New Algorithm Proposed for Dysbetalipoproteinemia Diagnosis

Artificial Intelligence in Medicine
Artificial Intelligence in Medicine
An algorithm using simple clinical variables was proposed for the diagnosis of dysbetalipoproteinemia.

An algorithm using simple clinical variables was proposed for the diagnosis of dysbetalipoproteinemia (DBL), in a study published in the Journal of Clinical Lipidology.

The lipid clinic research database of the Montreal Clinical Research Institute was mined for patients with dyslipidemia. Of the 12,434 eligible patients with available ultracentrifugation data samples, 4891 presented with mixed dyslipidemia (total cholesterol ≥200 mg/dL and triglycerides ≥175 mg/dL) and 188 patients reached the gold standard of the Fredrickson criteria for a DBL diagnosis, as defined by an elevated very-low-density lipoprotein-cholesterol (VLDL-C)/triglyceride (TG) ratio and were carriers of apolipoprotein (apo) E2/E2 genotype.

Patients with vs without DBL had a higher frequency of DBL-related xanthomas (25% vs 1%, respectively; P <.0001) and a greater prevalence of broad beta band (95% vs 4%, respectively; P <.0001).

The variables significantly associated with DBL diagnosis were: total cholesterol levels, 17 TGs, non-HDL-C, apoB, total cholesterol/apoB ratio, and TG/apoB ratio (P <.0001 for all). The investigators determined that the non-HDL-C to apoB ratio best predicted DBL diagnosis with an area under the receiving operator curve (AUC) of 0.93 (95% CI, 0.91-0.95; P <.0001). The mean non-HDL-C to apoB ratio significantly increased based on each individual’s APOE genotype and DBL status (P <.0001).

When a cut-off point of 3.69 mmol/g for the non-HDL-C to apoB  ratio was applied, and with which  95% of cases were identified,  and in the presence of an E2/E2 genotype, DBL was predicted with a sensitivity of 94.8%, specificity of 99.6%, accuracy of 99.4%, and an AUC of 0.97 (95% CI, 0.95-0.99).

To validate this algorithm, investigators compared it with previously proposed screening programs, which were all found to have low sensitivity (non-HDL-C/apoB ratio >6.72, 23.4%; the Sniderman criteria with apoB <1.2 g/L, 26.6%; apoB/total cholesterol ratio <0.15, 31.6%; the Sniderman criteria with TG >2.3 mmol/L; 44.9%; the Sniderman criteria, 46.1%; non-HDL-C/apoB ratio >4.91, 62.3%) or low specificity (remnant cholesterol/TG ratio >0.23, 1.5%).

Study limitations include a lack of screening for rare APOE mutations, possible exclusion of patients from the DBL cohort based on the absence of apoE2/E2 status, and a high prevalence of DBL diagnosis (87%) in individuals with E2/E2 genotypes compared with previously reported rates (5%-18%).

“Screening using the non-HDL-C/apoB ratio (cutoff ≥ 3.69 mmol/g [1.43 in conventional units]) followed by confirmation using the APOE genotype offered excellent sensitivity and specificity for DBL,” noted the study authors. “Furthermore, laboratory information systems could easily calculate this ratio and identify subjects with mixed dyslipidemia that could qualify for APOE genotyping therefore helping clinicians in diagnosing this rare disease.”

Disclosure: Multiple authors declared affiliations with industry. Please refer to the original article for a full list of disclosures.


Paquette M, Vernard S, Blank D, et al. A simplified diagnosis algorithm for Dysbetalipoproteinemia. [published online June 10, 2020] J Clin Lipidol. doi:10.1016/j.jacl.2020.06.004