Risk Factors of Complications After Uterine Artery Embolisation for Symptomatic Fibroids: A Case-Control Study.

Adeline de Wit is the main author of this study. She collected and sorted all information from charts and imaging and wrote the draft of the manuscript. Marc‐Florent Tassi is the main contributor to the results, performing all statistical work and analysis. Denis Herbreteau is the radiologist performing most part of the embolisation. His deep knowledge in the field led us to start this study. Henri Marret was the investigator of this study. He supervised and helped Adeline de Wit and Marc‐Florent Tassi in finding the focus and the best way to design this study. Adeline de Wit, Marc‐Florent Tassi, Henri Marret and Denis Herbreteau reviewed and corrected the manuscript.

This study and data collection were approved by the Commission Nationale de l'Informatique et des Libertés (CNIL) under number 2019‐029, beforehand on 2 July 2019.

We conducted a retrospective nested case–control study, in our University Hospital aiming to identify risk factors of UAE complications. Our centre performs around 85 embolisations for fibroids each year with an estimated complication prevalence of 5%. Moreover, since 2005, UAE procedures have been standardised and electronic charts have been generalised. All patients who underwent a UAE for symptomatic fibroids between January 2005 and June 2020 were eligible for inclusion. All patients had an appointment with a gynaecologist and a radiologist before undergoing UAE. These consultations aimed to evaluate magnetic resonance imaging (MRI) type, number and size of fibroids and to determine the most appropriate course of treatment for each patient. UAE indications followed the latest recommendations at the time of the consultation. All UAE were performed in the neuro‐vascular specialised unit of our teaching hospital. The same experienced operator and a few assistants have performed UAE routinely in our centre for more than 20 years. After the procedure, all patients were hospitalised overnight in the gynaecological ward. When leaving the hospital, prescriptions included level II analgesics (weak opiates) and anti‐inflammatory medication for a couple of weeks. An information sheet explaining the procedure and complications' symptoms was delivered to all patients. In case of complications, all patients were advised to call in our ward; and if indicated, they were advised to go to the closest emergency service. Being a specialised reference centre, a third of the patients come from other counties. Follow‐up was scheduled for all patients, including an MRI around 6–9 months after the date of UAE and a consultation by the UAE's practitioner following this MRI. An additional consultation was programmed 1 year after the procedure so that patients living in other counties were still followed in our teaching hospital until this point. Patients' targeting was performed with the help of the public health department of our hospital (Service d'Information Médicale, Epidémiologie et Economie de la Santé, SIMEES) to find the best method enabling us to minimise loss of information. Pooling a research by keywords in electronic charts, using diagnostic codes and the French medical act encoding systems, and performing a loss to follow up research, we were able to find 73 cases of complications. These methods were performed on all electronic charts, going back to the year 2000. For patients undergoing several uterine embolisation procedures, the earliest UAE performed in our hospital ward during the study period was considered. Cases were defined as patients undergoing UAE and presenting a complication known to our service. Based on the Society of Interventional Radiology Standards of Practice Committee (SIR) definitions, only severe complications or complications requiring hospitalisation were considered [ 19 ]. Thus, we considered infection or sepsis, expulsion, vaginal bleeding, pain and emergency surgery as complications. Pain was included if it needed hospitalisation for intra‐venous analgesia or level 3 analgesics (Morphine and related substances). Post‐embolisation syndrome was not considered as a complication unless hospitalisation was required to administer intravenous analgesia. Infection was considered if bacterial cultures were positive, intra‐venous antibiotic treatment was administrated or the patient was hospitalised. All clinically diagnosed fibroid expulsions were considered. Patients presenting with vaginal bleeding who consulted in emergency were considered as cases. Failure of the procedure or recurrence of symptoms requiring a second embolisation or a surgical procedure was registered as late complication. Recurrence was defined as reappearance of symptoms within a year after UAE. Failure of treatment was defined as persistence of symptoms despite undergoing UAE. Patients complaining of pollakiuria (abnormally high frequent urination), minimal vaginal bleeding or pelvic heaviness had an indication of a second treatment, but these symptoms were not considered as early complications. If they perpetuated, these symptoms could be a sign of failure of treatment; as such, lead to a second course of treatment. Following literature, and aiming to differentiate early and late complications mostly represented by failures of treatment, thus we chose ahead to separate complications into these two groups. Early complications were defined as occurring within the first month after UAE, whereas late complications occurred later than a month following UAE [ 20 , 21 ]. Complications were considered for the patient's time of follow‐up in our hospital, from the date of the UAE up to June 2020; whichever was the specialist who diagnosed the complication. For each case, two controls undergoing UAE who did not present any complications were selected. All controls were matched to cases on three parameters: (1) county of living, considering our county or outside; (2) patient's age at moment of the procedure, matching was done within a 2 years range; (3) year of the procedure was matched within a 2 years range as well. If several controls were corresponding to a single case, the closest control in age and date of procedure was selected. The ratio of two controls to one case is justified by the initial estimate of the number of identifiable cases and the time available for analysing patient files and extracting data. Exclusion criteria included pre‐operative embolisation. UAE for other indications than fibroids, such as endometriosis, arteriovenous malformation or adenomyosis associated to fibroids or diagnosed on pre‐procedure MRI, were excluded except if the primary reason of the procedure was a fibroid‐related symptom. In case of concomitant carcinoma or suspected carcinoma, patients were excluded. Patients presenting with a vascular complication of UAE were not considered. The same rule was applied for minor embolisation complications (classified as SIR grade 1) [ 22 ] or any hospitalisation not related directly to fibroid's treatment. Patients who did not perform the follow‐up MRI or consultation were excluded from the study and considered lost to follow‐up. All potential risk factors found in literature including those found in our first study on the subject in 2005 [ 20 ] were considered and, if available for data extraction, analysed in univariate analysis: body mass index (BMI), active smoking, presence of intra‐uterine device (IUD), adenomyosis, submucosal fibroids, number of fibroids, volume of uterus and of the larger fibroid, volume of particles injected, type of particles injected, associated treatment or known infection at time of embolisation. Pelvic MRI was performed to assess the presence and number of fibroids, measure uterine and fibroid volume and to classify fibroids using FIGO classification [ 23 ]. PACS imaging software was used for all patients' MRI. Volume of uterus was registered from imaging report by a radiologist. If only its width, length and depth were registered, uterus volume was calculated using Shiota (uterine volume = (length × width × height × 0.35) + 107) and Harb methods (uterine volume = length × height × width × 0.52) [ 24 , 25 ]. Equally, if the imaging team did not calculate the volume of the largest fibroid, an ellipsoid formula was applied to its wider measurement to estimate its volume [ 26 ]. Patient was considered having submucosal risk factor if at least one of her fibroids were described as submucosal by the radiologist (FIGO Type 0 or 1 or 2). Hereafter, a single gynaecologist practitioner reviewed all MRI and UAE images to verify the accordance of MRI reports to measurements. Number and size of major fibroid (estimated by the ellipsoid formula), size of the uterus (estimated by Shiota et Harb methods) and presence of IUD during UAE procedure were reviewed as well. Concerning embolisation procedure, all aspects were collected using operative reports. Patients and procedure characteristics were first described according to their complication group (controls, early cases and late cases). Qualitative variables were described by frequency and rate; quantitative variables with median, first and third quartiles. Number of fibroids was treated as a categorical variable, sorting single fibroid from multiple fibroids as done in literature [ 27 ]. Pairwise comparisons between each case group and controls were performed with Fisher's exact test for qualitative variables and Mann–Whitney test for quantitative variables. Imaging and embolisation characteristics with a significant threshold of p  < =0.30 in bivariate analysis were considered as potential risk factors and included for adjusted analysis in a regression model controlling for confounding effect. Prior to conducting an adjusted analysis, we draw a directed acyclic graph (DAG) to represent our prior beliefs concerning causal relations between potential risk factors, complications and other parameters (Table  S1 ; [ 28 ]). All independence relations involving up to three conditional variables under the DAG were tested against the study data. A multivariate imputation of missing data by chained equation was also carried out prior to modelling. This imputation, based on an assumption of data missing at random (MAR), was carried out using all the available variables presented in Tables  1 and 2 of the results section. One hundred datasets were generated for the modelling process. Comparison of women history characteristics between controls, early ( 1 month) complication groups. Note: Results are reported as frequencies and rate for categorical variables: N (%) and as median, first and third quartiles for quantitative variables: median [Q1–Q3]. p ‐values are calculated with Fisher's exact test for categorical variables and Mann–Whitney test for quantitative variables. Comparison of imaging and embolisation characteristics between controls, early ( 1 month) complication groups. Note: Results are reported as frequencies and rate for categorical variables: N (%) and as median, first and third quartiles for quantitative variables: median [Q1–Q3]. p ‐values are calculated with Fisher's exact test for categorical variables and Mann–Whitney test for quantitative variables. Abbreviation: NC, not calculated. We assessed the global effect of selected risk factors by building for each a multinomial logistic regression adjusted for confounders derived from DAG assumptions (Table  S2 ) and for the three variables used for matching cases to controls (age, residence county and year of the UAE). A random intercept for each pair of case and control was also implemented in the modelling process. Regressions were repeated on each of the hundred imputed datasets and estimators pooled according to Rubin's rule. The association of risk factors included with early and late complications is presented as odds ratio and their 95% confidence intervals. A significant threshold of p  < 0.05 was considered in multinomial analysis. Results of the regressions performed on complete case data are presented as sensitivity analysis in Supporting Information (Table  S3 ). All analyses were performed with R software (R version 4.2.2). In regard to ethical considerations and patient information, the Commission Nationale de l'Informatique et des Libertés (CNIL) approved our study and data collection under number 2019‐029. No further ethical approval was needed for a retrospective study on charts and imaging software.

A total of 1257 UAE were found over the 15 years period using the coding system; out of which 1172 were performed to cure symptomatic fibroids (Figure  1 ). In the retrospective search for patients, only 28 patients over the 1172 patients found did not perform the control MRI planned after UAE in our teaching hospital. The rate of lost to follow‐up before the control MRI was therefore of 2.4%. Flow chart. UAE: Uterine artery embolisation. We collect 73 complication cases, corresponding to a 6.2% rate of complication or failure of treatment. However, 4 patients could not be matched to two controls and were therefore excluded from the analysis. A total of 69 cases between 26 and 54 years old were included in our study after matching cases to controls. They were separated between 31 early complications and 38 late complications. The main overall complications presented by cases were isolated failure of treatment ( n  = 22/69, 31.9%), infection ( n  = 21/69, 30.4%) including sepsis ( n  = 7/21, 33.3%) and pain ( n  = 17/69, 24.6%) (Table  3 ). Overall need of second treatment reached 45% ( n  = 31/69) corresponding to 17 repetitive UAE and 15 surgeries divided in nine hysterectomies, five myomectomies, and one hysteroscopy. Type of early ( 1 month) uterine artery embolisation complications in our population. The infected expulsion of fibroid was considered an ‘infection’ complication being more serious. Abbreviation: UAE, uterine artery embolisation. Population characteristics showed a greater immunosuppressed population in late cases group compared to controls ( p  = 0.01 7.9%). The early cases showed more history of intra‐uterine evacuation 24% versus 8% ( p  = 0.02) whereas the late complication group took more chronic medications related to high blood pressure or hypothyroidism ( p  < 0.01). The two cases populations were similar to the controls on the other clinical criteria studied and fibroid‐related medication (Table  1 ). Median follow‐up was 8.7 years (Q1–Q3: [4.8–11.6]) for controls, 6.2 years [3.6–10.1] for early complications and 10 years [7.5–11.6] for late complications. For six cases and nine controls adenomyosis was diagnosed on the MRI performed to evaluate fibroids (Table  2 ). More subjects had IUDs in place in the early complication group ( n  = 8/31, 26%, p  < 0.01) but not in the late complication group ( n  = 3/38, 8.1%, p  = 0.7) compared to controls ( n  = 9/138, 6.6%). Volume of the uterus estimated by Shiota or Harb's method was significantly higher in the two cases groups (253 cc with Shiota's method for controls compared to 419 cc in the early complication group, p  < 0.01, and 395 cc in the late complication group, p  = 0.01) as well as the volume of the major fibroid estimated by its diameter (57 mm for controls, 70 mm in the early complication group, p  = 0.03, and 72 mm in the later complication group, p  = 0.01). Submucosal location and multiple fibroids were not found to be associated with early or late complications in bivariate analysis. Sixty‐six cases underwent UAE with this same operator, and the last three had a second operator. UAE was performed by puncture of the right femoral artery. Only if this artery was not accessible, the left femoral artery was punctured, which happened twice (2.8%). Two main particles were used: – Embosphere particles of size 500–700, 700–900 and 900–1200 μm, – Embozene (also referred as Celo Nova) particles of size 700, 900 and 1100 μm. Embosphere particles of size 500–700, 700–900 and 900–1200 μm, Embozene (also referred as Celo Nova) particles of size 700, 900 and 1100 μm. Variable methods were registered including mixing Embosphere and Embozene particles or adding Embogold particles to Embosphere particles. Only once injecting Curaspon intra‐arterially did not finish the procedure. Additional procedures in the case groups included micro‐catheters ( n  = 7/69, 10.1%). For controls, additional procedures consisted of two operative hysteroscopies on the same day of the UAE ( n  = 2/138, 1.4%), two IUD removals ( n  = 2/138, 1.4%) and use of micro‐catheters to embolise uterine arteries ( n  = 17/138, 12.3%). There was no evidence that additional procedures were protective against early ( p  = 0.6) or late complications ( p  > 0.9). Embosphere particles were more frequently used in the late complication group compared to the control group ( n  = 23/38, 64% vs. n  = 80/138, 58% p  = 0.01). No differences were found between the early complication group and the controls. On control MRI performed 5–10 months after UAE, 17 controls (12.3%) showed non‐devascularised fibroids compared to 10 in early complication cases (32.3%, p  = 0.01), 15 in late complication patients (39.5%, p  < 0.01). We considered imaging and embolisation factors highlighted in bivariate analysis: adenomyosis, presence of IUD during UAE, volume of the uterus evaluated by Shiota methods, presence of multiple fibroids, diameter of the largest fibroid, submucosal location, type and quantity of particles. Women characteristics and history were not considered in these analyses due to the important number of missing information in the control group. IUD was strongly related to early complications (OR early  = 4.4, IC 95% = 1.5–13.3), which are mostly represented by infections, but did not affect later complications. The same kind of association was observed between early complications and the existence of multiple fibroids (OR early  = 3.7, IC 95% = 1.2–11.3) and possibly the presence of submucosal fibroids (OR early  = 4.8, IC 95% = 0.9–25.6). The only factor that appeared to be associated with risk of early and late complications was the diameter of the major fibroid (for an increased size of 25 mm, OR early  = 1.7, IC 95% = 1.1–2.6; OR late  = 1.5, IC 95% = 1.04–2.2) (Figure  2 ). Results of adjusted analysis (imputed data). IUD, Intra‐uterine device, OR, Odds ratio. In bivariate analysis, the size of the uterus seemed correlated to complications. But this association is not found in the adjusted analysis. Since size of fibroids is strongly related to the size of the uterus, the additional risk found in bivariate analysis might not be causal and is more likely to be mediated by the confounding effect of the number of fibroids more than the size of the uterus itself. Given the very small number of patients who received mixed particles and the variability of these mixes, the association between this variable and the risk of long‐term complications must be considered with great caution. Women with no risk factors have a way smaller chance of presenting an adverse outcome compared to women cumulating risk factors (OR early  = 49.9 IC 95%: [6.9–360], OR late  = 4.5 IC 95%: [0.74–28], for women without IUD and a single myoma compared to women carrying an IUD, with multiple fibroids 25 mm bigger, all other characteristics being equal).

Our study found an increased risk of complications for major size of fibroids. The presence of IUD and multiple fibroids at the time of the procedure are risk factors of early complications. Our study was done in a specialised centre performing an important amount of UAE procedures yearly. We aimed to include all risk factors discussed in literature. Our study reflects the need to differentiate early from late UAE complications, which probably have different risk factors. The number of patients in our study remains scarce due to the low incidence of overall complications after UAE. Other studies need to be performed on the subject to confirm our findings, especially on low incidence risk factors (concomitant infection) or multiple variations of procedure (mix of particles). The early complications are mostly represented by infection [ 29 ]. Presence of IUD was not found a risk factor of early or late complications in several retrospective studies [ 30 ], whereas diameter of the main fibroid was found to be a risk factor of infection in one study [ 17 ]. A submucosal location was not found to be a complication's risk factor in a previous retrospective study [ 31 ]. Over the world, UAE recommendations don't mention a systematic pre‐procedure IUD ablation [ 3 , 32 ]. However, some gynaecologist consider a pre‐UAE removal of IUD to be the ‘best patient care’ [ 33 ]. We have to balance adding an infective risk factor by adding an intra‐cervical procedure or sometimes a hysteroscopy to UAE compared to the overall infectious risk of realising UAE with the presence of IUD because no study answers this question yet and preventive antibiotic injection does not limit the post‐UAE infectious risk [ 34 ]. Our study underlines the importance of presence of IUD as an infectious complication risk factor. ‘Best patient care’ could rely on IUD removal before ongoing with UAE procedure, especially in presence of other risk factors, such as submucosal, large or multiple fibroids. In parallel, size of the main fibroid as UAE complication's risk factor remains controversial in literature [ 3 , 34 , 35 , 36 ]. The Society of Obstetricians and Gynaecologists of Canada mentions number of fibroids and size of the uterus as factors of reintervention [ 32 ]. When a previous study found size of the major fibroid and number of fibroids to be a risk factor of UAE's failure [ 37 ]. Including failure of treatment as a long‐term adverse event could increase the overall complication rate. UAE performed on young patient has more chance to be repeated following the physiological evolution of fibroids during a woman's life until menopause [ 38 ]. Level of fibroid necrosis after UAE should be evaluated as an isolated risk factor of late complications (reintervention) [ 39 , 40 ]. The presence of a remaining vascularisation of fibroid after UAE was not found to be related to complications, failure of treatment included, in our study. Submucosal location appears to be associated mainly to a unique large fibroid than to numerous smaller fibroids ( p  = 0,015). When adding up risk factors of UAE complications, size of major fibroid or number of fibroids and submucosal location, surgery could become a suitable option to prevent an emergency hysterectomy [ 41 ]. In the actual recommendations, size of the main fibroid is decisive to choose the course of treatment or the way to perform surgery [ 2 , 5 ]. But size is not referred to as a complication factor. In French recommendations, myomectomy or UAE can be indifferently discussed for unique large fibroids. Hence, high intensity focused ultrasound (HIFU) could be suggested in such cases.

The gynaecologist should number the presence of UAE complications and potential risk factors presented by each patient to better counsel them on the treatment course. They must be taken into account to inform the patient about an individual possible risk taken by each possible treatment. IUD removal should be discussed before ongoing to UAE procedure. If potential risk factors add up, surgery could become a more suitable treatment plan.

Approximately one‐third of the female population suffers from uterine fibroids [ 1 ]. Uterine artery embolisation (UAE) is now a well‐established conservative treatment for symptomatic fibroids. According to French guidelines, UAE is a good alternative treatment to surgery in women with unique fibroids of less than 10 cm or multiple fibroids with a maximum cumulative size of 15 cm [ 2 , 3 ]. Recent recommendations counsel to discuss treatment with patients in the absence of superiority between UAE or any kind of surgical treatment [ 4 ]. Patients must be informed of potential complications regarding the different treatment courses [ 4 ]. Across the sea, recommendations also emphasise on discussing the fact that hysterectomy could be performed in case of severe complication after UAE, as well as on multidisciplinary team discussion prior to treatment [ 5 , 6 ]. UAE is well accepted, available in most countries and shows very few complications [ 7 , 8 ]. Its most frequent complication is a failure of treatment needing a second embolisation or a surgical procedure [ 9 , 10 , 11 ]. Exceptionally some complications could be life‐threatening, especially infection [ 12 , 13 ], or can justify surgery, like hysterectomy, sometimes in emergency. These complications are rare [ 14 ]. Only a few case reports are found in literature [ 15 ]. Retrospective studies suggest size or location of fibroids to be linked to complications [ 16 ]. Recently, a case–control study found uterine volume and BMI to be post‐operative infection risk factors [ 17 ]. Risk factors for complications are not yet clear and neither researched due to the limited incidence of UAE complications. A better understanding of these risk factors would help strengthen surveillance in exposed women and eventually help prevent potential emergency surgeries by providing an alternative treatment plan [ 18 ].

The ICMJE were provided. Pr Herbreteau and Pr Marret have a past of conflict of interest.

Table S1. Table S2. Table S3.