Mei Chung, MPH; Gowri Raman, MD; Thomas Trikalinos, MD, PhD; Joseph Lau, MD; Stanley Ip, MD
Disclaimer: The authors of this manuscript are responsible for its content. Statements in the review should not be construed as endorsement by the Agency for Healthcare Research and Quality.
Grant Support: From the Agency for Healthcare Research and Quality, U.S. Department of Health and Human Services (contract no. 290-02-0022).
Potential Financial Conflicts of Interest: None disclosed.
Requests for Single Reprints: Mei Chung, MPH, Center for Clinical Evidence Synthesis, Institute for Clinical Research and Health Policy Studies, Tufts Medical Center, 800 Washington Street, Box 63, Boston, MA 02111; e-mail, email@example.com.
Current Author Addresses: Ms. Chung and Drs. Raman, Trikalinos, Lau, and Ip: Tufts Medical Center, 800 Washington Street, Box 63, Boston, MA 02111.
Chung M., Raman G., Trikalinos T., Lau J., Ip S.; Interventions in Primary Care to Promote Breastfeeding: An Evidence Review for the U.S. Preventive Services Task Force. Ann Intern Med. 2008;149:565-582. doi: 10.7326/0003-4819-149-8-200810210-00009
Download citation file:
Published: Ann Intern Med. 2008;149(8):565-582.
Evidence suggests that breastfeeding decreases the risk for many diseases in mothers and infants. It is therefore important to evaluate the effectiveness of breastfeeding interventions.
To systematically review evidence for the effectiveness of primary careâ€“initiated interventions to promote breastfeeding with respect to breastfeeding and child and maternal health outcomes.
Electronic searches of MEDLINE, the Cochrane Central Register of Controlled Trials, and CINAHL from September 2001 to February 2008 and references of selected articles, restricted to English-language publications.
Randomized, controlled trials of primary careâ€“initiated interventions to promote breastfeeding, mainly in developed countries.
Characteristics of interventions and comparators, study setting, study design, population characteristics, the proportion of infants continuing breastfeeding by different durations, and infant or maternal health outcomes were recorded.
Thirty-eight randomized, controlled trials (36 in developed countries) met eligibility criteria. In random-effects meta-analyses, breastfeeding promotion interventions in developed countries resulted in significantly increased rates of short- (1 to 3 months) and long-term (6 to 8 months) exclusive breastfeeding (rate ratios, 1.28 [95% CI, 1.11 to 1.48] and 1.44 [CI, 1.13 to 1.84], respectively). In subgroup analyses, combining pre- and postnatal breastfeeding interventions had a larger effect on increasing breastfeeding durations than either pre- or postnatal interventions alone. Furthermore, breastfeeding interventions with a component of lay support (such as peer support or peer counseling) were more effective than usual care in increasing the short-term breastfeeding rate.
Meta-analyses were limited by clinical and methodological heterogeneity. Reliable estimates for the isolated effects of each component of multicomponent interventions could not be obtained.
Evidence suggests that breastfeeding interventions are more effective than usual care in increasing short- and long-term breastfeeding rates. Combined pre- and postnatal interventions and inclusion of lay support in a multicomponent intervention may be beneficial.
Human milk is the natural nutrition for all infants. According to the American Academy of Pediatrics, it is the preferred choice of feeding for all infants (1). The goals of Healthy People 2010 for breastfeeding are an initiation rate of 75% and continuation rate of 50% at 6 months and 25% at 12 months after delivery (2). A survey of U.S. children in 2002 indicated that only 71% had ever been breastfed, and the percentage of infants who continue to be breastfed to some extent is 35% at 6 months and 16% at 12 months (3). Although the breastfeeding initiation rate is close to the goal set by Healthy People 2010, according to this survey, the breastfeeding continuation rates at 6 and 12 months fall short.
Evidence suggests that breastfeeding decreases risks for many diseases in infants and mothers. In children, breastfeeding has been associated with a reduction in the risk for acute otitis media, nonspecific gastroenteritis, severe lower respiratory tract infections, atopic dermatitis, childhood leukemia, and the sudden infant death syndrome. In mothers, a history of lactation has been associated with a reduced risk for type 2 diabetes and breast and ovarian cancer (4). According to the American Academy of Pediatrics, some of the obstacles to initiation and continuation of breastfeeding include insufficient prenatal education about breastfeeding, disruptive maternity care practices, and lack of family and broad societal support (5). Effective interventions reported to date include changes in maternity care practices, such as those implemented in pursuit of the Baby-Friendly Hospital Initiative (BFHI) designation (6, 7), and worksite lactation programs (8). Some of the other interventions implemented include peer-to-peer support, maternal education, and media marketing (9).
Our review is based on an evidence report (10) that was requested by the Center on Primary Care, Prevention, and Clinical Partnerships at the Agency for Healthcare Research and Quality, on behalf of the U.S. Preventive Services Task Force, to support the Task Force's update of its 2003 recommendations on counseling to promote breastfeeding (11). Together with the Tufts Evidence-based Practice Center, these agencies jointly developed an analytic framework for study questions to evaluate the available evidence to promote and support breastfeeding (Figure 1). Five linked key questions were proposed in the analytic framework:
1. What are the effects of breastfeeding interventions on child and maternal health outcomes?
2. What are the effects of breastfeeding interventions on breastfeeding initiation, duration, and exclusivity?
3. Are there harms from interventions to promote and support breastfeeding?
4. What are the benefits and harms of breastfeeding on infant or child health outcomes?
5. What are the benefits and harms of breastfeeding on maternal health outcomes?
The contextual questions regarding the effectiveness of health care system influences on interventions to promote breastfeeding and the potential benefits and harms related to such interventions can be answered by synthesizing the available scientific evidence for each key question. To avoid redundant work, a joint decision was made to adopt results from our earlier Agency for Healthcare Research and Quality evidence report (4) to address questions 4 and 5 on the benefits and harms of breastfeeding for infants and mothers. Table 1 (12–84) presents a synopsis of that report's findings on questions 4 and 5. We address only questions 1 to 3 in this article. Specifically, we examine the effects of primary care–initiated interventions to support or promote breastfeeding on child and maternal health outcomes and breastfeeding rates, as reported in randomized, controlled trials (RCTs) from developed countries. We also document reported harms from interventions to promote and support breastfeeding.
This systematic review focuses on recent evidence (September 2001 to February 2008) and updates a previous systematic review (85) conducted for the U.S. Preventive Services Task Force to support its 2003 recommendation on counseling to promote breastfeeding (available at http://www.ahrq.gov/clinic/uspstf/uspsbrfd.htm). We searched for English-language articles in MEDLINE, the Cochrane Central Register of Controlled Trials, and CINAHL from September 2001 to February 2008 by using such Medical Subject Heading terms and keywords as breastfeeding, breast milk feeding, breast milk, human milk, nursing, breastfed, infant nutrition, lactating, and lactation. We also reviewed reference lists of a related systematic review (86) for additional studies.
We included RCTs published from September 2001 to February 2008 that included any counseling or behavioral intervention initiated from a clinician's practice (office or hospital) to improve the breastfeeding initiation rate or duration of breastfeeding among healthy mothers or members of the mother–child support system (such as partners, grandparents, or friends) and their healthy term or near-term infants (≥35 weeks' gestation or ≥2500 g). We focused our review on studies conducted in developed countries; however, because of the widespread interest in the BFHI, we also included RCTs of the BFHI that were conducted in Brazil and Belarus.
We considered interventions conducted by various providers (lactation consultants, nurses, peer counselors, midwives, and physicians) in various settings (hospital, home, clinic, or elsewhere) to be eligible as long as they originated from a health care setting. We considered maternity services to be primary care for this review. We also included such health care system interventions as staff training. We excluded community- or peer-initiated interventions. Control comparisons were any usual prenatal, peripartum, or postpartum care, as defined in each study. Studies needed to report rates of breastfeeding initiation, duration of breastfeeding, or exclusivity of breastfeeding to be included. Figure 2 shows our search and selection process.
One investigator extracted data from each study, and another confirmed them. The extracted data included study setting, population, control, description of intervention (type, person, frequency, and duration), definitions of breastfeeding outcomes (initiation, exclusivity, and duration), definitions of health outcomes in both mothers and children (when provided), and analytic methods.
Breastfeeding interventions can include a combination of individual components, such as structured breastfeeding education or professional or lay support. We defined 3 categories of breastfeeding intervention: those that included a component of formal or structured breastfeeding education, those that included a component of either professional or lay breastfeeding support, or those that did not include the aforementioned components. The first 2 categories are not mutually exclusive. Table 2 shows complete details.
We classified breastfeeding regimens as exclusive or nonexclusive. Studies used different definitions of exclusive breastfeeding (“no supplement of any kind,” “including water while breastfeeding,” or “occasional formula is permissible while breastfeeding”); we adopted all of those definitions. We classified all other breastfeeding regimens (full, partial, mixed, or nonspecified) as nonexclusive.
We defined breastfeeding initiation as any breastfeeding at discharge or up to 2 weeks after delivery. We also defined a priori breastfeeding durations of 1 to 3 months as short-term, 4 to 5 months as intermediate-term, 6 to 8 months as long-term, and 9 or more months as prolonged. We categorized studies with breastfeeding durations shorter than 1 month as “no breastfeeding” in our meta-analyses.
Two investigators assessed the methodological quality of all eligible studies by using criteria developed by the U.S. Preventive Services Task Force (87). We assigned each article a quality rating of “good,” “fair,” or “poor.” The criteria for quality assessment of primary studies included randomization techniques, allocation concealment, clear definitions of outcomes, intention-to-treat analysis, and statistical methods. A third investigator reviewed studies for which the first 2 investigators gave discordant quality ratings. We reached final grades for those studies via consensus. We performed subgroup analyses to examine the effects of study quality on the meta-analysis results. We also based our qualitative conclusions on good- or fair-quality studies.
We calculated the rates of breastfeeding initiation and short-term, intermediate-term, long-term, and prolonged breastfeeding for both the intervention and control groups in each study. We recorded the exclusivity of breastfeeding and did the same calculations for the exclusive breastfeeding rates.
We used the rate ratio (relative risk) as the metric of choice to quantify the effectiveness of each breastfeeding promotion intervention. We used the DerSimonian and Laird model for random-effects meta-analysis (88) to obtain summary estimates across studies. We tested for heterogeneity by using the Cochran Q test, which follows a chi-square distribution to make inferences about the null hypothesis of homogeneity (considered significant at P < 0.100) and quantified its extent with I2(89, 90). The I2 statistic ranges between 0% and 100% and quantifies the proportion of between-study variability that is attributed to heterogeneity rather than chance.
We used random-effects meta-regression (fitted with restricted maximum likelihood) to explore whether the effectiveness of breastfeeding interventions depends on breastfeeding duration, provided that at least 6 studies with relevant information were available (91, 92).
We performed subgroup analyses according to various study factors, such as study quality, timing of intervention (prenatal, postpartum, or combined prenatal and postpartum), and different components of breastfeeding interventions. We used a Z test to compare summary estimates between the subgroups.
We used Intercooled Stata, version 8.2 (Stata, College Station, Texas) for all analyses. All P values are 2-tailed and considered significant when less than 0.05 unless otherwise indicated.
The Agency for Healthcare Research and Quality and the U.S. Preventive Services Task Force helped formulate the initial study questions but did not participate in the literature search, determination of study eligibility criteria, data analysis or interpretation, or preparation of the manuscript.
We identified 4877 abstracts in our search and evaluated a total of 147 full-text articles. Thirty-eight RCTs met our eligibility criteria: 32 parallel RCTs described in 33 publications (93–125), 4 clustered RCTs (126–129), and 2 quasi-RCTs described in 3 publications (130–132) (Figure 2). Table 3 shows the 36 trials that were conducted in developed countries (Australia, Canada, Denmark, France, Italy, Japan, Netherlands, New Zealand, Scotland, Sweden, Singapore, United Kingdom, and United States). Two trials on BFHI were conducted in developing countries (Brazil and Belarus).
RCT = randomized, controlled trial.
* All RCTs for question 1 also included for question 2.
The interventions included system-level breastfeeding support (such as BFHI and training of health professionals), formal breastfeeding education, professional support (such as from lactation consultants, midwives, nurses, physicians, or other health professionals), lay support (such as peer support or counseling), motivational interviews, delayed or discouraged pacifier use, and skin-to-skin contact. Several components were often combined into a single, multifaceted breastfeeding intervention.
Eleven trials (29%) were of good quality, 14 trials (37%) were of fair quality, and 13 trials (34%) were of poor quality. The Appendix Table describes the criteria of quality assessment used to reach the overall quality rating for each RCT. Table 3 summarizes the study characteristics.
What are the effects of breastfeeding interventions on child and maternal health outcomes?
The effects of breastfeeding interventions on child health outcomes were reported in 3 RCTs published in 4 articles (93–95, 126). One of these RCTs also reported maternal health outcomes. One good-quality study (126), PROBIT (Promotion of Breastfeeding Intervention Trial), was conducted in Belarus, and 2 fair-quality studies (93–95) were conducted in low-income populations in the United States. We could not combine the results from these RCTs in a meta-analysis because the interventions were dissimilar.
The PROBIT study was a good-quality, cluster, multicenter RCT involving a total of 17 046 mother–infant pairs from urban and rural areas in Belarus. Infants in the intervention group (a modeled BFHI) had a significant reduction in the risk for 1 or more gastrointestinal infections (adjusted odds ratio, 0.60 [95% CI, 0.40 to 0.91]) and atopic dermatitis (adjusted odds ratio, 0.54 [CI, 0.31 to 0.95]) compared with those in the control group but had no significant reduction in respiratory tract infections (126). The 2 fair-quality RCTs, involving a total of 564 mother–infant pairs in low-income families in the United States, reported discordant results. The major drawbacks of these 2 RCTs were high rates of loss to follow-up or missing breastfeeding data. One study showed no significant differences between the 2 groups (hospital and home visits by 2 study lactation consultants vs. usual care) in the risk for gastrointestinal illnesses, respiratory tract diseases, or otitis media (94, 95), whereas the other study found that the risk for 1 or more diarrheal episodes during the study was decreased in the intervention group (home visits by trained breastfeeding peer counselors) compared with the control group (17.5% vs. 37.5%; P = 0.02) (93). The latter study also reported that mothers in the intervention group were less likely than those in the control group to have menses return at 3 months (47.6% vs. 66.7%; P = 0.03). Cessation of menstrual periods for the first few postpartum months during exclusive breastfeeding is a normal physiologic process.
What are the effects of breastfeeding interventions on breastfeeding initiation, duration, and exclusivity?
We found substantial heterogeneity across eligible trials in the actual breastfeeding promotion interventions (including many different combinations of “intervention components”) and their implementation, timing, and intensity (Table 3). Furthermore, the definition of “usual” or “routine” care varied substantially because of differences in background social support and health care systems in the various countries. The sociodemographic characteristics of the study populations also varied.
As shown in Figure 3, breastfeeding promotion interventions resulted in an increased rate of breastfeeding initiation (rate ratio, 1.04 [CI, 1.00 to 1.08]) and short-term breastfeeding (rate ratio, 1.10 [CI, 1.02 to 1.19]) compared with usual care, with significant statistical heterogeneity in both cases. It is questionable whether these trivial effects have any real-world effect. For short-term exclusive breastfeeding, the relative risk was 1.72 (CI, 1.00 to 2.97), again with evidence of statistical heterogeneity (Figure 4). When we excluded the 2 RCTs from developing countries (98, 126), the results for any breastfeeding initiation and short-term breastfeeding were no longer significant. However, intervention effects on short- and long-term exclusive breastfeeding were significant (rate ratios, 1.28 [CI, 1.11 to 1.48] and 1.44 [CI, 1.13 to 1.84], respectively), with evidence for statistically significant heterogeneity for short-term exclusive breastfeeding (I2 = 55%; P = 0.006).
Table 4 describes subgroup analyses performed according to the timing of breastfeeding promoting interventions (prenatal, postnatal, and combinations thereof). Overall, the direction of the effects favors breastfeeding promotion interventions over usual care and was statistically significant for some subgroups (Table 4). We found no clear pattern for the outcome of any breastfeeding with respect to intervention timing. However, for short-term exclusive breastfeeding, the summary point estimates of the corresponding rate ratios are larger for the combination of pre- and postnatal interventions (P = 0.01, Z test).
We performed subgroup analyses on the effects of different components of breastfeeding interventions on breastfeeding initiation, duration, and exclusivity compared with usual care. Again, multiple components were often combined into a single, multifaceted breastfeeding intervention. Our analyses compared only a specific component within a multifaceted intervention with usual care. Indirect comparison of the pooled effect sizes between different intervention components could be misleading, because other components in the intervention and control groups may not be the same across the different subgroups. Overall, we did not find that formal or structured breastfeeding education or individual-level professional support significantly affected the breastfeeding outcomes. We did find that lay support significantly increased the rate of any and exclusive breastfeeding in the short term by 22% (CI, 8% to 48%) and 65% (CI, 3% to 263%), respectively. Meta-regression suggested that larger effects (compared with usual care) were associated with longer duration for any breastfeeding (P = 0.04) (Table 5).
Finally, the summary rate ratios of breastfeeding initiation and duration did not statistically significantly differ across RCTs of different quality grades (data not shown).
Ten RCTs in 11 publications (100, 102, 105, 107, 110, 112, 119, 121, 125, 130, 131) reported the differences in the absolute breastfeeding duration between breastfeeding intervention and usual care groups. The follow-up durations ranged from 2 weeks to 1 year. We did not perform meta-analyses because the intervention components and units of analysis for the breastfeeding outcomes varied greatly across these trials. Seven of the 10 RCTs did not show a significant difference in absolute breastfeeding duration between the intervention and control groups. The other 3 RCTs, 2 of good quality and 1 of fair quality, showed that delayed pacifier use (>4 weeks) was more effective than early pacifier use (within 2 to 5 days) (102) and system-level professional support (105) and postpartum skin-to-skin care (107) were more effective than usual care in increasing breastfeeding duration.
We identified 2 poor-quality RCTs involving family members in breastfeeding intervention. These 2 RCTs were graded poor quality because of incomplete reporting of trial protocol (for example, randomization and blinding) and nonrigorous definitions of breastfeeding outcomes. One study compared the effects of breastfeeding classes for expectant fathers to control group classes of baby care and safety on rates of any breastfeeding initiation and any breastfeeding at 2 months (113). This study found that more women whose partners attended the breastfeeding classes initiated breastfeeding than did women whose partners attended the control class (74% vs. 41%; P = 0.02). However, the rate of any breastfeeding at 2 months did not significantly differ between the intervention and control groups. The other study examined the role of a grandmother (maternal mother) or a close female confidante (sister or friend) of the mother's own choice in supporting breastfeeding (122). This study found no significant difference in breastfeeding initiation or duration between the breastfeeding promotion with a female confidante and the routine prenatal care without a female confidante.
Are there harms from interventions to promote and support breastfeeding?
We did not identify any study specifically designed to examine harms from interventions to promote and support breastfeeding (regardless of design). None of the eligible RCTs reported harms from the breastfeeding interventions.
This systematic review summarizes the effects of primary care–initiated interventions to promote and support breastfeeding with respect to maternal and child health outcomes and breastfeeding outcomes. Although a large number of RCTs have been published since 2001, fewer than one third of them fulfilled most of our quality criteria and another one third had substantial methodological flaws (Appendix Table). We also found great heterogeneity among the actual interventions as well as the background social support and health care systems that constituted usual or routine care across studies. Nonetheless, the RCTs reviewed in this report showed consistent findings. The evidence suggests that breastfeeding interventions can be more effective than usual care in increasing short- and long-term breastfeeding rates. Combined pre- and postnatal interventions and inclusion of layperson support in a multicomponent intervention may be beneficial. Observational data from our previous report (4) showed a relationship between breastfeeding and many beneficial child and maternal health outcomes (Table 1). In summary, only a few RCTs directly examined the effectiveness of breastfeeding interventions on child and maternal health outcomes. Thus, our conclusions about the value of breastfeeding interventions on health outcomes are largely based on an indirect chain of evidence.
Our review has several limitations, which stem mainly from methodological shortcomings of the primary studies and the multitude of possible breastfeeding promotion interventions. First, we found substantial clinical and methodological heterogeneity across studies, which make our summary effects difficult to interpret. This variability in interventions, definitions, and outcomes is not surprising. Breastfeeding schedules and habits are determined by cultural norms, personal desires, and a plethora of socioeconomic factors. To the extent possible, we performed subgroup and sensitivity analyses on factors that may explain the observed heterogeneity. Second, trials of breastfeeding interventions included several individual components. It is impossible to reliably distinguish “independent” effects for these components without performing head-to-head comparisons between them because the effects of individual components cannot be considered independent or additive. Finally, we did not use strict criteria to categorize “primary care–initiated” interventions. Whether a study was classified as primary care–initiated was entirely dependent on the clarity of reporting of the individual studies.
We did not find interventions with formal breastfeeding education or individual-level professional support to be effective in increasing the rates of breastfeeding initiation or duration. However, some evidence suggests that interventions with lay support may be effective in increasing the rates of short- and long-term breastfeeding. This conclusion, however, is based on findings from indirect comparisons of different studies. To further understand the role of lay versus professional support in breastfeeding promotion, future studies should directly compare them in the same population.
Only 2 fair-quality RCTs in developed countries directly examined the effects of breastfeeding interventions on child health outcomes. In both trials, the effects of interventions on rates of exclusive breastfeeding matched the corresponding effects on child outcomes. Specifically, 1 RCT reported an increased exclusive breastfeeding rate at 3 months and a lower risk for diarrheal diseases in the breastfeeding intervention group than in the control group (93). The other RCT did not detect a significant difference in the exclusive breastfeeding rate at 3 months and also did not detect a difference in certain infant health outcomes between the intervention and control groups (94, 95). One may surmise from the above findings that the rate of exclusive breastfeeding may be an important determinant of certain health outcomes in infants. It is unclear whether differences in definitions of exclusive breastfeeding, health outcomes, and unknown factors that could interact with the intervention could also explain some of the different findings. However, these findings stressed the need to further examine the role of postnatal home support for breastfeeding from trained professionals or peer counselors.
Two good-quality RCTs conducted in developing countries (98, 126) provided good evidence that the BFHI is effective in increasing exclusive breastfeeding rates, at least up to 6 months after delivery. The PROBIT (126) also compared infants in the breastfeeding intervention group with those in the control group and showed a significant reduction in the risk for 1 or more gastrointestinal infections and atopic dermatitis. It is conceivable that a cluster randomized study similar to PROBIT in Belarus could be done in the United States, as the BFHI is not yet widely adopted; only 1.3% of the maternity units in this country are designated as baby-friendly (according to http://www.babyfriendly.org). Such a study is important to estimate the public health effect in a sociocultural environment that is not as breastfeeding-friendly as the one in Belarus. To assess the effectiveness of the complete BFHI, it is important to implement all 10 steps (Table 2); none of the studies conducted in developed countries did that.
More cluster RCTs with greater methodological rigor are needed to provide an understanding of the effectiveness of various breastfeeding interventions. Cluster RCTs allow random assignment of groups (such as families or primary care practices) rather than individuals. Cluster studies preempt exposures of intended interventions to nontargeted individuals, thus minimizing cross-contamination of interventions between groups. However, cluster RCTs are more complex to design, require more participants to obtain equivalent statistical power, and demand more complex analyses (133). In addition to proper randomization, the quality of the RCTs can be improved with clear and unbiased patient selection criteria, a common definition of exclusive breastfeeding, reliable collection of feeding data, definition of specific and quantifiable clinical outcomes of interest, and blinded assessments of the outcome. Any substantial differences in the degree of breastfeeding between the intervention and control groups as a result of the breastfeeding intervention will provide further opportunity to investigate any disparity in health outcomes between the 2 groups.
The In the Clinic® slide sets are owned and copyrighted by the American College of Physicians (ACP). All text, graphics, trademarks, and other intellectual property incorporated into the slide sets remain the sole and exclusive property of the ACP. The slide sets may be used only by the person who downloads or purchases them and only for the purpose of presenting them during not-for-profit educational activities. Users may incorporate the entire slide set or selected individual slides into their own teaching presentations but may not alter the content of the slides in any way or remove the ACP copyright notice. Users may make print copies for use as hand-outs for the audience the user is personally addressing but may not otherwise reproduce or distribute the slides by any means or media, including but not limited to sending them as e-mail attachments, posting them on Internet or Intranet sites, publishing them in meeting proceedings, or making them available for sale or distribution in any unauthorized form, without the express written permission of the ACP. Unauthorized use of the In the Clinic slide sets will constitute copyright infringement.
Results provided by:
Copyright © 2016 American College of Physicians. All Rights Reserved.
Print ISSN: 0003-4819 | Online ISSN: 1539-3704
Conditions of Use
This PDF is available to Subscribers Only