Association Between Dehydroepiandrosterone Sulphate Levels at 7 Years Old And Bone Mineral Density At 10 Years Old – a Prospective Cohort Study

Rita Santos-Silva (  ritasantossilva@gmail.com ) Universidade do Porto Faculdade de Medicina https://orcid.org/0000-0002-0338-3399 Manuel Fontoura Universidade do Porto Faculdade de Medicina Milton Severo Universidade do Porto Instituto de Saude Publica Raquel Lucas Universidade do Porto Instituto de Saúde Pública: Universidade do Porto Instituto de Saude Publica Ana Cristina Santos Universidade do Porto Instituto de Saude Publica


Introduction
The study participants are a subsample of children included in a prospective birth cohort, Generation XXI, whose full details have been published elsewhere [34; 35].
The recruitment of participants was conducted in 2005/2006 in all ve public maternities of the metropolitan area of Porto, Portugal. Of the invited mothers, 8495 agreed to participate (91%) and a total of 8647 newborns were enrolled. At ages 4, 7 and 10 years old, all Generation XXI participants were invited to a face-to-face follow-up evaluation. Of the 8647 initial cohort members, 7459 (86%), 6889 (80%), and 6392 (74%) were assessed at the 4-, 7-, and 10-year-old follow-up evaluations, respectively. The follow-up visits included a physical examination and fasting blood sample, according to standard procedures.
From those that attended the 7-year-old follow-up, 700 prepubescent children were randomly selected and their DHEAS levels were measured, as part of a study on adrenarche [10]. Among these 700 prepubescent children, 274 (139 girls and 135 boys) had complete information in all variables analyzed, including a full-body dual-energy X-ray absorptiometry (DXA) scan at 7 and 10 years old. A comparison between those of the 700 who were included in the complete information analysis and those who were not is depicted in Supplementary Table 1.

Data collection
At the baseline, data on maternal demographic and socioeconomic characteristics, lifestyle, obstetric history, pre-pregnancy anthropometrics, and personal history of diseases, were collected by trained interviewers, using structured questionnaires, during the hospital stay.
Data on delivery and newborn characteristics (including gestational age, birth weight, and length) were additionally extracted from clinical records [34; 35].
Birth weight and length were transformed in z-scores according to the Fenton growth charts [36].
Anthropometric measurements At 4,7, and 10 years of age, trained researchers performed anthropometric measurements, with the child in underwear and bare feet. Weight was measured to the nearest 0.1 kg using a digital scale (Tanita®, Arlington Heights, IL, USA), and standing height was measured to the nearest 0.1 cm using a wall stadiometer (Seca®, Hamburg, Germany). BMI was calculated by dividing weight (kg) by squared height (m 2 ). BMI was transformed into age and sex-speci c z-scores using World Health Organization (WHO) standards [37].

DXA-derived bone measures and body composition
At ages 7 and 10, whole-body DXA scans were performed using a Hologic Discovery QDR® 4500W device (software version 13.3.0.1; Hologic Inc., Bedford, MA, USA) according to standard manufacturer's protocol, while the child was in underwear and with the bladder emptied. Standard quality assurance tests were performed daily using the spine phantom according to the manufacturer's instructions. Scans were evaluated immediately after acquisition and later validated by a second technician. Total body less head (subtotal) bone mineral content (BMC) (g) and areal BMD (aBMD) (g/cm 2 ) were obtained, according to the International Society of Clinical Densitometry recommendation [38]. Fat and lean mass (g), and body fat percentage (%) were also assessed by DXA scan.

Sexual maturity evaluation
Sexual development evaluation was conducted by trained observers according to the sexual maturity ratings including breast changes in females, genital changes in males, and pubic hair changes in both females and males (Tanner stages) [39]. In girls, breasts were evaluated by inspection and palpation, and in boys, testicular volume was assessed by palpation using the Prader orchidometer. Included participants were classi ed as Tanner stage I, II, III, IV or V. Prepuberty was de ned as Tanner stage I. Biochemical analysis DHEAS levels were measured in blood (serum) collected at the 7-year-old follow-up visit, by electrochemiluminescence immunoassays on the Roche cobas e411 analyzer (Roche Diagnostics, Basel, Switzerland). An overnight fasting venous blood sample was obtained before 11:00 a.m., after applying topical analgesic with lidocaine/prilocaine (EMLA cream).

Statistical analysis
Categorical and continuous variables were presented as counts (proportions), and mean and standard deviation (SD). The chi-square and the t-test for two independent samples were used to evaluate differences between sexes. The partial correlation test adjusted for age, sex and BMI z-score was applied to analyze the associations between aBMD and anthropometric and hormonal indicators. This analysis was performed using SPSS® (v.24; SPSS, IBM Corp., Armonk, NY, USA).
Path analysis was used to estimate crude and adjusted linear regression coe cients (β) and 95% con dence intervals (95% CI), which represent the increase in aBMD (g/cm 2 ) at 10 years old for each 10 µg/dL increase in DHEAS at 7 years old. Path analysis was conducted based on the theoretical model depicted in Figure 1. Since bone size and aBMD increase with body height and weight [15; 16; 40; 41], BMI, along with age, sex, and the stage of puberty, should be considered when assessing determinants of BMD in children and adolescents [16 ; 19]. Thus, our adjusted model included BMI z-score at 7 years old, aBMD at 7 years old, and Tanner stage at 10 years old as explanatory variables. Considering sex differences in pubescent development timing and bone mass increase, boys and girls were analyzed separately.
Path analysis was performed with the lavaan [42] package from R software version 4.0.3; 95% CI was calculated by bootstrapping. Full information maximum likelihood estimation was used to handle missing values, assuming missing at random [43]. The t of the models was assessed using different indexes: the Comparative Fit Index (CFI) [44], the Tucker-Lewis Index (TLI) [45], and the Root Mean Square Error of Approximation (RMSEA) [46]. A good model t is indicated by a CFI and TLI values ≥ 0.90 and values of RMSEA lower than 0.08. The nal model had CFI 1.000, TLI 1.029, RMSEA 0.000 (girls) and CFI 1.000, TLI 1.003, RMSEA 0.000 (boys).

Sample characteristics
Characteristics of the 274 participants (139 girls and 135 boys), and the comparison between sexes, are shown in Table 1. Evaluations were conducted at a mean age of 7.1 (SD: 0.2) years old and 10.1 (SD: 0.2) years old, with no sex differences. Neonatal and maternal characteristics were similar in boys and girls. Table 1 Anthropometric, metabolic, and hormonal characteristics of the participants at birth, 7 and 10 years, and maternal characteristics, by sex

Partial correlations
Partial correlation coe cients between aBMD at 7 and 10 years old and independent variables, for the whole sample, are summarized in Table 2. Areal BMD at 7 years old correlated positively with birth length, height, fat and lean mass at 7 years old, and DHEAS levels at 7 years old, after adjustment for age, sex and BMI z-score. Areal BMD at 10 years old correlated positively with height, fat and lean mass at 7 years old, and with height and lean mass (but not fat mass) at 10 years old, after controlling for age, sex and BMI z-score. Areal BMD at 10 years old also correlated positively with DHEAS levels at 7 years old and aBMD at 7 years old, adjusted for age, sex and BMI z-score.  Path analysis A mediation analysis is depicted in Figure 1. It comprises the estimated total, direct, and indirect effects of 10 µg/dL increase in DHEAS at age 7 in aBMD (g/cm 2 ) at age 10, strati ed by sex.
In girls, crude analysis showed that higher DHEAS at age 7 was associated with higher aBMD at age 10 (β = 0.007 [95% CI: 0.004; 0.010], p<0.001) (Figure 1, ab + ed + c). This total association was mainly explained by indirect effects. Higher DHEAS at age 7 was associated with higher Tanner stage at age 10, and higher Tanner stage was associated with higher aBMD, and this indirect effect represented 21% of the total effect (p=0.001) (Figure 1, ed). Higher DHEAS at age 7 was also associated with higher aBMD at age 7, and higher aBMD at age 7 was associated with higher aBMD at age 10, and this indirect effect explained 61% of the total effect (p<0.001) (Figure 1, ab). No direct effect of DHEAS at age 7 in aBMD at age 10 was observed (Figure 1, c) ( Table 3).  (Figure 1, ab + ed + c), explained in 33% by the indirect effect of DHEAS on Tanner stage and Tanner stage on aBMD (Figure 1, ed). The indirect effect of DHEAS on aBMD at age 7 and aBMD at age 7 on aBMD at age 10 was attenuated and lost statistical signi cance (Figure 1, ab) ( Table 3).
BMI z-score at age 10 was also considered in the model, but it neither changed the results appreciably nor improved the t of the model (data not shown), and therefore was excluded from the nal model.

Discussion
The present study explores the effect of DHEAS at the age of 7 years on aBMD at the age of 10 years. Firstly, we found that aBMD at 10 years old correlated positively with DHEAS at 7 years old, after adjustment for age, sex, and BMI z-score. Secondly, using path analysis, we tried to distinguish a possible direct effect of DHEAS at age 7 on aBMD at age 10 from an indirect effect partially explained by sexual maturity or by aBMD at age 7. Although no direct effect of DHEAS at age 7 on aBMD at age 10 was observed, we found in girls, but not in boys, an indirect effect explained by sexual maturity, as higher DHEAS levels at 7 years old were associated with higher sexual maturity at 10 years old, which was further associated with higher aBMD, controlling for BMI.
To our best knowledge, this is the rst study to address the longitudinal effect of DHEAS on aBMD in prepuberty and early puberty. So far, only a few crosssectional studies have investigated the effect of circulating adrenal androgens on bone mass acquisition in mid-childhood, with mixed results, and a comparison with our ndings is di cult due to different populations and methodological approaches. In accordance with our results, a positive effect of adrenal androgens on BMD was found in premenarcheal girls [47] and in two populations of children aged 5-8 years [15] and 6-18 years [32]. On the other hand, no association was found between DHEAS and bone mineral density in 255 children aged 7-8 years [31] and in a population of boys aged 6-14.5 years [26]. In a large cohort involving 472 Finnish children aged 6-8 years, the positive association of DHEAS with BMD disappeared after adjustment for fat and lean mass [33].
In our analyses, we have decided to study girls and boys separately, as we recognize the large sex differences in bone mass increase and the timing of puberty. At age 7, no statistically signi cant difference was found in aBMD between boys and girls, while, at the age of 10, girls presented higher aBMD than boys. Furthermore, at the age of 10, most of the girls had started puberty (78%), while 71% of the boys were still prepubescent. In girls, the effect of DHEAS at age 7 on aBMD at age 10 was partially explained by sexual maturity, as higher DHEAS at 7 years old was associated with higher Tanner stage at 10 years old, which was further associated with higher aBMD. In boys, no such effect was found, and some explanations for this sex discrepancy can be pointed out. Firstly, during puberty, girls accrue more bone mass than boys, and they do it in earlier Tanner stages [16]. Secondly, previous studies have shown that higher serum DHEAS at 7 years old is associated with earlier pubescent development in girls, but not in boys [11; 12]. Therefore, the indirect effect of DHEAS on sexual development and sexual development on aBMD is less relevant in boys than in girls, at this age.
It should be noted that the found effect of DHEAS on BMD at age 10, partially explained by sexual maturity, could also be the result of other unmeasured sexual hormones, such as estrogens. A direct effect of DHEAS on BMD, independent of estrogens, was not established.
Bone size and aBMD increase with height and weight [15; 16; 40; 41]. Hence, BMI, along with sex and the stage of puberty, was considered in our analyses.
Obese and overweight prepubescent children present higher DHEAS levels [7; 9] and higher androgen levels are associated with changes in body composition, such as increased central adiposity and lean mass [19], which can affect the bone. The association between BMI and BMD in children is mostly determined by lean mass [33; 41], but adiposity also appears to play a role, despite contradictory ndings [33; 48; 49]. Adiposity may augment BMD through an increased mechanical load exerted on the skeleton by fat mass [18], or the aromatization of androgens in fat [16], or through unmeasured cytokines, growth factors or other hormones (leptin, insulin and estrogens) [33], which may exert direct stimulatory effects on osteoblasts [19]. Although we had other measures of adiposity, like waist circumference or body fat, they were not included in the model due to multicollinearity.
Bone modelling and growth in childhood and early pubescent years are in uenced by endogenous and exogenous factors. Exogenous factors include nutrition (mainly calcium and vitamin D) and weightbearing physical activity, while endogenous factors include hormones (growth hormone, sex steroids, and various growth factors), cytokines, and growth plate aging [50]. BMD is also affected by genetic and early growth. In a previous study involving 1853 participants from the same birth cohort, Generation XXI, weight and height velocities up to the age of 6 were associated with increased aBMD at 7 years with the strongest associations observed for growth in early childhood [51]. Moreover, in the same population, children that between zero and 4 years followed a trajectory of persistent weight gain, had clearly increased bone mass at 7 years old, and weight gain seemed slightly more bene cial when it occurred later than on a normal trajectory during the rst years of life [52].

Strengths and limitations
The strengths of our study include the novelty, as previous longitudinal data on the study subject is minimal. Furthermore, we have used a population-based cohort, with detailed information regarding birth and early childhood, physical examination, anthropometry, biochemical data, and DXA evaluation, according to standardized procedures, at ages 7 and 10 years, as well as DHEAS levels in prepuberty. Consequently, our results cannot be generalized to other age groups, as they would differ because of the effect of increased growth hormone and sex steroid levels on BMD during puberty.
Nevertheless, some limitations must be acknowledged. DXA is a two-dimensional estimate of volumetric bone density, so differences in bone size may confound the androgen-BMD association assessed by this technique. Nevertheless, adjustment for BMI partially attenuates this effect. Bone age evaluation was not part of the research protocol due to radiation exposure, and therefore no conclusions regarding skeleton maturation can be drawn. It is possible that the effect of DHEAS on BMD is not fully evident at the age we have assessed, especially among boys, who start puberty later than girls. As we continue to follow this cohort, we may carry on further investigation in different age ranges.
We have used path analysis to answer our main objective, but it is worth noting that path analysis is not intended to prove causation but rather to test if observed results are consistent with a priori hypothesis. Our statistical model is necessarily oversimpli ed, given the complex relationships between the variables analyzed. These variables may be in uenced by several genetic and environmental factors that were not measured in this study. Furthermore, it assumes that the observed relations follow a particular direction that may not be totally realistic. Thus, the observed statistical associations demand careful interpretation regarding causality.

Conclusion
In girls, DHEAS at 7 years old affected aBMD at 10 years old. This effect is indirect, as higher DHEAS levels were associated with more advanced sexual maturity at the age of 10, and more advanced sexual maturity was associated with higher aBMD. No direct effect of DHEAS on aBMD was observed. No effect of DHEAS at 7 years old on aBMD at 10 years old was seen in boys. Supplementarymaterial.docx