Summary

Exposure to airborne fibres in the workplace is measured using a counting technique based on phase-contrast optical microscopy. Fibre counting has a bias and is extremely variable, but these can be controlled by good quality control practices. In accordance with the requirements of official methods that have been established both in Quebec and internationally, laboratories (or counters) performing this type of analysis must, for the purposes of assessing the reliability of their results, participate in an inter-laboratory quality control program using samples that are representative of their activities.

This report provides an assessment of the IRSST Quality Control Program from 1992 to 2011. The program was developed from samples taken in various Quebec industrial sectors (textile; mining; sites for removing materials containing asbestos; brakes and the dismantling of industrial furnaces) and containing artificial mineral fibres (AMF) and asbestos (amosite or chrysotile or mixed fibres, i.e., chrysotile + AMF; chrysotile + fibrous gypsum).

The principal objective of the research was to review the IRSST Quality Control Program for Fibre Counting by making use of the database generated from 1992 to 2011. To this end, the specific objective was to identify, through longitudinal analysis, the various determinants associated with the samples and counters affecting the accuracy and variability of the fibre counting.

A slide bank, consisting of 132 samples, was circulated among 660 participants, including 52 reference counters, during the 79 rounds covered by the review. The 32,777 results generated were recorded in a database that facilitated linking the sample characteristics (industry, fibre type, fibre mix and fibre density) and the counter characteristics (status -- reference (R) or non-reference (NR); affiliation; experience; and involvement). Specific requests and cross analyses, achieved by combining characteristics relating to the samples and the counters, were used to investigate the influence of various determinants and to verify their statistical significance using analysis of variance.

The R counters obtained an average coefficient of variation (CV) of 17%, two times less than that obtained by the NR counters (35%), and this across all variables. The mean accuracy, obtained by dividing the results of the NR counters by their target value (results of the R counters) was 1.06 (indicating an overestimation of 6%).

Of all parameters having a statistically significant effect (p <0.001) on the accuracy of the results of fibre counting, fibre type was the most determining parameter, followed by industrial sector, fibre density and, lastly, counter affiliation.

Thus, for all fibre types, the CV of the R counters was constant (<20%), with the exception of the chrysotile + fibrous gypsum mixture (22%). For the NR counters, the CV ranged from 27 to 40% as follows: amosite < AMF  < chrysotile + AMF  < chrysotile < chrysotile + fibrous gypsum. The most significant biases were observed in samples containing only chrysotile (-18 to 12%) or in a mixture with fibrous gypsum (+ 15%), whereas the results were more accurate in the case of amosite fibres (1%). The counters generally had poorer performance with samples containing chrysotile. In fact, the chrysotile fibres were fine, curved and less visible under the optical microscope than were the amosite or AMF fibres.

Across the various industrial sectors, the variability (or CV) of the R counters remained relatively stable (16 to19%) while that of NR counters increased from 30 to 59%, as follows: textiles < brakes < removal < mines < dismantling. An average positive bias of 2 to 12% was observed depending on the sector involved: brakes < textiles ≈ removal < dismantling < mines. The textile sector is therefore the industry with the lowest level of difficulty of analysis. Samples from the textile sector were composed primarily of fibres and contained few interfering particles.

The variability and accuracy of the counts were also influenced by fibre density or load. The CV was higher for samples of low density (39%) than it was for those of high density (25%). The accuracy was greater for samples of high density (- 3%) than it was for those of low density (+ 13%). This is a known phenomenon: the counters tend to overestimate the number of fibres in a sample that is only lightly loaded, while the results are underestimated when the density is higher.

When the results of two significant determinants, fibre type and industry, were combined, the fibre type factor was always preponderant. Only the removal-chrysotile combination exhibited the greatest difficulty for analysis, in terms of variability and accuracy.

The counters’ experience in terms of their affiliation appears to have an impact on the variability and accuracy of the results. Thus, the counters from private firms (consultants), who accounted for 93.3% of all participants, resulted in a higher CV for all types of fibres and industries. However, a high hiring rate was observed for this group throughout the program, and this generated a significant number of inexperienced counters, something that might explain the increasing variability of the results. In fact, over 71% of all the counters had no experience at the time of their first involvement. Although the variability of the NR counters increased slightly over time, probably due to the poorer performance of the new counters, an improvement in accuracy was observed once these counters had gained actual experience.

The success rates of the IRSST program (95.1%) and the British program, AFRICA (90-95%), were similar, while those of the Spanish (85-87%), Belgian (81%) and French (85%) programs were lower. The conditions contributing to better performance were related to fibre type (MFA or amosite), industrial sector (textile) and density (> 800 f / mm²).

The high degree of variability in fibre counting reveals the need to participate in an inter-laboratory (or collaborative) program to improve the performance of this type of technique of analysis. The poor performance of less experienced counters highlights the need for additional training. Given the significant influence of the variables linked to samples (fibre type, industrial sector and fibre density), maintaining a bank of diverse samples that are representative of various work environments is strongly recommended in the continuation of the IRSST Quality Control Program for Fibre Counting.