Modeling Pathways From the Perinatal Factors to the Vascular Risk Phenotype at the End of the Second Decade of Life

HomeHypertensionVol. 76, No. 2Modeling Pathways From the Perinatal Factors to the Vascular Risk Phenotype at the End of the Second Decade of Life Free AccessResearch ArticlePDF/EPUBAboutView PDFView EPUBSections ToolsAdd to favoritesDownload citationsTrack citationsPermissionsDownload Articles + Supplements ShareShare onFacebookTwitterLinked InMendeleyReddit Jump toSupplemental MaterialFree AccessResearch ArticlePDF/EPUBModeling Pathways From the Perinatal Factors to the Vascular Risk Phenotype at the End of the Second Decade of LifeBirth Cohort, Brazil Marconi Satuf Amaral, Cecilia Claudia Costa Ribeiro, Maria Teresa Seabra de Britto e Alves, Maria Jacqueline Silva Ribeiro, Joelma Ximenes Prado Teixeira Nascimento, Vanda Maria Ferreira Simões, Alexandre Archanjo Ferraro, Marco Antônio Barbieri and Antônio Augusto Moura da Silva Marconi Satuf AmaralMarconi Satuf Amaral From the Public Health Department (M.S.A., C.C.C.R., M.T.S.d.B.e.A., J.X.P.T.N., V.M.F.S., A.A.M.d.S.), Federal University of Maranhão, São Luís, Brazil , Cecilia Claudia Costa RibeiroCecilia Claudia Costa Ribeiro Correspondence to Cecilia Claudia Costa Ribeiro, Graduate Program in Public Health, Public Health Department, Federal University of Maranhão, Rua Barão de Itapary, 155- Centro, São Luís (MA), CEP 65020-070, Brazil. Email E-mail Address: [email protected] https://orcid.org/0000-0003-0041-7618 From the Public Health Department (M.S.A., C.C.C.R., M.T.S.d.B.e.A., J.X.P.T.N., V.M.F.S., A.A.M.d.S.), Federal University of Maranhão, São Luís, Brazil Dentistry Department (C.C.C.R.), Federal University of Maranhão, São Luís, Brazil , Maria Teresa Seabra de Britto e AlvesMaria Teresa Seabra de Britto e Alves From the Public Health Department (M.S.A., C.C.C.R., M.T.S.d.B.e.A., J.X.P.T.N., V.M.F.S., A.A.M.d.S.), Federal University of Maranhão, São Luís, Brazil , Maria Jacqueline Silva RibeiroMaria Jacqueline Silva Ribeiro Ceuma University, São Luís, Maranhão, Brazil (M.J.S.R.) , Joelma Ximenes Prado Teixeira NascimentoJoelma Ximenes Prado Teixeira Nascimento From the Public Health Department (M.S.A., C.C.C.R., M.T.S.d.B.e.A., J.X.P.T.N., V.M.F.S., A.A.M.d.S.), Federal University of Maranhão, São Luís, Brazil , Vanda Maria Ferreira SimõesVanda Maria Ferreira Simões From the Public Health Department (M.S.A., C.C.C.R., M.T.S.d.B.e.A., J.X.P.T.N., V.M.F.S., A.A.M.d.S.), Federal University of Maranhão, São Luís, Brazil , Alexandre Archanjo FerraroAlexandre Archanjo Ferraro Department of Pediatrics, University of São Paulo School of Medicine, University of São Paulo, Brazil (A.A.F.) Department of Pediatrics, University of São Paulo School of Medicine, University of São Paulo, Brazil (A.A.F.) , Marco Antônio BarbieriMarco Antônio Barbieri Faculty of Medicine of Ribeirão Preto, São Paulo University, Department of Puericulture and Pediatrics, Ribeirão Preto, Brazil (M.A.B.). and Antônio Augusto Moura da SilvaAntônio Augusto Moura da Silva From the Public Health Department (M.S.A., C.C.C.R., M.T.S.d.B.e.A., J.X.P.T.N., V.M.F.S., A.A.M.d.S.), Federal University of Maranhão, São Luís, Brazil Originally published29 Jun 2020https://doi.org/10.1161/HYPERTENSIONAHA.119.14218Hypertension. 2020;76:359–365Other version(s) of this articleYou are viewing the most recent version of this article. Previous versions: June 29, 2020: Ahead of Print AbstractDownload figureDownload PowerPointRisk factors act around birth increasing future vascular risk. In this study, we analysed the pathways from perinatal factors to the vascular risk phenotype (VRP) in adolescents including indirect pathways mediated by obesity in adolescence. Data from a Brazilian cohort were collected at birth and at 18 to 19 years (follow-up). A theoretical model was constructed to analyze the association between variables at birth (socioeconomic status, prepregnancy body mass index, mother's age, history of maternal hypertension, maternal smoking, gestational age at birth, birth weight, sex, delivery type) and at follow-up (smoking and excess weight) with the VRP, using structural equation modeling. VRP was a continuous latent variable, representing the shared variance of blood pressure indictors and carotid-femoral pulse wave velocity. Males had higher VRP (standardized coefficient [SC], 0.561; P<0.001). Higher prepregnancy body mass index was associated with higher VRP (SC, 0.140; P=0.032). Gestational age <34 weeks had a total (SC, 0.259; P=0.002) and direct effect (SC, 0.354; P=0.018) on VRP. Cesarean delivery had a total effect, albeit borderline, on VRP (SC, 0.159; P=0.066). Excess weight at follow-up was the main determinant of a high VRP (SC, 0.470; P<0.001). Male sex, cesarean section, gestational age <34 weeks, pregestational excess weight, and excess weight in adolescents were associated with increased VRP at 18 to 19 years of age.IntroductionMalnutrition, overnutrition, and other risk factors during gestation and in the neonatal period are important factors that may underlie cardiovascular disease origins.1,2 Preterm birth and low birth weight have been shown to be associated with future high blood pressure3,4 and arterial stiffening.5 Conversely, a high birth weight has been associated with increased risk of cardiometabolic disease and obesity,6 although high birth weight has not been related directly to future blood pressure.7 Maternal obesity,8 delivery by cesarean section,9 and preterm birth10 have been associated with higher risks of obesity in the future11; thus, obesity may mediate the association between perinatal exposures and the vascular risk phenotype (VRP).A set of factors from the perinatal period (maternal obesity, maternal hypertension, cesarean section, and preterm birth), acts increasing future vascular risk, which have complex relationships of dependence and temporality with each other; that has not been adequately explored by multivariate regression models.3,4,12 Thus, modeling pathways between perinatal factors and vascular disease risk may add knowledge that will allow us to better analyze the multicausal structure through structural equation modeling.We estimated vascular pathology risk using the VRP, a latent variable representing the shared variance of systolic blood pressure, diastolic blood pressure, and carotid-femoral pulse wave velocity (cfPWV). Individually, these effect indicators do not measure the VRP well. However, using them together as a continuous variable may reduce the magnitude of error in vascular measurements; being advantageous in young people because there is no need to use a cutoff value to address its association with risk factors in this population.13To explore the structure of multicausality associated with the VRP, we modeled direct and indirect pathways between prenatal/perinatal factors and this outcome at the end of the second decade of life. In addition, we analyzed the pathways from perinatal factors to the VRP including indirect pathways mediated by obesity in adolescence.MethodsThe data that support the findings of this study are available from the corresponding author upon reasonable request.Study DesignThe studied population was derived from the RPS (Ribeirão Preto, Pelotas, and São Luís [using the initial letters of 3 Brazilian cities]) Birth Consortium. The base sample can be considered representative of the city of São Luís because it included 94.1% of all births in the city during this period (March 1997 to February 1998). From 2831 births, 2541 were from resident mothers. After excluding multiples (n=50), stillbirths (n=48), and infant deaths (n=65), the final sample size included 2378 infants alive at their first anniversary.In 2015, we attempted to locate the participants from the initial cohort for follow-up when they were 18 to 19 years of age. Ultimately, 653 adolescents from the initial birth cohort were identified in 2015 and agreed to participate in the follow-up evaluation (Figure 1).Download figureDownload PowerPointFigure 1. Flowchart of participants. 1997/1998 São Luís Birth Cohort, Brazil.The Research Ethics Committee at the University Hospital/Federal University of Maranhão approved the project in 2005 (n° 3104-476/2005) and in 2015 (n° 19/2015). During the follow-up period, all of the teenaged participants signed an informed consent form.Data CollectionBased on a parental interview questionnaire applied at birth, the following prenatal and birth characteristics were documented: socioeconomic status (ie, mother's education, father's education, household income, and head of the family's occupation), mother's age, mother's height, mother's prepregnancy weight, mother's smoking habits, history of gestational hypertension, date of last menstruation, date of delivery, type of delivery, and the birth weight and sex of the delivered neonate.The mother's prepregnancy body mass index was calculated based on her height and prepregnancy-weight (body mass index=weight [kg]/ height [m] squared). Birthweights were obtained on pediatric electronic scales (Fillizola, São Paulo, Brazil; calibrated with 10-g precision) shortly after birth, without clothing.Gestational age at birth was obtained by subtracting the date of delivery from the first day of the woman's last menstrual period. Gestational age was defined on weeks14: extremely preterm (<28 weeks), very preterm (28 to <32 weeks) or moderately preterm (32 to 33 weeks), late preterm (≥34 to <37 weeks), and term (≥37 weeks). In the analysis, gestational age was categorized into <34 weeks (including extremely and very preterm) or ≥34 weeks (including late preterm and term).During follow-up in 2016, data entry was performed in the Research Electronic Data Capture application (https://redcap.vanderbilt.edu/), which allows data to be entered at the point of collection. At the same time, the following data were collected from the adolescents: cfPWV, systolic blood pressure, diastolic blood pressure, weight, height, and tobacco use. The adolescents' systolic and diastolic blood pressure were measured using the oscillometric digital OMRON HEM-7421-INT device (Omron Healthcare, Inc, Kyoto, Japan). The mean of the three measurements was used in our analysis, made at 1-minute intervals, with the person seated position for at least 5 minutes and with the dominant arm held in a support so that the brachial artery stayed at heart level.15Next, cfPWV was measured by one trained examiner using a SphymoCor AtCor Medical automated computerized system (AtCor Medical, Sydney, Australia). The cfPWV was obtained after the participant had been at rest in a supine position for 10 minutes in a silent environment with controlled air temperature, with the participant wearing a garment allowed access to the femoral and carotid pulses. Two tonometers were placed on the skin of the most prominent areas of the right common carotid artery and right femoral, enabling the system to obtain the interval of time between the start of the carotid wave and the start of the femoral wave. The carotid-femoral distance was measured with a metric measuring tape, subtracting the distance between the sternum furcula and the carotid site from the distance between the sternum furcula and the femoral site.16 Thus, the cfPWV was calculated as the ratio of the distance between the 2 transducers and the interval of time between the 2 waves. Two cfPWV measurements were obtained and the mean calculated. When the difference between the 2 measurements was >0.5 m/s, a third one was obtained, and the median among them was used.The participants' height and weight were measured by trained staff following standard techniques with the participant barefoot and wearing light clothing. A Bod Pod Body composition measurement scale (CosMed USA, Inc, Chicago, IL) was used for anthropometric measurements; the anthropometer was supported on a wall to enable measurements to be taken in the orthostatic position. The teens' body mass indexes were calculated as above.The proposed theoretical model was constructed to analyze the association between perinatal factors and the VRP at the second decade of life. Perinatal exposures assumed as potential triggers of the VRP were: maternal age, pregestational obesity, gestational hypertension, cesarean section, gestational age at birth, and birth weight. As a potential confounder, the latent variable the socioeconomic status was considered a more distal determinant of perinatal exposures, the VRP, and smoking in adolescence. Obesity in the second decade of life was tested as a mediator because it may be part of the causal chain between perinatal exposures and the VRP (Figure 2).Download figureDownload PowerPointFigure 2. Proposed theoretical model and results to the association between factors of perinatal period to vascular risk phenotype at the end of the second decade of life. 1997/1998 São Luís Birth Cohort, Brazil.Maternal age was used as a discrete-continuous variable; mother's smoking habit (yes or no); mother's prepregnancy body mass index categorized as low (<18.5 kg/m2), eutrophia (18.5–24.9 kg/m2), overweight (25–29.9 kg/m2), or obesity (≥30 kg/m2); self-reported gestational hypertension was based on medical diagnosis (yes or no); type of delivery (cesarean or vaginal); low birth weight was categorized into <2500 g or ≥2,500 g and<3999 g; smoking during adolescence (yes or no); body mass index in adolescence was categorized as low (<18.5 kg/m2), eutrophia (18.5–24.9 kg/m2), overweight (25–29.9 kg/m2), or obesity (≥30 kg/m2); sex was recorded as a dichotomous variable (male or female).Low birth weight (<2500 g) was compared with adequate birth weight (between 2500 to 3999 g). Babies weighing ≥4000 g were excluded from this analysis.Statistical AnalysisStructural equation modeling is an epidemiological tool for testing a hypothetical causal structure of multiple variables, observable, and latent variables, minimizing measurement error in the process of estimation.17 Latent variables are derived from the combination of effect indicator variables, representing the common variance shared among them, resulting in an estimation of effects that are free from the bias originated from measurement errors.17The latent variables, socioeconomic status, and VRP were constructed using exploratory factor analyses to extract factors, representing the effect indicators of the latent variables. Next, both latent variables were submitted to confirmatory factor analysis, using the weighted least square mean and variance adjusted estimation method because the effect indicators of the socioeconomic status latent variable were categorical. In the confirmatory factor analysis, the fit indices values considered to judge model adequacy were: P>0.05 on the χ2; value <0.08 for the upper limit of the 90% CI for the root mean square error of approximation; comparative fit index and Tucker-Lewis index >0.90; and a weighted root mean square residual index value<1.0.17,18 The factor loadings >0.40 was adopted as the minimum value to consider that the indicator contributes something (reflects) to the latent variable.18 All these analyses were performed in MPLUS 7.0 software.The latent variable family socioeconomic status was deduced through the following indicator variables collected at birth: mother's education; father's education; head of the household's occupation; and monthly household income relative to the Brazilian national minimum wage in 1997/1998 (Figure 2).Initially, the latent variable VRP also included the heart rate. However, this indicator was excluded for not showing a factor loading >0.40.18 Thus, the VRP was deduced from the correlations among 3 indicators: cfPWV, diastolic blood pressure, and systolic blood pressure, used as continuous numerical variables (Figure 2).In the structural equation modeling, theta parameterization was used.17 The weighted least square mean and variance adjusted and the same previously described adjustment index values used in confirmatory factor analysis were considered to determine whether the model analyzed by structural equation modeling showed good fit.17,18 The final model evaluated total, direct, and indirect effects between latent and observable variables on VRP, with effect criterion of P<0.05.17Variables were compared between the participants who were followed-up and those who were not with the χ2 test and the nonparametric Mann-Whitney U test (Table S1 in the Data Supplement). To minimize spurious associations arriving from sampling losses, the sample was weighted by calculating the probability of a participant having attended the follow-up as a function of these variables in a logistic regression model. Therefore, the inverse of the probability of selection was calculated for each individual and then used to weight structural equation modeling estimates.ResultsOf the original birth cohort of 2378 participants, 653 (27.4%) participated in the follow-up evaluation at 18 to 19 years of age. Characteristics of the participants at birth and adolescence are reported in detail in Table S2. Their mean systolic blood pressure, diastolic blood pressure, and cfPWV were 116 mm Hg (SD ±12.36), 72 mm Hg (SD ±7.52), and 5.6 m/s (SD ±0.95), respectively (Table S2).The confirmatory factor analysis model of the latent variables socioeconomic status and VRP showed a good fit (Table S3). Similarly, the final model analyzed in structural equation modeling showed a good fit for all of the indexes considered (root mean square error of approximation, 0.036; comparative fit index, 0.952; Tucker Lewis index, 0.924; and weighted root mean square residual, 0.880; Table S3). The VRP showed factor loadings for the indicators of systolic blood pressure (factor loading, 0.937; P<0.001), diastolic blood pressure (factor loading, 0.647; P<0.001), and PWV (factor loading, 0.465; P<0.001) >0.40, indicating good convergent validity (Table 1).Table 1. Factor loadings, SE, and P Values of the Indicators of the Latent Variables Socioeconomic Status and Vascular Risk PhenotypeLatent VariableFactor LoadingsSEsP ValueSES Household income0.7030.030<0.001 Mother's education0.7050.036<0.001 Father's education0.7470.029<0.001 Head of the family's occupation0.7610.038<0.001Vascular risk phenotype SBP0.9370.021<0.001 DBP0.6470.024<0.001 PWV0.4650.026<0.0011997/1998 São Luís Birth Cohort, Brazil. DBP indicates diastolic blood pressure; PWV, pulse wave velocity; SBP, systolic blood pressure; and SES, socioeconomic status.Socioeconomic status did not have a total effect on the VRP (standardized coefficient [SC], 0.008; P=0.884; Table 2). Socioeconomic status had only an indirect borderline effect (P value between 0.05 and 0.10) on the VRP (SC, 0.132; P=0.082), with main pathways via cesarean section (SC, 0.086; P=0.060) and via adolescent obesity (SC, 0.086; P=0.057; Table 2).Table 2. Standardized Coefficients, SEs, and P Values for the Total and Direct Effects of the Structural Equation Model of the Association Between Prenatal and Perinatal Factors Associated With the Vascular Risk Phenotype in AdolescentsVascular Risk PhenotypeTotal EffectsDirect EffectsMain Indirect EffectsSCSEP ValueSCSEP ValueSCSEP ValueSocioeconomic status0.0080.0520.884−0.1240.0890.1640.0860.0470.066*0.0860.0450.057†Mother's age−0.1140.0540.035−0.0890.0710.211−0.0830.0310.008§Maternal smoking0.1410.1180.2340.0350.1240.778………Mother's prepregnancy BMI0.1400.0650.0320.0060.1180.9590.0900.0300.003‖History of gestational hypertension0.1040.1200.3860.1040.1200.386………Gestational age <34 wk0.2590.0850.0020.3540.1490.018………Cesarean section0.1590.0870.0660.1530.0820.061………Male sex0.5610.043<0.0010.5460.042<0.001………Low birth weight0.1460.1270.2480.2220.1440.123………Adolescent smoking−0.0730.0830.3810.0660.0800.409−0.1390.0450.002¶Adolescent BMI0.4700.061<0.0010.4700.061<0.001………1997/1998 São Luís Birth Cohort, Brazil. BMI indicates body mass index; SC, standardized coefficient; and SES, socioeconomic status.*SES to vascular risk phenotype: via cesarean section.†SES to vascular risk phenotype: via adolescent obesity.§Mother's age to vascular risk phenotype: via adolescent obesity.‖Prepregnancy BMI to vascular risk phenotype: via adolescent obesity.¶Adolescent smoking to vascular risk phenotype: via adolescent obesity.Mother's age had a protective total effect (SC, −0.114; P=0.035) on the VRP. High prepregnancy obesity had a total effect (SC, 0.140; P=0.032) on the VRP (Table 2).Gestational age <34 weeks had a total (SC, 0.259; P=0.002) and direct (SC, 0.354; P=0.018) effect on the VRP (Table 2). Cesarean delivery had a total effect, albeit borderline, on the VRP (SC, 0.159; P=0.066; Table 2).Male participants had higher VRP values than females (SC, 0.561; P<0.001). Being overweight at adolescence was strongly associated with the VRP, as evidenced by a very high SC of 0.470 (P<0.001; Table 2).We additionally tested more parsimonious models (using a smaller number of variables) to replicate our main findings that cesarean section and gestational age <34 weeks were associated with the VRP, including analyses stratified by sex (details as given in Tables S4 through S9).In parsimonious model including all individuals, cesarean section had a total effect on the VRP (SC, 0.166; P=0.060) replicating the results of the proposed theoretical model. The effect of cesarean section on the VRP was observed in boys (total effect SC, 0.238; P=0.036; direct effect [SC, 0.247; P=0.020]), but not in girls (Table S6).In the parsimonious models, including all individuals, gestational age <34 weeks had a total effect on the VRP (SC, 0.256; P=0.014) replicating the results of the proposed theoretical model. The effect of gestational age <34 weeks on the VRP was observed in boys (total effect SC, 0.295; P=0.029; direct effect SC, 0.295; P=0.029), but not in girls (Table S9).DiscussionPrenatal and birth factors, such as pregestational obesity, cesarean section, and gestational age <34 weeks were associated with higher VRP values at follow-up in adolescence. Obesity at 18 to 19 years old was directly associated with higher VRP values, as well as being an indirect pathway to a high VRP, through the ancestral variable pregestational obesity.Use of the latent variable VRP as a continuous variable has the advantage of obviating the need for a cutoff value in evaluations of population risk, especially for this sample of apparently healthy adolescents. Systolic/diastolic blood pressure values for our study sample at follow-up (116±12.36/72±7.52 mm Hg) were very near 115/75 mm Hg, the cutoff after which there is a continuous relationship with vascular pathology mortality risk.19 In terms of cfPWV, for each increase of 1 m/s of aortic cfPWV, there is a 14% increment in cardiovascular event risk.20 Hence, these data reinforce the notion that high VRP values reflect real vascular changes tied to future cardiovascular disease risk.Birth by cesarean section was directly associated with higher VRP values. Although the association with cesarean section was borderline (P value between 0.05 and 0.10), its coefficient was one of the highest among the perinatal variables associated with the VRP at follow-up. Previous studies have described an association between cesarean delivery and cardiometabolic risk.9 Cesarean delivery has been associated to increased risk of obesity.9,21,22 However, in our results, cesarean section was directly associated with higher values of the VRP, independently of obesity. As an explanatory mechanism, it may be that the physiological stress of labor during vaginal delivery, which does not occur with cesarean delivery, might be an important stimulus for programming the hypothalamic-pituitary-adrenal axis and supporting the development of adequate neonatal immunologic and cardiovascular systems.23 Furthermore, cesarean delivery may result in changes in microbiome composition and in immune and inflammatory responses, which may be one of the mechanisms triggering higher future cardiovascular risk.24The high cesarean section rate (40%) is in agreement with cesarean rates in Brazil at the 1990s,25 especially represented by the elective cesarean component.26 Our result that cesarean section is directly associated with vascular risk is worrisome because increasing cesarean section rates have been observed around the world.27Birth at a gestational age <34 weeks was associated with higher VRP values. The effect of gestational age was total and direct. Together, these findings are in line with meta-analysis results that showed that preterm birth was associated with subsequent higher systolic blood pressure, even in adolescence.3In more parsimonious models, our main findings of the originally proposed model were replicated. Cesarean section and also gestational age <34 weeks were associated with higher values of the VRP, and these effects were significantly observed in males, but not in females. A possible explanation to the fact that males had higher standardized coefficients of cesarean section and gestational age <34 weeks associated with the VRP than females is that the factors associated with the VRP may be easily observed in males because they tend to develop metabolic abnormalities earlier in life than females and they are more prone to changes related to vascular aging, as increased SBP and cfPWV.28,29Our finding of lower VRP values in females than in males is consistent with a strong cardiovascular protective effect of being female, even in adolescence. It has been noted that women have a lower propensity to develop hypertension and a lower incidence of cardiovascular events.30 This protective cardiovascular effect has been attributed to the actions of estrogen or X chromosome-associated factors on the functional properties of vascular walls.31 Thus, the inclusion of cfPWV as a latent variable may have reflected these properties, contributing to replicate the finding of a female protective effect even in adolescence, even if the arterial stiffness may be influenced by the cyclical fluctuations of female sex steroids over the menstrual cycle.32Higher maternal age had a total effect associated with lower VRP and also had a main indirect effect via protecting adolescent's obesity. As an explanation, in our study older women were more likely to have less obese adolescents (SC, −0.176; P=0.006; data not shown). Furthermore, younger mothers had been pointed as the highest consumers of snack, processed foods, sugar-sweetened cola, and lowest consumers of fruit and vegetable,33 sharing those lifestyle risk behaviors with their offsprings.34Pregestational obesity was associated with higher VRP values, with an indirect effect via adolescent's obesity, supporting the findings that maternal obesity predisposes offspring to obesity and is associated with elevated blood pressure in childhood and adolescence.10 Genetic, epigenetic, or environmental mechanisms, such as obesogenic eating habits and a family's sedentary lifestyle after birth,34,35 might help to explain why the offspring of mothers who were obese during pregnancy are at higher future cardiovascular risk.Obesity at follow-up had the strongest association with high VRP values, supporting a relationship between obesity and cardiovascular disease in adolescence. Obese adolescents have been shown to have increased arterial stiffness, particularly in the central arteries.36 In another study that followed participants for 5 years, obese adolescents had increased cfPWV values (25% versus 3%) and diastolic blood pressure (23% versus 3%), compared with nonobese controls.37Contrary to expectations, socioeconomic status did not have a protective total effect on the VRP.38 Possible high socioeconomic status–mediated protective effects on VRP may have been canceled out by pathways that link higher socioeconomic status to increased incidence of cesarean delivery or excess body weight, both factors were associated with higher VRP in our study. Pregnant women with high socioeconomic status values were more likely to have cesarean deliveries (SC, 0.561; P<0.001; data not shown). Moreover, higher family socioeconomic status at birth was associated with excess weight in adolescence (SC, 0.182; P=0.040; data not shown).The lack of a smoking effect on the VRP in this study may be consequent to the low smoking rates in the population studied given that northeast Brazil is an area of notably low tobacco use.39 Adolescent smoking was indirectly associated with lower vascular risk; although this was an unexpected finding, it may be explained via obesity reduction (adolescent smoking was associated with the vascular risk phenotype, via lower obesity in adolescence [SC, −0.139; P=0.002]; legend Table 2).This study explored, in a pioneering way, the complex interrelations of causality between prenatal and birth environmental factors and the VRP, here proposed as a latent variable to measure vascular changes already occurring in adolescence. This approach, which included blood pressure and arterial stiffness indicators, probably served as a good index of cardiovascular risk outcome in young people because the use of these indicators together may increase their predictive value of future cardiovascular risk.13One limitation of this study was loss of participants to follow-up between birth and the transition to adulthood. Therefore, we weighted structural equation modeling estimates by the inverse probability of follow-up at 18 to 19 years of age to minimize selection biases arising from sampling losses, thereby increasing the study's internal validity.As a recommendation, our findings need to be validated in an independent cohort study, including a test of the reflective indicators of the latent variable VRP.PerspectivesPrepregnancy obesity, gestational age <34 weeks, and cesarean section were the perinatal variables associated with the VRP. These findings are worrisome, since prepregnancy obesity and cesarean section are highly prevalent a