This was a prospective cohort study including live-born neonates in the city of Ribeirão Preto/São Paulo, from June 1978 to May 1979 
. During this period, 9,067 live neonates born in the eight Ribeirão Preto hospitals (98% of the total number of live newborns during the period) participated in the study. There were 3.5% losses due to refusal or early discharge from hospital. Babies whose mothers did not reside in the city and were not from Ribeirão Preto at the time of delivery were excluded, with 6,973 live newborns remaining, 6,827 of them singletons and 146 twin deliveries.
The cohort was re-evaluated between April 2002 and May 2004 when the individuals had completed 23–25 years of age. Of these, 246 died during the first year of life 
and 97 died by 20 years of age, for a total of 343 deaths 
, leaving 6,484 eligible subjects. Contact was sought with one in each three individuals based on the geo-economic characterization of the city, divided into four regions, according to family head income. Based on the records of the Unified Health System and of private health plans and on the contacts made in the 2nd and 3rd phase of the study, it was possible to locate 5,665 individuals. The losses due to refusal to participate in the study (209 cases), to death after 20 years of age (34 cases), imprisonment (31 cases) and failure to attend the interview (431 cases) corresponded to a total of 705 individuals. Losses were replaced using the same sampling frame, resulting in 2,063 young adults aged 23 to 25 years, corresponding to 31.8% of the 6,484 subjects, participating in the 4th phase of the study of the Ribeirão Preto cohort 
The final sample consisted of 2,063 participants. This sample size permitted us to detect a 6% difference in the increased prevalence of adiposity between CS and vaginal delivery, assuming a prevalence of about 30%, with an 80% power and a 5% probability of type I error. For prevalence around 10% this same sample size permits the detection of 4% differences with the same power and the same probability of type I error. Details of the methodology have been previously published 
The mothers were interviewed soon after delivery using a questionnaire with socioeconomic and demographic information. The newborns were weighed by trained personnel, using standardized techniques 
. Gestational age was calculated on the basis of the mother's information about the last normal menstrual period.
The young adults were interviewed in order to obtain socioeconomic, demographic and life habit information. The following anthropometric measurements were obtained: weight, height, waist and hip circumference, and tricipital and subscapular skinfolds using standardized techniques applied by trained personnel. All measurements were obtained with the subjects wearing light clothing and no shoes.
WC was measured at the midpoint between the last rib and the upper margin of the iliac crest using an inextensible metric tape 
and classified as increased when its value was ≥90 cm for men and ≥80 cm for women, as proposed by the International Diabetes Federation 
Height was measured with the individual standing up and barefoot, using a wood stadiometer with a wood support and an inextensible ruler. The subject stood up erect, with arms along the body and head on the Frankfurt plane 
WHtR was calculated as waist circumference in cm divided by height in cm and was defined as increased for men and women when its value was >0.5 
Hip circumference was measured at the point of greater circumference on the gluteal region using an inextensible tape 
. The waist-hip ratio (WHR) was calculated by dividing waist circumference in cm by the hip circumference in cm, and was considered to be increased when its value was ≥0.90 for men and ≥0.85 for women 
The tricipital skinfold (TSF) was measured in the posterior midpoint of the arm between the acromion and olecranon and the subscapular skinfold (SSF) was measured 2 cm below the margin of the lower angle of the scapula 
using a caliper (Holtain Ltd., Crynych, U.K.,) with a limit measurement of 40 mm. Values above the 90th percentile obtained for the study population were considered to be increased.
The birth variables selected were birth weight (<2500 g, 2500 |–3000 g, 3000 |–3500 g, 3500 |–4000 g and ≥4000 g), type of delivery (vaginal and cesarean), newborn's sex, maternal schooling in years of study (0–4, 5–8, 9–11 and ≥12), maternal smoking during pregnancy as number of cigarettes smoked per day (non-smoker, 1–10, >10), parity (1, 2–4, ≥5), maternal age (<20, 20–34 and ≥35 years) and gestational age as a continuous variable.
The association of type of delivery with increased WC, WHtR, WHR, TSF and SSF was estimated by Poisson regression with robust adjustment of variance, with the calculation of the incidence rate ratio (IRR) and its respective 95% confidence interval (95% CI) 
, and with the level of significance set at 0.05. The independent variables listed above were first submitted to non-adjusted analysis for each response variable; next, adjusted analyses were carried out, with the type of delivery being the explanatory variable and the remaining variables being possible confounders. Since there was selective attrition according to some birth variables, probabilities of selection for each individual were calculated in a logistic regression model. In this model those followed-up were coded 1 and those not followed up were coded 0. Maternal schooling, sex, maternal smoking during pregnancy and parity were predictors of the probability of participation in the follow-up. To verify if these different probabilities of selection would have biased the estimates, models using inverse-probability weighting were then fitted and compared with estimates derived from models without weighting 
Four models were fitted for each response variable. The first was the unadjusted model. The second was the unadjusted model using inverse-probability weighting. The third model was adjusted for birth variables (newborn's weight and sex, maternal schooling, maternal smoking during pregnancy, parity, maternal age and gestational age), and the last model was adjusted for birth variables using inverse-probability weighting. No significant interactions were detected between sex and the remaining adjustment variables. All analyses were carried out using Stata, version 12. Model fit was evaluated by the goodness of fit chi-squared test.
The study was approved by the Research Ethics Committee of the University Hospital, Faculty of Medicine of Ribeirão Preto, University of São Paulo (protocol HCRP n. 7606/99). All subjects gave written informed consent to participate in the study.