Objective
The objective of the study was to correlate the presence of major levator ani muscle (LAM) injuries on magnetic resonance imaging (MRI) with fecal incontinence (FI), pelvic organ prolapse (POP), and urinary incontinence (UI) in primiparous women 6-12 months postpartum.
Study Design
A published scoring system was used to characterize LAM injuries on MRI dichotomously (MRI negative, no/mild vs MRI positive, major).
Results
Major LAM injuries were observed in 17 of 89 (19.1%) women who delivered vaginally with external anal sphincter (EAS) injuries, 3 of 88 (3.5%) who delivered vaginally without EAS injury, and 0 of 29 (0%) who delivered by cesarean section before labor ( P = .0005). Among women with EAS injuries, those with major LAM injuries trended toward more FI, 35.3% vs 16.7% ( P = .10) and POP, 35.3% vs 15.5% ( P = .09), but not UI ( P = 1.0).
Conclusion
These data support the growing body of literature suggesting that both EAS and LAM are important for fecal continence and that multiple injuries contribute to pelvic floor dysfunction.
Vaginal birth increases the likelihood that women will have pelvic floor dysfunction. Recent advances in pelvic floor imaging and neuromuscular testing provide evidence to pinpoint the specific anatomical and functional abnormalities caused by vaginal birth that are associated with increased occurrence of fecal incontinence (FI), pelvic organ prolapse (POP) and stress urinary incontinence (SUI).
Visible damage to the pubic portion of the levator ani muscle (LAM) is 1 specific injury associated with vaginal delivery. Pelvic imaging using magnetic resonance imaging (MRI) can improve our understanding of anatomic observations and clinical conditions that occur after childbirth. In a case-control study, young women with de novo SUI, 9 months after a first vaginal birth, were more than twice as likely to have a LAM defect imaged on MRI compared with women without SUI (28% vs 11%). However, using a different case-control mix of middle-aged women with and without stress SUI, the same researchers found similar rates of LAM injury in each group.
The relationship of LAM injury to other known injuries is not entirely clear. In a case-control study of primiparous women with and without de novo postnatal stress incontinence, LAM injury was more common in those who experienced an anal sphincter rupture at delivery compared with those who did not (62% vs 17%; odds ratio, 8.1). The effect of this injury on FI, POP, and SUI is not fully known.
The Childbirth and Pelvic Symptoms (CAPS) study assessed the prevalence and incidence of FI, POP, and urinary incontinence (UI) in a population of primiparous women after primary repair of clinically diagnosed third- or fourth-degree sphincter tears, compared with women who underwent vaginal delivery without a tear and those who underwent a cesarean delivery prior to labor. Two hundred thirty-five women in CAPS (approximately 25% of the CAPS population) underwent a single pelvic MRI at 6-12 months after first delivery, for the primary purpose of evaluating the anal sphincter complex. Because of the association between anal sphincter injury and LAM defects in previously described studies, frequent occurrence of levator defects in this cohort would be expected.
The goal of the current study was to determine the presence of major LAM defects on MRI (MRI positive) in women with clinically recognized third- or fourth-degree sphincter lacerations and to correlate its occurrence with FI, POP, and UI at 6-12 months postpartum in the entire cohort of primiparous women that participated in the CAPS imaging study.
Materials and Methods
Study population
The CAPS study was a multicenter, prospective cohort study conducted by the Pelvic Floor Disorders Network (PFDN) and supported by the National Institute of Child Health and Human Development. This study of primiparous women evaluated FI and UI symptoms at 6 weeks and 6 months after delivery as has been previously described. The CAPS study included 3 cohorts of primiparous women: group 1, vaginal delivery with a clinically recognized third- or fourth-degree anal sphincter tear (sphincter tear); group 2, vaginal delivery without a clinically recognized anal sphincter tear (vaginal control); and, group 3, cesarean delivery without labor (cesarean control). All women with a sphincter tear were invited to participate; the next woman delivered at each institution without a sphincter tear was invited to be a control, and therefore, many women without sphincter tears were not invited to participate.
A subset of the CAPS subjects was invited to undergo additional evaluations at 6-12 months postpartum. This included pelvic floor imaging using both MRI and endoanal ultrasound, the techniques of which have been described previously. The women who participated in this MRI study were similar to the overall CAPS population who did not participate in age, rates of episiotomy and vacuum or forceps intervention, and fecal incontinence prevalence and severity. However, more African-American women participated in this MRI study.
Physical examination using the pelvic organ prolapse quantification (POP-Q) score was performed. Other outcome data gathered 6-12 months after childbirth included the Medical Epidemiologic and Social Aspects of Aging (MESA) and FI Severity Index (FISI) questionnaires to assess UI and FI. A separate informed written consent was obtained from all imaging subjects in this institutional review board-approved study.
For the purposes of this study, POP is defined as maximal vaginal descent at or below the hymen, and UI as at least 1 response of urine leakage “sometimes” or more frequently on all 15 MESA items, whereas stress UI is limited to at least 1 response of urine leakage “sometimes” or more frequently on the first 6 MESA SUI items. FI is present if the subject’s response on the FISI was anything other than “never” to questions regarding accidental bowel leakage of liquid, solid, or mucus stool within 1 month. We chose a cut point of maximal vaginal descent at or below the hymen to define prolapse because of an emerging body of research that suggests that women are more likely to be symptomatic at this level than above the hymen.
Image reader training
Three board-certified radiologists with 2-10 years of subspecialty experience in abdominal imaging participated in an on-line, case-based training led by the originator of the LAM defect scoring system (J.O.D.). On-line training and teleconferencing was facilitated using Microsoft Office Live Meeting (2009; Microsoft Corp, Richmond, CA), enabling collective discussion and image review. A centralized data coordinating center for the PFDN facilitated these meetings. The training and testing data set used pelvic floor MRI images from the Michigan Pelvic Floor Research Group’s imaging library, which contains approximately 1000 MRI scans of subjects with pelvic floor dysfunction and asymptomatic volunteers. The on-line training was performed during 6, 2 hour, web-based teleconferences.
In total, 75 examinations obtained from the Michigan data library were reviewed for the training portion of the study, in groups of 15 studies per training session. In the initial session, readers were taught to identify the relevant landmarks and anatomy and to understand and utilize a LAM injury grading system. This technique is known for high interrater reliability in categorical grading of LAM defects, with a reported overall weighted kappa coefficient of 0.86 (95% confidence interval, 0.83–0.89) as determined by investigators at 1 site.
For the next 2 reader review sessions, the readers prospectively reviewed studies that were collated in a Microsoft PowerPoint (2004; Microsoft Corp, Redmond, WA) presentation, with 12 images presented per slide, in both axial and coronal orientations. The readers were blinded to clinical data and outcomes. Scoring and assessment were discussed on a teleconference call, but a final consensus score was neither collected nor recorded.
For the final 3 sessions (45 studies), the readers reviewed and scored the studies prospectively and independently and then submitted the scores prior to the group review and discussion of the examinations. When it was determined that training was sufficient, based on an analysis of interrater agreement and agreement within 1 point difference, the 2 readers who had the highest agreement with the trainer and between themselves went on to score the MRI examinations that were obtained in the CAPS imaging study cohort.
MRI image evaluation
Two radiologist readers at different clinical sites independently scored injury severity to each side of the LAM complex, blinded to all clinical and research data. Each reader received the study images on DVD, and used the Open-Source Apple-Macintosh–based OsiriX DICOM viewer platform to review the studies. The readings occurred over a 3 month period, and no additional training or feedback was provided to the readers once the MRI review started.
The right and left LAM muscles were identified on both axial and coronal imaging, and a categorical variable was assigned to grade the muscle defect as follows: 0, no defect; 1, less than half of muscle missing; 2, greater than half of muscle missing; and 3, muscle absent. The scores for each side were then summed. The threshold for significant injury was met when either the total score was 4 or greater or a unilateral grade 3 defect was present. Lesser scores have not been associated with increased occurrence of pelvic organ prolapse or at-risk obstetrical parameters. A total score of 4 or greater or grade 3 on either side was classified as a major LAM injury (MRI positive); and anything less was scored MRI negative. The MRI-negative category included women with both no injury and minor injuries.
Statistical analysis
The LAM defect score was dichotomized into no major injury (MRI negative) vs major injury (MRI positive), as described in previous text. Kappa coefficients were computed to evaluate interrater reliability.
A consensus score was then determined: if the 2 raters agreed on the dichotomous outcome, then the outcome was accepted; otherwise, a third rater (the trainer) evaluated the image set, also blinded to all clinical data, and the dichotomous outcome was determined by the majority (ie, the trainer and 1 of the original raters). The consensus score was used in all the analyses.
Analyses were performed for all subjects and then separately for each cohort of the study (the sphincter tear, vaginal control, and cesarean control groups). Because a large majority of major LAM injuries was found in 1 cohort, it was determined that cohort status would confound tests of association. Therefore, tests were reported separately for each cohort. Fisher’s exact test (2 tailed) was used to test for statistical significance; P values were reported with no adjustment for multiple tests. P < .05 was considered statistically significant. Analyses were performed using SAS 9.1 (SAS Institute, Cary, NC).
Results
Population
The population has been previously described. Briefly, the mean age and SD across all cohorts is 27.8 ± 6.1, with no significant difference in age by cohort ( P = .30). Approximately 75% of the population was white. Fifty-six percent of the sphincter tear group had undergone a forceps or vacuum delivery, compared with 11% of the vaginal control group. Additionally, the birthweight was significantly larger in the sphincter tear group as compared with the vaginal control group or cesarean control group, at 3602 ± 433, 3410 ± 378, and 3415 ± 434 g ( P = .005), respectively.
Reader reliability
The images from 206 MRI examinations were available for review. Twenty-nine of the 235 examinations were not evaluated because either the original data had become unreadable or the identifier on the images could not be verified. Reliability results are presented in Table 1 . The readers agreed on 189 scores and disagreed on 17 scores, for overall moderate agreement (Cohen’s kappa = 0.59). One reader identified more MRI-positive scans than the other (McNemar’s test, P = .013). The third reader then evaluated the 17 examinations with disagreement between the first 2 readers. This reader classified 5 scans as MRI positive, agreeing with 4 of the 14 scans classified by the first reader as MRI positive and the second reader as MRI negative, and with 1 of the 3 scans classified by the second reader as MRI positive but by the first reader as MRI negative. There were 121 scans that received a score of 0 from both readers (of 174 that were classified as MRI negative by both readers). Of the 32 scans that were MRI positive, 28 examinations had at least 1 side with a score of 3 (complete muscle loss).
