Contraception and Breastfeeding at 4 Months Postpartum Among Women Intending to Breastfeed

Introduction

The many benefits of breastfeeding, including reduced infant mortality from infectious diseases, ¹ decreased maternal risk of breast cancer, ² and system-wide economic savings, ³ are well established. These benefits inform recommendations of the American College of Obstetricians and Gynecologists and the American Academy of Pediatrics for at least 6 months exclusive breastfeeding.^4,5 Current breastfeeding rates fall significantly below these recommendations. In 2013, 81% of mothers initiated breastfeeding, but only 52% were breastfeeding at 6 months. ⁶ Mothers also report failure to meet personal breastfeeding goals. One study found that 60% of women stopped breastfeeding earlier than desired. ⁷

Breastfeeding is one of several key health issues for postpartum women; adequate birth spacing is also important for the health of mothers, infants, and future children. Compared with shorter intervals, interpregnancy intervals (IPIs) of 18–23 months are associated with reduced risk of major maternal complications, including death, bleeding, and endometritis, ⁸ in addition to decreased perinatal risks of preterm birth and low birth weight. ⁹ The World Health Organization (WHO) recommends an IPI of at least 24 months. ¹⁰

Achieving such IPIs requires adequate contraception for women who are sexually active with men. However, one study indicated that up to two-thirds of postpartum women have unmet contraceptive needs. ¹¹ Breastfeeding itself is an effective temporary form of contraception. The lactational amenorrhea method requires that a woman be exclusively (or nearly exclusively) breastfeeding, amenorrheic, and less than 6 months postpartum ¹² ; if all three criteria are not met, another form of contraception is needed to prevent pregnancy.

The most popular contraceptive choice in the United States remains combined hormonal contraceptives (CHCs), selected by 25.9% of women aged 15–44 using contraception in 2011–2013. ¹³ CHCs have established benefits of familiarity, effectiveness, reversibility, and decreased menstrual bleeding and pain. Hormonal contraceptive use during breastfeeding is controversial. There are theoretical concerns that hormones (progesterone and particularly estrogen) diminish lactation through effects on prolactin, the hormone responsible for human milk production. Despite these concerns, recent systematic reviews have concluded that the evidence is limited, inconsistent, and inadequate to establish an effect of hormonal contraception on lactation.^14,15

If hormonal contraception interferes with milk production, then women using hormonal methods might be less likely to achieve their breastfeeding intentions. Maternal intention is an important predictor of breastfeeding,^16,17 and provider counseling can increase maternal awareness of breastfeeding benefits and strengthen intention to breastfeed. ¹⁸ Our study aimed to determine whether breastfeeding outcomes at 4 months postpartum differed by postpartum contraceptive method among women with a prenatal intention to breastfeed at least 4 months. Based on the results of prior randomized trials,^19–21 we hypothesized that women using hormonal contraceptives would have similar breastfeeding outcomes compared with those using nonhormonal methods.

Materials and Methods

We analyzed data from the Infant Feeding Practices Study II (IFPS II), a prospective cohort study of U.S. mothers followed during May 2005 through June 2007. The IFPS II surveyed women in the third trimester, monthly from 2 to 7 months postpartum, then every 7 weeks until 12 months postpartum. The surveys collected data on infant feeding, demographics, and health characteristics. The IFPS II sample was selected from a national consumer opinion panel of 500,000 U.S. households and restricted to women who were at least 18 years old, healthy at birth, and with a healthy full or nearly full-term singleton infant. Approximately 4,900 women completed the prenatal questionnaire and 2,000 completed each postnatal follow-up questionnaire. IFPS II study methods have been described in detail elsewhere. ²² We restricted our analyses to women who reported prenatal intention to breastfeed at least 4 months and contraception use at 3 months postpartum (n = 1,349).

Our primary outcome measures were any and exclusive breastfeeding at 4 months, where any breastfeeding was defined as feeding any amount of breast milk over the previous 7 days, and exclusive breastfeeding was defined as feeding only breast milk over the same period. Prenatal breastfeeding intention was assessed in the third trimester by asking, “What method do you plan to use to feed your new baby in the first few weeks?” If respondents intended any breastfeeding, we used responses to the following question to calculate total intended months of breastfeeding: “How old do you think your baby will be when you completely stop breastfeeding?” If respondents intended exclusive breastfeeding, we calculated total intended months of exclusive breastfeeding using the following: “How old do you think your baby will be when you first feed him or her formula or any other food besides breast milk?” Exclusive breastfeeding response options were categorized as “Less than 1 month,” “1–2 months,” “3–4 months,” “5–6 months,” “7–9 months,” or “More than 9 months.”

Contraceptive methods reported at 3 months postpartum were categorized as follows: nonhormonal contraceptives (NHCs) included any nonhormonal methods, including barrier methods, sterilization, and natural family planning; CHCs included the pill, contraceptive patch, and vaginal ring; and progestin-only pills (POPs) were categorized separately. Since women reporting intrauterine device (IUD) use did not specify hormonal versus nonhormonal IUD type, IUD users are presented separately for descriptive statistics and combined with NHC users for analyses. Due to the small number of women reporting depot medroxyprogesterone acetate (DMPA) use (n = 64 or 5%) and higher progestin dose in DMPA than POPs, descriptive statistics for DMPA users are presented separately and excluded from primary analyses. Women reporting multiple birth control methods were assigned the method with the largest dose of hormones. Sensitivity analyses excluded women who reported using multiple birth control methods and also considered alternate categorization by grouping IUDs and DMPA as progestin-only methods, along with POPs.

Using causal theory, we identified potential confounders for analysis: duration of prenatal any/exclusive breastfeeding intention (measured in months), age (measured in years), race/ethnicity (white, non-white), education (college degree, some college, high school, or less), income (≥$50,000, $20,000–49,999, <$20,000), marital status (married, unmarried), region (Northwest, Midwest, South, West), postpartum depressive symptoms (Edinburgh Postnatal Depression Scale Score >9 at 2 months postpartum), parity (multiparous, primiparous), and timing of postpartum return to work (≥12, 6 weeks through <12, <6 weeks).

For a covariate to be included as a confounder, it must (a) result in a change in the exposure–outcome effect estimate by >10% using backward elimination from the full model and (b) address residual confounding in the literature. To avoid overadjustment, we explored collinearity using a scatterplot matrix and correlation table for all covariates, considering variables with a p-value <0.0001 and a Pearson Correlation Coefficient >0.5 to be collinear. The only two covariates missing >5% of responses were postpartum depressive symptoms and timing of postpartum return to work. We believe that these data are missing at random because they were asked as part of a follow-up questionnaire, not as part of the core questionnaire like other covariates. As a result, we considered these two covariates identified through causal theory as potential confounders using our a priori change-in-estimate criterion and by comparing nested models where observations were not missing.

Using multivariable logistic regression, we estimated both the odds and predicted probabilities of any breastfeeding at 4 months postpartum by contraceptive category among women with prenatal intention to breastfeed at least 4 months. We repeated these analyses for exclusive breastfeeding at 4 months among women with prenatal intention to exclusively breastfeed at least 3–4 months. Given high prevalence of both breastfeeding outcomes in our sample, the odds ratio may be an inflated estimate of the risk of breastfeeding cessation associated with contraceptive use. Thus, predicted probabilities are presented as an alternate means of interpreting the risk of breastfeeding cessation associated with each contraceptive category. Predicted probabilities were calculated by setting each covariate to the category associated with higher likelihood of breastfeeding (i.e., white race, married status, multiparity, higher income, and education levels) and each continuous covariate to the median value. All analyses were conducted in SAS version 9.4 (Cary, NC). This study was deemed exempt by the Institutional Review Board at the University of North Carolina at Chapel Hill.

Results

Most women in the study sample used NHCs (n = 720 or 53%), although some groups were more likely to select certain methods (Table 1). Overall, we found that NHC use was more prevalent among women who were older, college graduates, married, multiparous, and returned to work by 6 weeks postpartum. Women using CHCs were more likely to earn <$20,000 per year, live in the South, and report depressive symptoms at 2 months postpartum. The majority (n = 1,116 or 83%) of the sample was white, and contraceptive choice did not differ by race/ethnicity. At 4 months postpartum, 810 (60%) reported any breastfeeding and 361 (27%) reported exclusive breastfeeding.

Among women with prenatal intention to breastfeed at least 4 months, women using POPs had greater odds of any breastfeeding at 4 months postpartum (adjusted odds ratio [aOR] 3.15, 95% confidence interval [CI] 1.42–7.02), compared with those using NHCs (Table 2). Women using CHCs had decreased odds of any breastfeeding at 4 months (aOR 0.17, 95% CI 0.10–0.29), compared with those using NHCs. For white, married multiparous women with high education and income, the probability of any breastfeeding at 4 months among those using NHCs was 90% (95% CI 85–94); it was 97% (95% CI 92–99) among those using POPs and 61% (95% CI 46–74) among those using CHCs.

Among women with prenatal intention to breastfeed exclusively at least 3–4 months, the odds of exclusive breastfeeding at 4 months postpartum did not differ significantly among women using POPs (aOR 1.15, 95% CI 0.69–1.91) compared with those using NHCs (Table 3). Women using CHCs had decreased odds of exclusive breastfeeding at 4 months (aOR 0.34, 95% CI 0.18–0.65), compared with those using NHCs. For white, married multiparous women with high education and income, the probability of exclusive breastfeeding at 4 months among those using NHCs was 53% (95% CI 45–62); it was 57% (95% CI 44–69) among those using POPs and 28% (95% CI 17–44) among those using CHCs.

In sensitivity analyses including DMPA and IUDs as progestin-only contraceptives, we found similar rates of any and exclusive breastfeeding among progestin-only contraceptive versus NHC users. CHC use remained associated with significantly lower odds of any and exclusive breastfeeding, compared with NHC use. In sensitivity analyses excluding women who reported using multiple birth control methods, results were not significantly altered.

Discussion

Our study found that, among women intending at least 4 months of breastfeeding, compared with those using NHCs, those using POPs were more likely and those using CHCs were less likely to achieve their prenatal breastfeeding intentions. Among those intending at least 3–4 months of exclusive breastfeeding, those using CHCs were less likely than NHC users to achieve their prenatal breastfeeding intentions, but those using POPs had similar likelihood of achieving their intentions.

While POP users had the highest odds of breastfeeding, sensitivity analyses grouping IUDs and DMPA with POPs produced an attenuated and nonsignificant difference between the expanded progestin-only group and the NHC group. It may be that use of IUDs or DMPA is associated with decreased odds of breastfeeding; it is also plausible that providers counsel women with strong breastfeeding intentions to use POPs, leading women especially motivated to breastfeed to select POPs. It is challenging to draw conclusions regarding IUDs and DMPA given the small number of women reporting use of these methods in our sample. However, given the large body of research suggesting that use of POPs does not adversely affect breastfeeding,^23,24 it is logical to conclude that women using POPs and providers recommending them are especially supportive of breastfeeding. This type of social support may be challenging to control for in studies, and may increase duration of lactation independent of any physiological effects of POPs.

Our findings confirm and extend prior studies reporting lower breastfeeding rates among women using CHCs and stand in contrast to other studies suggesting no change in breastfeeding outcomes among this group. A recent systematic review on this subject was published in 2015 by the Cochrane Collaboration. ¹⁴ This review identified 11 randomized trials, four focusing on CHCs, comparing hormonal contraception of any type versus another form of hormonal contraception, nonhormonal contraception, or placebo during lactation. Two older trials compared CHCs to placebo and reported conflicting results.^25,26 Neither study quantified outcomes, making interpretation challenging. Two other trials compared CHCs to POPs. A WHO study beginning in 1984 found that those using CHCs had lower milk volumes at 9–24 weeks postpartum, but infant weight gain was not significantly different between CHC and POP groups.^19,20

A double-blind randomized controlled trial by Espey et al. in 2012 compared infant and lactation outcomes among 127 women who intended to breastfeed and also intended to use oral contraceptives for family planning; these women were randomized to begin either CHCs or POPs at 2 weeks postpartum. This trial did not find significant differences in infant growth, rates of formula supplementation, or breastfeeding continuation between the groups at 8 weeks postpartum, with 64.1% in the CHC group and 63.5% in the POP group continuing to breastfeed at 8 weeks. ²¹ This study protocol differs from current Center for Disease Control (CDC) guidelines. The 2016 U.S. Medical Eligibility Criteria for Contraceptive Use classifies use of CHCs at less than 21 days postpartum as a level “4” for all women, indicating unacceptable health risk due to risk of thromboembolism with early exposure to exogenous estrogen. ²⁷ Nevertheless, Espey et al. did not find an adverse effect of CHCs on breastfeeding, compared with POPs. It is unlikely that later introduction would have a larger impact than early introduction of CHCs on lactation outcomes.

By restricting our analyses to women with a prenatal intention to breastfeed, our findings extend prior research by exploring how contraceptive method may influence the likelihood of fulfilling personal breastfeeding goals. Many external forces, such as employment characteristics and childcare resources, influence breastfeeding. ²⁸ Clinicians should aim to support women to fulfill their individual goals as informed by broad national or global breastfeeding targets. Such “patient-centered care” respecting patient preferences and values has been identified as a fundamental tool for improving the U.S. healthcare system by the Institute of Medicine. ²⁹

In addition to reporting on a patient-centered outcome, the strengths of our study include its large sample size, ability to adjust for multiple potential confounders, and prospective design. Our study also had several limitations to consider. Generalizability may be limited because the majority of women were white, married, with a college degree, and income ≥$50,000; as a result, we were not able to evaluate whether race or ethnicity modified associations between contraception and attainment of breastfeeding goals. Another limitation is that contraception was only measured once at 3 months postpartum. It is possible that women stopped breastfeeding and then switched to CHCs, rather than started CHCs and then stopped breastfeeding. We used breastfeeding outcomes at 4 months postpartum to preserve temporal order of exposure and outcome. However, this reduced our sample to women intending to breastfeed at least 4 months, further limiting generalizability.

Our study had small numbers of women reporting DMPA (n = 64 or 5%) or IUD (n = 92 or 7%) use, with no differentiation between hormonal and nonhormonal IUDs, and we were unable to draw conclusions regarding breastfeeding outcomes among women using these methods. As subdermal contraceptive implants were not available in the United States during most of the study period, ³⁰ we were not able to assess the association between subdermal etonogestrel and lactation outcomes.

Conclusion

Our study found that, in a cohort of women intending to breastfeed at least 4 months, women using POPs were more likely, and women using CHCs were less likely, to achieve their prenatal breastfeeding intentions. Given the importance of contraception and breastfeeding for both individual mother-infant dyads and for population health, further research is needed to evaluate timing of CHC initiation and unplanned weaning. These studies should ideally collect prospective longitudinal data on infant feeding, infant growth, and contraceptive use. As research continues to explore this topic, evolving information should be shared with women so that they can select the contraceptive that best aligns with their family planning and breastfeeding goals.