Evidence-Based Strategies for Reducing Cesarean Section Rates: A Meta-Analysis
Nils Chaillet PhD,
Nils Chaillet and Alexandre Dumont are at the Research Centre of Sainte-Justine Hospital and in the Department of Obstetrics and Gynaecology, University of Montréal, Montréal, Quebec, Canada.
Nils Chaillet, PhD, Centre de recherche de l’hôpital Sainte-Justine, Département Obstétrique et Gynécologie, Université de Montréal, 3175 chemin de la Côte Ste-Catherine, Local 4986-B, Montréal, Québec, Canada H3T 1C5.
This study was supported by the Research Centre of Sainte-Justine Hospital, Montreal, Quebec, Canada. The sponsors of the study had no role in study design, data collection, data analysis, data interpretation, or writing the report.
Nils Chaillet, PhD, Centre de recherche de l’hôpital Sainte-Justine, Département Obstétrique et Gynécologie, Université de Montréal, 3175 chemin de la Côte Ste-Catherine, Local 4986-B, Montréal, Québec, Canada H3T 1C5.
ABSTRACT: Background: Canada’s cesarean section rate reached an all-time high of 22.5 percent of in-hospital deliveries in 2002 and was associated with potential maternal and neonatal complications. Clinical practice guidelines represent an appropriate mean for reducing cesarean section rates. The challenge now lies in implementing these guidelines. Objectives of this meta-analysis were to assess the effectiveness of interventions for reducing the cesarean section rate and to assess the impact of this reduction on maternal and perinatal mortality and morbidity. Methods: The Cochrane Library, EMBASE, and MEDLINE were consulted from January 1990 to June 2005. Additional studies were identified by screening reference lists from identified studies and expert suggestions. Studies involving rigorous evaluation of a strategy for reducing overall cesarean section rates were identified. Randomized controlled trials, controlled before-and-after studies, and interrupted time series studies were evaluated according to Effective Practice and Organisation of Care Group criteria. Results: Among the 10 included studies, a significant reduction of cesarean section rate was found by random meta-analysis (pooled RR = 0.81 [0.75, 0.87]). No evidence of publication bias was identified. Audit and feedback (pooled RR = 0.87 [0.81, 0.93]), quality improvement (pooled RR = 0.74 [0.70, 0.77]), and multifaceted strategies (pooled RR=0.73 [0.68, 0.79]) were effective for reducing the cesarean section rate. However, quality improvement based on active management of labor showed mixed effects. Design of studies showed a higher effect for noncontrolled studies than for controlled studies (pooled RR = 0.76 [0.72, 0.81] vs 0.92 [0.88, 0.96]). Studies including an identification of barriers to change were more effective than other interventions for reducing the cesarean section rate (pooled RR = 0.74 [0.71, 0.78] vs 0.88 [0.82, 0.94]). Among included studies, no significant differences were found for perinatal and neonatal mortality and perinatal and maternal morbidity with respect to the mode of delivery. Only 1 study showed a significant reduction of neonatal and perinatal mortality (p < 0.001). Conclusions: The cesarean section rate can be safely reduced by interventions that involve health workers in analyzing and modifying their practice. Our results suggest that multifaceted strategies, based on audit and detailed feedback, are advised to improve clinical practice and effectively reduce cesarean section rates. Moreover, these findings support the assumption that identification of barriers to change is a major key to success. (BIRTH 34:1 March 2007)
The World Health Organization states that no region in the world is justified in having a cesarean section rate greater than 10–15 percent (1). The cesarean section rate in Canada increased steadily from 17.5 percent of deliveries in 1994–1995 to 21.2 percent in 2000–2001 (2). Canada’s cesarean section rate reached an all-time high of 22.5 percent of in-hospital deliveries in 2001–2002, according to a recent report by the Canadian Institute for Health Information (3). Moreover, cesarean delivery was associated with a high maternal and neonatal complication rate and increased health costs (4–8).
Clinical practice guidelines implement the best evidence into practice and represent an appropriate means for reducing cesarean section rates. The development of guidelines for obstetric care has increased in recent years in developed countries because of the leadership of nongovernmental, professional, and expert organizations and national and international agencies. The challenge now lies in implementing these guidelines (9–13).
Strategies for implementing guidelines produce different results and are classified into 3 levels of effectiveness (9,14): generally ineffective, mixed effects, and generally effective (9–13). With respect to general medical practice, ineffective interventions are mailing (i.e., using postal services) (15–17) and didactic, traditional continuing medical education (18–20). Mixed-effects interventions are opinion leader (21–23), audit and feedback (23–25), and continuous quality improvement (26). Interventions that are generally effective are manual or computerized reminders (27,28), academic detailing (29,30), and multifaceted interventions (31–33).
The premise of this study is that effective strategies for reducing cesarean section rates differ from those of other medical specialties because of different forces and variables that influence professional behavior in obstetrics. Moreover, each clinical environment presents specific challenges in the implementation of an intervention. The identification of specific barriers to change represents a new challenge in developing specific interventions adapted to specific clinical environments (4,8,13,33–35).
The objectives of this meta-analysis are, first, to assess the effectiveness of interventions for reducing the cesarean section rate, second, to determine if an identification of barriers to change can improve the effect of interventions, and, third, to assess the impact of cesarean section rate reduction on maternal and perinatal mortality and morbidity.
Data source and identification of studies
Together with a medical librarian, 1 investigator (N.C.) conducted multiple searches of the Cochrane Library, in the Registry of Controlled Trials, EMBASE, and MEDLINE databases in publication type category from January 1990 to June 2005, using the following MeSH terms: “cesarean section,”“trial of labor,” and “vaginal birth after cesarean.” These terms were then combined with text words: “mailing,”“education,”“continuous education,”“continuing education,”“audit,”“peer review,”“second opinion,”“opinion leader,”“academic detailing,”“educational outreach,”“continuous quality improvement,”“management of labor,”“community-based model,”“quality improvement,”“computer reminder,”“paper reminder,”“reminder,”“multifaceted strategy,”“multiple strategy,”“tailored intervention,”“multifaceted intervention,” or “multiple intervention.” Additional studies were identified by screening reference lists from identified studies and from expert suggestions.
Studies involving rigorous evaluation of a strategy for reducing overall cesarean section rates were identified. Randomized controlled trials, quasi-randomized controlled clinical trials, controlled before-and-after studies, and interrupted time series studies were evaluated. Usually, the best source of evidence is considered to be controlled trials; however, the rigorous, interrupted time series studies design may be a reliable source of information (36).
The standard Cochrane and Effective Practice and Organisation of Care criteria were used to assess inclusion and quality of studies with respect to design and to enhance internal validity of the analysis (9,14,37). The minimum inclusion criteria included objective measurement of performance and relevant and interpretable data presented or obtainable. Two additional minimum criteria for controlled before-and-after studies (contemporaneous data collection and appropriate choice of control site) and for interrupted time series studies (intervention time clearly defined and at least 3 data points before and 3 after the intervention) are included.
Eight quality criteria were assessed for controlled studies: protection against selection bias (randomized controlled trials and quasi-randomized controlled clinical trials), characteristics of study and control practitioners (controlled before-and-after studies), exclusion bias, follow-up of patients, detection bias, baseline performance, reliability of first outcome, and protection against contamination. Specific quality criteria were assessed for interrupted time series studies: protection against secular changes, appropriate data analysis, reason for the number of point preintervention and postintervention, specification of the shape of the intervention effect, protection against detection bias, completeness of data set, and reliable first outcome. Each criterion is noted “done,”“not clear,” or “not done.” Studies classified as “fair” and “good” quality, according to the Cochrane and Effective Practice and Organisation of Care quality scale (14,37), were included in this study.
Data extraction and analysis
After the first screening, 2 reviewers (N.C. and A.D.) independently abstracted specific information from full-text studies according to standardized data extraction checklist items derived from the Cochrane and Effective Practice and Organisation of Care checklist (14). Discordances between the 2 reviewers were resolved by consensus. When information about individual observations over time was reported only graphically in the original paper, the data set was derived by computer scanning the figures (14,38).
For interrupted time series, effect size was estimated reporting prerate and postrate interventions (14,37). Inclusion of interrupted time series in meta-analysis is a statistical challenge because the change might be just due to the trend and not due to the intervention. To control this bias and validate the intervention effect, data were analyzed using an autoregressive integrated moving average (ARIMA) model to isolate the effect of the intervention from existing time trends (14,38,45).
Begg’s statistics and funnel plot were computed for assessed publication bias according to Cochrane procedures (14,37). Q and I2 tests, meta-regression models, and subgroups analysis were selected for addressing heterogeneity from adjusted risk ratios (14,37). An ARIMA model was computed using SPSS version 11.0, and meta-analysis and meta-regression were computed using Stata version 7.0 (46,47).
A total of 831 studies corresponding to our search strategy were identified for the period 1990–2005 (Fig. 1). Of these, 810 were excluded based on eligibility criteria outlined in the Methods section and Fig. 1. The full-text articles for the remaining 21 citations were retrieved. Four additional articles were obtained from reference lists and expert suggestions, bringing the total number of identified studies to 25. After review of the full-text articles using the eligibility criteria, 11 studies remained and were evaluated further for quality. One interrupted time series (48) was excluded for “poor” quality based on Cochrane and Effective Practice and Organisation of Care quality criteria (assessment of reduction of cesarean section rate by regional statistics). The reasons for exclusion of the 14 studies were data not obtainable for overall cesarean section rates (23,49) (2 controlled studies), data not presented or study comprising fewer than 500 women (24,50–52) (3 studies), and lack of 3 data points before and after the intervention for interrupted time series (53–60) (9 studies). In all, 10 studies met all the inclusion and quality criteria.
Characteristics of interventions
Tables 1–3 present information about the characteristics and effect size of each strategy. There were 2 cluster randomized controlled trials (40,61), 3 randomized controlled trials (64–66), and 5 interrupted time series studies (62,63,67–69). A prospective identification of the barriers to change in order to improve implementation of intervention was performed in 40 percent of studies. Among the 10 included studies, 4 involved audit and feedback, 4 involved quality improvement, and 2 involved multifaceted strategies. The strategy based on a mandatory second opinion (40) was considered as an audit and feedback. Quality improvement programs were based on active management of labor (64,65,67) and community-based continuity of care (66). Multifaceted programs included physician and public education, physician peer review, and physician, hospital payment, and malpractice reform (68); the other program included physician education and audit and feedback (69). The total number of included women was 776,909, with 686,334 from randomized controlled trials.
Table 1. Nature and Description of Strategies Used for Reducing Cesarean Section Rate
Nature of Strategies
Description of Programs for Reducing Cesarean Section Rate
ACOG = American College of Obstetricians and Gynecologists; AML = active management of labor; EFM = electronic fetal monitoring; VBAC = vaginal birth after cesarean section.
Policy of mandatory second opinion systematically before cesarean section. Both physicians discussed the case in relation to the guidelines. After this process, the attending physician made the final decision. The guidelines were made available for all physicians at intervention hospitals.
External peer reviews were done by ACOG-trained teams of 4 physician and nurse reviewers, who visited the hospital, interviewed key staff members, and reviewed 100 labor and delivery records to assess the quality of care. Review teams provided feedback through an exit interview and a written summary of findings and recommendations.
The peer review included precesarean consultation and postcesarean surveillance. A weekly departmental cesarean indication conference was organized to present every cesarean case. A second opinion by a consultant was required for all cesarean sections. VBAC was encouraged. Feedback was provided by conferences.
The medical audit cycle takes the concept of medical audit a stage further. Clinical standards are established, current practice is compared, modification of management takes place, and medical audit is continued with emphasis on completing this “feedback loop.” Monthly medical audit and meetings where all results were reviewed were held.
Active management of labor protocol had 3 components: one-to-one nursing care, standardized criteria for the diagnosis of labor, and management of labor based on ACOG guideline that encourages early amniotomy, cervical examination performed every 2 hr, and early diagnosis of inefficient uterine action treated with oxytocin.
Program of active management of labor that encourages early amniotomy, early diagnosis of slow progress in labor, and the use of higher than usual doses of oxytocin. The fetal heart was monitored electronically. Use of intrauterine pressure catheters was encouraged. Umbilical cord arterial and venous blood was obtained routinely.
Continuity of midwifery care was a focus of the community-based model. The emphasis was on continuity of care (a consistent team approach) rather than caregiver (the same midwife). An informal evening at which the women could meet the midwives was held every 2 mo. One midwife was always on call for women in labor.
Three initiatives were undertaken. VBAC was more strongly encouraged, and EFM was routinely used. Quality management initiative was introduced because of suspected differences among physicians, and annual report was distributed. A program of active management of labor for term nulliparous women was undertaken.
Proposed strategies include physician and public education about maternal and fetal benefits of vaginal delivery; practice guidelines for management of labor, with physician peer review and feedback, and corrective measures when guidelines are not followed; and physician, hospital payment, and malpractice reform.
Clinical guidelines’ changes were made to facilitate lowering the cesarean section rate. The nursing staff was given extensive education in the techniques of AML and EFM. Monthly summary statistics were generated by computer, analyzed, and reported to the medical and nursing staffs. A confidential report was delivered to physicians every 6 mo.
Table 2. Study Settings
Study and Country
Study of Barriers
No. of Participants
AF = audit and feedback; C-RCT = cluster randomized controlled trial; EDUC = educational strategy; ITS = interrupted time series; mixed = medical and paramedical practitioners; MULTI = multifaceted strategy; OG = obstetricians and gynecologists; QUAL = quality improvement; RCT = randomized controlled trial; REF = physician, hospital payment, and malpractice reform.
Number of cesarean sections and deliveries of cluster trials were determined taking into account the cluster effect (event/design effect). IF = 1 + (m− 1) × ICC. IF is the design effect, m is the average cluster size in the trial, and ICC is the intracluster correlation coefficient.
For cluster trials: ARR (absolute risk reduction) = (rate change in intervention group) − (rate change in control). ARR for randomized controlled trials (RCTs) = (rate change in intervention group) − (rate change in control group).
For cluster randomized controlled trials, risk ratio was adjusted for baseline imbalance. Adjusted risk ratio = (mean postrate intervention/mean postrate control)/(mean baseline rate intervention/mean baseline rate control). For RCTs, risk ratio = (mean rate intervention/mean rate control).
Four controlled trials and 1 interrupted time series study met all criteria and were rated as “good quality”(40,64–66,68). The remaining studies were rated as “fair.” The main reasons for fair quality rating for randomized controlled trials were that baseline measures were not reported or unclear (61) and protection against contamination was not clear (61). Four interrupted time series studies were rated fair because no ARIMA model or time series regression models were used for data analysis (62,63,67,69).
Adjusted from an existing time trend, ARIMA model analysis showed a significant difference between baseline and intervention for the 5 included interrupted time series (p < 0.05; Table 3). Thus, time trends did not explain the observed changes, and the interrupted time series studies were included in the random meta-analysis.
Among included studies, we found a significant reduction of cesarean section rates (pooled RR, 0.81; 95% CI, 0.75–0.87; p < 0.00001; Fig. 2). A formal test for heterogeneity gave a significant result (I2= 87.6%, p < 0.00001), and a random effects model was used. We noted a relative diminution of cesarean section rate of 19 percent (relative risk reduction, 19%; 95% CI, 13–25%).
The Begg’s funnel plot showed no striking evidence of publication bias. Neither Begg’s adjusted rank correlation test (p = 0.86, continuity corrected) nor Egger’s regression asymmetry test (p = 0.83) gave a statistically significant result.
Heterogeneity and subgroups analysis
A meta-regression was performed for exploring heterogeneity among studies. Various covariables, such as publication and study years, study design, strategy used, country, identification of barriers to change, type of providers, study quality, source of guidelines, total number of women, study and intervention duration, number of centers, and initial cesarean section rates were assessed. Three covariables were highly significant and explained 61–80 percent of the variation (strategy used, 61%, p = 0.003; study design, 72%, p < 0.001; and identification of barriers, 80%, p < 0.001). Most of the variation was explained by these 3 covariables (p < 0.001).
Subgroups analyses were performed for these 3 covariables (Table 4). Pooled risk ratios for audit and feedback, quality improvement, and multifaceted strategy were significant for reducing cesarean section rate. However, a significant heterogeneity, due to design, was found for audit and feedback (I2= 81.7%, p < 0.01). Indeed, subgroups analysis for audit and feedback stratified by design showed a significant homogeneity (I2 for audit and feedback [controlled] = 0%, p = 0.97; I2 for audit and feedback [interrupted time series] = 0%, p = 0.32). Multifaceted strategy presented a highly significant reduction of cesarean section rates (pooled RR, 0.73; 95% CI, 0.68–0.79; p < 0.00001) with a relative risk reduction of 27 percent (95% CI, 21–32%).
Table 4. Pooled Risk Ratios (RR) Stratified by Strategy, Design, and Identification of Barriers, Random Effect Model
Pooled risk ratio for controlled and noncontrolled studies was in favor of a reduction of cesarean section rates (Table 4). However, noncontrolled studies had a greater reduction in cesarean section compared with controlled studies (pooled RR = 0.92 [0.88, 0.96], p = 0.00037 for controlled studies and 0.76 [0.72, 0.81], p < 0.00001 for noncontrolled studies). A significant heterogeneity, explained by the study of barriers, was found for noncontrolled studies (I2= 83.8%, p = 0.02). Indeed, subgroups analysis for noncontrolled studies stratified by study of barriers showed a significant homogeneity (I2 for noncontrolled studies with identification of barriers = 46.3%, p = 0.13; only 1 noncontrolled study had no identification of barriers).
Pooled risk ratios were more significantly in favor of a reduction of cesarean section rate among the interventions including an identification of barriers to change, compared with other interventions (Table 4). Pooled risk ratio was 0.74 [0.71–0.78] (p < 0.00001) for interventions with identification of barriers and 0.88 [0.82, 0.94] (p < 0.00042) for standard interventions. A test for heterogeneity did not give a significant result for interventions including an identification of barriers (I2= 46.3%, p = 0.13). For interventions with no identification of barriers, a significant heterogeneity, due to 1 noncontrolled study, was detected (I2= 64.0%, p = 0.02). Indeed, subgroup analysis for controlled studies with no identification of barriers showed a significant homogeneity (I2= 0%, p = 0.49).
Morbidity associated with reduction of cesarean section rates
In this study, the relative cesarean section rate was reduced by 19 percent (pooled RR = 0.81; 95% CI, 0.75–0.87). The major part of this reduction was explained by a significant reduction of cesarean section for the following indications: dystocia (40,62,63,67–69), repeat cesarean section (62,63,67,68), fetal distress (40,62,69), and maternal indications (40).
Among included studies, no significant differences were found for stillbirth rate (40), perinatal and neonatal mortality (40,68,69), neonatal and maternal admission to intensive care unit (40,63,66–68), and perinatal and maternal morbidity (66,68,69). Only 1 study showed a significant reduction of neonatal and perinatal mortality (p < 0.001) associated with a reduction of cesarean section rates (67). For this study, overall cesarean section rate was reduced by 7 percent, the neonatal mortality was reduced from 10.3 to 3.8 per 1,000 live births between 1986 and 1991 (RR = 0.37; 95% CI, 0.21–0.64), and the perinatal mortality was reduced from 19.5 to 10.3 per 1,000 in the same period (RR = 0.53; 95% CI, 0.37–0.75).
The findings of this meta-analysis showed that audit and feedback, quality improvement, and multifaceted strategies were effective in changing clinical practice and reducing cesarean section rates. Pooled risk ratio was more effective to reduce the cesarean section rate for multifaceted strategies than for audit and feedback. However, both multifaceted interventions included audit and feedback as an essential part of their program.
Audit and feedback seeks to improve patient care and outcomes by systematic review of care against explicit criteria and the implementation of change with feedback (13). Audit and feedback is an effective strategy to reduce the cesarean section rate, but this diminution is limited when it is used alone. Cesarean section rate was only reduced by 13 percent when audit and feedback was used alone but reduced by 27 percent when audit and feedback was used in a multifaceted strategy. This difference can be partially explained by a contamination bias for 1 cluster randomized controlled trial (61) and by the lack of identification of barriers to change for 3 out of 4 studies.
Quality improvement meta-analysis presented a significant pooled risk ratio. However, this effect was due only to a noncontrolled study based on active management of labor (67). Although the 3 randomized controlled trials were in favor of a reduction of cesarean section rate, the 2 randomized controlled trials based on active management of labor were nonsignificant (64,65). One study based on a continuity-of-care program showed a significant reduction of overall cesarean section rates but no significant differences in subgroups analysis because the study was potentially underpowered (66). Despite significant results in the meta-analysis, quality improvement based on active management of labor seems to present mixed effects, but these results could be explained by a potential lack of statistical power for the randomized controlled trials. In a systematic review, quality improvement based on nonrandomized studies can improve outcomes of care; but the few randomized studies suggest no impact on clinical outcomes and no evidence of organization-wide improvement in clinical performance (26). Quality improvement interventions have highly variable effectiveness and are extremely dependent on the context in which they are used and the way they are implemented (26).
Multifaceted strategies were effective and showed the strongest reduction of cesarean section rate. No significant heterogeneity was found. A multifaceted intervention has been defined as one that involves 2 or more interventions targeting different barriers to change (13). Included multifaceted interventions were based on guideline education, audit and feedback, hospital payment and malpractice reform, and identification of barriers to change. Detailed feedback has been identified as a major part of the tailored programs for changing clinical practice and reducing cesarean section rates (68,69). The findings of this study confirmed that audit and feedback strategy must be included in a change program, targeting different barriers and promoting a detailed feedback, to be fully effective.
The relative reduction of the cesarean section rate was higher for noncontrolled studies than for controlled studies. However, of the 5 controlled trials, 1 presented a contamination bias (61) and 3 based on quality improvement presented a lack of statistical power (64–66). When controlled for potential confounding variables, the effect size of 1 randomized controlled trial became significant (OR = 0.57; 95% CI, 0.36–0.95) (65). Moreover, the 2 cluster randomized controlled trials were based on audit and feedback, and no controlled trial was based on multifaceted strategies. These biases could explain the smaller reduction of cesarean section rate for controlled trials in this meta-analysis. Additional cluster randomized studies are necessary to validate the significant effect of multifaceted strategies.
The identification of facilitators and barriers is a necessary stage to improve local definition and adaptation of the final structure and processes of the intervention. These barriers and facilitators are related to, first, external changes influencing the practice environment; second, the practice environment, including unit leadership, policy, and availability of equipment; third, the potential adopters; and, fourth, strategies used to promote uptake of the guideline recommendations (35). In this meta-analysis, studies including identification of barriers improved the implementation of interventions, and a higher reduction of cesarean section rate was noted. These findings demonstrated that those wishing to reduce the cesarean section rate should consider using such influences.
Finally, 2 robust controlled studies, based on opinion leader and computerized audit and feedback, were not included because data were not obtainable (23,49). These studies showed a significant reduction of overall cesarean section rates. Lomas et al noted no significant changes for maternal death, stillbirth rate, and Apgar scores less than 7 at 1 minute between the control and the opinion leader groups and an Apgar score at 5 minutes that was significantly better (p < 0.001) in the opinion leader group (23). One case of ruptured uterus occurred in the opinion leader group but not in the planned cesarean group (23). These results seem to confirm that cesarean section rates can be safely reduced.
Many included studies were published from the middle 1990s. During recent years, obstetric practices may have changed. Today, obstetric practitioners seem more risk averse about policies concerning induction or vaginal birth after cesarean, and the efficiency of interventions could be different. However, 2 recent included studies (40,62) showed that cesarean section rates could be safely decreased with interventions based on audit and feedback. Nevertheless, the type of interventions should be adapted considering the current trends in obstetrics practice.
Moreover, in this review, perinatal morbidity was assessed by 3 noncontrolled studies and measured by birth trauma, low Apgar scores and thick meconium (68,69), and Apgar scores and birthweight (66). Maternal morbidity was assessed by 1 study (66). However, several authors showed a significant increase in uterine ruptures for women undergoing a trial of labor after cesarean delivery (6,70). In the literature, the risk of urinary incontinence is higher among women who have had vaginal deliveries than among women who have had cesarean sections (71,72), but the risk of complications in a future pregnancy is higher among women who have had cesarean sections (6). Thus, the findings in this review about maternal morbidity must be considered with caution. Further studies are necessary to assess the evolution of perinatal and maternal morbidity with respect to the mode of delivery.
The cesarean section rate can be safely reduced by complex interventions that involve health workers in analyzing and modifying their practice. Our results suggest that multifaceted strategies, based on audit and detailed feedback, are advised to improve clinical practice and effectively reduce the cesarean section rate. Moreover, these findings support the assumption that identification of barriers to change is a major key to success.
The authors would like to thank the team of the Clinical and Evaluative Research Unit in Perinatality, Sainte-Justine Hospital, Department of Obstetric and Gynaecology, Montréal, Quebec, Canada, for their counsel, input, and feedback in the development of this study and especially Caroline Tourigny, François Audibert, Marie-Cécile Fière, Emmanuel Bujold, Zhong-Cheng Luo, Marie Hatem, and Hairong Xu. We also thank William Fraser, Sainte-Justine Hospital, Department of Obstetric and Gynaecology, University of Montreal, for methodological expertise.