Lipoprotein(a) analytical variation and cardiovascular thresholds

The study was designed to evaluate the analytical performances of six automated immunoturbidimetric lipoprotein(a) (Lp(a)) assays, each corresponding to a combination of a reagent and an analyzer (Binding Site Optilite Lp(a) kit, Diagam Lp(a) reagent on Mindray BS480, Diagam Lp(a) reagent on Roche c502, Randox Lp(a) reagent on IDS-ISYS, Roche Tinaquant LPA2 on Roche c702 and Siemens Atellica CH Lp(a)/Atellica^® CH 930) using results expressed in mg/dL or nmol/L.

Imprecision was assessed on manufacturer controls according to the CLSI EP15-A3 protocol. One hundred and 45 patient serum samples from the cardiology department were assessed to compare each method using Spearman's correlation, Passing-Bablok regression and Bland-Altman analyses. Percent agreement was calculated according to the European Atherosclerosis Society and American College of Cardiology/American Heart Association Lp(a) cut-offs for cardiovascular risk.

All immunoturbidimetric methods achieved the desirable imprecision of 5.1 % based on biological variation. Highly significant correlations were found across multiple comparisons (p<0.001). The absolute mean of the difference according to Bland-Altman ranged between 0.4 and 2.6 for Lp(a) assays in mg/dL and between 0.6 and 6.5 for Lp(a) assays in nmol/L. The percent agreement between two assays varied from 87.6 to 96.6 % and from 82.8 to 93.8 % using the EAS or the ACC/AHA thresholds.

Despite acceptable correlation among the six Lp(a) turbidimetric assays, a risk of "misclassifications" between two assays up to 12.4 % with the EAS recommendations and 17.2 % with the ACC/AHA guideline remains a major concern. Standardization of methods, including expression of results in nmol/L is needed.