Cesarean delivery and metabolic risk factors in young adults: a Brazilian birth cohort study

ABSTRACT

Background: Cesarean delivery (CD) perturbs the assembly of the neonatal gut microbiome and has been associated with child and adult obesity. However, it is still unknown whether CD is associated with metabolic risk factors in young adults.

Objective: We investigated the association of CD and metabolic risk factors in young adults in a cohort study who were 23–25 y of age at follow-up.

Design: We used data from a cohort study in Ribeirão Preto, Brazil. Baseline data on 6827 singleton pregnancies were collected in 1978–1979, and a sample of 2063 subjects were followed up 23–25 y later (2002–2004). Information on the type of delivery, birth weight, maternal age, parity, maternal schooling, and maternal smoking was obtained after birth. Anthropometric data, biochemical measurements, and information on participant schooling and smoking history were collected at 23–25 y of age. A linear regression was performed to assess the association between CD and biochemical measurements in early adulthood, controlling for a minimum set of confounders that were identified in a directed acyclic graph.

Results: The mean ± SD age of the subjects was 23.9 ± 0.71 y, and 51.8% of the sample were women. The CD rate was 32.0% and was more common in older (P < 0.001) and more educated mothers (P < 0.001). Compared with vaginal delivery, CD was associated with higher body mass index (BMI) after multivariable adjustment (P < 0.001) but not with glucose, homeostasis model assessment of insulin resistance, the quantitative insulin-sensitivity check index, total cholesterol, LDL cholesterol, HDL cholesterol, or triglycerides (all P > 0.05).

Conclusion: In our sample of Brazilian adults, CD was associated with higher BMI but not with other metabolic risk factors.

INTRODUCTION

Cesarean delivery (CD)⁸ was introduced into clinical practice as a life-saving procedure for the mother and her infant (1). However, in recent decades, an increasing number of CDs have been performed in low-risk pregnancies without medical indication (2–4). The overuse of CD has implications for the health of the mother and also, possibly, the long-term health of the offspring (5–8).

Emerging literature supports the hypothesis that CD is associated with immune-mediated diseases such as asthma (9), celiac disease, allergic rhinitis, gastroenteritis (10, 11), and type 1 diabetes (12). Studies have also found a positive association between CD and BMI in early childhood, adolescents, and young adults (13–16). A systematic review and meta-analysis showed that, compared with vaginal delivery (VD), CD was associated with a 33% increase in overweight and obesity in adulthood (17). Furthermore, CD has been associated with increased peripheral and central adiposity (18, 19).

However, to our knowledge, only one study has investigated CD in relation to metabolic and cardiovascular disease risk factors (blood pressure, blood glucose, HDL cholesterol, triglycerides, C-reactive protein, fat mass, waist circumference, and BMI). The study showed that CD was associated with small increases in systolic blood pressure, BMI, and fat mass but not with other metabolic or cardiovascular disease risk factors (14). Our study aimed to leverage data from a birth cohort study in Brazil with 23–25 y of follow-up to test the hypothesis that young adults born by CD compared with VD have higher BMI and greater presentation of other metabolic risk factors.

METHODS

Sample and participants

We analyzed prospective data from the Ribeirão Preto Birth Cohort Study collected in 1978–1979 (first phase) and 2002–2004 (fourth phase). The initial recruitment occurred between June 1978 and May 1979. A total of 9067 births were registered, corresponding to 98% of all live births in Ribeirão Preto (São Paulo, Brazil) over that period. Participants who were not from Ribeirão Preto or did not reside in the city at the time of delivery (n = 2094) were excluded, leaving 6973 liveborn infants. We also excluded 146 twin deliveries, which left 6827 singleton births. After delivery, trained personnel collected data from mothers and newborns including medical histories and anthropometric data.

From April 2002 to May 2004, when the individuals were 23–25 y of age, attempts were made to invite participants for a medical examination. A total of 5665 participants from the original birth cohort were located. Attempts were then made to invite a nonrandom sample (1 in 3 individuals of this group) for a medical examination. The first of every 3 names was selected from a list sorted by birth date in each geographic region and, if unavailable, the next name down was selected. Through this process, we were able to enroll 2063 young adults (31.8% of all singletons) to participate in the follow-up. Full descriptions of sample characteristics and the methodology were published elsewhere (20, 21).

Outcomes

Anthropometric measures

Weight and height were measured by trained staff with the participant standing barefoot and wearing light clothing. BMI (in kg/m²) was calculated by dividing weight by squared height (22).

Biochemical measurements

A 40-mL blood sample was collected in the interview (2002–2004) after subjects fasted ≥12 h for the measurement of glucose (mg/dL), insulin (mg/dL), total cholesterol (mg/dL), LDL cholesterol (mg/dL), HDL cholesterol (mg/dL), and triglycerides (mg/dL). Fasting glucose was determined by using the glucose oxidase/peroxidase-4-aminophenazone-phenol human-diagnostic colorimetric enzymatic method (Chronolab AG), and fasting insulin was determined by using a radioimmunoassay (insulin kit; DPC) with a CV of 7.9% (23). The HOMA-IR glucose was calculated as

The quantitative insulin-sensitivity check index (QUICKI) (24) was calculated as

Total cholesterol, LDL cholesterol, HDL cholesterol, and triglycerides were determined by using an enzymatic colorimetric method with the Dade Behring XP test and apparatus (Dade Behring) and reagents from Dade Behring Dimension clinical chemistry (Dade Behring).

Covariates

Maternal and perinatal variables

After birth, we collected information on the type of delivery (defined as CD or VD), birth weight (<2500 compared with ≥2500 g), sex (defined as male or female), maternal age (defined as <19, 19–34, or ≥35 y), maternal parity (in number of births defined as 1, 2–4, or ≥5), maternal schooling [defined as low (0–8 y) or high (>8 y)], and maternal smoking [defined as any time during pregnancy for nonsmoker or smoker of 1–9 or ≥10 cigarettes].

Offspring adult variables

In early adulthood, we collected information on schooling ([defined as low (0–8 y) or high (>8 y)], smoking status (defined as nonsmoker, ex-smoker, and smoker), and physical activity by using the International Physical Activity Questionnaire–Short Form and defined as high, moderate, or low according to the 3 levels of physical activity indicated by International Physical Activity Questionnaire guidelines (25).

Statistical analysis

The number of participants (n) and relative frequencies (percentages) were provided for categorical variables. Continuous variables were expressed as arithmetic or geometric means (±SDs). The chi-square test was performed to determine whether categorical variables (Table 1) were different by mode of delivery. Student’s t test (Table 2) was used to compare continuous variables.

TABLE 1.

Participant characteristics of mothers and offspring in the Ribeirão Preto birth cohort (São Paulo, Brazil), according to type of delivery¹

Variable	Vaginal (n = 1402), n (%)	Cesarean (n = 661), n (%)	P
Maternal
Age, y			<0.001
<19	124 (8.9)	29 (4.4)
19–34	1172 (84.2)	555 (84.2)
≥ 35	96 (6.9)	75 (11.4)
Schooling, y			<0.001
≥12	118 (8.6)	97 (14.9)
5–11	580 (42.3)	308 (47.2)
0–4	673 (49.1)	247 (37.9)
Smoking during pregnancy, cigarettes/d			0.762
Nonsmoker	1018 (75.0)	491 (76.5)
1–9	220 (16.2)	98 (15.3)
≥10	120 (8.8)	53 (8.3)
Parity, births			0.300
1	511 (37.2)	260 (39.9)
2–4	739 (53.9)	344 (52.8)
≥5	122 (8.9)	47 (7.2)
Offspring
Sex
M	676 (48.2)	319 (48.3)	0.985
F	726 (51.8)	342 (51.7)
Birth weight, g
<2500	88 (6.3)	40 (6.1)	0.843
≥2500	1314 (93.7)	621 (93.9)
Physical activity			0.496
High	566 (40.4)	285 (43.1)
Moderate	152 (10.8)	69 (10.4)
Low	684 (48.8)	307 (46.4)
Schooling, y
>8	1163 (83.0)	580 (87.7)	0.005
0–8	239 (17.0)	81 (12.3)
Smoking status
Nonsmoker	1027 (73.4)	503 (76.2)	0.316
Ex-smoker	120 (8.6)	55 (8.3)
Smoker	153 (18.1)	102 (15.5)

¹Totals may not add up to 2063 because of missing values. P values were determined by using the chi-square test.

TABLE 2.

Metabolic variables in young adults from the Ribeirão Preto birth cohort (São Paulo, Brazil) according to type of delivery¹

Variable	Total, n (%)	Total cholesterol,2 mg/dL	HDL cholesterol,2 mg/dL	LDL cholesterol,3 mg/dL	Triglycerides,3 mg/dL	Glucose,3 mg/dL	Insulin,3 mg/dL	HOMA-IR3	QUICKI3,4	BMI,3 kg/m2
Type of delivery
Vaginal	1402 (68.0)	167.1 (165.2, 169.1)	48.3 (47.7, 49.0)	95.4 (93.8, 96.9)	82.0 (79.8, 84.2)	83.0 (82.5, 83.5)	5.29 (5.08, 5.50)	1.08 (1.04, 1.13)	0.38 (0.38, 0.38)	23.6 (23.3, 23.8)
Cesarean	661 (32.0)	169.6 (166.8, 172.4)	48.1 (47.1, 49.1)	97.6 (95.4, 99.9)	83.8 (80.4, 87.3)	83.7 (82.7, 84.7)	5.31 (4.98, 5.65)	1.10 (1.02, 1.17)	0.38 (0.38, 0.39)	24.5 (24.2, 24.9)
P	—	0.154	0.722	0.107	0.381	0.154	0.918	0.749	0.987	<0.001

¹ P values were determined by using Student’s t test.

²All values are unadjusted arithmetic means; 95% CIs in parentheses.

³All values are unadjusted geometric means; 95% CIs in parentheses.

⁴QUICKI, quantitative insulin-sensitivity check index.

A multivariable linear regression analysis was performed to examine the association between CD and metabolic risk factors in early adulthood before and after controlling for covariates. The minimum set of variables to enter the adjusted model was identified in a directed acyclic graph (26, 27) in the DAGitty program (version 2.0 alpha, Johannes Textor) (28). Models were adjusted for maternal schooling at birth, maternal age, maternal smoking during pregnancy, birth weight, sex, and parity (Supplemental Figure 1).

Glucose, insulin, HOMA-IR, QUICKI, LDL cholesterol, triglycerides, and BMI were ln transformed for greater symmetry before undertaking linear regression. Back transformation was used to express the geometric means and 95% CIs of these variables. Data were analyzed with Statistical Package for the Social Sciences 18.0 software (SPSS Inc.). Significance levels for all statistical tests were 2-sided with α = 0.05.

Ethical aspects

This study was approved by the Research Ethics Committee of the Hospital das Clinicas de Ribeirão Preto, Faculdade de Medicina de Ribeirão Preto, Universidade de São Paulo (protocol Hospital das Clínicas de Ribeirão Preto n. 7606/99). All subjects gave written informed consent to participate in the study.

RESULTS

A total of 2063 subjects were included in the study, and the total sample size varied for each outcome. The mean ± SD age of subjects was 23.9 ± 0.7 y, and 51.8% (n = 1068) of the cohort was women. The CD rate was 32.0% (n = 661), which was similar to that of those not included in the follow-up (29.7%) (19).

Table 1 shows maternal and offspring characteristics according to type of delivery. CD was more common in older women (P < 0.001), more-educated mothers (P < 0.001), and more-educated offspring (P = 0.005). There were no other significant (all P > 0.05) differences in baseline characteristics according to delivery mode.

Geometric and arithmetic means (95% CIs) of metabolic variables in our participants at 23–25 y of age according to their type of delivery are shown in Table 2. In crude analyses, compared with VD, participants born by CD had higher BMI (P < 0.001). No other metabolic risk factors were significantly associated with type of delivery (all P > 0.05).

Table 3 shows the multivariable-adjusted geometric and arithmetic means (95% CIs) of metabolic variables according to the type of delivery. Similar to crude analyses, CD (compared with VD) was associated with higher BMI (P < 0.001) but not with other metabolic risk factors (P > 0.05). We did not find that associations differed by sex (all P-interactions > 0.05).

TABLE 3.

Regression coefficients for metabolic variables in young adults in the Ribeirão Preto birth cohort (São Paulo, Brazil) born by cesarean delivery compared with vaginal delivery¹

		P [n]		B (95% CI)
Variable	All individuals	Crude	Adjusted2	Crude	Adjusted2
Total cholesterol, mg/dL	167.9 (166.4, 169.5)3	0.154 [2037]	0.230 [1997]	0.03 (−0.93, 5.87)	0.03 (−1.36, 5.65)
HDL cholesterol, mg/dL	48.3 (47.7, 48.8)3	0.722 [2037]	0.206 [1997]	−0.01 (−1.42, 0.98)	−0.03 (−1.91, 0.41)
LDL cholesterol, mg/dL	96.1 (94.8, 97.4)4	0.107 [2029]	0.094 [1989]	0.04 (−0.00, 0.02)	0.04 (−0.00, 0.02)
Triglycerides, mg/dL	82.6 (80.7, 84.4)4	0.381 [2037]	0.468 [1997]	0.02 (−0.01, 0.03)	0.02 (−0.01, 0.03)
Glucose, mg/dL	83.2 (82.8, 83.7)4	0.154 [2043]	0.143 [2002]	0.03 (−0.00, 0.01)	0.03 (−0.01, 0.01)
Insulin, mg/dL	5.3 (5.1, 5.5)4	0.918 [2022]	0.857 [1982]	0.00 (−0.03, 0.03)	0.00 (−0.03, 0.04)
HOMA-IR	1.1 (1.0, 1.1)4	0.749 [2022]	0.689 [1982]	0.01 (−0.03, 0.04)	0.01 (−0.03, 0.04)
QUICKI5	0.4 (0.4, 0.4)4	0.987 [2022]	0.900 [1982]	0.00 (−0.01, 0.01)	0.00 (−0.00, 0.00)
BMI, kg/m2	23.9 (23.7, 24.1)4	<0.001 [2056]	0.001 [2015]	0.10 (0.01, 0.03)	0.12 (0.01, 0.03)

¹ P values were obtained by testing if the coefficient was zero in linear regression models.

²Model was adjusted for maternal schooling at birth, maternal age at childbirth, maternal smoking during pregnancy, birth weight, sex, and parity, variables identified as a minimum adjustment set in a directed acyclic graph.

³Geometric mean; 95% CI in parentheses.

⁴Arithmetic mean; 95% CI in parentheses.

⁵QUICKI, quantitative insulin-sensitivity check index.

DISCUSSION

In this Brazilian cohort followed up 23–25 y after birth, CD was significantly associated with higher BMI in young adulthood (P < 0.001) but not with other metabolic risk factors (all P > 0.05).

The association between CD and higher BMI may pose a public health burden for societies such as Brazil that have experienced considerable increases in CD rate over the past decades. Following our pioneer work on this topic (16, 18, 19), many studies, including 2 meta-analyses (17, 29), provided largely consistent confirmatory evidence. Moreover, it has been shown that this association persists after the inclusion of important confounders including maternal BMI (30). With the increase in risk of overweight and obesity, in addition to peripheral and central adiposity levels, we hypothesized that CD might have an important role in the development of cardiometabolic disease risk factors. However, in the current study, we did not find evidence to support this hypothesis.

There are several plausible explanations as to why we did not observe an association between the delivery mode and metabolic variables in young adulthood. First, CD may not alter metabolic disease risk. Another study from Brazil observed that CD was associated with BMI and blood pressure but not with other metabolic outcomes (14). Second, we may not have observed an association because we included nonelective CD deliveries that were due to a premature rupture of membranes or other biological causes. If only elective CD deliveries that were not performed because of biological reasons are associated with increased risk of metabolic disease, we may not have detected an association by also including nonelective CD deliveries in our explanatory variable. However, a recent meta-analysis showed that when data (from 5 studies) were stratified by type of CD, there were no significant differences in adult BMI, overweight, and obesity (17). Third, there is the possibility that a selection bias influenced our results. If subjects born by CD, who were more likely to develop deranged metabolic risk factors, were less likely to have been followed up, we may have missed an association. We also could not rule out the possibility that negative confounding by unmeasured maternal factors during pregnancy obfuscated a true association. Finally, it may be that the association between CD and metabolic risk factors is more pronounced in prosperous societies where breastfeeding is less common and where there is limited exposure to diverse bacteria in the immediate postnatal period. As such, studies on this topic in more economically prosperous settings are warranted.

Although our findings are largely consistent with the only other study that examined the delivery mode and metabolic outcomes (14, 31), there are differences in methods worth noting. We examined more metabolic variables, and unlike the previous study, blood glucose and lipids concentrations in our study were obtained from fasting samples and measured by using the glucose oxidase/peroxidase-4-aminophenazone-phenol human diagnostic colorimetric enzymatic method (32, 33). Our study also had the distinction of the use of a directed acyclic graph as a method for better understanding relations between the variables and to determine the minimum set of variables that should enter the adjusted model, which reduced the chances of overadjustment and collider bias (27, 34).

We showed that CD was more common in older mothers and mothers who had more schooling. CD offspring were also more likely to have greater educational achievement. This finding is in accordance with a study in Brazil, which showed that there has been a change in maternal age distribution toward older mothers accompanied by an increasing low-birth-weight rate (35). Other researchers proposed that the introduction of live saving technologies, such as CD, reaches initially the most-privileged social strata (e.g., high-schooling mothers) (36, 37).

In our study, BMI was significantly associated (P < 0.001) with higher log insulin and log HOMA-IR (data not shown). Our hypothesis was that consequences of CD start with increased BMI and later influence glucose, insulin, and consequently HOMA-IR and QUICKI regulation. Other studies have shown that metabolic disorders are preceded by BMI increases (38, 39). This effect might explain why, in this population of young adults, we showed that CD was associated with increased BMI but not with other metabolic risk factors.

One of the mechanisms by which CD may increase risk of obesity is by negatively affecting the developing gut microbiota in infants (11, 40–42). Early infancy differences in microbial profiles associated with the delivery mode (43) appear to have an impact on the development of some inflammatory and autoimmune allergic disorders such as obesity and allergic diseases later in life (44–47). It remains to be seen whether early prebiotic or probiotic supplementation to mimic the intestinal microflora of vaginally delivered infants might prevent long-lasting health consequences of an altered early life gut microbiota (43, 48). The mode of delivery has also been associated with DNA methylation (49, 50), which may also mediate the association between CD and obesity.

The main strengths of our study were the relatively large sample size in a unique population and outcomes were measured in a close age range of a healthy young adults, which minimized the confounding effect of age and age-dependent covariates.

But there were important limitations to consider. The main limitation of our study was the lack of data regarding CD indication, especially maternal prepregnancy BMI, which is an important risk factor for emergency CD (51, 52) and offspring health outcomes, including offspring BMI (53–55). Another major limitation was our lack of data on the response rate, which handicapped our ability to rule out the influence of a selection bias on the observed associations. We also did not have information on whether CD occurred before labor was started (56) or whether mothers were administered antibiotics before or during labor (30).

We also did not know whether or how long infants were breastfed in our study. Breastfeeding is known to stimulate the proliferation of a well-balanced and diverse microbiota (Bifidobacteria, Lactobacillus, and Bacteroides) (57), and may ameliorate possible adverse metabolic consequences of CD. In addition, reviews showed that CD is a risk factor for early weaning (58, 59). Although a lack of breastfeeding data was clearly a limitation, in Brazil, at the time of the current study, the type of delivery was shown to have little effect on the breastfeeding duration in 6-mo-old infants (60). If this were the same in Ribeirão Preto, estimates of this study would have changed little had we included a breastfeeding variable. Moreover, in Brazil, breastfeeding rates are associated with the maternal education level (61), and estimates were controlled for maternal education in this study.

In conclusion, in our sample of adult Brazilians aged 23–25 y, CD was associated with higher BMI but not with glucose, total cholesterol, LDL cholesterol, HDL cholesterol, triglycerides, insulin, HOMA-IR, and QUICKI. Future research on this topic from prospectively followed birth cohort studies that have collected information on potential confounding factors during pregnancy and have minimal participant attrition during follow-up are needed to confirm these findings.

ABBREVIATIONS

CD

cesarean delivery

QUICKI

quantitative insulin-sensitivity check index

VD

vaginal delivery