obesity reviews
doi: 10.1111/obr.12267
Review
Association between caesarean section and childhood obesity: a systematic review and meta-analysis S. Kuhle, O. S. Tong and C. G. Woolcott
Perinatal Epidemiology Research Unit,
Summary
Departments of Obstetrics and Gynaecology
Birth by caesarean section has been recently implicated in the aetiology of childhood obesity, but studies examining the association have varied with regard to their settings, designs, and adjustment for potential confounders. We conducted a systematic review and meta-analysis to summarize the available evidence and to explore study characteristics as sources of heterogeneity. A search of Medline, EMBASE, and Web of Science identified 28 studies. Random effects meta-analysis was used to calculate pooled risk ratios (RR) with 95% confidence intervals (CI). Caesarean section had a RR of 1.34 (CI 1.18–1.51) for obesity in the child compared with vaginal birth. The RR was lower for studies that adjusted for maternal pre-pregnancy weight than for studies that did not (1.29, CI 1.16–1.44 vs. 1.55, CI 1.11–2.17). Studies that examined multiple early life factors reported lower RRs than studies that specifically examined caesarean section (1.39, CI 1.23–1.57 vs. 1.23, CI 0.97–1.56). Effect estimates did not vary by child’s age at obesity assessment, study design or country income. Children born by caesarean section are at higher risk of developing obesity in childhood. Findings are limited by a moderate heterogeneity among studies and the potential for residual confounding and publication bias.
and Pediatrics, Dalhousie University, Halifax, NS, Canada
Received 5 November 2014; revised 7 January 2015; accepted 14 January 2015
Address for correspondence: Dr CG Woolcott, Perinatal Epidemiology Research Unit, IWK Health Centre, 5980 University Avenue, Halifax, NS B3K 6R8, Canada. E-mail:
[email protected] Keywords: Child, mode of delivery, mother, obesity. obesity reviews (2015)
Introduction The prevalence of childhood and adult obesity in Western countries has increased in the past several decades (1). Besides the poor lifestyle behaviours that adversely affect energy balance, perinatal factors such as maternal obesity, excess gestational weight gain and high birth weight have emerged as risk factors for offspring obesity (2–5). Another perinatal factor that has recently been implicated in the aetiology of childhood obesity is birth by caesarean section (6,7). It is speculated that because of differences in intrapartum bacterial exposure, infants born by caesarean section develop a gut microbiome that differs from that of infants born vaginally (8,9), which contributes to the development of obesity later in life (10). A number of studies have examined the relationship between delivery by caesar© 2015 World Obesity
ean section and childhood overweight and/or obesity, with conflicting results. Studies have varied with regard to their settings, designs and ability to adjust for potential confounders such as maternal pre-pregnancy body mass index (BMI) (4). The objective of the current study was to summarize the available evidence and to explore study characteristics as sources of heterogeneity in results of the association between caesarean section and childhood obesity through a systematic review and meta-analysis.
Methods Literature search A literature search was conducted using Medline, EMBASE and Web of Science spanning the inception of these 1
2
obesity reviews
Caesarean section and childhood obesity K. Stefan et al.
databases to July 2014 using subject headings and free-text keywords. A PubMed alert was set up to identify other potentially eligible studies published during the conduct of this research. The search strategy had a broad scope designed to capture not only studies in which caesarean section and its related terms could be found in searchable fields, but also all studies that assessed associations between early life factors and childhood obesity. The detailed search strategy can be found in the online supporting information. Reference lists from eligible studies were cross-checked to identify additional studies. No language restrictions were applied.
Study selection Studies were eligible for inclusion if they reported effect estimates for the association between the mode of delivery (caesarean section compared with vaginal delivery) and overweight or obesity in children as defined by the International Obesity Task Force (11), Centers for Disease Control (12) or World Health Organization reference (13), and derived from measured or reported BMI. The age range defining childhood was 2–18 years; studies that only presented results for children and adults combined were excluded. All human observational studies were considered for inclusion. Two reviewers (SK and OT) independently screened the titles and abstracts of the records retrieved by the literature search. The full-text of potentially relevant papers was retrieved and assessed independently by two reviewers (SK and OT). Disagreement over eligibility of a study was resolved by third-party adjudication.
Data extraction Two investigators (SK and CW) independently extracted data from the eligible studies using an electronic form. Unadjusted and adjusted effect estimates were recorded if they were reported; if not, unadjusted effect estimates were calculated if the data to do so were available. In addition, the following participant and study characteristics were recorded: bibliographic details, country, study design, sample size, main exposure of interest (caesarean section vs. early life factors in general), child’s age at BMI assessment, prevalence of childhood obesity, definition of overweight/obesity, caesarean section rate, and adjustment for maternal pre-pregnancy weight or BMI.
Statistical analysis The generic inverse variance method was used to determine the relative effects and their standard errors from the published findings (14). In the following, we will use the term ‘risk ratio’ (RR) to refer to the different effect measures (odds ratio, prevalence ratio, relative risk) reported in the
included studies. Random effects meta-analysis was conducted to estimate the pooled RR for the risk of obesity or overweight/obesity in offspring delivered by caesarean section. The primary outcome was childhood obesity; secondary outcomes included overweight and obesity combined, and overweight. We performed sensitivity analyses for the primary and secondary outcomes after excluding studies in which the outcome prevalence in the analysis sample was >10% as odds ratios will overestimate the RR when the rare disease assumption is not met. In studies where several multivariable models were presented, the estimate from the most comprehensively adjusted model was used. As some studies assessed the association at multiple ages, we used the median RR in the main analysis and conducted sensitivity analyses using the minimum and maximum RR. For studies that presented only unadjusted estimates, the univariate RR was used in lieu of the adjusted RR. Where possible, the unadjusted risk estimate was calculated as a relative risk (as opposed to an odds ratio) from the cross-tabulations provided in the publication. Statistical heterogeneity was assessed using Cochrane’s Q statistic and the I2 statistic. I2 percentages were interpreted as follows: 0% to 40% may not be important; 30% to 60% may represent moderate heterogeneity; 50% to 90% may represent substantial heterogeneity; 75% to 100% represents considerable heterogeneity (14). The presence of potential publication bias was examined visually using a funnel plot and statistically using Egger’s test (15). To explore potential sources of heterogeneity, we conducted subgroup analyses for study design (cohort vs. case control or cross-sectional), adjustment for maternal prepregnancy BMI (yes vs. no), age at BMI assessment (10%, the pooled effect estimate decreased slightly (RR 1.13, 95% CI 1.04–1.23). Pooled RR estimates from sensitivity analyses using the minimum or maximum reported effect estimates from studies in which children were assessed at multiple ages were comparable (RR 1.12, 95% CI 1.03–1.23, and RR 1.17, 95% CI 1.06–1.30, respectively). The pooled RR from six studies for the outcome overweight (excluding obesity) was 1.16 (95% CI 1.06–1.27)
Cohort – Cohort Cohort – Cross-sectional – Cross-sectional – Cohort – Cross-sectional – Cohort Cohort Cohort Cohort Cohort Case control Cohort – Cohort Cross-sectional Cohort Case control Cohort – Cohort Cohort – Cross-sectional –
Brazil – Brazil
Canada – Australia – Australia – United States – Greece – United Kingdom Brazil
Brazil
China Hong Kong Sri Lanka United States – United Kingdom Poland Canada Europe United States – United States Germany – Iran –
181,380 7,809 142 917 – 13,518 812 614 1,024 1,843 – 436 1,198 1,574 635 –
673
2,988 – 1,746 – 1,339 – 1,237 – 2,294 – 5,811 790
1,237 4,349 3,756
6,828 359 28,354 1,687 162 4,742
n*
4.9 8.0 4.0 11.9 – 3.0 8.7 9.0 6.0 3.5 – 7.0 6.0 2.0 9.0 –
8.0
10.5 – 12.0 – 6.0 – 3.0 – 11.0 – 9.0 10.5
4.0 15.0 4.0
3.5 4.5 7.0 8.0 4.5 4.0
Mean child age (years)
21.3 26.2 NA 20.3 – 9.9 21.4 14.5 29.6 33.7 – 22.7 17.0 17.0 64.6 –
31.6
18.9 – 20.9 – 22.8 – 22.6 – 28.7 – 9.1 55.3
30.5 30.5 45.4
NA 8.5 12.4 8.6 11.1 27.2
C-section rate (%)
IOTF IOTF WHO CDC CDC IOTF IOTF CDC IOTF CDC CDC CDC WHO WHO WHO WHO
CDC
IOTF IOTF IOTF IOTF IOTF IOTF CDC CDC IOTF IOTF CDC CDC
WHO Other WHO
WHO CDC IOTF IOTF IOTF WHO
BMI reference
OW/OB OW/OB OB OB OW/OB OW/OB OW/OB OW/OB OB OB OW/OB OB OB OW OB OW/OB
OB
OB OW OB OW OB OW OB OW OB OW OW/OB OB
OB OW/OB OB
OB OW/OB OW/OB OW/OB OB OB
Outcome
6.6 NA 7.4 18.5 34.0 23.6 14.4 29.4 6.4 8.7 18.9 25.2 2.2 16.2 17.6 26.6
2.2
9.8 23.1 8.2 20.4 5.9 12.9 9.4 17.2 11.6 30.5 20.6 13.2
11.3 8.7 12.4
3.2 22.6 1.3 13.9 3.5 7.9
Outcome prevalence (%)
OR OR OR OR OR Rel. Rel. Rel. OR OR OR Rel. OR Rel. OR Rel.
OR
risk
risk
risk
risk risk risk
OR OR OR OR OR OR OR OR Rel. risk Rel. risk OR OR
PR PR PR
OR OR OR Rel. risk OR PR
Effect measure (1.19–1.83) (1.10–5.62) (0.82–1.24) (1.10–2.19)† (1.24–22.04) (0.97–1.53)
(0.85–1.67) (1.03–1.64) (0.70–1.77) (0.89–1.57) (1.34–4.00) (0.77–1.69) (1.36–3.23) (0.86–1.77) (0.76–1.26)† (0.99–1.26)† (1.01–2.20) (1.04–2.92)
1.13 0.98 0.70 1.87 1.86 1.04 0.85 1.13 1.17 0.88 0.92 1.46 1.49 1.17 2.45 1.49
(1.08–1.18) (0.77–1.25) (0.25–2.02) (1.19–2.95) (1.27–2.73) (0.94–1.15)† (0.55–1.31)† (0.82–1.57)† (0.87–1.57) (0.61–1.27) (0.71–1.19) (1.08–1.98) (0.55–4.05) (0.88–1.56)† (1.03–5.84) (1.09–2.03)†
1.66 (0.44–6.33)
1.19 1.29 1.11 1.18 2.32 1.14 2.10 1.24 0.98 1.11 1.49 1.74
1.31 (0.90–1.92) 1.23 (0.98–1.54) 1.21 (0.99–1.48)
1.47 2.49 1.01 1.56 5.23 1.22
Effect estimate
Yes No No No No No No No Yes Yes Yes Yes Yes Yes No No
Yes
Yes Yes No No No No Yes Yes No No Yes No
Yes Yes Yes
Yes No Yes No No Yes
Maternal weight adj.
Caesarean section and childhood obesity K. Stefan et al.
*Study sizes represent the effective sample size used in the analyses. † Effect estimate not adjusted. adj., adjusted; BMI, body mass index; CDC, Centers for Disease Control; C-section, caesarean section; IOTF, International Obesity Task Force; OB, obesity; OR, odds ratio; OW, overweight; PR, prevalence ratio; Rel. risk, relative risk; WHO, World Health Organization.
Case control Cohort Cohort Cohort Case control Cohort
China United States Denmark Netherlands China Brazil
Li 2008 (30) Rooney 2010 (25) Ajslev 2011 (38) Steur 2011 (23) Zhou 2011 (19) Barros 2012 (35) (1982 cohort) Barros 2012 (35) (1993 cohort) Barros 2012 (35) (2004 cohort) Flemming 2012 (39) – Gopinath 2012 (31) (12-year olds) Gopinath 2012 (31) (6-year olds) Huh 2012 (7) – Birbilis 2013 (34) – Blustein 2013 (33) Goldani 2013 (32) (High SES cohort) Goldani 2013 (32) (Low SES cohort) Li 2013 (28) Lin 2013 (29) Rathnayake 2013 (26) Wang 2013 (22) – Weng 2013 (21) Zadzinska 2013 (20) Azad 2014 (37) Bammann 2014 (36) Costantine 2014 (40) – Mueller 2014 (18) Pei 2014 (27) – Salehi-Abargouei 2014 (24) –
Design
Country or region
Study
Table 1 Characteristics of studies that examined the association between caesarean delivery and offspring weight status
4
obesity reviews
© 2015 World Obesity
obesity reviews
Caesarean section and childhood obesity K. Stefan et al.
n
RR (95% CI)
Zhou 2011
162
5.23 (1.24, 22.05)
Flemming 2012
2988
1.19 (0.85, 1.67)
Gopinath 2012 (12−year olds)
1746
1.11 (0.70, 1.77)
Gopinath 2012 (6−year olds)
1339
2.32 (1.34, 4.01)
Huh 2012
1237
2.10 (1.36, 3.24)
Birbilis 2013
2294
0.98 (0.76, 1.26)
Bammann 2014
1024
1.17 (0.87, 1.57)
Pei 2014
1198
1.49 (0.55, 4.04)
Study
5
− Reference: Normal weight
1.40 (1.08, 1.80)
Subtotal − Reference: Normal weight/overweight Li 2008
6828
1.47 (1.19, 1.82)
Barros 2012 (1982 cohort)
4742
1.22 (0.97, 1.53)
Barros 2012 (1993 cohort)
1237
1.31 (0.90, 1.91)
Barros 2012 (2004 cohort)
3756
1.21 (0.99, 1.48)
Goldani 2013 (High SES cohort)
790
1.74 (1.04, 2.92)
Goldani 2013 (Low SES cohort)
673
1.66 (0.44, 6.30)
Rathnayake 2013
142
0.70 (0.25, 2.00)
Wang 2013
917
1.87 (1.19, 2.94)
Costantine 2014
1843
0.88 (0.61, 1.27)
Mueller 2014
436
1.46 (1.08, 1.98)
Salehi−Abargouei 2014
635
2.45 (1.03, 5.83)
Subtotal
1.33 (1.17, 1.51)
Overall
1.34 (1.18, 1.51)
.5
1
2
5
Figure 2 Forest plot of studies examining the association between caesarean section (compared with vaginal delivery) and offspring obesity. CI, confidence interval; RR, risk ratio; SES, socioeconomic status.
for caesarean section compared with vaginal birth. The I2 statistic (0%) indicated no heterogeneity. All but two studies (27,34) reported odds ratios in samples where the overweight prevalence was >10% and the pooled RR was 1.12 (95% CI 1.00–1.26). Pooled RR estimates from sensitivity analyses using the minimum or maximum reported effect estimates from studies with multiple BMI assessments were comparable (RR 1.16, 95% CI 1.05–1.28, and RR 1.17, 95% CI 1.07–1.29, respectively). Subgroup analyses Results from the subgroup analyses are shown in Table 2. The pooled RR for studies that adjusted for maternal prepregnancy weight was lower than the pooled RR for studies that did not adjust for maternal pre-pregnancy weight (RR 1.29, 95% CI 1.16–1.44 vs. RR 1.55, 95% CI 1.11–2.17), but this difference was not statistically significant. Studies that adjusted their multivariable models for maternal prepregnancy weight and provided both the unadjusted and © 2015 World Obesity
adjusted estimates showed a drop in the pooled unadjusted RR from 1.57 (95% CI 1.43–1.71) to 1.29 (95% CI 1.16– 1.43) after adjustment, corresponding to an 18% change in the effect estimate. No appreciable difference in the pooled RR was observed between: studies assessing BMI at less than 6 years compared with 6 years and older (RR 1.31, 95% CI 1.08–1.57 vs. RR 1.38, 95% CI 1.16–1.64); cohort studies compared with case control and cross-sectional studies (RR 1.34, 95% CI 1.17–1.53 vs. RR 1.36, 95% CI 1.06–1.75); and studies conducted in high-income compared with middle-income countries (RR 1.33, 95% CI 1.08–1.62 vs. RR 1.35, 95% CI 1.18–1.53). Studies where the caesarean section was the main exposure of interest reported slightly higher effect estimates than studies that examined multiple early life factors in general (RR 1.39, 95% CI 1.23–1.57 vs. RR 1.23, 95% CI 0.97–1.56), and the pooled effect estimate for the latter group was not statistically significant. Studies with a caesarean section rate of 30% or higher had
6
obesity reviews
Caesarean section and childhood obesity K. Stefan et al.
Subgroup
Adjusted for maternal pre-pregnancy weight Yes No Age at BMI assessment (years)