- Question: Does labor epidural analgesia influence breastfeeding rates at 6 weeks among women delivering vaginally?
- Finding: The overall breastfeeding rate at 6 weeks for women delivering with labor epidural analgesia was significantly lower compared to that among women delivering without labor epidural analgesia; however, delivering with or without labor epidural analgesia did not alter breastfeeding outcomes among multiparous women with previous breastfeeding experience.
- Meaning: Labor epidural analgesia reduces the likelihood of successful breastfeeding, although this does not appear to be a causal association; lactation support may be a simple strategy to improve breastfeeding outcomes tailored to women with no previous breastfeeding experience who deliver with labor epidural analgesia.
The World Health Organization recommends breastfeeding because it reduces child mortality and has health benefits that extend into adulthood.1,2 Benefits for the breastfed neonate include enhanced immune system and reduced risks for diabetes mellitus and leukemia; for the mother, breastfeeding increases bone mineral density, decreases chances of developing diabetes mellitus, and may decrease the risk of cancer.3
In the United States, breastfeeding is initiated in 79% of newborns and maintained up to 6 months in 49% of infants.4 While the awareness and programs to educate Israeli women about the benefits of exclusive breastfeeding are increasing,5 breastfeeding initiation over 10 years ago was 78%, declining at 2 months (58%) and resulting in <30% of women breastfeeding by 6 months.6
The effect of labor epidural analgesia (LEA) on successful breastfeeding has been evaluated in several studies with divergent results. A systematic review reporting on 23 studies published between 1994 and 2013 found conflicting results7: 10 studies found no effect of LEA on breastfeeding outcomes, 12 studies found a negative association, and 1 reported positive outcomes. In one of the largest randomized clinical trials evaluating breastfeeding patterns after neuraxial labor analgesia,8 1043 nulliparous women were randomly assigned to one of 3 regimens: a combined spinal epidural or an epidural with a low-dose mixture of bupivacaine and fentanyl or to an epidural control group receiving a high-dose epidural infusion with no fentanyl. Breastfeeding initiation and long-term outcomes at 12 months were similar, including those randomly assigned in a matched control group of women receiving systemic pethidine but no neuraxial analgesia.8 In a recent observational study, delivering vaginally with LEA seemed to decrease the odds for postpartum depression and increased the likelihood of breastfeeding at 6 weeks postpartum,9 suggesting that failed lactation and postpartum depression may have common mechanisms, including altered pain perception and modulation.10,11
Heterogeneity in study designs and LEA regimens may explain discrepancies in findings. In addition, only a few studies have taken into account women’s intention to breastfeed, impact of parity, and previous breastfeeding experience.12–14 Finally, while sociodemographic factors, including marital status, ethnicity, work status, and parity, are factors that influence women’s desire to breastfeed and ability to maintain breastfeeding, mothers’ working status is seldom reported. In Israel, women are given a 12-week paid maternity leave providing a supportive environment for maintained breastfeeding.
We designed this prospective observational cohort study to examine the influence of LEA on breastfeeding outcomes at 6 weeks postpartum in a mixed-parity cohort. We hypothesized that LEA would not influence breastfeeding status 6 weeks postpartum in women who intended to breastfeed.
The study was conducted at the Beilinson Hospital, Israel, a tertiary university hospital with 10,000 annual deliveries and a LEA rate of 70%, between June 2015 and March 2016.
The Institutional Review Board approved this study, and all participants provided written informed consent. Before patients were enrolled, the prospective observational cohort study was registered at clinicaltrials.gov (NCT02495350, principal investigator: S.O.Z., date of registration: June 17, 2015). The study was designed to concomitantly assess the relationship among LEA with 2 long-term outcomes after vaginal delivery, postpartum depression, and breastfeeding. Our findings on postpartum depression rates according to women’s intention to deliver with LEA and actual use of LEA was recently published.15
All women at 37 weeks’ gestation or more who delivered vaginally were eligible for participation and were approached consecutively for participation within 2–18 hours postpartum, and as close to delivery as possible. Women were not enrolled if <18 years of age, with a medical contraindication for LEA, receiving bromocriptine for a medical reason, unable to fill out a questionnaire or if they did not understand the Hebrew language, or if their baby was hospitalized in the neonatal intensive care unit.
Demographic data included maternal age, parity, gravity, marital status (married yes/no), religion (Jewish/other), and country of birth (Israeli-born/other). Detailed medical history including comorbidities (hypertension, diabetes mellitus, thyroid disease, and chronic anxiety requiring medication during pregnancy), obstetric conditions (preeclampsia and gestational diabetes), anesthetic outcomes including LEA use (yes/no), total epidural fentanyl dose (μg), obstetric outcomes (use of oxytocin yes/no, spontaneous versus instrumental vaginal delivery), and neonatal data were recorded at the time of enrollment.
A standardized LEA protocol is used in our labor room, which consists of initiation of epidural analgesia with bupivacaine 0.1% 15 mL and fentanyl 100 μg in 5-mL increments, followed by an epidural infusion of bupivacaine 0.1% 10 mL and fentanyl 2 µg/mL, with a patient-controlled epidural analgesia modality with 5-mL bolus with a lock-out time of 15 minutes.
Breakthrough pain (pain score >4) is routinely managed with an epidural bolus of bupivacaine 0.25% (5–8 mL) or lidocaine 1% (5–8 mL). Additional epidural fentanyl is not part of our management of breakthrough pain.
Breastfeeding Initiation Protocol
A standard breastfeeding initiation protocol is used in our labor room, which consists of both immediate skin-to-skin contact and postpartum education and support. Specifically, in the labor room, all newborns are placed immediately after delivery on the mother’s chest for skin-to-skin contact for 20 minutes. In the postpartum unit, a lactation specialist visits every woman daily to observe breastfeeding and improve technique. In addition, a 1-hour focused lecture on breastfeeding is given daily by a lactation specialist; women are encouraged to attend the lecture at least once during hospitalization. Because women receive a breastfeeding bracelet at the lecture to be worn on the wrist on the same side of the last breastfeeding session, having a bracelet or not allows nurses to ensure that all women attended the class. Therefore, it is expected that 100% of women attended the class. Finally, rooming-in is available for women interested in having their babies with them at all times. To date, there are no antenatal information classes or written pamphlets educating women about breastfeeding in our institution.
On enrollment, to avoid influencing the measured outcomes (breastfeeding and depression) and possibly stigmatizing women, information about the study and outcomes of interest was provided with a standardized script; women were told “We are performing a study to evaluate the influence of having delivered with or without an epidural on how women feel at 3 days and 6 weeks after delivery.”
The primary outcome of maintained breastfeeding at 6 weeks was selected based on previous studies evaluating breastfeeding outcomes after vaginal delivery with and without LEA at this specific time point.16,17
Assessments were conducted at 3 time points: (1) immediately after delivery within the first 18 hours with an in-person interview, (2) on the third day postpartum via phone call, and (3) at 6 weeks postpartum via phone call.
Women are typically discharged 2 days/nights after an uncomplicated vaginal delivery; consequently, the third-day postpartum assessment was done by phone interview.
Initial Postpartum Assessment
Within 2 hours after delivery, all women were asked about their intention to breastfeed (yes/no) and their assessment of the importance of breastfeeding (0: not important at all, 10: most important). In addition, multiparous women were asked if they had a previous successful breastfeeding experience after their previous delivery(ies) (>6 weeks) (yes/no). Use of LEA was reported by the mother and confirmed from data recorded in the medical record (Supplemental Digital Content 1, Appendix, http://links.lww.com/AA/C380).
Data from women not intending to breastfeed were initially recorded but subsequently not included in final analysis.
Assessment on Postpartum Day 3
Women were contacted via phone by a different member of the research team unaware of labor analgesia received (epidural yes/no) and asked if they were breastfeeding. If women were breastfeeding, they were asked if they were exclusively breastfeeding (yes/no) or also feeding in combination with bottled milk formula (yes/no).
Assessment at 6 Weeks
Women were contacted again via phone at 6 weeks by a member of the research team not aware of previous assessments to record breastfeeding status (yes/no) and exclusivity of breast milk feeding (yes/no).
The Edinburgh postnatal depression scale, with a 10-item questionnaire, was used, with a score ≥10 considered suggestive of postpartum depression.
To reduce loss to follow-up at 6 weeks, if a woman did not answer the phone, calls were made 5 times a day over the next 6 days by a research investigator. If a woman was not reached within this time period, she was considered lost to follow-up.
The study was powered to evaluate the primary outcome: breastfeeding at 6 weeks (yes/no). We hypothesized that there would be no difference in breastfeeding rates at 6 weeks between women delivering with LEA (70% of women at our institution) and women delivering without LEA (30%), and the breastfeeding rate in both groups would be similar to the concurrent estimated breastfeeding rate at 6 weeks in women delivering at our institution (75%).
To rule out an absolute change of 7.5% (−7.5% to 7.5%) in the proportion of women breastfeeding at 6 weeks at a significance level of 5% and power of 80% between delivering with and without LEA, thus performing a 2-sided noninferiority test, using χ2 test, with no predetermined difference between populations, both with 75% level breastfeeding, we calculated a required sample size of 824 for women delivering with LEA and 354 for women delivering without LEA, or total of 1178. To account for possible loss to follow-up (10%) and women not determined to breastfeed (which we estimated in our cohort to be in the order of 10%), we aimed for 1500 women.
The primary independent variable was LEA use during the delivery, and the primary outcome was any breastfeeding at 6 weeks (with a dichotomous outcome of yes/no). Comparisons between the groups for the primary outcome were performed using 1-way analysis of variance with post hoc Bonferroni correction. P value < .05 was considered significant for the primary outcome (breastfeeding versus not breastfeeding at 6 weeks among women delivering with or without LEA).
A multivariable regression analysis used to evaluate the relationship between LEA use and the primary outcome (breastfeeding at 6 weeks) included the entire cohort of women intending to breastfeed (N = 1204), and controlled for age, neonatal birth weight, 11 dichotomous variables (nulliparity, being Jewish, being Israeli born, marital status, gestational diabetes, pregnancy-induced hypertension, hypothyroidism, chronic anxiety requiring medication, intrapartum oxytocin use, instrumental delivery, and LEA use), and an interaction term between nulliparity and LEA use (Supplemental Digital Content 2, Table 1, http://links.lww.com/AA/C381).
To specifically explore the effect of previous breastfeeding experience, rather than parity itself (because not all multiparous women necessarily breastfed after their previous delivery), an additional multivariable regression analysis for the primary outcome (breastfeeding at 6 weeks) was conducted, replacing parity with previous breastfeeding experience, both as a covariate and in the interaction term (resulting in the interaction between previous breastfeeding experience and LEA).
Proportions between categorical covariates (such as nulliparity) and the primary outcome were compared using the χ2 test. Correlation for continuous variables (such as epidural fentanyl dose) was measured (given that distributions were not normal) with Spearman correlation coefficient. We also examined the continuous discrimination by level of epidural fentanyl dose of the outcomes using area under the curve (AUC) (thus a value of 1 would represent total prediction of the outcome by dose level, such as a perfect cutoff, and 0.5 totally random or no association).
For the secondary outcomes, which included breastfeeding at 3 days, breastfeeding exclusively at 3 days, and breastfeeding exclusively at 6 weeks, comparisons were performed using χ2 test; the P value for statistical significance was adjusted (Bonferroni correction was applied, and P < .017 was considered significant).
Statistical analysis was performed using R version 3.4.0 (S statistical package, IBM).
There were 1580 women initially approached for participation, with 1497 signing informed consent within 2–18 hours after delivery. A total of 1326 women were reached for follow-up of postpartum depression and breastfeeding success at 3 days and 6 weeks (Figure). However, 122 women had indicated not actually intending to breastfeed; therefore, data for these women were ultimately excluded from this study, even though some of these women were breastfeeding. A total of 1204 women were included in the final analysis, of which 836 delivered with LEA (69.4%) and 368 (30.6%) delivered without LEA.
There was no difference in demographic data or LEA use between women lost to follow-up (N = 171) and those reached at 6 weeks (N = 1326) except for women being younger in the lost to follow-up group (P = .009) (Supplemental Digital Content 3, Table 2, http://links.lww.com/AA/C382).
Findings According to LEA Use
Use of LEA was associated with maternal age, parity, oxytocin use, instrumental delivery (Table 1), and breastfeeding at 6 weeks (Table 2). Specifically, the difference in breastfeeding rates at 6 weeks (74.0% vs 83.4%) exceeded the prespecified clinically important difference of 7.5% (9.4%; 95% confidence interval [CI], 4.4%–14.4%; P < .001).
Multivariable logistic regression including the entire cohort (N = 1204) identified LEA use as the only variable associated with reduced breastfeeding at 6 weeks (odds ratio [OR], 0.60; 95% CI, 0.40–0.90; P = .015), after controlling for age, Jewish ethnicity, Israeli natality, marital status, medical conditions (gestational diabetes, pregnancy-induced hypertension, hypothyroidism, and anxiety), parity, oxytocin use, instrumental vaginal delivery, neonatal birth weight, and the interaction between nulliparity and LEA use.
Similar differences were identified when comparing breastfeeding rates at 3 days (87.6% among those delivering with LEA versus 94.0% among those delivering without), and exclusive breastfeeding at both 3 days and 6 weeks (Table 2).
Findings According to Parity
Multiparous women were less likely to deliver with LEA than nulliparous women (61.8% vs 84.9%; P < .001), were more likely to breastfeed at 6 weeks (80% vs 70%; P < .001), and more likely to exclusively breastfeed at 6 weeks (61% vs 41.7%; P < .001) (Table 3).
Findings According to Breastfeeding Experience
Among the multiparous women (N = 806), 719 women had previously breastfed (89.2%), which represented 59.7% of the entire cohort; the breastfeeding rate at 6 weeks was significantly higher in women with a previous experience of breastfeeding compared to women, nulliparous or multiparous, with no breastfeeding experience (84.3% vs 66.0%, respectively; OR, 2.9; 95% CI, 2.1–3.6; P < .0001). Among women with previous breastfeeding experience (N = 719), the breastfeeding rate at 6 weeks was not significantly different between women delivering with or without LEA (82.3% with LEA versus 87.3% without LEA; OR, 1.48; 95% CI, 0.97–2.29; P = .08).
Using a modified multivariable logistic regression model to specifically evaluate the impact of previous breastfeeding experience rather than parity itself in the entire cohort (N = 1204), age, being Jewish (yes/no), being Israeli born (yes/no), marital status (yes/no), gestational diabetes (yes/no), pregnancy-induced hypertension (yes/no), hypothyroidism (yes/no), chronic anxiety requiring medication (yes/no), intrapartum oxytocin use (yes/no), LEA (yes/no), instrumental delivery (yes/no), and neonatal birth weight, with parity being replaced by previous breastfeeding experience (yes/no) and the interaction term now being the interaction between previous breastfeeding experience and LEA use, only previous breastfeeding experience was associated with increased breastfeeding at 6 weeks (OR, 3.17; 95% CI, 1.72–5.80; P < .001).
Intrapartum oxytocin for induction or augmentation of labor was used in 646 cases (54.0%). Oxytocin had been given to 52% of women who were breastfeeding at 6 weeks and 62.1% of women who were not breastfeeding at 6 weeks (P < .003; Table 4).
No significant differences were identified in breastfeeding rates between those with and those without oxytocin, when stratified by LEA use (P = .14). Among women delivering with LEA (N = 827), the breastfeeding rate was 72.3% with oxytocin versus 77.5% without oxytocin (N = 242). Among women delivering without LEA (N = 363), the breastfeeding rate was 81.9% with oxytocin and 84.2% without oxytocin.
Epidural Fentanyl Dose.
Because delivering with LEA seemed to be associated with lack of breastfeeding at 6 weeks, we subsequently (post hoc) assessed whether the total amount of epidural fentanyl was associated with reduced likelihood of breastfeeding.
Including only women delivering with LEA, the total epidural fentanyl dose (median, 210 μg; interquartile range [IQR], 160–280) was not associated with breastfeeding status at 6 weeks (yes/no) (AUC, 0.47; 95% CI, 0.42–0.53). However, total epidural fentanyl dose was associated with breastfeeding status at 3 days (yes/no) (AUC, 0.61; 95% CI, 0.52–0.7). The total epidural fentanyl dose among women breastfeeding at 3 days was lower (160 μg; IQR, 140–200) than that among women who were not breastfeeding (median, 210 μg; IQR, 160–280; P < .001; 95% CI, 26–82).
There were only 13 women in the entire cohort reporting anxiety requiring medication during pregnancy, of which 10 delivered with LEA (Table 1). Of these 13 women, 7 were breastfeeding at 6 weeks (53.8%; Table 4).
Eighty women met criteria for postpartum depression, based on a cutoff score ≥10 on the Edinburgh questionnaire. Among these women, the breastfeeding rate at 6 weeks 72.5% (58/80 women) was not different from the rest of the cohort (P = .39).
In this large prospective study evaluating breastfeeding maintenance among nulliparous and multiparous women delivering vaginally with or without LEA, the overall breastfeeding rate was 83.3% at 3 days and 70.7% at 6 weeks. Contrary to our expectations, breastfeeding rates both at 3 days and at 6 weeks were significantly lower among women delivering with LEA compared to women delivering without. However, breastfeeding success was significantly associated with previous breastfeeding experience in multiparous women.
We believe that our study provides novel important information that has not previously been reported.8,16,17 The recent randomized clinical trial assessing breastfeeding success according to epidural fentanyl dose, which only enrolled women with previous breastfeeding experience, demonstrated that fentanyl dosing did not influence breastfeeding success.17 This is an important finding because this disproves the purported association between neonatal exposure to fentanyl and suckling ability. Our findings are consistent with this premise because, among multiparous women with previous breastfeeding experience, which constituted 59.7% of our cohort (N = 719), delivering with or without LEA did not impact breastfeeding success at 6 weeks. Our study design offered a more pragmatic and natural setting because women were not even aware that breastfeeding status would be assessed at 6 weeks; we demonstrate here with no Hawthorne effect, nor stigmatizing pressure, that with previous breastfeeding experience and 1:1 lactation support, >5 of 6 multiparous women are likely to successfully breastfeed at 6 weeks if intending to do so, irrespective of labor epidural use.
Therefore, women “needing” LEA are more likely to also “fail” at breastfeeding. Psychophysical factors predisposing women to request LEA may also affect breastfeeding outcomes. For several possible reasons, breastfeeding outcomes are worse in women who deliver with LEA, even though they initially intend to breastfeed. First, women delivering with LEA are more likely to be nulliparous, and regardless of parity, they may experience more difficult and prolonged labors with oxytocin augmentation than women who deliver without LEA. Second, women with multiple intrapartum interventions, including induction of labor and operative vaginal deliveries, are likely to experience more pain, fatigue, and stress during labor and delivery.14 Specifically, intrapartum exogenous oxytocin administration and LEA are associated with a reduction of endogenous oxytocin levels postpartum,18 and breastfeeding rates have been shown to be reduced among women receiving exogenous oxytocin,19–21 although this was not the case in our cohort. Third, although untested in our study, increased pain thresholds and coping strategies among women who deliver without LEA may allow them to better tolerate pain and discomfort associated with breastfeeding. Women with higher pain thresholds and enhanced endogenous pain inhibition may be those women more likely to withstand pain during labor and maintained breastfeeding.
The overall breastfeeding rate (76%) in this mixed-parity cohort is consistent with the published literature.7 Population-level differences in breastfeeding rates are attributed to a multitude of factors, including that of women’s motivation when participating in a randomized clinical trial specifically designed to evaluate breastfeeding outcomes with LEA and that of having previously breastfed. Factors associated with maintained breastfeeding in the Israeli society include ethnicity5,22 and religion,6 spouses’ opinion about breastfeeding,5 higher education,6 and active lactation counseling. Women delivering in our hospital are considered to belong to a secular middle-upper class with a good social support, although we are aware that such characteristics may have different influences on breastfeeding choices in different cultures with different social norms. Interestingly, we noted a higher number of women with exclusive breastfeeding at 6 weeks compared with the 3-day time point, which may be due to the fact that women choose (or prefer) to sleep on the first night after their vaginal delivery, and nurses offer to give 1 formula (bottle) feeding. For this reason, women consider themselves at 3 days not to have exclusively breastfeed even though they intend to breastfeed (likely exclusively) after this 1 bottle. The impact of that initial bottle feeding on overall outcome was not specifically investigated here.
The strengths of this study include its large size, entirely prospective study design, and the fact that, because there was no intervention, the enrollment rate was high, reducing the likelihood for bias. This cohort is representative of all women delivering in our institution and may be generalized to women delivering in a breastfeeding-friendly environment. Women were enrolled immediately after delivery, and answered a phone interview at 3 days and 6 weeks; therefore, there is no recall bias, and research staff evaluating the breastfeeding outcomes was unaware of LEA use. In addition, standardized hospital protocols for LEA and prompt skin-to-skin and breastfeeding initiation should have eliminated certain confounders, although we did not specifically record the time or duration of first breastfeeding session. Last, because of the 3 months paid maternity leave policy in Israel, difference in working status should not affect our findings.
We acknowledge several limitations. This is an observational cohort design, because randomly assigning women to deliver with LEA or not seems unreasonable nowadays; nonetheless, women’s self-allocation provides important information in itself, because some of the psychophysical factors that will influence women’s request for pain relief may also be factors that impact their ability to breastfeed, and our study, therefore, provides a pragmatic and realistic snapshot. We did not specifically record whether women were feeding directly from the breast rather expressing milk via pump, and exclusive direct feeding during the first 24–48 hours has been shown to be strongly correlated with maintained breast milk feeding 6 months postpartum.23 Similarly, we did not evaluate rooming status (partial versus complete) or why breastfeeding was eventually stopped. Because our primary outcome was breastfeeding at 6 weeks, other results may be underpowered, such as the effect of parity and/or previous breastfeeding experience on breastfeeding maintenance at 6 weeks. Among multiparous women, we did not record the mode of delivery in previous pregnancies; therefore, we cannot report on the effect of current vaginal birth after previous cesarean delivery on breastfeeding success. As for psychosocial factors, while we did record several sociodemographic parameters that are important in the Israeli social construct (eg, being Jewish and born in Israel), we only enrolled Hebrew-speaking women, which may have introduced a certain bias. Last, our study is underpowered to evaluate the effect of anxiety or depression on breastfeeding outcomes (only 13 women in the entire cohort reported anxiety), and while there were 80 women with a possible diagnosis of postpartum depression, this was not associated with breastfeeding status.
In conclusion, our study findings in a mixed-parity cohort show that delivering vaginally with LEA is associated with reduced likelihood of breastfeeding at 6 weeks. However, because previous breastfeeding experience was associated with higher likelihood of successful breastfeeding, our findings taken together suggest that other psychophysical factors predisposing women to request LEA also affect breastfeeding outcomes. Further studies targeting women at risk for reduced breastfeeding success, such as women with no breastfeeding experience who delivered with LEA are needed to evaluate the effect of tailored lactation services as 1 simple approach to improve breastfeeding success in the future.
Name: Sharon Orbach-Zinger, MD.
Contribution: This author helped design and conduct the study, collect and analyze the data, and prepare the manuscript.
Name: Ruth Landau, MD.
Contribution: This author helped design the study, analyze the data, and prepare the manuscript.
Name: Atara Davis, BSc.
Contribution: This author helped design and conduct the study, and collect the data.
Name: Oren Oved, MD.
Contribution: This author helped conduct the study and collect the data.
Name: Liron Caspi, MD.
Contribution: This author helped conduct the study and collect the data.
Name: Shlomo Fireman, MD.
Contribution: This author helped design the study and analyze the data.
Name: Shai Fein, MD.
Contribution: This author helped prepare the manuscript.
Name: Alexander Ioscovich, MD.
Contribution: This author helped design the study.
Name: Danielle Bracco, BA.
Contribution: This author helped collect the data.
Name: Moshe Hoshen, PhD.
Contribution: This author helped analyze the data.
Name: Leonid A. Eidelman, MD.
Contribution: This author helped design and conduct the study.
This manuscript was handled by: Jill M. Mhyre, MD.