Accuracy and repeatability of weighing for occupational hygiene measurements: results from an inter-laboratory...

Gravimetric analysis is a fundamental technique frequently used in occupational hygiene assessments, but few studies have investigated its repeatability and reproducibility. Four inter-laboratory comparisons are discussed in this paper. The first involved 32 laboratories weighing 25 mm diameter glassfibre filters, the second involved 11 laboratories weighing 25 mm diameter PVC filters and the third involved eight laboratories weighing plastic IOM heads with 25 mm diameter glassfibre filters. Data from the third study found that measurements using this type of IOM head were unreliable. A fourth study, to ascertain if laboratories could improve their performance, involved a selected sub-group of 10 laboratories from the first exercise that analysed the 25 mm diameter glassfibre filters. The studies tested the analytical measurement process and not just the variation in weighings obtained on blank filters, as previous studies have done. Graphs of data from the first and second exercises suggest that a power curve relationship exists between reproducibility and loading and repeatability and loading. The relationship for reproducibility in the first study followed the equation log s(R) = -0.62 log m + 0.86 and in the second study log s(R) = -0.64 log m + 0.57, where s(R) is the reproducibility in terms of per cent relative standard deviation (%RSD) and m is the weight of loading in milligrams. The equation for glassfibre filters from the first exercise suggested that at a measurement of 0.4 mg (about a tenth of the United Kingdom legislative definition of a hazardous substance for a respirable dust for an 8 h sample), the measurement reproducibility is more than +/-25% (2sigma). The results from PVC filters had better repeatability estimates than the glassfibre filters, but overall they had similar estimates of reproducibility. An improvement in both the reproducibility and repeatability for glassfibre filters was observed in the fourth study. This improvement reduced the reproducibility RSD at the lowest loading of 0.1 mg from 30 to 18%. Sub-groups of data from the better performing laboratories in the first glassfibre filter study produced similar curves to the robust data from the fourth study. These data are useful in estimating the likely accuracy of measurements from a laboratory where good analytical practices exist.