Tamara van Donge

Reference intervals for kidney biomarkers in pediatrics 203 10 After excluding outlying observations according to Tukey’s method (Table S1) and log transforming SCR and KIM-1 due to non-normality, Pearson’s correlation test was performed between the marker values and age as a continuous variable (Table S5). The strongest positive correlation with age as a continuous variable (no categories) was observed for SCR (r = 0.864 and p < 0.001), and the strongest negative correlation for BTP (r = -0.544 and p < 0.001). ALB showed a mild positive correlation (r = 0.261 and p < 0.01), and both B2M and KIM-1 showed a negative (r = -0.276 and p < 0.01, r = -0.217 and p < 0.05) correlation with continuous age, respectively. The other evaluated markers did not show any significant correlation with age as a continuous variable (p > 0.05). Five kidney function and injury markers ALB, B2M, BTP, SCR and KIM-1, showed an association with age (Figure 2). Based on the multivariable linear model (Table S6), the mean and 95% prediction intervals for those five markers were calculated for four different age groups (Table 4). The predicted B2M concentrations from the fitted model (Eq. 1) for a child of 2, 5, 10 or 15 years were 1.9 mg/L (95% CI (1.3–2.5)), 1.7 mg/L (95% CI (1.1–2.3)), 1.6 mg/L (95% CI (1.0–2.1)) or 1.7 mg/L (95% CI (1.1–2.3)), respectively (Figure 2). BTP showed a trend to decrease with age (p < 0.001), and the predicted concentrations were estimated at 0.9 mg/L (95% CI (0.6–1.1)), 0.7 mg/L (95% CI (0.6–1.0)), 0.6 mg/L (95% CI (0.4–0.9)) or 0.6 mg/L (95% CI (0.4–0.9)) for a 2, 5, 10- or 15-year-old child (Eq. 2, Figure 2). SCR showed a strong positive association with age and increased from 22.7 µmol/L (95% CI (15.2 – 33.9)) for a 2-year old infant to 58.9 µmol/L (95% CI (39.4 – 88.1)) for a 15- year old adolescent (Eq. 3, Table 4). KIM-1 decreased with age (p < 0.05), and showed a 50% decrease between a two-year-old child and a 15-year old adolescent (Table 4). CYSC was significantly associated with sex, showing lower concentrations in female pediatric population as compared to male participants (p < 0.01, Table S6). Piecewise regression with a breakpoint at approximately 1.8 years was a better fit compared to linear regression for CYSC in terms of AIC criteria (Table S6). In the period before 1.8 years, CYSC appears to show a decrease, and after this breakpoint a slight increase in CYSC is observed until the age of 15 years (Eq. 6a & 6b). Overall, girls show a lower CYSC concentration (median 0.86 mg/L (IQR 0.78–0.94)) compared to boys (median 0.92 mg/L (IQR 0.81–1.04, p = 0.01)), with the predicted CYSC concentration from the fitted model at 1.5 years of age being 0.96 mg/L (95% CI (0.71–1.22)) for boys and 0.91 mg/L (95% CI (0.66–1.17)) for girls, respectively (Table 4).