What they found

van der Kooij and colleagues conducted a thorough search of databases to find applicable studies conducted from September 1995 to November 2010, a period in which UAE went from a new technology to an established technique. They also examined citations attached to relevant studies and review articles. In the end, their metaanalysis included 11 articles stemming from 4 major RCTs. Together, these included 515 premenopausal women with heavy menstrual bleeding due to symptomatic uterine fibroids.

The findings, which included short-term intraprocedural, early postprocedural, and late postprocedural results, showed that UAE was associated with a shorter procedure time, less blood loss, less febrile morbidity, fewer blood transfusions, less pain on post-procedural days 0 and 1, and a shorter hospital stay. Long-term results include data from 6 months to 5 years postprocedure. These showed that UAE was associated with an increased number of re-interventions (including 23 hysterectomies). However, there was no difference in health-related quality of life or in satisfaction between groups.

Statistical quirks

In a metaanalysis like this one, heterogeneity, the difference in results between studies, can be measured informally by visual inspection or by statistical analysis. In general, a combination of the 2 should be performed. Statistical tests, such as the I ² index used in this study, are routinely performed to test for the presence of heterogeneity. However, these tests have important limitations, and indiscriminate use can introduce problems. When the comparison involves few studies, the power of the test is low, and heterogeneity may be interpreted as being nonsignificant when significant heterogeneity is actually present. Conversely, the tests can also have excessive power when there are a greater number of studies, and they may show statistically significant heterogeneity when no important heterogeneity is actually present.

An excellent example of the first situation is in the authors’ comparison of readmission rates between 30 days and 6 months. Here, the EMbolization vs HysterectoMY (EMMY) study showed an odds ratio (OR) for UAE of 19.8, and the Pinto study showed an OR of 1.0. A formal test for heterogeneity showed an insignificant P value, despite the huge and obvious disparity between the results.

Reintervention is another example where a visual inspection of the data is useful. The 4 included studies had disparate study populations and interventions, and results varied widely. While some of the important longer-term outcomes showed tight confidence intervals (CIs), particularly the quality-of-life comparisons, others showed moderately wide CIs. The pooled data indicated that at 5 years, re-intervention was more common among patients who had undergone UAE. In one of the studies, 34.6% of the participants who had undergone UAE required reintervention, compared with 10.7% of women who had a hysterectomy. In another, 26.4% of those treated with UAE required reinterventions, compared with 2% of women who underwent hysterectomy. The differences were significant. However, the OR for reintervention at 5 years showed an OR of 5.4 with a 95% CI ranging from 2.5–11.8. Despite the width of the CI, which indicates a corresponding relative imprecision of the point estimate, it is convincing that the reintervention rate was significantly higher after UAE than after surgery.

And another thing …

Journal Club members pointed out that 1 study included in the metaanalysis had an unusual study design that raised ethical questions. Subjects were randomized prior to consent, and the control group was assigned to hysterectomy without being informed of the study or the availability of the research. Subjects in the study arm, who were randomized to UAE, were given the choice of undergoing UAE or hysterectomy, and those selecting UAE were asked to acknowledge informed consent. Essentially, with this design, people in 1 arm are involved in research without their knowledge or consent and an intervention is withheld. Most ethical arguments in favor of this design rely on the investigational intervention being truly investigational and limited in availability, but that was not the case with UAE. Still, participants thought that ignoring the study results and excluding them from the metaanalysis would only necessitate further research.

Overall, though, the metaanalysis indicated that UAE is a reasonable alternative to abdominal myomectomy or hysterectomy for the treatment of symptomatic uterine myomas, as it had more favorable short-term outcomes and similar long-term results. There was a higher need for reintervention in women undergoing UAE, but patients may find this an acceptable risk, particularly if they want to avoid abdominal surgery or are poor surgical candidates. Journal Club members noted that the role of UAE compared to other minimally-invasive options, including endometrial ablation, vaginal hysterectomy, and laparoscopic or robotic hysterectomy, was not addressed in this study and deserves further investigation. In addition, they suggested that the outcome data in future studies extend beyond a 5-year follow-up, since many women have myomas treated more than 5 years before the menopause.