Serum neurofilament light chain reference database for individual application in paediatric care: a retrospective modelling and validation study

Background Neurological conditions represent an important driver of paediatric disability burden worldwide. Measurement of serum neurofilament light chain (sNfL) concentrations, a specific marker of neuroaxonal injury, has the potential to contribute to the management of children with such conditions. In this context, the European Medicines Agency recently declared age-adjusted reference values for sNfL a top research priority. We aimed to establish an age-adjusted sNfL reference range database in a population of healthy children and adolescents, and to validate this database in paediatric patients with neurological conditions to affirm its clinical applicability. Methods To generate a paediatric sNfL reference dataset, sNfL values were measured in a population of healthy children and adolescents (aged 0–22 years) from two large cohorts in Europe (the Coronavirus Antibodies in Kids from Bavaria study, Germany) and North America (a US Network of Paediatric Multiple Sclerosis Centers paediatric case-control cohort). Children with active or previous COVID-19 infection or SARS-CoV-2 antibody positivity at the time of sampling, or a history of primary systemic or neurological conditions were excluded. Linear models were used to restrospectively study the effect of age and weight on sNfL concentrations. We modelled the distribution of sNfL concentrations as a function of age-related physiological changes to derive reference percentile and Z score values via a generalised additive model for location, scale, and shape. The clinical utility of the new reference dataset was assessed in children and adolescents (aged 1–19 years) with neurological diseases (epilepsy, traumatic brain injury, bacterial CNS infections, paediatric-onset multiple sclerosis, and myelin oligodendrocyte glycoprotein antibody-associated disease) from the paediatric neuroimmunology clinic at the University of California San Francisco (San Francisco, CA, USA) and the Children's Hospital of the University of Regensburg (Regensburg, Germany). Findings Samples from 2667 healthy children and adolescents (1336 [50·1%] girls and 1331 [49·9%] boys; median age 8·0 years [IQR 4·0–12·0]) were used to generate the reference database covering neonatal age to adolescence (target age range 0–20 years). In the healthy population, sNfL concentrations decreased with age by an estimated 6·8% per year until age 10·3 years (estimated multiplicative effect per 1 year increase 0·93 [95% CI 0·93–0·94], p<0·0001) and was mostly stable thereafter up to age 22 years (1·00 [0·52–1·94], p>0·99). Independent of age, the magnitude of the effect of weight on sNfL concentrations was marginal. Samples from 220 children with neurological conditions (134 [60·9%] girls and 86 [39·1%] boys; median age 14·7 years [IQR 10·8–16·5]) were used to validate the clinical utility of the reference Z scores. In this population, age-adjusted sNfL Z scores were higher than in the reference population of healthy children and adolescents (p<0·0001) with higher effect size metrics (Cohen's d=1·56) compared with the application of raw sNfL concentrations (d=1·28). Interpretation The established normative sNfL values in children and adolescents provide a foundation for the clinical application of sNfL in the paediatric population. Compared with absolute sNfL values, the use of sNfL Z score was associated with higher effect size metrics and allowed for more accurate estimation of the extent of ongoing neuroaxonal damage in individual patients. Funding Swiss National Science Foundation, US National Institutes of Health, and the National Multiple Sclerosis Society.