative explanation for the observed results. However, there are inherent limitations to using EMR and administrative data. Potential sources of bias are described below.
In this study we analyzed secondary data, which were not collected for research purposes. This may have introduced information bias due to inconsistencies in coding practices, as well as coding that reflected “rule out” diagnoses or “upcoding” practice related to maximizing reimbursement. Also, substantial data were missing for key variables of interest, especially race (a common problem encountered in analyses of EMR and claims data sources), which may or may not be missing at random. As it is not anticipated that missing race or possible “upcoding” of diagnoses would be related to which specific uterine-sparing procedure patients received at baseline, the direction of impact of this bias on the observed associations would be expected to be toward the null; however, this cannot be directly verified.
The underlying cause of symptoms identified in this study could not be determined in either Q-EMR or COMPASS, as would be typical of most EMR data. This may have resulted in inclusion of some symptoms assumed to be uterine fibroid related but that are due to other causes. Also, only symptoms reported to and entered by the health care provider in the EMR were captured in this study. Thus, a patient was likely to have experienced additional symptoms that were not captured in the EMR. Misclassification of symptoms at baseline as well as subsequent symptoms as an outcome would not be expected to be associated with initial procedure, and thus would most likely be expected to bias associations involving either the baseline symptoms measure or the subsequent symptoms outcome measure toward the null.
Some important potential confounders, including such markers for disease severity as number, size, location of fibroids, knowledge of a woman's childbearing goals, and other reasons for choosing one procedure over another for treatment of uterine fibroids were not available within the EMRs. This unmeasured confounding may have led to overestimates or underestimates of the true associations of specific uterine-sparing procedures with onset of new symptoms and time to subsequent procedures, if reasons underlying the choice of initial procedure also influenced willingness to undergo an additional procedure, or if specific procedures were more likely to be used in women with more severe disease burden who also had a greater need to consider undergoing subsequent procedures (confounding by indication).
We used statistical methods including multivariable regression methods to account for measured confounders in analyses that compared different treatments for uterine fibroids. We had originally planned to perform subgroup analyses defined by race and other important designations (eg, age), in part to account for potential confounding but also to explore the comparative benefits of the different procedures within these subgroups. However, the sample sizes for the subgroups were relatively small and not sufficient for performing multivariable regression analyses within subgroups. Tests of interaction with race also would have been underpowered due to the small size of the subgroups. This is particularly true for race given the missing data for this variable. Although we did not conduct subgroup analyses, we included race and other important demographic and clinical covariates in the regression models and we provided discussion about differences observed among the groups for these variables (eg, differences found among the racial groups).
We encourage researchers in future studies with sufficient sample size to explore the comparative effectiveness of uterine fibroid therapies in clinically meaningful subgroups.
While the study made specific efforts to ensure regional representativeness and racial diversity through selection of the linked Q-EMR and claims and COMPASS data sources, the patterns of care, including frequency of specific procedures for treatment of uterine fibroids and characteristics of the patient population within these networks may not be fully generalizable to the US population. In addition, only patients who sought medical treatment for their uterine fibroids were reflected in these data sources and only commercially insured patients were available in the linked Q-EMR and claims data. Finally, we focused solely on surgical interventions for uterine fibroids; we did not consider medicinal treatments. Future studies involving comparative analyses of both surgical and medicinal treatments for this condition would clearly be of value.
Use of COMPASS as a second independent data source served as some additional validation of the findings from the Q-EMR data source, but it also identified some additional variation that may reflect regional care patterns. For example, while overall distribution of procedures for treatment of uterine fibroids was relatively similar between the 2 data sources, Q-EMR data reflected a higher prevalence of myomectomy and uterine artery embolization, potentially due to differences in the underlying populations as Q-EMR data focus primarily on outpatient care while COMPASS data also include inpatient data. Further, Q-EMR data include more family practices and fewer obstetrics and gynecology practices than COMPASS, which may underlie the different frequencies of specific procedures. Finally, Q-EMR data include those primarily insured through the commercial market, while COMPASS data also include those insured by Medicaid and those not insured.
Missing information is a common issue with EMR data, especially for certain variables (eg, race) that are not consistently collected by health care providers. Most of the missing information in the present study (for both data sources) related to race. While imputation strategies are reasonable for handling missingness among certain types of variables (eg, clinical outcomes), this is not necessarily the case when the missing information is a variable such as race. We excluded patients from the multivariable analysis who were missing race (n = 765 in Q-EMR) and other covariates or who had less than 2 years of follow-up in the EMR system (n = 14 missing other covariates or with less than 2 years of follow-up in Q-EMR). We recognize that, to the extent those with missing race information may differ from the rest of the sample, the external validity of the findings from the regression models may have been affected. However, the underlying mechanism (ie, missing completely at random, missing at random, missing not at random) of missing race in these databases, especially the Q-EMR database that was used for the multivariable analyses, is uncertain. Moreover, the optimal method for managing missing data for similar databases is unclear.24,25 Missing outcome data due to loss to follow-up is not anticipated because of the requirement that patients have at least 2 years of follow-up from the time of their index procedure. We performed some analyses that capped the follow-up for patients at 2 years, thus adjusting for differential follow-up times across patients. We also performed time-to-event analyses using all follow-up data and censoring patients who were lost to follow-up.
The SPC provided valuable feedback on both the protocol and the analyses of the data. These recommendations were discussed during face-to-face meetings and webinars and the investigators implemented the majority of recommendations that were feasible within the study design and scope. These recommendations were critical in shaping the study design and analyses. The SPC provided numerous recommendations of potential avenues for dissemination of study findings. At the end of the study, the stakeholder engagement evaluation interview assessed the stakeholders’ perspective of their involvement in the study. Overall, there was unanimous agreement about the value of patient advocate involvement, appreciation for the diversity of categories represented within the SPC, and appreciation for the opportunities to work with and learn from one another. Some of the participants expressed that they would have appreciated more opportunity to interact with the core research team. The limitation of the retrospective data source was noted by many of the stakeholders, who felt that important questions could not be addressed in the context of this study design.
This research provides additional new evidence about expectations for the durability of symptom relief and time to subsequent procedures among women with uterine fibroids that may be helpful to women and their health care providers when making treatment choices. Also, the broad agreement between descriptive results from both Q-EMR and COMPASS provides additional evidence to support the descriptive trends identified in this report.
In this analysis of linked Q-EMR and claims data, women who had uterine artery embolization as an index procedure were somewhat less likely to seek medical care for recurrent symptoms and were substantially less likely to have a subsequent procedure during the study period than women who had endometrial ablation, with women who had myomectomy falling in between. African American race was associated with a 20% greater risk of seeking medical care for recurrent symptoms of uterine fibroids, but it was not statistically significantly associated with an increased risk of subsequent procedures compared with White women.
There are multiple possible explanations for the observed associations of measures of treatment durability with specific uterine-sparing procedures. In addition to the possibility of true underlying differences in effectiveness of symptom relief between the procedures, confounding by indication, in which specific procedures were most commonly selected for women with more severe uterine fibroid disease who were also likely to require additional treatment sooner than women with less severe uterine fibroid disease, should also be considered. Other explanations include more complex relationships between the choice of procedure, desire to preserve fertility, and desire to postpone subsequent procedures. For example, endometrial ablation destroys the uterine lining and is used as a treatment for heavy menstrual bleeding due to multiple etiologies; because the treatment does not physically affect fibroids, recurrence of symptoms, or appearance of symptoms unrelated to bleeding, is not unusual. In randomized trials of endometrial ablation compared with hysterectomy for bleeding for all causes, recurrence rates are high.26 As was discussed at an in-person SPC meeting for this study, it is often understood when choosing uterus-conserving procedures that they are not a permanent or curative treatment but will improve symptoms and quality of life in the short term while preserving fertility or postponing the option of a hysterectomy for other reasons. Thus, it is expected that a subsequent procedure will likely be required in the future and without the initial procedure being considered a treatment “failure.” Future studies, namely prospective studies, should collect data on quality-of-life measures, as well as patient preferences and rationale behind treatment selection for uterine fibroids.
In addition to providing additional information that can inform patient and provider treatment decisions, some of the current study findings can also be applied to inform development of the new COMPARE-UF study, a large prospective patient registry of women with uterine fibroids, which will not share the same limitations of this study related to retrospective analysis of existing EMR and claims data.
Overall, the stakeholders provided critical input that shaped the study design, data analysis, and dissemination of key findings. After self-assessment of their involvement, the stakeholders agreed with the value of a diverse stakeholder committee, particularly for eliciting patient perspectives.
Research reported in this report was funded through a Patient-Centered Outcomes Research Institute® (PCORI®) Award (CE-12-11-4430). Further information available at: https://www.pcori.org/research-results/2013/which-treatments-uterine-fibroids-have-best-results
Medical Codes (PDF, 567K)
Stakeholder Partnership Council (SPC) Recommendations for Protocol (PDF, 154K)
Stakeholder Partnership Council (SPC) Recommendations for Analyses (PDF, 136K)
Myers E, Messner DA, Velentgas P. (2018). Which Treatments for Uterine Fibroids Have the Best Results? Patient-Centered Outcomes Research Institute (PCORI). https://doi.org/10.25302/5.2018.CE.12114430
The [views, statements, opinions] presented in this report are solely the responsibility of the author(s) and do not necessarily represent the views of the Patient-Centered Outcomes Research Institute® (PCORI®), its Board of Governors or Methodology Committee.
This book is distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License which permits noncommercial use and distribution provided the original author(s) and source are credited. (See https://creativecommons.org/licenses/by-nc-nd/4.0/
Your browsing activity is empty.
Activity recording is turned off.
See more...
