Introduction

Based on the Developmental Origins of Health and Disease (DOHaD) theory, early life events can increase the risk of noncommunicable diseases in adulthood [1]. Nutrition-related and other stressors during pregnancy may result in low birth weight (LBW), considered a proxy of an adverse intrauterine environment [2] and commonly the result of intrauterine growth restriction or prematurity [3]. Given the short- and long-term adverse outcomes in LBW neonates, this condition remains a major public health concern. The World Health Organization (WHO) estimates that LBW occurs in 15–20% of all births, with rates varying widely according to country income level [4]. In Brazil, according to the Information System on Live Births–SINASC [5], LBW rates range from 7.4% in the Midwest to 8.4% in the Southeast region of the country. Compelling evidence has indicated that LBW may predict cardiometabolic and renal disorders [6, 7].

Perinatal distress has been associated with blood pressure elevation in adulthood [8]. Underlying mechanisms for the association of early life events with blood pressure elevation have been proposed in animal and human models [9, 10]. Changes in renin-angiotensin system gene expression, and activity of the angiotensin-converting enzyme and sympathetic nervous system, may contribute to the development of hypertension in adult life [11,12,13]. Sustained increased blood pressure levels with gradual glomerular damage may precipitate chronic renal failure.

Nephrogenesis begins at early stages of embryonic life with an increase in glomeruli number in the last weeks of gestation [14]. Insults during intrauterine life (undernutrition, toxic exposures and/or placental abnormalities, as well as maternal health-related factors) [15] can negatively impact renal functional reserve. These adverse events can induce fetal programming with resultant reduction in nephron formation and hypertrophy of the remaining nephrons, leading to adaptive hyperfiltration and increased risk of kidney dysfunction later in life [16, 17]. An early sign of hypertension-dependent glomerular damage is urinary protein loss [18]. Issues regarding whether LBW adults born at full-term or prematurely are prone to exhibit different long-term consequences, and whether these changes can be detected early in the natural history of hypertension and chronic kidney disease, remain understudied. Even with the available data pointing to kidney damage, such as a decrease in the number of nephrons or compensatory glomerular hypertrophy observed in the offspring of mothers who had insults during pregnancy, more evidence on their impact in functional reserve and structural renal changes are needed. The Brazilian Longitudinal Study of Adult Health (ELSA-Brasil) is a large cohort that provides an opportunity to investigate the relationship of exposures with several outcomes in adulthood [19, 20]. In particular, this allows the testing of associations of early life events with blood pressure levels and renal function markers in adults without overt kidney diseases. We examined the association of LBW and prematurity with blood pressure levels and kidney function markers in middle-aged, non-diabetic participants of ELSA-Brasil without kidney disease.

Methods

Study design and population

A cross-sectional analysis was carried out of the baseline cohort from the ELSA-Brasil, a multicenter cohort study aimed at investigating biological, behavioral, environmental, occupational, psychological, and social factors related to incidence and progression of diabetes and cardiovascular disease. Methodological details have been reported elsewhere [19, 20]. Briefly, from August 2008 to December 2010, ELSA-Brasil recruited employees aged 35–74 years working in universities and research institutions located in the Northeast, Southeast and Southern regions of Brazil, with varied socioeconomic conditions and skin color distribution. Sample size of ELSA-Brasil was calculated based on estimations of incidence of type 2 diabetes mellitus and myocardial infarction for the Brazilian population. The current study involved a random sample of 998 individuals without diabetes or cardiovascular disease aged 35–54 years, drawn from 5,061 participants at the São Paulo center of the ELSA-Brasil for whom information regarding serum cystatin C (sCys C) was available [21]. The sample showed the same demographic and socioeconomic distributions. The study was approved by the local Ethics Committee and informed consent was obtained from all participants.

Eligibility criteria

Of the initial 998 participants, 17 with newly diagnosed diabetes and 16 with low estimated glomerular filtration rate (eGFR < 60 mL/min/1.73 m2) were not included. Fifty individuals born with macrosomia, 18 with thyroid dysfunctions (TSH < 0.1 or ≥ 10.0 µUI/mL), eight underweight individuals due to consumptive syndrome (unwanted weight loss higher than 10% without apparent cause) and two with liver dysfunctions (ALT or AST levels three-fold the normal range) were excluded. Since macrosomia confers potential risk of metabolic disorders and our aim was to specifically analyze low birth weight and prematurity, macrosomia was an exclusion criterion. Participants with missing data regarding exposure (birth weight) or outcomes "(sCys C, estimated glomerular filtration rate [eGFR] and albumin-to-creatinine ration [ACR]) were also excluded (n = 119), as shown in Fig. 1. The excluded participants did not change the main composition of the final sample. A total of 768 participants were included in the present analysis.

Fig. 1
figure 1

Flowchart of selection process of ELSA-Brasil participants in the present study

Full size image

Clinical and laboratory data

Interviews and anthropometric data were collected by trained personnel using standardized questionnaires [22]. Prematurity (yes or no) and birth weight were self-reported since birth records were not available for the entire sample. Participants were asked to report their weight at birth and the variable was classified into “ < 2.5 kg”, “2.5–4.0 kg”, “ > 4.0 kg” or “unknown”. All participants were asked to provide their specific birth weight and to recall their body weight at 20 years of age.

Sociodemographic and health factors of interest for this study were: age (years), sex (male or female), self-reported skin color (black, white, brown, yellow, and indigenous), parental history of diabetes and hypertension (yes or no) and educational levels of the participant and their mother. The latter was used as a socio-economic trajectory marker since it was available in the first wave of ELSA-Brasil. For the purpose of the present study, skin colors were grouped into white and non-white categories.

Blood pressure was measured using an Omron HEM 705CPINT device (Omron Co, Kyoto, Japan) after a 5-min rest in a sitting position. Three measurements were taken at 1-min intervals and mean values calculated. Body mass index (BMI) was calculated as weight in kilograms divided by height in meters squared. After overnight fasting, blood and urine samples were obtained, and participants then underwent a 2-h 75 g oral glucose tolerance test. Aliquots of biological samples were frozen at − 80 °C for further analyses [23, 24].

Fasting and 2-h post-load plasma glucose levels were measured using the hexokinase method (ADVIA Chemistry; Siemens, Deerfield, Illinois, USA) and glycated hemoglobin was evaluated by high-pressure liquid chromatography (HPLC) (Bio-Rad Laboratories, Hercules, California, USA), according to the National Glycohemoglobin Standardization Program (NGSP) certified method. Glomerular filtration rate was estimated using the equations proposed by the Chronic Kidney Disease Epidemiology Collaboration (eGFR CKD-EPI) [25]. Albuminuria was determined in a 12-h overnight urine sample by nephelometry and expressed as ACR. Urinary creatinine was measured using the kinetic Jaffe method (Advia 1200 Siemens, USA). Participants were instructed to avoid strenuous exercise and alcohol consumption for 24 h preceding the clinical visit. Serum cystatin C was measured using a human cystatin C ELISA kit (Elabscience Biotechnology, Houston, Texas, USA). Intra- and inter-assay coefficient of variability ranges were 5.05–5.38% and 3.29–6.48%, respectively.

Definitions

Prematurity was defined by an affirmative answer to the question: “Were you a premature baby, that is, were you born earlier than expected?”. Birth weight was classified into three categories: low birth weight (< 2.5 kg), normal birth weight (2.5–4.0 kg) and macrosomia (> 4.0 kg). Outcomes were blood pressure levels, eGFR, ACR and sCys C, and were analyzed as continuous variables.

Hypertension was diagnosed when systolic or diastolic blood pressure levels were ≥ 140 or 90 mmHg, respectively, or when the individual was on antihypertensive medication. Kidney function was considered altered when eGFR < 60 mL/min/1.73 m2 or urinary albumin excretion > 1000 mg/g creatinine, according to KDIGO 2012 [26]. Normal ranges of sCys C were 0.64–0.84 mg/L for men and 0.57–0.74 mg/L for women [27].

Diabetes and pre-diabetes were defined according to the ADA (American Diabetes Association) criteria [28]. Diabetes was diagnosed when fasting plasma glucose was ≥ 126 mg/dL or 2-h post 75-g glucose load > 200 mg/dL or glycated hemoglobin ≥ 6.5%. Prediabetes (yes or no) was defined as fasting plasma glucose between 100 and 125 mg/dL or 2-h post 75-g glucose load between 140 and 199 mg/dL or glycated hemoglobin 5.7–6.4%. Nutritional status was stratified by BMI into underweight (< 18.5 kg/m2), normal weight (18.5 and 24.9 kg/m2), overweight (25 and 29.9 kg/m2) and obese (> 30 kg/m2). Insulin sensitivity was assessed using the Homeostasis Model Assessment of Insulin Resistance (HOMA-IR) and considered a continuous variable [29].

Statistical analysis

Continuous variables with normal distribution were expressed as mean ± standard deviation (SD) and compared using Student’s t-test. Non-normal distributed parametric variables were expressed as median and interquartile range and compared using the Wilcoxon rank test. Categorical variables were expressed as absolute and relative frequencies and compared using the chi-squared test.

Associations of LBW (exposure) with outcomes (blood pressure levels and kidney function parameters) were initially analyzed by simple linear regression. Directed Acyclic Graphs (DAGs) were used to build theoretical models to analyze independent associations of exposure with outcomes in multiple linear regression analyses. The DAG is a causal diagram which allows the input of scientific evidence regarding the relationships among variables in graphics software to reach the ideal set of covariables (minimum sufficient adjustment) for the model to prevent bias and overadjustments [30, 31]. For the construction of the DAG for blood pressure and kidney function, several confounder variables were considered, such as age, sex, skin color, weight at age 20, maternal education level, prematurity, income, current BMI, smoking and others as shown in Figure S1 of the Supplementary material (Panels A and B). The figures were created using DAGitty software, version 3.0 (www.dagitty.net).

Having applied the DAG, the covariables selected for adjustments in the linear regression models regarding the association of LBW with blood pressure levels were: race/skin color, sex, parental history of hypertension, and prematurity. Linear regression models for the associations of LBW with kidney function parameters (eGFR, ACR and sCys C) were adjusted for skin color and prematurity.

Considering the contrasting sample sizes of the groups with normal and LBW, as well as possible selection bias due to the observational nature of the study, propensity score matching was employed to create more comparable groups [32, 33]. The nearest neighbor-matching algorithm, within a caliper of 0.1 SD of logit function of propensity score was used. Firstly, to apply the propensity score matching, a multiple logistic regression model was used, adjusted for DAG-based covariates (blood pressure was adjusted for skin color, sex, parental history of hypertension and prematurity, while kidney function markers were adjusted for skin color and prematurity), and the probability of each participant being LBW versus normal birth weight was estimated. Balance between the groups was assessed by comparing each covariate. When standardized mean difference lay in the –0.1 to 0.1 range, groups were considered balanced. This matching was able to reduce all covariate imbalance in the sample. Using the matched sample, multiple linear regression analyses were then performed to test associations of LBW with blood pressure and kidney function markers (outcomes) adjusted for the same DAG-based covariates. Tests were performed using “MatchIt”, “rbounds”, “Matching”, “twang” and “survey” packages in the R statistical environment.

Additionally, sensitivity analyses were performed by selecting the participants with self-reported preterm birth (prematurity). Also, when blood pressure levels were outcomes, sensitivity analyses were carried out excluding participants who reported use of antihypertensive agents (Supplementary Table S1). Since similar results were obtained in both analyses, preterm and hypertensive participants were retained for subsequent analysis.

All analyses were performed using R Project for Statistical Computing (R version 3.5.2) and statistical significance was set at a p value of 0.05.

Results

In the sample of 768 participants, mean age was 45.5 ± 4.6 years, 56.8% were female, and 60% reported white skin color. A total of 64 (8.3%) participants reported being LBW and 39 (5.0%) born preterm. The LBW group comprised participants that were predominantly female (54.6%), white skinned (56.0%) and had mothers with low educational level (71.9%). These characteristics showed a similar pattern in the preterm subgroup (67.0% female and 61.0% white skin color).

A total of 133 participants fulfilled the diagnostic criteria for hypertension, which was more frequent in the LBW than in the normal birth weight group (16.2% versus 29.7%, p = 0.024). Among the hypertensive participants in the normal and LBW groups, 8.1% and 14.0% were receiving antihypertensive treatment, respectively.

Overall, cardiometabolic risk profile and kidney function of the total sample were within normal ranges, except for being generally overweight (mean 26.0 kg/m2 ± 4.15) and having borderline fasting plasma glucose (102.0 mg/dL ± 7.5). Mean values of systolic (116.7 ± 14.7 mmHg) and diastolic (74.9 ± 10.5 mmHg) blood pressure, eGFR (89.6 ± 13.8 mL/min/1.73 m2), ACR (11.0 ± 39.0 mg/g creatinine) and sCys C (0.74 ± 0.61 mg/L) were within normal ranges.

Low birth weight participants were, on average, older than normal birth weight subjects (47.0 ± 5.5 years versus 45.3 ± 4.8 years, p = 0.027) and had a higher prevalence of low educational level (6.3 versus 3.7%, p = 0.025). Mean BMI values were lower at age 20 (21.2 ± 2.81 versus 20.7 ± 3.0 kg/m2, p = 0.196) than at current age (26.2 ± 4.1 versus 26.0 ± 4.4 kg/m2, p = 0.688) for both normal and LBW participants, respectively, with no difference between subgroups.

Participants with LBW had higher mean blood pressure and ACR, and lower eGFR, than individuals with normal birth weight, but there were no group differences for sCys C (Table 1).

Table 1 Main data of ELSA-Brasil participants born with normal and low birth weight
Full size table

The preterm subgroup also had significantly higher mean systolic and diastolic blood pressure levels, but no difference in kidney function markers was detected (Supplementary Table S2).

In multiple linear regression analysis, LBW was associated with blood pressure levels, but this association did not persist after adjustment for prematurity, which remained associated with systolic and diastolic blood pressure (Table 2). Associations of LBW with eGFR and ACR in the fully adjusted models had borderline significance (p = 0.05) (Table 3). In a separate analysis of only preterm participants, prematurity was independently associated with both systolic and diastolic blood pressure (Supplementary Table S3).

Table 2 Association of low birth weight with systolic and diastolic blood pressure
Full size table
Table 3 Association of low birth weight with estimated glomerular filtration rate and albumin-to-creatinine ratio
Full size table

Imbalance of variables

Initially, the sample had 64 LBW participants. Propensity score matching was applied, yielding sub-samples with matched participants (normal and LBW) as follows: (a) 48 matched participants for systolic and 61 matched participants for diastolic blood pressure; (b) 44 matched participants for eGFR; and (c) 49 matched participants for ACR with the same proportion of normal and LBW participants in each sub-sample. The matching approach rendered all covariates appropriately balanced (standardized mean difference of between –0.1 and 0.1) for further analyses. Imbalance for all variables was assessed before and after matching to compare improvement by matching (Supplementary Table S4).

Associations of LBW with blood pressure and kidney function markers after propensity score matching

The multiple linear regression model, adjusted for skin color and prematurity, showed that being born with low weight was directly associated with ACR values (β 1.34; 95%CI 0.47–2.20; p = 0.003). Even when reaching adequate balance, LBW was not associated with eGFR or with systolic and diastolic blood pressure levels (Table 4).

Table 4 Estimates of associations of LBW with blood pressure and kidney function markers after propensity score matching in ELSA-Brasil participants
Full size table

Discussion

We hypothesized that being born with LBW is associated with higher blood pressure levels and impaired kidney function markers in adulthood, indicating an insult during intrauterine life. Our findings corroborate this hypothesis since we found that all LBW participants showed higher blood pressure levels, and exhibited lower eGFR and higher ACR than those with normal birth weight. Enhancing our analysis by propensity score matching and DAG-based adjustments on multiple linear regression boosted confidence in the associations found between early-life events and outcomes in adulthood. We suggest that individuals with LBW may be at increased risk of renal abnormalities even when routine parameters are within normal values.

Hypertension is a major cause of chronic kidney disease (CKD) and both are prevalent disorders globally [34, 35]. Additionally, these conditions increase the risk for cardiovascular events and deaths [36]. Therefore, several studies have sought to identify predictors for the prevention of these diseases. There is evidence that the pathophysiological process starts in early life, congruent with the DOHaD theory [37, 38]. Our main results are in line with the DOHaD, which holds that insults in fetal life induce programming and that LBW is a surrogate of this condition. An interesting finding of the present study was that prematurity was associated with blood pressure levels, yet LBW was not. Previous studies have reported an association of prematurity with elevated blood pressure levels [39, 40]. Proposed mechanisms linking prematurity to higher blood pressure levels in adult life involve pre- and postnatal factors. Premature birth has been associated with increased vascular resistance [41, 42], endothelial dysfunction [43], immature autonomic blood pressure regulation [44] and high sympathetic nervous system activity [13]. The extent of these abnormalities is dependent on the fetal adaptations and cardiovascular system programming in response to an adverse intrauterine environment during a critical period of development [45]. It is known that the full number of nephrons is achieved at between 32 and 36 weeks of gestation and that glomerular filtration starts during intrauterine life [14, 46]. Therefore, prematurity may cause morphological and functional renal changes that lead to hemodynamic disturbances and hypertension [47]. Most preterm infants have accelerated postnatal growth during the first years of life [48], where this compensatory catch-up growth is commonly associated with rapid body weight gain and obesity in childhood [49]. We speculate that the overweight participants in the present study have an increased risk of hypertension later in life.

Several nutritional, toxic or emotional (maternal anxiety or depression) determinants of LBW – regardless of gestational age at delivery – can impact adult kidney function [50,51,52], assessed by a number of parameters such as GFR, serum creatinine, sCys C and albuminuria in clinical practice. Elevated ACR is a recognized biomarker of kidney dysfunction, even when eGFR is within normal ranges [53]. In the present study, mean values of eGFR and ACR were significantly worse in the LBW group compared to the other groups.

Furthermore, having employed propensity score matching, a strong association of LBW was found with albuminuria, but not with eGFR or sCys C. We emphasize that ACR might be a useful marker for early detection of kidney dysfunction in adults born with LBW. The higher albuminuria seen in LBW individuals might have been due, in part, to fetal glomerular morphological alterations and prenatal programming. This notion is supported by evidence demonstrating that LBW individuals have fewer nephrons [54], lower glomerular volume at birth [55], abnormal glomeruli structure with enlarged Bowman’s capsule and altered glomerular tufts [56, 57], factors that, in the long-term, could result in increased glomerular permeability and albuminuria. The present sample of LBW individuals without overt nephropathy may have had glomerular hypertrophy of the remaining nephrons and hyperfiltration. The current mean eGFR of this group was within the normal range, as was albuminuria, but remained significantly lower than the group with normal birth weight. Whether ACR, periodically measured, serves as a suitable marker of early kidney dysfunction in LBW individuals requires investigation by appropriately designed studies. This is relevant considering the synergistic impact of the nephron mass on the development of hypertension favoring CKD and cardiovascular disease later in life.

We elected to measure sCys C based on previous reports suggesting its greater accuracy than creatinine for detecting decreased renal function [58,59,60]. SCys C determination, combined with creatinine, can be used in GFR estimation (KDIGO 2012) [26]. However, in the present sample with normal eGFR, sCys C failed to detect kidney dysfunction in LBW individuals.

This study has limitations in relation to recall bias, given that data about early life events as well as body weight at 20 years were collected retrospectively. These self-reported data could not be validated due to the non-availability of birth records or other validation methods, such as medical records, for all participants. However, our sample cohort was middle-aged and previous studies have shown that perinatal-related events can be accurately reported during adult life [61,62,63]. We also used the exposure variable (LBW or prematurity) as a categorical variable, thus this may have reduced the statistical power of the analysis. This variable was chosen so as to minimize information bias from participants unable to accurately report their birth weight. The frequency of self-reported prematurity in our sample was comparable to the rate reported in the Brazilian population at large [5]. Despite being non-representative of the general Brazilian population, employees enrolled in ELSA-Brasil represent a heterogeneous sample regarding socioeconomic and ethnic-racial background [19]. As far as ACR is concerned, the lack of a direct measurement of muscle mass in which creatinine is generated could be a limitation. Taking into consideration that the distributions of BMI values in each subgroup were very similar, and the lack of difference between mean urinary creatinine between normal birth weight and LBW groups, it is unlikely that urinary concentrations of creatinine interfered in our ACR results. Therefore, we considered ACR measurement as a reliable method for the assessment of proteinuria in our sample [64].

A strength of the present study was its methodological approach. DAG was employed to avoid overadjustments [30, 31] when performing the regression analysis. However, although several covariates were incorporated into our theoretical model, including body weight at age 20, other exposures throughout the participants’ life course were unavailable. Size difference of the groups with normal birth weight and LBW was also a concern; therefore, propensity score matching was applied to reduce potential selection confounders from observational studies [32, 33] and sufficiently balanced groups were achieved.

In conclusion, prematurity was found to be associated with higher blood pressure levels and LBW with albuminuria in adulthood. Considering the magnitude of the ELSA-Brasil and potentialities of its cohort [20, 24], our findings are important to raise awareness of early life events and to be alert to a subset of healthy adults born with LBW who are at increased risk for hypertension and kidney dysfunction. Long birth cohort studies could help confirm the hypothesis raised in this study.