Variability and bias in the analyses of industrial hygiene samples

Information on variability and bias of analyses of industrial hygiene samples is needed for quality control, for design of sampling programs, and for interpretation of results. Data was assembled from interlaboratory programs conducted by PAT-NIOSH, CDC, West Allis Memorial Hospital, and the Toxicology Center of Quebec for 12 types of samples. At the geometric mean concentrations of the reference samples the relative standard deviations of reported results were as follows: Pb, Cd, Zn on membrane filters--4.7%; solvents on charcoal tubes--6.5%; Pb in blood--(CDC) 19%, (West Allis) 13%, (Quebec) 10%; Cd in blood--52%; Cd in urine--40%; As in urine--30%; Hg in urine--25%; F in urine--13%. For free silica on membrane filters, at the geometric mean value the geometric standard deviation was 1.29. For asbestos on membrane filters the statistical analysis was made on the square roots of the values, and the range of one standard deviation was from 68% to 138% of the average value. Four methods for Pb in blood were in close agreement, but three for As in urine differed substantially. Most methods did not achieve the reproducibility required by the NIOSH criteria. The probability of false positive and negative interpretation of results from each group of participating laboratories may be calculated from these data.