Development of Pediatric Diagnostic Scores Enhances Familial Hypercholesterolemia Detection
Overview
This study evaluated existing pediatric familial hypercholesterolemia (FH) diagnostic criteria and developed two novel tools: FH-PeDS, a semi-quantitative score, and ML-FH-PeDS, a machine learning model. Both new tools outperformed established criteria in identifying genetically confirmed FH cases in Slovenian and Portuguese cohorts, improving early detection where genetic testing is limited.
Background
Familial hypercholesterolemia is a common inherited disorder causing elevated LDL cholesterol from birth, significantly increasing premature cardiovascular disease risk. Despite its prevalence, FH remains underdiagnosed in children, partly due to limitations of adult-based diagnostic criteria and restricted access to genetic testing. Early detection and intervention are critical to reduce long-term cardiovascular complications. Existing clinical scores like DLCN and Simon Broome have suboptimal performance in pediatric populations, highlighting the need for tailored diagnostic tools.
Data Highlights
Diagnostic Tool
Cohort
AUC
Performance Notes
Established Criteria (all)
Combined
Not specified
Identified 47.4% of genetically confirmed FH cases; 10.9% missed entirely
FH-PeDS
Combined
0.897
Outperformed DLCN (0.857); P < 0.01
ML-FH-PeDS (Training)
Slovenian
0.932
Superior predictive power vs. DLCN (0.852); P < 0.01
ML-FH-PeDS (Testing)
Slovenian
0.904
39.7% sensitivity, 87.7% PPV at 98% specificity
ML-FH-PeDS
Portuguese
0.867
Maintained strong performance vs. DLCN (0.815); P < 0.01
Key Findings
Only 47.4% of genetically confirmed pediatric FH cases were detected by all established diagnostic criteria, with 10.9% missed entirely.
FH-PeDS, a novel semi-quantitative score, significantly outperformed the Dutch Lipid Clinics Network (DLCN) score in combined cohorts (AUC 0.897 vs. 0.857; P < 0.01).
The machine learning model ML-FH-PeDS demonstrated superior diagnostic accuracy with AUCs of 0.932 (training) and 0.904 (testing) compared to DLCN (0.852; P < 0.01).
ML-FH-PeDS showed high positive predictive value (87.7%) at 98% specificity, making it effective as a confirmatory test despite moderate sensitivity (39.7%).
External validation in the Portuguese cohort confirmed ML-FH-PeDS’s robust performance (AUC 0.867 vs. 0.815 for DLCN; P < 0.01), despite population differences.
Both FH-PeDS and ML-FH-PeDS can guide genetic testing decisions, especially in resource-limited settings where genetic testing is not widely available.
Clinical Implications
The FH-PeDS and ML-FH-PeDS tools provide clinicians with improved methods to identify children at high risk for familial hypercholesterolemia, facilitating earlier diagnosis and intervention. These tools are particularly valuable in settings with limited access to genetic testing, enabling targeted screening and more efficient use of resources. Incorporating these scores into pediatric practice may enhance cardiovascular risk reduction by promoting timely lipid-lowering therapies.
Conclusion
Current pediatric FH diagnostic criteria inadequately detect many affected children. The newly developed FH-PeDS and ML-FH-PeDS tools significantly improve diagnostic accuracy and can support earlier identification and management of FH in children, potentially reducing long-term cardiovascular morbidity.
References
Development of a Pediatric Diagnostic Score for Familial Hypercholesterolemia (FH-PeDS), 2024
by Jan Kafol, Beatriz Miranda, Rok Sikonja, Jaka Sikonja, Albert Wiegman, Ana Margarida Medeiros, Ana Catarina Alves, Tomas Freiberger, Barbara A Hutten, Matej Mlinaric, Tadej Battelino, FH-PeDS Collaborators, Steve E Humphries, Mafalda Bourbon, Urh Groselj
