Uncertainty estimation and in-house method validation of HPLC analysis of carotenoids for food composition data production

• Autor(a) principal: Dias, M. Graça
• Data de Publicação: 2008
• Outros Autores: Camões, M. Filomena G.F.C., Oliveira, Luísa
• Tipo de documento: Artigo
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10400.18/73
• topic: Composição dos Alimentos; HPLC; Validation; Tomato; Analytical and pre-analytical uncertainties

The method for separation and quantitative determination of the main carotenoids in food by high-performance liquid chromatography (HPLC) was in-house validated. Tomato (Lycopersicon esculentum M.) as food matrix was used to demonstrate its linearity, repeatability, intermediate precision, detection and quantification limits, sensitivity and bias. In addition, stability of carotenoids was studied in function of temperature and time. Method accuracy was quantified through measurement uncertainties estimate based on this validation study. Furthermore, a study was conducted to evaluate variability coming from location in an experimental field composed by 12 subfields. The use of two metal free reverse phase columns and an organic mobile phase based on acetonitrile, methanol and dichloromethane enabled the separation of the six target compounds (trans-a-carotene, trans-b-carotene, b-cryptoxanthin, all-lycopene, lutein, zeaxanthin) within a 30 min run; detection at 450 nm and external calibration allowed the quantification of the analytes. Carotenoids concentration and measurement uncertainty, in mg/100 g, in tomato varieties ‘‘lido” and ‘‘for salad” were, respectively, 1.0 ± 0.14 and 0.39 ± 0.056 for trans-b-carotene, 8 ± 2.0 and 2.3 ± 0.57 for all-lycopene and 0.10 ± 0.017 and 0.08 ± 0.015 for lutein; trans-a-carotene, b-cryptoxanthin and zeaxanthin were not detected in both varieties (detection limits, in lg/100 g, 0.81, 0.57 and 0.77, respectively). For b-carotene and lutein, uncertainty associated with the entire process including small-scale within-region variation was statistically different, at a significance level of 5%, from measurement uncertainty (which includes sampling in the laboratory).