Keywords
antibiotics - neonates - obesity - childhood - gut microbiome
Obesity has reached epidemic proportions in the developed world, with one in three
children in the United States considered overweight or obese.[1] It is estimated that 8.1% of infants and toddlers and 16.9% between the age of 2
to 19 years have a body mass index (BMI) greater than the 85th percentile,[2] making it one of the greatest public health threats in the country. South Bronx,
a socially disadvantaged and medically underserved area in New York City, is known
to have some of the highest burdens of various medical conditions and a much higher
prevalence of obesity (47%)[3] among children and adolescents compared with a national rate of 19.7%.[4] Aside from the significant risk of developing cardiometabolic conditions later in
life, the increased prevalence of childhood obesity is associated with comorbidities
previously considered “adult” diseases, including type 2 diabetes mellitus, hypertension,
nonalcoholic fatty liver disease, obstructive sleep apnea, and dyslipidemia.[5]
The development of obesity in children is multifactorial and influenced by diet, physical
activity, and socioeconomic factors. Lifestyle modifications, such as balancing calories
and increasing physical activity, continue to be first-line strategies for preventing
obesity in the pediatric population. However, this strategy has yielded a modest impact.
Therefore, it is vital to seek and understand other modifiable risk factors leading
to childhood obesity.[6] Research has identified that more intrinsic risk factors, including genetics and
gut microbiome, play a role in the development of obesity.[7] Specifically, the human gut microbiome, which comprises more than 100 trillion microbial
cells, has a wide-reaching metabolic, nutritional, and immunological effect.[8] It is known to influence overall metabolism through the regulation of energy expenditure,
fat storage, and maintaining immune and gut barrier functions.[8] Development of the gut microbiome starts at birth and is influenced primarily by
the mode of delivery and feeding practices, among other factors.[9] Studies have shown that microbiome composition is stable starting around 6 months
of age. The neonatal period is a critical time point in the development of host and
gut microbiome symbiosis, as it is the time of intense development and maturation
of the immune system and microbial imprinting.[10] Alterations of the microbiome increase the risk of developing chronic medical conditions
such as asthma, inflammatory bowel disease, allergic and inflammatory disorders, and
obesity. Association between the gut microbiome composition and the development of
these chronic medical conditions has been established in previous research.[11]
[12]
[13]
In animal models, changes in the microbiome brought about by antibiotics resulted
in weight gain through direct effects on growth and metabolic pathways associated
with fat production, metabolism, and storage.[14]
[15] Research in humans has produced mixed results, with some studies showing a significant
association between early-life antibiotic exposure and increased risk of development
of obesity.[7]
[14]
[16]
[17]
[18]
[19]
[20] In contrast, others report either marginal or no risk.[21]
[22]
[23]
However, studies are sparse on neonatal antibiotic exposure and the risk of overweight
or obesity in early childhood. Dysbiosis in early infancy could potentially have long-lasting
effects on weight trajectory later in childhood.[14] Hence, it is critical to explore the probability of antibiotic-induced dysbiosis
within the first month of life, particularly in infants exposed to antibiotics in
the neonatal intensive care unit (NICU), as it may provide additional impetus to limit
unnecessary antibiotic use.
In this study, we assessed the association between neonatal antibiotic exposure and
the risk of childhood obesity with a secondary dose–response assessment and comparison
of antibiotic exposure effect between those exposed only in the first 72 hours of
life and beyond 72 hours of life.
Materials and Methods
Study Design and Population
This is a retrospective cohort study on children born at BronxCare Health System or
extramural deliveries admitted to the NICU/nursery who were exposed to antibiotics
(ampicillin, gentamicin, or a cephalosporin) as well as a control group of infants
who were unexposed during the birth admission and followed up annually with respect
to the weight for length or BMI percentile from the second year of life through the
age of 5. We reviewed medical records from 2011 to 2020 and extracted data for exposed
and unexposed neonates.
Our study focused on children delivered, or that received care, at BronxCare Health
System within the first 28 days of life and followed up in the BronxCare Health System
outpatient clinics consistently through the age of 5. We included infants born at
34 to 41 completed weeks of gestation and excluded preterm neonates less than 34 weeks'
gestational age, as they received antenatal steroids for fetal lung maturity. Postterm
neonates and preterm neonates were excluded because of their possible confounding
effect.[24]
[25] Children with congenital conditions affecting growth (i.e., Prader–Willi), congenital
hypothyroidism, cystic fibrosis, and congenital immune deficiencies were also excluded.
Measurements
Measurement of Exposure
The exposure variable, neonatal antibiotic exposure, was extracted from electronic
medical records (EMR) from the medication administration section indicating actual
administration, including dose and length of antibiotic treatment. Ampicillin, gentamicin,
cefotaxime, ceftazidime, antifungal, and antiviral agents were extracted for analysis.
However, because ampicillin and gentamicin constituted greater than 98% of antibiotics
used in our NICU, they were the primary exposure variables considered. The predominant
reason for antibiotic administration was to rule out sepsis in neonates with fever,
hypothermia, respiratory distress, and/or maternal chorioamnionitis. The exposure
period was between 2011 and 2015, allowing a 5-year follow-up of children until 2020.
Similarly, the control group, neonates in the NICU or nursery admitted at the same
time and did not receive antibiotics, were extracted from the EMR, and followed up
for the same duration. The total number of live births delivered in our hospital including
extramural deliveries is 2,388 out of which 1,447 met our inclusion criteria. Preterm
neonates (<34 weeks) and postterm neonates accounted for 399 of the excluded data
and 542 did not have 5-year follow-up data.
Covariates
Our covariates included variables shown in previous literature to be significantly
associated with childhood obesity or based on biological plausibility. These included
maternal age at delivery, ethnicity, pregravid maternal weight, mode of delivery,
intrapartum prophylaxis against group B streptococcal disease (GBS), birth weight,
neonatal gender, type of feeding (breastfeeding, formula, or both), and mother and
child comorbidities. For the early feeding practices, we reviewed charts at health
care maintenance visits at 0 to 2, 3 to 4, and 6 months of age to check if the infants
were exclusively breastfed, formula fed, or both. For the intrapartum antibiotic prophylaxis
(IAP) against GBS, adequate IAP was defined as administration of penicillin >4 hours
before delivery and inadequate refers to antibiotics other than penicillin or penicillin
administration for <4 hours before delivery. These variables were extracted by carefully
reviewing the maternal and child records in the EMR database.
Measurement of Outcome
Ambulatory clinics visit records were reviewed for the identified exposed and unexposed
cohorts. The documented CDC BMI percentile at the 5-year-old health care maintenance
visit was used in the analysis. We also included those that missed the 5-year-old
visit but were seen for acute ambulatory care or in the emergency department 5 months
before or after their fifth birthday. Overweight was defined as BMI percentile at
85th to less than the 95th percentile for age and sex, and obesity was defined as BMI at or greater than the
95th percentile for age and sex. Some studies reviewed operationalized outcomes of relative
body mass using BMI-for-age z-score,[16] weight-for-length z-score,[21] or weight-for-age z-score and categorized these outcomes based on set references. In this study, we used
dichotomized BMI percentile (BMI percentile ≥ 95th percentile or BMI percentile < 95th percentile) at 5 years as the primary outcome
variable.[26]
[27]
[28]
Statistical Analysis
Our sample consisted of 1,447 neonates admitted to the NICU or nursery and followed
up in our outpatient clinics for at least 5 years. The maternal–child demographics,
anthropometric measures, and other clinical characteristics were summarized using
descriptive statistics. Categorical variables were presented in frequencies and percentages,
and continuous variables were presented as mean (standard deviation). The normality
of the variables was tested using the Kolmogorov–Smirnov test. A bivariate analysis
was done to examine if the risk of the outcome variable (childhood obesity) significantly
differs between the two cohorts based on different predictor variables. Chi-square
statistics analyzed the categorical variables, and the continuous variables were analyzed
by Student's t-test since they are normally distributed. Statistical significance was set at a p < 0.05. All the statistically significant variables in the bivariate analysis were
modeled in a stepwise multivariate logistic regression analysis to determine whether
neonatal ampicillin–gentamicin exposure is an independent predictor of early childhood
obesity. In our analysis, we used the stepwise selection method, a flexible method
that can be directly applied to mixed continuous and categorical predictors and allows
the selection of significant variables at a p < 0.05. The stepwise selection method is very attractive because of its tractability
and gives a good sequence of models that helped us identify predictive factors associated
with early childhood obesity. A receiver operator characteristic (ROC) curve was plotted
for the final models predicting our outcome variable. The goodness of fit of the multivariate
logistic model using the Hosmer–Lemeshow test, Cox and Snell R-square, and max rescaled
R-square were also evaluated. The multicollinearity of variables was checked to identify
and adjust for colinear independent variables in the regression model. We also did
an ad hoc analysis, where we stratified neonates that were exposed to antibiotics
into two groups; exposure less than or equal to 72 hours and greater than 72 hours
and we did a bivariate analysis to determine whether there is an association with
childhood obesity based on the duration of exposure. We chose 72 hours because, during
the time frame when our exposure cohorts were extracted, 2011 to 2015, we administered
prophylactic antibiotic treatment for 72 hours pending blood culture results for neonates
with suspected neonatal sepsis, unlike 48 hours of treatment with negative blood culture
result that we currently practice. Statistical analysis was conducted using SAS version
9.4 software (Cary, NC), with statistical significance set at p < 0.05.
Results
Characteristics of Study Participants
A total of 1,447 neonates were examined. A total of 749 (51.8%) were exposed to ampicillin–gentamicin,
and 698 (48.2%) were not exposed to antibiotics. Of the total subjects, 333 (23%)
were obese, and 1,114 (77%) were not obese at 5 years of age. Other chronic inflammatory
conditions like childhood asthma and atopic diseases constituted 12%. Demographic
characteristics were similarly distributed based on maternal age, gestational age,
and gender. A total of 54% of mothers were of Hispanic origin. Vaginal delivery was
the prevalent route of delivery (61.3%). Only 7.4% were exclusively breastfed for
at least 4 months, and most of the early feeding practice was formula (49.2%) and
mixed (43.5%). Of the 350 (25%) mothers who required GBS IAP, 244 (69.7%) received
adequate IAP. Approximately 9, 8, and 33% of the mothers had diabetes mellitus, hypertension,
and obesity, respectively ([Table 1]).
Table 1
Baseline and maternal–child characteristics by obesity status
|
Characteristics
|
All, N (%)
|
Obese
|
p-Value
|
|
|
Yes, n = 333
|
No, n = 1,114
|
|
|
Gender
|
|
|
|
0.48
|
|
Male
|
732 (50.7)
|
163 (22.3)
|
569 (77.7)
|
|
|
Female
|
713 (49.3)
|
170 (233.8)
|
543 (76.2)
|
|
|
Maternal race
|
|
|
|
<0.01[a]
|
|
Black/African
|
562 (38.9)
|
76 (13.5)
|
486 (86.5)
|
|
|
Hispanics
|
782 (54.1)
|
226 (28.9)
|
556 (71.1)
|
|
|
Others
|
101 (7.0)
|
31 (30.7)
|
70 (69.3)
|
|
|
Maternal age (y)
|
1,445
|
28.56 (6.61)
|
28.32 (6.68)
|
0.56
|
|
Birth weight (g), mean (± SD)
|
1,437
|
3,318.22 (688.9)
|
3,132.93 (619.0)
|
<0.01[a]
|
|
Length of stay (d), mean (± SD)
|
1,447
|
5.80 (6.6)
|
5.93 (7.55)
|
0.77
|
|
Delivery route
|
|
|
|
0.12
|
|
C-section
|
553 (38.8)
|
140 (25.3)
|
413 (74.7)
|
|
|
Vaginal
|
874 (61.3)
|
190 (21.7)
|
684 (78.3)
|
|
|
Ampicillin–gentamicin
|
|
|
|
0.01[a]
|
|
Yes
|
749 (51.8)
|
193 (25.8)
|
556 (74.2)
|
|
|
No
|
698 (48.2)
|
140 (20.1)
|
558 (79.9)
|
|
|
Ampicillin count[b]
|
|
7.81 (3.87)
|
8.21 (4.11)
|
0.23
|
|
Gentamicin count[b]
|
|
3.60 (1.88)
|
3.85 (1.92)
|
0.12
|
|
Early feeding
|
|
|
|
0.05[a]
|
|
Breastfed
|
104 (7.4)
|
16 (15.4)
|
88 (84.6)
|
|
|
Formula
|
690 (49.2)
|
157 (22.8)
|
533 (77.2)
|
|
|
Mixed
|
610 (43.5)
|
158 (25.9)
|
452 (74.1)
|
|
|
Intrapartum GBS prophylaxis
|
|
|
|
0.38
|
|
Adequate
|
244 (17.1)
|
65 (26.6)
|
179 (73.4)
|
|
|
inadequate
|
106 (7.4)
|
23 (31.7)
|
83 (78.3)
|
|
|
None
|
1,076 (75.5)
|
244 (22.7)
|
832 (77.3)
|
|
|
Childhood asthma
|
|
|
|
0.39
|
|
Yes
|
185 (12.8)
|
38 (20.5)
|
147 (79.5)
|
|
|
No
|
1,261 (87.2)
|
295 (23.4)
|
966 (76.6)
|
|
|
Child atopic disease
|
|
|
|
<0.01[a]
|
|
Yes
|
307 (21.2)
|
99 (32.2)
|
208 (67.8)
|
|
|
No
|
1,139 (78.8)
|
234 (20.5)
|
905 (79.5)
|
|
|
Maternal diabetes
|
|
|
|
<0.01[a]
|
|
Pregestational
|
28 (1.9)
|
10 (35.7)
|
18 (64.3)
|
|
|
Gestational
|
106 (7.3)
|
40 (37.7)
|
66 (62.3)
|
|
|
None
|
1313 (90.7)
|
283 (21.6)
|
1,030 (78.4)
|
|
|
Maternal hypertension
|
|
|
|
0.05[a]
|
|
Chronic hypertension
|
63 (4.4)
|
14 (22.2)
|
49 (77.8)
|
|
|
Gestational hypertension
|
48 (3.3)
|
18 (37.5)
|
30 (62.5)
|
|
|
None
|
1,336 (92.3)
|
301 (22.5)
|
1,035 (77.5)
|
|
|
Maternal asthma
|
|
|
|
0.83
|
|
Yes
|
41 (2.8)
|
10 (24.4)
|
31 (75.6)
|
|
|
No
|
1,406 (97.2)
|
323 (23.0)
|
1,083 (77.0)
|
|
|
Maternal obesity
|
|
|
|
<0.01[a]
|
|
Yes
|
473 (32.7)
|
138 (29.2)
|
335 (70.8)
|
|
|
No
|
974 (67.3)
|
195 (20.0)
|
779 (80.0)
|
|
|
Maternal BMI, kg/m2 mean (± SD)
|
1,447
|
29.85 (6.46)
|
28.25 (6.36)
|
<0.01[a]
|
Abbreviations: BMI, body mass index; GBS, group B streptococcal disease; SD, standard
deviation.
Notes: Descriptive and bivariate analysis. Comparing baseline characteristics and
other clinical factors of mother–child dyad based on child obesity status at 5 years.
Results for categorical data are presented as percentage and frequency. Pearson's
chi-square is used to compare differences between demographic and clinical characteristics
among the children based on their obesity status, and the result of normally distributed
continuous variable are presented as mean ± SD. Student's t-test was used to compare differences between clinical characteristics among the children
based on their obesity status. Clinical diagnoses mentioned are based on entries in
the medical record and not on any prespecified diagnostic criteria.
a Statistical significance (p < 0.05).
b The ampicillin/gentamicin counts refers the number of times ampicillin at 50 mg/kg/dose
and gentamicin at 4 mg/kg/dose was given.
Factors Associated with Childhood Obesity at 5 Years
[Table 2] shows the baseline demographics and other clinical characteristics associated with
childhood obesity at 5 years based on ampicillin–gentamicin exposure status. Compared
with the children without antibiotic exposure, those exposed had a higher incidence
of obesity (25.8 vs. 20.1%, p < 0.01). Infants that were formula-fed or got mixed feeding had a higher incidence
of obesity (22.8 and 25.9%, respectively) compared with breastfed infants (15.4%,
p < 0.05). Compared with no atopy, those with childhood atopic disease were more likely
to be obese (32.2 vs. 20.5%, p < 0.01). The incidence of obesity was higher among children exposed to maternal pregestational
and gestational diabetes compared with no diabetes (35.7 and 37.7 vs. 21.6%, p < 0.01). Also, children exposed to maternal gestational hypertension had a higher
incidence of obesity (37.5%) compared with chronic hypertension or no hypertension
(22.2 and 22.5%, respectively, p = 0.05). Compared with vaginal delivery, more infants delivered via cesarean section
were obese at 5 years (25.3 vs. 21.7%), but the difference was not statistically significant.
The maternal age, neonatal length of hospital stay, gestational age, ampicillin, and
gentamicin dose, GBS IAP, and maternal and childhood asthma were not statistically
different among the children with and without obesity ([Table 1]).
Table 2
Obesity at 5 years based on the duration of antibiotic exposure
|
Obese
|
p-Value
|
|
Yes
|
No
|
|
|
Ampicillin–gentamicin use duration (n = 749)
|
≤ 72 h
|
200 (26.7)
|
549 (73.3)
|
0.31
|
|
> 72 h
|
173 (23.2)
|
576 (76.8)
|
Note: Ampicillin–gentamicin exposure stratified into ≤ 72 and > 72 hours and obesity
status at 5 years.
[Table 3] shows factors independently associated with childhood obesity at 5 years. Following
a stepwise multivariable logistic regression analysis, ampicillin and gentamicin exposure
remains an independent predictor of childhood obesity, with a 37% increase in the
risk of obesity among the exposed compared with the unexposed group (adjusted odds
ratio: 1.37, 95% confidence interval: 1.055–1.775).
Table 3
Factors independently associated with childhood obesity
|
Variable
|
Adjusted odds ratio (aOR)
|
95% Confidence interval
|
p-Value
|
|
|
Lower
|
Upper
|
|
|
Ampicillin–gentamicin
|
1.368
|
1.055
|
1.775
|
0.02[a]
|
|
Maternal obesity
|
1.664
|
1.274
|
2.174
|
<0.01[a]
|
|
Child atopy
|
1.744
|
1.302
|
2.336
|
<0.01[a]
|
|
Maternal diabetes
|
|
|
|
<0.01[a]
|
|
None
|
Reference
|
|
Gestational
|
2.295
|
1.474
|
3.573
|
<0.01[a]
|
|
Pregestational
|
2.037
|
0.867
|
4.788
|
0.10
|
|
Maternal race
|
|
|
|
<0.01[a]
|
|
Hispanics
|
Reference
|
|
Black/African
|
0.357
|
0.265
|
0.480
|
<0.01[a]
|
|
Others
|
1.131
|
0.712
|
1.797
|
0.60
|
Abbreviations: aOR, adjusted odds ratio; OR, odds ratio.
Notes: Factors associated with childhood obesity at 5 years. Adjusted OR were presented.
Adjusted OR < 1 denotes factors that are associated with a lesser risk of childhood
obesity while OR > 1 denotes factors that are associated with a higher risk of childhood
obesity. Hosmer–Lemeshow test (p = 0.7431), indicates that our models fit the data. Max-rescaled R-square (R
2 = 0.1073) was used to assess how well our model explains the variability in the ambulatory
outcome.
a Statistical significance (p < 0.05) with a 95% confidence interval that excludes null value of 1.
We further stratified the antibiotic-exposed group by the duration of exposure, less
than or equal to 72 versus greater than 72 hours. Obesity among the ampicillin and
gentamicin exposed group was 26.7 versus 23.1% for exposure less than or equal to
72 hours and greater than 72 hours, respectively. However, this difference was not
statistically significant ([Table 2]).
Discussion
In our retrospective study of a large cohort of children born in a community with
a high prevalence of obesity, we found that those exposed to antibiotics during the
birth admission were more likely to be obese at 5 years of age compared with those
who were not exposed. To our knowledge, this is the first study to report an association
between neonatal exposure to penicillin-based and aminoglycoside antibiotics and obesity
in early childhood. Our study is consistent with some studies, which showed a similar
association between antibiotic exposure in early childhood and an increased risk of
obesity.[6]
[13]
[15]
[19]
[22]
[25] Scott et al found a similar association to our study, reporting that antibiotic
exposure before the age of 2 years increased the odds of childhood obesity at 4 years
by 21%, consistent with our study, where antibiotic exposed neonate showed a 37% increase
in the risk of obesity at 5 years. An earlier study found the obesity rate among children
and adolescents (3–21 years) to be about 47% in the South Bronx.[3] However, our study presents the prevalence of obesity at age 5, which may explain
the relatively lower rate (23%).
A few studies have shown a negative association between early antibiotic exposure
and childhood obesity.[20]
[21]
[22]
[26]
[27]
[28] The differing methodologies and exposure time to outcome evaluation used in these
studies could explain the inconsistent findings. For example, Rogawski et al and Gerber
et al extracted their antibiotic exposure variable from the prescription database,[21] caregiver verbal report, and prescription data (20) with no means of verification,
which could lead to misclassification bias. We have minimized this bias by utilizing
data that was extracted directly from the EMR, including the record of actual administration
of antibiotics with standardized weight-based dosing and duration of treatment. Pyle
et al and Kamphorst et al reported decreased weight gain with neonatal antibiotic
exposure. It is important to highlight the differences in the population characteristics,
wherein preterm (<32 weeks) and very low birth weight infants were included[29] and exposure time to outcome evaluation was 12 months.[30] We believe the differences in population characteristics and exposure time to outcome
evaluation resulted in the differing findings. Li et al also suggested that the differences
among studies may result from a lack of differentiation between exposure to maternal
infection (especially GBS) versus antibiotic administration, for which superior study
designs or a propensity score analysis may be required to further explore the association.[31]
Increased exposure to antibiotics through longer or multiple courses is thought to
lead to more pronounced alterations in the gut microbiome and potentially more effects
on the metabolic pathways. Previous studies[7]
[19] have demonstrated a dose-dependent association between antibiotic use and childhood
obesity when utilizing data with multiple antibiotic exposures over a prolonged time
throughout childhood. However, we did not find a significant difference between the
incidence of obesity between infants exposed for less than or equal to 72 versus greater
than 72 hours, suggesting that mere exposure, regardless of duration and dose, is
impactful in the neonatal period. However, the effect we found in our study may also
be limited to the timing of antibiotic exposure in our cohort, as animal models suggest
that earlier exposure may lead to more significant alterations.[32] The risk of developing obesity with antibiotic exposure in the first month of life
emphasizes that the neonatal microbiome is a dynamic environment that is extremely
inducible by external factors like antibiotics.[21]
[22]
[23]
This emphasis on exposure timing is fundamental. For instance, some studies[16]
[17] found a positive association in children exposed to antibiotics before 6 months.
Scott et al found increasing risk of obesity as the timing of exposure shortens from
2 to 1 year to <6 months.[7] Others, like Block et al, reported a marginal association between antibiotic exposure
in infants exposed under 24 months of age and being overweight or obese at 5 years
of age.[14] It is important to highlight that in the population we studied, antibiotics were
administered in the first few days, if not first hours, of life. By focusing solely
on the neonatal period, we were able to explore the effects of antibiotic exposure
on the formative period of gut microbiome development and confirm that alterations
at this time may have been impactful on the future weight trajectories of the children
included in the study. Hence, neonatal antibiotic exposure is a modifiable risk factor
for obesity in our population, and cautious use of antibiotics at this time may have
critical importance in preventing childhood obesity.
In our study, the association between antibiotic exposure and childhood obesity remained
significant after controlling for covariates, some of which include maternal comorbidities
and feeding practices. We found that maternal obesity and gestational diabetes were
significantly associated with childhood obesity at 5 years. This association is intuitive
as maternal genetic or environmental predispositions that led to maternal comorbidities
can easily influence weight trajectories in their offspring.[33] Additionally, in our study, the unadjusted analysis showed an association between
feeding practices and childhood obesity; the variable was removed in the stepwise
multivariate analysis model due to nonsignificance at 0.05 significance level. This
finding is inconsistent with previous studies, which suggested that breastfeeding
is an important early contributor to gut microbiota.[20]
[21] These studies surmised that antibiotic exposure might attenuate the beneficial effect
of breastfeeding. However, since only 7.4% of our study population were exclusively
breastfed, that conclusion is difficult to ascertain from our study population.
The initial development and maturation of the neonatal microbiome is closely related
to the maternal microbiome, which is influenced by diet, infection, antibiotic use,
and a host of other factors. During the perinatal period, there is transfer of the
maternal microbiota, especially from vaginal fluid and gastrointestinal tract to the
neonate's gut during vaginal delivery, forming the core of the neonatal gut microbiome.
Skin bacteria colonize infants born via cesarean section.[34] Delivery route was not significantly related to child obesity status in our bivariate
analysis and was not included in our final model. The noninclusion of the delivery
route as a confounder in our model is a potential limitation.
Strengths and Limitations
Strengths and Limitations
Strengths of this study include large sample size and the retrospective cohort design,
which provides a means of estimating risk, as opposed to a cross-sectional or case–control
design that suggests only association. Additionally, the direct extraction of our
primary independent variable (ampicillin/gentamycin exposure) from the administered
medication section of the EMR reduced misclassification bias, added to the exposure
classification's validity, and improved our ability to detect a true effect. We also
included multiple covariates (including variables previously established as risk factors
or based on biological plausibility) in our analysis. This helped us to account for
their possible effect on our outcome variable, obesity at 5 years.
Some potential limitations of this study include a single-center analysis and the
study design, which only suggested association but was not equipped to validate causation.
Hence, prospective studies will be needed to establish the potential risk of neonatal
antibiotic exposure in the development of early childhood obesity. The lack of exploration
of additional antibiotic use after the birth admission and the failure to evaluate
the possible effect of such exposure on the outcome variable is another potential
limitation. In addition, selection bias, which is inherent in retrospective studies,
may have influenced our result. Therefore, we tried to strengthen our design by adopting
a retrospective cohort design, which provides a higher level of evidence for the association
between early antibiotic exposure and obesity compared with the case–control approach
employed by Azad et al,[35] which was lacking in the precision of the effect size (wide confidence interval),
albeit statistically significant. Although we adjusted for many potentially confounding
variables, variables such as GBS IAP and race or ethnicity had poor documentation,
which may have impacted our findings.
Conclusion
Neonatal antibiotic exposure (specifically ampicillin–gentamicin) was associated with
early childhood obesity at 5 years. However, we did not find a significant impact
from the duration of antibiotic exposure. Our findings lend credence to previous studies
on the effect of antibiotics on weight gain. Given the high rate of neonatal antibiotic
exposure in our cohort, cautious use of antibiotics in this critical period cannot
be overemphasized, especially among the vulnerable population in the South Bronx.
This will be an area of intervention that might have long-term implications for this
population. Finally, with the high incidence of obesity among our cohort compared
with the national average, augmenting current efforts at controlling and preventing
traditional obesity risk factors with cautious antibiotic use may reduce the prevalence
of childhood obesity and the attendant cardiovascular comorbidities.
