Avaliação de critérios estatísticos utilizados em programas interlaboratoriais para pesquisa de chumbo em sangue Evaluation of statistical criteria applicable to interlaboratory comparisons for lead in blood

The participation in Interlaboratory Comparison Programme (IP) is a tool used by laboratories to ensure reliability of analytical results. In IPs the selection of a statistical treatment applied to the results can affect the estimation of the “true” value (assigned value) and its uncertainty, the acceptance range and the performance assessment of the participants laboratories. Using the results of the participant laboratories in the Proficiency Test Programme for lead in blood, coordinated by Instituto Adolfo Lutz, Sao Paulo, for different levels of concentration, seven statistical techniques were compared. The assigned value was calculated using the median, or the average value after the elimination of the dispersed results, employing Hampel, Grubbs, Dixon and Z-score’s criteria. The acceptance range was calculated according to Zscore, making use of the standard deviation or using Horwitz’s equation or the Horwitz’s one modified by Thompson. It was also used the current criterion applied to the Program, which considers X designated ± 6, for concentrations [Pb-s] ≤40 µg/100mL and X designated ± 15% X designated for concentrations > 40 µg/ 100mL. It was verified that the assigned values obtained through the statistical techniques used did not present significant differences when the confidence level was 5%. However, the acceptance ranges obtained showed they were influenced by the statistical technique used.