664711

research-article2017
JHLXXX10.1177/0890334416664711Journal of Human LactationChen et al.

Original Research
Journal of Human Lactation

Association Between Infant Feeding

111
The Author(s) 2017
Reprints and permissions:
Modes and Gastroesophageal Reflux: sagepub.com/journalsPermissions.nav
DOI: 10.1177/0890334416664711

A Repeated Measurement jhl.sagepub.com

Analysis of the Infant Feeding

Practices Study II

Pei-Lin Chen, MPH1, Nels Soto-Ramrez, PhD, MS, MPH2,

Hongmei Zhang, PhD, MS1, and Wilfried Karmaus, MD, MPH1

Abstract
Background: Gastroesophageal reflux in neonates is frequently reported by parents, potentially motivating changes in infant
feeding mode and/or addition of solid food.
Objective: The authors prospectively analyzed associations between repeated measurement of feeding modes and reflux
in infancy.
Methods: The Infant Feeding Practices Study II, conducted between 2005 and 2007 (2,841 infants), provides data on reflux
and feeding modes at nine time points from months 1 to 12. Feeding modes were defined based on direct breastfeeding,
feeding of bottled human milk, formula feeding, their combinations, and use of solid food. Repeated measurements were
investigated using 1-month delayed models to estimate risk ratios (RRs) and their 95% confidence intervals (CIs). Risk ratios
of different feeding modes were estimated for reflux; addressing a reverse association, RRs for feeding mode were estimated
as responses to prior reflux.
Results: Compared to direct breastfeeding, combinations with formula feeding showed a statistically significant risk for reflux
(bottled human milk plus formula feeding: RR = 2.19, 95% CI [1.11, 4.33]; formula feeding: RR = 1.95, 95% CI [1.39, 2.74];
and mixed breastfeeding plus formula feeding: RR = 1.59, 95% CI [1.40, 2.42]). Addition of solid food was not protective (RR
= 1.21, 95% CI [0.86, 1.70]). Analyses of reverse association (reflux
feeding) showed fewer breastfed infants among those
with reflux in the prior month.
Conclusion: Any combination of infant feeding with formula seems to be a risk for reflux. Although breastfeeding was
protective, mothers with a child with reflux were more likely to wean their child.

Keywords
breastfeeding, feeding mode, formula, gastroesophageal reflux, solid food

Background discuss with their pediatricians during the routine 6-month

Gastroesophageal reflux (GER) is considered to occur several
times per day in healthy infants and is frequently reported by
parents as health concerns in neonates (Lightdale, Gremse, &
Section on Gastroenterology, Hepatology, & Nutrition, 2013).
Gastroesophageal reflux is defined as the passage of gastric
contents into the esophagus with or without regurgitation and
vomiting (Czinn & Blanchard, 2013; Hegar & Vandenplas,
2013). Gastroesophageal reflux disease (GERD) is consid-
ered when, in addition to GER, the infant develops complica-
tions such as respiratory problems, esophagitis, and failure to
thrive and needs intensive care (Vandenplas etal., 2009).
Gastroesophageal reflux and feeding difficulties and GER-
related symptoms are among the main concerns that parents
infant visit (Lightdale etal., 2013). However, 95% of these
symptoms are considered to spontaneously resolve by 12
months (Czinn & Blanchard, 2013; Jadcherla etal., 2012). It
1
Division of Epidemiology, Biostatistics, and Environmental Health, School
of Public Health, University of Memphis, Memphis, TN, USA
2
College of Social Work, University of South Carolina, Columbia, SC, USA
Date submitted: April 24, 2015; Date accepted: July 27, 2016.
Corresponding Author:
Wilfried Karmaus, MD, MPH, Division of Epidemiology, Biostatistics, and
Environmental Health, School of Public Health, University of Memphis,
Memphis, TN 38152, USA.
Email: karmaus1@memphis.edu
2 Journal of Human Lactation
has been reported that the transition to solid food may be
associated with such a spontaneous resolution; however, this
is also the time when infants learn how to sit up (around 6
months of age) (Hadders-Algra, 2013). Thus, gravitational
influence may contribute to a decline in regurgitation.
Studies have demonstrated that partially breastfed babies
may suffer from more frequent regurgitation than exclusively
breastfed babies, especially between 2 and 6 months of age,
and infants who were exclusively breastfed were less likely to
experience regurgitation than those who were partially breast-
fed (Heacock, Jeffery, Baker, & Page, 1992; Hegar etal., 2009).
In addition, infants who experienced regurgitation were char-
acterized by reduced success in breastfeeding compared to
babies who did not (Yaln & Kukonmaz, 2011). These find-
ings support the recommendation of the American Academy of
Pediatrics (2012), which emphasizes that human milk is an
optimal source of nutrition through the first year of life and
recommends exclusive breastfeeding for at least 6 months.
There are also concerns whether infants with reflux gain
sufficient weight (Hegar, Satari, Sjarif, & Vandenplas, 2013).
It has been suggested that the addition of thickening agents
or solid food may be necessary to improve the weight status
of these infants (Lightdale etal., 2013; Nevo, Rubin, Tamir,
Levine, & Shaoul, 2007), for instance, through the addition
of rice cereal and bean gum to formula. Other recommenda-
tions for parental management of GER include parental edu-
cation, positional treatment, feeding techniques, and proper
diet, drugs, and surgery, in this order (Hegar etal., 2013;
Horvath, Dziechciarz, & Szajewska, 2008; Jadcherla etal.,
2012; Khoshoo, Ross, Brown, & Edell, 2000).
Focusing on GER, we hypothesize that, compared to
direct breastfeeding (i.e., at the breast), other modes of infant
feeding may pose a risk for developing GER during the first
12 months of life (mode of feeding is a risk factor for GER).
It has been estimated that in the first 6 months, about 50% of
infants need a change in feeding mode because of regurgitation
or vomiting (Nevo etal., 2007). However, there is a lack of
knowledge on whether reflux motivates parents to switch to for-
mula or to use solid food as a thickening agent to prevent reflux.
We hypothesize that GER in 1 month will be associated with
a different mode of feeding in the subsequent investigation
(reverse association: GER as a risk factor for mode of feeding).
The Infant Feeding Practices Study II (IFPS II) collected
information on modes of feeding at nine time points during
the first 12 months of life, including the presence of reflux.
By examining the association between different feeding
modes and reflux over time, the repeated measurement data
will first provide information about the temporal sequence of
feeding modes and reflux. Second, it will improve our under-
standing of whether addition of solid food will reduce the
prevalence of reflux. For the observation period of 12
months, we considered two delayed models. First, in a
1-month-delayed model for reflux, we linked information on
infant feeding inquired in the previous month with the occur-
rence of reflux in the next month. Second, in a delayed model
considering reverse associations, we associated reflux in the
Key Messages
There is a dispute on whether feeding of solid food is
protective against gastroesophageal reflux. In addi-
tion, little is known on whether specific feeding
modes pose a higher risk of gastroesophageal reflux.
Using the Infant Feeding Practices Study II, we
applied repeated measurement analyses and con-
sidered the time order of feeding and reflux. Our
results suggest that any formula feeding poses a
risk for reflux in the next month. Though protec-
tive, fewer children with reflux were breastfed in
the following month.
Adding solid food into the diet seems not to be
beneficial for infants with reflux but feeding
directly at the breast seems to be protective.
prior month with subsequent mode of feeding and the addi-
tional use of solid food.
Methods
The IFPS II data were collected in the United States by the
Food and Drug Administration in collaboration with the
Centers for Disease Control and Prevention and co-funded by
the Office on Womens Health of the Department of Health
and Human Services, the National Institute of Child Health
and Human Development, the Office of Dietary Supplements,
the National Cancer Institute, and the Maternal and Child
Health Bureau of the Health Resources and Services
Administration (Fein etal., 2008; Grummer-Strawn, Scanlon,
& Fein, 2008; Soto-Ramrez etal., 2013). Analysis of the
IFPS II data was exempt from institutional review board
approval by the University of Memphis.
Participants
The sampling frame for the IFPS II was drawn from a nation-
ally distributed consumer opinion panel of 500,000 house-
holds. About 4,900 pregnant women were prenatally enrolled
from May 2005 to June 2007. The criteria for enrollment
were as follows: (1) mother was at least 18 years of age and
in the third trimester of pregnancy; (2) mother was moder-
ately literate and English proficient; (3) mother had a stable
address for at least 11 months; (4) neither mother nor infant
had a medical condition at birth that would affect feeding; (5)
infant was born full-term or near-term (> 35 weeks gesta-
tion); (6) infant weighed at least 5 pounds (2.26 kg); and (7)
infant did not attend intensive care for more than 3 days
(Fein etal., 2008; Soto-Ramrez etal., 2013).
A total 3,003 of 4,900 women filled out the neonatal ques-
tionnaire (the first month questionnaire). Of these 3,003
mothers, 2,988 mothers provided information on feeding
modes for at least one to nine questionnaires that were
Chen et al. 3
Figure 1. Repeated model analyses with delayed effects for infants who have reflux symptoms in the following month.
administered during the study (Fein etal., 2008; Soto-Ramrez
etal., 2013). Of the latter, 2,841 (91.5%) answered questions
about the infants reflux in the questionnaire administered in
months 2, 3, 4, 5, 6, 7, 9, 10, and 12.
Questionnaires
All mailed questionnaires were completed by the mother. In
addition, a short telephone interview was conducted at the
time of the infants birth and a neonatal questionnaire was
sent when the infant was around 1 month old. Nine question-
naires about infant feeding, health care, and related issues
were mailed to the mother, one per month until the infant was
7 months old and then twice about every 7 weeks until month
12 (Fein etal., 2008). All time-dependent variables (reflux,
solid food, and modes of feeding) were corrected for the
childs age when the questionnaire was returned.
Exposures
Infant feeding information was collected nine times during
infancy using postnatal questionnaires and ascertaining feed-
ing practices in the past 7 days. Based on the questionnaire
information, we classified feeding modes into seven catego-
ries (see appendix): (1) direct breastfeeding (feeding from
the breast); (2) bottled human milk feeding (mother expressed
or pumped milk); (3) formula feeding; (4) mixed breastfeed-
ing (combination of 1 and 2); (5) direct breastfeeding and
formula feeding (combination of 1 and 3); (6) feeding of
bottled human milk and formula (combination of 2 and 3);
and (7) mixed breastfeeding and formula (combination of 1,
2, and 3). In addition to this, we considered feeding of solid
food: How often is your baby fed which food listed below?:
baby cereal, other cereals and starches, vegetables, French
fries, fruit, meat, fish or shellfish, eggs, peanut food items,
dairy food items, and soy food items.
Outcome Variable
Reflux information was collected at nine time points and ques-
tions asked about experience in the past 2 weeks. The IFPS II
included a question to address health outcomes among infancy:
Which of the following problems did your baby have during
the past 12 months? A list of 14 health outcomes was provided.
Reflux is 1 of 14 health items and it was assumed to be present
when the mother answered yes to the reflux question.
Covariates
Studies show that factors such as maternal education level,
gestational age, maternal smoking status, maternal body mass
index, and birth weight of the infant may influence maternal
judgment on how to feed their offspring. For instance, studies
have demonstrated that mothers with higher education are
less likely to feed formula (Wojcicki etal., 2010). To estimate
the adjusted association between modes of feeding and reflux,
we considered the following covariates as potential con-
founders: sociodemographic factors (i.e., maternal race,
maternal age, and education), maternal factors (maternal pre-
pregnancy body mass index, and smoking status), infants
birth weight, season of birth, and solid food.
Information on maternal race was based on the initial tele-
phone interview and included race/ethnicity: Caucasian,
African American, Latina/Hispanic, or other. Infant birth
weight was collected in the Birth Screener. Other questions
in the prenatal questionnaire addressed employment status,
breastfeeding attitudes and experiences, infant feeding plans,
sources of information regarding diet and infant feeding, and
participation in the Special Supplemental Nutrition Program
for Women, Infants, and Children (WIC).
Statistical Analysis
SAS 9.4 (SAS Institute Inc., Cary, North Carolina, USA) was
used for our analysis. Sample characteristics focused on all
infants with information on reflux (n = 2,841). Descriptive
analyses were performed for all infants with information on
reflux to calculate the prevalence of reflux and proportion of
children who received solid food at different time points
throughout the first year of the infants life. We considered
delayed models from months 1 to 12 (see Figure 1): first, dif-
ferent feeding modes were linked to reflux in the subsequent
month. Second, in the reverse association model, reflux was
used as a predictor for feeding mode in the next month, allow-
ing a delayed effect (see Figure 1). To examine repeated modes
of feeding on repeated reflux events, we used a generalized
4 Journal of Human Lactation
linear model estimate using generalized estimating equations
(GEE) (PROC GENMOD with repeated statement and log
link function) to estimate risk ratios (Skov, Deddens, Petersen,
& Endahl, 1998; Soto-Ramrez etal., 2013; Zhang & Yu,
1998). We adjusted for within-participant effects using the
regular maximum likelihood method and with a first-order
autoregressive working correlation structure. The first-order
autoregressive covariance matrix improved the fit based on
the evaluation of the quasi-likelihood under the independence
model criterion (QIC) (Cui, 2007). All potential confounders
were simultaneously entered into the models. A backward
elimination process was used to retain confounders in the final
model. Covariates that changed the effect of the main associa-
tion by 10% or more when omitted from the model were con-
sidered as confounders.
To address the potential reverse association between reflux
and various modes of infant feeding, we additionally used
multinomial logistic regression categorical modeling (PROC
CATMOD). The multinomial logistic regression model han-
dles multiple dependent feeding modesdirect breastfeeding,
feeding bottled human milk, formula feeding, mixed breast-
feeding, direct breastfeeding plus formula feeding, bottled
human milk plus formula feeding, and mixed breastfeeding
plus formula feedingand estimates odds ratios (ORs) of
reflux related to these various responses. The CATMOD pro-
cedure applies maximum likelihood estimations of general-
ized logits defined as the logarithm of the frequency ratios of
various different categorical and mutually exclusive outcomes
(Agresti & Liu, 2001). Again, to understand the time order of
events, a 1-month delayed model was tested to determine
whether reflux was associated with a change in infant feeding
mode in the following month (reverse association). In all
tables, we provide risk ratios (RRs) or ORs and their 95% con-
fidence intervals (CIs). For the central analysis linking differ-
ent infant feeding modes in different months to repeated
measurements of reflux, we also adjusted for multiple testing
(six feeding modes), controlling for false discovery rate (FDR)
(Benjamini & Hochberg, 1995), and additionally provide the
adjusted p values.
Results
Approximately 95% (2,841 of 2,988) of the participants pro-
vided information on their offspring reflux for at least one
time point of the questionnaire. About 93% of the participants
answered the 3-month questionnaire, and 59% answered the
one at 12 months (see Table 1). About 10% of the mothers
reported smoking during pregnancy, and 70% of the mothers
had at least some college or had graduated from college.
For all the modes of feeding, the prevalence of reflux was
highest in month 3, ranging from 8.1% to 19.4% at 3 months
(see Table 2). Higher proportions of reflux were found among
infants who received bottled human milk plus formula feed-
ing, followed by those who were fed bottled human milk or
formula (months 3-9). With the exception of formula feeding
and direct breastfeeding, the probability of all the modes of
feeding decreased from month 2 to month 10 (see Table 3).
In contrast, the proportion of infants who received solid food
increased from month 2 to month 10.
Delayed Model
When modeling the role of infant feeding modes on reflux in
the subsequent investigation (month), the reference was
direct breastfeeding. Without adjusting for multiple testing,
three modes of feeding were associated with a higher RR of
reflux: bottled human milk plus formula, formula feeding,
and mixed breastfeeding plus formula feeding (RR = 2.19,
1.95, and 1.59, respectively) (see Table 4), as their confi-
dence limits do not include the null-value of 1. When we
adjusted for multiple testing, only the p value for formula
feeding remained statistically significant.
Since there is a possibility of a reverse association
(reflux different feeding mode), we first analyzed this
option with a month-by-month approach using categorical
modeling and estimating ORs. Infants who experienced
reflux in the preceding month were more likely to be formula
fed in months 3, 6, 7, and 9 compared to infants without
reflux (OR = 1.12, 1.77, 2.19, and 2.00, respectively) (see
Table 5). Also, infants who experienced reflux in the preced-
ing month were 7.74 times more likely to be fed with bottled
human milk plus formula in the subsequent month 6.
Next, we tested a reverse association using a repeated mea-
surement approach. Compared to infants without reflux, hav-
ing an episode of reflux in the preceding month resulted in a
reduced RR of direct breastfeeding (RR = 0.79) (see Table 6),
indicating that mothers stopped direct breastfeeding. Among
the confounders, African American mothers were less likely to
directly breastfeed compared to non-Hispanic white mothers
(RR = 0.42). Mothers who smoked during pregnancy were
more likely to feed formula than mothers who did not smoke
during pregnancy (RR = 1.64) (see Table 6).
Discussion
The prevalence of reflux was highest in month 3 (8.1% to
19.4%) and slowly decreased thereafter. The proportion of
infants who were directly breastfed slightly decreased from
month 2 to month 6, and the tendency of formula feeding
increased from month 2 to month 12. Using a repeated mea-
surement approach and compared to direct breastfeeding,
any feeding mode combined with formula feeding was a risk
factor for the occurrence of reflux in the next month. Our
analysis showed that the addition of solid food was not pro-
tective against reflux. The findings of the reverse association
analysis, also using a repeated measurement approach,
showed decreased direct breastfeeding in infants who had
reflux in the prior month, indicating that mothers did not
continue direct breastfeeding. In addition, infants with reflux
were more likely to be formula fed in the following month,
Chen et al. 5

Table 1. Characteristics of the Study Population.

Sample with reflux Sample with reflux Sample with reflux data
data at any month data at 3 months at 12 months
(n = 2,841) (n = 2,782) (n = 1,771)

Variable n (%) n (%) n (%)

Maternal race/ethnicity
Caucasian 2347 (85.0) 2300 (85.0) 1496 (86.6)
African American 122 (4.4) 121 (4.5) 58 (3.4)
Latina/Hispanic 168 (6.1) 163 (6.1) 97 (5.6)
Other 126 (4.5) 122 (4.5) 77 (4.5)
Maternal age (years)
18-24 635 (22.4) 615 (22.2) 290 (16.4)
25-29 965 (34.0) 942 (33.9) 604 (34.2)
30-34 787 (27.8) 776 (28.0) 533 (30.1)
35 448 (15.8) 443 (16.0) 341 (19.3)
Maternal education status
< High school 81 (3.1) 80 (3.1) 47 (2.9)
High school 454 (17.3) 446 (17.3) 246 (14.7)
College/college graduate 1828 (69.6) 1789 (69.4) 1170 (69.9)
Masters degree 266 (10.1) 264 (10.2) 211 (12.6)
Preconception maternal body mass index (kg/m2)
Underweight (< 18.5) 127 (4.5) 122 (4.4) 67 (3.8)
Normal 1265 (45.1) 1245 (45.4) 794 (45.5)
Overweight (25 to < 30) 734 (26.2) 722 (26.3) 459 (26.3)
Obese ( 30) 677 (24.2) 656 (23.9) 426 (24.4)
Infants gender
Male 1408 (49.6) 1380 (49.7) 874 (49.4)
Female 1430 (50.4) 1399 (50.3) 895 (50.6)
Mode of delivery
Vaginal, not induced 1077 (38.0) 1058 (38.1) 698 (39.5)
Vaginal, induced 957 (33.8) 933 (33.6) 561 (31.8)
Planned cesarean 466 (16.5) 460 (16.6) 306 (17.3)
Unplanned cesarean 333 (11.8) 323 (11.6) 201 (11.4)
Birth weight
< 2500 g 48 (1.7) 48 (1.7) 25 (1.4)
2500-4000 g 2467 (86.9) 2414 (86.9) 1535 (86.8)
> 4000 g 323 (11.4) 317 (11.4) 209 (11.8)
Maternal smoking during pregnancy
Yes 269 (9.6) 262 (9.5) 143 (8.1)
No 2559 (90.4) 2507 (90.5) 1621 (91.9)
Season of birth
Fall 1096 (38.6) 1076 (38.7) 660 (37.3)
Spring 31 (1.1) 30 (1.1) 16 (0.9)
Summer 661 (23.3) 652 (23.5) 429 (24.3)
Winter 1050 (37.0) 1021 (36.7) 664 (37.5)

compared to those without reflux. In the first 5 months,
reflux had a nonsignificant but suggestive effect on solid
food, indicating that infants with reflux were more likely to
receive additional solid food compared to those without
reflux. Overall, these findings suggest that direct breastfeed-
ing was discontinued after the infant experienced reflux and
the mother switched to formula feeding.
The limitation of this study is that, although the sample is
nationally distributed, minority groups were underrepresented
(Grummer-Strawn etal., 2008). The socioeconomic status and
education level in this study population were higher than the
national average and participants had to be English speaking
and moderately literate (Fein etal., 2008; Grummer-Strawn
etal., 2008). Mothers who were less educated, were younger,
had a lower socioeconomic status, participated in WIC, were
nonwhite, and lived in the southern region were more likely to
drop out of the study (Fein etal., 2008; Grummer-Strawn
etal., 2008; Soto-Ramrez etal., 2013). Another limitation is
6
Table 2. Prevalence of Maternal Reports of Infants Reflux Episode in Different Modes of Feeding.

Month 2 Month 3 Month 4 Month 5 Month 6 Month 7 Month 9 Month 10.5 Month 12
(n = 1,704) (n = 2,216) (n = 1,998) (n = 2,090) (n = 2,007) (n = 1,959) (n = 1,872) (n = 1,680) (n = 1,018)
Prevalence of reflux 9.2 10.1 8.9 7.7 6.6 5.4 4.9 3.8 1.6
Stratified by mode of feeding
Direct 6.6 (22/334) 8.3 (34/410) 6.1 (22/360) 4.4 (17/387) 5.0 (19/383) 3.2 (12/376) 4.3 (16/369) 2.4 (8/329) 2.2 (7/325)
breastfeeding
Bottled human 4.8 (1/21) 15.8 (3/19) 7.1 (1/14) 13.3 (2/15) 6.7 (1/15) 7.7 (1/13) 6.7 (1/15) 0 (0/10) 0 (0/5)
milk feeding
Formula feeding 10.9 (63/580) 11.5 (101/875) 11.3 (100/886) 9.9 (92/962) 8.2 (83/1,007) 6.5 (69/1,062) 5.1 (56/1,092) 3.9 (41/1,047) 2.1 (12/571)
Mixed 9.6 (34/354) 9.5 (42/440) 9.7 (34/351) 7.2 (24/330) 6.0 (15/250) 3.8 (8/209) 3.3 (5/152) 6.0 (6/100) 0 (0/49)
breastfeeding
Direct 8.8 (15/171) 8.1 (18/221) 4.3 (7/164) 3.3 (6/180) 5.5 (10/182) 6.3 (12/189) 6.4 (10/157) 4.4 (6/137) 0 (0/47)
breastfeeding plus
formula feeding
Bottled human 7.5 (3/40) 19.4 (7/36) 11.4 (4/35) 14.7 (5/34) 12.5 (3/24) 8.3 (1/12) 12.5 (2/16) 0 (0/11) 0 (0/4)
milk plus formula
feeding
Mixed 9.8 (20/204) 9.3 (20/215) 9.0 (17/188) 7.7 (14/182) 2.1 (3/146) 3.1 (3/98) 4.2 (3/71) 6.5 (3/46) 0 (0/17)
breastfeeding plus
formula feeding
Solid food 15.3 (33/216) 11.7 (52/444) 10.3 (90/875) 7.5 (119/1,579) 6.5 (121/1,872) 5.4 (105/1,950) 5.0 (94/1,892) 3.8 (66/1,740) 1.5 (27/1,763)

Note. N = 2,841. Data are given as % (number of infants who experienced reflux/total number of infants at each feeding mode).
Chen et al. 7

Table 3. Probability of Different Infant Feeding Modes and Addition of Solid Food Stratified by Month.

Month 2 Month 3 Month 4 Month 5 Month 6 Month 7 Month 9 Month 10.5 Month 12
(n = 1,704) (n = 2,216) (n = 1,998) (n = 2,090) (n = 2,007) (n = 1,959) (n = 1,872) (n = 1,680) (n = 1,018)
Direct 334 (19.6) 410 (18.5) 360 (17.9) 387 (18.5) 383 (19.1) 376 (19.2) 369 (19.7) 329 (19.6) 325 (31.9)
breastfeeding
Bottled human 21 (1.2) 19 (0.9) 14 (0.7) 15 (0.7) 15 (0.8) 13 (0.7) 15 (0.8) 10 (0.6) 5 (0.5)
milk feeding
Formula feeding 580 (34.0) 875 (39.5) 886 (44.2) 962 (46.0) 1007 (50.2) 1062 (54.1) 1092 (58.4) 1047 (62.3) 571 (56.1)
Mixed 354 (20.8) 440 (19.9) 351 (17.8) 330 (15.8) 250 (12.5) 209 (10.7) 152 (8.1) 100 (6.0) 49 (4.8)
breastfeeding
Direct 171 (10.0) 221 (10.0) 164 (8.3) 180 (8.6) 182 (9.1) 189 (9.7) 157 (8.4) 137 (8.5) 47 (4.6)
breastfeeding
plus formula
feeding
Bottled human 40 (2.4) 36 (1.6) 35 (1.7) 34 (1.6) 24 (1.2) 12 (0.6) 16 (0.9) 11 (0.7) 4 (0.4)
milk plus formula
feeding
Mixed 204 (12.0) 215 (9.7) 188 (9.5) 182 (8.7) 146 (7.3) 98 (5.0) 71 (3.8) 46 (2.7) 17 (1.7)
breastfeeding
plus formula
feeding
Solid food 210 (12.3) 408 (18.4) 776 (38.8) 1443 (69.0) 1725 (85.9) 18181 (92.8) 1773 (94.7) 1599 (95.2) 976 (95.9)

Note. N = 2,841. Data are given as n (%).

Table 4. Risk Ratios Estimated in a Delayed Model Using Repeated Measurements of Reflux in Infancy From 2 to 12 Months.

Delayed modela

Risk factor Risk ratio 95% CI False discovery rate

Feeding modes
Direct breastfeeding [reference] 1
Bottled human milk feeding 1.26 [0.40, 3.97] 0.4783
Formula feeding 1.95 [1.39, 2.74] 0.0054
Mixed breastfeeding 1.08 [0.73, 1.58] 0.0526
Direct breastfeeding plus formula feeding 1.51 [0.98, 2.31] 0.6069
Bottled human milk plus formula feeding 2.19 [1.11, 4.33] 0.0526
Mixed breastfeeding plus formula feeding 1.59 [1.40, 2.42] 0.0543
Solid food 1.21 [0.86, 1.70]
Gender [female as reference] 1.24 [0.97, 1.58]
Maternal race/ethnicity
African American 0.67 [0.35, 1.28]
Latina/Hispanic 0.56 [0.28, 1.14]
Other 0.51 [0.26, 0.99]
Caucasian [reference] 1
Maternal smoking during pregnancy [nonsmoking as reference] 1.29 [0.84, 2.02]
Maternal age (years)
18 to 25 0.56 [0.32, 0.99]
25 to < 30 [reference] 1
30 to 35 1.23 [0.77, 1.96]
> 35 1.95 [0.84, 4.52]
Parity 0.97 [0.92, 1.02]

Note. CI = confidence interval. Generalized estimating equations (PROC GENMOD with repeated statement and log link function) were used to estimate
relative risks.
a
Events/trials = 357/12,399.
8 Journal of Human Lactation

Table 5. Reverse Association: Odds Ratios of Reflux Related to Different Modes of Feeding With Direct Breastfeeding as Reference.

Direct Bottled Mixed

Direct breastfeeding human milk breastfeeding
breastfeeding Bottled human Formula Mixed plus formula plus formula plus formula Solid
as reference milk feeding feeding breastfeeding feeding feeding feeding food
Month 2 reflux 1 0.74 1.12 0.86 0.98 1.60 0.49 1.48
month 3
feeding
Month 3 reflux 1 0.97 1.51 1.32 1.05 1.35 1.28 1.41
month 4
feeding
Month 4 reflux 1 1.22 1.56 1.28 0.85 2.43 0.98 1.32
month 5
feeding
Month 5 reflux 1 1.29 1.77 1.40 0.54 1.77 0.83 1.44
month 6
feeding
Month 6 reflux 1 2.46 2.19 1.82 2.39 7.74 1.00 0.85
month 7
feeding
Month 7 reflux 1 2.07 2.00 1.32 1.48 2.42 1.58 0.57
month 9
feeding
Month 9 reflux 1 1.50 1.20 1.76 1.27 0.004
month 10
feeding
Month 10 1 1.01 1.03 0.59 0.12
reflux
month 12
feeding

Note. The reflux occurred in the preceding month and the mode of feeding in the subsequent month. Multinomial logistic regression categorical
modeling (PROC CATMOD with the direct and response statement) was used. A dash indicates that data could not be estimated because of the
small sample size.

that we could not adjust for medications used to treat gastro-
esophageal reflux, since this information was not specifically
collected in the IFPS II.
The IFPS II conducted repeated investigations; hence, there
are missing data. Table 2 shows that missing information is
more frequently found in the first survey on breastfeeding and
reflux (month 2) and in the last three surveys (months 9-12).
Missingness seems to result from temporary nonparticipation
or temporary incompleteness. Since GEE models can handle
missing completely at random or missing conditionally at
random (by conditioning on potential confounder, e.g., WIC
and race), we believe that GEE is an appropriate statistical
approach for the longitudinally collected repeated data. Hence,
we do not consider that missingness has biased our results.
A strength of our study was the use of seven categories of
infant feeding modes, which provide a more detailed classifi-
cation of direct (at the breast) breastfeeding, bottled human
milk, formula, and different combinations of feeding strate-
gies. Such a consistent classification was facilitated since the
same information was collected at different points in time
(Soto-Ramrez etal., 2013). These seven categories were
more likely to present a realistic picture of feeding practice
than a simple classification into breastfeeding or formula
feeding (Soto-Ramrez etal., 2013). Another strength was the
use of repeated measurements to analyze the longitudinal data
(Lee, Herzog, Meade, Webb, & Brandon, 2007). First, this
established the correct time order of exposure and response.
Second, the GEE approach did not require the same number
of observations per participant. This was very useful in our
analysis since some participants dropped out and joined again.
The IFPS II is the largest study of infant feeding practices
conducted in the United States that follows all participants
through 12 months of age (Fein etal., 2008). As the IFPS II
questionnaire asked mothers about feeding mode during the past
7 days and reflux in the past 2 weeks, our analysis applied a
1-month delayed model to test the association between infant
feeding mode and reflux and reverse association (reflux feed-
ing), minimizing recall biases (Grummer-Strawn etal., 2008).
It is surprising that our analysis showed that feeding bottled
human milk plus formula feeding has a 119% higher risk of
reflux (see Table 4). Other groups with formula feeding also
had a higher risk of occurrence of reflux (direct breastfeeding
 Table 6. Reverse Association: Risk Ratios Estimated in Delayed Models Using Repeated Measurements of Each Feeding Mode Associated With Reflux in Infancy From 2 to 12
Months.

Direct Bottled human Mixed Direct breastfeeding Bottled human milk Mixed breastfeeding
breastfeeding milk feeding Formula feeding breastfeeding plus formula feeding plus formula feeding plus formula feeding Solid food
(events/trials = (events/trials = (events/trials = (events/trials = (events/trials = (events/trials = (events/trials = (events/trials =
Risk factor 2,652/12,543) 117/12,543) 5,354/12,543) 1,982/12,543) 1,239/12,543) 187/12,543) 1,012/12,543) 7,400/12,399)
Reflux 0.79 [0.66, 0.94] 1.01 [0.62, 1.64] 1.03 [0.96, 1.11] 0.82 [0.65, 1.02] 1.08 [0.80, 1.44] 1.30 [0.70, 2.41] 0.97 [0.72, 1.32] 1.15 [0.82, 1.63]
Gender [female 0.95 [0.83, 1.08] 1.44 [0.74, 2.79] 0.96 [0.89, 1.03] 1.09 [0.95, 1.26] 0.98 [0.83, 1.16] 1.24 [0.83, 1.86] 0.98 [0.83, 1.17] 1.40 [1.19, 1.65]
as reference]
Maternal ethnicity [Caucasian as reference]
African 0.42 [0.24, 0.72] 2.33 [0.77, 7.06] 1.20 [1.04, 1.38] 0.60 [0.36, 1.01] 1.39 [0.93, 2.08] 1.77 [0.6, 3.65] 1.38 [0.89, 2.11] 2.74 [1.77, 4.24]
American
Latina/ 0.64 [0.46, 0.89] 2.63 [0.97, 7.09] 1.02 [0.87, 1.21] 0.72 [0.51, 1.01] 1.62 [1.20, 2.18] 0.85 [0.34, 2.15] 1.32 [0.97, 1.79] 0.77 [0.52, 1.12]
Hispanic
Other 1.03 [0.70, 1.52] 1.15 [0.32, 4.13] 0.80 [0.63, 1.00] 0.68 [0.47, 0.99] 1.85 [1.32, 2.59] 0.83 [0.28, 2.12] 1.12 [0.79, 1.58] 0.60 [0.40, 0.91]
Maternal 0.40 [0.26, 0.62] 1.64 [1.48, 1.80] 0.47 [0.30, 0.73] 0.92 [0.66, 1.27] 0.62 [0.18, 2.08] 0.94 [0.63, 1.39] 1.45 [1.04, 2.03]
smoking during
pregnancy
Parity 0.96 [0.84, 1.09] 1.34 [0.63, 2.45] 1.05 [0.97, 1.13] 1.05 [0.91, 1.21] 0.99 [0.84, 1.18] 1.41 [0.94, 2.11] 0.98 [0.82, 1.17] 0.98 [0.95, 1.02]
Maternal age (years) [25 to < 30 as reference]
18 to < 25 0.71 [0.57, 0.88] 0.64 [0.28, 1.46] 1.46 [1.31, 1.62] 0.63 [0.49, 0.81] 0.95 [0.73, 1.24] 1.01 [0.57, 1.78] 0.99 [0.75, 1.33] 1.30 [0.92, 1.85]
30 to < 35 0.79 [0.67, 0.92] 0.46 [0.19, 1.14] 1.07 [0.96, 1.19] 0.86 [0.73, 1.01] 1.26 [1.02, 1.55] 1.27 [0.79, 2.04] 1.11 [0.90, 1.37] 0.90 [0.67, 1.22]
> 35 0.80 [0.66, 0.96] 0.45 [0.14, 1.43] 1.08 [0.95, 1.22] 0.79 [0.64, 0.97] 1.37 [1.07, 1.75] 0.88 [0.47, 1.68] 1.31 [1.03, 1.67] 0.97 [0.55, 1.70]

Note. Data are given as risk ratio (95% confidence interval). Exposure: reflux; outcome: infant feeding modes (infants who do not have reflux at each feeding mode as reference). Generalized estimating
equations (PROC GENMOD with repeated statement and log link function) were used.

9
10 Journal of Human Lactation

plus formula feeding: 51% higher risk; and mixed breastfeed-
ing plus formula feeding: 59% higher risk). Compared to direct
breastfeeding, formula feeding had a 95% higher risk of reflux.
Wenzl etal. (2003) and Chao and Vandenplas (2007) found
that formula feeding with a thickening agent is an efficient
therapy for uncomplicated GER and regurgitation in infants
(Horvath etal., 2008). However, the sample sizes of these
studies were small (Wenzl etal. n = 14 and Chao & Vandenplas
n = 80). A systematic review based on 14 randomized con-
trolled trials, all with smaller sample sizes, showed that, com-
pared to standard formula, adding thickening agents to formula
may result in a significant decrease in the prevalence of regur-
gitation (Horvath etal., 2008). In contrast, a joint pediatric
recommendation concluded that use of a thickened formula
may decrease visible regurgitation but does not result in a
measurable decrease in the frequency of esophageal reflux
episodes (Vandenplas etal., 2009). Nevertheless, as our
reverse-association delayed-models show, mothers discontin-
ued direct breastfeeding once their infant developed reflux.
However, the above studies failed to address the effect of
direct breastfeeding. A study by Hegar etal. (2013) in
Indonesia demonstrated that infants who received formula
compared to those who were exclusively breastfed had a 5 to
10 times higher frequency of regurgitation and vomiting com-
pared to those exclusively breastfed. Likewise, a Turkish study
showed that babies who had regurgitation had experienced
less breastfeeding than those who did not (Yaln &
Kukonmaz, 2011). Hence, our findings are in agreement with
observational studies showing that direct breastfeeding is a
protective factor against reflux.
Conclusion
Our repeated measurement analysis of nine ascertainments of
reflux during infancy showed that direct breastfeeding, not
including feeding of bottled milk, protected against reflux
events. Early introduction of solid food did not guard infants
against reflux and formula feeding poses a higher RR of reflux.
The reverse association analyses showed that infants who had
reflux experienced higher odds of consuming solid food in
consecutive months, suggesting that parents may be faced
with misleading information on how to manage GER in their
infants. It is essential to inform pediatricians and health pro-
viders that adding solid food in the diet is not beneficial for
infants with reflux but that direct breastfeeding is beneficial.
Further study on gastroesophageal reflux needs to con-
sider factors such as volume, time, and rate of swallowing
required for feeding, interruption of breathing, or frequency
of gastric emptying (Goldfield, Richardson, Lee, & Margetts,
2006; Heacock etal., 1992; Koenig, Davies, & Thach, 1990;
Mizuno & Ueda, 2006). It is possible that short-term intake
of a larger volume may increase the risk of reflux.

Appendix. Classification of Feeding Modes.

Does your baby How many times in the In the past 7 days,
usually feed from Does your baby past 7 days was your In the past 7 days, how how often was your
both breasts at usually let go of the baby fed expressed or often was your baby fed baby fed each food
each feeding? breast himself or pumped breast milk? each food listed below? listed below? (N40b):
Mode of feeding (N56) herself? (N57) (N61) (N40a): breast milk formula
Direct Valid answer No 0 >0 0
breastfeeding Valid answer Yes 0 0
Valid answer Missing 0 0
Bottled breast Baby is fed only Jumped 1 >0 0
milk feeding pumped milk
Formula feeding Instructed to skip if Instructed to skip if Instructed to skip =0 1
not breastfeeding not breastfeeding
0 1
Bottled breast Valid answer No 1 >0 0
milk and direct Valid answer Yes 1 0
breastfeeding Valid answer Missing 1 0
Direct Valid answer No 0 >0 1
breastfeeding Valid answer Yes 0 1
plus formula Valid answer Missing 0 1
feeding
Breast milk plus Valid answer Missing 1 >0 1
formula feeding Valid answer Missing 0 1
Mixed: direct Valid answer Yes 1 >0 1
breastfeeding,
bottled breast Valid answer Missing 1 1
milk feeding, Valid answer No 1 >0 1
formula feeding

Note. The numbers in parentheses (e.g., N56) represent the question numbers used in the questionnaire of the Infant Feeding Practices Study II.
Chen et al. 11
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect
to the research, authorship, and/or publication of this article.
Funding
The authors disclosed receipt of the following financial support for
the research, authorship, and/or publication of this article: This
work was supported by a grant from the Centers for Disease Control
and Prevention (HRSA MCHB grant R40 MC 17165), United
States.
References
Agresti, A., & Liu, I. (2001). Strategies for modeling a categori-
cal variable allowing multiple category choices. Sociological
Methods & Research, 29(4), 403-434.
American Academy of Pediatrics. (2012). Breastfeeding and the use
of human milk. Pediatrics, 129(3), e827-e841. doi:10.1542/
peds.2011-3552
Benjamini, Y., & Hochberg, Y. (1995). Controlling the false discov-
ery rate: A practical and powerful approach to multiple testing.
Journal of the Royal Statistical Society: Series B. Statistical
Methodology, 57, 289-300.
Chao, H. C., & Vandenplas, Y. (2007). Effect of cereal-thickened
formula and upright positioning on regurgitation, gastric emp-
tying, and weight gain in infants with regurgitation. Nutrition,
23(1), 23-28.
Cui, J. (2007). QIC program and model selection in GEE analyses.
The Stata Journal, 7(2), 209-220.
Czinn, S. J., & Blanchard, S. (2013). Gastroesophageal reflux dis-
ease in neonates and infants: When and how to treat. Paediatric
Drugs, 15(1), 19-27.
Fein, S. B., Labiner-Wolfe, J., Shealy, K. R., Li, R., Chen, J., &
Grummer-Strawn, L. M. (2008). Infant Feeding Practices
Study II: Study methods. Pediatrics, 122(Suppl. 2), S28-S35.
doi:10.1542/peds.2008-1315c
Goldfield, E. C., Richardson, M. J., Lee, K. G., & Margetts, S.
(2006). Coordination of sucking, swallowing, and breathing
and oxygen saturation during early infant breast-feeding and
bottle-feeding. Pediatric Research, 60(4), 450-455.
Grummer-Strawn, L. M., Scanlon, K. S., & Fein, S. B. (2008). Infant
feeding and feeding transitions during the first year of life.
Pediatrics, 122(Suppl. 2), S36-S42. doi:10.1542/peds.2008-
1315d
Hadders-Algra, M. (2013). Typical and atypical development of reach-
ing and postural control in infancy. Developmental Medicine and
Child Neurology, 55(Suppl. 4), 5-8. doi:10.1111/dmcn.12298
Heacock, H. J., Jeffery, H. E., Baker, J. L., & Page, M. (1992).
Influence of breast versus formula milk on physiological gas-
troesophageal reflux in healthy, newborn infants. Journal of
Pediatric Gastroenterology and Nutrition, 14(1), 41-46.
Hegar, B., Dewanti, N. R., Kadim, M., Alatas, S., Firmansyah, A.,
& Vandenplas, Y. (2009). Natural evolution of regurgitation in
healthy infants. Acta Paediatrica, 98(7), 1189-1193.
Hegar, B., Satari, D. H., Sjarif, D. R., & Vandenplas, Y. (2013).
Regurgitation and gastroesophageal reflux disease in six to nine
months old Indonesian infants. Pediatric Gastroenterology,
Hepatology & Nutrition, 16(4), 240-247.
Hegar, B., & Vandenplas, Y. (2013). Gastroesophageal reflux:
Natural evolution, diagnostic approach and treatment. Turkish
Journal of Pediatrics, 55(1), 1-7.
Horvath, A., Dziechciarz, P., & Szajewska, H. (2008). The effect of
thickened-feed interventions on gastroesophageal reflux in infants:
Systematic review and meta-analysis of randomized, controlled tri-
als. Pediatrics, 122(6), e1268-e1277. doi:10.1542/peds.2008-1900
Jadcherla, S. R., Chan, C. Y., Moore, R., Malkar, M., Timan, C. J.,
& Valentine, C. J. (2012). Impact of feeding strategies on the
frequency and clearance of acid and nonacid gastroesophageal
reflux events in dysphagic neonates. Journal of Parenteral and
Enteral Nutrition, 36(4), 449-455.
Khoshoo, V., Ross, G., Brown, S., & Edell, D. (2000). Smaller
volume, thickened formulas in the management of gastro-
esophageal reflux in thriving infants. Journal of Pediatric
Gastroenterology and Nutrition, 31(5), 554-556.
Koenig, J. S., Davies, A. M., & Thach, B. T. (1990). Coordination
of breathing, sucking, and swallowing during bottle feedings in
human infants. Journal of Applied Physiology, 69(5), 1623-1629.
Lee, J. H., Herzog, T. A., Meade, C. D., Webb, M. S., & Brandon,
T. H. (2007). The use of GEE for analyzing longitudinal bino-
mial data: A primer using data from a tobacco intervention.
Addictive Behaviors, 32(1), 187-193.
Lightdale, J. R., Gremse, D. A., & Section on Gastroenterology,
Hepatology, & Nutrition. (2013). Gastroesophageal reflux:
Management guidance for the pediatrician. Pediatrics, 131(5),
e1684-e1695. doi:10.1542/peds.2013-0421
Mizuno, K., & Ueda, A. (2006). Changes in sucking performance
from nonnutritive sucking to nutritive sucking during breast-
and bottle-feeding. Pediatric Research, 59(5), 728-731.
Nevo, N., Rubin, L., Tamir, A., Levine, A., & Shaoul, R. (2007).
Infant feeding patterns in the first 6 months: An assessment in
full-term infants. Journal of Pediatric Gastroenterology and
Nutrition, 45(2), 234-239.
Skov, T., Deddens, J., Petersen, M. R., & Endahl, L. (1998).
Prevalence proportion ratios: Estimation and hypothesis test-
ing. International Journal of Epidemiology, 27(1), 91-95.
Soto-Ramrez, N., Karmaus, W., Zhang, H., Davis, S., Agarwal,
S., & Albergottie, A. (2013). Modes of infant feeding and
the occurrence of coughing/wheezing in the first year of life.
Journal of Human Lactation, 29(1), 71-80.
Vandenplas, Y., Rudolph, C. D., Di Lorenzo, C., Hassall, E., Liptak, G.,
Mazur, L.,... European Society for Pediatric Gastroenterology,
Hepatology, and Nutrition. (2009). Pediatric gastroesophageal reflux
clinical practice guidelines: Joint recommendations of the North
American Society for Pediatric Gastroenterology, Hepatology, and
Nutrition (NASPGHAN) and the European Society for Pediatric
Gastroenterology, Hepatology, and Nutrition (ESPGHAN). Journal
of Pediatric Gastroenterology and Nutrition, 49(4), 498-547.
Wenzl, T. G., Schneider, S., Scheele, F., Silny, J., Heimann, G.,
& Skopnik, H. (2003). Effects of thickened feeding on gastro-
esophageal reflux in infants: A placebo-controlled crossover
study using intraluminal impedance. Pediatrics, 111(4, Pt. 1),
e355-e359. doi:/content/111/4/e355.full.html
Wojcicki, J. M., Gugig, R., Tran, C., Kathiravan, S., Holbrook, K., &
Heyman, M. B. (2010). Early exclusive breastfeeding and mater-
nal attitudes towards infant feeding in a population of new mothers
in San Francisco, California. Breastfeeding Medicine, 5(1), 9-15.
Yaln, S. S., & Kukonmaz, B. B. (2011). Relationship of lower
breastfeeding score and problems in infancy. Breastfeeding
Medicine, 6(4), 205-208.
Zhang, J., & Yu, K. F. (1998). Whats the relative risk? A method of
correcting the odds ratio in cohort studies of common outcomes.
Journal of the American Medical Association, 280(19), 1690-1691.