McAteer JP, Ledbetter DJ, Goldin AB. Role of Bottle Feeding in the Etiology of Hypertrophic Pyloric Stenosis. JAMA Pediatr. 2013;167(12):1143-1149. doi:10.1001/jamapediatrics.2013.2857
Bottle feeding has been implicated in the etiology of hypertrophic pyloric stenosis (HPS). Further data are needed to define the nature of this relationship and the clinical variables that influence it.
To determine if bottle feeding after birth is associated with the development of HPS in infants. We hypothesized that bottle feeding is associated with an increased risk of HPS and that this risk is modified by other risk factors.
Design, Setting, and Participants
Population-based case-control study of births from January 1, 2003, to December 31, 2009, using Washington State birth certificates linked to hospital discharge data. Cases included all singleton infants born within the study period and subsequently admitted with both a diagnostic code for HPS and a procedure code for pyloromyotomy (n = 714). Controls were randomly chosen among singleton infants who did not develop HPS and were frequency matched to cases by birth year.
Feeding status (breast vs bottle) was coded on the birth certificate as the type of feeding the infant was receiving at birth discharge.
Main Outcome and Measure
Diagnosis of HPS.
Hypertrophic pyloric stenosis incidence decreased over time, from 14 per 10 000 births in 2003 to 9 per 10 000 in 2009. Simultaneously, breastfeeding prevalence increased from 80% in 2003 to 94% in 2009. Compared with controls, cases were more likely to be bottle feeding after birth (19.5% vs 9.1%). After adjustment, bottle feeding was associated with an increased risk of HPS (odds ratio [OR], 2.31; 95% CI, 1.81-2.95). This association did not differ according to sex or maternal smoking status but was significantly modified by maternal age (<20 years OR, 0.98; 95% CI, 0.51-1.88; ≥35 years OR, 6.07; 95% CI, 2.81-13.10) and parity (nulliparous OR, 1.60; 95% CI, 1.07-2.38; multiparous OR, 3.42; 95% CI, 2.23-5.24).
Conclusions and Relevance
Bottle feeding is associated with an increased risk of HPS, and this effect seems to be most important in older and multiparous women. These data suggest that bottle feeding may play a role in HPS etiology, and further investigations may help to elucidate the mechanisms underlying the observed effect modification by age and parity.
Hypertrophic pyloric stenosis (HPS) is an idiopathic thickening of the circular smooth muscle layer of the pylorus typically occurring within the first several weeks of life.1 The resultant gastric outlet obstruction is treated surgically by pyloromyotomy.2 With an incidence of about 2 cases per 1000 births, the condition is a common indication for surgery in early infancy.1,3Quiz Ref IDDespite the frequency with which HPS occurs, its etiology remains unknown.4 It is known to be more common among firstborns, males, and white individuals, and although most cases are sporadic, the disease tends to cluster in families.2,5 The biological factors driving these patterns, however, remain elusive.
Because of its almost exclusive occurrence within the first 2 months of life, early environmental exposures have been a natural focus of inquiry. Infant feeding practice is one such exposure that may play a role. Early descriptive ecological studies in Europe suggested that increased breastfeeding rates may be related to increased HPS incidence.6- 8 More recent studies in Canada, Italy, Nigeria, and Taiwan have suggested the opposite—that increased breastfeeding rates may be related to the decreasing HPS incidence observed in those populations.9- 12 Another recent study found that any exposure to bottle feeding was associated with HPS in a national Danish cohort.13 Taken together, these studies have sparked interest in the potential role of breastfeeding vs bottle feeding in the etiology of HPS.
The existing literature draws from a diverse international experience and generally consists of small case numbers, which has impaired the ability to perform subgroup analyses. Given that other HPS risk factors have been noted to be modified by known demographic factors, it is important that larger case numbers be used to assess such effects.14Quiz Ref IDTo assess the effect of bottle feeding vs breastfeeding on the subsequent risk of HPS, as well as to assess the variability in that effect across strata of other known risk factors, we conducted a population-based case-control study using birth records linked to hospital discharge data in Washington State.
The study was approved by the University of Washington institutional review board (approval 43226). We performed a population-based case-control study using state birth records linked to the Washington State Comprehensive Hospital Abstract Reporting System, a statewide inpatient hospital discharge database that provides deidentified patient data regarding age, sex, payer status, hospital diagnoses, procedures performed, length of stay, and discharge disposition. The Washington State Department of Health created and maintains this linked database. We hypothesized that the odds of HPS would be increased in children who were not breastfed at birth discharge, even after controlling for other exposures and demographic factors. Further, we hypothesized that this association would vary according to other known risk factors described later.
Using hospital discharge records from January 1, 2003, to December 31, 2009, HPS cases were identified as any infant (<6 months) with both a diagnostic code for HPS (International Classification of Diseases, Ninth Revision code 750.5) and a procedure code for pyloromyotomy (43.3). We required both diagnostic and procedure codes to minimize case misclassification. Patients were restricted to singleton infants. Additionally, infants who were diagnosed and treated for HPS during the birth hospitalization were excluded, since exposure status (feeding status) was recorded during the birth hospitalization and would not have clearly preceded the development of HPS in these cases. To assure that nonbreastfeeding at birth discharge was representative of bottle feeding rather than requirement for other nutritional supplementation (eg, tube feedings or hyperalimentation), we also excluded all patients with diagnoses of major congenital gastrointestinal anomalies (esophageal atresia/tracheoesophageal fistula, intestinal atresia, anorectal atresia, Hirschsprung disease, congenital diaphragmatic hernia, gastroschisis, omphalocele, necrotizing enterocolitis, and neonatal intestinal perforation). This is similar to the approach taken in previous studies.9
Quiz Ref IDControls were randomly chosen in a 10:1 ratio relative to cases from among all singleton infants without a diagnosis of pyloric stenosis born during the same time frame. Case-control ratio was determined a priori based on prestudy sample size calculations, which were powered for a minimal detectable odds ratio (OR) of 1.30 for bottle feeding. Controls were frequency matched to cases by year of birth to account for changes in HPS incidence over time.
The exposure of interest was bottle feeding, defined as any infant not being exclusively breastfed during the birth admission. Feeding status was recorded on the birth record as the type of feeding that the infant received after birth during the birth admission. The variable is coded as breastfed or nonbreastfed on the birth record. Washington State began recording feeding status after birth beginning in January 2003. To assure minimal misclassification of the exposure, in addition to the exclusions mentioned earlier, we ran the regressions both with and without infants younger than 37 weeks’ gestational age, since nonbreastfeeding in these infants might be more likely to represent hyperalimentation or tube feeding rather than bottle feeding. Since no substantial difference in results was noted, we included younger gestational ages in the final analysis.
Potential confounders identified a priori were those clinical factors previously found to be risk factors for HPS or that might be expected to otherwise confound the relationship between feeding practice and HPS.14- 16 These factors were collected from the birth record and included infant sex, maternal age (<20 years, 20-34 years, and ≥35 years), maternal race (white, black, Native American, Hispanic, and Asian/Pacific Islander), maternal parity, maternal body mass index (calculated as weight in kilograms divided by height in meters squared) (<19, 19-25, >25-30, and >30), gestational age (<37 weeks, 37-41 weeks, and ≥42 weeks), birth weight (<2500 g, 2500-<3000 g, 3000-<3500 g, 3500-<4000 g, 4000-<4500 g, and ≥4500 g), and maternal tobacco smoking during pregnancy (yes/no). Maternal parity was categorized as nulliparous (no previous births by mother), primiparous (1 previous birth), or multiparous (≥2 previous births). Additionally, given the strong association of HPS with infant sex and birth order, as well as the fact that recent studies have found certain risk factors to be modified by these demographic characteristics, we elected a priori to assess for effect modification of the feeding practice–HPS association by infant sex, maternal age, maternal parity, and maternal smoking status.
Descriptive statistics were used to compare characteristics between cases and controls. Information on total births per year, overall breastfeeding prevalence, and smoking prevalence among pregnant women was obtained from vital statistics data from the Washington State Department of Health. Hypertrophic pyloric stenosis incidence was calculated as number of incident cases of HPS in a given year divided by the total number of singleton births that year.
A multivariate logistic regression model was used to quantify the association between bottle feeding at birth discharge and subsequent admission for HPS. Risk estimates were adjusted for the covariates listed earlier. A P value <.05 was considered statistically significant. To assess effect modification, 4 more logistic regression models were created with interaction terms between the exposure of interest (feeding status) and each of the 4 potential effect modifiers listed earlier. Likelihood ratio tests were used to test the overall significance of the interaction model compared with the original model with no interaction terms. Linear combination tests were used to compare the differences in risk estimates across strata of effect modifiers with more than 2 categories (maternal age and maternal parity). Effect modification was considered to be present if the P value for an interaction term was <.05. To obtain values for the ORs of the feeding practice–HPS association for each stratum of the effect modifiers, separate logistic regression models were run for each stratum (eg, 3 models for maternal age and 2 models for infant sex).
Statistical analysis was performed using Stata version 12 (StataCorp).
We identified 825 infants with a diagnostic code for HPS over the study period. One hundred three patients did not have a procedure code for pyloromyotomy, 6 were treated during the birth admission, and 2 had concomitant gastrointestinal anomalies. After these exclusions, 714 cases were included in the analysis. Gestational age, birth weight, and maternal body mass index were similar for both cases and controls. Smoking during pregnancy was nearly twice as common in cases as in controls. More than twice as many cases as controls were bottle fed (19.5% vs 9.1%) (Table 1). Relative to controls, cases were more likely to be male, born to younger mothers, and firstborn. Cases were also more likely to be born to white mothers and less likely to be born to mothers of Asian descent.
The overall incidence of HPS in Washington State over the study period was 12 cases per 10 000 births. This incidence decreased from 14 per 10 000 in 2003 to 9 per 10 000 in 2009 (Figure, A). Over the same period, the prevalence of breastfeeding at birth discharge increased from 80% in 2003 to 94% in 2009. The prevalence of maternal smoking remained stable between 10% and 11% (Figure, B). Similarly, the proportion of total births with other known risk factors (male sex, younger maternal age, and nulliparity) did not change appreciably over the 7-year study period (data not shown).
Quiz Ref IDIn the multivariate analysis, bottle feeding at birth discharge was found to be independently associated with HPS (OR, 2.31; 95% CI, 1.81-2.95), even after adjusting for sex, maternal age, maternal race, maternal body mass index, birth order, gestational age, birth weight, and maternal smoking (Table 2). In assessing for potential effect modification, neither infant sex (male OR, 2.20; female OR, 2.60; P = .26 for interaction) nor maternal smoking status (smoker OR, 2.15; nonsmoker OR, 2.40; P = .73 for interaction) were found to modify the bottle feeding effect. Quiz Ref IDMaternal age, however, was noted to significantly modify the bottle feeding effect. While there was no significant association noted in mothers younger than 20 years (OR, 0.98; 95% CI, 0.51-1.88), the association became stronger for increasing age categories (ages 20-34 years: OR, 2.58; 95% CI, 1.93-3.44, and 35 years and older: OR, 6.07; 95% CI, 2.81-13.10). Modest modification of the bottle feeding–HPS association was also noted according to birth order, with an OR of 1.60 for infants of nulliparous women (95% CI, 1.07-2.38) and an OR of 3.42 for those born to multiparous women (95% CI, 2.23-5.24) (Table 3).
Our understanding of HPS has advanced greatly since the condition was first described by Hirschsprung in 1888, and diagnostic and treatment paradigms are well established.2 In spite of this, the cause of this disease remains an enigma. It is generally accepted that the etiology is multifactorial, but the relative contributions of genetic and environmental factors, and how those factors interact to cause pyloric muscle hypertrophy within a narrow window in early infancy, are unknown.4 Changes in disease incidence in certain countries have prompted studies to pinpoint the underlying environmental impetus. Epidemiologic patterns in Scotland and Germany suggested a potential role for sleeping position, but these data have not been replicated.17,18 Breastfeeding became a target of interest in the 1980s when several ecological studies suggested that increasing breastfeeding rates might be related to observed increases in HPS incidence.6- 8 Subsequent studies showed more clearly that increased breastfeeding rates correlated with a decreased incidence of HPS.9- 13 In this case-control study, we have analyzed the largest group of patients with HPS with data on infant feeding practice yet studied, to our knowledge, and have shown not only that bottle feeding is associated with subsequent development of HPS but that this risk varies according to other well-known risk factors.
Given the pathophysiology of HPS, early feeding practice is a natural exposure to consider in its etiology. As noted by previous authors, our results may either indicate a protective effect conferred by breast milk or an increase in risk due to bottle feeding.13 Other studies have suggested possible mechanisms by which bottle feeding of formula might confer an increased risk of HPS. Infant formula is noted to have higher osmolarity than breast milk and decreases gastric emptying.19 Breast milk contains high levels of endogenous vasoactive intestinal peptide, which may help to mediate pyloric relaxation and facilitate gastric emptying, while infants fed by bottle have higher serum gastrin levels, which can be associated with pylorospasm.10,20 Other authors have suggested that the larger volume of feeds taken by formula-fed infants may account for the effect and that the HPS sex ratio may be partially explained by the fact that boys consume larger volumes of formula, thus magnifying the effect in males.13 The lack of effect modification by sex in our study, however, argues against this latter hypothesis.
While formula osmolarity and gastrointestinal protein content provide possible mechanisms by which feeds may impact the development of HPS, these theories do not explain the strong effect modification by maternal age noted in this study. Overall, while children of younger and nulliparous women are at increased risk of HPS, this study shows that the association between bottle feeding and HPS is more pronounced in older and multiparous mothers. If one assumes bottle feeding to be a causal factor in the etiology of HPS, then according to population attributable risk calculations in our study, about 11% of HPS cases overall are attributable to bottle feeding. Among mothers 35 years or older, that figure is 32%, compared with essentially 0% in mothers younger than 20 years. This, along with the greater effect in multiparous women, suggests a possible hormonal role for the effect of feeding practice. Sex hormones, especially testosterone, have previously garnered interest in HPS research, but recent studies suggest that maternal and fetal androgen levels have little impact on disease incidence.21,22 The modification of effect by age and parity in our study suggests that estrogen may play an important role. Estrogen levels during pregnancy vary according to maternal age, with older mothers having higher serum estradiol levels.23 Further, estrogen has been shown to slow gastric emptying, in part through down-regulation of nitric oxide–mediated intestinal smooth muscle relaxation.24,25 Decreased nitric oxide activity itself has been previously implicated in the pathogenesis of HPS.26,27 Bottle feeding may further add to the effects of maternal estrogens via other mechanisms, such as the intrinsic estrogenic activity present in isoflavones in soy-based formulas or the estrogen-modulating effects of bisphenol A in baby bottles.28- 30 It may be that higher estrogen levels in utero in older and multiparous mothers “prime” the fetus’s pyloric muscle via a decrease in nitric oxide–mediated relaxation, making it more susceptible to the effects of bottle feeding after birth.
The potential mechanisms outlined earlier are only speculative, and further investigations are needed to test these hypotheses. Although the effects of bottle feeding are most prominent among older and multiparous women, the greatest baseline risk for HPS is in younger and nulliparous women. It may be that bottle feeding has similar effects on the infants of all mothers but that the higher baseline risk in younger, primiparous women simply masks this effect, thus accounting for the observed modification of effect by older age and increasing parity. As such, further research is needed to address the mechanisms underlying these associations to better define the effects of bottle feeding on the pylorus in early infancy.
This study has several limitations. First, the study design is observational and associations should be interpreted with caution. The administrative data source is subject to potential misclassification, and there may be confounding factors that are not adequately controlled for in the study. Because we only knew the feeding status during the birth admission, some infants who were breastfed initially could have switched to bottle feeding (or vice versa) at home prior to the development of HPS. Since prior studies have suggested that any exposure to bottle feeding increases the risk of HPS, however, we expect that this limitation would bias our results toward the null.13 Additionally, compared with other studies that used retrospective surveys to determine bottle feeding status, our definition of exposure, determined during the birth admission, avoids the potential for recall bias. Another potential limitation is that babies bottle fed expressed breast milk in our data were coded as bottle fed, so bottle feeding does not purely capture formula feeding alone.
This study has a number of strengths including a large number of cases enabling subgroup analysis, a population-based sample, and reliable coding of state birth records. Our results present a more complete picture of the influence of feeding practice on the development of HPS than has previously been available, showing not only a clear association between bottle feeding and HPS, but an association that varies according to certain maternal factors. Our data also suggest that in Washington State the increase in breastfeeding may be a driver of the decreasing incidence of HPS. If the association is indeed causal, these data certainly further emphasize the benefits of breastfeeding, especially in certain patient groups, but more importantly point to further potential etiologies in the genesis of HPS. Further studies are warranted to validate these findings and to look more closely at the speculative mechanisms, including possible hormonal effects, underlying the bottle feeding–HPS association.
Corresponding Author: Jarod P. McAteer, MD, MPH, Division of Pediatric General and Thoracic Surgery, Seattle Children’s Hospital, 4800 Sand Point Way NE, Seattle, WA 98105 (email@example.com).
Accepted for Publication: April 27, 2013.
Published Online: October 21, 2013. doi:10.1001/jamapediatrics.2013.2857.
Author Contributions: Dr McAteer had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: All authors.
Acquisition of data: McAteer, Goldin.
Analysis and interpretation of data: All authors.
Drafting of the manuscript: McAteer, Goldin.
Critical revision of the manuscript for important intellectual content: Ledbetter, Goldin.
Statistical analysis: McAteer, Goldin.
Administrative, technical, or material support: Goldin.
Study supervision: Ledbetter, Goldin.
Conflict of Interest Disclosures: None reported.