Our study set out to assess the link between BMI z-score of female adolescents with WD and myocardial function. Our main results were that (i) LV morphological remodeling was significantly and positively correlated with BMI z-score, (ii) GLS and longitudinal strains recorded at the apical and medial levels but not at the base were significantly correlated with BMI z-score, the highest strains being observed in adolescents with the lowest BMI z-scores, (iii) from short-axis planes, both circumferential strains and rotations, and so twist and untwisting rates were not correlated with BMI z-score, and (iv) from a multivariate stepwise regression analysis, fibrinogen and SBP were the main variables explaining the alteration of regional longitudinal strains with BMI z-score.
Cardiac remodeling and BMI z-score
We found a positive correlation between the BMI z-score and the LV mass. These results are in agreement with current knowledge on the cardiac remodeling induced by WD, namely a decrease in measurements in AN [3, 25, 26] and conversely an increase in OB [2, 6]. These morphological adaptations are linked not only to the nutritional state itself but also to the load conditions , which vary with BMI, as demonstrated in our study by positive correlation between BMI z-score and clinical variables of HR and BP. In AN, possible causes of hypotrophy therefore include decreased afterload from relative hypotension leading to down-regulation of the LV mass , reduced preload leading to ventricular remodeling and a direct effect of malnutrition as for skeletal musculature . By contrast, in OB, the excess adipose tissue is responsible for insulin resistance with deleterious effects on the heart: insulin may aggravate hypervolemia through salt and water retention, act as a myocardial growth factor, activate the sympathetic nervous system, and therefore ultimately increase the size of the heart . On the other hand, the increase in preload and afterload (linked to arterial stiffness) also seems to be involved in the morphological changes . Moreover, in line with the literature , we showed that these morphological changes were associated with thickening of epicardial fat, which is closely correlated with the BMI z-score. Interestingly, this epicardial fat is considered as a marker of cardiovascular risk , in particular because in excess it has direct effects on coronary atherosclerosis and causes fatty infiltration, inflammation and myocardial fibrosis [8, 23].
Regional myocardial function and BMI z-score
The EF, assessed from conventional echocardiographic variables, remained unchanged whatever the BMI z-score. The strength of our study was to evaluate systolic function from the GLS measured by STE. Interestingly, GLS was negatively correlated with the BMI z-score (i.e., lower strains with higher BMI z-score). Moreover, the regional assessment of LV longitudinal strains gave additional information by demonstrating that these correlations were observed only at the mid and apical levels, but not at the base. Our results thus robustly demonstrate a continuum of adaptation of longitudinal strains at mid and apical levels in a population of adolescents from very low to high BMI z-score. In AN patients, only one study used STE and found no alteration in GLS compared with controls, except in a very small subgroup of patients with purging behavior in whom apical longitudinal strains were reduced. They demonstrated that the only predictor of the apical strains was the BMI z-score . In adolescents with obesity, it is well-established that GLS is lower than in controls [4, 6, 11, 12]. Of note, from regional analysis of longitudinal strains, Binnetoglu et al. (2015)  also recently reported a more pronounced decrease in strains at the mid and apical segments compared to the base.
To gain insight into the underlying mechanisms of these alterations, we performed multivariate stepwise regression analyses and observed that fibrinogen and SBP were the two main independent predictors of LV mid or apical strains. Fibrinogen was considered to be a high-risk marker for developing vascular inflammatory diseases, such as arterial hypertension and atherosclerosis . Our results demonstrated close correlations between fibrinogen and not only the BMI z-score but also the SBP. A chronic inflammatory state may therefore appear with increasing BMI z-score mediated in particular by fibrinogen, this latter promoting the appearance of higher SBP, afterload, and in fine reduction in strains. It has been recently demonstrated that longitudinal strains are heavily influenced by afterload . The inflammatory state could also induce an adverse effect on strains via an occurrence of myocardial fibrosis, well-established in OB .
Twisting mechanics and BMI z-score
For a comprehensive evaluation of the impact of BMI z-score on the myocardial mechanics, we also assessed circumferential strains and rotations. Another important (and unexpected) finding was that both circumferential strains and rotations (and so twist)  were normal regardless of the BMI z-score. Interestingly, the discrepancies between findings obtained on longitudinal and circumferential strains suggested a specific impact of BMI z-score on the longitudinal deformations. The greater vulnerability of longitudinally orientated myocardial fibers may be explained by their major distribution in the subendocardium, leading them to be more susceptible to wall stress, ischemia and fibrosis .
The LV twist helps to create a uniform distribution of LV fiber stress and fiber shortening across the wall and is essential for the course of systole. Its disappearance has been clearly shown to increase oxygen demand and reduce the efficiency of LV systolic function . LV twist is affected by loading conditions of the heart; an increase in afterload and/or a decrease in preload induce a decrease in twist in animal studies . In physiological states in which pre- and afterload were decreased, twist remained unchanged, or was increased . In our study, the wide range of BMI z-scores, accompanied by changes in SBP, did not impact the twist. We note that other factors could impact the twist and allow its conservation, such as an increase in intrinsic myocardial contractility or an increase in sympathetic activity, which have been suggested to increase the twist [13, 24].
The energy stored during LV twist in elastic components is restored very early in diastole, creating an intraventricular pressure gradient that favours LV filling, thus linking systole to diastole (systolic-diastolic coupling) . Here we observed no significant correlation between peak untwisting rate and BMI z-score, suggesting that this mechanism was preserved over a wide range of WD. Taken together, our results strongly support that twisting mechanics are unaffected by variation in BMI z-score, probably partly explaining the preservation of systolic and diastolic function in adolescents with WD, regardless of their BMI z-score.
To obtain a broad, continuous scale of BMI z-score, we decided to include AN patients, who were not constitutionally thin. This could therefore introduce bias in our results, linked to the pathophysiology specific to this disease.