**4. Discussion**

To our knowledge, this is the first study to describe the association of growth-related changes in body size during early, intermediate, late, or global growth with hemodynamic (central and peripheral) and arterial (structural and functional) properties in early childhood and beginning of adulthood, adjusting for characteristics at birth, at the time of the CV study and for CRFs exposure. Associations were assessed considering three interrelated comparative analysis: (1) strength and (2) independence of the associations and (3) explanatory power of anthropometric parameters and factors associated with CV risk. Our main findings were:


were mainly explained by growth-related body changes and/or by their interaction with current z-BMI (Tables S4 and S6). Similar results were observed when the associations were analyzed considering the exposure to CRFs (e.g., hypertension, dyslipidemia) (Table 9). In turn, current z-BMI was the anthropometric variable with the greatest explanatory capacity for CV conditions and variations at 18 y.In summary, body size changes during childhood and/or adolescence contributed to explain arterial variations through the interaction with current z-BMI or BWH z-score at birth (Tables S8, S10 and S12). Among factors associated with CV risk, z-BMI and/or z-pSBP were the main explanatory variables for CV z-scores (Table 10).

The explanatory capacity of growth-related body changes was reduced or lost with, as the age at which the CV system was evaluated increased. In this regard, at least two issues must be analyzed. On the one hand, the impact (association) of anthropometric changes on the CV properties evaluated at a given age, would vary, depending on the period of "body change or gain" considered (e.g., 0–2 y vs. 0–6 y). On the other hand, the association between CV properties and the anthropometric changes observed in a given period (e.g., 0–6 y) could depend on the age at which the CV system is evaluated (e.g., 6 vs. 18 y, like in this work). Studies suggested that BW gain patterns in very early infancy (e.g., 0–6 postnatal mos.) [22–24] would be particularly important as determinants of future pBP levels, while other studies showed that BW gain during childhood would be a stronger predictor of pBP [25,26]. The exact "timing of the BW gain" associated with middle or long-term CV risk is still debated, highlighting the need for additional research to clarify and/or reconcile mixed findings. In this work, in general terms we did not find di fferences in the strength of association when comparing 0–2 vs. 0–6 y data, but some CV characteristics assessed at 6 y showed greater association with 0–6 y anthropometric changes. Evelein et al. (2013) reported that postnatal BW for length gain (0-3 mos.) was associated with carotid IMT (but not sti ffness) in children (5 y) [21]. However, when data about growth in later infancy (3–6, 6–9 and 9–12 mos.) were considered, no associations with arterial properties were found [21]. Skilton et al. (2013) reported that BW gain, BH-adjusted-BW gain and ΔBWH z-score 0-18 mos. were positively associated with carotid IMT assessed at 8 y [27]. Unfortunately, the impact of changes in di fferent periods was not analyzed. Linhares et al. (2015) found that the "adverse" long-term e ffect of accelerated growth in infancy varied depending on the time of growth acceleration. Particularly, carotid IMT at ~30 y was associated with 2–4 y BW-gain, rather than with early BW gain [7]. Additionally, Vianna et al. (2016), found that the relative BW-gain between 2 and 4 y was associated with increased aortic sti ffness (evaluated by cfPWV) at ~30 y, whereas birth weight, BW-gain within the first 2 years of life (0–2 y) and linear growth (length/height gain) in childhood were not associated with cfPWV [28]. Pais et al. (2016) assessed PWV in children (8–9 y) and analyzed data considering and categorizing growth trajectories. The highest arterial sti ffness levels were observed in groups with accelerated body growth during childhood, with adequate early growth pattern [29].

The dependence on the age at the time of CV study ascribed to the association between an anthropometric parameter and CV properties has been previously described, mainly for birth weight. About this, body size at birth has been associated with pBP levels, andit was reported that the relation becomes progressively stronger with increasing age, being hypothesized that the initiating process occurs in uterus and amplifies throughout life [11,30,31], satisfying theories that seek to explain the detriment of the CV system related to low birth weight and/or catch-up growth [32]. It was even postulated that interactions between increased arterial sti ffness, increased pPP, stretching of vascular smooth muscles and synthesis of collagen may contribute to the amplification phenomenon through a feedback loop [33]. By contrast, Lule et al. analyzed data from studies that measured pBP at di fferent ages and did not find an age-related increase in the strength of the association between birth weight and pBP [6]. Furthermore, the relationship between birth weight and later pBP varied depending on the age of the participants: neonates showed consistent positive association; mainly negative associations were seen in children, and studies in adolescents showed inconsistent results [6]. Then, as age increases, the positive association observed in neonates could become negative, non-existent or even positive, which is in agreemen<sup>t</sup> with our findings. This could be explained, at least in theory by the fact that as age increases the exposure-time to already present factors capable of impacting on the CV system also increases. As age increases, subjects could become exposed to factors (i.e., CRFs) capable of modifyingCV properties. Then, the association between anthropometricchanges and CVproperties could be modified by exposure to co-factors.

As mentioned in 6 y children the "hierarchical order" among explanatory variables for CV variations would be: current z-BMI >ΔBWH z-score 0–2 y or Δz-BMI 0–6 y<sup>&</sup>gt; BWH z-score at birth (Table 2). Conversely, only current z-BMI showed significant association with arterial properties at 18 y (Table 3). Birth weight showed almost no association with CV properties, disregard of the subjects' age at the time of the CV study (Tables 2 and 3), and most of the associations between ΔBWH z-score 0–2 y or Δz-BMI 0–6 y and CV characteristics at 6 ywere independent of birth conditions (Tables 5 and 6). Then, the association between body-size changes during infancy or childhood and the CV system at 6 y, would not depend on having been born with low, normal or elevated BWH. When current z-BMI was considered some associations between bodily changes in childhood and CV properties at 6 y were no longer significant. In adolescents, the associations between body changes and CV variables were always dependent on z-BMI at the time of CV study (Tables 6–8).

The dependence (or independence) of the association between CV parameters and growth–related anthropometric changes on bodysize at birth and/or on current z-BMI has been previously assessed, with dissimilar findings. A positive association was observed between BW gain or adiposity accumulation during childhood and later pBP levels [23,34–37]. However, the extent to which birth size modifies the associations between postnatal growth and future pBP levels and/or arterial properties remains unclear. Belfort et al. (2007) found that infants who were thinner at birth were more susceptible to adverse effects on pBP at 3 y of accelerated BWH gain within the first 6 postnatal mos. [22]. Whether the finding is extensive to mid-childhood when BP is highly correlated with adult BP [38] is to be clarified. Leunissen et al. (2012) showed that regardless of birth-size, adiposity accumulation during childhood is a risk factor for later (~20 y) development of high BP levels [37]. Accordingly, Kelishadi's review (2014) concludes that early growth, rather than birth weight, would be important as a determinant of later BP levels [32]. Supporting a BMI-independent association between body size changes and CV properties, Thiering et al. studied children (*n* = 1127, age≤10 y) and reported that higher BW peak (velocity) in infancy was associated with an increase in pSBP and pDBP after confounders adjustment [39]. In contrast, it has been proposed that the association between BMI at adiposity peak and BP at 6 y would be mediated by current BMI [40]. Marinkovic et al. (2017) observed that infant peak BW velocity and BMI at adiposity peak associationwith childhood pSBP and pDBP (at 6 y), which could be explained by current BMI [12].

Our results support the proposal that the association between anthropometric parameters and pBP depends on current BMI, at the same time as they provide original information showing that unlike what was described for pBP, the association between growth-related body size changes (0–2 y) and cBP at 6 y would not depend on current z-BMI. This is further supported by the fact that reflection parameters, which are main determinants of the di fferences between cBPand pBP, also showed associations not explained by current BMI (Tables 5 and 6). It is to note that compared to pBP, the cBP would be of greater value in terms of association with CVchanges and risk prediction [41].

As stated above, the association between growth-related body-size changes and arterial thickness has been previously described [21,27,42]. Our results provide additional information, showing that at 6 y, the association between body size gain and thickness is independent of body size at birth and current z-BMI at 6 y, and that it is statistical significant for both carotid (elastic) and femoral (muscular) arteries. In turn, Evelein et al. (2013) described interaction between birth size and postnatal weight for length by analyzing the impact on arterial sti ffness (i.e., distensibility and arterial elastic modulus). The thinner the children were at birth, the lower the distensibility (greater the elastic modulus) with increasing weight for length gain [21]. Then, the impact of birth-size and or growth-related changes would vary depending on the CV properties considered.

Finally, as mentioned, current z-BMI was the anthropometric parameter with the greatest explanatory capacity (power) for the CV variations observed at 6 and 18 y. However, interindividual variations in some hemodynamic and arterial parameters at 6 and 18 y were mainly explained by growth-related body changes and/or by their interaction with current z-BMI (Tables S4 and S6 for children; Tables S8, S10 and S12 for adolescents). Similar results were observed when the associations were analyzed taking into account the exposure to CRFs (Tables 9 and 10). In children, body change during growth, independently or by means of an interaction with current z-BMI, allowed to explain to a greater extent some CV characteristics (i.e., arterial thicknesses and diameters). In other words, disregard of birth size, exposure to CRFs and/or z-pBP, arterial wall thickness and/or diameters at 6 y could be explained by body growth between 0–2 or 0–6 y. Thus, although CV properties at 6 y would be associated with current z-BMI, knowing the history of BW gain could contribute to a better understanding of the CV characteristics of a specific child. Two children with similar z-BMI, could present CV differences associated with their "history" of body size changes (e.g., between 0–2 or 0–6y). Furthermore, for variables such as wall thickness in children, the history of weight gain would havegreater explanatory capacity than current z-BMI or factors with recognized impact on the CV system (e.g., CRFs). In adolescents, the history of BW gain would not be a primary explanatory variable for CVvariables (i.e., for IMT), but due to variables interactions it could contribute or complement data obtained from current z-BMI and/or BP.

#### **5. Strengths and Limitations**

This work has several strengths that should be considered. First, the population-based prospective cohort design, including a large number of subjects studied from early life. Repeated measures during growth-period enabled us to study the impact of growth profiles or patterns on CV properties, assessed at two specific times: early childhood (6 y) and onset of adulthood (18 y). Second, we used our own specific "reference populations" to define CV z-scores (Supplementary Tables S1 and S2). Third, many potential confounders were considered in order to isolate the effect of BW gain in the statistical models. Fourth, taking into account that the impact of body change on the CV system may depend on the period in which it occurs, we studied different periods of body gain (0–2, 0–6, 0–18, 6–18 y). Fifth, the relationship between BW gain and adult pBP is one of the most studied, based on the "fetal origin" hypothesis, but pBP is a particular variable and does not inform about central hemodynamic conditions, or about structural and/or functional arterial changes (e.g., associated with early vascular aging or atherosclerosis development). Thus, we designed an integral approach in which multiple CV parameters (e.g., pBP, cBP, arterial diametersand thicknesses, local and regional stiffness) and different arterial pathways (i.e., elastic and muscular) were evaluated. Sixth, unlike most works that analyzed the associations between body changes and the CV system considering a single age, we studied children and young adults. Up to now, most studies included premature, small for gestational age, obese and/or hypertensive subjects and data about the CVimpact of growth-related bodychanges in healthy pediatric and/or adolescent populations werescarce. In this work, healthy children and adolescents were studied.

Some limitations should be considered. First, we did not have information about blood biomarkers measured by our technicians. Therefore data about some conditions (i.e., existence of dyslipidemia) was obtained from reference physicians, registers and/or self-reports. Second, although we adjusted for several potential confounders, residual confounding factors may persist, as in any observational study. Third, in this work we chose to use change in BWH z-score or z-BMI between two time points as growth-indicators. This approach is a simple practical (clinical) method for quantifying a "change"; although more detailed growth patterns could be derived from longitudinally collected anthropometric measures in both cohorts. Fourth, we did not perform an analysis discriminating by sex; despite we are aware of data suggesting that the impact of childhood growth on the CV system may differ between boys and girls [12]. Fifth, we included subjects born at term and preterm, but as most of them belonged to the first condition (98% and 92% inchildren and adolescent cohorts, respectively) the results should be assigned to term-born subjects. Sixth, comparative analysis of the associations between anthropometric data and CV (hemodynamic and/or arterial) variables measured at 6 and 18 y was done considering two di fferent cohorts, instead of a single cohort followed for more than 20 years. Although obtaining similar data for di fferent cohorts could be considered as strength of the work, as a limiting factor it should be noted that for some variables data were not obtained in both cohorts and some aspects of the associations could only be evaluated in one of them. Finally, we did not analyze growth considering body composition (e.g., fat mass) and its changes as was previously done [43].
