anal incontinence; endoanal ultrasonography; grading systems; obstetric sphincter tears; three-dimensional

Ultrasound Obstet Gynecol 2012; 40: 207 214 Published online in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/uog.10154 Incontinence after primary repair of obstetric anal sphincter tears is related to relative length of reconstructed external sphincter: a case control study S. NORDERVAL*, K. RØSSAAK, A. MARKSKOG* and B. VONEN *Department of Gynaecology and Obstetrics, Ålesund Hospital, Ålesund, Norway; Department of Gastroenterological Surgery, University Hospital of North Norway, Tromsø, Norway; Department of Gynecology and Obstetrics, Copenhagen University Hospital, Herlev, Denmark; Department of Surgery, Institute for Clinical Medicine, Medical Faculty, University of Tromsø, Tromsø, Norway KEYWORDS: anal incontinence; endoanal ultrasonography; grading systems; obstetric sphincter tears; three-dimensional ABSTRACT Objective To determine if anatomic primary repair with end-to-end reconstruction of the external anal sphincter (EAS) in its full length combined with separate repair of coexisting internal anal sphincter (IAS) tear, when present, results in less incontinence and better anal sphincter integrity compared with conventional primary end-to-end repair in which the IAS is not actively reconstructed. Methods Women who sustained third- or fourth-degree obstetric tears were included prospectively in the study following anatomic primary repair. Women treated with conventional primary repair prior to the study period comprised the control group. Three-dimensional endoanal ultrasonography (3D-EAUS) images were classified according to the EAUS defect score, and incontinence according to St Mark s score. Results Sixty-three women were included in the study group and 61 in the control group, with mean follow-up times of 11 and 21 months, respectively. Among women who had not delivered vaginally prior to the tear, St Mark s score 3 was reported by 9.6% (5/52) in the study group and 37.5% (15/40) in the control group at follow-up (P = 0.002). The corresponding numbers among women who had previously delivered vaginally were 36.4% (4/11) and 42.9% (9/21), respectively (nonsignificant). St Mark s score correlated with the EAUS defect score (P = 0.017). An EAS defect exceeding 50% of the sphincter length was significantly less common in the study group, and in a multivariable logistic regression model, mode of repair (anatomic vs conventional) was the only factor explaining the difference in EAS sphincter length between the two groups (P = 0.007). Conclusion Improved continence status after anatomic primary repair was associated with a better longitudinal reconstruction of the EAS, while the integrity of the IAS did not differ between the groups. Women with a history of vaginal delivery prior to the sphincter tear had an inferior outcome regardless of mode of repair. Copyright 2012 ISUOG. Published by John Wiley & Sons, Ltd. INTRODUCTION Obstetric anal sphincter tears occur in 0.4 7.5% of vaginal deliveries 1 4, and remain a major cause of anal incontinence. Until recently primary repair usually consisted of reconstruction only of the external anal sphincter (EAS) without recognition or intended repair of coexisting injury to the internal anal sphincter (IAS). After such conventional primary repairs 40% or more of women report incontinence when assessed in a systematic manner recording symptoms over a 4-week period or longer 2,4 6. A study including deliveries from Ålesund Hospital showed that 42% of the women reported moderate to severe incontinence, with symptoms occurring once a week or more 3. Encouraging results have been obtained after obstetric sphincter tears where injuries to the IAS were identified and repaired in addition to overlap or endto-end reconstruction of the EAS 7 9. In a prior study by our group 9 moderate to severe incontinence was reported by 15% of women following end-to-end repair of the EAS combined with reconstruction of coexisting IAS tear (anatomic primary repair). At the time of initiation of the present study there had been no randomized studies of primary repairs consisting of either end-to-end or overlap repair of the EAS combined with separate reconstruction Correspondence to: Dr S. Norderval, Department of Gastroenterological Surgery, University Hospital of North Norway, N-9038 Tromsø, Norway (e-mail: stig.norderval@unn.no and norde@online.no) Accepted: 10 November 2011 Copyright 2012 ISUOG. Published by John Wiley & Sons, Ltd. ORIGINAL PAPER

208 Norderval et al. of a coexisting IAS tear. A randomized trial comparing primary end-to-end repair with overlap repair of the EAS but without recognition of IAS tears failed to show any differences between the techniques 10. Later randomized studies in which the IAS was included in the repair have provided conflicting results 11,12. Based on available knowledge we decided to introduce anatomic primary repair at the Department of Gynaecology and Obstetrics, Ålesund Hospital, Norway, and to compare the outcome with a historic cohort of women treated with conventional primary repair at the same hospital. The main aim was to determine if anatomic primary repair would lead to a reduced incidence of moderate to severe anal incontinence, defined as St Mark s score 3. The secondary aim was to determine if incontinence at follow-up is related to the extent of postrepair sphincter defects revealed by three-dimensional endoanal ultrasonography (3D-EAUS). PATIENTS AND METHODS In the period from August to December 2004, all consultants and registrars at the department were taught the method of anatomic primary repair. An instruction video showing the repair technique was distributed to all doctors. The video was also available at various locations in the ward, and several instruction lectures were held during the period. By the end of December 2004 all registrars and consultants had participated in or performed at least one primary repair where injury to the IAS was looked for and separately repaired if present. The study was approved by the Regional Committee for Medical Research Ethics, North-Norway, and by the Norwegian Social Science Data Service. Women who underwent primary repair for an obstetric sphincter tear at Ålesund Hospital from January 2005 until December 2006 were prospectively included in the study group, while women who had been treated with conventional primary repair at the same hospital prior to August 2004 were invited to a follow-up study serving as a control group. Written informed consent was obtained before inclusion. Exclusion criteria were previous obstetric sphincter tear, inflammatory bowel disease, neurological disease and previous or subsequent anorectal surgery. Women who were pregnant at followup with a gestational age > 12 weeks were also excluded from the study. Tears repaired prior to 2004 had been classified into Grade 3a, 3b and 4 according to the former commonly used classification system, in which a Grade 3a rupture implies that some fibers of the EAS remain intact in contrast to a Grade 3b rupture, while a Grade 4 tear implies rupture of the sphincters and anal mucosa 3,13,14. In order to be able to compare results between the two groups we had to continue with this classification for the study group, with additional recording of tears to the IAS. Details regarding the anatomic repair (study) group and conventional repair (control) group are given below. For both groups the anal mucosa in Grade 4 ruptures was repaired with 3-0 absorbable polyfilament polyglycolic suture (Vicryl, Johnson and Johnson, Hamburg, Germany) with the knots placed in the anal canal. Antibiotics were not given except for cases of Grade 4 rupture. The wound was examined before discharge. Stool softener (lactulose) was prescribed for 2 weeks in all cases, and diclofenac 50 mg two to three times daily was given routinely for 3 days. Study group When a sphincter tear was suspected the wound was explored thoroughly in order to expose the complete injury. A tear to the IAS was reconstructed with interrupted 3-0 monofilament polyglyconate suture (Maxone TM, Covidien, Mansfield, MA, USA) while the EAS was repaired end-to-end with interrupted 2-0 absorbable monofilament polyglyconate suture (Maxone). A partially intact EAS (Grade 3a rupture) was never divided prior to reconstruction. If a coexisting IAS tear was identified, the intact part of the EAS was gently pulled distally when reconstructing the IAS. Furthermore, special attention was paid to re-establishing the whole length of the EAS from the superficial to the deep part, based on our previous study showing that a short EAS at follow-up manometry is associated with an inferior outcome 9. Data regarding parity, incontinence symptoms and dyspareunia prior to the sphincter tear were recorded before discharge. A follow-up assessment was planned for 9 months after the primary repair. Control group All women who sustained an obstetric sphincter tear during delivery at Ålesund Hospital in the period from January 2002 to July 2004 were invited to a follow-up assessment. We planned to extend the period back in time if the number of women included did not reach that required based on the initial power calculation. Data regarding delivery, primary repair and previous medical history were retrieved from the hospital s electronic medical records. Primary repairs were restricted to the EAS only, with end-to-end reconstruction using 2-0 absorbable polyfilament polyglycolic acid suture. No specific attempt was made to re-establish the complete length of the EAS, and tears to the IAS were neither looked for nor actively repaired. Follow-up Follow-up was identical for the two groups. Incontinence was classified according to the validated St Mark s score, where a score of zero reflects complete continence and a score of 24 reflects maximum incontinence. This score system records frequency of incontinence episodes present in the last 4 weeks prior to assessment, including symptoms of fecal urgency, use of constipating medication and pads in addition to impact on daily activities. The presence of symptoms of dyspareunia at follow-up was

Obstetric tears and sphincter length 209 recorded. Symptoms of incontinence and dyspareunia prior to the sphincter tear were recorded retrospectively with respect to the control group. 3D-EAUS was performed with the women in the lithotomy position using a Falcon ultrasound scanner 2101 EXL with a probe containing a 360 rotating 10-MHz 1850 crystal (B-K Medical, Gentofte, Denmark). Using a motorized external puller EAUS pictures were taken every 0.25 mm from the puborectal muscle to the anal opening, allowing 3D reconstruction. Each patient s complete EAUS dataset was stored electronically for later assessment. All interviews and EAUS investigations were performed by one of the authors (S.N.), who had neither participated in the primary repairs nor seen the patients previously. Ultrasound assessment All 3D-EAUS datasets were assessed by one of the authors (S.N.), who had performed more than 600 such assessments prior to this study. The 3D-EAUS datasets were transferred to a personal computer and displayed on a high-resolution 14-inch LCD-screen using BK 3D-viewer software version 7.0 (B-K Medical). The datasets from both groups were mixed, and assessment was performed blinded to all other data. In all cases the datasets were assessed more than 2 years after the EAUS investigation had been performed. The datasets from the control group had been assessed once previously 15, and the present assessment was performed more than 2 years after the former assessment. The EAUS assessment was standardized as previously described 15 : the puborectal muscle (PRM) represented the proximal limit for assessment of the IAS. The proximal limit for assessment of the EAS was set at 5 mm distal to the distal border of the PRM, as the anterior part of the normal female EAS is shorter than are the lateral and posterior parts. In this way the likelihood of falsely classifying a normal EAS as having a proximal defect was reduced. An EAS defect that was total at one level and partial at another level was classified as total throughout the whole extent of the defect. Sphincter defects were classified according to the EAUS defect score 15. The scoring system classifies defects in the IAS and EAS in three dimensions. A score of zero reflects no defect while a maximum score of seven reflects complete defects in both IAS and EAS extending more than 50% of the sphincter length and more than 90 in the radial plane (Table 1). This EAUS defect score has recently been validated, showing good inter- and intraobserver agreement for the experienced observer 16. The various steps of the assessment are carried out as follows, and illustrated in Figure 1: The level of the distal border of the PRM is identified in the axial plane and marked with electronic calipers (dotted line 1, Figure 1a). This defines the proximal limit for the assessment, and eventual defects are recorded when assessing the scan moving distally in the axial plane. If some of the EAS fibers join anteriorly to leave a partial defect (arrows, Figure 1b), this level is marked with calipers (dotted line 2, Figure 1b). If the EAS at some level appears normal, then it is marked (dotted line 3, Figure 1c). The distal border of the anterior EAS is finally identified (dotted line 4, Figure 1d). In the lateral view the various corresponding levels from the axial assessment are identified (Figure 1e). The distal border of the PRM is illustrated with a continuous vertical line. The normal female EAS is shorter anteriorly than laterally with a proximal anterior gap, hence the assessment of the EAS starts 5 mm distal to the PRM (dotted vertical line, Figure 1e). The longitudinal extension of a complete EAS defect is now seen between the dotted vertical line and the continuous arrow (space A), the longitudinal extension of a partial EAS defect between the continuous arrow and dotted arrow (space B) and the longitudinal extension of a normal-looking EAS is seen between the dotted arrow and double-lined arrow (space C) (Figure 1e). Finally the maximal radial extension of the EAS defect is measured in the axial plane, in this case 70 (Figure 1f). The IAS is assessed in a similar way from the distal border of the PRM, but only complete IAS defects are recorded. In the case shown there is no IAS defect. The length of the EAS defect (both complete and partial) exceeds 50% of the total EAS length (space A + B + C), with a radial extension of less than 90, resulting in an EAUS defect score of 3 (Figure 1f). Table 1 Endoanal ultrasonography defect score for classification of sphincter defects 15 Defect characteristic 0 1 2 3 Score External sphincter Length of defect 50% > 50% Depth of defect None Partial ( 50%) Total but 90 radial defect Total and > 90 radial defect Internal sphincter Length of defect 50% > 50% Depth of defect None Total but 90 radial defect Total and > 90 radial defect Total score is calculated by adding score from each row: no defect, score = 0; maximal defect, score = 7.

210 Norderval et al. Figure 1 Three-dimensional (3D) endoanal ultrasound (EAUS) images showing assessment of the EAUS defect score. (a) Dotted line 1 shows level of distal border of puborectal muscle (PRM) in the axial plane. (b) Dotted line 2 shows level of distal border of PRM when some of the external sphincter (EAS) fibers join anteriorly to leave a partial defect (arrows). (c) Dotted line 3 shows level at which EAS appears normal. (d) Dotted line 4 shows distal border of anterior EAS. (e) Corresponding levels in the lateral view; continuous vertical line indicates distal border of PRM, dotted vertical line indicates assessment of EAS starting 5 mm distal to PRM, space A indicates longitudinal extension of a complete EAS defect, space B that of a partial EAS defect and space C that of a normal-looking EAS. (f) View in axial plane, showing maximal radial extension of EAS defect (in this case 70 ). IAS, internal anal sphincter. Statistical analysis Data were analyzed using SPSS version 16.0 (SPSS Inc., Chicago, IL, USA). The χ 2 -test or Fischer s exact test was used as appropriate for comparison of categorical data, while continuous data were analyzed using the t-test. Correlation was assessed using Spearman s test. Multivariable linear regression or logistic regression was performed when appropriate to adjust for the effects of covariates, and two-tailed P < 0.05 was considered to be statistically significant. An initial power calculation was performed based on the results of our previous studies 3,9. In order to show a reduction in the incidence of moderate to severe incontinence from 42 to 15% with a power of 0.90 using a two-sided test and alpha of 5%, 57 patients would be needed in each group. The incontinence data from the control group were analyzed when 61 women had been included. Moderate to severe incontinence was reported by 39% at followup. Based on this result a new power calculation was performed, which showed that 80 women would be needed in the intervention group to show a reduction in incontinence from 39 to 15% with a power of 0.90 and alpha of 5% using a two-sided test. Our intention was to include this number of women. However, in 2006 the department of Obstetrics and Gynaecology joined a project that aimed at reducing the incidence of obstetric sphincter tears by changing the midwives practice according to the Finnish method 1, focusing on manual assistance during the final part of the second stage of labor. This project turned out to be very successful, and the incidence of obstetric sphincter tears dropped immediately from about 5 to 1.5% 17. Based on this new low incidence of tears the study period would have had to be extended by 2 years in order to reach the required inclusion number. For practical and economic reasons the head of the hospital s research program decided to stop inclusion by the end of 2006. RESULTS A total of 155 women sustained a sphincter tear between 1 January 2002 and 31 July 2004, and 61 women (39.4%) were included in the control group. The results from the control group have been published in part previously 15. Eighty-five women sustained a sphincter tear between 1 January 2005 and 31 December 2006, of whom 63 (74.1%) were included in the study group. Causes of exclusion in the two groups are shown in Table 2. There were no significant differences in the frequencies of third-

Obstetric tears and sphincter length 211 Table 2 Reasons for exclusion from study group (anatomic primary repair) and control group (conventional repair) Parameter Study group Control group Total number of women with tears 85 155 during inclusion period Number of women included 63 (74.1) 61 (39.4) Number of women excluded 22 (25.9) 94 (60.6) Reasons for exclusion Declined participation 3 (3.5) 28 (18.1) Subsequent pregnancy/delivery 2 (2.4) 20 (12.9) Previous obstetric sphincter tear 8 (9.4) 9 (5.8) Previous or subsequent anorectal 0 (0.0) 3 (1.9) surgery No response to follow-up requests 3 (3.5) 34 (21.9) Locum obstetrician on call 4 (4.7) Communication difficulties 1 (1.2) Unknown 1 (1.2) Data given as n or n (%). Table 3 Degree of sphincter tears in included and excluded women eligible for study group (anatomic repair) and control group (conventional repair) Grade of rupture Group 3a 3b 4 Study group Included women (n = 63) 33 (52.4) 24 (38.1) 6 (9.5) Excluded women (n = 22) 12 (54.5) 8 (36.4) 2 (9.1) Control group Included women (n = 61) 37 (60.7) 16 (26.2) 8 (13.1) Excluded women (n = 94) 58 (61.7) 24 (25.5) 12 (12.8) Data given as n (%). There were no statistical differences between included and excluded women in the study group or in the control group, or between the included women in the two groups. and fourth-degree sphincter tears between the control group and the study group, or between included and not included women in either of the cohorts (Table 3). In the study group an IAS tear was identified during primary repair in 11 of 57 women (19.3%) with third-degree tears. The time from primary repair to follow-up was shorter for the study group than for the control group, with a mean of 11 (range, 9 17) months compared with 21 (range, 9 34) months, respectively (95% CI of the difference, 7.8 12.4; P < 0.001). A total of 52 out of 63 women (82.5%) in the study group had not delivered vaginally prior to the sphincter tear compared with 41 of 61 women (67.2%) in the control group (P = 0.031). The tears were repaired in the operating theater under regional or general anesthesia to a greater extent in the study group than in the control group. Obstetric and demographic data are shown in Table 4. Postoperative complications Wound complications were seen in three women in the control group. Two women developed wound infection, Table 4 Obstetric and demographic data among women who sustained obstetric sphincter tears and were treated with anatomic repair (study group) or conventional primary repair (control group) Parameter Study group (n = 63) Control group (n = 61) One or more vaginal deliveries prior to 11 (17.5) 20 (32.8)* sphincter tear Cesarean section only prior to sphincter 7 (11.1) 9 (14.8) tear Epidural during delivery 13 (20.6) 16 (26.2) Birth weight (g) 3801 3849 Episiotomy 22 (34.9) 22 (36.1) Vacuum extraction 13 (20.6) 14 (23.0) Forceps delivery 3 (4.8) 2 (3.3) Repaired in operating theater 22 (34.9) 3 (4.9) Repaired under regional or general 36 (57.1) 19 (31.1) anesthesia Repair performed or assisted by consultant 36 (57.1) 44 (72.1) Data given as n (%) or mean. *P = 0.031. P < 0.001. P = 0.004. P not given if > 0.05. which was successfully treated with antibiotics and wound irrigation. One woman developed an anovaginal fistula that healed spontaneously after 2 months. No complications were recorded in the study group. Incontinence Three women (4.9%) in the control group and three (4.8%) in the study group reported some incontinence symptoms prior to the sphincter tear. Incontinence, defined as St Mark s score > 0, was reported at follow-up by 19/63 women (30.2%) in the study group and 32/61 (52.5%) in the control group (P = 0.012). Moderate to severe incontinence, defined as St Mark s score 3, was present in nine women (14.3%) in the study group and 24 women (39.3%) in the control group (P = 0.002). Mean St Mark s score was 1.1 in the study group and 2.2 in the control group (95% CI of the difference, 0.2 2.1; P = 0.017). Among women who had not delivered vaginally prior to the sphincter tear, incontinence was reported by 14/52 (26.9%) in the study group and 21/40 (52.5%) in the control group (P = 0.012), with St Mark s score 3 in five (9.6%) and 15 (37.5%), respectively (P = 0.002). Among women who had delivered vaginally at least once prior to the sphincter tear, incontinence at follow-up was reported by 5/11 (45.5%) in the study group and 11/21 (52.4%) in the control group (P = 0.71), with St Mark s score 3 reported by four (36.4%) and nine (42.9%) women, respectively (P = 0.72). Dividing the control group into two equal subgroups based on the time from primary repair to follow-up, incontinence was reported by 16/30 women (53.3%) in the subgroup with a mean follow-up of 13 (range, 9 20) months and by 16/31 women (51.6%) in the subgroup with a mean follow-up of 28 (range, 21 35) months. Incontinence was present in two of the three women who

212 Norderval et al. Table 5 Ultrasonographic findings at follow-up related to peroperative classification of tears in women who had not delivered vaginally prior to the sphincter tear Peroperative assessment Follow-up ultrasonographic findings Degree of tear n No defect (n) EAS defect (n) EAS and IAS defects (n) Study group (anatomic repair) 43 4 34 5 Partial EAS tear (Grade 3a) 24 2 21 1 Without IAS tear 18 2 15 1 With IAS tear 6 0 6 0 Complete EAS tear (Grade 3b) 14 2 12 0 Without IAS tear 10 1 9 0 With IAS tear 4 1 3 0 Grade 4 tear 5 0 1 4 Control group (conventional repair) 35 1 28 6 Partial EAS tear (Grade 3a) 20 1 16 3 Complete EAS tear (Grade 3b) 11 0 10 1 Grade 4 tear 4 0 2 2 EAS, external anal sphincter; IAS, internal anal sphincter. sustained wound complications in the control group, with a St Mark s score of 8 in both cases. Excluding these three women from the analysis altered the above findings only marginally. Ultrasonographic findings Owing to technical problems 3D scans were available in only 55 of the 61 women (90.2%) in the control group. In the study group, 52 of 63 women (82.5%) were willing to undertake EAUS at follow-up. Hence 3D scans from a total of 107 women, of whom 78 had not delivered vaginally prior to the sphincter tear, were available for assessment. The distribution of the various EAUS findings in the two groups at follow-up is shown in Figure 2, and the relationship between degree of tear and ultrasonographic findings in women who had not delivered vaginally prior to the tear is shown in Table 5. The mean EAUS defect score was 2.4 in the study group and 3.1 in the control group (95% CI of the difference, 0.07 1.23; P = 0.029). In the study group mean EAUS score was 3.0 in women who had delivered vaginally prior to the sphincter tear and 2.3 in women who had not delivered vaginally prior to the tear (95% CI of the difference, 0.37 to 1.82; P = 0.19, while the corresponding scores in the control group were 3.7 and 2.7 (95% CI of the difference, 0.18 1.85; P = 0.017). In a multivariable linear regression model including both groups and adjusting for factors affecting the EAUS score, prior vaginal delivery and degree of sphincter tear (third-degree vs. fourth-degree) contributed significantly to the equation (P = 0.007 and 0.003, respectively), while mode (conventional vs. anatomic) and location (delivery room vs. theater) for repair did not (P = 0.13 and 0.99, respectively). Separate analysis of the various EAUS defect score parameters revealed no differences between the study and control groups with respect to size or length of IAS defects. However, 22/52 women (42.3%) in the study group had an EAS defect exceeding 50% of the sphincter Distribution of EAUS findings after anal sphincter repair (%) 90 80 70 60 50 40 30 20 10 0 No defect EAS defect EAS and IAS defect Figure 2 Relative distribution of endoanal ultrasonographic (EAUS) findings after conventional primary repair ( )(n = 55) and anatomic repair ( )(n = 52). Differences between the two groups were not significant. EAS, external anal sphincter; IAS, internal anal sphincter. length compared with 39/55 (70.9%) in the control group (P = 0.003). Among women who had not delivered prior to the sphincter tear, an EAS defect exceeding 50% of the sphincter length was seen in 18/43 women (41.9%) in the study group and 23/35 (65.7%) in the control group (P = 0.036). In multivariable logistic regression analysis adjusting for factors affecting the EAS sphincter length as a dichotomous variable (defect more or less than 50% of the sphincter length), mode of repair significantly contributed to the equation (P = 0.007) while previous vaginal delivery or degree of tear (third-degree vs. fourth-degree) did not (P = 0.35 and 0.13, respectively). No other significant associations between mode of repair (conventional vs. anatomic) and the various EAUS defect score parameters were found when adjusted for vaginal delivery prior to the sphincter tear. Ultrasonographic findings related to incontinence There was an overall significant correlation between the degree of incontinence expressed by the St Mark s

Obstetric tears and sphincter length 213 score and the extent of ultrasonographic sphincter defects expressed by the EAUS defect score (P = 0.017). Mean St Mark s score was 1.1 in women with EAUS defect score 0 2 and 2.4 in women with EAUS defect score 3 7 (95% CI of the difference, 0.27 2.93; P = 0.014). Mean St Mark s score was 1.1 in women with EAS defects 50% of the sphincter length and 2.3 in women with EAS defects > 50% (95% CI of the difference, 0.15 2.13; P = 0.025). When follow-up EAUS revealed combined EAS/IAS defects the mean St Mark s score was 3.4 vs. 1.2 when only EAS defects were seen (95% CI of the difference, 0.12 4.25; P = 0.039). However, the extent of the EAS defects expressed as the EAS part of the EAUS defect score (with scores ranging from 0 to 4 (Table 1)) were 3.1 in women with coexisting IAS defects and 2.4 when only EAS defects were seen (95% CI of the difference, 0.35 1.14; P < 0.001). Other findings Ten of the women in each group (15.9 and 16.4% of the study and control groups, respectively) reported some degree of dyspareunia prior to the sphincter tear. At follow-up dyspareunia was reported by 14/61 women (23.0%) in the study group (two women in the study group had not resumed sexual intercourse at follow-up) and 19/61 (31.1%) in the control group (P = 0.31). DISCUSSION Overall incontinence rate and incontinence rate with St Mark s score 3 were significantly lower in the study group than in the control group, indicating the superiority of anatomic primary repair. Although the study was underpowered according to the second power calculation, the difference reached significance because of a better result than anticipated in the study group. Unexpectedly, this difference was not present in the subgroup of women who had delivered vaginally at least once prior to the sphincter tear, where the incontinence rate was high in both groups. The inferior result in this subgroup was associated with a higher EAUS defect score, and vaginal delivery prior to the sphincter tear was the only factor that significantly affected the EAUS score in a multivariable regression model. It is known that undiagnosed sphincter injuries occur in 6 35% of vaginal deliveries 18 20. Although none of the women in our study had sustained a clinically revealed obstetric sphincter tear during prior deliveries, some may have sustained undiagnosed sphincter tears. A pre-existing sphincter defect could affect the quality of the primary repair, resulting in inferior reconstruction and functional outcome. Previous vaginal delivery may also affect pudendal nerve function, which in turn may increase the risk of incontinence after a subsequent sphincter tear 20. We have no data that can shed further light upon this, as our study did not involve evaluation of the pudendal nerves. However, the higher EAUS defect scores in this subgroup indicate a structural rather than neurological explanation for the unfavorable outcome. The incidence of IAS defects at follow-up EAUS did not differ between the groups (Figure 2). This was also the result when only EAUS from the subgroup of women who had not delivered vaginally prior to the repair was analyzed, thereby excluding the possibility of preexisting IAS tears (Table 5). Persisting IAS defects at EAUS follow-up despite the intention to reconstruct a coexisting IAS tear during primary repair has been reported previously 7,9. In their study Sultan et al. 7 reported IAS defects in 44% of women at EAUS followup, and most of the cases represented IAS tears overlooked during primary repair. In the present study, overlooked IAS defects during primary repair were only revealed in 3% of the women in the study group following a third-degree rupture (Table 5). Furthermore, the IAS was intact in all women in whom an IAS tear had been identified during primary repair, indicating an adequate operating technique for third-degree tears. The somewhat disappointing result regarding IAS patency following the attempt at anatomic repair of fourth-degree tears may be explained by inadequate identification of the IAS defect or by inappropriate surgical technique. Our finding probably reflects the complexity of fourth-degree tears, with edema and bleeding complicating precise identification of the anatomical structures. However, the improved outcome after anatomic primary repair was not associated with the patency of the IAS but rather with the relative length of the reconstructed EAS. In a multivariable logistic regression model the mode of repair (anatomic vs. conventional) was the only factor explaining the difference in EAS sphincter length between the two groups. An inverse correlation between the degree of incontinence and anal sphincter length at anal manometry has been reported by our group 9. Hence, reconstruction of the whole length of the EAS was a crucial step in the anatomic repair. Our present ultrasonographic findings highlight the importance of reconstructing the entire length of the EAS during primary repair. It is possible that the outcome would have been slightly better if a higher degree of IAS patency had been obtained, but this remains conjecture. There are several limitations to this study. More women in the control group than in the study group had delivered vaginally prior to the sphincter tear. This difference affected the outcome, and has been dealt with by analyzing the subgroups separately or by controlling for the difference using multivariable regression analysis. Furthermore there are at least three important factors that may have biased our results. First, only 39% of the women sustaining obstetric tears in the given period were included in the control group. It is possible that women who were continent after the sphincter tear were less likely to participate in the follow-up study, resulting in a false high incontinence rate in our control group. This is countered by the fact that the incontinence rate was quite similar to the incidence revealed in a previous study including 83% of all sphincter tears over 2 years,

214 Norderval et al. and where most of the included women had delivered at Ålesund Hospital 3. Several other Scandinavian studies report an overall incontinence rate of between 46 and 60% following conventional primary repair 13,14,21,22.We therefore argue that the overall incontinence rate of 52% in the control group reflects the true incidence following conventional primary repair at the study hospital. Second, time to follow-up was 10 months longer in the control group than in the study group. Starck et al. 14 reported a slight increase in the incidence of incontinence from 1-year follow-up to 4-year follow-up after primary repair, but many of the women in this study had a subsequent vaginal delivery that may have affected the outcome. In the control group, incontinence rates in the subgroups with 13 and 28 months follow-up were the same. It therefore seems less likely that the somewhat longer time to follow-up in the control group affected the incontinence rate in this group. Third, more women in the study group had the repair performed in the operating theater under regional or general anesthesia. Even though this was an intentional change, the better operative environment provided by a theater could impact positively on the quality of the repair. However, the location of the operation did not affect the EAUS defect score at all in multivariable linear regression analysis. As repairs performed in theater were so strongly associated with the use of regional or general anesthesia, our data do not show any benefit in itself of performing the repairs under regional or general anesthesia. In conclusion, the incidence of incontinence was lower after anatomic primary repair than after conventional repair. Improved continence status was associated with a better reconstruction of the EAS, while the integrity of the IAS did not differ between the groups. There was an overall correlation between the degree of incontinence and the extent of the ultrasonographic defects. Women with a history of vaginal delivery prior to the sphincter tear had an inferior outcome whatever the mode of repair. REFERENCES 1. Pirhonen JP, Grenman SE, Haadem K, Gudmundsson S, Lindqvist P, Siihola S, Erkkola RU, Marsal K. Frequency of anal sphincter rupture at delivery in Sweden and Finland result of difference in manual help to the baby s head. Acta Obstet Gynecol Scand 1998; 77: 974 977. 2. Nordenstam J, Mellgren A, Altman D, Lopez A, Johansson C, Anzen B, Li ZZ, Zetterstrom J. Immediate or delayed repair of obstetric anal sphincter tears a randomised controlled trial. BJOG 2008; 115: 857 865. 3. Norderval S, Nsubuga D, Bjelke C, Frasunek J, Myklebust I, Vonen B. Anal incontinence after obstetric sphincter tears: incidence in a Norwegian county. Acta Obstet Gynecol Scand 2004; 83: 989 994. 4. Sultan AH, Kamm MA, Hudson CN, Bartram CI. Third degree obstetric anal sphincter tears: risk factors and outcome of primary repair. BMJ 1994; 308: 887 891. 5. Pinta TM, Kylänpää ML, Salmi TK, Teramo KA, Luukkonen PS. Primary sphincter repair: are the results of the operation good enough? Dis Colon Rectum 2004; 47: 18 23. 6. Williams A, Adams EJ, Bolderson J, Tincello DG, Richmond DH. Effect of a new guideline on outcome following third-degree perineal tears: results of a 3-year audit. Int Urogynecol J Pelvic Floor Dysfunct 2003; 14: 385 389. 7. Sultan AH, Monga AK, Kumar D, Stanton SL. Primary repair of obstetric anal sphincter rupture using the overlap technique. Br J Obstet Gynaecol 1999; 106: 318 323. 8. Kairaluoma MV, Raivio P, Aarnio MT, Kellokumpu IH. Immediate repair of obstetric anal sphincter rupture: medium-term outcome of the overlap technique. Dis Colon Rectum 2004; 47: 1358 1363. 9. Norderval S, Oian P, Revhaug A, Vonen B. Anal incontinence after obstetric sphincter tears: outcome of anatomic primary repairs. Dis Colon Rectum 2005; 48: 1055 1061. 10. Fitzpatrick M, Behan M, O Connell PR, O Herlihy C. A randomized clinical trial comparing primary overlap with approximation repair of third-degree obstetric tears. Am J Obstet Gynecol 2000; 183: 1220 1224. 11. Fernando RJ, Sultan AH, Kettle C, Radley S, Jones P, O Brien PM. Repair techniques for obstetric anal sphincter injuries: a randomized controlled trial. Obstet Gynecol 2006; 107: 1261 1268. 12. Rygh AB, Körner H. The overlap technique versus end-to-end approximation technique for primary repair of obstetric anal sphincter rupture: a randomized controlled study. Acta Obstet Gynecol Scand 2010; 89: 1256 1262. 13. Nazir M, Carlsen E, Jacobsen AF, Nesheim BI. Is there any correlation between objective anal testing, rupture grade, and bowel symptoms after primary repair of obstetric anal sphincter rupture?: an observational cohort study. Dis Colon Rectum 2002; 45: 1325 1331. 14. Starck M, Bohe M, Valentin L. The extent of endosonographic anal sphincter defects after primary repair of obstetric sphincter tears increases over time and is related to anal incontinence. Ultrasound Obstet Gynecol 2006; 27: 188 197. 15. Norderval S, Markskog A, Rossaak K, Vonen B. Correlation between anal sphincter defects and anal incontinence following obstetric sphincter tears: assessment using scoring systems for sonographic classification of defects. Ultrasound Obstet Gynecol 2008; 31: 78 84. 16. Norderval S, Dehli T, Vonen B. Three-dimensional endoanal ultrasonography: intraobserver and interobserver agreement using scoring systems for classification of anal sphincter defects. Ultrasound Obstet Gynecol 2009; 33: 337 343. 17. Hals E, Oian P, Pirhonen T, Gissler M, Hjelle S, Nilsen EB, Severinsen AM, Solsletten C, Hartgill T, Pirhonen J. A multicenter interventional program to reduce the incidence of anal sphincter tears. Obstet Gynecol 2010; 116: 901 908. 18. Sultan AH, Kamm MA, Hudson CN, Thomas JM, Bartram CI. Anal-sphincter disruption during vaginal delivery. NEnglJMed 1993; 329: 1905 1911. 19. Kolodziejczak M, Sudol-Szopinska I, Stefanski R, Panorska AK, Gardyszewska A, Krasnodebski I. Anal endosonographic findings in women after vaginal delivery. Eur J Radiol 2011; 78: 157 159. 20. Zetterstrom J, Mellgren A, Jensen LL, Wong WD, Kim DG, Lowry AC, Madoff RD, Congilosi SM. Effect of delivery on anal sphincter morphology and function. Dis Colon Rectum 1999; 42: 1253 1260. 21. Starck M, Bohe M, Valentin L. Results of endosonographic imaging of the anal sphincter 2 7 days after primary repair of third- or fourth-degree obstetric sphincter tears. Ultrasound Obstet Gynecol 2003; 22: 609 615. 22. Lindqvist PG, Jernetz M. A modified surgical approach to women with obstetric anal sphincter tears by separate suturing of external and internal anal sphincter. A modified approach to obstetric anal sphincter injury. BMC Pregnancy Childbirth 2010; 10: 51.