The Brazilian Study of Gestational Diabetes (EBDG) is a multicenter cohort study with principal objective to evaluate American Diabetes Association (ADA) and World Health Organization (WHO) diagnostic criteria for gestational diabetes mellitus (GDM) against pregnancy outcomes 
. It was conducted in health care units of the National Health System in six Brazilian cities: Porto Alegre, São Paulo, Rio de Janeiro, Salvador, Fortaleza and Manaus. The study methodology has been previously reported 
. In brief, the study consecutively enrolled women 20 years or older from general prenatal clinics who were between 20 and 28 weeks of pregnancy and had no history of diabetes outside of pregnancy. The ethics committees of each institution involved approved the study. All participants provided their verbal informed consent after being informed about the studýs nature. All clinical investigations were conducted according to the principles expressed in the Declaration of Helsinki. This study occurred before the first regulatory guidelines for research involving humans in Brazil (National Council Resolution 196/96) that instituted mandatory written informed consent for all participants from the year 1996.
Our current investigation uses data from study phases I to III. Phase I consisted of standardized interviews and examinations, and glucose tolerance testing. The interview, performed at the prenatal clinicat enrollment, obtained information on maternal age, skin color, parity and education, as well as alcohol consumption and smoking. Weight and height were measured in duplicate according to standard protocol, and mean values were used in analyses 
. Pre-pregnancy BMI was calculated using the reported pre-pregnancy weight and height measured at enrollment. Pre-pregnancy nutritional status was classified according to the current classification of the Institute of Medicine 
. A 75 g oral glucose tolerance test was then performed between 24 and 30 weeks of pregnancy. Gestational diabetes was defined as blood glucose greater than or equal to 140 mg/dl two hours after intake, according to WHO criteria 
. Hypertensive disorders were classified according to the National High Blood Pressure Education Program 
. Data on clinical evolution, gestational weight gain and delivery were obtained through a review of medical records in study phases II and III. Phase II comprised all prenatal care, including maternal weight data from each prenatal consultation. Phase III involved data collection related to delivery, immediate postpartum period and the infants’ first hours of life.
From a total of 5,564 enrolled pregnant women, 73 did not have their weight and height measured at enrollment, 248 did not report pre-pregnancy weight, 1,123 had no clinic visit with recorded weight after the 28th week of gestation and 1,006 had insufficient data to calculate weight gain in the third trimester, leaving 3,114 pregnant women with calculated gestational weight gain. We excluded an additional 51 participants due to multiple gestation and 819 due to not having information, which permitted the calculation of weight gain separately in both the second and third trimesters, resulting in a total of 2,244 for the weight gain analysis ().
The trimesters were defined as first (less than 14 complete weeks), second (14–27 complete weeks) and third (28 complete weeks until delivery). The mean weekly weight gain in the second and third trimesters was estimated using the difference between the first and last weight record in the trimester divided by the number of weeks between the two observations, as previously reported 
Gestational weight gain was considered adequate in the 2nd
trimesters if the woman was within the range recommended by the 2009 IOM/NRC based on pre-pregnancy BMI: below 18.5 kg/m2
, a gain between 0.44 and 0.58 kg/week; from 18.5 to 24.9 kg/m2
, a gain from 0.35 to 0.50 kg/week; from 25 to 29.9 kg/m2
, a gain from 0.23 to 0.33 kg/week; and greater than or equal to 30 kg/m2
, a gain from 0.17 to 0.27 kg/week 
. A total weight gain from 12.5 to 18 kg was considered adequate for women with pre-pregnancy BMI below 18.5 kg/m2
; a gain from 11.5 to 16 kg for those with pre-pregnancy BMI from 18.5 to 24.9 kg/m2
; a gain from 7 to 11.5 kg for those with pre-pregnancy BMI from 25 to 29.9 kg/m2
; and a gain from 5 to 9 kg for those with pre-pregnancy BMI greater than or equal to 30 kg/m2
We estimated gestational age at delivery using an ultrasound exam performed before the 26th week of gestation. When this information was not available, we used a hierarchical clinical criteria in the following order: any other ultrasound exam consistent with neonatal age estimation or reported last menstrual period; reported last menstrual period consistent with neonatal age estimation or uterine height; neonatal age estimation; ultrasound after week 26; uterine height; and finally, when no other criteria was available, last menstrual period.
Preterm birth outcome was considered as less than 37 weeks of gestation. Small for gestational age (SGA) was defined as birth weight below the 10th percentile for gestational age in the EBDG study, considering those born alive with over 34 weeks of gestation and large for gestational age (LGA) as birth weight greater than the 90th percentile in relation to gestational age.
Categorical characteristics of the sample are presented as absolute and relative frequencies. Pearson’s chi-square was used to test crude associations with weight gain per trimester, categorized as insufficient, adequate and excessive. Weight gain is expressed as a continuous variable with differences in weekly gains between the 2nd and 3rd trimesters being tested using the Wilcoxon matched-pairs signed ranks test.
To characterize the association of weight gain with each dichotomous obstetric outcome (cesarean section, preterm birth, SGA and LGA), Poisson regression models with robust variance were constructed with progressive inclusion of covariates. The covariates considered in the models were pre-pregnancy BMI, trimester-specific weight gain, age, height, skin color, parity, education, smoking, alcohol consumption, gestational diabetes and hypertensive disorders in pregnancy.
The criterion used to include these covariates was a p value <0.20 in the univariate analysis. Variables with p value <0.05 (Wald test) were maintained in the model, and those with higher p values were removed in decreasing order. Additionally, any potential confounder which changed the estimate of the relative risk for the association between weight gain and the outcome variable by >10% was kept in the model. Results were expressed as relative risk (RR) and 95% confidence interval (CI). We performed the data analyses using SPSS version 18 (SPSS Inc., Chicago, IL). The significance level was considered as 0.05.