Identificador: TDX:765
Autores: Maroto Sánchez, Maria Alicia
Resumen:
According to the ISO Norm 17025, the accredited analytical laboratories should obtaintheir results accompanied by two basic quality parameters: traceability and uncertainty.Therefore, nowadays the analysts are increasingly urged to assess the traceability of theirresults and to estimate the uncertainty of these results. In this thesis, we have proposeddifferent methodologies to calculate uncertainty in analytical methods that are usedroutinely in the laboratory. These methodologies calculate uncertainty as a whole (i.e. bygrouping terms of uncertainty whenever possible) and use the information generated duringthe validation process and, specially, the information obtained in the assessment oftraceability. Moreover, the uncertainty can also be calculated using the informationgenerated during precision studies, during robustness studies and during the internal qualitycontrol.The advantage of the proposed methodologies is that they require few additional work tocalculate uncertainty because the analytical methods should always be validated beforeapplying them to the analysis of routine samples. We have developed the calculation ofuncertainty at one concentration level when traceability is assessed with several types ofreferences such as certified reference materials (CRM), reference methods andinterlaboratory studies. This uncertainty is calculated using the information obtained in theassessment of traceability. Moreover, this uncertainty can be calculated by using theinformation generated from control charts, precision studies and robustness studies. Thecalculation of uncertainty at one concentration level should only be applied in a restrictedconcentration range because it is assumed that both the method bias and precision are thesame in the whole concentration range. However, many analytical methods are applied inwide concentration ranges and to several kind of matrices. In these cases, we havedeveloped methodologies to calculate uncertainty at several concentration levels and takinginto account the matrix variability in the uncertainty budget. To calculate this uncertainty, itis necessary to verify traceability with several reference samples whose concentrations arecomprised within the concentration range of the routine samples. Moreover, the absence oftwo types of biases should be verified: proportional bias (i.e. that varies with concentrationand is expressed in terms of recovery) and constant bias. We have developedmethodologies for calculating uncertainty when traceability is assessed using spikedsamples. However, these methodologies can also be applied when traceability is assessedusing other type of references such as certified reference materials.Finally, we have studied if the uncertainty of results is underestimated when it is wronglyconcluded in the assessment of traceability that the method bias is not significant. We haveobserved that uncertainty can be very underestimated when the uncertainty of biasrepresents more than 30 % of the final uncertainty. Therefore, it is necessary to havecontributions of the uncertainty of bias lower than 30 % so that the uncertainty of results isnot underestimated. This can be achieved by analysing the reference samples at least tentimes and by using references with lower uncertainty.
Repositori URV