Local diagnostic reference levels for paediatric computed tomography based on European guidelines: anatomy- and indication-based approaches.

Diagnostic reference levels (DRLs) guide dose optimisation, yet paediatric computed tomography (CT) DRLs remain scarce, heterogeneous, largely anatomy-based. This study aimed to establish local paediatric DRLs for head, chest, and abdomen-pelvis CT according to European guidelines, compare them with European DRLs, and assess the feasibility of indication-based DRLs. This single-institution study included 604 of 829 CT acquisitions performed in children aged ⩽16 years in 2017 or 2022. DRLs, defined as the 75th percentiles of volumetric CT dose index (CTDIvol) and dose length product, were derived by age for head CT and by weight for body CT. Indication-based DRLs were explored for five indications. Local DRLs were compared with European and other recent values using 95% confidence intervals. Quantile regression was used to estimate DRL curves and identify predictors for cervical spine CT dose. Head CT DRLs were up to 33% lower than European levels (p< 0.05), and abdomen-pelvis CT DRLs up to 44% lower (p< 0.05), while chest CT DRLs did not differ significantly (p> 0.05). Compared with recent studies, head CT DRLs were lower (p< 0.05), whereas chest and abdomen-pelvis CT DRLs were higher for some groups. Higher doses were observed with older scanner. DRLs for craniosynostosis and chronic sinusitis CT were tenfold lower than those for head trauma or stroke CT. For cervical spine CT, weight was a stronger predictor of CTDIvol than age (pseudoR2= 0.72 vs 0.54). In conclusion, paediatric CT DRLs can be established at institutional level in accordance with European guidelines. The observed DRLs suggest successful implementation of optimisation strategies for head, whereas chest and abdomen-pelvis CT may offer further optimisation potential. Latest-generation scanners and paediatric protocols across the paediatric spectrum should be standard practice, and greater emphasis on indication-based DRLs can enable more precise and clinically relevant optimisation. Weight-based stratification is recommended for cervical spine CT DRLs.