Dear Editor,

We read with interest the recent report by Boulanouar et al., entitled “Diagnostic reference levels based on clinical indications for paediatric computed tomography examinations at the Mohammed VI University Hospital Center in Marrakech-Morocco” published in Radioprotection (Boulanouar et al., 2025). The investigators retrospectively analysed 1,730 pediatric CT examinations across head, chest, abdomen-pelvis, and chest-abdomen-pelvis, proposing age-stratified diagnostic reference levels (DRL-CI) using CTDIvol and DLP percentiles and comparing results with international benchmarks.

Although this work fills a gap, as DRL-CI for pediatric CT is valuable for dose optimization and benchmarking across institutions, we identify issues that warrant clarification before the findings can be translated into practice. Key concerns include inconsistent reporting of age stratification between the methods and results and discordant statements between the abstract and the main text regarding statistical significance for CTDIvol versus DLP.

Age stratification and denominators

The methods describe three age bands, yet the results summarize a different grouping (including an infant category). A brief reconciliation of inclusion criteria and denominators (e.g., whether ≤1-year scans were excluded or merged) would improve interpretability, particularly for DRLs where size effects are most pronounced.

Statistical interpretation of DRL variation

The article states that DRLs were defined by medians and third quartiles, while hypothesis testing is justified as median comparison. However, DRLs are typically operationalized as the 75th percentile; testing medians does not directly test differences in the 75th percentile. A more coherent report would have been preferable to improve interpretability and would have included (i) a single, consistent DRL definition; (ii) effect sizes with confidence intervals to manage uncertainty; and (iii) sensitivity analyses focused on upper-tail measures (e.g., quantile regression or 75th percentile bootstrap comparisons), as well as an adjustment for scan length and multiphase acquisition, the main drivers of DLP.

Clinical translation in an evolving technology landscape

Population-level modeling underscores why CT dose optimization remains clinically important (Smith-Bindman et al., 2025), yet contemporaneous correspondence highlights how post-2020 reconstruction and protocol advances may materially change dose distributions (Hsu et al., 2025). Reporting scanner generation, reconstruction approach, and protocol standardization, and presenting size surrogates (if weight is unavailable) would strengthen external generalizability.

Examining these points would have allowed readers to better judge whether the observed differences reflect a clinical indication, the structure of the protocol, or a reporting artifact, thereby strengthening confidence in the proposed DRL-CI, facilitating safer and more generalizable implementation, i.e., improving application to practice, and paving the way to future DRL initiatives.

References

© A.K Gupta et al., Published by EDP Sciences 2026