Agreement and bias between direct LDL-C measurement and estimation equations in routine clinical lipid samples at Sebbi Hospital, Uganda

Abstract

Low-density lipoprotein cholesterol (LDL-C) is central to cardiovascular risk assessment, yet its accurate determination remains challenging. The Friedewald formula performs poorly in hypertriglyceridemia, while newer formulas such as Martin/Hopkins and Sampson show improved accuracy. In Uganda and sub-Saharan Africa, contemporary comparisons of these equations against direct assays are lacking, increasing the risk of misclassification. We conducted a retrospective method comparison study at Sebbi Hospital, Uganda, analyzing 1,289 lipid profiles (2022–2025). Directly measured LDL-C on the Mindray BS-230 analyzer was compared with Friedewald, Martin/Hopkins, and Sampson formula-calculated LDL-C. Agreement, correlation, root mean square error (RMSE), mean absolute difference (MAD), and misclassification at European Society of Cardiology/European Atherosclerosis Society (ESC/EAS) treatment thresholds were assessed across triglyceride strata (<150, 150–399, ≥400 mg/dL). For TG <150 mg/dL, all methods showed comparable agreement with the Sampson equation (84.13 – 85.03%, k=0.696 – 0.706), with misclassification rates of 14.97 – 15.87% and MAD ranging from 13.87 to 15.54 mg/dL. For TG 150 - 399 mg/dL, agreement declined markedly for Friedwald (67.75%), but remained substantially higher for directly measured LDL-C(86.07%) and Martin/Hopkins (82.12%). Misclassification was highest for Friedwald (32.43%) compared with Martin/Hopkins (17.88%) and directly measured LDL-C (13.93%). For TG≥400 mg/dL, agreement was nominally high for directly measured LDL-C (78.79) but weighted kappa was poor and non-significant for all three methods. MAD was 57.56mg/dL for directly measured LDL-C, 103.45mg/dL for Friedwald and 64.96mg/dL for Martin/Hopkins, with over 93% of directly measured LDL-C values differing from Sampson-calculated values by more than 20mg/dL. Agreement and misclassification varied by triglycerides level, with all methods showing comparable concordance with Sampson equation in the low-TG stratum. Friedwald equation showed the poorest agreement and highest misclassification at moderately elevated triglycerides level. In severe hypertriglyceridemia, large differences were observed across all methods, though conclusions were limited y small sample size. Findings highlight the importance of transparent reporting of LDL-C measurement method and careful interpretation near ESC/EAS treatment thresholds.