The variability of measured and calculated low-density lipoprotein (LDL) cholesterol in statin-treated diabetes patients

Background: The Sampson-NIH and Martin-Hopkins low-density lipoprotein cholesterol (LDL-C) equations are advocated as being superior to the Friedewald calculation. However, their mathematical complexity means they may have different biological and analytical variation when tracking LDL-C in the same patient. This study has established the biological variation (BV) of calculated and directly measured LDL-C (dLDL-C) in patients taking equivalent doses of a long (atorvastatin) and short (simvastatin) half-life statin. It also modelled how analytical imprecision might add to these BVs.

Methods: In a crossover study of lipid BV involving 26 patients with type 2 diabetes (T2DM) initially taking either simvastatin 40 mg or atorvastatin 10 mg, fasting lipids were measured 10 times over 5 weeks after a 3 month run-in. The same procedure was then followed for the alternate statin. Outlier removal and CV-ANOVA established the BV of dLDL and each formula. Analytical measurement uncertainty was estimated from 6 months of real-world data.

Results: The intra-individual BV of dLDL-C measurement was considerably lower with atorvastatin than simvastatin (CV 1.3%(95% CI 1.1-1.5%) vs. 11.1%(10.2-12.2%), respectively). No equation could distinguish this difference (Friedewald 11.0%(95% CI 10.0-12.1%) vs. 12.9%(11.8-14.2%), Sampson-NIH 10.4%(9.5-11.5%) vs. 11.7% (10.7-12.8%) and Martin-Hopkins 9.3%(8.5-10.3%) vs. 11.3%(10.3-12.4%)). Real-world analytical CVs were 2.6% (Sampson-NIH), 2.6% (Martin-Hopkins) 2.8% (Friedewald) and 2.0% (dLDL-C).

Conclusions: Inherent biological LDL-C variability using these formulae is substantially greater than direct measurement in T2DM patients taking atorvastatin. Typical analytical imprecision was also greater. Together, this may fundamentally limit these equations' ability to track true LDL-C changes in patients taking popular statin treatments.