This study is part of the FINNCARE study (23), which is a follow-up of the Finnish Genetics of Pre-eclampsia Consortium (FINNPEC) multicenter study cohort (24). In 2008–2011, 1,450 nulli- or multiparous PE women and 1,065 non-PE women were prospectively recruited along with their partners and newborns. PE was defined as hypertension and proteinuria occurring after 20 weeks of gestation [systolic BP (SBP) ≥ 140 mmHg and/or diastolic BP (DBP)  ≥ 90 mmHg, urinary excretion of ≥0.3 g protein in a 24-h specimen or 0.3 g/L, or two ≥1+ readings on dipstick].

In the FINNCARE study (NCT04676295), PE-exposed and non-PE-exposed families from the FINNPEC cohort living in the Hospital District of Helsinki and Uusimaa were recalled and examined 8–12 years after delivery in a prospective cohort study setting to evaluate their CVD risk profile during preadolescence (Supplementary Figure S1). The exclusion criteria for all mothers included ongoing pregnancy or lactation, multiple pregnancy, and inability to communicate in Finnish. For non-PE-exposed families, the exclusion criteria also included PE, gestational hypertension or chronic hypertension, gestational diabetes, and/or diabetes during or following the index pregnancy. There were no statistical differences in major maternal gestational and child perinatal background characteristics between participating (N = 192) and non-participating (N = 118) PE dyads from the Hospital District of Helsinki and Uusimaa FINNPEC cohort precluding potential recruitment bias (results not shown). Data were collected from 182 PE-exposed and 85 non-PE-exposed children examined in a tertiary care setting at the Clinical Trial Unit located at Children's Hospital, Helsinki University Hospital, Finland, between June 2019 and June 2022. The FINNCARE study received ethical approval from the Ethics Committee of the Hospital District of Helsinki and Uusimaa in December 2018 (HUS/3347/2018), and participation was confirmed with a signed informed consent.

One skilled investigator (TS) obtained the vascular ultrasound images using the Vevo MD system (VisualSonics, Toronto, ON, Canada). The Vevo MD was equipped with electronic transducers (UHF48 and UHF70) corresponding to 30- and 50-MHz center frequencies, and the highest frequency able to image the far wall without compression was applied. Common carotid artery images were obtained 1 cm proximal to the carotid bulb, and femoral artery images were obtained at the inguinal fold. Both were examined bilaterally. Right radial artery images were assessed 1 cm proximal to the skin fold that separates the palma manus from the anterior antebrachium, and right brachial artery images were assessed 2–5 cm proximal to the cubital skin fold. The left radial artery was assessed, in substitution for the right radial artery, in six children because of right radial artery hypoplasia related to an arterial line during the neonatal stage. Bilateral radial arteries in one prematurely born child were hypoplastic, precluding radial artery analyses. Another investigator (MR) blinded to study subject characteristics analyzed the vascular images offline with electronic calipers using the VevoLAB software. IMT and LD during systole and diastole were measured for all blood vessels. Intima-media-adventitia thickness (IMAT) was measured in radial, brachial, and femoral arteries. Adventitia thickness (AT) was calculated as the mathematical difference of IMAT − IMT. We used the leading-edge technique, and all measurements, except systolic LD, were acquired in end-diastole (25). The mean of three caliper measurements was used in analyses. The right and left vessels provided similar results, and means are reported and used in analyses.

We calculated the common carotid artery β-stiffness index (CBSI, local carotid artery stiffness) (26), common carotid artery distensibility coefficient (CDC, local carotid artery distensibility) (26), and common carotid artery wall stress (CWS, local carotid artery wall stress) (27) using the following formulas:CBSI=ln(SBPDBP)/(CCALDS−CCALDDCCALDD)CDC=1000∗((CCALAS−CCALADCCALAD)SBP−DBP)CWS=(MAPxCCALDD)/(2xCCAIMT)

SBP, DBP, and MAP are office systolic BP, diastolic BP, and mean arterial pressure. Office BP was measured following a 1-h rest in the sitting position from the non-dominant arm with the Omron HBP-1300 and HBP-1320 devices. MAP was calculated as DBP + 1/3(SBP − DBP). CCALDS and CCALDD are common carotid artery lumen diameters in peak systole and end-diastole, respectively. CCALAS and CCALAD are common carotid artery lumen areas in systole and diastole, respectively. CCAIMT is common carotid artery IMT.

Child arterial wall layers and CBSI, CDC, and CWS variables are in the analyses used as early proxies of arterial health in our PE-exposed and non-PE-exposed children. Intra-variability coefficients of variation (CVs) were 1.7% (IMT) and 2.13% (LD) for the carotid artery, 5.1% (IMT) and 3.8% (IMAT) for the brachial artery, 3.2% (IMT) and 3.4% (IMAT) for the radial artery, and 3.0% (IMT) and 3.1% (IMAT) for the femoral artery. Inter-variability CVs were 5.2% (IMT) and 2.0% (LD) for the carotid artery, 9.3% (IMT) and 4.3% (IMAT) for the brachial artery, 4.1% (IMT) and 3.1% (IMAT) for the radial artery, and 3.9% (IMT) and 4.5% (IMAT) for the femoral artery.

Body height and weight were measured to the nearest 0.1 cm and 0.05 kg, respectively (Seca 285, Seca GmBH & Co, Hamburg, Germany). We measured waist, hip, thoracic, and head circumferences; arm and leg lengths; midpoint brachial, antebrachial, and calf circumferences; and thigh circumference (at the proximal one-third of the femur) to the nearest 0.1 cm using a tape measure. Bioelectrical impedance analysis (InBody 720, InBody Bldg, Korea) was performed to assess LBM, skeletal muscle mass, fat mass, and body fat percentage. The waist–hip ratio and BMI were calculated, as well as the body surface area (BSA) based on the Haycock formula. Z-scores for height, BMI and weight in relation to height, and weight in relation to age were generated based on a recent Finnish population dataset (28).

BP was measured at 30-min intervals during daytime and 1-h intervals at night for 24 h using an ambulatory BP oscillometric Schiller BR-102 plus device (29). The 24-h BP z-scores were calculated for height (30). Daytime and nighttime were corrected according to individual BP diaries, and 16 daytime and 18 nighttime registrations were excluded because less than 65% of the measurements were valid during the 24-h ambulatory recording (29). Office BP and ambulatory BP devices provided similar results when simultaneously compared in 36 children (results not shown).

Pulse wave velocity (PWV) was assessed at rest from the carotid–femoral (CF-PWV) and the carotid–radial (CR-PWV) regions with the mean of two measurements used in further analyses (Complior Analyse, Alam Medical, Saint-Quentin-Fallavier, France). The carotid–femoral distance was multiplied by 0.8. The Complior device automatically generated the central SBP, DBP, and pulse pressure (PP) based on the carotid waveform, and diastolic and mean office brachial BPs were used for calibration.

Information on current household income (annual), child and parental background diseases, and parental smoking and alcohol intake was collected using standard questionnaires. We acquired maternal and perinatal data from the original FINNPEC database related to index pregnancy (24). Early-onset PE was defined as a PE diagnosis or delivery prior to 34^0/7 gestational weeks and late-onset PE as a PE diagnosis or delivery at or after 34^0/7 gestational weeks (31). Small for gestational age (SGA) was defined as birth weight below −2SD and prematurity as delivery before 37^0/7 gestational weeks. We generated z-scores for birth height, birth weight, and birth head circumference based on a recent Finnish population dataset (32).

We present mean and standard deviation (SD), median and interquartile range (IQR), and count and percentage for normally distributed numerical data, non-normally distributed numerical data, and categorical data, respectively. Histograms, Kolmogorov–Smirnov and Shapiro–Wilk normality tests, Q–Q plots, and skewness were used to check for normality distribution. An independent-sample t-test or Mann–Whitney U-test was used to assess the differences between groups for continuous variables and Pearson χ² or two-tailed Fisher's exact test to assess differences between groups for categorical variables.

We conducted univariate linear regression analyses to evaluate potential predictors for the vascular parameters of children. Potential predictors were organized into different domains in the tables: (1) child sex, (2) child age, (3) child anthropometrics, (4) child adiposity, (5) child ambulatory BP, and (6) child PWV. Only significant associations (p < 0.010) are reported because multiple tests in univariate regression analyses are prone to type 1 error.

Multiple linear regression models were then built to evaluate the influence of BP, PE, and other potential predictors on children's carotid, brachial, radial, and femoral arterial IMT, AT, and LD, as well as CBSI, CDC, and CWS. We included two models for peripheral artery variables to assess the influence of overall body size and local organ size. The unstandardized beta coefficients [and 95% confidence intervals (CIs)] were multiplied by 1,000 for all arterial wall layers (IMT and AT) in univariate and multiple linear regressions, showing the change in micrometers. We checked multiple linear regression models for normality, linearity, homoscedasticity, and independence. Multicollinearity was assessed with variance influence factor (VIF) and collinearity tolerance (CT), and VIF < 2.5 and CT > 0.3 were considered appropriate. Two-tailed tests were used in analyses, and a p-value of <0.05 was considered significant. SPSS v. 27 (IBM, New York, USA) was used for statistical analysis.

A total of 182 PE-exposed children and 85 non-PE-exposed children attended the follow-up visit, with four children being 8 years old; 24 children 9 years old; 61 children 10 years old; 81 children 11 years old; 75 children 12 years old; and 22 children 13 years old (mean age 11.4 years, results not shown). In total, 46 children were classified as early-onset PE based on diagnosis and 25 children based on the birth criteria (Table 1). PE-exposed children were more often premature and born SGA compared with non-PE-exposed children, which was accentuated in early-onset PE-exposed children.

The LD, IMT, and AT values of carotid, brachial, radial, and femoral arteries were not statistically significantly different between PE-exposed (including early-onset and late-onset PE-exposed) and non-PE-exposed children (Table 2). CBSI, CDC, and CWS values were not different between PE-exposed (including PE-exposed subgroups) and non-PE-exposed children. Early-onset and late-onset PE-exposed children by delivery definition provided similar results for vascular structure and function (results not shown).

Tables 3, 4 and Supplementary Tables S1, S2 show the univariate analyses exploring potential predictors of the arterial structure and function of all children. Brachial, femoral, and radial artery IMTs were associated with both ambulatory (24-h, daytime, and nighttime) and office central SBP and PP, while there were no associations with common carotid artery IMT and ambulatory BP or office central BP (Table 3). Radial artery IMT was also correlated with CF-PWV. Brachial and femoral artery IMTs were significantly related to general (height, weight, and thoracic circumference) and local (arm/leg lengths and circumferences) child anthropometric measures, with LBM and muscle mass showcasing the strongest associations. Radial artery IMT was associated with height, weight, BSA, head circumference, and leg and arm lengths and was strongest with LBM and skeletal muscle mass. Common carotid artery IMT was significantly associated with head circumference and height. Brachial and radial artery IMTs were higher among males. Brachial and femoral artery IMTs showed significant associations with adiposity measures BMI and BMI z-score, and brachial artery IMT with fat mass and waist–hip ratio. Radial artery IMT was only associated with fat mass percentage. However, no associations between adiposity and common carotid artery IMT were found. The relationship between peripheral artery AT and anthropometric measurements was weaker compared with IMT. However, significant associations were found between the femoral artery AT and BMI z-score, brachial artery AT and central SBP and PP, and radial artery AT and 24-h (including daytime and nighttime) SBP and central SBP and PP (Supplementary Table S1). Similar to IMTs, arterial lumen dimensions were strongly related to general and local child anthropometric measures, LBM and muscle mass, and BMI, waist–hip ratio, fat mass, and fat mass percentage (Supplementary Table S2). However, consistent relations between lumen dimensions and ambulatory BPs were not found.

CDC and CWS parameters were associated with adiposity parameters, including waist circumference, waist–hip ratio, BMI, BMI z-score, fat mass, and fat mass percentage (Table 4). CDC was also associated with anthropometrics, including body weight, BSA and head circumference, and CWS with head circumference only. CDC further showcased borderline significance with child age. CWS was the only local carotid artery parameter associated with CF-PWV. CBSI was only borderline associated with waist circumference and waist–hip ratio, while no consistent associations with other child anthropometrics were found. However, when data were analyzed in children exposed to PE only, CBSI was associated with BMI, fat mass, and fat mass percentage (results not shown). Parameters for local carotid artery stiffness were not associated with sex.

Table 5 and Supplementary Table S3 show multiple linear regression models assessing the influence of multiple predictors on arterial structure and local carotid artery function in PE-exposed and non-PE-exposed children combined. All peripheral arteries’ IMTs, including brachial, radial, and femoral, were independently predicted by child LBM and 24-h SBP at the follow-up (Table 5). LBM displayed the highest standardized coefficients in all peripheral artery IMT models (brachial artery IMT model's adjusted R² 0.217, radial artery IMT model's adjusted R² 0.208, and femoral artery IMT model's adjusted R² 0.214). The local organ size performed similarly to LBM, or slightly less, and remained significantly related to the IMTs of all peripheral arteries. Furthermore, brachial and radial artery IMTs were higher in males, and radial artery IMT was related to child age at follow-up. PE was non-significant in all models. The femoral artery IMT model was stronger using 24-h SBP compared with office central SBP (LBM model: adjusted R² 0.214 vs. 0.127; local organ size model: R² 0.172 vs. 0.070). Like IMTs, peripheral artery LDs were independently predicted by sex, LBM, and local organ size (Supplementary Table S3; brachial artery model adjusted R² 0.431, radial artery model adjusted R² 0.211, and femoral artery model adjusted R² 0.394). PE was not related to LDs in any models. Central SBP was not a significant predictor of vascular wall layer thickness or LD (results not shown).

CBSI, CDC, and CWS were all independently predicted by the waist–hip ratio at follow-up (Table 5, model's adjusted R² 0.024, 0.066, and 0.047, respectively). CDC and CWS were also independently predicted by body fat percentage with model's adjusted R² 0.036 and 0.043, respectively (results not shown). CDC was independently predicted by child age at follow-up, and for CBSI, age reached borderline significance as a predictor. CWS was predicted by the child head circumference.

Multiple significant associations were observed between carotid and peripheral artery LDs and IMTs and different birth anthropometrics (height, weight, and head circumference), but birth anthropometrics did not remain significant when adjusting for child anthropometrics at follow-up (results not shown). Similarly, we found no independent associations between LDs and IMTs at follow-up and maternal pre-pregnancy BMI, age at delivery, parity, maternal gestational BPs, child prematurity, or SGA (results not shown). Similarly, no relations were seen for CBSI, CDC, and CWS (results not shown).