Diagnosis of Levator Ani Muscle Avulsion in Instrumented Delivery: Meta-analysis

Objective : The objective of this meta-analysis was to establish the rates of levator ani muscle (LAM) avulsion in patients after forceps delivery (FD) and vacuum delivery (VD) as assessed by 3D/4D transperineal ultrasound. Methods : A systematic search was performed in the MEDLINE, PubMed, Google Scholar


Introduction
Levator ani muscle (LAM) avulsion is defined as the disconnection of levator muscle fibers from their insertion on the inferior pubic ramus [1].It is present in 10-35% of women after a vaginal delivery [2].The most critical moment for the appearance of LAM avulsion is when the vertex is at station + 3 or + 4 [3], where the area of the levator hiatus acquires its greatest size [4].LAM avulsion is a diagnostic challenge: although it is sometimes exposed by a vaginal tear [5], most often is that it goes unnoticed during childbirth.Therefore, diagnostic imaging tests have been gaining relevance for the detection of LAM avulsion, such as magnetic resonance imaging (MRI) [6,7] and threedimensional (3D) ultrasound [8].Of the different modes of 3D ultrasound (introital and transperineal), transperineal or translabial ultrasound has shown a correlation with the diagnosis of LAM avulsion both clinically and by MRI [5].The abdominal probes used in transperitoneal ultrasound also have clear advantages over vaginal transducers due to their larger size and better ergonomic properties of the perineal structures [9].MRI is a less accessible test, is more expensive, and must be done in less anatomical postures, which leads us to think that transperineal ultrasound is the most desirable imaging test for the study of LAM avulsion.
As for the risk factors for LAM avulsion, instrumented delivery promotes it more than normal vaginal delivery (NVD).However, there are very different rates of LAM avulsion under forceps delivery (FD) vs. vacuum delivery (VD), possibly due to the lack of uniformity of the different study designs.Therefore, the objective of this metaanalysis was to establish the rates of LAM avulsion in patients after instrumented delivery (FD and VD), as assessed by 3D/4D transperineal ultrasound.

Study Protocol
For this meta-analysis, the guidelines collected for systematic reviews and meta-analysis of PRISMA 2020 were followed.The systematic search was performed in the MEDLINE, PubMed, Google Scholar, and Embase databases up to December 31, 2021.Our search was performed using the following medical subject heading (MeSH) terms: (Delivery, Obstetric/adverse effects OR Vacuum Extraction, Obstetrical/adverse effects OR Obstetrical Forceps/adverse effects OR Extraction, Obstetrical/adverse effects OR Delivery, Obstetric/methods OR Delivery, Obstetric/instrumentation) AND (Pelvic Floor/Diagnostic imaging OR Pelvic Floor/injuries OR Pelvic Floor Disorders/diagnostic imaging OR Soft Tis- sue Injuries/etiology OR Perineum/injuries OR Anal Canal/injuries.No language restrictions were applied.No ethical approval was needed since it was based on previously published medical literature.

Selection of Studies
Those studies that met the following criteria were included: -They determined the rate of LAM avulsion with VD or FD.
-They included only primiparous patients.
-The data are presented in such a way that we could calculate the odds ratios (ORs) and the 95% confidence intervals (CIs).

Data Extraction
The information extracted from the selected articles (type of study; time of postpartum ultrasound; number of NVDs, VDs, and FDs; and the avulsion rates in NVD, VD, and FD) were included in an Excel calculation.Possible discrepancies arose in the selection of the studies as well as the information extracted from them, and a final consensus was reached between authors JAGM, AFP, and JASB.Those articles belonging to the same working group were assessed for the time periods in which the studies were carried out to verify that there was no bias due to the involuntary reanalysis of the same study.

Statistical Analysis
The statistical software jamovi version 2.2 (The jamovi project, https://www.jamovi.org) was used to analyze the data.The number of avulsions was summarized as incidence rates depending on the type of delivery performed (NVD, VD, or FD).A binary controlled study was done to calculate the number of LAM avulsions in the NVD group, the VD group, and the FD group.ORs, their logarithms, and their 95% CIs were determined to measure effect sizes.The results of all studies (ORs) were aggregated using a random-effects model.The ORs and prediction intervals were also obtained for the real results for the study of heterogeneity, and the estimation of τ 2 is provided, as are the Q value for heterogeneity and the I 2 statistic.Forest graphs were drawn.Results were analyzed using the logarithm of the OR.If heterogeneity was observed (τ 2 > 0), the prediction interval for the real results was assessed.If the studies were atypical and/or influential in the context of the model, the studentized residuals and Cook's distances were applied.Those studies with a standard normal distribution and a studentized residual greater than the (100 × (1 -0.05/(2k)) × r) th percentile were considered atypical.Those studies with Cook distances greater than the median plus 6 times the interquartile range of the Cook distances were considered influential.To check for asymmetry in the funnel plot, the rank correlation test and the regression test were used (the standard error of the results was used as a predictor).

Assessment of Risk of Bias
The data on the study design, the patients included, the selection process, the characteristics of the patients, and the statistical method were analyzed.Two independent reviewers (JAGM and AFP) assessed the risk of bias and the quality of the included studies, interpreted the findings of the systematic review, and determined the potential bias resulting from the heterogeneity of the study.

Results
The search included 1225 studies that met the MeSH criteria, of which 26 were included in the review because they met the inclusion criteria (Fig. 1).Of all the excluded studies, we highlight the study by Greenbaum et al. [10] which, although it met the inclusion criteria, was excluded because it did not mention the type of instrumentation performed during vaginal delivery.

Fig. 2. Analysis of the general levator ani muscle (LAM) avulsion rates between vacuum delivery (VD) (including the exploration of Halle et al. [35] at 6 weeks postpartum) and normal vaginal delivery (NVD).
Of the 13 studies included, log odds ratios ranged between -0.41 and 3.14, the majority being positive (85%).The estimated average log OR was 0.70 (95% CI: 0.29-1.11), the mean being different from 0 (z = 3.37, p = 0.0008).The Q test showed heterogeneous results (Q = 23.87,p = 0.02, τ 2 = 0.25, I 2 = 48.53%),with a prediction interval of 95% from -0.36 to 1.76.Therefore, although the average result is estimated to be positive, in some studies the true result may be negative.The studentized residuals were all within ± 2.89 (without indications of outliers).Evaluating Cook's distances, none of the studies could be considered too influential.There was no asymmetry in the funnel plot in the rank correlation (p = 0.95) or in the regression test (p = 0.54).
with a joint OR for FD vs. VD of 2.36 (95% CI: 1.46-3.84).Based on our findings, we agree with the results of a previous systematic review [37], where it was specified that FD was associated with a higher incidence rate and severity of LAM avulsion compared to NVD [37].However, there is controversy about whether FD presents higher rates of LAM avulsion than VD.In fact, a recent meta-analysis found Kielland's forceps to be safe and favored their use over vacuum-assisted delivery [38].However, that metaanalysis did not study LAM.One of the main problems that we found in the previous reviews is that they included and compared different imaging techniques for the study of LAM avulsion [39,40], such as MRI and 3D introital ultrasound.In addition, the vast majority of studies included in those reviews did not specify the type of instrument used (the type of forceps or vacuum) or how it was applied during delivery.
In this meta-analysis, we analyzed the type of instrument specified in the study.The most commonly used vacuum was the 50 mm bird's cup [19,[24][25][26]29,30,[32][33][34].Only one study made reference to the rate of LAM avulsion in deliveries with another type of vacuum (Kiwi vacuum) [32], so there were no differences in the LAM avulsion rates compared to those produced by Malmstrom's vac-

Fig. 3. Analysis of the general levator ani muscle (LAM) avulsion rates between vacuum delivery (VD) (including the exploration of Halle et al. [35] 1 year after delivery) and normal vaginal delivery (NVD).
Of the 13 studies included, log odds ratios ranged between -0.41 and 3.14, with the majority being positive (85%).The estimated average log OR was 0.66 (95% CI: 0.27 to 1.05), the mean being different from zero (z = 3.31, p = 0.0009).The Q test showed nonsignificant heterogeneous results (Q = 20.91,p = 0.05, τ 2 = 0.19, I 2 = 41.36%), with a prediction interval of 95% from -0.29 to 1.61.Therefore, although the average result is estimated to be positive, in some studies the true result may be negative.The studentized residuals were all within ± 2.89 (without indications of outliers).Evaluating Cook's distances, none of the studies could be considered too influential.
There was no asymmetry in the funnel plot in the rank correlation (p = 0.59) or in the regression test (p = 0.39).uum, with an OR of 0.977 (0.426; 2.241; p = 0.957) and an adjusted OR of 2.90 (0.691; 12.20; p = 0.146).Something similar was true of the studies that specified the type of forceps used during the FD, as most describe the rate of LAM avulsion caused by only one type of forceps (Kielland's forceps) [28][29][30][31]33,34,36].Only one study included different forceps [19], with a small number of participants (20 Anderson forceps and 22 Wrigley forceps), and did not compare the rate of LAM avulsion between the two types.
Only five studies studied the rate of LAM avulsion according to the way in which the instrument was applied during instrumented delivery [26,28,29,33,36].It was determined that the number of vacuum tractions necessary to complete the fetal extraction was not associated with a higher rate of LAM avulsion [26], nor was the position of the fetal head (anterior, posterior, or transverse) at the time of vacuum application [26].In the case of FD, there are controversies.It has been described that the rate of LAM avulsion is not increased when rotational forceps are used (OR: 1.5 [0.6-3.6];p: not significant), in the correction of asynclitism (OR: 0.8 [0.3-1.9];p: not significant), or according to the station of the fetal head at the time of placement of the forceps (OR: 2.0 [0.8-5.1];p: not significant) [28].However, another study established that rotational forceps were associated with avulsion, with an OR of 2.45 (95% CI 1.22- Of the 8 studies included, log odds ratios ranged between 1.28 and 2.62, the majority being positive (100%).The estimated average log OR was 1.71 (95% CI: 1.33 to 2.09), the mean being different from zero (z = 8.78, p < 0.0001).The Q test showed that there was no significant amount of heterogeneity (Q = 3.86, p = 0.796, τ 2 = 0.00, I 2 = 0.00%).The study by Caudwell-Hall et al. [23] had a relatively large weight compared to the rest.
The studentized residuals were all within ± 2.73 (without indications of outliers).Evaluating Cook's distances, none of the studies could be considered too influential.There was no asymmetry in the funnel plot in the rank correlation (p = 0.39) or in the regression test (p = 0.86).4.93) [36].One study analyzed whether the disarticulation of the forceps before the delivery of the fetal head could benefit the rate of LAM, and it found a crude OR (without disengagement vs. disengagement) for avulsion of 0.90 (95% CI 0.49-1.67;p = 0.757) and an adjusted OR (adjusted for maternal age, induced labor, epidural period, second stage of labor, perineal tear, and fetal head circumference) of 0.82 (95% CI 0.40-1.69;p = 0.603) [31].However, when comparing the rates of LAM avulsion between VD and FD considering the position (anterior or other) and the station (low or medium instrumentation) of the fetal head at the time of instrument placement, no differences were found [29,33].
Another noteworthy aspect of the current literature is the time when transperineal ultrasound is performed to diagnose LAM avulsion.It is advisable not to make a diagnosis of LAM avulsion until 3 months after delivery to avoid diagnostic errors [41], and this recommendation was qualified by a recent meta-analysis: Rusavy et al. [37] advised that the diagnosis of LAM avulsion be made 6 months after delivery, or 12 months after FD.On the other hand, different authors have mentioned that LAM avulsion can disappear over time [42][43][44][45].In fact, we recently reported that partial avulsions can present an improvement toward an intact LAM within 9 months postpartum [46].There are two types of avulsions: Type I LAM avulsion is a lesion that can recover over time, and type II LAM avulsion Therefore, although the average result is estimated to be positive, in some studies the true result may be negative.The studentized residuals were all within ± 2.89 (without indications of outliers).Evaluating Cook's distances, none of the studies could be considered too influential.There was no asymmetry in the funnel plot in the rank correlation (p = 0.68) or in the regression test (p = 0.28).
is not recoverable [47].This aspect has not only been described by transperineal ultrasound [47] but has also been established in anatomical studies and by MRI [48].In this meta-analysis, the vast majority of studies performed ultrasound examinations up to 6 months after delivery [11][12][13][14][15][16][17]19,20,22,23,[25][26][27][28][29][30][31][32][33][34][35], but they did not assess its evolution over time.LAM avulsion produces an alteration in the support of the pelvic floor structures, however this fact should not influence the decision to implement a delivery.In fact, a recent Cochrane review establishes that there is no instrumentation that guarantees maternal and fetal safety [49].However, instrumentation at delivery will depend on the skill of the operator, the selection of available instruments, and the clinical setting [49].
The main strength of this meta-analysis lies in its inclusion criteria: Only those studies with primiparous women after instrumented deliveries were included, and LAM avulsion was diagnosed only by 3D/4D transperineal ultrasound.One shortcoming is that the time of the ultrasound exam was not the same in all the articles, and most studies do not specify the type of instrument and the techniques used.In addition, there are no clinical trials that could provide solid evidence for the question that motivated this meta-analysis.

Conclusions
In conclusion, delivery instrumented with forceps or vacuum favors LAM avulsion.Current data cannot establish whether this injury is attributable to the specific instrument or to the type of delivery involved in the instrumentation itself.