Materials and Methods
This study protocol was approved by the institutional review board of the University of Rochester School of Medicine and Dentistry. Because of a lack of empiric data about formula advertising materials and effects on breast-feeding initiation, sample size calculations were based primarily on data from studies of discharge pack effects on breast-feeding duration.4,7 On the basis of local data, we expected about 70% of the study population would choose to breast-feed. Anticipating prenatal losses to the cohort of approximately 10%, we randomized 547 women to achieve a sample size of 490 for analyses of breast-feeding initiation (85% power [α = .05] to detect a 13% different in initiation). Larger than expected prenatal losses resulted in a sample size of 444 women (235 in the commercial and 209 in the research group) and provided 80% power (α = .05) to detect a 15% difference in breast-feeding initiation between groups. For the postnatal study of breast-feeding duration, a sample size of 160 dyads in each study group (n = 320) was desired to provide 90% power (α = .05) to detect a mean difference of 18 days (standard deviation [SD] 20 days) between groups (and specified subgroups).4,7
Five hundred forty-seven consecutive women presenting for their first prenatal visits at any one of six obstetric outpatient settings in Rochester, New York, were randomized to receive one of two distinct educational packs about infant-feeding methods. Both packs contained information about breast-feeding and formula feeding but emphasized breast-feeding as optimal infant nutrition. One pack contained formula company–produced materials (commercial, n = 277), and the other contained materials with similar subject matter, but selected to eliminate all forms of formula advertising (research, n = 270). Potential participants were identified by regularly reviewing first prenatal appointments scheduled at each clinical site. Using computer-generated random-number lists, randomly sized blocks of women were stratified further by obstetric practice before assigning the intervention group. On a weekly basis, study packs, commercial and research, prelabeled with the participant's name and a study identification sticker, were dispensed to the clinical sites. Packs were distributed at the end of the prenatal visit by an office nurse who was masked to the study hypothesis.
To maintain masking of the office personnel, every effort was made to ensure uniformity in appearance (size, shape, and outer packaging) as well as volume of the packs. The commercial pack contained a formula company diaper bag; formula company–produced educational materials on pregnancy, infant feeding, and infant growth and development; a can of powdered formula; a business reply card to join a so-called baby club redeemable for a case of infant formula; and several infant formula discount coupons. The research pack contained a generic diaper bag; noncommercial educational materials on pregnancy, infant feeding, and infant growth and development; a coupon redeemable for $5.00 worth of infant items at a local department store; and a package of electrical socket outlet covers.
Until the start of this trial, participating offices distributed routinely formula company–produced diaper bags containing infant-feeding information and formula samples as well as similar pregnancy information to all obstetric patients at their first prenatal visit.1,2 Because of the popularity of these materials, participating obstetricians requested that both intervention groups receive packs of similar subject content, monetary value, and aesthetic appeal. Thus, the research packs were designed to provide these, but specifically to eliminate formula promotion items and samples.
Informed consent was not obtained for the antenatal portion of the study; however, a letter contained in both packs informed women that the office was participating in an evaluation of educational materials. A short questionnaire contained in the pack, precoded with the study identification number, and office logs of pack distribution were used to verify randomization. No randomization errors were detected.
After randomization, women were tracked throughout pregnancy. A total of 444 women presented for delivery at Rochester General Hospital, the inpatient obstetric study site. During their postpartum stay, participants were approached and asked to participate in an interview about how they decided to feed their infant. Standardized interviews were administered by a research assistant who was masked to the hypothesis and to group assignment. Interviews included questions regarding family composition, work force participation, past infant-feeding experiences, familial support, and personal beliefs/choices about infant-feeding methods. Mothers who chose to breast-feed their infants were asked additional questions about their breast-feeding experience in the hospital, problems, and goals. Maternal (prenatal and hospital) and infant (hospital) chart reviews were conducted for all participants. Demographic data were collected on all participants from birth certificates.
Women who chose to breast-feed their infants and delivered term, healthy infants were recruited to participate in a postpartum follow-up study of breast-feeding. Informed consent was obtained. Telephone interviews were conducted by trained nurse interviewers masked to the study hypothesis and group assignment at 2, 6, 12, and 24 weeks postpartum. Information was gathered on infant feeding, including breast-feeding frequency, duration, and maternal and infant problems; proportion of diet composed of breast milk; and timing of introduction and use of various supplemental foods and/or liquids. Using protocols from the hospital's mother-baby telephone support line, interviewers also offered assistance with maternal, infant, or breast-feeding concerns at the end of each contact.
Losses to the cohort were 19% (103 of 547) in the prenatal phase of the study and were due to miscarriage (n = 71), randomization of a small number of nonpregnant women (n = 3), family relocation (n = 26), and multiple gestations (n = 3). Peripartum interviews were completed for 437 of 444 participants who presented to Rochester General Hospital for delivery (231 commercial group, 206 research group). In seven cases, we failed to identify participants during their hospital admissions. Although data are not complete for these participants, hospital chart reviews (maternal and infant) were conducted and birth certificate data were obtained so that sociodemographic and infant-feeding information is available (initial decision and feeding method at hospital discharge). Attrition from the study did not vary significantly by study group.
A cohort of 311 breast-feeding women of 444 participants was identified postpartum. Two breast-feeding women were missed in the hospital, 15 completed peripartum interviews but refused follow-up, and 15 were ineligible for follow-up because of maternal or infant complications at birth. Eleven additional women were lost to follow-up before 24 weeks postpartum. Overall, losses to the cohort from first prenatal visit to 6 months postpartum were 26.7% (146 of 547). Participants excluded from analyses of breast-feeding duration (n = 17; two missed, 15 ineligible) did not differ from those included in the analyses for race, parity, mode of delivery, marital status, previous breast-feeding experience, education, age, or infant gestational age. However, they tended to have lower overall goals for breast-feeding (21.1 compared with 36.8 weeks, P = .05) and lower birth weight infants (3244 ± 775 compared with 3576 ± 447 g, P = .09)
In this study, exclusive breast-feeding was defined as receiving only breast milk. Full breast-feeding included the infrequent use of water, juice, or ritualistic feeds, and overall breast-feeding was defined as the length of time an infant received any breast feedings.8
Data were entered and analyzed using SPSS (SPSS Inc., Chicago, IL) and Epi Info 6 (Centers for Disease Control and Prevention, Atlanta, GA). The effect of the two infant-feeding information packs (commercial or research) on breast-feeding initiation, duration, and other outcomes was tested. Survival analyses (Kaplan Meier and Cox proportional hazard models) were used to evaluate continuous outcomes (eg, breast-feeding duration in days), and χ2 and logistic regression analyses were used to evaluate discrete outcomes (eg, breast-feeding initiation, in-hospital termination). Reported P values are two-sided and were considered statistically significant if ≤ .05.
Previous research identified women who became ill postpartum as a vulnerable group.6 In these data, having undergone a cesarean was used to identify at-risk women because it was correlated highly with peripartum illness and problems associated with labor and delivery.
In adjusted analyses, potentially confounding factors were selected as those that showed some difference between the two intervention groups and were judged to have a potentially independent effect on breast-feeding, but not to be part of the causal pathway. Each potentially confounding variable was tested to assess any meaningful effects on the size of the odds ratio (OR) or hazard ratio for the intervention variable. Candidate variables were then tested in increasingly complex models, eliminating those that failed to have an effect on the OR of the intervention variable in tandem with other covariates. Unadjusted results are presented unless there were substantial changes to the effect size of the intervention after adjustment for potential confounders.
Four hundred forty-four women are included in the analyzes of breast-feeding initiation (Commercial intervention = 235 and research intervention = 209), whereas analyses of in-hospital breast-feeding termination include women who chose to breast-feed their infants (n = 311). Analyses of breast-feeding duration (n = 294) excluded 15 mother-infant dyads with medical complications at birth and two dyads with incomplete data (no peripartum interview or follow-up conducted).
The obstetric practice where women received care was tested as a potentially confounding factor and did not exert a significant effect on any outcome. Table 1 presents group characteristics of the 444 women who completed the prenatal phase of the study protocol. Intervention groups did not vary for maternal age, education, parity, ethnicity, mode of delivery, previous breast-feeding experience, plans to return to work, and infant gestational age or birth weight. Participants were largely white (94.4%) and privately insured (96.8%), and most had plans to return to work within 6 months (60%).
Overall, 43.5% of the 311 women who choose to breast-feed their infant were uncertain about how long they wanted to breast-feed or had personal goals of 12 weeks or less (commercial 42%, research 46%, P = .55). First-time breast-feeding women were more likely to have uncertain goals or goals for breast-feeding of 12 weeks or less compared with those who had previously breast-fed a child (51.4% compared with 36.2%, P = .01). Primiparous women also were more likely to have uncertain or personal goals for breast-feeding of 12 weeks or less compared with multiparous women (51.6% compared with 37.6%; P = .02). Most multiparous women who choose to breast-feed their infant had previously breast-fed a child (92.1%). Women with plans to return to work within 6 months also were more likely to have uncertain goals or goals for breast-feeding of 12 weeks or less (71.9 compared with 48.8, P < .001)
Eighty-seven percent of participants reported that they had read the materials they received from their obstetrician about infant feeding and found them to be helpful (commercial, 84% compared with research, 90%, P = .12). In the commercial intervention group, 21% of women reported the use of a formula company business reply card to receive a case of formula before the birth of their infant.
The small difference in breast-feeding initiation between the two groups (1.4% difference) was not statistically significant, suggesting no effect of the intervention on women's choice of an infant-feeding method (Table 2). After adjusting for women's plans for work, previous breast-feeding experience, and marital status, the OR was 0.99 (95% CI 0.64, 1.55). Subgroup analyses also showed no significant effects on breast-feeding initiation among primiparous women (OR 1.04, 95% CI [0.57, 1.92]), those with low educaitonal attainment (OR 0.93, 95% CI 0.49, 1.75), or those who delivered by cesarean (OR 0.93, 95% CI [0.38, 2.27]). Eleven percent of women reported changing their minds during pregnancy about their choice of infant-feeding method. The likelihood of changing methods did not vary by study group (P > .5).
Among women who chose to breast-feed their infants, the intervention did not significantly affect the length of personal goals (commercial 26.5 weeks ± 17.4 compared with research 26.6 ± 18.7 weeks, P = .54) or the likelihood of having an uncertain goal (commercial 16.1%, research 17.4%, P = .88). Goal attainment also was unaffected (relative risk [RR] 1.12; 95% CI 0.86, 1.47) and was not altered by adjustment for previous breast-feeding experience, maternal education, cesarean delivery, and maternal age (OR 1.01 [95% CI 0.75, 1.35]). For goal attainment, women who refused (n = 15), were lost to follow-up (n = 11), terminated breast-feeding before hospital discharge (n = 14), or had uncertain goals (n = 42) were excluded from the analysis.
Breast-feeding cessation before hospital discharge and before 2 weeks postpartum were used to assess effects of the intervention on early breast-feeding cessation. Peripartum breast-feeding cessation (in-hospital) was significantly higher in women exposed to the commercial intervention. Relative risks for early cessation were approximately five-fold higher in the commercial group (Table 2). Adjustment for potential confounders (maternal age, planned return to work, previous breast-feeding experience, marital status, and socioeconomic status) increased the strength of the associated OR to 10.3 (95% CI 1.3, 81.2; P = .03).
Breast-feeding termination within the first 2 weeks postpartum was higher in the commercial intervention group across all categories of breast-feeding, with findings approaching statistical significance for overall and exclusive breast-feeding in unadjusted analyses (overall breast-feeding RR 1.58 [95% CI 0.97, 2.56; P = .06], full RR 1.26 [95% CI 0.91, 1.74; P = .16], and exclusive breast-feeding RR 1.25 [95% CI 0.98, 1.59; P = .07]) (Table 2). After adjustment for potential confounders (maternal age, planned return to work, previous breast-feeding experience, marital status, and socioeconomic status), results for both overall and exclusive breast-feeding were statistically significant ([overall OR 1.91, 95% CI 1.02, 3.55; P = .04] [exclusive OR 1.65, 95% CI 1.02, 2.68; P = .04]), and results for full breast-feeding approached statistical significance (OR 1.66, 95% CI 0.995, 2.77; P = .05).
Subgroup analyses demonstrated that women exposed to the commercial intervention, with uncertain goals for breast-feeding or goals of 12 weeks or less had significantly increased risks for cessation at 2 weeks (overall breast-feeding, RR 1.55 [95% CI 1.12, 2.14; P = .02], full breast-feeding, RR 1.60 [95% CI 1.12, 2.26; P = .008], and exclusive breast-feeding, RR 1.66 [95% CI, 1.11, 2.48; P = .01]). Adjustment for previous breast-feeding experience, plans to return to work, and parity did not alter these results. Results for all other subgroups demonstrated nonstatistically significant differences by treatment group.
Small decreases in mean and median duration were noted in each category of breast-feeding among women exposed to the commercial intervention (Table 3). None of these differences, however, reached statistical significance (Table 4). There were no significant differences between the study groups with respect to reported reasons for breast-feeding termination (P > .5).
Differences in breast-feeding duration among primiparas, women of low educational attainment, and women who underwent cesarean did not demonstrate significant effects due to the intervention. Subgroup analyses, however, demonstrated statistically and clinically significant declines in duration as a result of the commercial intervention among women with goals of 12 weeks or less, or undefined personal goals for breast-feeding (Table 4). In this group, which constituted 43.5% of the breast-feeding women in this study, women exposed to the commercial packs compared with those exposed to the research packs experienced a 35-day decline in mean overall breast-feeding duration. Mean full duration was reduced by 20 days, and mean exclusive duration was reduced by 11 days (Figure 1).
Breast milk is widely recognized as the optimal way to nourish an infant9–11 and is of such importance to the health of mothers and infants that United States national health goals for the year 2000 specify that 75% of new mothers will initiate breast-feeding and 50% will continue to breast-feed their infants to 6 months.12 Currently, only about 59% of postpartum women in the United States initially breast-feed their infants, and as few as 22% are still breast-feeding at 6 months.13
It is well documented that both advertising and attitudes of medical care providers influence women's choice of infant-feeding methods,10,14–17 placing obstetricians in a critical position to influence the choice to breast-feed.10,15,18 Whereas other studies have shown that the postnatal distribution of commercial formula materials affects adversely breast-feeding duration,4–6 this trial was conducted to evaluate the effects of distributing such materials prenatally.
Women's choice of an infant-feeding method was not affected by the two types of infant feeding materials. Only minimal differences in breast-feeding initiation were noted between the study groups (1.4% difference). In this study, 61% of women had made their decision before pregnancy, and only 11% reported changing their minds during pregnancy. Moreover, the intervention did not appear to influence those who changed their minds. It appears likely that compared with other sociocultural factors, patient education materials of the type tested in this study exert little influence on choice of infant-feeding method.
Breast-feeding cessation in the early peripartum period, however, was significantly higher in women exposed to the commercial intervention. Relative risks for cessation before hospital discharge were approximately five-fold higher. The attributable risk in the group exposed to the commercial packs was 6% (the rate in the commercial [12 of 163 = 7.4%] minus the rate in the research group [two of 148 = 1.4%]). For every 100 women exposed prenatally to commercial materials in this study, six women quit breast-feeding before hospital discharge because of that exposure (one of every 17 women).
Breast-feeding cessation before 2 weeks also was significantly higher in women exposed to the commercial packs. Formula supplementation is known to interfere with the successful establishment and maintenance of mothers' milk supply and may explain the effects of the intervention on early lactation success.10 Women in the commercial intervention received at least one sample of formula, and 21% of breast-feeding women in this group reported using a business reply card to obtain additional formula supplies before delivery. Ready access to a supply of formula could result easily in increased formula supplementation. Taken as a whole, the observed increases in early breast-feeding cessation support the conclusion that the potential for formula promotion materials to undermine breast-feeding is great.
With respect to long-term breast-feeding, the commercial intervention did not significantly affect breast-feeding duration either in the larger study group or among primiparas, those who underwent cesarean, or women of low educational attainment. One might speculate that timing of the intervention, at first prenatal visit, rather than later in pregnancy dampened potential effects on breast-feeding duration. This study, however, was designed to mimic obstetric practices in our community, and we believe it to be a valid evaluation of the effect of obstetric distribution of formula promotion products on breast-feeding.1,2 Effects on longer-term breast-feeding duration were thus minimal in this setting and within this population of largely white, married, middle-class, well-educated women.
Despite the social advantages of this study population, long-term breast-feeding duration, however, was shortened significantly among women with uncertain breast-feeding goals or goals of 12 weeks or less, a group that constituted 43.5% of the study population. Effects were noted across all breast-feeding categories in these women who experienced average declines of 35 days in overall, 20 days in full, and 11 days in exclusive breast-feeding. Declines of this magnitude among women with uncertain goals or goals of 12 weeks or less are not surprising. Formula promotion materials deliver equivocal messages about the desirability and ease of breast-feeding and are designed to enhance the use of formula as a sole or complementary feeding method.19 Such messages may be enhanced when provided by a trusted medical authority.20,21 Additionally, the return of formula company business reply cards enrolls women in so-called new baby clubs whose benefits include the repeated delivery of formula supplies and promotional materials throughout the prenatal and early postpartum periods.
Although we hypothesized that exposure to the commercial intervention would affect adversely women's personal goals for breast-feeding duration, there do not appear to be any significant effects on either the length of goals or the likelihood of attaining a stated goal. In secondary analyses of these data, women's personal goals were demonstrated to be very strong predictors of breast-feeding duration.22 Declines in breast-feeding duration among women with uncertain goals or goals of 12 weeks or less suggest that for some women, commitment to breast-feeding may be affected by exposure to the commercial materials.
Limitations of this study include a lack of socioeconomic and racial diversity in the study population. Generalizability of the results to all women of child-bearing age is limited, and results in more vulnerable populations may differ. Women with uncertain or relatively short breast-feeding goals may be even more prevalent in populations of impoverished and minority women, populations for which breast-feeding duration is known to be shorter and initiation is lower.13 Further research is needed to confirm or refute the results of this study in such populations. In assessing the results of this study, readers also should note that breast-feeding education and support were widely available for this population. Many women attended prenatal breast-feeding classes or read prenatally about breast-feeding (97% of cohort). This study was conducted in a hospital that complies with baby-friendly recommendations and employs lactation consultants to provide in-hospital and postpartum breast-feeding support. Furthermore, women in this study had contact with study nurses who offered assistance at each postpartum contact. Results of similar interventions in communities with less supportive environments may vary.
Advertising materials from nonobstetric sources (eg, direct mailings to the home) also may have exposed both intervention groups to formula promotion materials and lessened the effects of the intervention. Fifty-six percent of women in the research group reported the prenatal receipt of formula company promotion items from sources other than their obstetric provider. Postnatally, approximately 40% in both intervention groups reported the receipt of such items. Such exposures could lessen differences between the groups (cointervention in both groups biasing results toward the null) and make detection of effects of the intervention more difficult.
Our findings are limited to a comparison of the effects of educational materials that include formula advertising and samples and educational materials that eliminate such items and do not distinguish between detrimental effects of the commercial intervention and beneficial effects of the research intervention. The isolated postpartum use of breast-feeding educational materials (noncommercial), however, has not been shown to prolong breast-feeding,23 and the distribution of formula discharge packs in the peripartum setting is known to shorten breast-feeding.4–7 Infant-feeding educational materials that comply with the WHO international code are widely available at low or no cost from state health departments, ACOG, and the American Academy of Pediatrics.