Continuous reference intervals for high-sensitivity cardiac troponin T in children: a closed-form approach with zlog transformation.

内容摘要

Age-dependent physiology makes interpreting high-sensitivity cardiac troponin T (hsTnT) challenging in early childhood. Values below the analytical limit create left-censored data, complicating reference interval (RI) estimation. We combined two paediatric cohorts (CALIPER, n=277; LIFE Child, n=1,358; ages 0-10 years) to derive continuous, age-dependent hsTnT RIs. Empirical quantiles were estimated using a sliding-window approach and fitted with four-parameter logistic functions. To address the high proportion of values below the analytical limit of detection (LOD; 3 ng/L), the 2.5th percentile was inferred from the 65th and 97.5th percentiles assuming lognormal distribution. A closed-form zlog transformation was established to standardise hsTnT values. Pointwise 95 % bootstrap confidence bands (100,000 stratified resamples) were computed for both reference limits. HsTnT concentrations declined steeply during infancy and stabilised at low levels thereafter, with 46 % of samples below the analytical limit. The closed-form approach modelled reference limits according to the following logistic equation f x = 1 1 + e - a + b · x · maxv - minv + minv with parameters: a=-1.15 and 0.47, b=-4.13 and -6.64, minv=1.13 and 7.50 and maxv=124.13 and 210.3 for lower and upper limits, respectively. Bootstrap confidence bands were widest in the first six months of life and narrowed substantially thereafter. The zlog transformation effectively removed age dependency, with >95 % of values within the theoretical RI (-1.96 to +1.96). We present the first continuously modelled, closed-form RIs and corresponding zlog-transformation for paediatric hsTnT (ages 0-10 years). This framework enables direct LIS/EHR deployment without specialized software and facilitates age-independent interpretation and longitudinal monitoring in paediatric cardiology.