Serum versus lithium-heparin plasma for lipid testing: matrix-related bias and calculated LDL-C classification

To assess whether serum and lithium-heparin plasma can be used interchangeably for routine lipid testing and whether matrix-related bias affects calculated low-density lipoprotein (LDL)-cholesterol (LDL-C) classification near clinical decision thresholds.

This prospective paired-sample method-comparison study included 104 fasting adult ambulatory patients referred for routine lipid testing. Serum and lithium-heparin plasma were collected during the same venipuncture and analysed on a Cobas 6000 c501 analyser. LDL-C was calculated using the Friedewald equation when triglycerides (TGs) were <4.5 mmol/L (n=103). Agreement was assessed using paired testing, Passing-Bablok regression, Bland-Altman analysis, intraclass correlation coefficients (ICCs), Cohen kappa and McNemar testing at selected decision thresholds. Interchangeability was interpreted using analytical bias, measurement uncertainty, limits of agreement and threshold discordance.

Lithium-heparin plasma yielded lower total cholesterol and calculated LDL-C than serum, with mean biases (plasma - serum) of -0.15 mmol/L (-3.06%) and -0.14 mmol/L (-4.77%), respectively. TGs and HDL-cholesterol did not differ significantly. Serum-plasma correlations and ICCs were high. LDL-C classification was discordant in 3/103, 2/103 and 7/103 paired samples at thresholds of 1.4, 1.8 and 2.6 mmol/L, respectively; all discordances classified plasma below and serum above the threshold.

Under these local analytical and pre-analytical conditions, the mean bias was modest but directional and sufficient to alter calculated LDL-C classification in a minority of samples near decision thresholds. Laboratories using lithium-heparin plasma for lipid testing should verify matrix agreement locally, document the specimen matrix and avoid alternating serum and plasma during longitudinal monitoring unless equivalence has been demonstrated.