Preoperative assessment of submucous fibroids by three-dimensional saline contrast sonohysterography

Authors


Abstract

Objective

Submucous fibroids are a common cause of menstrual disturbance that can be treated by hysteroscopic resection. Preoperative patient selection is critical for the success of surgery and it is usually based on hysteroscopic assessment of fibroid protrusion into the uterine cavity. Three-dimensional saline contrast sonohysterography (3D-SCSH) provides additional information about the size and location of submucous fibroids, but it has not yet been evaluated for its usefulness in preoperative assessment. The aim of this study was to examine the potential value of various demographic and ultrasound variables for the prediction of successful submucous fibroid resection.

Methods

This was a prospective study of symptomatic women diagnosed with submucous fibroids who underwent 3D-SCSH prior to hysteroscopic resection. The women's age and parity and fibroid position, diameter, protrusion ratio, size of the intramural component and distance from the internal cervical os were all recorded. The outcome measure was success of a single hysteroscopic resection in achieving a complete excision of the fibroid. Univariate analysis and multivariate logistic regression analysis using a training set and a testing set were performed to investigate the prediction of successful fibroid resection.

Results

A total of 61 women with 67 fibroids were included in the study. There was a statistically significant difference between women who had complete and those who had incomplete resection in the mean protrusion ratio (67.8 (SD 14.5)% vs. 47.0 (SD 17.8)%; P = 0.001), the median maximum fibroid diameter (24.0 (interquartile range (IQR), 19.0–30.5) mm vs. 45.0 (IQR, 28.8–51.3) mm; P = 0.001) and the median size of the intramural component of the fibroid (8.0 (IQR, 4.0–11.5) mm vs. 16.5 (IQR, 12.5–29.3) mm; P = 0.001). Stepwise multivariate logistic regression analysis retained size of intramural component (odds ratio (OR) 0.511), parity (OR 0.002) and fibroid diameter (OR 0.843) as significant independent predictors of a complete fibroid resection. The model had an area under the receiver–operating characteristics curve of 0.975 (SE 0.039) for the training set (n = 39) and 0.864 (SE 0.090) for the testing set (n = 28).

Conclusion

Submucous fibroid protrusion ratio, fibroid diameter and size of the fibroid's intramural component are significantly associated with the likelihood of successful fibroid resection. A logistic regression model can calculate individual probability of complete resection and may improve preoperative counseling of patients. Copyright © 2011 ISUOG. Published by John Wiley & Sons, Ltd.

INTRODUCTION

Submucous fibroids are a common cause of heavy and irregular menstrual periods1. They have also been identified as a potential cause of subfertility and early pregnancy loss2. In the past, the only available surgical treatment for fibroids was open myomectomy. In recent years a hysteroscopic procedure, transcervical resection of fibroid (TCRF), has been introduced into clinical practice. The procedure enables the removal of submucous fibroids without the risks associated with open surgery. It is usually performed as a day-case operation and postoperative recovery is rapid. A significant drawback of hysteroscopic surgery, however, is the inability to successfully remove all submucous fibroids3. A critical aspect of the procedure is careful selection of patients to avoid treatment failure. In an attempt to estimate the likelihood of submucous fibroids being amenable to complete hysteroscopic resection, a classification system that categorizes fibroids into three groups depending on the degree of protrusion into the uterine cavity as assessed at diagnostic hysteroscopy has been developed and adopted by the European Society of Hysteroscopy4. It is assumed that the likelihood of successful removal of the fibroid increases with increasing proportion of its volume protruding into the cavity. Although this classification system is widely used in clinical practice it is not very accurate in predicting the success of hysteroscopic resection4. This indicates that other factors may also play a role in determining the success of surgery, but these have not yet been explored.

Diagnostic hysteroscopy is a subjective method that enables visualization and assessment only of the part of the fibroid that is visible inside the uterine cavity. It cannot be used to measure accurately the size of the fibroid or its component confined to the myometrium. Ultrasonography has some advantages over hysteroscopy for the assessment of submucous fibroids: it can be used in the outpatient clinic as part of the initial assessment of women complaining of menstrual problems, it is simple and safe and is free from surgical or anesthetic complications. In addition ultrasound allows for accurate measurement of fibroid size. Enhancement of the ultrasound image by infusion of saline into the endometrial cavity (saline contrast sonohysterography (SCSH)) facilitates the detection of submucous fibroids and improves diagnostic accuracy5. A combination of SCSH and three-dimensional ultrasound (3D-SCSH) adds further to the assessment of submucous fibroids. By using this technique it is possible not only to measure the size of fibroids but also to determine with accuracy the proportion of the fibroid confined to the myometrium6, 7. The aim of this study was to compare various characteristics of submucous fibroids on preoperative 3D-SCSH with the outcome of hysteroscopic resection in an attempt to identify variables that can be used to predict the likelihood of successful complete fibroid resection.

METHODS

This was a prospective observational study conducted at the Gynaecology Outpatient Department at King's College Hospital NHS Foundation Trust. The study was approved by the research ethics committee of the hospital.

All women presenting with a history of heavy and/or irregular periods between August 2006 and February 2008 had a detailed clinical history taken and were assessed by attending clinicians. They were offered an ultrasound scan as part of the initial investigation of their complaint. The study inclusion criteria were: symptoms of heavy and/or irregular periods and diagnosis of a submucous fibroid on two-dimensional transvaginal ultrasound. We excluded women on hormonal contraception and those who had had a previous operation for their fibroids. Women who consented to take part in the study were then referred for 3D-SCSH to examine their fibroids in greater detail.

The technique for 3D-SCSH has been described previously7. Briefly, using a sterile Cuscoe speculum, the cervix was visualized and cleaned with sterile chlorhexidine solution. A 3.3-mm soft plastic pediatric nasogastric suction catheter was then passed through the cervix into the uterine cavity without grasping the cervix. The speculum was removed, taking care not to disturb the catheter, and a 5-MHz transvaginal 3D ultrasound probe inserted into the vagina (Voluson 730 Expert; GE Medical Systems, Milwaukee, WI, USA). The uterine cavity was visualized and the position of the catheter within it confirmed. A longitudinal view of the uterus was obtained and the catheter was withdrawn to a level just above the internal cervical os. A volume of 5–10 mL of sterile saline solution was then instilled into the uterine cavity and a 3D volume was generated by the automatic sweep of the mechanical transducer. The acquired volume was the shape of a truncated cone, with a depth of 4.3–8.6 cm and a vertical angle α = 90°. The volumes were stored digitally and analyzed using multiplanar visualization. With this technique it is possible to examine the uterine cavity and submucous fibroid in three orthogonal planes to identify the maximum degree of fibroid protrusion into the cavity8. For each woman we recorded the number of submucous fibroids and the location of the fibroid with respect to the endometrial cavity (anterior/posterior/fundal). The widest diameter of each fibroid was then visualized in a plane perpendicular to the endometrium.

The plane of the endometrial–myometrial junction was determined by subjective assessment of the ultrasound image and two measurements were recorded, taking care to exclude the overlying endometrium from the measurement (Figure 1). The section of the fibroid protruding into the cavity (A) and the intramural component (B) were both measured. A protrusion ratio—expressed as a percentage (A/(A + B)) × 100—was then calculated to describe the degree of fibroid protrusion into the uterine cavity. The fibroid diameter was determined as the sum of A + B. The fibroid was then classified according to the European Society of Hysteroscopy Classification of Submucous Fibroids as Type 0 (fibroid polyp), Type 1 (< 50% confined in the myometrium or protrusion ratio > 50%) or Type 2 (≥ 50% confined in the myometrium or protrusion ratio ⩽ 50%)4. We also measured the distance between the lowermost part of the fibroid and the internal cervical os (C, Figure 1). All 3D-SCSH examinations were performed by a single experienced operator and all records were made using a computerized database (PIA Fetal Database, version 5.5.4.152, Viewpoint Bildverarbeitung GmbH, Munich, Germany).

Figure 1.

Ultrasound image of the endometrial–myometrial junction showing a fibroid protruding into the uterine cavity. A, protruding part of fibroid; B, intramural component of fibroid; C, distance between lowermost part of fibroid and internal os.

After the examination women were counseled regarding treatment options. Those who opted for surgery were booked for TCRF irrespective of fibroid diameter and protrusion ratio. TCRF was performed under general anesthesia with a rigid 30° resectoscope with loop wire electrodes (Storz Endoscopy, Germany). Aseptic technique was observed throughout the procedure, which was performed with monopolar or bipolar diathermy with the use of either glycine or normal saline, respectively, to distend the uterine cavity. Infusion pressure was elevated by a pneumatic cuff under manometric control at a pressure of 100–120 mmHg. A high-intensity cold light source and fiber-optic cable were used to illuminate the uterine cavity. The procedure was monitored with a single-chip video camera with the image displayed on a monitor that was visible to the operator. An attempt was made to resect all fibroids completely. Throughout the operation the balance between infused and collected fluid was monitored. The procedure was stopped when all the fibroid had been removed4 or the resection had gone 1 cm deep into the myometrium, as measured from the endometrial surface to the deepest point of the space in the myometrium from where the fibroid had been removed. The procedure was also stopped if the fluid deficit exceeded 1.5 L or it was not possible to visualize the uterine cavity owing to excessive bleeding. In women with multiple fibroids resection was considered complete if all fibroids were removed completely. All surgical procedures were performed by a single surgeon, who was an expert in minimally invasive surgery and who was not blinded to the ultrasound examination results.

Statistical analysis

The aim of the study was to assess the value of various variables for the prediction of complete resection of submucous fibroids at TCRF. The Kolmogorov–Smirnov test was used to test for normal distribution. Dichotomous variables were expressed as percentage (95% CI), non-normally distributed variables as median (interquartile range) and normally distributed variables as mean (SD). We used the chi-square test to compare proportions, the Mann–Whitney U-test to compare the medians of non-normally distributed variables and Student's t-test to compare the means of normally distributed variables. We compared age, parity, median maximum fibroid diameter, median size of the intramural component, median distance from the internal cervical os and mean protrusion ratio between women who had complete and those who had incomplete fibroid resection at TCRF. We then split the data into two sets: a training set and a testing set. We programmed the statistical package to randomly select 60% of cases (with analysis performed per fibroid), which produced the training set on which logistic regression was carried out. The remainder of the cases were used as a testing set on which the derived regression equation was cross-validated. The randomly selected training set contained 39 cases consisting of 27 complete and 12 incomplete resections, while the testing set contained 28 cases—22 complete and six incomplete resections. There was no significant difference in all the variables between the training and testing sets. We performed forward conditional multivariate logistic regression analysis with completeness of resection as the dependent variable and all other variables, including parity (nulliparous (0) vs. parous (1)), fibroid position (anteroposterior (1) vs. fundal (0)) and number of fibroids (single (1) vs. multiple (0)) as independent variables. For each variable included in the analysis we plotted receiver–operating characteristics (ROC) curves and selected the optimal cut-off. All statistical analyses were performed using SPSS v. 16.0 (Chicago, IL, USA) and for all tests two-tailed P < 0.05 was considered statistically significant.

RESULTS

Between August 2006 and February 2008 80 women with symptomatic submucous uterine fibroids were assessed with 3D-SCSH and were provisionally booked for hysteroscopic resection. Nineteen women were not operated on or did not have a complete dataset. 13/19 (68.4 (95% CI, 47.5–89.3)%) women were lost to follow-up, 3/19 (15.8 (95% CI, 0.0–32.2)%) did not attend for their operation and 3/19 (15.8 (95% CI, 0.0–32.2)%) opted for abdominal myomectomy. Out of 61 women who were included in the data analysis 55 had a single submucous fibroid and six women had two fibroids each. The total number of fibroids operated on was 67. Forty-nine of 67 (73.1 (95% CI, 62.5–83.7)%) fibroids were completely resected at TCRF. There were no significant differences in the mean women's age (P = 0.435), the proportion of parous women (P = 0.141), the proportion of women with multiple fibroids (P = 0.281) and the median distance between the distal part of the fibroid and the internal cervical os (P = 0.472) between women who had complete and those who had incomplete resection. The mean protrusion ratio, the median fibroid diameter and the median size of the intramural component (B, Figure 1) were all significantly different in women with complete and incomplete resections (Table 1).

Table 1. Univariate analysis of characteristics of women with complete and incomplete hysteroscopic fibroid resection (n = 67)
ParameterComplete resection (n = 49)Incomplete resection (n = 18)P
  1. Analysis performed on a per fibroid basis, so women with two fibroids have been included twice.

  2. IQR, interquartile range.

Age (years, mean (SD))42 (5.6)39 (7.5)0.435
Nulliparous (n (%; 95% CI))19 (38.8; 25.1–52.4)3 (16.7; 0.0–33.7)0.141
Multiple fibroids (n (%; 95% CI))7 (14.3; 4.5–24.1)5 (27.8; 7.1–48.5)0.281
Mean protrusion ratio (% (SD))67.8 (14.5)47.0 (17.8)0.001
Median maximum diameter of fibroid (mm (IQR))24.0 (19.0–30.5)45.0 (28.8–51.3)0.001
Median size of intramural component of fibroid (mm (IQR))8.0 (4.0–11.5)16.5 (12.5–29.3)0.001
Median distance from fibroid to cervical os (mm (IQR))15.5 (9.0–20.0)14.0 (8.0–24.0)0.472
Fundal location (n (%; 95% CI))15 (30.6; 7.8–43.5)4 (22.2; 3.0–41.4)0.559

The results of multivariate analysis for the training set (n = 39) are presented in Table 2. Parity, size of the fibroid's intramural component and fibroid diameter were significantly associated with complete fibroid resection. The logistic regression equation was: p = 1/(1 + e−logit(p)) where logit(p) was: 2.75 × 109 + (0.002 × parity) + (0.511 × intramural component) + (0.843 × fibroid diameter) (measurements in mm).

Table 2. Results of multivariate analysis to identify significant predictors of completeness of fibroid resection at transcervical resection (training set, n = 39)
ParameterOdds ratio (95% CI)P
Age0.418
Parity0.002 (0.0–0.035)0.035
Size of intramural component0.511 (0.277–0.943)0.032
Fibroid diameter0.843 (0.655–1.0)0.05
Degree of protrusion0.132
Distance from internal cervical os0.597
Fibroid position0.496
Fibroid multiplicity0.148

In the training set the model had an area under the ROC curve (AUC) of 0.975 (SE 0.039) with 96.3 (95% CI, 81.0–99.9)% sensitivity and 91.7 (95% CI, 61.5–99.8)% specificity based on a cut-off of 35.9% calculated probability. When this model was applied to the testing set (n = 28) the AUC was 0.864 (SE 0.090) with 86.4 (95% CI, 65.1–97.1)% sensitivity and 83.3 (95% CI, 35.9–99.6)% specificity based on a cut-off of 9.3% calculated probability (P > 0.05) (Figure 2). We also plotted ROC curves for degree of protrusion, fibroid diameter and size of the intramural component for the testing set (Table 3).

Figure 2.

Receiver–operating characteristics curves for training set (n = 39) (equation image; area under curve (AUC), 0.975) and testing set (n = 28) (equation image; AUC, 0.864) for calculated probability of prediction of success of hysteroscopic resection of uterine fibroids.

Table 3. Receiver–operating characteristics (ROC) curve analysis for application of logistic regression model, degree of protrusion, fibroid diameter and intramural depth using the testing set (n = 28)
ParameterArea under ROC curve (SE)Best cut-off valueSensitivity (%) (95% CI)Specificity (%) (95% CI)
Logistic regression model0.864 (0.090)> 9.3%86.4 (65.1–97.1)83.3 (35.9–99.6)
Degree of protrusion0.777 (0.114)43%86.4 (65.1–97.1)66.7 (22.3–95.7)
Fibroid diameter0.867 (0.070)38.0 mm86.4 (65.1–97.1)83.3 (35.9–99.6)
Size of intramural component0.833 (0.090)10 mm59.1 (36.4–79.3)100.0 (54.1–100.0)

During the study period one woman with incomplete fibroid resection had significant bleeding that necessitated insertion of a Foley catheter to achieve hemostasis and needed overnight admission. Another 8/18 (44.4 (95% CI, 21.5–67.4)%) women with incomplete resection had the procedure abandoned because the fluid deficit exceeded 1.5 L. In the remaining 9/18 women (50.0 (95%, CI 26.9–73.1)%) the procedure was abandoned when resection had proceeded more than 1 cm deep into the myometrium. One woman who had complete resection of her fibroid had fluid overload that necessitated diuretic treatment.

DISCUSSION

Our study showed that the outcome of hysteroscopic resection of submucous fibroids is significantly influenced by parity, the overall size of the submucous fibroid and the size of its intramural component. This study is the first to correlate demographic and ultrasound findings with surgical outcome, and thus gives valuable information for the preoperative counseling of patients. Our results confirm the initial observation by Wamsteker et al.4, who suggested that the degree of protrusion of submucous fibroids into the uterine cavity can be used to assess the probable success of their hysteroscopic resection. The classification of Wamsteker et al. does not include size of the fibroid, which is a possible weakness, as most surgeons would agree that the likelihood of successful fibroid resection tends to decrease with increasing fibroid size. We hypothesize that the omission of fibroid size from the current classification could be explained by limitations of diagnostic hysteroscopy, which does not allow for the accurate measurement of fibroid size. However, previous studies have alluded to the importance of fibroid size in determining the outcome of resection. Vercellini et al.9 suggested that fibroids measuring > 30 mm in diameter should only be operated on by expert surgeons if complete resection is to be accomplished in a single procedure. This recommendation, however, is not based on scientific data but appears to be based on the authors' personal experience. Although Emanuel et al.10 did not assess fibroid size in their analysis of fluid deficit at TCRF, they did suggest that size may be an important determinant of the amount of fluid loss. Hart et al.11 found that fibroids over 30 mm in diameter are more likely to require two-stage procedures. In our study the risk of incomplete resection was significantly higher in fibroids > 38 mm in size, which is in agreement with the previously mentioned studies.

Our results show that the size of the intramural component of the fibroid is also important. This may be because in smaller fibroids resection can be successful even if the protrusion ratio is small. With increasing size of the intramural component the procedure becomes more difficult irrespective of the protrusion ratio. This may explain why in our multivariate analysis size of the intramural component is more significant than the protrusion ratio in predicting complete resection.

To assess the relative importance of the various significant predictors of complete resection and to create a clinically useful tool we performed multivariate analysis. This resulted in a logistic regression equation that can be used preoperatively to estimate the probability of successful fibroid resection for individual women. Apart from improving patient counseling, preoperative knowledge of a low probability of a successful single-stage operation would reduce the pressure on the operating surgeon to remove all of the fibroid in a single session, which may in turn lead to a reduction in complications. An attempt at a novel submucous fibroid classification system has been previously made by Lasmar et al.12, who proposed a complex classification based on a number of parameters (penetration of the nodule into the myometrium, extension of the base of the nodule with respect to the wall of the uterus, size and topography) determined using a variety of diagnostic methods (transvaginal ultrasound, magnetic resonance imaging and hysteroscopy). The system may not be easy to apply in clinical practice and reproducibility and accuracy have not been assessed. Our model is much simpler to use and can generate a probability score for individual women.

While our methodological design eliminates the problem of interobserver variability by using a single operator for ultrasound and operative procedures, it does represent conditions that are difficult to reproduce in routine clinical practice, where operations are performed by different surgeons. In this study we used intraoperative estimation of complete fibroid resection as our endpoint. We chose this endpoint because it allows direct comparison with the initial study in this field by Wamsteker et al.4. However this endpoint is susceptible to operator bias, which may have been exacerbated by the fact that it was not possible to blind our surgeon to the ultrasound results. In order to mitigate this risk we used strict criteria to define completeness of resection, which should improve the reliability of our results. Our single-center set-up inevitably resulted in a relatively small sample size and it is therefore essential that our classification is subjected to prospective audit in individual units before it can be used in patient selection and counseling.

original image

Ancillary