Baseline and 30-Day Data

Patient Characteristics

• Patient age: Age is related to ovarian function, likelihood of pregnancy, and possibly recurrence risk for fibroids.
• Patient race: As outlined above, there are clear differences between white and African-American women in incidence and severity of fibroids. Differences in the size and number of fibroids contribute to racial differences in the risk of perioperative complications for surgical therapies for fibroids. Race/ethnicity is collected using classifications approved by the National Institutes of Health (NIH).
• Parity: The relationship between fibroids, infertility and pregnancy outcomes is still somewhat unclear.
• Body mass index (BMI): Calculated from height and weight. Obesity is a risk factor for complications from surgical procedures; if obesity is not a predictor of poor outcome for UAE, then this would be important for patient and physician decision making.
• Menstrual cycle regularity: Establishing a preprocedure baseline is important for determining any effects of UAE on subsequent menstrual function.
• History of infertility: This may be a marker of disease severity, as well as a potentially important covariate for subsequent analysis of pregnancy rates.
• Symptoms (bleeding, intermenstrual or menstrual pain, bulk-related symptoms such as urinary frequency, other): Previous studies of fibroid therapies that preserve the uterus have not characterized the types of symptoms, which makes assessment of response difficult. For example, treatments that result in amenorrhea but do not result in significant shrinkage of fibroids may not result in relief of bulk-related symptoms such as urinary frequency. Knowing the response of specific symptoms to UAE is helpful for subsequent outcome analyses.
• Patient self-reported symptom severity: Reported on a scale of 0-100 (100 most severe).
• UFS-QOL symptom and total scores: Measure symptom status and health-related quality of life.
• Medical therapy for fibroids within 3 months prior to procedure (nonsteroidal anti-inflammatory drugs, oral contraceptives, depot medroxyprogesterone acetate, oral progestins, GnRH agonists, narcotics, other).
• Prior surgical procedures: Useful as a potential surrogate measure for severity of disease and possible predisposition to recurrence or persistence. In addition, since prior surgery may increase the risk of complications from other surgical procedures (through development of adhesions), documentation of any relationship between prior surgery and outcomes of UAE is important.
• Uterine size (measured in three dimensions with ultrasound or magnetic resonance imaging), number of fibroids, largest fibroid, location of largest fibroid within the uterus, morphology of largest fibroid, size and location of fibroid most likely to be causing symptoms (if not the largest fibroid). The intent of gathering this data is to classify in a general way the extent of disease to determine its possible impact on outcome.

Procedure Characteristics

• Procedure length: Procedure length is an intermediate outcome that is related to resource utilization, and may also be a surrogate for degree of difficulty of the case.
• Use of prophylactic antibiotics and prophylaxis against DVT: Prophylactic antibiotics have been shown to reduce the risk of infectious complications of hysterectomy. Although the efficacy of prophylactic antibiotics for prevention of infectious morbidity after UAE can be best demonstrated through a randomized trial, collection of data in a large cohort will provide useful pilot data to help estimate sample size for such a trial. Similarly, DVT prophylaxis has been shown to be cost-effective in some hysterectomy patients.⁶⁰ Because DVT and pulmonary embolus are rare events, a large cohort is needed to determine risk prior to consideration of a randomized trial.
• Periprocedural pain management: As with antibiotics and DVT prophylaxis, the efficacy of various postprocedure pain control methods can best be determined in a randomized trial. Comparison of outcomes by type of pain management used in this cohort is useful for hypothesis generation and sample size estimation for randomized trials.
• Procedure technique (number of vessels embolized, presence of anomalous vessels, type of embolic agents used): Again, data on differences in outcome based on procedural details, especially by type of agent used, is primarily helpful for the design of future studies.
• Fluoroscopy time, number of angiographic images: These data, which will inevitably be correlated with procedure length, are recorded to provide data for possible future estimation of radiation exposure during the procedure.

In-Hospital Outcomes

• Length of stay.
• Patient self-rating of maximal pain level prior to discharge from the hospital.
• Hospital adverse events (groin hematoma, vessel injury, symptomatic non-target embolization, contrast reaction, adverse drug reaction, pain/fever/nausea requiring greater than 48 hours in-hospital, other complication) and severity of adverse event (described in the adverse events definitions section).

30-Day Outcomes

• Self-reported number of days missed from work.
• Self-reported number of days until return to usual activity.
• Reintervention (myomectomy, hysterectomy, repeat UAE, hysteroscopy, D&C, myolysis, endometrial ablation, other).
• Adverse events (recurrent pain, sloughing of fibroids, new hot flashes/night sweats, infection, thromboembolism, persistent bleeding, other) and severity.
• Unanticipated visit to the emergency room, or readmission to hospital. These are important not only for estimation of short-term complication rates and resource utilization, but also because complications requiring readmission have been shown to be a predictor of long-term dissatisfaction with hysterectomy.⁶¹

Long-term Outcomes (6 and 12 months)

• Regularity and length of menstrual cycle: Changes in menstrual cycle regularity and length may be one of the first signs of ovarian failure, either in natural menopause or, possibly, as a consequence of UAE.
• Amenorrhea (yes/no), and reason for amenorrhea (medication, pregnancy, hysterectomy, spontaneous amenorrhea, or other): Pregnancy and hysterectomy are important endpoints. Use of medication to induce amenorrhea indicates persistent or recurrent symptoms. Spontaneous amenorrhea may be related to natural menopause, or caused by the procedure.
• Use of medications specifically for relief of symptoms related to fibroids (nonsteroidal anti-inflammatory drugs, oral contraceptives, oral progestins, depot medroxyprogesterone acetate, GnRH agonists, other). Recurrence or persistence of symptoms requiring medication is an important endpoint.
• Procedures for recurrent or persistent fibroid symptoms (myomectomy, hysterectomy, repeat UAE, hysteroscopy, D&C, myolysis, endometrial ablation, other): Persistent or recurrent symptoms are an important endpoint, reflecting the durability of the procedure.
• Changes in sexual function: Changes in sexual function may be related to changes in ovarian function; prospective studies of hysterectomy for benign disease indicates that sexual function is excellent after the procedure for most women⁶²; data on sexual function after UAE would be important for comparison pending randomized trials.
• Persistent vaginal discharge: Persistent vaginal discharge has been noted in some series after UAE. Data on the long-term incidence and prevalence of this complaint will be useful for patient counseling.
• Unplanned ER visit or hospitalization for fibroid-related symptoms or side effects of UAE: These are important endpoints for detecting late complications or symptom recurrence that might result in additional resource use.
• At risk for pregnancy (heterosexually active without contraception): This variable defines the denominator for estimation of pregnancy rates.
• Pregnancy status (currently pregnant, pregnancy completed).
• Pregnancy outcomes (miscarriage, elective abortion, ectopic pregnancy, live birth).
• Actual date of delivery, expected date of delivery, and birth weight: The majority of women are aware of their "due date" (estimated date of confinement, or EDC) prior to delivery. Since it is likely that many, if not most, women who become pregnant after UAE would be considered high risk on the basis of their history, it is highly likely that most women in the cohort who become pregnant will seek early prenatal care and undergo ultrasound during the first or second trimester. By comparing the patient's self-reported EDC to the actual date of delivery, a gestational age for the time of delivery can be estimated. From this and the birth weight, both premature and small for gestational age infants should be able to be identified.
• Method of delivery (vaginal, planned cesarean section, cesarean section in labor): It is unclear how a history of UAE will affect obstetric decision making; for example, it is possible that obstetricians would recommend cesarean section for women who have undergone UAE, as they often do for women who have undergone extensive myomectomy with entry into the uterine cavity at the time of surgery. A history of UAE may lead to a lower threshold for performing cesarean section in labor if labor progress is inadequate, or if fetal heart rate tracing is not reassuring.
• Indication for cesarean section (history of UAE, breech, fetal distress, other).
• Pregnancy complications (uterine rupture, retained placenta, stillbirth, other): If UAE results in weakening of the uterine wall, an increased risk for rupture during labor is possible. Retained placenta may be more likely if UAE results in changes in the relationship between the uterine wall and the uterine lining, as may be seen after cesarean section or other procedures where scarring of the interior uterine cavity can occur. Stillbirth risk might be increased by persistent changes to uterine blood flow.

We are specifically not measuring uterine size, presence of fibroids on radiological examination, amount of menstrual blood lost or other anatomic or physiological measures of response to therapy during the followup interval. The rationale for this is:

• Controlling for variability in readings or scans performed across multiple sites, potentially even in the same patient, would be difficult.
• Resources were not available to pay for radiological studies that would not necessarily be clinically indicated.
• Most importantly, we believe that patient self-report of symptoms and QOL are the most clinically important and valid indicators of response to therapy.

Return to Contents

Imaging Measures

Imaging confirmation of the diagnosis was required in order for a case to qualify for therapy and be entered in the registry. Details of these imaging studies were recorded, including the type of imaging used, the uterine dimensions (including the cervix), dimensions and location of the dominant (largest) fibroid, dimensions and location of a secondary fibroid that the investigator may have believed to be a significant cause of the symptoms, and the number of fibroids (1, 2, 3, 4, 5, or greater than 5). Once entered into the registry, the volumes of the uterus and listed fibroids were calculated using the following formula for a prolate ellipse (length x width x depth x 0.5233).

Return to Contents

Definitions

Technical Success

Technical success was defined as completion of a bilateral uterine embolization (embolization of either both uterine arteries or the dominant arterial supply on both sides, whether it be uterine, ovarian or another collateral vessel).

Clinical Failure

For purposes of analyzing mid-term outcomes, a definition of clinical failure was needed, and no established definition was available. For this analysis, a clinical failure is defined as (1) failure to achieve at least a 10 percent improvement from baseline in the UFS-QOL symptom score at 12 months after therapy, or (2) the patient having any gynecologic surgical intervention including hysterectomy, myomectomy, D&C, hysteroscopy, and endometrial ablation.

Adverse Events

An adverse event (AE) was defined as any event that was unexpected and resulted in an unanticipated physician office visit, observation period, emergency room visit, or therapy (even if the therapy was only medication). The frequencies of re-interventions, readmissions, and emergency room and physician office visits were recorded. We also recorded whether the adverse events resolved by 30 days after treatment.

The categories of AEs are listed in Table 2. AEs were scored according to the SIR scale for severity.²⁰ In addition, AEs were classified using a system of complication definitions derived from American College of Obstetricians and Gynecologists (ACOG) quality indicators, which has been used previously to classify complications of both hysterectomy and myomectomy.²¹

Return to Contents

Methods for Data Collection

Identification of Eligible Subjects

Subjects treated at core sites who consented to participate in followup beyond the 30-day followup from the site had their contact information and informed consent sent to the followup group; this data was kept separate from the short-term outcome database. At this point, registry sites were not directly involved in obtaining information as part of the routine followup. All followup was conducted centrally at DCRI.

Mailing of Questionnaires

Eligible subjects were mailed questionnaires 6 and 12 months after the date of the UAE. For subjects enrolled in the proposed study, additional mailings will take place at 24 and 36 months, with a possible extension to 48 and 60 months if additional funding is secured. For the 6-month followup, a trained interviewer telephoned patients at their identified contact phone number if questionnaires were not returned within 3 weeks. For longer-term followup, nonresponders were sent a second mailed questionnaire after 3 weeks. If there was no response after 4 weeks, patients were phoned. Using a standard script, patients were reminded about the questionnaire and were offered the opportunity to have another questionnaire sent, or to answer questions over the phone.

Methods for Contacting/Locating Patients

The followup group at the DCRI used standard methods for maximizing followup rates, which have proven successful in numerous registries for cardiovascular outcomes. These methods include confirming at each followup interval patient contact information, including emergency contact and referring physician. If this was unsuccessful, the site interventional radiologist was contacted to obtain contact information.

Return to Contents

Data Management for Questionnaires

After return of mailed questionnaires to the followup group, data were entered into an Oracle® database. Questionnaires were retained in case needed as source documents. Routine checks were performed for range and completeness, although patients were not recontacted for incomplete data.

Return to Contents

Data Analysis: General Approach

Prior to analysis, data in the operational database were transferred to SAS® software (SAS Institute, Cary, NC) for analysis. The following is an overview of the analysis process. Methods for specific analyses are presented along with the results.

Description of Baseline Characteristics

Descriptive statistics are used to characterize the patient population, using the baseline characteristics described above. Continuous variables are summarized using means and standard deviations or medians and interquartile ranges, as appropriate. Categorical variables are presented as proportions.

Description of Outcomes

We calculated annual rates of the following outcomes (using Kaplan-Meier methods to account for loss to followup):

• Proportion of women undergoing therapy for recurrent symptoms.
• Proportion of women undergoing therapy for new symptoms.
• Of those attempting pregnancy, the proportion of women who become pregnant.
• Of those achieving pregnancy, the proportion who experience a live birth.
• Of those women experiencing live birth,
  • Gestational age at delivery.
  • Method of delivery (vaginal vs. planned cesarean vs. unplanned cesarean).
  • Birth weight.
  • Other complications.

Distributions of the following are evaluated at yearly intervals after UAE:

• UFS-QOL score.
• Change from preceding UFS-QOL score.
• Change from preprocedure UFS-QOL score.

Time-to-event is also of interest and is examined for the following outcomes, provided the rates are high enough:

• Time to recurrence of primary symptoms.
• Time to recurrence of primary symptoms.
• Time to therapy for recurrent symptoms.
• Time to therapy for new symptoms.

As with baseline characteristics, summary measures are presented depending on the type of variable (continuous versus categorical), and type of distribution. Point estimates and 95 percent confidence intervals are also presented.

Analysis of Adverse Events

SIR Class A and B AEs were considered to be minor complications, and Class C, D, E, or F AEs were considered to be major complications. With each report of a major complication, the site was asked to provide additional details of the AE, including its nature, duration, treatment provided, preliminary classification of the AE and whether the AE had resolved. An Adjudication Committee consisting of three physician members of the Registry Steering Committee—two interventional radiologists (JBS and RWK) and a gynecologist/epidemiologist (ERM)—reviewed all major AEs and decided by consensus the final assignment of SIR and ACOG classifications.

For event rate calculations, denominators included all patients with AE reports during hospitalization and up to 30 days, even if all baseline data were not complete (n = 3,041 for in-hospital events, n = 2,729 for 30-day events). Rates were also calculated including those with missing AE data (no definite indication of whether or not an AE occurred) assuming that lack of AE reporting indicated no AE. Given that the combined major and minor 30-day events were very low, we examined the combined adverse events for predictors.

For the descriptive analysis, mean/median values were used to describe continuous variables and percentages were reported for categorical variables. The Wilcoxon rank-sum test for continuous variables and the Chi-square test for categorical variables were used to test for differences. To determine factors that predict "any (one or more) adverse event at 30 days" a stepwise Generalized Estimating Equations (GEE) method was used. The GEE method is used to adjust for the correlation between responses and outcomes of patients from the same hospital. The correlation occurs due to factors such as hospital peculiarities, or peculiar characteristics of patients visiting that hospital. The technique renders not only unbiased but also more efficient estimates of the parameters associated with the variable of interest and covariates.

Initially, the investigator's opinion was sought to determine what variables are likely to be associated with occurrence of outcome. Then, univariate analyses were conducted to determine which variables were likely to show an independent effect in the full model. These were again discussed with the investigator and possibilities for existence of interactions and of correlations between covariates were considered. All variables were then tested in multivariate analyses and results were evaluated at each step, sequentially-manually eliminating non-significant variables that were not considered important by the investigator and retaining significant variables (p < 0.05) that were deemed important. This resulted in a predictive model for any adverse event at 30 days.

General Methodology for Predictors of Outcomes

The effect of baseline characteristics on the above outcomes is assessed using multivariable analysis. Because patients from the same provider will tend to have more similar results than patients across providers, hierarchical modeling is used to account for this clustering. For categorical outcomes, hierarchical logistic regression is used with provider as a random effect, while hierarchical linear models are used for assessing the impact of baseline characteristics on UFS-QOL scores. For time-to-event analyses, frailty models are attempted. If lack of convergence is an issue, standard Cox regression models are used.

A primary hypothesis of interest is whether preprocedure uterine volume and number of fibroids are significant determinants of length of time to recurrence of symptoms. More general hypotheses address the factors that influence quality of life over time, especially as they relate to increasing quality of life.

For each outcome, models are constructed using the outcome as the dependent variable. Univariable analyses are performed to identify baseline variables that demonstrate an association with a given outcome. Because the number of events is limited, we wish to identify the minimal number of variables needed for the initial model. Variables are included in the initial model if (a) the variable is shown to be a significant predictor of outcome in the univariable analysis, (b) the variable is expected to have a significant impact based on clinical grounds (variables such as age or race), and/or (c) the variable is one for which testing of interactions is important.

The variables considered for determining significance as predictors were: patient demographic variables (age, BMI, race, parity, uterine volume), medical history variables (smoking history, other comorbidities, prior medical and invasive therapies for leiomyomas), presenting signs and symptoms (predominant presenting symptom—bleeding, pain, bulk symptoms, number of fibroids, size, location and morphology of largest or symptom-causing fibroid), medications used (prophylactic antibiotics, DVT prophylaxis), procedure-related variables (duration of procedure, material used, which vessels were embolized) and site level variables (core, academic, experience).

Transformations for continuous variables that do not meet the linearity assumption (on the logit scale for dichotomous outcomes) are considered and initial models, with carefully selected interaction terms, are constructed. Variables are sequentially removed from the model, starting with interaction terms. Variables whose removal does not significantly change the predictive value of the model were discarded from the final model.

The effect of the final variables in the model on the likelihood of dichotomous outcomes, adjusting for the presence of other variables, is expressed in terms of the adjusted odds ratio with 95 percent confidence limits.

A longitudinal analysis with provider as a random effect and repeated measures on patients over time was used to model UFS-QOL. Patient and process level variables are included in the model, along with a time trend, allowing changes in QOL over time to be assessed. This analysis will be applied when 24- and 36-month outcomes are available.

Predictors of Time to Events

It is important to understand not only whether certain events occur, but also the time frame in which they occur. Development of new and recurrent symptoms, as well as therapy for new and recurrent symptoms, was analyzed using time-to-event analyses. Initially, unadjusted estimates were inspected with Kaplan-Meier curves, applying Greenwood's formula for confidence intervals. For multivariable risk-adjustment, similar techniques to those described above were employed when developing the Cox proportional hazards regression models for time-to-event. Additionally, we verified important model assumptions, such as proportional hazards and linearity of predictor variables.

Missing Data

Prospective registries offer a way to obtain important information about outcomes of procedures that cannot be obtained using other methods, including clinical trials. In order to be useful, nonrandomized studies need to have sufficient information on relevant patient, provider, and procedure characteristics to perform meaningful multivariable analyses to control for potential confounders. Early in the development of a procedure, it may be difficult to identify all of the key variables needed for multivariable analysis, so the "conservative" approach is to collect as much data as possible. This effort requires substantial resources for collecting, managing and analyzing the data. Especially for registries that seek to collect data from a wide range of practitioners, a balance must be found between collecting enough information to address confounding, and making data collection and management as efficient as possible.

So that observations with missing data did not have to be deleted, thereby incurring bias, multiple imputation methods were used. These methods essentially simulate the missing values and draw inferences based on the average over the simulated values, thereby attaching the appropriate significance level. As a sensitivity analysis, we used both the multiple imputation and a more straightforward substitution method, such as assuming certain missing values to be null and others to be at the median of the nonmissing values, and compared results. By identifying those variables where missing values may potentially change the inferences that can be drawn from the data, these sensitivity analyses can help guide future observational studies of uterine fibroids. Imputations used for these initial analyses are shown in Table 3.

A p-value of < 0.05 was established as the level of statistical significance for all tests. All statistical analyses were performed using SAS software (version 8.2, SAS Institute, Cary, NC).

Return to Contents

Chapter 3. Results

Site and Patient Enrollment

A total of 101 sites expressed interest in the registry and were provided materials. Sixteen sites never submitted completed materials or did not obtain institutional review board approval. At the conclusion of patient enrollment, the registry had a total of 85 activated sites: 27 core sites and 58 participating sites.

Table 4 illustrates site enrollment by core vs. participating status, academic vs. non-academic status, and U.S. vs. international. Seventy-two sites contributed data to the registry: 26 core and 46 participating. Twenty-six of the core, or high-volume, sites logged 2,922 (88 percent) patients into the registry of which 2,782 (95 percent) were enrolled. Forty-six of the participating sites logged in 397 (12 percent) patients of which 378 (97 percent) were enrolled. Cases were contributed by 65 United States sites and 7 international sites (4 in the United Kingdom, one in Canada, one in Hong Kong, and one in Australia).

The number of cases per site ranged from 1 to 514. Five sites enrolled over 150 patients, 7 sites enrolled between 75 and 150, 11 sites between 75 and 25, and the remainder less than 25. Core sites were significantly more likely to be academic centers (48 percent vs. 34 percent, p < 0.0001) and to have greater experience with UAE prior to enrolling patients in the registry. The mean (± SD) number of patients for core sites prior to entry into the registry was 9.4 (± 6.2), vs. 0.99 (± 1.19) for participating sites (p < 0.0001).

A diagram of patient flow through the initial part of the study is summarized in Figure 1 (16 KB). The first patient was enrolled December 13, 2000. When the registry closed on December 31, 2002, 3,319 patients had been logged into the registry. Of these patients, 3,166 (95.4 percent) consented to participation in the short-term outcomes registry, with complete data on core variables available for 3,005 (94.9 percent.) Of the 3,160 enrolled patients, 30-day data were received on 2,729 (86 percent) cases.

At the end of patient enrollment, 2,112 patients consented to and were eligible for long-term followup. Patient eligibility criteria included complete symptom and quality-of-life scores (regardless of value of the scores) at enrollment, and patient willingness to participate in followup and to provide contact information. As defined in the methods section, only those patients enrolled at core sites were eligible for long-term followup.

Overall data quality was excellent. Only 2 percent or less of patient baseline characteristics were missing, except for symptom and quality-of-life scores (14 percent), and descriptions of uterine anatomy (4-5 percent missing description of number of fibroids and location of largest fibroid). Consistently, however, missing data for all variables was more common at participating sites than core sites.

Return to Contents

Baseline Characteristics

Table 5 shows the baseline characteristics of the entire patient group. Almost half of the patients were African-American. Although overall comorbidity was relatively uncommon, overweight and obesity were not, with 12 percent of subjects having a BMI greater than or equal to 35. Mean BMI was 27.9, median 26.5.

Table 6 summarizes baseline symptom status and treatment history. Heavy menstrual bleeding (84.5 percent), pain (62.1 percent), and bulk-related symptoms (83.9 percent) were all common, but heavy menstrual bleeding was the single most bothersome symptom in 64.3 percent of patients. Less than half of patients had either medical therapy within 3 months of the procedure (45.3 percent) or a previous procedure (34.7 percent).

Table 7 illustrates the baseline uterine anatomy characteristics as determined by the preprocedure imaging study. In most cases, MRI was the imaging used (58.3 percent). Mean uterine volume was 683.4 cc; 43.4 percent of patients had one or two fibroids, while 32.6 percent had five or more. The most common fibroid type was intramural (42.8 percent). In 10 percent of patients, a second fibroid other than the largest was identified as the primary symptomatic fibroid.

Patient characteristics at core and participating sites were for the most part similar; however, core sites had higher proportions of African-American patients (49 percent versus 39 percent, p = 0.0002), nulliparous patients (40 percent versus 35 percent, p = 0.004), and patients with preexisting diabetes or hypertension (31 percent versus 23 percent, p = 0.0017), and lower proportions of patients with fundal fibroids (25 percent versus 36 percent, p < 0.0001) and large uteri (mean uterine volume at core sites 683 + 524 cc, versus 629 + 492 cc, p = 0.02).

Table 8 shows select temporal trends in patient demographics over the course of the registry. The proportion of cases from participating sites and nonacademic sites increased over time. The proportion of women from racial/ethnic groups other than African-American or Caucasian also increased, as did the proportion of women with a single leiomyoma. Fewer women with BMIs of less than 25 or smoking histories and fewer nulliparous women underwent the procedure.

Table 9 compares the baseline characteristics of African-American and white women. African-American women enrolled in the registry are younger and more likely to be obese and have larger uteri and more numerous fibroids than white women. Bulk-related symptoms such as bloating or urinary frequency were less likely to be the predominant symptom in African-American women, despite their larger uteri. Both symptom scores and QOL results suggest that, on average, symptoms are more severe in African-American women.

Return to Contents

The UAE Procedure

Prophylactic antibiotics were used in most patients, with 74 percent receiving preprocedure antibiotics, and 25 percent receiving antibiotics both pre- and post-procedure. Nearly all patients were given conscious sedation intravenously during the procedure. This was supplemented with narcotics via patient-controlled analgesia (PCA) pump in 67.6 percent of patients, and non-PCA narcotics in 34.8 percent. In addition, 82 percent received nonsteroidal anti-inflammatory medications and 6.7 percent received acetaminophen. Regional analgesia (spinal or epidural) was used in 11.9 percent of patients. Almost all of these were patients in one practice where spinal analgesia is commonly used, in addition to IV sedative and/or hypnotics. DVT prophylaxis was used in 5.1 percent of patients, all in the form of automated intermittent compression devices after the procedure.

The average procedure time was 56 minutes. Average fluoroscopy time was 16.7 minutes. In those cases where angiographic exposures were made, the mean number of exposures was 83.6 and the median was 51.

Technical success was achieved in 96.2 percent of patients. In 92.7 percent there was embolization of both uterine arteries, and in another 3.5 percent of patients there was embolization of one uterine artery and a contralateral ovarian artery in patients with an anatomic variant. Of the 4 percent of cases that were technical failures, there were 25 cases where only one uterine artery was identified and no second vessel (uterine, ovarian, or other) was identified and only unilateral embolization was completed. In the remaining cases two uterine arteries were present but one or both could not be successfully catheterized. A total of 18 patients had no embolization.

The primary embolic agent used was calibrated microspheres (either Embosphere® or Embosphere Gold®, Biosphere Medical, Rockland, MA, U.S.A.). These were used in 73 percent of cases. Particulate PVA (Contour, Boston Scientific/Medi-Tech, Natick, MA, U.S.A.; Biodyne, Cook, Inc, Bloomington, IL, U.S.A. or TRUFILL®, Cordis, Inc, Miami, Florida, U.S.A.) was used in 31 percent of cases, and gelatin sponge (usually Gelfoam, Pharmacia, Kalamazoo, MI, U.S.A.) in 3 percent of cases. In some cases more than one primary embolic agent was used. Embolization was supplemented in 6 percent of cases by either gelatin sponge (5 percent) or embolic coils (1 percent, e.g. Gianturco Embolic Coils, Cook Inc, Bloomington, IL, U.S.A.). Microcatheters were used in 56 percent of cases, with an average of 1.2 microcatheters per case.

Most patients were kept in the hospital overnight and discharged the next day, with a mean stay of 1.7 days (40 hours) from pre-procedure admission to the hospital discharge. Patients returned to normal activities in a mean of 14 days. Of those women who worked outside the home (N = 2,404), the mean number of days lost from work (including the day of the procedure) was 10.

Adverse Events

Adverse events were recorded in hospital and at 30 days after the procedure. These are summarized in Table 10 (severity of complications) and Table 11 (complication by type).

There were 94 adverse events in 90 patients (3 percent of the total population) during the hospitalization for UAE. Four patients experienced 2 AEs. Of these, 74 were SIR Class A or B (minor). There were 20 major AEs (SIR Class C, D or E) that required the patient to stay longer than 48 hours. These were primarily for prolonged pain or nausea (10 events) requiring hospitalization longer than 48 hours. All resolved without sequelae but one. A single Class E complication occurred, which was a femoral nerve injury resulting in leg pain that was judged permanent. There were no deaths (Class F).

Because of the low number of serious adverse events, the power to detect predictors was limited. At the bottom of Table 10, the complications meeting the definitions of the ACOG quality indicators are listed. Because these descriptors are not inclusive of all potential complications, the number listed is smaller than the total in the upper section of the table.

AEs occurring between discharge from the hospital and 30 days after UAE were reported in 710 women (26 percent), of whom 191 (7 percent of total of 2,729) had more than 1 AE. Major events occurred in 111 patients (4 percent), with the most common emergency room care or readmission for recurrent pain in 65 (2.1 percent). The next most common event was evaluation for possible infection in 17 (0.62 percent).

There was one Class E complication at 30 days. A patient developed pelvic pain while on vacation abroad 10 days after embolization. The cause was unclear and exploratory laparotomy was performed. There was inflammation involving both ovaries, although the uterus was normal. Bilateral oophorectomy was performed. The patient had an uncomplicated subsequent recovery. Pathologic report was not available for review.

The most common minor adverse events were hot flashes (5.7 percent) and pain requiring additional therapy (9.6 percent). The classification of minor events at 30 days was limited by incomplete data in 228 cases; it was unclear from the submission whether these required any therapy or not and thus their status as Class A or B complications could not be distinguished. These are listed separately in Table 10.

The complications according to the ACOG classification are listed in the lower portion of Table 10 and represent a subset of only those events meeting the definitions. These standard peri-operative definitions provide description of the complications in terms commonly used for operative morbidity. Using those definitions, the overall morbidity was 0.23 percent in hospital and 7.8 percent at 30 days. The most common of these was febrile morbidity (3.88 percent).

During the period between hospital discharge and 30 days after UAE, 32 (1.17 percent) of patients had a surgical intervention. One patient had a repeat embolization (for initially failed procedure). Three patients had myomectomies, 5 had hysteroscopy with resection/removal of a sloughing fibroid, 9 had D&C, and 3 had hysterectomies. Eleven patients had another type of gynecologic intervention (one discussed above, 10 not specified). Of these interventions, the majority was for management of a sloughing fibroid or passage of fibroid tissue. Indications for hysterectomy were not recorded.

Analysis of Predictors of Adverse Events

In-hospital adverse events. In univariable analyses, there were few predictors of an in-hospital adverse event and included length of procedure (OR 1.012, 95% CI 1.005, 1.019), core site status (OR 0.3341, 95% CI 0.15, 0.76), and size of fibroid (OR 1.073, 95% CI 1.013, 1.138). Patient presenting symptoms, demographic factors, and the type of embolic material used did not predict adverse events. While the use of antibiotics or DVT prophylaxis also did not alter in-hospital complications, the very brief hospitalization associated with the procedure would not normally allow associated complications to develop prior to discharge. There was a "U-shaped" relationship between site prior experience and AE rates: 0-117 cases, 4.75 percent AE rate; 118-200 cases, 1.89 percent; 201-390 cases, 1.13 percent; 391-750 cases, 2.11 percent, and greater than 750 cases, 4.77 percent; p < 0.0001).

Multivariable analysis only demonstrated a relatively small alteration in the odds of an adverse event in hospital. These included length of procedure (OR 1.10, 95% 1.005, 1.01), size of fibroid (OR 1.11, 95% CI 1.028, 1.20) and uterine volume (OR 0.9994, 95% CI 0.998, 0.999).

Events during first 30 days. The univariable analysis for any 30-day event is presented in Table 12. The primary presenting symptoms before UAE (abnormal bleeding, bulk/pressure, pain, other) did not influence the likelihood of any AE, nor did the position of the dominant fibroid (subserosal, intramural/transmural, or submucosal). Those with prior procedures or prior medical therapy, African-Americans, smokers, and those with comorbidities were at increased risk. Use of DVT prophylaxis decreased the odds of an AE, while longer procedures had a minor associated increase in risk. Thirty-day serious events were less common at academic sites (3.1 percent versus 4.7 percent, p = 0.036), but there was no clear relationship with site experience. Again, embolic material and other procedure-related factors did not influence risk nor did presenting symptoms, uterine or fibroid morphology, or operator experience.

Multivariable analysis of adverse events is presented in Table 13. After adjustment for potential confounding, only smoking status (OR 1.14, 95% CI 1.007, 1.29), African-American race (OR 1.129, 95% CI 1.01-1.25), history of prior procedures (OR 1.23, 95% CI 1.02, 1.38), and duration of procedure (OR 1.0037, 95% CI 1.0009, 1.006) were associated with an increased risk of 30-day AEs. Use of DVT prophylaxis decreased risk (OR 0.76, 95% CI 0.62, 0.92).

Return to Contents
Proceed to Next Section