Direct measurement vs. Friedewald, Martin-Hopkins and Sampson-NIH equations for LDL-C estimation in Polish primary care population
M Golawski, T O Osadnik, M G Gierlotka, M L Lejawa, M F Fronczek, N P Pawlas, J J Jozwiak, M B BanachAbstract
Introduction
Friedewald (FE), Martin-Hopkins (M-HE), and modified Sampson-NIH (mSNE) equations are used to estimate LDL cholesterol (LDL-C) when direct measurement (DM) is unavailable.
Purpose
To assess the differences in LDL-C estimates obtained from three different equations compared to direct measurement.
Methods
The LIPIDOGRAM2015 study was conducted in a primary care setting. Patients had blood lipids measured, anthropometric measurements taken and filled questionnaires on comorbidities and lifestyle. LDL-C was measured directly, and LDL-C values were also calculated using the FE, M-HE, and mSNE equations; the Friedewald equation was used only for triglyceride levels below 4.52 mmol/l (400 mg/dl). Cohen's Kappa score assessed the accuracy of assigning patients into LDL-C intervals based on guideline cut-offs (1.4, 1.8, 2.6, 3.0 mmol/l). Cardiovascular risk categories were assigned to each patient according to the ESC2025 guidelines, using available patient records. We compared measured and calculated LDL-C values to target LDL-C levels, noting that modern targets differ from those used in 2016. Failure was defined as equation-based classification of LDL-C target achievement when DM indicated non-achievement.
Results
After excluding patients for whom the risk category could not be reliably assigned, a total of 13,585 patients were included in the analysis. Invalid (negative) equation estimates were individually removed. LDL-C estimates varied significantly across the methods (p < 0.001; Welch-type ANOVA). Statistics for each LDL-C estimate are presented in the Table. The analysis of mean differences, mean absolute errors (MAE), root mean square errors (RMSE), and Cohen's Kappa indicates that the mSNE equation outperformed both FE and M-HE. Plots illustrating the differences between the LDL estimates and DM LDL-C are shown in the Figure below. In our population, the Friedewald equation tended to underestimate LDL-C levels the most, while other equations showed a tendency to overestimate LDL-C at very low HDL-C, and all equations slightly underestimated LDL-C at moderately high total cholesterol (TC) levels. Failure rates indicated a trend of FE > M-HE > mSNE (all p < 0.001; McNemar test, corrected for multiple comparisons).
Conclusions
The modified Sampson-NIH equation provided estimates of LDL-C closest to direct measurement in the Polish primary care population. Further studies are needed to quantify errors in populations treated under modern guidelines.Table 1.Estimate and error metrics.Figure 1.Estimate vs. measured LDL-C.