To perform three-dimensional endoanal ultrasound (EAUS) after primary repair of fourth-degree anal sphincter rupture (ASR) and correlate the sonographic defects with anal incontinence (AI); to measure the axial and sagittal thickness and angle of the puborectal muscle (PRM) as well as the length of the anal canal, and then correlate these measures with AI; and to assess the interobserver measurement agreement between an inexperienced and an experienced sonologist.
EAUS was offered to 84 consecutive women, who were asked to answer a validated questionnaire after fourth-degree ASR. AI was graded according to the Wexner score and EAUS defects were graded according to the Starck score.
Sixty-one women (73%) answered the questionnaire. The median (range) follow-up time was 5.1 (1.3–8.7) years. Thirty-three (54%) of these women underwent EAUS and were included in the study. There was no difference in the incontinence scores between women who underwent EAUS and those who did not. Eleven of the women who underwent EAUS (33%) were continent, 22 women (67%) had flatus incontinence at least once a month, of whom 12 also had incontinence for liquid stool and two had incontinence for solid stool. The median Wexner score was 2 (range, 0–12). Five of the patients (15%) had no ultrasound defects. All of the patients with Wexner scores ≥ 4 had a Starck score of ≥ 10. No association between ultrasound defects and AI was demonstrated, however, the angle of the PRM and parity were associated with Starck score. No clear association between the measurements of the PRM and AI was shown. The experienced observer detected more of the small defects than did the inexperienced observer.
The reported prevalence of clinically acknowledged anal sphincter ruptures (ASR) in Denmark is approximately 3.8%1. Follow-up studies of women after ASR, although different in design, are relatively consistent in their findings2–7. Women who have suffered from ASR are more likely to develop anal incontinence (AI) later in life than those with no ASR. Persistent AI subsequent to ASR occurs in 20 to 60% of cases6–19, whereas AI in general has a prevalence of 4%9–11. The mechanism of continence depends on multiple factors, but the puborectal muscle (PRM) is known to be important in keeping women continent11, 12. Some studies with long-term follow-up of women following delivery have found that only a small fraction of cases of incontinence are attributable to ASR13, 14.
Endoanal ultrasound (EAUS) has been used in obstetric research since the early 1990s and is comparable with magnetic resonance imaging (MRI) in showing defects15. Most studies on EAUS have documented an association between defects demonstrated on ultrasound and AI after third- and fourth-degree sphincter rupture16–21. Some research suggests that combined defects of the external anal sphincter (EAS) and the internal anal sphincter (IAS) as well as isolated defects of the IAS are more likely to cause problems7, 8, 19. Furthermore, the length of the EAS is thought to be important in preserving continence22.
Different scoring systems are used for grading AI. The Wexner score grades problems relating to incontinence for flatus and liquid and solid stool, as well as lifestyle changes and the need to wear a pad, on a scale from zero to 2023. Some studies have also used scoring systems when reporting ultrasound defects in the sphincter complex21, 22. The Starck score assesses defects of the EAS and the IAS on ultrasound examination on a scale from zero to 16 (Table 1)24.
Table 1. Scoring system for defects in the external and internal anal sphincters (Starck score)24
Length of defect
Depth of defect
Size of defect
Length of defect
Depth of defect
Size of defect
The aim of this study was to look for a possible correlation between AI and ultrasonographic defects of the anal sphincter in women after primary treatment for ASR. We focused on fourth-degree obstetric ASR to avoid diagnostic bias and used the Starck score to grade defects on ultrasonography. We also investigated whether or not the axial and sagittal thickness and the angle of the PRM had an influence on AI. Another aim was to calculate the agreement between an inexperienced and an experienced observer in the assessment of the Starck score and in measurement of the PRM.
The study was carried out at Hvidovre University Hospital, Copenhagen, which deals with approximately 5500 deliveries per year, with a prevalence of third- and fourth-degree ASR of 3% and 0.3%, respectively. The standard suturing procedure at the hospital is as follows: suturing of the rectal mucosa and the IAS with continuous 3-0 Vicryl (polyglactin 910; Johnson & Johnson, New Brunswick, NJ, USA) in separate layers, and suturing of the EAS with an end-to-end anastomosis using Vicryl 2-0, and closing of the vaginal skin with continuous Vicryl 2-0, followed by a reconstruction of the perineum in three layers, including intracutaneous closure of the perineal skin, using Vicryl 2-0. All patients with ASR were routinely offered instruction in pelvic floor exercises by a specialist physiotherapist while in hospital, and as outpatients 1 month later.
In December 2005, 84 women consecutively treated for fourth-degree ASR between January 1 1999 and December 31 2004 received an invitation to answer a validated, symptom-specific questionnaire25. The letter also contained an invitation to undergo an EAUS examination, which included three-dimensional (3D) imaging. All medical records were scrutinized to ensure that the diagnosis of fourth-degree ASR was correct. The questionnaire was followed by a telephone interview to ensure that the data used to calculate the Wexner score were valid. The study was approved by the local ethics committee (reference number 10328), and the women were only included after written informed consent had been obtained.
A B-K Medical Pro Focus ultrasound machine (B-K Medical, Herlev, Denmark) with an endoanal probe (type 2050) was used for EAUS, which enabled us to record and evaluate sphincter complex and puborectalis recordings in 3D. All women were examined in the lithotomy position by the main author. The probe was introduced into the anal canal until the PRM was visible. The probe was then held in place while the ultrasound crystal was activated and moved caudally, allowing a 3D volume of the full length of the anal canal to be obtained. Recordings were obtained using a frequency of 13 MHz, along a length of 60 mm with 0.25 mm between each image.
Evaluation of the 3D recordings was performed before the analysis of the data from the questionnaire. All measurements and evaluations were executed on a laptop with the use of B-K software. The images were independently analyzed by two of the authors, one with experience of performing fewer than 150 EAUS scans and the other with experience of more than 1000 EAUS scans. In the event of disagreement, the scans were re-evaluated and a consensus was sought. The Starck defect scores for the IAS and the EAS were used for evaluation of the sphincter complex.
A method for performing measurements of the PRM was invented in collaboration with the late Bjørn Fortling from B-K Medical. The thickness in the axial and sagittal planes and the angle of the PRM were measured from a point on a horizontal line passing through the top part of the probe and a line along the outer boundary of the PRM at the 6 o'clock position (Figure 1). The anal canal was defined from the proximal part of the PRM, where the puboanalis muscle was visible, to the distal level of the EAS.
Data on the age and parity of the study population were collected from the local obstetric database. Data from the questionnaire were registered in Excel (Microsoft Excel 2000, Redmond, WA, USA) and analyzed along with data from the EAUS analyses.
SAS 9.1 for Windows (SAS Institute Inc., Cary, NC, USA) was used for statistical analysis. We constructed Bland–Altman plots and limits of agreement (LOA) of interobserver differences for both the Starck score and the measurements of the PRM, and a weighted kappa value (Cicchetti–Allison weights26) was also calculated for interobserver agreement of the Starck score. Spearman's correlation coefficient was used to assess correlations between variables. A regression analysis was also performed on the data set, in which values of P ≤ 0.05 were considered statistically significant.
Of the 84 women contacted, 61 (73%) returned the questionnaire and 33 of these consented to EAUS. No significant differences in reports of AI were found between the group that underwent EAUS and the group that did not. Before the study, three of the 84 women had undergone a secondary anal sphincter reconstruction owing to AI and were therefore excluded. The median follow-up interval between ASR and interview was 5.5 (range, 3.1–8.7) years. None of the women included had delivered vaginally before the occasion on which they suffered ASR. Of the 33 women who underwent EAUS, 22 had had a further pregnancy after ASR and of these 13 had delivered by Cesarean section.
Eleven of the women who underwent EAUS (33%) had no AI. Twenty-two (67%) had flatus incontinence, and of these 12 had incontinence for liquid stool and two had incontinence for solid stool. The median Wexner score was 2 (range, 0–12). All subjects with a Wexner score ≥ 4 (n = 7) had a Starck score of ≥ 10. No significant difference regarding AI was found between women with combined defects and those with isolated defects of the sphincter complex on EAUS. Twenty-eight (85%) of the EAUS recordings showed defects of the sphincter complex, with a median Starck score of 10 (range, 0–14). Of these, 18 had combined defects of the IAS and EAS (Figure 2) and five had isolated defects of the EAS. Sixteen (57%) of the defects in the EAS were located in the proximal part of the EAS only. Data on the PRM are presented in Table 2. We found no association between the thickness of the PRM in the sagittal plane and the length of the anal canal.
Table 2. Characteristics of the puborectal muscle (PRM) and anal canal of women who underwent endoanal ultrasound
Thickness of the PRM in the sagittal plane (mm)
Thickness of the PRM in the axial plane (mm)
Angle of PRM (°)
Length of the anal canal (mm)
Comparing the Starck score with the Wexner score, no correlation was found (r = 0.30, P = 0.08). Using regression analysis we found an association between the Starck score, parity and the angle of the PRM. The two associations were positive, indicating that the Starck score increased with parity (P = 0.008, 95% CI, 1.11–6.65) and the angle of the PRM (P = 0.01, 95% CI, 0.13–0.90). Parity and the angle of the PRM were tested for interaction, but this was found to be insignificant. No association was found between the Wexner score and our measurements of the PRM.
Bland–Altman plots were used to assess interobserver differences in Starck scores for sphincter defects and measurements of the thickness in the axial and the sagittal planes as well as the angle of the PRM (Figure 3). With regard to the Starck score, a systematic difference was seen between the two observers (mean −1.88; LOA −6.29 to 2.53), with the experienced observer classifying the defects as more severe than the inexperienced observer. A weighted kappa value of 0.66 (95% CI, 0.52–0.80) was calculated for the interobserver agreement of the Starck scores. The defects detected by the experienced observer and not by the inexperienced observer were small defects in the EAS.
With regard to the measurements of the PRM, the angle and thickness in the axial plane were the parameters with the smallest disagreement between the two observers (mean −0.16, LOA −7.6 to 7.4; and mean 0.55, LOA −2.96 to 4.06, respectively). The measurements of the angle of the PRM varied within 15°, and measurements of the thickness in the axial plane varied within 7 mm. The interobserver LOA for measurement of the thickness of the PRM in the sagittal plane were −5.1 to 4.70 mm.
This is the first study to focus on ASR in combination with the PRM and the anal canal using 3D EAUS imaging. The prevalence of AI was high in our study population (67%), which was not surprising since we only focused on fourth-degree sphincter rupture. Furthermore, our follow-up time was long compared to those of other studies of this kind. Tetzschner et al.6 reported an incidence of AI of 42% 2–4 years after ASR, and Starck et al.20 reported an incidence of 61% 4 years after ASR. Both these studies assessed women who had suffered third-degree and fourth-degree ASR. The prevalence of defects on EAUS was 85% in our study, nearly the same as the values of 90% and 93% that have been reported in earlier studies20, 21.
No correlation was found between the Starck and Wexner scores in our population. Previous studies have demonstrated a correlation between defects detected on EAUS and AI following obstetric sphincter tears. Norderval et al.21 could not detect a statistically significant correlation between the Starck score and AI at a median of 21 (range, 9–35) months after delivery in a data set larger than ours (n = 61), but Starck et al.20 did show a significant association (P ≤ 0.01) between the Starck score following primary repair and the Wexner score after 4 years in 41 patients. The lack of correlation observed in our study is most probably related to our lack of power; the limitation of our data set to fourth-degree ASR gave us less variation and therefore less power. We did, however, demonstrate an association between defects, parity and the angle of the PRM.
Our Bland–Altman plots revealed several problems in our data. The analysis of interobserver agreement regarding the Starck score showed that the experienced rater classified the defects as more severe than did the inexperienced rater, with the difference between the two raters primarily related to smaller defects in the external sphincter. However, the weighted kappa value of 0.66 indicates good agreement27. We found no association between the length of the anal canal and the thickness of the PRM in the sagittal plane, therefore indicating that the anal canal could not be a confounding factor in our measurement of the PRM.
Tankova et al.11 showed that women with AI had a significantly larger anorectal angle than women without AI. In our data set, measurement of the PRM was not sufficiently reproducible to conclusively test the true relationship between Starck score, AI and the dimensions of the PRM. However, the interobserver agreement was best for measurement of the angle of the PRM, and it was the only measurement of the PRM to be associated with defects on EAUS. The PRM is attached to the perineal body, which is a part of the sphincter complex. This might explain the association between the angle of the PRM and Starck defects. The larger the defect in the sphincter complex, the less caudal the attachment of the PRM, which could result in a wider angle of the PRM.
We were unable to demonstrate a significant difference in AI between the group with isolated defects and the group with combined defects, as has been reported previously7, 16. However, 18 (64%) of the women had combined defects of the EAS and the IAS, and it is possible that a study with more power would be able to detect a difference in AI.
We used 3D EAUS to study the sphincter complex, the PRM and the anal canal after fourth-degree sphincter rupture. The endoanal technique has been used in a large number of studies, and is the generally preferred technique for examining the sphincter complex. Studies which examine the PRM have used MRI or perineal ultrasound methods, primarily in describing defects on the PRM28–30. Up until now, no studies have compared different ultrasound techniques for visualization of the pelvic floor. At this stage it is not possible to recommend one particular technique over another. When studying the posterior part of the PRM, the IAS and the EAS, EAUS certainly has an advantage; the close and parallel relationship to the anal canal makes it possible to use high-frequency transducers, which results in high-resolution images. Another advantage is the fixation of the transducer in the anal canal, which thereby facilitates performance of the examination for the inexperienced examiner. However, it is not possible to use the endoanal technique for studying the levator ani muscle. Here, one would use the transvaginal or transperineal technique.
We hope that future studies, utilizing the various ultrasound techniques available, will provide us with a better understanding of the pelvic floor as a whole functional unit, with regard to both imaging and symptoms.
We are indebted to the Toyota Foundation, which financed the ultrasound equipment to support this study.