INTRODUCTION: Analysis of fatty acids (FAs) in biological matrices using GCMS is highly complex and made more so when one considers calibration, analytical accuracy, and what is truth. This paper describes efforts taken to create a 5-log calibration system which covers the dynamic range of, and accurately measures, each clinically relevant FA analyte (C8-C26 chain-length). Challenges discussed include matrix effect differences in calibrator diluents, a lack of well characterized reference material for all analytes, and variable inter-laboratory method comparisons. Performance characteristics explored within are inaccuracy and imprecision of calibrators, bias between laboratories and reference materials, and clinical concordance.
OBJECTIVE: Accurately and precisely measure 34 clinically relevant fatty acids from serum and plasma samples in a single GC-MS injection.
METHODS: Working calibrators were prepared by spiking purified standards for all analytes into calibrator matrix at concentrations to include all clinically significant measurements. To liberate fatty acids, serum or plasma samples were combined with a working IS solution containing 17 stable isotope labeled IS and 2 odd-chain FA dopants and underwent acid and base hydrolysis. Following the addition of 6mL hexane, liquid-liquid extraction was performed. Samples were evaporated and subsequently derivatized with a 50µL pentoflourobenzylbromide derivative solution. Lastly, samples were dissolved in 350µL hexane and injected on a ZB-1 plus (30m x 0.25mm x 025um) column at 0.5µL. The detection and analysis of 34 FA’s was performed by negative chemical ionization and selected ion monitoring on a GC-MS system.
RESULTS: Multiple calibration systems were evaluated for matrix effects as well as harmonization to external CLIA-certified laboratories. Matrix effects were alleviated using ethanolic calibrators (matrix effect <10%) but did not provide inter-laboratory agreement (bias -34%-238%).
As such, establishing traceability to standard reference materials (SRMs) was the next logical approach towards defining trueness. Unfortunately, NIST SRM materials include certified/reference values for 22 of 34 FAs, with measurement uncertainties (U) as high as 66.7%. This propagated to varying results when analyzed by multiple laboratories - stearic acid SRM1980 (U=6.4%) 1.9%-6.9% bias to the target value and erucic acid SRM1980 (U=38.2%) 47.1%-223.5% bias to the target value.
Due to the lack of methodological agreement and poor utility of SRM target values, the next course of action was to ensure repeatable preparation or value assignment of calibrators that are traceable to reference intervals (RI). As such, the available SRMs were measured with low imprecision (1.5%-10.3% CV) over seven batches (N=12-18), contemporaneously to measurement of the reference interval specimens, to establish reference values traceable to the generated RIs. These reference values can then be used for future value verification/assignment of calibration preparation.
To benchmark the performance of the assay to other CLIA-certified laboratories, twenty serum aliquots were analyzed on three discrete occasions. Each discrete analysis was plotted against the mean of all three. The repeatability within each laboratory was demonstrated by the regression slopes and mean CV for a given FA. For this test 95% and 97% of FAs demonstrated slopes between 0.9-1.1 and mean CV < 10%, respectively. The same was true for 95% and 96% of FAs measured at one external laboratory, but only 40% and 13% of FAs measured at a second external laboratory. Nonetheless, clinically annotated specimens shared between labs demonstrated 100% clinical concordance when interpreted against the in-house generated RIs.
CONCLUSION: Providing accurate FA measurement was earnestly explored, and impossible to define, however clinical correlation is possible using precise analytical practices.