Objectives. Pacifiers are related to a shorter duration of breastfeeding. However, it is unclear whether this association is causal, because confounding, reverse causality, and self-selection of mothers may play a role. These issues were investigated through a combination of epidemiologic and ethnographic research in southern Brazil.
Methodology. A population-based cohort of 650 mothers and infants were visited shortly after delivery and at 1, 3, and 6 months. The rate of complete follow-up was 96.8%. A subsample of 80 mothers and infants was selected for the ethnographic study, which included in-depth interviews and participant observations in the age range of 2 to 6 months with a mean of 4.5 visits.
Results. The epidemiologic study showed that pacifier use was common with 85% of users at 1 month. However, this was a dynamic process, with many infants starting or abandoning the pacifiers in any age range. Children who stopped breastfeeding in a given period were likely to take up the pacifier during that period. Further analyses excluded all infants not breastfed at 1 month of age and those who reportedly had breastfeeding problems, leaving 450 infants with full data. Intense pacifier users at 1 month (children who used the pacifiers during most of the day and at least until falling asleep) were four times more likely to stop breastfeeding by 6 months of age than nonusers. Users also had fewer daily breastfeedings than nonusers. After adjustment for several confounding variables, logistic regression showed that pacifier use was still associated with an odds ratio of 2.5 (95% confidence interval, 1.40 to 4.01) for stopping breastfeeding. The ethnographic analysis showed that pacifier use was widely regarded as a positive behavior and that mothers often strongly stimulated the infants to accept it. Although few mothers openly admitted that pacifiers might shorten breastfeeding, a considerable group effectively used pacifiers to get their infants off the breast or to increase the interval between feedings. The latter also had rigid breastfeeding styles that increased maternal-infant distance, had important concerns about objective aspects of infant growth and development, and were highly sensitive to infant crying. These behaviors were linked to intense comparison between themselves and other mothers and to a lack of self-confidence. Nonwhite mothers, those who delivered vaginally, and mothers of infant girls seemed to be more confident and less affected by these difficulties. The epidemiologic analysis confirmed that pacifier use was more closely associated with breastfeeding duration among nonwhite mothers and for normally delivered infants.
Conclusions. Pacifiers may be an effective weaning mechanism used by mothers who have explicit or implicit difficulties in breastfeeding, but they are much less likely to affect infants whose mothers are confident about nursing. Breastfeeding promotion campaigns aimed specifically at reducing pacifier use will fail unless they also help women face the challenges of nursing and address their anxieties. The combination of epidemiologic and ethnographic methods was essential for understanding the complex relations between pacifier use and breastfeeding.
The avoidance of pacifiers was included as 1 of 10 steps for successful breastfeeding in the 1990 Innocenti Declaration,1,2 although there was no epidemiologic evidence of an association at that time. Pacifiers are often used by infants in many less developed countries,3,4 and a short duration of breastfeeding may lead to increased morbidity and mortality attributable to infectious diseases.5 It is therefore important to establish whether pacifier use may indeed affect breastfeeding, because this will influence promotion policies.
The epidemiologic association between pacifier use and breastfeeding duration seems to have been initially documented in a cross-sectional study in 19936 and has been confirmed by other studies since then.7-9 However, none of these studies were specifically designed to test this association, and relevant data for its interpretation may not have been collected. In addition to the issue of confounding, two other major problems may be present: (1) reverse causality—breastfeeding difficulties may lead to pacifier use rather than the reverse; and (2) self-selection bias—behavioral characteristics of the mother or infant that are often impossible to assess using standard epidemiologic techniques may lead both to pacifier use and to shorter breastfeeding.
In this study, we have attempted to tackle these complex issues through a combination of epidemiologic and ethnographic research in an urban setting in southern Brazil. In the former, a birth cohort of children were followed up prospectively: (1) to describe pacifier use and breastfeeding patterns; (2) to investigate the association between pacifier use and subsequent breastfeeding; (3) to check reverse causality; (4) to understand the mechanisms mediating this association; (5) to rule out a large number of possible confounding variables; and (6) to identify factors that may modify the relation of pacifiers to breastfeeding.
The simultaneous ethnographic study was aimed at further understanding issues related to causality by specifically exploring: (1) how much and why mothers value pacifier use; (2) how mothers stimulate pacifier use and, if so, whether they are aware of their own stimulating behaviors; (3) how readily the infants actually take the pacifiers; and (4) the presence of self-selection.
Pelotas is a city of 300 000 in the extreme south of Brazil, a relatively developed part of the country. The five maternity hospitals in the city were visited daily from January to December 1993 in an attempt to interview all women giving birth. A total of 5304 interviews were carried out; there were only 16 losses and refusals (0.3%). A systematic sample of 655 infants were selected for follow-up at the ages of 1, 3, and 6 months. It was not possible to locate 6 of these children at the age of 1 month (2 refusals and 4 whose families moved away). At the 3- and 6-month visits, respectively, 11 and 28 children could not be located (2 and 4 refusals, 9 and 24 who moved away). Therefore, 96.8% of the original cohort were successfully traced throughout the 6 months of the study.
Mothers were interviewed regarding pacifier use, breastfeeding patterns, and socioeconomic, environmental, and reproductive variables. Birth weight was obtained from hospital records. Breastfeeding duration refers to the total duration of any breastfeeding. Infants were considered pacifier users if they sucked pacifiers every day. Full-time users were infants reported to use the pacifiers during the whole day as well as either during the whole night or until falling asleep. The remaining users were considered part-time. Information was also collected on other variables that might affect breastfeeding duration, including the intake of teas, water, and nonhuman milk, obtained through a 24-hour recall, and bottle feeding.
Eighty-eight infants (13.4%) were no longer breastfed at the age of 1 month, when the information on pacifier use was collected. They were excluded from further analyses of breastfeeding duration.
Because pacifier use might be an indication of breastfeeding problems, we also decided to exclude mothers reporting such problems when the infants were 1 month old (19.8% of the sample). Only 42.2% of the latter were still breastfeeding at 3 months, compared with 73.6% of those who did not report difficulties (P < .001). Pacifier use was slightly higher (18.0%) among the former than the latter (11.7%; P = .17). The analyses were repeated after also excluding infants whose mothers believed that pacifiers affect breastfeeding, but this did not change the previous results.
Variables were considered as potential confounders when previous research showed that they might be associated with both breastfeeding duration and pacifier use or when the ethnographic findings suggested that this might be the case. Information on 15 potential confounders was collected at the hospital interview and during the 1-month visit. The variables included: (1) how long the mother intended to breastfeed (months); (2) whether colostrum was given (yes or no); (3) whether breastfeeding was started in the hospital (yes or no); (4) whether there were difficulties in establishing breastfeeding (yes or no); (5) breastfeeding style (demand or schedule); (6) whether the mother reported difficulties in getting the infant off the breast after feeding (yes or no); (7) how the mother takes the child off the breast after feeding (uses finger, pacifier, bottle, or none); (8) family income (minimum wages); (9) maternal schooling (highest grade completed); (10) maternal age (years); (11) low birth weight (yes or no); (12) birth order (first born or others); (13) gender (male or female); (14) maternal skin color (white or nonwhite, the latter being African-Americans); and (15) type of delivery (vaginal or cesarean). Only 5 of these variables were associated (P < .20)10 both with pacifier use at 1 month and with breastfeeding at 3 or 6 months: maternal age, low birth weight, gender, breastfeeding at the time of hospital discharge, and breastfeeding on demand or schedule at 1 month of age. These five variables therefore fulfilled the criteria for a possible confounding effect and were adjusted for in all analyses. The final analyses were repeated including all 15 variables listed above, yielding very similar results.
Maternal skin color and type of delivery were also included in the multivariate analyses because there was a suggestion from the ethnographic study that these variables—along with gender—might modify the association between pacifier use and breastfeeding.
Associations between confounding variables and both pacifier use patterns and breastfeeding prevalences at 3 and 6 months were assessed through the χ2 test. This test was also used in the crude analyses of breastfeeding prevalences. Multivariate models were carried out separately for three different outcomes; breastfeeding prevalences at 3 and 6 months were analyzed through unconditional logistic regression, and total breastfeeding duration (in days) was studied through the Cox regression, a type of survival analysis. Information on infants who were still breastfeeding at 6 months of age were treated as censored information. Because the three approaches resulted in virtually identical results, only the latter will be presented, because it makes the best use of the available data.
The ethnographic study included a subsample of 80 mothers from those visited at the 1-month follow-up. Using a quota sampling frame, mothers were chosen according to schooling and age groups. Only mothers from four large predetermined neighborhoods were chosen to ensure a variation in access to health services and migration patterns. Within each quota and neighborhood, mothers were selected randomly. A comparison of this subsample with all women who gave birth in 1993 showed that they were highly similar. First interviews were conducted by four field workers (including D.P.B.) at approximately 2 months and then periodically from 3.5 to 6 months. The number of visits ranged from 3 to 10 (mean, 4.5). Methods included informal conversation, semistructured interviewing, participant observation, and semistructured observations of breastfeeding and pacifier use behaviors in the home at 2 months. These observations were conducted in a “natural” setting rather than through formal observations in which the interviewer does not interact with the mother. Detailed ethnographic descriptions were written after the observations. This method was chosen, because previous field work in the area showed that this was less likely to bias behaviors. At the end of the field work, behavioral aspects showing greatest heterogeneity were quantified with a precoded observational questionnaire filled out by the field worker with the help of previously written field notes. The analysis included a mixture of qualitative and quantitative methods. Despite the small sample size, some associations were strong enough to reach statistical significance, assessed through Pearson's correlation coefficients and χ2 and t tests.
A grounded theory approach was used throughout the analysis.11,12 This approach is an inductive, systematic data collection process relying on semistructured interviewing and observational methods that adapt to the data collection setting. Relevant variables were identified as those that most strongly emerged from the observations themselves rather than from the researchers' preconceptions. This process elicits data of higher validity, closer to the reality experienced by a society's member.
Informed consent was obtained from all mothers participating in the study, and the research protocol was approved by the Faculty of Medicine's ethical review board.
The study sample was made up of 329 boys and 326 girls. Their mean birth weight was 3199 g, and 8.1% weighed less than 2500 g at birth. The mean schooling level of their mothers was 6.8 years, and only 2.8% had not attended school. The average maternal age was 26.1 years, and 14.8% of them were younger than 20 years. Most families (64.8%) had a family income of US$300 per month or less.
The median duration of breastfeeding was 102 days, and only 15 children (2.3%) were never breastfed. At the ages of 1, 3, and 6 months, the percentages of breastfed infants were, respectively, 86.4%, 58.2%, and 34.4%. When this analysis was restricted to infants who were breastfed at 1 month and whose mothers did not report breastfeeding difficulties at that age, the prevalence at 3 and 6 months increased to 73.6% and 42.6%. The median duration then became 143 days.
Pacifiers were introduced very early. Almost half of the mothers took them to the hospital. Table 1 shows that, at 1 week of age, about 7 of every 10 infants were already using pacifiers. This proportion increased to approximately 85% at 1 month of age and remained at this level at 3 and 6 months. Table 1 also shows the frequencies of pacifier use patterns. The proportion of full-time users increased slightly with age.
Changes in pacifier use patterns between 1 and 3 months were common. Of nonusers at 1 month, 36.5% became users at 3 months. Of part-time users at 1 month, only 6.4% quit the habit by 3 months, and 11.5% evolved to full-time users. Of full-time users, none abandoned the habit, but 47.5% became part-time users.
The effect of stopping breastfeeding on the initiation of pacifier use was also investigated (Fig 1). All 12 nonusers at 1 month who stopped breastfeeding between 1 and 3 months became pacifier users, compared with 26% of those who continued breastfeeding during this interval. The respective percentages for children 3 to 6 months of age were 75% and 26%. Of the 12 children mentioned above, 6 started using the pacifier after full weaning, and the other 6 started before. This suggests that causality operates in both directions, implying that reverse causality is taking place.
Maternal opinion on whether pacifiers affected breastfeeding was asked directly in the epide-miologic questionnaire; most mothers (86.3%) of 1-month-old infants answered negatively. There were fewer pacifier users (70.6%) among children whose mothers believed in this association, compared with 88.2% among the remaining infants (P < .001). The corresponding use rates at 3 months were 63.2% and 88.7%, respectively (P< .001). Among infants who were breastfed at the age of 1 month, those whose mothers believed that pacifiers affected breastfeeding were less likely to be fully weaned by 3 months (19.2% versus 34.7%;P = .01). Reported belief in the association between pacifiers and breastfeeding, therefore, was associated both with less intense pacifier use and with longer breastfeeding duration. This variable was included in the multivariate analyses as a confounder.
Other factors that might directly affect breastfeeding were also addressed. Bottles were used by 84.6% of the infants at 1 month and 90.1% at 3 months. Exclusive breastfeeding was uncommon: only 15.1% at 1 month of age. Herbal teas or water were being given to 66.4% of all infants at the age of 1 month. At 3 months, water was being given to all but 19 infants, and herbal teas were being given to 65.6% of all children. In addition, nonhuman milk was consumed by 38.2% of the infants at 1 month (14.5% formula and 24.3% cow milk; a few infants received both). All but 2 infants consuming liquids or nonhuman milk received them through bottles.
The forthcoming analyses were restricted to 450 infants who were breastfed at the age of 1 month and whose mothers did not report breastfeeding difficulties. At this age, the association between pacifier use patterns with the frequency of breastfeeding and the intake of liquids and nonhuman milk was investigated. Among breastfed infants, the number of feeds decreased with more intense pacifier use (Table 2). The proportions of breastfed infants who also received nonhuman milk increased from 12.3% among nonusers to 37.0% among full-time users, and a similar pattern was observed for the use of teas or water. Among infants who received both human and nonhuman milk, intake of the latter also tended to be lower for full-time pacifier users than for nonusers, although this difference did not reach statistical significance. On the other hand, the intake of teas and water tended to be greatest among full-time users. Therefore, the frequency of breastfeeding and the intake of nonhuman milk tended to decrease with intense pacifier use, whereas the reverse was observed for liquid intake.
There was a very strong association between pacifier use at 1 month of age and breastfeeding duration (Table 3). Nonusers were four times more likely to be breastfed at 6 months than full-time users. Figure 2 shows the life table analysis of breastfeeding duration according to pacifier use.
In the unadjusted analyses, the use of teas and water, cow milk, or formula and of bottles at the age of 1 month were all associated with subsequent breastfeeding duration (Table 4). The association with use on nonhuman milk was particularly strong. Both formula and cow milk use were equally associated with breastfeeding duration. Because bottles were used for feeding either liquids or milk, this variable is not included in the multivariate analyses to avoid redundancy.
Breastfeeding was longer for older mothers, infants with birth weights of 2500 g or more, girls, those breastfed while still in the maternity ward, and infants breastfeeding on demand. On the other hand, pacifier use was more frequent among infants of young mothers, those with low birth weights, boys, those not breastfed in the maternity ward, and those breastfeeding on schedule. These confounding variables, along with maternal opinion on whether pacifiers affect breastfeeding, were included in the forthcoming multivariate analyses.
Table 5 shows the results of the Cox regression, expressed as hazard ratios or relative risks of stopping breastfeeding. The second column from the left shows the unadjusted ratios for each of the three risk factors. Both full-time pacifier use and intake of cow milk or formula were associated with fourfold increases in the rate of weaning, whereas the use of teas and water increased the risk by 44%. When these variables were adjusted for one another, the hazard ratio associated with nonhuman milk remained about fourfold, whereas full-time pacifier use was associated with a ratio of approximately 2.5, and teas and water were associated with a 30% increase. The last column in Table 5 shows that these effects remained almost unchanged after further adjustment for confounding variables.
The ethnographic study confirmed that pacifier use was considered a normal behavior. Pacifiers were often stated to be soothing, “cute,” “pretty,” and a symbol of social status (“a luxury”). Pacifiers are even regarded as growth promoters by leading to the earlier acceptance of foods other than breast milk, because liquid and semisolid foods such as soups and black beans are often first introduced by dipping the pacifier in them.
Pacifiers were offered to virtually every infant in early life, and nonuse seemed largely the infants' decision. Infants who refused were labeled as “choosy,” and their parents were blamed for lack of discipline, that is, for not stimulating acceptance of the pacifier. A rejection of the pacifier also implied that the mother was most likely allowing the child to use her breast as a pacifier, because it is widely believed that most children prefer the breast if not taught otherwise. A general notion is that one should make the child wait to be breastfed rather than to give in to the demands of a whiny or overly needy child.
Although most mothers acknowledged that pacifiers might affect dental development, only 13 of 80 were explicitly aware of the association with breastfeeding. These mothers were both more likely to breastfeed at the time of the interview, and their children were less likely to use pacifiers. These results are similar to the epidemiologic questionnaire findings. However, the ethnographic study identified a group of 17 mothers who, although not openly admitting the association, reported using the pacifiers to make the infants wait to be breastfed either “on schedule” or once the mothers had free time. These mothers were labeled as having “implicit” awareness of the association. They were more likely to express anxiety and difficulty in making their infants accept the pacifiers and commonly reported withholding breastfeeding with the purpose of making the infants hungry and more likely to take the pacifiers. Women with implicit awareness also gave greater importance to the use of a pacifier, often praising its benefits. Investigating explicit awareness, therefore, did not seem to be the most adequate way of determining the role of the pacifier in weaning.
The ethnographic study measured intensities of pacifier use and maternal stimulation toward its use. Pacifier use was characterized as a score ranging from 0 to 4, according to how many of the following behaviors were observed: (1) the infant was using the pacifier (in the mouth) when the interviewer arrived; (2) the infant was rocked to sleep with the help of the pacifier; (3) the infant was observed sleeping with the pacifier in the mouth; and (4) the infant did not refuse the pacifier when the mother offered it. The intensity of maternal stimulation ranged from 0 to 6, according to whether the mother was observed actively introducing the pacifier with (1) honey or (2) juices, (3) after the child's refusal, (4) or constantly throughout the observation period; (5) whether the pacifier was always seen near to the child; and (6) whether the mother's first reaction after the infant cried was to give the pacifier. Pearson's correlation coefficient was calculated to assess the magnitude of the association between the two scores, and the result was highly significant (r = .42; P < .001). Breastfeeding prevalence was inversely related to both the maternal stimulation score (r = −.41; P< .001) and the child pacifier acceptance score (r = −.29; P = .01). A composite score based on the 10 items described above reflected the observed intensity of pacifier stimulation and acceptance. The means of these three scores (maternal acceptance, child acceptance, and intensity) showed a linear association with the variables on pacifier use at 1 and 3 months use (coded none, partial, and intense), which had been collected in the epidemiologic study.
The ethnographic component was specifically aimed at assessing whether pacifier use might be an indicator of the presence of other behavioral, social, and cultural factors constraining breastfeeding. The grounded theory approach was used to construct a behavioral and cultural profile characterizing the social circumstances of mothers who promote the pacifier intensely and wean early. Two general areas were addressed: the maternal-infant relationship and mothers' social status. The maternal-infant link will be described according to maternal breastfeeding styles, mothers' approaches to infant crying, and mothers' expectations of infant development.
The 40 mothers who were observed breastfeeding showed marked differences in breastfeeding behavioral styles. Mothers whose infants used pacifiers more intensely exercised a higher degree of behavioral control while breastfeeding, as shown by the following characteristics: (1) lower number of breastfeeds during the total observation period; (2) not allowing the infant to fall asleep on the breast (P = .06); (3) not allowing the infant to leave and take the breast again several times throughout feeding; (4) less infant interaction with the mother during feeding (ie, looking, smiling, and holding the breast); (5) sitting in a fixed, stiff position during feeding; (6) not allowing the infant to decide when the feeding is over; and (7) not leaving breasts exposed after feeding.
The above variables were all significantly associated with the mean pacifier use intensity score (t test,P ≤ .05 unless stated otherwise). These behavioral patterns show that mothers who are embarrassed by breastfeeding, who take a more mechanical approach to feeding, and who do not let their infants control feeding pace and termination also tended to use pacifiers more intensely.
Reaction to Crying
In this analysis, several variables emerged as relevant to women's approaches to infant crying. The following observed variables were related both to more intense pacifier use and to shorter breastfeeding (P > .05): (1) the mother does not pick up the infant immediately when crying starts; (2) the infant was rated by observer to cry often relative to others in the sample; and (3) the infant spends most of the time in the bed or stroller rather than in the mother's arms.
Expectations of Development
At every visit, mothers were asked to report on the infants' growth and development. Based on their answers, two sets of developmental characteristics were created. The first referred to more “mechanical” or “culture-centered” items and included: (1) the infant is larger and heavier; (2) clothes no longer fit; (3) the infant has learned to hold the pacifier and take it well; (4) the infant is smart and likes the bottle better than the breast now; and (5) the infant likes tea and can take the bottle well.
The second set refers to more “naturalistic” or “child-centered” items, including: (1) the umbilical cord fell off early; (2) teeth are coming out; (3) the infant is learning to move around on the bed alone; (4) the infant has learned to smile; (5) the infant makes “conversation” noises; and (6) the infant has more hair now.
Women who defined their infant's development with reference to more mechanical indicators were less likely to breastfeed (χ2test, P = .05), and their infants used pacifiers more intensely (t test, P = .03). On the other hand, reference to naturalistic items was not associated with either breastfeeding or pacifier use.
In short, a number of aspects of the maternal-infant relationship were related to both pacifier use and breastfeeding. The next question was why some women exhibited these characteristics, whereas others did not. Textual and observational analyses showed that women who seemed more nervous and preoccupied with following the “correct” care-giving rules (regarding growth, accustoming, disciplining, and pacifying) also seemed more concerned with their social environment as well as being more sensitive to social criticisms. They actively compared themselves and their situations (in many regards) with other women, often criticizing other mothers and praising their own child-raising abilities.
Ethnographic analyses showed that these women used their maternity to impose their power and knowledge within their social circles and possibly even to acquire a higher social status. Statistical analyses showed that women who were more competitive in this way tended to leave their infants on the bed more often, to react negatively to infant crying, to be more concerned with aspects of infant growth and development (such as weight and length), and to have a more rigid approach to breastfeeding, and their infants used pacifiers more intensely. These women adhered to these care-giving rules more intensely, because they represented avenues through which their social control could be implemented.
Relevant maternal and infant characteristics identified in the ethnographic analysis were correlated with a number of demographic variables. Three of these came out as potentially relevant: type of delivery, sex of the infant, and maternal skin color. Mothers who were white, who underwent cesarean sections, and who had boys were more likely to present a number of the problem behaviors described above. These ethnographic findings led to the hypothesis that the association between pacifiers and breastfeeding might be modified by these traits. This possibility was tested using the epidemiologic data set (Table6). Pacifiers seemed to affect girls more intensely than boys, but this interaction was not significant (P = .15). On the other hand, both maternal color and type of delivery were significant effect modifiers. Pacifier use was only associated with breastfeeding duration among white mothers. It also had a stronger influence on infants delivered by cesarean section. Because cesarean sections and skin color might reflect socioeconomic status, effect modification by family income and maternal education was also investigated, but both analyses were negative.
This population-based study had a high rate of follow-up, covering a representative sample of mothers and infants from a middle-sized city in southern Brazil. Other health, demographic, and economic statistics for the city show that it is fairly typical of this region of the country. The possibility of selection bias is therefore unlikely. The prospective nature of the study also avoided recall bias. Because the study was specifically designed to investigate the association between pacifier use and breastfeeding, it was possible to collect data on relevant variables not included in previous studies. For example, mothers expressing difficulties with breastfeeding at 1 month could be identified and excluded from the analyses, because pacifier use was already associated with such problems, and their inclusion would have biased the results. In fact, when the analyses were repeated without excluding such mothers, the association between pacifier use and breastfeeding termination was even stronger.
The ethnographic component proved to be invaluable and markedly influenced the interpretation of the epidemiologic results. It should be noted, however, that these results were used to answer an essentially epidemiologic question, and that given the scope and length of this article, it was not possible to fully address the implications of the ethnographic findings.
As in other Brazilian studies,6,7 the frequency of pacifier use was high, and the duration of breastfeeding was short. Pacifier use was a dynamic process, with many children changing status between 1 and 3 months of age. As the ethnographic data show, pacifier use is seen as a desirable behavior, and mothers will go to great lengths to make their children take it up.
Mothers who reported believing that pacifiers affected breastfeeding were more likely to breastfeed longer, and their children were less likely to use pacifiers. This might suggest that less knowledgeable mothers would be more likely to wean early, so that providing adequate information would be required. However, the ethnographic analysis suggested that women who did not state awareness when directly asked may indeed have been aware but were unwilling to admit it. Case studies of the women with explicit awareness show that these women were simply more apt to talk openly of the pacifier's weaning effects, because they were comfortable with breastfeeding and were not feeling guilty about their infants' pacifier use.
An original finding was the suggestion that breastfeeding interruption may lead to pacifier initiation (Fig 1). Because bottle feeding requires little active suction, this may be explained by the infant's continued need for oral stimulation and comfort. This suggests that reverse causality may take place. The lower prevalence of breastfeeding among pacifier users in cross-sectional studies may be partly explained by greater initiation of pacifier use after full weaning.
Another interesting finding was the association between pacifier use and a decreased number of breastfeedings. Spending more time sucking a pacifier thus led to a lower frequency of feedings and to reduced prolactin production. A related problem was “nipple confusion.”13 Pacifier users have different sucking patterns than nonusers,14 and these patterns may interfere with the acquisition of oral motor skills necessary for nutritive sucking. An additional issue raised by the ethnographic study was that the duration of each breastfeeding may be shorter in pacifier users, whose mothers are less likely to let the infants decide when to stop feeding. Because the fat content of foremilk is lower than that of hindmilk,15 shorter feedings may leave the infant hungrier and more likely to cry, thereby prompting mothers to introduce nonhuman milk.
The results shown in Table 2, all restricted to infant who were breastfed, deserve further discussion. In addition to showing a lower frequency of breastfeedings, the volume of nonhuman milk ingested by full-time pacifier users was 30% lower than that ingested by nonusers, although this difference was not statistically significant. An interesting group consisted of the 10 nonusers at 1 month who were partially breastfed. Their average nonhuman milk intake was 502 mL, compared with an expected total milk intake of about 600 mL; it might be argued that these infants used the breast as a pacifier, rather than for nutrition. In general terms, it is likely that infants who used pacifiers more did not get as much nonhuman milk because they were sucking on the pacifiers, whereas those who were not full-time pacifier users needed more nonhuman milk. This would also explain why those who got more nonhuman milk would take less tea or water, because their caloric intake was satisfied by more nonhuman milk. These findings contradict the widely held maternal belief that pacifiers are growth promoters. Further research is needed to clarify these associations and to investigate whether, by decreasing overall milk intake, pacifier use may result in slower growth.
In the epidemiologic study, pacifier use, nonhuman milk intake, and the ingestion of teas and water were all independently associated with the duration of breastfeeding. The effect of nonhuman milk was the strongest, whereas that of teas and water was the weakest. There was a clear dose-response effect for pacifier use, with full-time users having shorter durations. Control of confounding variables, although hardly changing the effects of nonhuman milk or teas and water, led to an important reduction in the relative risks associated with pacifier use—from about 4.0 to less than 2.5. Previous epidemiologic studies6,7 showed slightly higher relative risks, but the difference may have been attributable to tighter control of confounding variables in our investigation as well as the exclusion of children who, while still on the breast at 1 month of age, had breastfeeding difficulties.
Classic epidemiologic methods are characterized by the study of large samples using questionnaires or laboratory measurements. Confounding variables are usually assessed through standardized questions. This has obvious limitations, because questionnaire answers may fail to elicit true knowledge or behaviors, as shown above in relation to the awareness of harmful pacifier effects. Despite these limitations, adjustment for confounders resulted in a substantial reduction in the magnitude of the effect of pacifiers on weaning. This raises the possibilities of residual confounding and that tighter adjustment might further reduce the relative risk.
On the other hand, ethnographic methods added a new depth to the epidemiologic results. Mothers who actively stimulated pacifier use had a number of behavioral and sociocultural traits that were also associated with shorter breastfeeding. These included a rigid breastfeeding style, strong expectations about objective aspects of infant growth and development, and anxious reactions to the infants' crying. These behaviors, in turn, were more often observed among mothers concerned about what others think of their mothering abilities, among those giving strong credence to strict care-giving rules, and among those who seemed more competitive with other mothers, indicating a lack of self-confidence and strong concern about their social roles. Although not explicitly recognizing that pacifiers could be used to shorten breastfeeding, many mothers who are less comfortable breastfeeding resort to pacifier use (among other methods) to reduce the discomfort they feel when breastfeeding.
It might be argued that that the maternal behaviors blamed for increased pacifier use may be consequences rather than causes of pacifier use; mothers would introduce pacifiers first and then acquire the habits of not allowing the infants to fall asleep on the breast and of feeding less frequently. The fact remains, however, that the mothers had particular motives for using pacifiers in the first place. These motives, in turn, were shown to be related to maternal characteristics (eg, competition, concern with rules, and relationship with the body) that are deeply ingrained and unlikely to have been affected by whether the infant was an intense pacifier user.
The ethnographic data set does not allow formal control of confounding variables, because of the small number of breastfeeding mothers. However, these results strongly suggest that in addition to the confounders measured in the epidemiologic study, several others—probably more important ones—cannot be assessed through standard questionnaires but might further account for the observed association. If one accepts that the operational difference between confounding and (self) selection bias is that the former can be measured and controlled in the statistical analysis, whereas the latter cannot,16 the ethnographic analysis revealed an important degree of self-selection toward pacifier use and early weaning that the epidemiologic study could not detect.
Both the ethnographic and epidemiologic data suggest that some mothers and infants seem to be less affected by pacifier use. By possibly being more self-confident and less sensitive to social pressures, by being less concerned with rapid growth and mechanical development of the child, and by handling their infants in a more flexible and relaxed way, these mothers continue to breastfeed despite their infants' pacifier use. These include nonwhite mothers, those who have given birth vaginally, and, to a lesser extent, mothers of infant girls. Previous studies in our population described the shorter duration of breastfeeding among infants delivered by cesarean sections,17 as well as differentials according to skin color and gender.18 It was remarkable that the three variables pointed out by the ethnographic study as potential effect modifiers were confirmed in the epidemiologic analyses (or almost so, for the infants' sexs), because a number of other variables were tested and found negative.
The issue is not that pacifiers do not contribute to stopping breastfeeding—in fact, the very high relative risks in the susceptible groups (Table 6) suggest that they do so—but whether these mothers would not stop breastfeeding anyway and are just using pacifiers as weaning tools. Our findings suggest that breastfeeding promotion campaigns aimed specifically at reducing pacifier use will fail. Such campaigns should also provide support for women facing the challenges of nursing and address their anxieties. Further analyses from the ethnographic study on these issues will be published separately.
Previous epidemiologic reports6-8 on the association between pacifier use and breastfeeding raised the possibility that pacifier use might be a marker rather than a determinant of breastfeeding and, more recently, also of adult intelligence.9 The present results confirm that self-selection plays a role in this association, but the full picture is more complex. Pacifiers seem to contribute to earlier weaning among a group of women uncomfortable with breastfeeding but do not seem to affect breastfeeding duration among self-confident mothers. Pacifiers, therefore, might be seen as a contributing rather than a sufficient cause of early breastfeeding termination. Regardless of the discussion on causality, intense pacifier use is still an excellent marker for identifying mothers and infants facing breastfeeding difficulties toward whom supportive actions should be aimed. Although the ethnographic results refer to maternal behaviors that are mostly specific to the society under study, it is plausible that mothers from other societies who stimulate pacifier use may also have similar behaviors that curtail breastfeeding. Research in other settings, however, is needed to confirm our findings.
This research was supported by the Division of Control of Diarrhea and Acute Respiratory Diseases of the World Health Organization, the European Economic Commission, and the Fundação de Amparo a Pesquisa do Rio Grande do Sul, Brazil.
- Received May 2, 1996.
- Accepted October 8, 1996.
Reprint requests to (C.G.V.) Departamento de Medicina Social, Universidade Federal de Pelotas, CP 464, Pelotas RS 96001-970, Brazil.
- ↵World Health Organization. Protecting, Promoting and Supporting Breastfeeding: the Special Role of Maternity Services. A Joint WHO/UNICEF Statement. Geneva, Switzerland: World Health Organization; 1989
- ↵UNICEF/World Health Organization. Innocenti Declaration on the Protection, Promotion and Support of Breastfeeding. Florence, Italy: UNICEF/World Health Organization; 1990
- Tomasi E,
- Victora CG,
- Olinto MTA
- Barros FC,
- Victora CG,
- Semer TC,
- Tonioli Filho S, Tomasi E, Weiderpass E
- ↵Ford RPK, Mitchell EA, Scragg R, et al. Factors adversely associated with breast feeding in New Zealand. J Paediatr Child Health. In press
- Maldonado G,
- Greenland S
- ↵Bernard HR. Research Methods in Anthropology: Qualitative and Quantitative Approaches. London, United Kingdom: Sage Publications; 1994
- ↵Glaser B, Strauss A. The Discovery of Grounded Theory. Chicago, IL: Aldine; 1967
- ↵Riordan J. A Practical Guide to Breastfeeding. Boston, MA: Jones and Bartlett; 1991:47
- Wolff PH
- ↵Akre J, ed. Infant feeding: the physiological basis. Bull WHO. 1989;67(suppl):26(chapter 2)
- ↵Rothman KE. Modern Epidemiology. Boston, MA: Little, Brown & Co; 1986;.
- Victora CG,
- Huttly SRA,
- Barros FC,
- Vaughan JP
- Copyright © 1997 American Academy of Pediatrics