The association between caesarean section delivery and later life obesity in 21-24 year olds in an Urban South African birth cohort.

BACKGROUND: Obesity is an important public health problem and rates have reached epidemic proportions in many countries. Studies have explored the association between infants delivered by caesarean section and their later life risk of obesity, in many countries outside Africa. As a result of the increasing caesarean section and obesity rates in South Africa, we investigated the association in this country.
METHODS: This was a retrospective analysis of data that were collected from a prospective South African birth cohort (Birth to Twenty Plus), established in 1990. A total of 889 young adults aged 21-24 years were included in the analysis. Poisson regression models were fitted to assess the association between mode of delivery and early adulthood obesity.
RESULTS: Of the 889 young adults, 106 (11.9%) were obese while 72 (8.1%) were delivered by caesarean section; of which 14 (19.4%) were obese. Caesarean section delivery was significantly associated with obesity in young adults after adjusting for potential confounders like young adults' sex and birth weight, mothers' parity, and education (incidence rate ratio 1.64, 95% CI 1.01-2.68, p = 0.045).
CONCLUSION: The association of caesarean section with early adulthood obesity should be interpreted with caution because data on certain key confounding factors such as mothers' pre-pregnancy body mass index and gestational diabetes were not available. Further research from Africa, with larger sample sizes and databases with useful linking of maternal and infant data, should be conducted.

Introduction

Obesity is an important public health problem and rates have reached epidemic proportions in many countries—one in every five young people was estimated to be obese in 2012 in high and middle income countries [1]. Globally, 39% and 13% of individuals aged 18 years and older were overweight and obese, respectively, in 2016 [2]. In South Africa, the rates of overweight or obesity in 2016 were higher among women (61% in women and 31% in men) [3]. Obesity has been associated with adverse health outcomes such as type 2 diabetes, cardiovascular disease, cancer, and premature mortality, in both adults and children [4-8]. Diet (high energy foods), physical inactivity, birth weight, genetics, and parity are commonly explored risk factors of obesity [9-11]. Mode of delivery at birth has also been suggested to be associated with obesity in later life.

Due to the rising rates of obesity, any hypothesized risk factor, such as caesarean section (CS) as a mode of child delivery, is worth exploring. Acknowledging that CSs are sometimes performed to prevent birth complications, biased motives have been identified [12,13] and the use of the procedure is increasing globally [14]. As many as one in three births was reported to be by CS in high-income countries such as the United States of America (USA) in 2017 [15]. The rates of CS in South Africa in 2014, reported in the 2016 South African Health Review, were estimated to be 70.8% in the private health sector, and 24.7% in the public health sector (as reported by the District Health Information System (DHIS) which uses routinely collected health information to manage health services) [16]. An underlying mechanism for the proposed association between CS and later life obesity is the reported limited microbial diversity of offspring delivered by CS [17,18]. This is presumed to persist to adulthood [19].

A number of studies have explored the association between CS and later life obesity. Three systematic reviews and meta-analyses reported increased pooled effect size estimates in young adults (YAs) delivered by CS (such individuals were more likely to be obese) [20-22]. Five additional studies reported similar findings [19,23-26]. However, some studies did not find significant associations, leading to inconsistent findings [27,28]. All previous studies have been conducted in countries outside Africa [19,23-35] and, to our knowledge, nothing is known in the African context. The aim of the study reported in this paper was to investigate the association between CS delivery and early adulthood obesity among singletons in an ongoing longitudinal birth cohort (Birth to Twenty Plus–Bt20+) with more than two decades of follow-up in an urban region in South Africa [36].

Materials and methods

Study design and setting, exposure and outcome variables and potential confounders

This was a retrospective analysis of data that were collected from a prospective cohort study established in 1990. Birth to Twenty Plus (formerly Birth to Ten) is an ongoing prospective South African birth cohort established in Soweto, Johannesburg, Gauteng in 1990 [36]. The cohort comprises 3 273 singleton children of mothers who were recruited from antenatal clinics and had an expected delivery date from 23 April to 8 June 1990. Consent to participate in the study was provided by the mothers at enrolment; all subsequent data were also collected with signed consent. Participants have been followed up through administered questionnaires, contact with parents or caregivers, telephone calls, and field visits. Further information on the cohort has been published elsewhere [36,37].

Information on mode of delivery was copied by the investigators of the Bt20+ cohort from the official birth notification forms compiled at the local authority. Information on CS and vaginal deliveries, i.e. normal (NVD) and assisted (AVD—forceps and vacuum) was available for this study. A digital scale and wall mounted stadiometer were used to measure weight and standing height, to the nearest 0.1 kg and 0.1 cm, respectively. These measurements were taken by trained research personnel. Body Mass Index (BMI) was calculated using the formula weight/height2 (kg/m2). We defined obesity (BMI ≥30 kg/m2), as per the World Health Organization (WHO) [38].

We reviewed published literature on factors associated with obesity and mode of delivery, and identified variables of interest from similar studies. Young adults’ characteristics included gestational age (at delivery), sex, ethnicity, age, education, alcohol consumption, cigarette smoking, birth weight (low (<2.5 kg), normal (2.5–4.0 kg), macrosomic (>4 kg)), and breastfeeding duration. Mothers’ characteristics included parity, age at delivery, and education. We generated infants’ birth weight corrected for gestational age and sex (in centiles) by using the INTERGROWTH-21st calculator to compare these variables with an international standard [39]. Participant’s age was calculated as the difference between date of birth and date of data collection.

Statistical analysis

We assessed the differences in participants’ characteristics (early life, young adult and maternal) across the modes of delivery and BMI categories, and compared the sex-stratified prevalence of obesity for each mode of delivery. With a non-rare outcome (obesity) with prevalence of 11.8% for our overall sample, we evaluated the association between CS and early adulthood obesity using Poisson regression models, with robust standard errors. Incidence rate ratio (IRR) and 95% CIs were computed. Although, lifestyle and behavioural characteristics, such as YAs’ diet, physical activity, smoking habits, and alcohol consumption, have been associated with obesity, it has been suggested that they are not true confounders in the analysis of the association between mode of delivery and later life obesity [19,40]. Young adults’ sex and birth weight in kg, mothers’ parity, and education at YA’s birth were included in the adjusted models. Although, ethnicity was associated with both the mode of delivery and BMI, it was not included in the analysis. This is because including it did not significantly change the effect estimate, in our final regression model.

Furthermore, we conducted a post-estimation (Wald) test to investigate the differences between obesity rates across sex categories (heterogeneity). This was done by introducing cross product (interaction) terms between mode of delivery and YAs’ sex in the final adjusted model. Sex stratified models were then computed. In addition, we analysed the combined VD data to interrogate the potential effect of differential exposure of infants to maternal vaginal and faecal microflora, implicated in the subsequent genesis of childhood obesity, between those born by CS and VD.

Missing data

In the primary cohort, both BMI (outcome variable) and mode of delivery (primary predictor variable) had missing data of 51.6% and 49.1%, respectively. This was mainly because of loss to follow up and we excluded these individuals from our study. We however, estimated the difference between the primary cohort and our study participants by comparing socio-demographic characteristics between the two groups. Kruskal-Wallis test, Pearson’s Chi-square test, and Fisher’s exact test were used, where appropriate. Some of the covariates in our final analytic sample had missing values. These variables were young adults’ alcohol intake (7.3%), education (1.1%), and birth weight (0.1%); mother’s post-school education (3.5%). We assumed that covariates with incomplete data were missing at random, so the probability of a participant having missing data in these covariates depends on other observed data in the analysis and not on the values of the missing data itself or on unmeasured variables. Multiple imputation, using multivariate normal imputation (MVNI) was performed in Stata to impute missing data for these covariates in the final regression model. The assumption that all variables in the imputation model jointly follow a multivariate normal distribution was not plausible, due to the presence of categorical variables in the model. However, it has been suggested that reasonable inference can still be drawn, even if the assumption of multivariate normality is not held, as this has been demonstrated in different studies [41-43].

The imputation models included variables highlighted for Poisson regression model adjustment, under the statistical analysis section. Twenty imputations were performed and results of the analyses were pooled using Rubin’s rules [44]. We compared the observed and imputed data to observe any difference in the mean in both groups. Apart from birth weight (kg), which was a normally distributed continuous variable, all other covariates were in count or categorized form.

Sensitivity analysis

To investigate the robustness of the inference from the imputed regression analysis and also to address the possibility of residual confounding, sensitivity analysis was conducted. Complete case analysis was performed for regression models–individuals with missing data before imputation were excluded. Models adjusting exclusively for early life factors were computed, as well as those adjusting for lifestyle and behavioural characteristics. We also adjusted for YAs’ birth weight as a continuous covariate, and YA’s breastfeeding duration at infancy, mother’s parity and mother’s age at delivery as categorical covariates, in subsequent analyses. P values <0.05 were considered statistically significant. All analyses were conducted using Stata® IC 14 (StataCorp LP College Station, TX).

Ethical considerations

Prior to analysing the data, we obtained a written ethical approval from the Human Research Ethics Committee of the University of the Witwatersrand (clearance certificate number M161184). Anonymised data were received from the gatekeeper of the primary study and a data sharing agreement was signed to ensure confidentiality and to limit access to the data.

Results

The study participants comprised 889 21–24 year old YAs after excluding those in the primary cohort with missing data; one cohort member was excluded for being older than 24 (Fig 1). Of the final analytic sample, 72 (8.1%) were delivered by CS of which 14 (19.4%) were obese. The means of the observed and imputed data were similar, hence multiple imputation provided appropriate data for analysis.

Fig 1

Flowchart showing the study population and selection of study participants.

To examine the differences in the primary dataset and study participants, Tables 1 and 2 compare the variables in both cohorts. There were significant differences in the distribution of certain characteristics between the primary cohort and study participants, with regards to the YAs’ ethnicity, breastfeeding duration at infancy, and smoking, as well as mothers’ education, gestational age, and age at YAs birth (P<0.05). With no post school education in both groups, although the rate was higher among the study participants.

Table 1

Comparison between primary cohort and study participants (categorical variables).

Variables	Primary cohort		Study participants		P value
	N = 3273		N = 889
	n	%	n	%
Young adult characteristics
Sex					0.511
Male	1594	48.7	444	49.9
Female	1679	51.3	445	50.1
Ethnicity					0.000
Black	2568	78.5	803	90.3
Others	705	21.5	87	9.7
Mode of delivery					0.473
NVD	1470	44.9	793	89.2
AVD	39	1.2	24	2.7
CS	157	4.8	72	8.1
Unknown	1607	49.9	-
Smoking					0.000
No	2640	80.7	530	59.6
Yes	633	19.3	359	40.4
Alcohol intake					0.845
No	1004	30.7	560	63.0
Yes	482	14.7	264	29.7
Unknown	1787	54.6	65	7.3
Young adult education					0.895
<grade12	625	16.8	349	39.3
≥ grade 12	960	29.3	531	59.7
Unknown	1688	51.6	10	1.1
Early life and maternal characteristics
Mother’s post-school education					0.004
No	2604	79.6	792	92.2
Yes	328	10.0	67	7.8
Unknown	341	10.4	30	3.4

NVD–normal vaginal delivery, AVD—assisted vaginal delivery, CS—caesarean section

P values approaching or <0.05 shows a difference between the cohorts; Pearson’s Chi-square

Missing observations were excluded from the inferential statistics i.e. P value estimation

Table 2

Comparison between primary cohort and study participants (continuous variables).

Variables	Primary cohort		Study participants		P value
	N = 3273		N = 889
	Med	IQR	Med	IQR
Young adults age at visit (years)	23	22–23	23	22–23	0.402
Young adults BMI (kg/m2)	22.1	19.7–26.1	22.2	19.9–26.3	0.570
Unknown n(%)	1728(52.8)		-
Mothers gestational age (weeks)	38.0	38.0–40.0	38.0	38.0–39.0	0.000
Unknown n(%)	102(3.1)		-
Young adults birth weight	3.1	2.8–3.4	3.1	2.8–3.4	0.711
Unknown n(%)	6(0.2)		1(0.1)
Mother’s parity at YA birth	2.0	1.0–3.0	2.0	1.0–3.0	0.147
Young adult breastfeeding duration at infancy (months)	3.5	0.0–17.0	8.0	1.0–20.0	0.000
Mother’s age at YA delivery (years)	25.0	21.0–30.0	24.0	20.0–30.0	0.002
Unknown n(%)	2(0.1)		-

YA—young adult, Med—median, IQR—interquartile range

P values approaching or <0.05 shows a difference between the cohorts; Wilcoxon rank-sum test

Missing observations were excluded from the inferential statistics i.e. P value estimation

Some of the individual distributions across the categories of these covariates followed similar trends between the primary cohort and study participants, although p values show differences statistically (P<0.05). This is because high proportions of Blacks (YA) and non-smokers (YA) were observed in both groups, but they were higher among study participants. High proportions of mothers with no post school education were observed in both groups, and this was also higher among the study participants.

The study participants’ characteristics by mode of delivery category are presented in Tables 3 and 4. As seen in Table 3, CSs were more frequent than NVDs in those who had low birth weight or macrosomia, and in Indians and Coloureds (non-Black participants).

Table 3

Socio-demographic characteristics by mode of delivery.

	Mode of delivery
	Total		NVD		AVD		CS		P value
	N = 889		n = 793		n = 24		n = 72
	n	%	n	%	n	%	n	%
Young adult characteristic
Sex									0.890
Male	444	49.9	398	50.2	11	45.8	35	48.6
Female	445	50.1	395	49.8	13	54.2	37	51.4
Ethnicity									<0.001
Black	803	90.3	727	91.7	17	70.8	59	81.9
Non-Black	86	9.7	66	8.3	7	29.2	13	18.1
Alcohol intake									0.617
No	560	63.0	502	63.3	16	66.7	42	58.3
Yes	264	29.7	231	29.1	8	33.3	25	34.7
Unknown	65	7.3	60	7.6	0	-	5	6.9
Smoking									0.479
No	530	59.6	472	59.5	12	50.0	47	59.5
Yes	359	40.4	321	40.5	12	50.0	26	40.5
Education									0.304
<grade 12	349	39.3	305	38.5	12	50.0	32	44.4
≥ grade 12	530	59.6	479	60.4	11	45.8	40	55.6
Unknown	10	1.1	9	1.1	1	4.2	0	-
Early life and maternalCharacteristic
Mother’s post-school education									0.187 f
No	792	92.2	709	89.4	23	95.8	59	81.9
Yes	67	7.8	59	7.4	0	0.0	8	11.1
Unknown	30	3.4	25	3.2	1	4.2	5	6.9
Birth weight (kg)									0.018 f
LBW (<2.5)	74	8.3	63	7.9	4	16.7	7	9.7
Normal (2.5–4.0)	798	89.8	718	90.5	20	83.3	60	83.3
Macrosomia (>4)	16	1.8	11	1.4	0	-	5	6.9
Unknown	1	0.1	1	0.1	0	-	0	-
Birth weight (centile)
SGA (<10th)	79	8.9	68	8.6	5	20.8	6	8.3	0.139 f
AGA (≥10th-≤90th)	680	76.5	610	76.9	18	75.0	53	72.2
LGA (>90th)	129	14.5	114	14.4	1	4.2	14	19.4
Unknown	1	0.1	1	0.1	0	-	0	-

Non-black–Indians and Coloured, LBW–low birth weight, SGA–small for gestational age, AGA–appropriate for gestational age, LGA–large for gestational age, NVD/AVD–Normal/Assisted vaginal delivery

Fisher’s exact (f)

Table 4

Maternal characteristics of the study participants by mode of delivery.

	TotalN = 889		NVDn = 793		AVDn = 24		CSn = 72		P value
	Med	IQR	Med	IQR	Med	IQR	Med	IQR
Gestational age (weeks) at delivery	38.0	38.0–39.0	38.0	38.0–39.0	38.0	38.0–40.0	38.0	37.5–39.0	0.074
Parity at delivery	2.0	1.0–3.0	2.0	1.0–3.0	1.0	1.0–3.0	2.0	1.0–3.0	0.136
Breastfeeding duration (months)	8.0	1.0–20.0	9.0	1.0–20.0	6.5	2.5–24.0	6.0	0.6–18.0	0.489
Age at delivery (years)	24.0	20.0–30.0	24.0	20.0–30.0	23.5	19.0–30.0	25.5	21.0–30.0	0.541

Med–median, IQR–interquartile range, NVD/AVD–Normal/Assisted vaginal delivery, CS–caesarean section

No statistically significant differences were observed for gestational age, parity, breastfeeding duration or age at delivery (Table 4).

Compared to the primary cohort, the study participants were more likely to be Black (78.5% vs 90.3%), smokers (19.3% vs 40.4%), have mothers who had no post school education (79.6% vs 92.2%). They were also breast fed for a longer period, and had mothers who had longer gestational age and were younger when they gave birth, than those in the primary cohort. There were no differences between YAs sex, mode of delivery, education, alcohol intake, BMI and birth weight, and mothers’ parity, between the primary cohort and the study participants (p>0.05). Certain characteristics were associated with BMI, such as YA’s sex (p = <0.001), ethnicity (p = 0.039), smoking habit (p = <0.001), and education (p = <0.001); and mothers’ parity at delivery (p = 0.019) (supplementary–S1 and S2 Tables).

Sex-stratified prevalence of obesity are presented in Table 5. Overall, the prevalence of obesity was higher in CS-delivered participants than in those delivered through either AVD or NVD, but there were no differences when the results were stratified by sex.

Table 5

Prevalence of obesity in each mode of delivery category among study participants, by sex.

	N	n	%	95% CI	P value
Overall					0.042
NVD	793	88	11.1	9.1–13.5
AVD	24	4	16.7	6.4–37.0
CS	72	14	19.4	11.9–30.2
Total	889	106	11.9	9.9–14.2
Male					0.190
NVD	398	9	2.3	1.2–4.3
AVD	11	1	9.1	1.3–44.0
CS	35	3	8.6	2.8–23.5
Total	444	13	2.9	1.7–5.0
Female					0.133
NVD	395	79	20.0	16.3–24.2
AVD	13	3	23.1	7.6–52.2
CS	37	11	29.7	16.8–46.2
Total	445	93	20.9	17.4–24.9

CI–confidence interval, VD–vaginal delivery, CS–caesarean section

As seen in Table 6, birth by CS was associated with an almost two-fold increase in the risk of obesity among YAs aged 21–24 years, in the crude analysis. After adjusting for YAs’ sex and birth weight, and mothers’ parity and education at delivery as potential confounders, a similar estimate was observed.

Table 6

The association between mode of delivery and early adulthood obesity, stratified by sex and birth weight (using normal BMI as reference)-Imputed models.

Variable	IRR	95%CI	P value	AdjIRR	95%CI	P value
Main analysis
crude
NVD	1.00	reference		1.00	reference
AVD	1.50	0.60–3.76	0.384	1.41	0.57–3.49	0.460
CS	1.75	1.05–2.92	0.031	1.64	1.01–2.68	0.045
Male
NVD	1.00			1.00	reference
AVD	4.02	0.56–29.10	0.168	4.90	0.65–37.17	0.124
CS	3.79	1.07–13.38	0.038	4.01	1.14–14.09	0.031
Female
NVD	1.00			1.00	reference
AVD	1.15	0.42–3.18	0.782	1.12	0.41–3.07	0.717
CS	1.49	0.87–2.54	0.146	1.44	0.85–2.44	0.173

N = 889; Poisson regression

OR–odds ratio, CI–confidence interval, NVD/AVD–Normal/Assisted vaginal delivery, CS–caesarean section. Adjusted for YAs’ sex and birth weight; mothers’ parity and education at YA’s birth in all models.

Although no interaction was found between mode of delivery and either in the final adjusted models, obesity rates were statistically different across the sex (p = <0.001) categories, indicating heterogeneity (see supplementary section, S1 Stata code).

The results of the complete case analysis for regression models are presented in S3 Table in the supplementary information–conclusions were unchanged from these analyses. To address the possibility of residual confounding and to determine if the results held up under different scenarios, a series of sensitivity analyses were conducted and results are presented in S4 Table in the supplementary information. The risk of being obese if delivered by CS were 1.62 times the risk of being obese if delivered by AVD or NVD. The additional adjustment for YA’s breastfeeding duration, smoking habit and alcohol intake, as well as treating certain covariates as linear or categorical in the sensitivity analyses, did not change the conclusions reached from the results of the original analyses. The observed association was not statistically significant when only early life factors (YAs’ sex, birth weight and breastfeeding duration; and mothers’ parity, gestational age, age and education at delivery) were adjusted for.

Discussion

Findings and evidence from previous studies

In this analysis of 21–24 year old South Africans from a longitudinal urban birth cohort, CS was associated with obesity in early adulthood. Among males, CS was statistically associated with an increased risk of obesity. The magnitude of the association among females is of particular clinical concern. The confidence intervals of the IRRs in the stratified analyses were wide, indicating low precision, due to the small sample size. Most of the sensitivity analyses supported the finding of an association between CS and early adulthood obesity. A marginal statistical significance was observed when early life and maternal characteristics were adjusted for, regardless of their collinearity or correlation with other variables and their effect on the primary predictor.

Significant associations between CS and obesity in early adulthood have been previously reported, despite the inconsistencies in some studies [27,28]. In a systematic review and meta-analysis by Darmasseelane et al. (2014), comprising studies published prior to 31 March 2012, the odds of being obese among CS-delivered YAs (aged 18 years or older) from 11 studies, was 22% higher than the odds of being obese among YAs delivered vaginally (OR 1.22, 95% CI 1.05–1.42) [20]. Results from a meta-analysis by Li et al. (2013) support this association among YAs aged 19 or older (OR 1.50, 95% CI 1.02–2.20) [21]. The studies included in these reviews were conducted independently in different countries (Finland, England, Scotland, Sweden, Brazil, Netherlands, China, India, and Denmark). Results from a study by Mesquita et al. (2013) showed an increase in fat accumulation due to delivery by CS: measured by waist-hip ratio (IRR 1.45, 95% CI 1.18–1.79 [23]. More recently, Yuan et al. (2016) and Barros et al. (2017) reported increased risks of obesity in CS-delivered YAs (RR 1.15, 95% CI 1.06–1.26, n = 8 486; β coefficient 0.15, 95% CI 0.08–0.23, n = 1 794, respectively) in Boston (USA) and Pelotas (Brazil) [19,26]. Additionally, Hansen et al. 2018 reported over a two-fold increase in odds of overweight or obesity in CS–delivered at age 20 years (OR 2.17, 95% CI 1.10, 4.27), after adjustment for potential confounders. This is similar to the risk observed in our study [25].

The aforementioned studies were all conducted outside Africa. However, some of the study sites were in upper-middle income countries, such as Brazil and China which are similar, socio-economically, to South Africa, and hence their results are comparable to those from our study. Also, findings from South Africa, regarding certain established potential confounders, such as socioeconomic status (SES), are different from those reported in high income countries. For example, while most high income countries have reported that high SES leads to higher body mass index, those conducted in South Africa have found that low SES was associated with higher body mass index [45,46].

Underlying mechanism

It has been suggested that infants born via VD are exposed mainly to microorganisms in the birth canal or vaginal environment, and that those delivered by CS are exposed to micro flora on their mother’s skin [47-49]. The abundance of Coprococcus and Ruminococcus of the family Lachnospiraceae was also reported by Tun et al. (2018) in infants delivered through CS–signifying dysbiosis in early life [48]. The investigators further explained the impact of family Lachnospiraceae in promoting adiposity.

Differences in intestinal bacterial colonisation; such as certain Bifidobacteria spp. that contribute to digestion and infant intestine development have been reported to be absent in infants delivered by CS [50]. Kalliomäki et al. (2008) reported that, compared to those of normal BMI, overweight seven-year old participants had lower Bifidobacteria counts at age one and six years [51]. In addition, infants delivered by CS had almost no Bifidobacteria spp. in their faecal samples in studies by Huurre et al. (2008) and Biasucci et al. (2008) [52,53]. Nonetheless, it is possible that the underlying reason or indication for CS delivery could be a cause of later life obesity. This is because certain clinical states resulting in CS, as well as antibiotics administered during CS, have been suggested to increase the tendency of offspring obesity [40,54,55].

Limitations and strengths

A limitation of this study is that the exposure and outcome assessments were dependent on the accuracy of the data collected from the Bt20+ cohort. In addition, the current surge of CS in South Africa was not reflected in the results; only 72 (8.1%) women gave birth through CS. Consequently, the 95% CIs were wide, indicating low precision of the ORs. We did not have the relevant data to examine the proposed mechanism underlying the association between CS birth and obesity in later life (i.e. deprivation of CS-born infants of microorganisms essential for regulating digestion). Caesarean delivery was not recorded as being elective or an emergency. It has been proposed that differences in intestine microbiota composition might arise due to prolonged delivery or ruptured fetal membranes [30,56], but this could not be investigated. In addition, the proposed effect (obesity) of antibiotics administered during CS on offsprings delivered through the procedure, could not be assessed due to lack of data.

Management practices in labour and delivery units differ across hospitals and has been reported by Plough et al. (2017) to be associated with risk of primary cesarean delivery [57]. The differences in CS rate between public and private hospital could not be explored, as all participants were recruited from public hospital. Finally, key pre-pregnancy information was not available, e.g. mothers’ BMI (height and weight); gestational diabetes, preeclampsia, or pregnancy-induced hypertension; smoking habits; information about previous CS; and family income. The absence of these variables might have resulted in residual confounding, leading to higher point estimates than observed in previous studies.

The strength of this analysis was the availability of data to estimate associations between CS and obesity in later life. We were also able to demonstrate temporality as the exposure (CS) preceded the outcome (obesity in early adulthood). Additional strengths of the study were the availability of information on important early life factors of YAs, and prospectively collected data in the cohort. Being able to provide different estimates for normal versus assisted VD is a particular strength of this study. Infant umbilical vein cortisol and other stress molecule concentrations differ significantly between those born by normal and assisted VD, which might influence longer term offspring energy metabolism [58]. Also, higher birth weight, which has been reported to be associated with CS birth, is a risk factor for instrumental VD [59]. Finally, the outcome measure (BMI) and primary predictor (mode of delivery) does not differ between our study participants and primary cohort. Hence, any major effect of selection bias was less likely.

Conclusions

Caesarean section as a mode of delivery was statistically associated with obesity in the study participants. Further research is required in South Africa, and Africa in general, using routinely collected data that provide useful linking of maternity data with information in other databases. This will help to identify larger study populations and minimize costs, while investigating the association across BMI categories and exploring the underlying mechanisms for the association. Sex-stratified analyses, taking into account the potential interaction with mode of delivery, and differences in obesity rates, should be performed between populations. The reported increased odds of obesity in later life after CS, including those found in this study, support the plausibility of a biological link and should be considered as a motivating factor to reduce CS as a mode of delivery, unless clinically indicated.

showing heterogeneity in obesity rates between male and female young adults.

(PDF)

Click here for additional data file.

Comparing the mean of observed and imputed data.

(PDF)

Click here for additional data file.

Body mass index categories of study participants by socio-demographic characteristics.

(PDF)

Click here for additional data file.

Body mass index categories of study participants by maternal characteristics (continuous).

(PDF)

Click here for additional data file.

The association between mode of delivery and early adulthood obesity, stratified by sex (using normal BMI as reference) – complete case analysis.

(PDF)

Click here for additional data file.

Sensitivity analysis: Examining the association between mode of delivery and early adulthood obesity under different scenarios.

(PDF)

Click here for additional data file.

30 Aug 2019

PONE-D-19-21640

The association between caesarean section delivery and later life obesity in 21-24 year olds in an Urban South African birth cohort

PLOS ONE

Dear Dr Sogunle,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

We would appreciate receiving your revised manuscript by Oct 14 2019 11:59PM. When you are ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter.

To enhance the reproducibility of your results, we recommend that if applicable you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). This letter should be uploaded as separate file and labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. This file should be uploaded as separate file and labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. This file should be uploaded as separate file and labeled 'Manuscript'.

Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.

We look forward to receiving your revised manuscript.

Kind regards,

David Meyre

Academic Editor

PLOS ONE

Journal Requirements:

1. When submitting your revision, we need you to address these additional requirements.

Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

http://www.journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and http://www.journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2. We note that you have indicated that data from this study are available upon request. PLOS only allows data to be available upon request if there are legal or ethical restrictions on sharing data publicly. For information on unacceptable data access restrictions, please see http://journals.plos.org/plosone/s/data-availability#loc-unacceptable-data-access-restrictions.

In your revised cover letter, please address the following prompts:

a) If there are ethical or legal restrictions on sharing a de-identified data set, please explain them in detail (e.g., data contain potentially identifying or sensitive patient information) and who has imposed them (e.g., an ethics committee). Please also provide contact information for a data access committee, ethics committee, or other institutional body to which data requests may be sent.

b) If there are no restrictions, please upload the minimal anonymized data set necessary to replicate your study findings as either Supporting Information files or to a stable, public repository and provide us with the relevant URLs, DOIs, or accession numbers. Please see http://www.bmj.com/content/340/bmj.c181.long for guidelines on how to de-identify and prepare clinical data for publication. For a list of acceptable repositories, please see http://journals.plos.org/plosone/s/data-availability#loc-recommended-repositories.

We will update your Data Availability statement on your behalf to reflect the information you provide.

3. We note that you have included the phrase “data not shown” in your manuscript. Unfortunately, this does not meet our data sharing requirements. PLOS does not permit references to inaccessible data. We require that authors provide all relevant data within the paper, Supporting Information files, or in an acceptable, public repository. Please add a citation to support this phrase or upload the data that corresponds with these findings to a stable repository (such as Figshare or Dryad) and provide and URLs, DOIs, or accession numbers that may be used to access these data. Or, if the data are not a core part of the research being presented in your study, we ask that you remove the phrase that refers to these data.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Partly

Reviewer #2: Yes

Reviewer #3: Yes

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: No

Reviewer #2: Yes

Reviewer #3: Yes

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: No

Reviewer #2: Yes

Reviewer #3: No

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

Reviewer #3: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: Sogunle et al. present a study that examines the association between birth by Caesarean section and obesity in 889 young adult members of a South African birth cohort in Soweto, Johannesburg. The manuscript is well written but it is limited by the inability to adjust for the key confounder maternal pre-pregnancy weight. I also have some concerns with the methodology:

## Major concerns

The authors performed a post-hoc power calculation. Power calculations should always be performed *before* a study is done, not afterwards. Since the estimates from the study have an error, using them in the power calculation will result in a power estimate that also has an error and is thus meaningless. Please remove entirely.

I do not understand why the authors split the vaginal delivery group into "normal" and "assisted". From the microbiome hypothesis standpoint, there is no reason to split this group. Others have splitted the C-section group into "before 2nd stage of labor" and "at 2nd stage of labor" to capture exposure to the vaginal flora in the latter group, but I am not aware of any other paper on the CS/OB association that split the vaginal delivery group. I would strongly suggest that the authors remove that analysis and instead present the combined vaginal delivery group analysis (Table 6, top) as their main analysis.

The prevalence of obesity is 11.8% for the overall sample and 20.9% in the women. With a non-rare outcome, logistic regression is not the best choice of model because of the overestimation of the RR by the OR and the non-collapsibility of the OR (Shrier and Pang 2015). The authors should re-run the analysis with either a Poisson model with robust standard errors or a log-binomial model (Knol et al. 2012). The former is easily done in Stata using

`poisson y x1 x2 x3, robust irr`

Knol MJ, Le Cessie S, Algra A, Vandenbroucke JP, Groenwold RH. Overestimation of risk ratios by odds ratios in trials and cohort studies: alternatives to logistic regression. CMAJ.2012 May 15;184(8):895-9.

Shrier I, Pang M. Confounding, effect modification, and the odds ratio: common misinterpretations. J Clin Epidemiol. 2015 Apr;68(4):470-4.

The authors created separate categories for missing covariates, with some of these categories having n=1 obeservations. While this is a quick and easy fix, it is not appropriate anymore in this day and age where all statistical software packages have easily accessible multiple imputation commands:

https://stats.idre.ucla.edu/stata/seminars/mi_in_stata_pt1_new/

Please use MI to impute the missing values.

Why was ethnicity not included as a confounder in the models? It seems like an obvious choice, since it affects both CS rates and obesity risk.

Please remove the number needed to harm calculation, as you are not able to assume causality when one key confounder was not available (maternal pre-pregnancy weight).

There are several typos in the tables:

Table 1 - The n's for Ethnicity, Education do not add up to 889 horizontally and vertically.

Table 4 - Either the BMI median (23.3) or the IQR (20.7, 23.3) for the overall group, CS stratum are not correct. Also, the authors should check the p value for the male group - it appears to be falsely low.

The cohort originally included some 3400 people. The authors should state if and how the remaining 889 adults differ from the original cohort.

The authors claim that their study is the first to examine the association between C-section and obesity in the African context. They should discuss (in the Discussion) how and why they think the association would be different in the Southafrican context compared to high income countries or non-African middline income countries.

## Minor concerns

Table 6 could be omitted as all the relevant information is already in the text.

Line 250: The effect estimate (1.50) is outside the 95% CI (1.02, 1.20)

Line 255: The beta coefficient (0.12) is outside the 95% CI (0.01, 0.03)

Line 258: Please do not cite a commentary as evidence. There is a systematic review by the same authors that could be cited if need be:

Kuhle S, Tong OS, Woolcott CG. Association between caesarean section and childhood obesity: a systematic review and meta-analysis. Obes Rev. 2015 Apr;16(4):295-303.

Line 276: should read "seven-year" (with hyphen)

Line 289: give % instead of n ("72 women")

Reviewer #2: This study, using the data from a cohort in South Africa, investigated the association between caesarean section delivery and risk of obesity in 21-24 years. This topic is not new. However, the study design was good, and the manuscript is well written.

Page 6 line 145-148, I think it is better to remove this part in the “potential cofounders”

Page 9 line 179-187, the author says, ”compared with primary cohort,…….”, but I did not see the table or figure illustrate this result.

You can consider to merge table 3 and table 4, as they showed similar results.

Reviewer #3: The manuscript by Sogunle and colleagues describes the results of a study evaluating the relation of mode of delivery - vaginal vs. cesarean - with obesity among young adults. As pointed out by the authors, this manuscript follows a growing but consistent literature linking birth by cesarean section to increased adiposity to childhood obesity. Also as pointed out by the authors, there is less literature addressing the question of whether this risk persists later in life and there is no literature from Africa so far. This paper addresses these two important gaps. Despite the importance of providing information that can help close these knowledge gaps, the paper has some important limitations - many of them already acknowledged by the authors - that require some revisions prior considering further this manuscript. My main concerns are the following.

1. Possibility of confounding by indication. As is the case of most of the literature addressing this topic, the most important limitation of this paper is that it cannot distinguish between cesarean deliveries performed based on a clear obstetric/medical indication from cesarean deliveries performed due to maternal request, physician’s convenience, repeat cesarean or similar reasons. The introduction, however, mentions a potential instrumental variable for non-indicated cesareans. Specifically, the enormous gaps in cesarean delivery rate between hospitals in the public sector and hospitals in the private sector is unlikely to be explained by any reason different to convenience - wether it is convenience of the mother, the physician or both. Can the authors present data on mode of delivery by type of hospital in Table 1. Also, please consider type of hospital as a potential confounder and effect modifier. Also, while the authors do not have data on indication of cesarean, there is information of normal vs assisted vaginal delivery which could serve as a proxy for indication for medical intervention. In other words, the same events that lead to an instrumented vaginal delivery in South Africa, could in o different context lead to a cesarean delivery. Being able to provide different estimates for natural vs instrumented vaginal is a particular strength of this paper and I would suggest that the authors highlight further this particular strength of their paper in the discussion.

2. Possibility of selection bias. The authors mention that only a fraction of the original participants could be assessed for outcome. They also mention that the frequency of cesarean delivery does not differ between participants included and not included in the analysis. This is important as it makes it any major effect of selection bias less likely. I would suggest adding a supplemental table showing key characteristics of individuals included and not included in the analysis to further highlight this point and to mention this issue briefly in the discussion emphasizing how this may or may not result in selection bias.

3. Possibility of residual confounding. I agree with the authors that a major weakness of the study is the lack of key data on potential confounders, particularly for maternal prepregnancy BMI. While the authors do mention this as a potential issue in the discussion, quantifying the potential effect of residual confounding would be ideal. A simple and easy to interpret method to provide this quantification is the one described by VanderWeele and colleagues (Ann Intern Med 2017; JAMA 2019). I would encourge the authors to use this or other methods to quantify the effect of unmeasured/residual confounding in this paper.

4. Please justify the split of 3100g to test EM by birthweight

5. Please describe briefly the methods used to estimate the number needed to harm

6. Lines 244-259. Please also mention ref 25 in this paragraph as it appears to be the one paper not included in the meta-analysis that is not explicitly mentioned in this paragraph.

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: No

Reviewer #3: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files to be viewed.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email us at figures@plos.org. Please note that Supporting Information files do not need this step.

23 Oct 2019

REVIEWERS’ COMMENTS

Please find the point-by-point responses to the reviewers’ comments below.

REVIEWER #1

Sogunle et al. present a study that examines the association between birth by Caesarean section and obesity in 889 young adult members of a South African birth cohort in Soweto, Johannesburg. The manuscript is well written but it is limited by the inability to adjust for the key confounder maternal pre-pregnancy weight. I also have some concerns with the methodology.

Major concerns

Comment 1: The authors performed a post-hoc power calculation. Power calculations should always be performed *before* a study is done, not afterwards. Since the estimates from the study have an error, using them in the power calculation will result in a power estimate that also has an error and is thus meaningless. Please remove entirely.

Response: We have removed the power calculation, as requested.

Comment 2: I do not understand why the authors split the vaginal delivery group into "normal" and "assisted". From the microbiome hypothesis standpoint, there is no reason to split this group. Others have splitted the C-section group into "before 2nd stage of labor" and "at 2nd stage of labor" to capture exposure to the vaginal flora in the latter group, but I am not aware of any other paper on the CS/OB association that split the vaginal delivery group. I would strongly suggest that the authors remove that analysis and instead present the combined vaginal delivery group analysis (Table 6, top) as their main analysis.

Response: We agree with this criticism which echoes Reviewer 3 (comment no. 1), that we did not adequately explain the rationale behind our splitting of the vaginal delivery group into "normal" and "assisted" and the strengths thereof. We believe that our explanation justifies our reason for retaining the ‘split’ analysis. We have thus revised the texts in the methods (Page 6, Line 133-136) and discussion (Page 18, Line 369-373) sections, as follows:

“Being able to provide different estimates for normal versus assisted VD is a particular strength of this study. Infant umbilical vein cortisol and other stress molecule concentrations differ significantly between those born by normal and assisted VD, which might influence longer term offspring energy metabolism (Mears et al. 2004). Also, higher birth weight, which has been reported to be associated with CS birth, is a risk factor for instrumental VD (Aiken et al. 2014). Nevertheless, we did, in addition, analyse the combined VD data to interrogate the potential effect of differential exposure of infants to maternal vaginal and faecal microflora, implicated in the subsequent genesis of childhood obesity, between those born by CS and VD.”

Mears K, McAuliffe F, Grimes H, Morrison J. Fetal cortisol in relation to labour, intrapartum events and mode of delivery. Journal of Obstetrics and Gynaecology. 2004;24: 129–132. doi:10.1080/01443610410001645389

Aiken CE, Aiken AR, Brockelsby JC, Scott JG. Factors influencing the likelihood of instrumental delivery success. Obstet Gynecol. 2014;123: 796–803. doi:10.1097/AOG.0000000000000188

Comment 3: The prevalence of obesity is 11.8% for the overall sample and 20.9% in the women. With a non-rare outcome, logistic regression is not the best choice of model because of the overestimation of the RR by the OR and the non-collapsibility of the OR (Shrier and Pang 2015). The authors should re-run the analysis with either a Poisson model with robust standard errors or a log-binomial model (Knol et al. 2012). The former is easily done in Stata using

`poisson y x1 x2 x3, robust irr`

Knol MJ, Le Cessie S, Algra A, Vandenbroucke JP, Groenwold RH. Overestimation of risk ratios by odds ratios in trials and cohort studies: alternatives to logistic regression. CMAJ.2012 May 15;184(8):895-9.

Shrier I, Pang M. Confounding, effect modification, and the odds ratio: common misinterpretations. J Clin Epidemiol. 2015 Apr;68(4):470-4.

Response: We agree that logistic regression is not the best choice of model, and have re-analyzed the data using the Poisson regression model with robust standard errors on Page 5, line 118-121. The results are shown on page 13 line 255-262 and Table 6 on page 14.

Comment 4: The authors created separate categories for missing covariates, with some of these categories having n=1 observations. While this is a quick and easy fix, it is not appropriate anymore in this day and age where all statistical software packages have easily accessible multiple imputation commands:

https://stats.idre.ucla.edu/stata/seminars/mi_in_stata_pt1_new/

please use MI to impute the missing values.

Response: We have imputed the missing data, as suggested. The previous OR was 1.99 (95% CI 1.00 – 3.94); the new IRR is 1.64 (95% CI 1.01 - 2.68).

Comment 5: Why was ethnicity not included as a confounder in the models? It seems like an obvious choice, since it affects both CS rates and obesity risk.

Response: Although, ethnicity was associated with both the mode of delivery and BMI, it was not included in the analysis. Because including it did not significantly change the effect estimate, in our final regression model. We have clarified this in the manuscript (see Page 6, line 126-128).

Comment 6: Please remove the number needed to harm calculation, as you are not able to assume causality when one key confounder was not available (maternal pre-pregnancy weight).

Response: We have removed the calculation.

Comment 7: There are several typos in the tables:

a. Table 1 - The n's for Ethnicity, Education do not add up to 889 horizontally and vertically.

Response: The numbers have been corrected in Table 3 (previously Table 1) on page 11.

Please note: Table 1 is now table 3 because we have added Tables 1 and 2 for comparison between the primary cohort and the study participants (see response to Reviewer 2, comment 2).

b. Table 4 - Either the BMI median (23.3) or the IQR (20.7, 23.3) for the overall group, CS stratum are not correct. Also, the authors should check the p value for the male group - it appears to be falsely low.

Response: We have removed Table 4 from the report, based on Reviewer 2’s comment that Table 3 (now Table 5 on page 13) shows similar results, i.e. information on the prevalence of obesity, stratified by sex.

Please note: Table 3 is now Table 5 because we have added Tables 1 and 2 for comparison between the primary cohort and the study participants (see response to Reviewer 2, comment 2).

Comment 8: The cohort originally included some 3400 people. The authors should state if and how the remaining 889 adults differ from the original cohort.

Response: We have now included Tables 1 and 2, comparing the characteristics of the members in the primary cohort (n = 3273) with those in our study sample (n = 889), in the results section. Most of the participants’ characteristics (including our outcome measure (BMI)) and the predictor (mode of delivery) were similar between the two groups. Other similar characteristics were: young adults’ sex, education, alcohol habit, birth weight; and mothers’ parity. Characteristics that differed, as well as the inserted tables, are briefly discussed on pages 8-10, line 190-212.

Comment 9: The authors claim that their study is the first to examine the association between C-section and obesity in the African context. They should discuss (in the Discussion) how and why they think the association would be different in the South African context compared to high income countries or non-African midline income countries.

Response: We have added a paragraph justifying this in the discussion section on page 16, line 318-322.

“Findings from South Africa, regarding certain established potential confounders, such as socioeconomic status (SES), are different from those reported in high income countries. For example, while most high income countries have reported that high SES leads to higher body mass index, those conducted in South Africa have found that low SES was associated with higher body mass index [Cois et al. 2015; Micklesfield et al. 2013].”

Cois A, Day C. Obesity trends and risk factors in the South African adult population. BMC Obes. 2015;2: 42–42. doi:10.1186/s40608-015-0072-2

Micklesfield LK, Lambert EV, Hume DJ, Chantler S, Pienaar PR, Dickie K, et al. Socio-cultural, environmental and behavioural determinants of obesity in black South African women. Cardiovasc J Afr. 2013;24: 369–375. doi:10.5830/CVJA-2013-069

Minor concerns

Comment 1: Table 6 could be omitted as all the relevant information is already in the text.

Response: Table 6 containing the sensitivity analysis in our previously submitted manuscript has been added as S4 Table in the supplementary information.

Comment 2: Line 250: The effect estimate (1.50) is outside the 95% CI (1.02, 1.20)

Response: There was a typographical error in the 95% CI. It has been corrected to 1.02 - 2.20 (see line 304, page 15).

Comment 3: Line 255: The beta coefficient (0.12) is outside the 95% CI (0.01, 0.03)

Response: This has been removed. There was an error in the published manuscript which showed a significant p value and a point estimate outside the 95% CI (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6546227/).

Comment 4: Line 258: Please do not cite a commentary as evidence. There is a systematic review by the same authors that could be cited if need be:

Kuhle S, Tong OS, Woolcott CG. Association between caesarean section and childhood obesity: a systematic review and meta-analysis. Obes Rev. 2015 Apr;16(4):295-303.

Response: The text referring to the commentary has been removed.

Comment 5: Line 276: should read "seven-year" (with hyphen)

Response: This has been corrected in line 335, page 17.

Comment 6: Line 289: give % instead of n ("72 women")

Response: This has been corrected in line 346, page 17.

REVIEWER #2

This study, using the data from a cohort in South Africa, investigated the association between caesarean section delivery and risk of obesity in 21-24 years. This topic is not new. However, the study design was good, and the manuscript is well written.

Comment 1: Page 6 line 145-148, I think it is better to remove this part in the “potential cofounders”

Response: The text to which the reviewer is referring is: “Although, lifestyle and behavioral characteristics, such as YAs’ diet, physical activity, smoking habits, and alcohol consumption, have been associated with obesity, it has been suggested that they are not true confounders in the analysis of the association between mode of delivery and later life obesity”. We have not removed the variables from the list, as we think it is important to mention them, as suggested by Yuan et al. 2016, and Mueller et al. 2016) (see Page 6, line 121-124).

Mueller NT, Mao G, Bennet WL, Hourigan SK, Dominguez-Bello MG, Appel LJ, et al. Does vaginal delivery mitigate or strengthen the intergenerational association of overweight and obesity? Findings from the Boston Birth Cohort. International journal of obesity (2005). 2017;41: 497–501. doi:10.1038/ijo.2016.219

Yuan C, Gaskins AJ, Blaine AI, Zhang C, Gillman MW, Missmer SA, et al. Cesarean birth and risk of offspring obesity in childhood, adolescence and early adulthood. JAMA pediatrics. 2016 Nov 1;170(11):e162385

Comment 2: Page 9 line 179-187, the author says, ”compared with primary cohort,…….”, but I did not see the table or figure illustrate this result.

Response: We have now included the Tables 1 and 2 in the results section on Page 9-10 (see response to Reviewer 1, comment 8).

Comment 3: You can consider to merge table 3 and table 4, as they showed similar results.

Response: As Table 3 and Table 4 showed similar results regarding the prevalence of obesity, we removed Table 4 (see response to Reviewer 1, comment 7). Table 3 (now Table 5) shows the prevalence of obesity for each mode of delivery category - page 13.

Please note: Table 3 is now table 5 because we have added Tables 1 and 2 for comparison between the primary cohort and the study participants (see response to comment 2).

REVIEWER #3

The manuscript by Sogunle and colleagues describes the results of a study evaluating the relation of mode of delivery - vaginal vs. cesarean - with obesity among young adults. As pointed out by the authors, this manuscript follows a growing but consistent literature linking birth by cesarean section to increased adiposity to childhood obesity. Also as pointed out by the authors, there is less literature addressing the question of whether this risk persists later in life and there is no literature from Africa so far. This paper addresses these two important gaps. Despite the importance of providing information that can help close these knowledge gaps, the paper has some important limitations - many of them already acknowledged by the authors - that require some revisions prior considering further this manuscript. My main concerns are the following.

Comment 1: Possibility of confounding by indication. As is the case of most of the literature addressing this topic, the most important limitation of this paper is that it cannot distinguish between cesarean deliveries performed based on a clear obstetric/medical indication from cesarean deliveries performed due to maternal request, physician’s convenience, and repeat cesarean or similar reasons. The introduction, however, mentions a potential instrumental variable for non-indicated cesareans. Specifically, the enormous gaps in cesarean delivery rate between hospitals in the public sector and hospitals in the private sector is unlikely to be explained by any reason different to convenience - whether it is convenience of the mother, the physician or both. Can the authors present data on mode of delivery by type of hospital in Table 1? Also, please consider type of hospital as a potential confounder and effect modifier. Also, while the authors do not have data on indication of cesarean, there is information of normal vs assisted vaginal delivery which could serve as a proxy for indication for medical intervention. In other words, the same events that lead to an instrumented vaginal delivery in South Africa, could in a different context lead to a cesarean delivery. Being able to provide different estimates for natural vs instrumented vaginal is a particular strength of this paper and I would suggest that the authors highlight further this particular strength of their paper in the discussion.

Response:

a. We cannot present results by type of hospital as we do not have these data. The primary cohort comprised recruited participants from public health facilities (antenatal clinics). Although we have no evidence, it is highly unlikely that any of the women went on to give birth in a private facility due to their relatively low SES.

b. We appreciate the suggestion that not having data on type of hospital could be a limitation of the study. This has now been added: “Management practices in labour and delivery units differ across hospitals and has been reported by Plough et al. (2017) to be associated with risk of primary cesarean delivery.” Line, 356-357, page 17.

c. We have further highlighted the strength of the paper regarding the availability of data on normal and assisted vaginal delivery. See response to Reviewer 1, comment 2.

Plough AC, Galvin G, Li Z, Lipsitz SR, Alidina S, Henrich NJ, Hirschhorn LR, Berry WR, Gawande AA, Peter D, McDonald R. Relationship between labor and delivery unit management practices and maternal outcomes. Obstetrics & Gynecology. 2017 Aug 1;130(2):358-65.

Comment 2: Possibility of selection bias. The authors mention that only a fraction of the original participants could be assessed for outcome. They also mention that the frequency of cesarean delivery does not differ between participants included and not included in the analysis. This is important as it makes it any major effect of selection bias less likely. I would suggest adding a supplemental table showing key characteristics of individuals included and not included in the analysis to further highlight this point and to mention this issue briefly in the discussion emphasizing how this may or may not result in selection bias.

Response: We have now done this. Please refer to response to Reviewer 1, comment 8.

Comment 3: Possibility of residual confounding. I agree with the authors that a major weakness of the study is the lack of key data on potential confounders, particularly for maternal prepregnancy BMI. While the authors do mention this as a potential issue in the discussion, quantifying the potential effect of residual confounding would be ideal. A simple and easy to interpret method to provide this quantification is the one described by VanderWeele and colleagues (Ann Intern Med 2017; JAMA 2019). I would encourage the authors to use this or other methods to quantify the effect of unmeasured/residual confounding in this paper.

Response: Although we agree with your concern, we did not quantify the potential effect of residual confounding by calculating the E-value because we would be making, in the context of our study, assumptions about unmeasured confounders. This has been discussed in a paper by Ioannidis et al. (2019).

Ioannidis JPA, Tan YJ, Blum MR. Limitations and Misinterpretations of E-Values for Sensitivity Analyses of Observational Studies. Annals of Internal Medicine. 2019;170: 108–111. doi:10.7326/M18-2159

Comment 3: Please justify the split of 3100g to test EM by birth weight.

Response: We understand that it is unusual to use 3100g for classification, hence, we have removed the EM analysis for birth weight. The classification was done because the sample size for birth weight was low for some mode of delivery and BMI categories. Therefore, the median birth weight i.e. 3100g was used for classification for the EM analysis.

Comment 4: Please describe briefly the methods used to estimate the number needed to harm

Response: We have removed this analysis. Please refer to Reviewer 1, comment 6.

Comment 6: Lines 244-259. Please also mention ref 25 in this paragraph as it appears to be the one paper not included in the meta-analysis that is not explicitly mentioned in this paragraph.

Response: We have included ref 25 in the paragraph in line 311-313 on page 16.

Submitted filename: Response to Reviewers.docx

Click here for additional data file.

31 Oct 2019

The association between caesarean section delivery and later life obesity in 21-24 year olds in an Urban South African birth cohort

PONE-D-19-21640R1

Dear Dr. Sogunle,

We are pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it complies with all outstanding technical requirements.

Within one week, you will receive an e-mail containing information on the amendments required prior to publication. When all required modifications have been addressed, you will receive a formal acceptance letter and your manuscript will proceed to our production department and be scheduled for publication.

Shortly after the formal acceptance letter is sent, an invoice for payment will follow. To ensure an efficient production and billing process, please log into Editorial Manager at https://www.editorialmanager.com/pone/, click the "Update My Information" link at the top of the page, and update your user information. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to enable them to help maximize its impact. If they will be preparing press materials for this manuscript, you must inform our press team as soon as possible and no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

With kind regards,

David Meyre

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

6 Nov 2019

PONE-D-19-21640R1

The association between caesarean section delivery and later life obesity in 21-24 year olds in an Urban South African birth cohort

Dear Dr. Sogunle:

I am pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please notify them about your upcoming paper at this point, to enable them to help maximize its impact. If they will be preparing press materials for this manuscript, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

For any other questions or concerns, please email plosone@plos.org.

Thank you for submitting your work to PLOS ONE.

With kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr David Meyre

Academic Editor

PLOS ONE