Ann. occup. Hyg., Vol. 46, No. 3, pp. 329-339, 2002
© 2002 British Occupational Hygiene Society
Published by Oxford University Press
Developments in the RICE Asbestos Fibre Counting Scheme, 1992–2000
Institute of Occupational Medicine, 8 Roxburgh Place, Edinburgh EH8 9SU, UK
Received 10 July 2001; in final form 29 October 2001
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONBACKGROUNDDATA AND ANALYSISRESULTSDISCUSSIONCONCLUSIONS AND RECOMMENDATIONSREFERENCES |
|---|
The UK’s Regular Interlaboratory Counting Exchanges (RICE) scheme provides proficiency testing for laboratories counting asbestos fibres by phase contrast optical microscopy, as in the method for measurement of airborne fibre concentrations. From 1984 to 1992, the scheme used circulations of industrial samples containing mostly chrysotile, and reference values were obtained from fibre counts by automated image analysis. In 1992, lower density (<100 fibres/mm2) samples from asbestos clearance operations were added and the new reference values were medians of the laboratories’ determinations. In extensive data from 28 recent rounds of sample exchanges, the new reference values are shown to be more reliable than the old. Average counting levels have changed, with different trends according to fibre density. In low density samples, after initial increases, the levels appear to have stabilized. Counting levels on the higher density samples show a continuing trend of ~0.5% decrease per round. Widening the density range may have reduced the influence of counters’ preconceptions of what values are expected and so their counts on the reference samples may now better reflect their routine counting. The implications of these findings and of other new developments, such as expected new counting rules, are discussed.
Keywords: asbestos; fibre counting; proficiency testing; optical microscopy
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONBACKGROUNDDATA AND ANALYSISRESULTSDISCUSSIONCONCLUSIONS AND RECOMMENDATIONSREFERENCES |
|---|
Concentrations of airborne asbestos fibres are measured by sampling onto membrane filters, which are subsequently mounted on glass slides. Deposited fibres are then counted by phase contrast optical microscopy, following the Health & Safety Executive’s (HSE) standard method MDHS 39/4 (HSE, 1995). This counting is prone to substantial variation between analysts both within and between laboratories. In circumstances where substantial variation can occur, there is a clear need for quality control. Proficiency testing is a valuable method of quantifying performance and it helps counters and laboratories to improve. In the UK, the proficiency of laboratories counting asbestos fibre samples is tested by the Regular Interlaboratory Counting Exchanges (RICE) scheme.
Currently, ~200 laboratories participate in the scheme. Every year, each laboratory is sent three or four batches of eight samples, permanently mounted on glass slides, for counting. Laboratories are classified as satisfactory if their results over four rounds are within target bands around previously assigned reference values for the samples. Satisfactory performance in RICE fulfils requirements for laboratory accreditation by the United Kingdom Accreditation Service (UKAS) and for quality control in MDHS 39/4.
RICE was launched as a publicly available scheme in 1984, and the initial development and operation from 1984 to 1992 have been described previously (Crawford and Cowie, 1984; Crawford et al., 1992). In 1992, several significant technical changes were introduced (Brown et al., 1994) to improve the effectiveness of the scheme. The data generated by the subsequent fibre counting on the RICE samples allow us to examine the impact of those changes on the assessment of proficiency. We also investigate the extent, nature and causes of any changes in the level of fibre counts over this period.
RICE will be affected by the introduction of the World Health Organization (WHO) all-fibre method (WHO, 1997) in the next few years. We describe preparations being made to ensure continuity of the operation of RICE during the transition to the new method.
International standards in quality are as relevant to the conduct of RICE as they are to all other facets of asbestos monitoring and this is more visibly the case since the publication of a standard specifically for proficiency testing schemes—the International Organization for Standardization (ISO/IEC) Guide 43 (ISO, 1997). We discuss how that standard’s guidance on statistical criteria for comparing laboratory results to reference values translates to the specific situation for fibre counting.
Finally, we anticipate moves towards harmonization of proficiency testing schemes in countries in the European Union.
|
BACKGROUND |
|---|
|
TOPABSTRACTINTRODUCTIONBACKGROUNDDATA AND ANALYSISRESULTSDISCUSSIONCONCLUSIONS AND RECOMMENDATIONSREFERENCES |
|---|
The operation of the RICE scheme
RICE is operated by the Institute of Occupational Medicine (IOM), on behalf of the UK regulatory body, the HSE, and with membership arranged through the Health & Safety Laboratory. RICE is overseen by the HSE’s Committee on Fibre Measurement (CFM) and its technical sub-committee on proficiency testing.
Each reference sample has been assigned a reference value for its density and up to 1992 these were means of densities from counts produced by Magiscan, an image analyser. Magiscan counted on average ~10% lower than experienced laboratories; therefore, the target bands defining laboratory performance were centred around 1.1 times the Magiscan reference value.
In each round of sample exchanges, a laboratory counts a batch of eight reference samples and returns its results—numbers of fibres and fields with the derived estimates of fibre density (fibres/mm2)—to the IOM. If enough (75%) of its counts are within the defined target bands, then the laboratory is awarded a satisfactory ‘rating’. The rating is then either 1 (good performance, i.e. 75% of counts in an inner band) or 2 (less good, i.e. 75% within a wider band). If insufficient counts are within the defined bands, then the laboratory is rated 3 (falling short of satisfactory performance) for that round.
From the results of four consecutive rounds, each laboratory is awarded a ‘category’. Categories 1 and 2 are achieved by average performance equivalent to ratings 1 and 2, respectively. Categories 1 and 2 are regarded as satisfactory and lists of laboratories with satisfactory performance in RICE are published each round.
The initial improvements: 1984–1992
Following the introduction of the RICE scheme, variation in counting performance reduced substantially. The initial improvement is seen (Fig. 1) as a large increase, over the first five rounds, in the percentage of laboratories with rating 1; there is a corresponding drop in the percentages with ratings 2 and 3. The subsequent trends were influenced by key events (at rounds 10 and 23) that are described below.
|
At round 10 (1987), the limits of satisfactory performance were tightened (by ~30%, see Table 1), to reflect the improved standards achieved by laboratories. This narrowing of the target caused the drop at round 10 in the percentage of rating 1 scores.
|
Changes implemented in 1992
Introduction of low density samples and new reference values
Ideally, proficiency testing should include the range of sample types and densities that a laboratory might encounter in its routine work. Most of RICE’s pre-1992 samples came from industrial situations, typically with higher fibre densities and containing chrysotile fibres. In contrast, from the 1990s the samples evaluated in the UK mostly came from asbestos removal and clearance operations and contained amphibole asbestos. Therefore, RICE moved to introduce samples from asbestos removal work. These samples, taken mainly to confirm satisfactory completion of clearance, generally have low densities.
A major difficulty was that the Magiscan image analyser could not produce reliable counts for low density samples. For these, the only reliable source of reference values would be visual counts by experienced microscopists, so the decision was taken to dispense with Magiscan and to define new reference values from a set of visual counts. The scheme had already produced slides with up to 60 counts. Slides with a minimum of 15 counts were now used, with the median density being taken as the reference value to minimize the effect of any outlying counts. For the new lower density samples, there were no existing counts, so visual counts had to be generated before the samples could be included in the scheme. In a series of supplementary circulations in 1988–1990, RICE laboratories counted candidate samples, generating 15–25 counts on most, from which median densities were calculated and assigned as initial reference values. Over 560 new samples, plus 45 blank samples, were introduced to the sample stock at the start of round 23 in 1992 (Brown et al., 1994).
A legacy of the supplementary circulations continues in the system for replacing breakages and losses. After round 23, a candidate sample was added to the eight reference samples in each batch. The candidate sample was labelled in the same way as the reference sample. So it is counted by all the laboratories receiving the batch; within four rounds it acquires 15 or more counts and becomes a reference sample. Approximately 15–20 samples are assigned reference values after each round.
New performance assessment criteria
In proficiency testing schemes, the achievement of a performance target should not be unduly influenced by the choice of sample. This implies that measures of discrepancy between counts and reference values should be standardized for the uncertainty associated with the reference values; either differences are standardized by their standard error to form ‘z-scores’ which are then referred to constant limits, or (equivalently) the limits of the target bands are made to depend on the standard error. (The standard error may vary with the reference value.) RICE results are presented in the latter format.
If the only source of variation in fibre counts were from the random distribution of fibres across the filter area, then the counts would behave as Poisson variables. Miller (1984) discusses the effect on the Poisson standard deviation of scaling from the counted numbers of fibres to densities. Under the stopping rule in MDHS 39/4 (i.e. stop counting after 100 fibres or 200 fields), for densities above ~64 fibres/mm2 the standard deviation increases proportionally with the reference value, while for lower densities it is proportional to the square root of the reference value.
In fact, variation in assessments of a slide are much more variable than predicted under the Poisson assumption (e.g. Miller, 1984). However, over years of RICE data, Poisson-based uncertainty inflated by a fixed factor (say x2 or x3 on the standard deviation scale) has been a reasonable fit throughout the density range. Others have suggested different approximations (e.g. MDHS 39/4).
The performance limits (Table 1) were set such that approximately the same proportion of counts fell within them at all densities, based on the data available prior to round 23 (Brown et al., 1994). If that estimation was correct, then a sample’s density would not affect the difficulty of achieving a count within the target range. In RICE, laboratories are sent similar mixes of sample densities (over any four-round period) to prevent any bias that might arise from this approximation.
Performance bands
To help the laboratories identify readily where improvements are needed, the reports from RICE show each count in relation to the performance bands. Outer performance limits (defined in Table 1) separate what is accepted as ‘satisfactory’ and what is not; and inner performance limits subdivide the ‘satisfactory’ range into ‘A’ and ‘B’ bands, as illustrated in Fig. 2. If a laboratory achieves 75% of its counts in band A, then it is assigned a rating 1 for that round; less than that but 75% within bands A and B gives rating 2; more than 25% in band C and it is assigned rating 3.
|
Since the replacement of Magiscan counts by the medians of visual counts removed one source of error in the reference values, albeit a small error, it was expected to create scope for some tightening of the performance limits (Brown et al., 1994).
|
DATA AND ANALYSIS |
|---|
|
TOPABSTRACTINTRODUCTIONBACKGROUNDDATA AND ANALYSISRESULTSDISCUSSIONCONCLUSIONS AND RECOMMENDATIONSREFERENCES |
|---|
Objectives of the data analysis
The specific objectives of this analysis of the counts for RICE rounds 23–50 were:
for high density samples, to examine the impact of changing the reference values from those based on counts by Magiscan image analyser to the median of (a minimum of 15) densities from visual counts by the participating laboratories;
for the new low density samples, to evaluate the assessment of proficiency against the new performance limits.
The data for this study
The database held 1117 reference samples with 72 244 individual evaluations (after exclusion of withdrawn samples). These samples are grouped by type and by reference density in Table 2.
|
Data analysis
From the numbers of counts in performance bands A, B and C, we assessed how the average performance in the scheme was affected by the 1992 changes. We excluded counts made while the samples were candidates, that being consistent with the evaluations reported to the participants.
To examine the relative level of counts, we have calculated median densities for counts in each round and compared them to medians from all counts on that sample from rounds 23–50.
Counts on blank filters were used only to analyse the laboratories’ performances on this type of sample.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONBACKGROUNDDATA AND ANALYSISRESULTSDISCUSSIONCONCLUSIONS AND RECOMMENDATIONSREFERENCES |
|---|
Ratings on high density samples
While Fig. 1 indicates that the percentage of laboratories with rating 1 increased after round 23, that trend may be influenced by data for the new low density samples. Therefore, the comparison using just the high density samples (in Table 3) focuses on the influence of the changed derivation of the reference values. In the four rounds prior to round 23 (where all samples were high density), 1219 ratings were reported, of which 78% were rating 1. In the four rounds after 23, about half the samples were high density and the percentage of rating 1 (if the 1165 ratings were calculated only from those high density samples) would have been 82%. This increase from 78 to 82% is less than the ~10% improvement that we had estimated prior to the change (Brown et al., 1994) based on using the two reference values (Magiscan and visual) for similar comparisons on counting data from rounds 10–16. However, these results support the expectation that the new derivation of reference values would lead to a small improvement in ratings.
|
Effect of the changes on consistency in counts
A few rounds after the introduction of the low density samples, an unexpected consequence became apparent. Figure 3 shows a distinct ‘step’ in the relationship at ~100 fibres/mm2. For the samples with density below 100, the new counts appear to be generally slightly higher than the ones which had been made in the supplementary exchanges. In contrast, for density greater than 100 fibres/mm2, the new counts appear lower than the old counts by ~10–15 fibres/mm2. So for samples in the range 100–200 fibres/mm2, the difference is of the order of 12%. This difference is not large enough to have a major effect on the ratings.
|
The explanation for the difference may be that prior to round 23 the microscopists knew that they were counting samples with density almost always >100 fibres/mm2 for regular RICE rounds, or <100 in the supplementary rounds. In contrast, after round 23 the reference values ranged from 0 to 600 fibres/mm2.
The discovery of this discontinuity between old and new counts led us to the supposition that the more recent counts were better estimates of the values that competent analysts should be expected to produce. Therefore, revised reference densities were calculated from data obtained after round 23 and these came into operation in round 36 (1996), again being based on 15 or more counts per sample.
Blank filters
In RICE rounds 23–50, ignoring excluded samples, there were 40 blank filter samples which accumulated 2204 counts, an average of ~55 counts per sample. Thirty-nine of these samples had reference densities between 0.3 and 2.5 fibres/mm2. However, one had a reference density of 18.8 fibres/mm2, presumably because of some unusual characteristic of the filter such as artefacts resembling fibres or contamination by fibres. Apart from that sample, there were remarkably few outlying counts. Excluding the atypical sample, only seven counts lay outside band B. These seven counts include three extreme outliers with densities ~90–140 fibres/mm2. It is difficult to explain why any counter reported such results—indeed, they may be from clerical errors in reporting. However, nearly all the RICE microscopists obtained acceptable results on these samples.
Sample characteristics and performance limits
The stringency of the performance limits at high density was compared with that for the limits at low density by totalling the counts in each performance band (A, B, C). These results are shown in Table 4 for low (<64 fibres/mm2) and high density (>64 fibres/mm2) samples. Clearly, most of the counts (on all samples) fall into band A and only a few into band C.
|
The low density samples have a higher percentage of band A counts (95%) compared with the high density samples (83%). Correspondingly, the low density samples have a lower percentage of band C counts, 1%, compared to ~5%. The latter difference appears substantial and important because it might affect performance classification.
For samples with density <64 fibres/mm2, the relative stringency of the limits was compared for samples subdivided into three ranges of density: 0–10, 10–30 and 30–64 fibres/mm2. The percentage of band A counts increased from ~91% to nearer 98% as density decreased. Since the samples with density <64 fibres/mm2 are mostly of the same type, these differences in the percentage are therefore probably due to the positions of the limits. Similar subdivision of samples >64 fibres/mm2 into four density ranges produced percentages close to their mean of 83%.
These data suggest that the current definition of the performance limits overcompensates for the larger relative variation at the lowest densities, producing a disproportionately high level of apparent agreement. The recent data will provide a basis for us to revise the target band widths at lower densities. In the interim, it remains appropriate and necessary to stratify the batches of samples by density such that all participant laboratories encounter a similar mix of densities over any four-round period.
Density is not the only factor that differs between the clearance samples and the industrial samples. The former samples contain mostly amosite asbestos and the latter are mostly chrysotile, thus confounding the influence of density and asbestos type in most comparisons. However, some of the clearance samples have densities >64 fibres/mm2 and for that part of the density range comparisons by sample type (in Table 4) show more frequent band C counts (5%) for the industrial chrysotile samples than for the asbestos removal samples (at ~3%). This is consistent with the industrial samples being more difficult to assess.
Temporal trends in level of counts
During the rounds since round 23, the mix of densities and types in the RICE samples remained approximately constant, but a plot of percentages of counts in bands shows an apparent drift over these 28 rounds. In Fig. 4, the percentage of A band counts decreased from ~90% towards 80%. The bands B– and C– show gradually increasing percentages, respectively, from 4 to 8% and 2 to 4%. These B– and C– bands correspond to counts below the reference density. In contrast, the percentages for the B+ and C+ bands show, respectively, a slight decrease and negligible change. Such a small drift has had only a small effect on performance assessments of laboratories.
|
A direct indication of the drift in the counting level was obtained from the median densities. These medians were calculated from the usually three, four, five or six counts on each sample in each round, but excluded medians from fewer than three counts. That median was then divided by the current reference value to give a ratio for that sample in each round. Finally, for each round, we calculated the average ratio for all samples in the round.
This average ratio from 376 samples indicates the level of counts in each round as compared to the reference values. The average ratios are plotted in Fig. 5 for samples subdivided by density. The total and subsets follow broadly similar trends. Closer inspection suggests that the low density samples showed a high ratio when first introduced (round 23), then a gradual downward trend from 1.07 at round 24 to 1.0 at round 35. Then, from round 35, the average ratios fluctuate mainly within the range 0.95–1.0 and appear to have reached an approximately stable level. The high density samples show a continuing decrease from an average ratio of 1.05 at round 23 to 0.9 at round 50.
|
Change and stability in counts on low density samples
For the low density samples, the results described above (Figs 3 and 5) show two changes of similar magnitude, but in opposite directions (i.e. an increase at round 23, then a gradual decrease between rounds 23 and 35). Hence, the counts from round 35 onwards are likely to have been of similar level to those in the supplementary exchanges.
To test this deduction, we used the data from reference samples which had been in the scheme throughout the period from round 23 to round 50 (318 samples) and looked at how the level of counts (as median densities) changed among four time periods:
before round 23—before the changes (using the round 23 reference value);
rounds 23–26—immediately after the changes;
rounds 27–35;
rounds 36–50.
We calculated the median of the counts made on each sample within each of the above four periods. Then these medians were divided by the round 23 reference density. Thus, the medians for the first period (the supplementary exchanges and routine rounds prior to round 23) were normalized to 1. Then, averaging over all samples in each period, we obtained a mean ratio for that period relative to the initial period. By calculating these mean ratios for samples subdivided by density band (as estimated at round 23), we obtained the results in Fig. 6. For each density range, the trends are from substantial numbers of samples (32–81) except for the range 64–100 fibres/mm2 (only seven samples from asbestos removal).
|
The pattern of change is readily apparent from Fig. 6. For the samples in the ranges 10–30 and 30–64 fibres/mm2, the average ratios for the final period are 0.94 and 0.98, respectively, supporting the deduction from Figs 3 and 5 that counting levels on most low density samples had returned to approximately the same level. For the relatively few samples in the range 64–100 fibres/mm2, the mean ratio for the final period was also close to 1. However, for the lowest density range (0–10), the average ratio for the final period was only 0.84. For the high density industrial chrysotile samples (100–300 fibres/mm2), the counts appear to have dropped on average by ~15% compared to those made prior to round 23. For samples with density >300 fibres/mm2, the decrease is ~10%. Possible reasons for these trends are suggested later.
Stability of counting level on blank samples
Since the blank samples had given median densities in the range 0.3–2.5 fibres/mm2, we compare the trends for those samples with the trend (of decreasing counting levels) observed for the clearance samples in the lowest density range (0–10 fibres/mm2) in Fig. 6. The few blank samples that had been present throughout also showed the median densities tending to decrease over the four time periods. This finding is interesting, considering that blank filters should contain no fibres.
The counts on the blank samples may be due to imperfections in the clarity of the filter, rather than contamination with fibres. So a decrease in the count on blanks may be due to a gradual change in the clarity of the filter and the disappearance of false ‘fibres’. Similar changes in the background of the lowest density clearance samples might explain why the counts on those samples showed a downward trend unlike those for the clearance samples with higher fibre loading.
Another possible explanation for changes in level of fibre counts could be drift in the operators’ performance when their expectations for the fibre densities of samples within the scheme had changed. As discussed below, this may be the explanation of the trends for the denser samples. However, a combination of effects seems likely when the trends differ so markedly among bands of sample density.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONBACKGROUNDDATA AND ANALYSISRESULTSDISCUSSIONCONCLUSIONS AND RECOMMENDATIONSREFERENCES |
|---|
Visual reference counts instead of Magiscan counts
The results obtained after switching reference values from Magiscan counts to visual counts produced better ratings, as had been expected. By redefining the reference values, the target was moved closer to the counts that competent analysts produce. This was not simply a circular comparison of counts to reference values derived from the same counts. In RICE, the slide exchanges are organized such that the proficiency testing counts are mainly from laboratories other than those who produced the counts contributing to the original reference value. So the achievement of better ratings reflects a general consistency between analysts.
Effect of presenting a wider range of reference densities
The relative levels of counts before and after the change at round 23 showed an unexpected but explicable pattern. The microscopists are probably influenced by knowledge of the density range of the samples. Inclusion of the low density samples extended the density range of the RICE samples and removed a reason for some analysts to prefer, perhaps subconsciously, higher counts when counting RICE batches. (Conversely, their subconscious approach to the low density samples was likely to be somewhat different when the samples were in routine RICE rather than in the supplementary exchanges where all samples had been low density.) Such a change in the subjective approach to the counting would explain the transient nature of the changes in counting level just after the introduction of the low density samples in RICE. This seems to be the only feasible explanation for counts apparently increasing on one subset (the low density samples) and simultaneously decreasing on the other subset (the high density samples).
As an additional benefit, the target satisfactory range became a smaller fraction of the widened range of possible densities, so counts become less likely to fall in the target range merely by chance.
Temporal stability of counting level
The trend seen in the counting levels for most low density samples suggests that the counts on these samples may well have reached a stable level. For example, these counts appear to be at approximately the same level in the recent RICE rounds as in the supplementary exchanges, as indicated in Fig. 6.
In contrast, the counts on the high density samples appear to be following a trend of continuing decrease, albeit very slowly. Over the 28 rounds, the counting levels (median densities) decreased by ~15%, i.e. <0.5% per round on average. So the difference over 10 rounds is small compared with the width of the performance bands for acceptance of counts as satisfactory. Nevertheless, it clearly needs to be considered, for example in reviewing the possibility of tightening the performance bands.
A practical implication of the temporal drift in counts on high density samples is that reference values (median densities) derived from counts produced in RICE rounds 48–50 would be, on average, ~5% lower than if the same samples had been counted during rounds 38–40. The performance bands therefore need to allow for this uncertainty (temporal bias). Adequate allowance—perhaps more than adequate allowance—is given by the present band width and definition of satisfactory performance, since most of the laboratories regularly achieve counts in band A and almost all regularly achieve satisfactory performance. Nevertheless, the current findings indicate the importance of continuing to monitor the relative level of counts in future rounds.
International guidance on proficiency testing schemes
International standards have been published recently for the operation of proficiency testing schemes, for example ISO Guide 43 (ISO, 1997). This guide includes recommendations on the estimation and assigning of reference values and on the assessment of laboratory agreement with the assigned reference values.
ISO Guide 43 recommends that where the participants’ results are used to determine the assigned values, techniques should be in place to minimize the influence of extreme results. In RICE this is achieved by using a median from a minimum of 15 counts; a median is not readily influenced by extreme results and 15 counts ensures a stable estimate.
Extreme or outlying values in small data sets have a distorting influence on estimates of variability, such as variances or standard deviations, much greater than their effects on averages. RICE avoids this problem by basing its estimates of variability on a common underlying relationship between reference values and their standard deviations, derived from large amounts of data. This satisfies the recommendation of ISO Guide 43 to use reliable estimates of variability.
Preparation for WHO fibre counting rules
The WHO all-fibre counting rules are expected to be introduced into UK regulations by ~2004. There are consequences for the operation of RICE and for laboratories’ internal quality control schemes, and there is a need to ensure that the proficiency of fibre counting can be tested and assured during the period of change.
The current method (MDHS 39/4) for counting asbestos fibres excludes fibres touching non-respirable particles, which are defined for this purpose as particles with diameter >3 µm. In contrast, the new WHO rules (WHO, 1997) will include those fibres. Counting levels will therefore rise, although only on samples where there are both fibres and non-respirable particles. Consequently, this change in counting rule will give rise to a need for appropriate changes in the assigned reference values.
Many of the RICE samples have low levels of non-respirable particulate dust and would be unaffected, so it was necessary to identify which samples would be affected. Initially this was done for most of the samples by one microscopist counting fibres in 60 fields on each sample and keeping tallies of the fibres counted under the current rules and the additional fibres to be counted under WHO rules. For samples where this initial estimate indicated that the increase would exceed 10%, further similar counts of current fibres and additional fibres are being produced. The results so far (from two further counts on most of these samples) demonstrate that for some samples the density will double. However, further work continues to confirm these estimates and their applicability to counts from the RICE laboratories in general.
A change to the WHO all-fibre counting rules in the UK will be a step towards harmonization of counting procedures in EU countries, since some countries already use the equivalent of the WHO all-fibre rules (as regards the treatment of fibres touching non-respirable particles) for asbestos fibre counting. For those laboratories that operate the two fibre counting methods, MDHS 59 for man-made mineral fibres and MDHS 39/4 for asbestos fibres, the switch to WHO all-fibre rules will reduce the workload for microscopist training and for participation in proficiency testing and in-house quality control procedures.
International harmonization of fibre counting proficiency testing schemes
Within Europe, there are several national fibre counting proficiency testing schemes and a recent comparison of these schemes (Arroyo and Rojo, 2001) has looked at the stringency of the performance limits for counts in each scheme. Then, based on the assumption that intrinsically the average consistency of counting performance of laboratories would not differ substantially between countries, the authors proposed the changes in limits that would create similar percentages of counts falling within the target bands in all the schemes. Their suggestion is similar to moving the RICE criterion for satisfactory performance closer to Category 1 (instead of Category 1 or 2). The results presented here suggest that some such change is feasible, but also reveal the influence of some of the main factors (such as stability of the reference values) that affect the scope for tightening the criteria.
|
CONCLUSIONS AND RECOMMENDATIONS |
|---|
|
TOPABSTRACTINTRODUCTIONBACKGROUNDDATA AND ANALYSISRESULTSDISCUSSIONCONCLUSIONS AND RECOMMENDATIONSREFERENCES |
|---|
The study showed that there have been changes in the median level of counts by RICE laboratories and that these followed a quite complex pattern, as follows:
The counting levels showed some distinct changes over the 28 rounds of RICE following the introduction of the clearance samples and other associated changes.
The trends in counting levels indicate that counters’ prior knowledge of the density range, not surprisingly, may have influenced the level of counts in the past.
Counts on the asbestos clearance (low density) samples showed an initial, temporary increase (compared to counts in supplementary exchanges of low density samples on their own), but appear to have returned to an approximately stable level.
The very low density (<10 fibres/mm2) clearance samples and also the blank samples showed a continuous decrease in counts, unlike the other clearance samples. An explanation might be gradual change in the background of mounted samples causing false ‘fibres’ to disappear.
On the high density, industrial chrysotile samples: (i) counting levels (observed as the median densities) decreased gradually, by ~0.5% per round; and (ii) over 28 rounds, this drift amounted to 15%, which is still small compared to the target bands for acceptable counts.
The study also indicated recommendations regarding any future decisions on tightening the performance bands. In particular:
the ratings achieved in rounds 23–50 support the expectation that the switch to reference values based on visual counts would make the targets easier to achieve and therefore make more stringent limits appropriate;
the temporal drift in median values will contribute to the uncertainty in the reference values and therefore should be taken into account when the performance bands are reviewed;
the approximation that variation is constant on the square root scale for low density samples, although supported by our earlier study (Brown et al., 1994), overcompensates for the relatively greater variation in counts on low density samples and therefore a minor tightening of the limits for the lower density samples becomes appropriate.
Finally, the finding of the above overcompensation in the performance bands for low density confirms that the stratification of the mix of densities of RICE samples (in the batches sent to each participant laboratory over any series of four rounds) is necessary. Even if this overcompensation were negligible, inclusion of a range of types and densities of samples would still be desirable for reliable assessment of counting proficiency.
Acknowledgements—This study was funded by the HSE. The conclusions and opinions are those of the authors and do not necessarily reflect HSE policy.
|
FOOTNOTES |
|---|
* Author to whom correspondence should be addressed.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONBACKGROUNDDATA AND ANALYSISRESULTSDISCUSSIONCONCLUSIONS AND RECOMMENDATIONSREFERENCES |
|---|
Arroyo MC, Rojo JM. (2001) A proposal for harmonising laboratory performance criteria in national asbestos fibre counting schemes. Ann Occup Hyg; 45: 447–55.
Brown PW, Crawford NP, Jones AD, Miller BG, Maclaren WM. (1994) Development of the Regular Interlaboratory Counting Exchanges (RICE) Scheme to include visual reference counts and samples from asbestos clearance operations. Ann Occup Hyg; 38: 687–703.
Crawford NP, Cowie AJ. (1984) Quality control of airborne asbestos fibre counts in the United Kingdom—the present position. Ann Occup Hyg; 28: 391–8.
Crawford NP, Brown PW, Cowie AJ. (1992) The RICE and AFRICA schemes for asbestos fibre counting. Ann Occup Hyg; 36: 59–69.
HSE. (1995) MDHS 39/4—Asbestos fibres in air: sampling and evaluation by phase contrast microscopy (PCM) under the Control of Asbestos at Work regulations. London: HSE.
ISO. (1997) Guide 43. Proficiency testing by interlaboratory comparisons—Part 1. Development and operation of proficiency testing schemes. ISO/IEC.
Miller BG. (1984) Statistical method for analysis of membrane filter samples of airborne asbestos. Ann Occup Hyg; 28: 217–27.
WHO. (1997) Determination of airborne fibre number concentrations. Geneva: WHO.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  T. OGDEN Annals of Occupational Hygiene at Volume 50: Many Achievements, a Few Mistakes, and an Interesting Future Ann. Hyg., November 1, 2006; 50(8): 751 - 764. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. GRZEBYK, E. KAUFFER, and L. FREVILLE Proposal to Adapt the Workplace Analysis Scheme for Proficiency (WASP) Programme to Fibre Counting Tests Ann. Hyg., June 1, 2005; 49(4): 325 - 334. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||