Associations among clinical data, BMD measures and bone metabolism
markers
In Spearman correlation analysis (Table 3 ), LSBMD z-score was
correlated positively with age (p = 0.03), BA (p = 0.003), BMI z-score
(p = 0.005) and Tanner stage, (p = 0.001) and negatively with Dkk-1 (p =
0.019). Similarly, TBBMD z-score was correlated positively with age (p =
0.03), BA (p = 0.02), BMI z-score (p = 0.01) and Tanner stage, (p =
0.02) and negatively with Dkk-1 (p = 0.001). Dkk-1 correlated positively
with the history of delayed puberty (p = 0.038) and negatively with
blood transfusion interval (p = 0.038). However, LSBMD and TBBMD
z-scores showed no correlation with the history of delayed puberty, age
at first blood transfusion, transfused blood volume, pre-transfusion Hb,
markers of iron overload, 25OHD, and OC and CTX z-scores.
In multivariable linear regression analysis (Table 4 ), Dkk-1
was inversely associated with TBBMD z-score adjusting for BA, sex and
BMI z-score (model 1 , unstandardized β ± SE: -0.002 ±0.001, p = 0.002). The association remained significant when further
adjusting for average pre-transfusion Hb and 25OHD (model 2 , β
± SE: -0.002 ± 0.001, p = 0.005), and the history of delayed
puberty, type of iron chelator and prednisolone use (model 3 , β
± SE: -0.002 ± 0.001, p = 0.009). Nevertheless, there was no
significant association between Dkk-1 and LSBMD z-score in all
regression models.