Development of methods for the determination of vitamins A, E and β-carotene in processed foods based on supercritical fluid extraction: a collaborative study

New methodologies based on supercritical fluid extraction (SFE) have been developed for the determination of fat-soluble vitamins in processed foods. The results obtained so far indicate that SFE is well suited to extraction of fat-soluble vitamins from food products, although validation work is required to establish accuracy and precision. The vitamins investigated were A, E and β-carotene, and the processed foods were UHT milk, milk powder, minced meat, liver paste, infant formula, canned baby food and margarine. Extraction equipment employed analyte collection on either a solid-phase trap or in a solvent. After extraction, the samples were saponified and the vitamins determined using reversed-phase liquid chromatography with ultraviolet or fluorescence detection. Sample throughput was at least 12 samples day -1 , i.e. at least twice the number achievable with a conventional extraction methodology. The detection limits for the vitamins in different processed foods were well below 0.1 µg g -1 . Recoveries (in comparison with vitamin levels obtained using conventional solvent extraction) were close to 100% for experienced personal with access to modern automatic equipment. To reach this level, it was necessary to protect the vitamins with an antioxidant during the different steps of the analysis procedure, to add methanol or ethanol to the extraction cell to facilitate the analyte extraction from the food matrix, and when using a solid-phase trap, to employ a fractionated extraction-elution procedure to prevent breakthrough losses. The developed methods were tested in a validation exercise between five laboratories, which had taken part in the method development, and in an intercomparison between 10 laboratories including laboratories with less experience of vitamin determination. The within-laboratory RSD was generally ≤11%. The average of the between-laboratory relative standard deviation (RSD) was about 23% in the validation, and increased to about 40% in the intercomparison. Ruggedness tests performed at different steps of the project showed that different types and models of equipment did not give large differences in recoveries. Thus, the increasing RSD can largely be ascribed to differences in experience in vitamin analysis of the participants.