Validation of a reference method for total cholesterol measurement in human serum and assignation of reference values to proficiency testing samples

Our objective was to develop a reference method to measure total cholesterol in human serum, in order to assign values and assess the accuracy of field methods in French clinical laboratories. A reference method based on gas chromatography coupled with mass spectrometry and isotope dilution (GC–IDMS) was developed and validated. It was then used to assign reference values to five frozen serum samples from voluntary proficiency testing schemes gathering 170 French clinical laboratories. Three peer groups were defined and bias against the reference method target value was calculated. Accuracy of the reference method was assessed against NIST SRM 1951b. Bias of the reference method was less than 0.5% and imprecision was less than 1.0%. Our study indicated that field methods tended to overestimate total cholesterol concentration, mean bias being +5.02%±1.02%. The most popular methods (phenolic chromogen with spectrophotometric detection, 80% of participants) exhibited the highest bias (peer group mean bias: +5.51±1.24%). Neither these methods nor those using a non-phenolic chromogen with reflectometric detection (10% of participants, peer group mean bias: +4.20±1.44%) met NCEP recommendations according to which bias should be less than 3%. Only the methods using a non phenolic chromogen with a spectrophotometric detection met these recommendations (10% of participants, peer group mean bias: +1.39±2.75%). As all three peer groups provided positively biased results, the consensus mean usually used to assess the trueness of routine methods is biased as well, which results in an erroneous estimation of method bias. Therefore, this study highlights the value added by reference method target values to assess trueness of field methods and monitor performance of clinical laboratories. ► A GC–IDMS reference method for the measurement of total cholesterol in human serum was implemented. ► The reference method was used to assign reference values to frozen serum samples of five French PT schemes. ► Our results show that cholesterol measurements are positively biased, mean bias being +5.02%±1.02%. ► The most widely used methods did not meet NCEP recommendations. ► This study highlights the input of reference method target values when the consensus mean is biased.