Randomized Comparison of Uterine Artery Embolization (UAE) with Surgical Treatment in Patients with Symptomatic Uterine Fibroids (REST Trial): Subanalysis of 5-Year MRI Findings
- First Online:
- Cite this article as:
- Ananthakrishnan, G., Murray, L., Ritchie, M. et al. Cardiovasc Intervent Radiol (2013) 36: 676. doi:10.1007/s00270-012-0485-y
- 756 Views
To report 5-year contrast-enhanced magnetic resonance imaging findings of the REST trial recruits who underwent either uterine artery embolization (UAE) or myomectomy.
A total of 157 patients were randomized to UAE or surgery (hysterectomy or myomectomy). Ninety-nine patients who had UAE and eight patients who had myomectomy were analyzed. MRI scans at baseline, 6 months, and 5 years were independently interpreted by two radiologists. Dominant fibroid diameter, uterine volume, total fibroid infarction (complete 100 %, almost complete 90–99 %, partial <90 %), and new fibroid formation were the main parameters assessed and related to the need for reintervention.
In the UAE group, mean ± standard deviation uterine volume was 670 ± 503, 422 ± 353, and 292 ± 287 mL at baseline, 6 months, and 5 years, respectively. Mean dominant fibroid diameter was 7.6 ± 3.0, 5.8 ± 2.9, and 5 ± 2.9 cm at baseline, 6 months, and 5 years. Fibroid infarction at 6 months was complete in 35 % of women, almost complete in 29 %, and partial in 36 %. Need for reintervention was 19, 10, and 33 % in these groups, respectively (p = 0.123). No myomectomy cases had further intervention. At 5 years, the prevalence of new fibroid was 60 % in the myomectomy group and 7 % in the UAE group (p = 0.008).
There is a further significant reduction in both uterine volume and dominant fibroid diameter between 6 months and 5 years after UAE. Complete fibroid infarction does not translate into total freedom from a subsequent reintervention. New fibroid formation is significantly higher after myomectomy.
Uterine artery embolization (UAE) has become a well-established minimally invasive alternative to surgery for women with symptomatic uterine fibroids . Although the pathophysiology of fibroids is poorly understood, it is believed that UAE works by causing ischemic infarction of the fibroids while preserving surrounding tissues (in particular the uterus). In 2004, Pelage et al.  and others subsequently [3, 4] have suggested that complete fibroid infarction is not guaranteed after UAE and that the need for further treatment or reintervention is related to the completeness of infarction of the total fibroid mass. The randomized EMMY  and REST  trials have recently reported 5-year reintervention rates ranging from 23 to 33 %. Patients choosing UAE clearly wish to avoid these reinterventions, which include hysterectomy and repeat embolization.
Contrast-enhanced magnetic resonance imaging (CEMRI) of the pelvis is the established method of imaging fibroids before and after UAE. It offers detail on morphology, vascularity, infarction rates, new fibroid growth, and other pathologies .
The aim of this study was to report the CEMRI findings of women who underwent either UAE or myomectomy within the REST trial and to relate these to long-term clinical outcome, particularly the need for reintervention.
Material and Methods
Imaging Protocol and Analysis
All patients were to undergo a baseline CEMRI scan that was repeated at 6 months and again at 5 years. A standard protocol was used at all the participating centers, which included T2-weighted scans in the sagittal and coronal oblique planes, and fat-saturated pre- and postcontrast T1-weighted scans in the sagittal plane. Two experienced senior radiologists with a special interest in gynecological imaging (F.B. and S.L.) reviewed all available images independently and in a blinded manner.
Imaging parameters included uterine volume, diameter of the dominant fibroid, and degree of fibroid infarction. The entire fibroid burden was assessed for infarction on the postcontrast T1-weighted sequences, and a simple visual scoring system (or eyeballing) judgment used to put the patient into one of three groups: complete (100 %), almost complete (90–99 %), and partial (<90 %) infarction. New fibroid formation and any incidental findings were recorded. Any difference in opinion between the readers was settled by a consensus opinion. For the volume and diameter calculations, archived images were used if available. However, when they were not available, the formal radiology report was used. Only archived images (6-month images) were used for categorizing degree of infarction. For volume calculations, a prolate formula measurement (length × width × height × 0.523) was used. A junctional zone thickening of more than 14 mm was used to diagnose adenomyosis as a part of incidental findings.
Although the main clinical outcome analyses for the REST trial used the intention-to-treat principle, the treatment received by the patient was used for this imaging review. Confidence intervals for the percentage reduction in uterine volume and fibroid diameter were based on a one-sample t test. Analysis of variance was used to examine the relationship between age and degree of infarction, and the Kruskal–Wallis test was used for uterine volume and fibroid diameter. Kaplan–Meier plots and a log rank test were used to illustrate and examine time till reintervention in the three infarction groups. A Mann–Whitney test was used to compare baseline uterine volume and fibroid diameter in those who had a reintervention within 5 years and those who did not; and a two-sample t test was used to compare ages. The incidences of new fibroid formation in the embolization and in the myomectomy cases were compared by Fisher’s exact test.
Reintervention was defined as either repeat UAE, hysterectomy, myomectomy, or endometrial ablation. The indications were for either complications (uterine abscess, n = 1; pain, n = 1) or recurrent and/or persistent symptoms (menorrhagia or bulk symptoms, n = 14). The type of reintervention is described in detail elsewhere [6, 8] and included hysterectomy (n = 10), repeat UAE (n = 5), and endometrial ablation (n = 1).
A total of 157 patients were enrolled onto the REST trial; 99 underwent successful embolization and 8 underwent successful myomectomy. In the UAE group, 72 baseline, 74 six-month, and 68 five-year CEMRI scans were available, and in the myomectomy group, 5 baseline, 4 six-month, and 5 five-year scans were available. In addition to the 72 patients with baseline CEMRI scans, an additional 14 patients had morphological information available from the original baseline radiology report that was used (Fig. 1).
Uterine Volume and Fibroid Diameter
Dominant fibroid diameter and uterine volume at baseline 6 months and 5 years
Time, mean ± SD
Percentage decrease, mean (95 % CI)
Baseline to 6 months
Baseline to 5 years
Uterine volume (mL)
670 ± 503
n = 85
422 ± 353
n = 84
292 ± 287
n = 68
34 % (29–40)
n = 83
53 % (45–60)
n = 67
Dominant fibroid diameter (cm)
7.6 ± 3.0
n = 85
5.8 ± 2.9
n = 82
5.0 ± 2.9
n = 58
27 % (21–32)
n = 82
39 % (31–46)
n = 58
Uterine volume (mL)
616 ± 453
n = 6
109 ± 58
n = 4
161 ± 102
n = 5
78 % (49–100)
n = 4
67 % (50–84)
n = 5
Dominant fibroid diameter (cm)
8.0 ± 4.1
n = 6
Mean ± standard deviation volume was 670 ± 503 mL at baseline, 422 ± 353 mL at 6 months, and 292 ± 287 mL at 5 years. The percentage reduction from baseline in uterine volume at 6 months and 5 years was 34 and 52 %, respectively, indicating that by 6 months, uterine volume had already reduced by over a third.
Mean baseline fibroid diameter was 7.6 ± 3.0 cm, 5.8 ± 2.9 cm at 6 months, and 5.0 ± 2.9 cm at 5 years. The percentage reduction in dominant fibroid diameter was 27 and 39 % at 6 months and 5 years, respectively.
Mean ± standard deviation volume was 616 ± 453 mL at baseline, 109 ± 58 mL at 6 months, and 161 ± 102 mL at 5 years. Baseline fibroid diameter was 8.0 ± 4.1 cm at baseline; measurement at 6 months and 5 years was not relevant after myomectomy.
Fibroid Infarction After UAE
At 6 months, of the 74 available scans, 26 (35 %) had complete, 21 (28 %) almost complete (90–99 %) and 27 (36 %) partial (<90 %) infarction. There was no evidence of an effect of age (p = 0.77), baseline uterine volume (p = 0.50), or fibroid diameter (p = 0.57) on the degree of infarction at 6 months.
Reintervention Relative to Degree of Infarction and Other Parameters
New Fibroid Formation
At 6 months, one patient in the UAE group and none in the myomectomy group had developed a new fibroid. At 5 years, 3 (60 %) of 5 in the myomectomy group and 5 (7 %) of 68 in the UAE group had developed new fibroids (p = 0.008). All new fibroids displayed contrast enhancement.
Incidental findings on CEMRI at 6 months and 5 years
(n = 74)
(n = 4)
(n = 68)
(n = 5)
Ovarian cystic disease
4 (5.4 %)
4 (5.9 %)
2 (2.7 %)
1 (25.0 %)
3 (4.4 %)
1 (20.0 %)
2 (2.7 %)
1 (1.4 %)
1 (1.4 %)
1 (1.5 %)
Gartner duct cyst
1 (1.5 %)
The principal findings of this study relate to changes in fibroid infarction rates and the relationship, if any, between these and the need for subsequent reintervention. Additional information includes new fibroid formation and coincidental new pelvic pathology over a 5-year period.
The additional 5-year CEMRI has demonstrated that the known reduction [3, 4, 9] in uterine volume and dominant fibroid diameter (typically measured at 6 months) is a process that continues to evolve up to potentially 5 years. It is not clear whether fibroid involution is complete by that time, but patients can be reassured that shrinkage will continue beyond the first 6 months, reinforcing the relatively slow and progressive nature of this treatment option.
No relationship was found between the reintervention rate and factors such as baseline uterine volume and baseline fibroid diameter. Others have also failed to identify any relationship between baseline uterine volume and the need for reintervention [3, 4].
CEMRI at 6 months revealed that complete fibroid infarction occurred in a minority (35 %) of patients. Although a further 29 % had almost complete infarction (99–90 %) there were still 36 % with only partial infarction (<90 %). Reported infarction rates vary widely in the literature, and there are few that have adhered to the classification we have used. However, Katsumori et al. , in a cohort of 221, reported a complete infarction rate of 64 %, incomplete infarction of 33 %, and partial infarction rate of 2.3 %. Kroencke et al. , in a cohort of 115, found 52 % with complete infarction, 28 % incomplete infarction, and 20 % with partial infarction.
Factors governing infarction rates are poorly understood but may include fibroid morphology, collateral blood supply, and technical factors such as embolic agent, embolization end point, and operator experience. The REST trial  recruited from 2000 to 2006 and included early experience in some of the centers. However, any technical failures (e.g., unilateral embolization) have been excluded from this analysis. The particle size used in the REST trial was prespecified at 500–700 μm, and most operators used noncalibrated polyvinyl alcohol. Katsumori et al. , with the lowest partial infarction rates, used gelfoam exclusively, and Kroencke et al.  used a wide variety of different products, but not gelfoam.
These differences between studies are difficult to understand but may be partly the result of different study design and methodology. Independent image review by two radiologists with an interest in gynecologic imaging (not an interventional radiologist) may set a higher standard than the normal self-review.
In the literature, partial infarction rates ranged 2.3  to 36 % in the REST trial . Our partial infarction group had the highest reintervention rate (33 %), consistent with findings in other studies. Kroencke et al.  reported a partial infarction rate of 20 %, and Katsumori et al.  reported a rate of 2.3 %, although the latter group contained only five patients.
However, we also found that those with complete infarction had a reintervention rate of 19 %. This is relatively high and contradicts other studies, which have demonstrated very low reintervention rates with complete infarction: 3  and 0 % . Again, this is difficult to explain, and differences in follow-up and study methodology may play a part. All the REST trial patients were followed to 5 years, whereas the other studies had a mean follow-up of 30  and 36 months .
The 5-year outcomes of the REST trial  reported a continued need for reintervention beyond 3 years and indeed did not plateau at 5 years. We also included reinterventions for both complications and recurrent or persistent symptoms. Review of the literature indicates that there is variability in both inclusion criteria and indications for reintervention. Katsumori et al. included reinterventions such as hysterectomy, myomectomy, hysteroscopic intervention, dilatation and curettage, repeat embolization, and angiography for fibroid-related symptoms (including failure of symptom control, complication-related symptoms, or new uterine disease). Kroencke et al. included, hysteroscopic resection, myomectomy, hysterectomy, and repeat UAE for leiomyoma-related symptoms.
The pathophysiology of fibroids is poorly understood, and it is well recognized that the majority cause no symptoms. The link between bulk and pressure symptoms is perhaps easier to understand than menorrhagia. Menorrhagia is more common with submucous fibroids but occurs with other anatomical locations. Production of cytokines and vasoactive agents differ in women with fibroids than those without, which may result in an increase in the menstrual blood loss. There is no evidence for a relationship between fibroid vascularity and menstrual blood loss, and it is oversimplistic to assume that complete fibroid infarction will cure menorrhagia. In any fibroid group, there will in addition always be some patients with dysfunctional uterine bleeding where the fibroids are simply a coincidental association.
Our findings in the myomectomy patients should be interpreted with caution because only eight patients underwent successful myomectomy, and just five of these had imaging at 5 years. However, there are some interesting observations. At 6 months, there was a larger reduction in uterine volume (78 %) than that observed after UAE (34 %). However, some of this advantage was lost at 5 years with the UAE group, indicating a continuing mean volume reduction beyond 6 months (422–292 mL), but the myomectomy group had a slight increase in mean volume (109–161 mL) from 6 months to 5 years.
The above observations may be partially explained by the difference in new fibroid formation between the two groups at 5 years—myomectomy (60 %) compared to UAE (7 %). New fibroid formation after both UAE and myomectomy has been poorly studied, but there are several possible explanations. First, during a myomectomy procedure, small fibroids (<2 cm) may go undetected or may simply be left alone. These could then grow and become labeled as new fibroids. In the REST trial, fibroids <2 cm were ignored at both surgery and magnetic resonance imaging. However, the patients undergoing UAE would have had these small fibroids embolized as the entire uterus is treated. A second possibility is that embolization and the subsequent ischaemia of the uterus may somehow protect against new fibroid development.
Finally, a number of new incidental findings were identified on the CEMRI scans. None was present on the baseline images, and none was of any clinical significance. The most common finding was benign ovarian cystic disease (cysts, endometrioma). Perhaps of most importance was the absence of any interval cancers in the pelvis in these women at 5 years. One of the perceived advantages of hysterectomy is protection against future malignancy in the uterus, cervix, and adnexae. This will reassure patients as they decide between surgery and UAE.
There are some limitations to our study. The overall numbers are small—especially so in the myomectomy group. The study also lacked power to arrive at statistically significant conclusions when comparing outcomes and reintervention rates between myomectomy and UAE. A number of CEMRI scans were lost over the duration of the trial. This was due to advances in image storage and retrieval, with the loss of some original data. This disparity between scans performed and analyzed decreased significantly at 5 years as a result of electronic archiving. Overall, there was only a minor difference in the number of patients scanned at 6 months (n = 74) and 5 years (n = 68).
Further research is clearly necessary into the causes of incomplete infarction and reintervention. Complete infarction, although the aim of UAE, ought not imply freedom from further intervention in the future.
Funded by Chief Scientist Office, Scottish Government Health Directorates, St. Andrew’s House, Regent Road, Edinburgh, EH1 3DG, Scotland, UK.
Conflict of interest
The authors declare that they have no conflict of interest.