Article Text

Cross-shift changes in FEV1 in relation to wood dust exposure: the implications of different exposure assessment methods
1. V Schlünssen1,
2. T Sigsgaard1,
3. I Schaumburg2,
4. H Kromhout3
1. 1Department of Environmental and Occupational Medicine, University of Aarhus, DK-8000 Aarhus C, Denmark
2. 2Department of Occupational and Environmental Medicine, Skive Hospital, Resenvej 25, DK-7800 Skive, Denmark
3. 3Institute for Risk Assessment Sciences (IRAS), Division of Environmental and Occupational Health, Utrecht University, NL 3512 Utrecht, Netherlands
1. Correspondence to:  Dr V Schlünssen  Department of Environmental and Occupational Medicine, University of Aarhus, DK-8000 Aarhus C, Denmark; vivi_schlunssendadlnet.dk

## Abstract

Background: Exposure-response analyses in occupational studies rely on the ability to distinguish workers with regard to exposures of interest.

Aims: To evaluate different estimates of current average exposure in an exposure-response analysis on dust exposure and cross-shift decline in FEV1 among woodworkers.

Methods: Personal dust samples (n = 2181) as well as data on lung function parameters were available for 1560 woodworkers from 54 furniture industries. The exposure to wood dust for each worker was calculated in eight different ways using individual measurements, group based exposure estimates, a weighted estimate of individual and group based exposure estimates, and predicted values from mixed models. Exposure-response relations on cross-shift changes in FEV1 and exposure estimates were explored.

Results: A positive exposure-response relation between average dust exposure and cross-shift FEV1 was shown for non-smokers only and appeared to be most pronounced among pine workers. In general, the highest slope and standard error (SE) was revealed for grouping by a combination of task and factory size, the lowest slope and SE was revealed for estimates based on individual measurements, with the weighted estimate and the predicted values in between. Grouping by quintiles of average exposure for task and factory combinations revealed low slopes and high SE, despite a high contrast.

Conclusion: For non-smokers, average dust exposure and cross-shift FEV1 were associated in an exposure dependent manner, especially among pine workers. This study confirms the consequences of using different exposure assessment strategies studying exposure-response relations. It is possible to optimise exposure assessment combining information from individual and group based exposure estimates, for instance by applying predicted values from mixed effects models.

• AM, arithmetic mean
• GM, geometric mean
• BWsy2, between-workers variance
• BGSy2, between-group variance
• WGSy2, within-group variance
• GSD, geometric standard deviation
• FEV1, forced vital capacity in one second
• SD, standard deviation
• SE, standard error
• WWsg2, within-worker variance in each group
• BWsg2, between-workers variance in each group
• j, mean number of repeated samples in each group
• Bg, weighting factor  =  (WWsg2 × (j−1)) / (BWsg2 + WWgs2 × (j−1))
• exposure assessment
• wood dust
• lung function

## Statistics from Altmetric.com

Exposure-response analyses in occupational studies rely heavily on the ability to reliably and accurately distinguish workers with regard to exposures of interest. Different approaches have been used, in order to maximise the validity of exposure assessment. Using an individual based approach for non-acute effects in general increases the precision of the exposure-response relation, but at the expense of introducing bias (attenuation).1,2 Conversely, a group based approach is in general unbiased but introduces a significant amount of error in the analyses.3 Seixas and Sheppard4 introduced an alternative exposure assessment strategy that uses the advantages of both an individual and group based approach. This strategy can however, only be applied when individual exposure measurement data are available for each study subject. In a recent paper by Heederik and Attfield5 on FEV1 (forced expiratory volume in the first second) and cumulative dust exposure in the mining industry, it was shown that using the same measurement data in different ways had considerably impact on the exposure-response relations observed. Van Tongeren and colleagues6 found large effects on the slope and standard error of the exposure- response relation between FEV1 and current dust exposure using different grouping schemes for estimating exposure in the carbon black industry.

Cross-shift FEV1 by definition takes place during one specific day. Therefore one may argue, that exposure measured on that specific day will be the most appropriate estimate of the relevant exposure. We have earlier revealed an exposure-response relation between bronchial hyperresponsiveness and wood dust exposure, despite a time lag of several months between exposure assessment and clinical examination, suggesting bronchial hyperresponsiveness to be related to average current exposure rather than acute exposure.7 Several studies have investigated the relation between cross-shift FEV1 and wood dust exposure,8–14 and most of them revealed an acute obstructive decrease in lung function by exposure to wood dust. As far as we are aware, no one has explored the implications of using average exposure measures compared to concurrent individual exposure measures.

Recently, a cross-sectional epidemiological study has been undertaken to investigate the relation between wood dust exposure and respiratory impairment among workers in the furniture industry in Denmark. Among other health parameters, baseline lung function and cross-shift decline in FEV1 were investigated.15 Until now, an individual based exposure assessment has been used in the data analysis.

The aim of this analysis was to evaluate the impact of alternative ways of estimating average exposure in an exposure-response analysis of dust exposure and cross-shift decline in FEV1 among wood dust exposed workers.

### Main messages

• There seems to be an association between average exposure and cross-shift changes in FEV1 among non-smoking woodworkers, especially among pine workers.

• It is possible to combine information from an individual and a group based exposure assessment approach, in order to maximise the validity of exposure estimates in terms of both attenuation and precision.

## METHODS

### Exposure assessment

A total of 2303 woodworkers from 54 furniture industries in Viborg County, Denmark, participated in the study, which took place between October 1997 and April 1998. For 1560 persons dust samples as well as data on lung function parameters were available. For 21 persons, repeated exposure measurements but no health data were available. They were included in the exposure assessment as well, making a total of 1581 persons and 2217 measurements. Figure 1 shows a flow chart of the investigation. The time interval between each sampling round was between 5 and 9 days.

Figure 1

Flow chart of the study.

Personal dust sampling was carried out with passive dust monitors described earlier.16,17 The method is based on measuring light extinction before and after sampling on transparent foils. The light extinction increase was reported as dust covered foil area, converted into equivalent inhalable dust concentration by linear regression models, based on earlier and present calibration samples.17 The mean (SD) duration of sampling was 255 (51) minutes.

On the day of dust sampling, information was collected on work task (sanding, cutting, sanding + cutting, assembly/packing, other tasks, mixed tasks), degree of automation (full automatic, semi-automatic, manual) and use of compressed air (yes/no).

### Policy implications

• Cross-shift changes in lung function among woodworkers should not focus solely on current exposure, but should also take into account long term average exposure.

In total, 42% of woodworkers used mainly pinewood, 13% particle board or fibre board, and 6% different kinds of hard wood, mainly beech. The rest, 39%, used a mixture of different wood species.

Exposure to wood dust for each worker was calculated in several ways.

#### Method 1

Using the dust sample collected for each individual on the same day as lung function was measured (individual, 1. round). For 1541 individuals, dust measurements from first sampling round were available, see fig 1.

#### Method 2

Calculating an individual arithmetic mean exposure per person using all samples from individuals with a first sampling round measurement available, a total of 1541 individuals and 2181 measurements (individual, all rounds), see fig 1.

#### Method 3

Using the quartiles of the distribution of the individual arithmetic mean exposure from method 2 to create four groups, with similar number of samples (individual, four groups).

#### Methods 4–5

Estimation of the most efficient grouping strategy.

The following grouping variables were explored: task, degree of automation, factory size (dichotomised <200/⩾200 employees), use of compressed air, combination of task and factory size, and combination of task and factory. For each group, the median (geometric mean) of the individual arithmetic mean distribution was calculated.

Additionally, five exposure groups were formed, by ranking all combinations of factory and task categories according to the arithmetic mean exposure. The quintiles of this distribution were chosen as cut off points.

Contrast in mean exposure levels between groups was calculated as described by Kromhout and Heederik18 using formula 1:

$Math$

where BGSy2 is the between-group and WGSy2 the within-group variance.

#### Method 6

For the grouping based on the combination of task and factory size a weighted estimate was calculated based on the James-Stein shrinkage estimator, first described by James and Stein,19 and further developed to be group specific by Seixas and Sheppard.4 For each of the 12 groups based on a combination of task and factory size, a weighting factor Bg was calculated using formula 2:

$Math$

where WWsg2 is the within-worker variance, BWsg2 the between-workers variance, and j the mean number of repeated samples in each group. A weighted mean of group based and individual based exposure estimates were calculated using the following formula:

$Math$

Where “group mean” is the group mean estimated under method 4 and “individual mean” is the mean estimated under method 2.

#### Methods 7–8

Dust exposure was modelled with two mixed effects models, and predictions from each models were used as measures of exposure. The first model included significant determinants of dust exposure as fixed effects and worker and factory as random effects. The following variables were included: task, degree of automation, use of compressed air, work at night shift and safety representative elected within the last two years. Additionally, the interaction term “assembly/handling × manual work” was included in the model. The second model included the combination of task and factory size as fixed effect and worker as random effect.

### Lung function

Pulmonary function was tested using a dry spirometer (Vitalograph, Buckingham, UK) on a random day during the week at the same day as dust measurements were performed. Among others, FEV1 was measured. Testing was performed in accordance with international guidelines.20 Height and weight were measured before the lung function test. For 1293 subjects, lung function was measured both before and after work. The percentage difference (cross-shift FEV1) was calculated as pre-shift minus post-shift, divided by the largest measured FEV1, multiplied by 100.

From a self administered questionnaire, information about smoking, age, gender, sideline occupation, lung diseases, previous dusty non-woodworking jobs, and seniority in the wood industry, was collected.

### Statistical analyses

As the data were lognormal distributed, analyses were performed with log transformed data with SAS System for Windows version 8.0.

Nested one way analysis of variance (ANOVA) was used to estimate within (WWSy2) and between worker (BWSy2) components of variance.21 Nested two way ANOVA was used to estimate the within (WGSy2) and between group (BGSy2) variability.18

The mixed models were elaborated with PROC MIXED.

The relation between exposure and lung function variables was analysed in SPSS for Windows, version 10.0, using least square regression. As dependent variable, the adjusted (for age, height, gender) residuals of cross-shift FEV1 were used. Dust exposure, smoking, body mass index, and wood species were included in the final models.

All participating subjects gave informed consent, and the protocol has been approved by the Ethics Committee for Viborg County, Denmark.

## RESULTS

Table 1 presents the number of measurements, the GM (geometric mean), and GSD (geometric standard deviation) for dust exposure in the three sampling rounds. No difference was seen between the three rounds, with an overall GM (GSD) of 0.96 mg/m3 (2.05).

Table 1

Inhalable dust concentration (mg/m3) for measurements in three sampling rounds

The overall ratio of within- to between-worker variance was: 0.275/0.269 = 1.02. In table 2, variance components and contrast for different grouping strategies are given. Grouping according to the combination of task and factory size (“12 categories”) and the quintiles of the combination of task and factory (“5 categories”) revealed highest contrast in exposure between groups, 0.51 and 0.64 respectively. Grouping by automation and use of compressed air both revealed contrast below 0.1.

Table 2

Variance components and contrast using different grouping strategies

Figure 2 presents mean dust level and number of persons in each group for the combination of task and factory size (fig 2A) and the quintiles of the combination of task and factory (fig 2B). The number of persons in each group was highly unequally distributed, and ranged from 5 to 429 for the combination of task and factory size and between 179 to 390 in the quintiles for the combination of task and factory.

Figure 2

Mean dust level and number of persons in each group for grouping “task + factory size” (A), and “task + factory, quintiles” (B).

Table 3 shows the correlation coefficients between the eight exposure parameters. High correlation coefficients were seen between the three methods using individual samples (r > 0.84). Additionally, high correlations were seen between individual samples, weighted mean, mixed model 1, and mixed model 2 (r > 0.81). The lowest correlation was seen between the individual samples and “12 categories” (r < 0.43). All correlations were highly significant at the p < 0.001 level.

Table 3

Pearson’s correlation coefficient r for eight exposure estimates for current average exposure

Table 4 gives results of linear regression on exposure estimates and cross-shift FEV1. No statistically significant (p < 0.05) relations were shown, irrespective of which exposure estimate was used. However, the size of the effect (slope) was 70% higher for “12 categories”, than for individual estimates on the day of the lung function testing. As expected, the standard error (SE) of coefficients was increased in grouping estimates compared to individual estimates. The weighted estimate and mixed model 2 had slopes and standard errors in between individual and grouping estimates. Mixed model 1 and “5 categories” had the lowest slopes.

Table 4

Adjusted linear regression on cross-shift decline in FEV1 (the residuals adjusted for age, gender, and height) and dust exposure (1156 individuals)

Table 5 shows results of linear regression on exposure estimates and cross-shift FEV1, stratified by smoking and further by smoking and use of pine. A statistically significant positive association between dust exposure and cross-shift FEV1 was revealed for non-smoking woodworkers for all exposure estimates except “5 categories” and “mixed model 1”. Further stratification by pine revealed a stronger association for pine workers compared to woodworkers using other species, although the difference between the slopes did not reach the level of 95% significance.

Table 5

Adjusted linear regression on cross-shift decline in FEV1 (the residuals adjusted for age, gender, and height), and dust exposure

No relation between dust exposure and cross-shift FEV1 was seen for smokers, irrespective of the chosen measures of exposure.

Among non-smokers, the highest slope and SE was revealed for “12 categories”. The lowest slope and SE was seen for the individual estimates and for “5 categories”, with the weighted estimate and mixed model 2 having slopes and SE in between. Among non-smoking pine workers the same tendency was seen, except for “12 categories” with a slope comparable to the individual estimates. The weighted estimate and mixed model 2 achieved the highest slopes.

The distribution of dust among smokers and non-smokers, and among pine workers and non-pine workers was explored, revealing no overall significant differences.

Using different estimates of exposure only marginally changed the slopes for other variables in the models—that is, smoking, wood species, and body mass index (data not shown). The non-pine workers used quite different wood species. No difference was observed when the relation between dust exposure and cross-shift FEV1 was stratified by these wood species.

## DISCUSSION

We revealed associations between exposure to dust in wood manufacturing and cross-shift decline in FEV1 among non-smoking woodworkers, especially among pine workers.

Several studies revealed an acute obstructive decrease in lung function by exposure to wood dust, but only Mandryk and colleagues8 performed a substantial number of measurements. They found an exposure-response relation between cross-shift decline in FEV1 and 163 inhalable dust measurements from mixed wood species. Lung function and dust measurements were performed on the same work day. Beritic-Stahuljak and colleagues9 found an exposure-response relation between cross-shift decline in FEV1 and exposure to softwood (mainly pine) but not to hardwood. In some studies cross-shift decline in FEV110–12 was found among workers exposed to wood dust, compared to non-exposed subjects. A few studies did not find any relation between exposure to wood dust and cross-shift FEV1.13,14

We have earlier shown increased coughing among woodworkers, most pronounced in non-smokers.15 It may be that the massive effect of smoking “masks” the effect of wood dust. In fact, a “healthy smoker effect” has been suggested earlier.22

Differences in slope and SE of exposure-response relations were revealed, depending on how exposure measurements were treated in the exposure assessment. Increased slopes and increased SE were found for the group based estimates using task and factory size as grouping variable. Despite a larger contrast in the grouping with five categories formed by ranking all combinations of task and factory, this estimate resulted in a decreased slope, compared to the other group based exposure estimates. This is most likely due to unequal distribution of samples in each of the 246 groups, with 80 groups only containing one sample, and the largest group comprising 65 samples. This will have resulted in very imprecise estimates of average exposure for a considerable amount of the 246 groups.23 Even though a grouping strategy was used in this approach, considerable misclassification error could not be prevented.

Grouping by “12 categories” revealed a decreased slope for pine workers compared to non-pine workers, suggesting a less efficient grouping for pine workers. One explanation may by an impact of terpene exposure on cross-shift decline in lung function among pine exposed workers. Terpenes are naturally occurring substances in pine wood.24 Irritation to mucous membranes and impairment in lung function has been documented.25–27 The documentation on terpene exposure in the dry wood industry (for example, furniture industry, joineries) is sparse. Eriksson et al found chronic impairment in lung function among joinery workers compared to controls, and ascribe the effect mainly to terpenes.28 Unfortunately, we were not able to explore the impact of terpene exposure in this study.

The weighted estimate based on the group specific (a combination of task and factory size) James-Stein shrinkage estimator and predictions from mixed model 2 (including task and factory size as fixed effects and worker as random effect) resulted in similar estimates of slope and SE. This is not surprising, since the predicted values using PROC MIXED are achieved using empirical Bayes’ estimates including fixed and random effects,29 which is basically similar to the James-Stein shrinkage method as described by Seixas and Sheppard.4 By using information that distinguishes the individual from the group and information about the group that makes it an identifiable unit, it seems possible to maximise the validity of measurements in terms of both attenuation and precision.

Estimates of average exposure including non-acute determinants for exposure resulted in highest slopes. Our results indicate that cross-shift changes in FEV1 are not only determined by actual exposure level, but also are determined by chronic exposure. This suggests an “irritation effect” of wood dust on the respiratory mucosa not disappearing at the end of the working day, and this may “prime” the lungs for the acute effect.

In order to be able to generalise the model outside the actual study, factory was included as a random effect in mixed model 1. As a consequence, it was not possible to include factory size as an explanatory variable. This was the most influential difference between mixed model 1 and 2, which probably explains why mixed model 1 seems to predict exposure in a less appropriate way, even though more explanatory variables were included. Thus, by leaving out factory specific determinants (that is, factory size), the generalisation of the model is increased, but results in a less sufficient model.

The study by Scheeper and colleagues30 is to our knowledge the only other study in the wood industry, where the ratio of within- to between-worker variance has been explored. In that study, exposure was measured for two randomly selected days or five consecutive days. Within-worker variance was 1.11 times the between-worker variance. The within- to between-worker ratio of variance in this study was 1.02, which was in accordance with our previous findings among the 293 woodworkers with three repeated samples.31 Repeated measurements with short time lag of one week might be influenced by autocorrelation and a possible bias of the variances towards smaller values may be present. Symanski and Rappaport32 however showed that this trend was only presented in a minor part of the series analysed.

Both within- and between-worker variability in exposure concentrations in this study were relatively low compared to earlier studies comprising data on variance components,2,3 suggesting a relatively uniform exposure to wood dust. The overall GM and GSD in the present study were lower compared to an earlier cross-sectional study in the Danish furniture industry,17 suggesting that control measures introduced in the recent past have successfully reduced mean dust levels, possibly by controlling the most extreme exposure situations.

The contrast for grouping by “task” and by “task and factory size” was 0.30 and 0.51, respectively. Kromhout and colleagues3 estimated contrast between 0.15 (job) and 0.57 (plant) for wood manufacturing in one furniture factory and three joineries, based on data from Scheeper and colleagues.30 Van Tongeren and colleagues6 found a high contrast (0.75) in the carbon black industry grouping inhalable dust samples by job and plant. Contrasts in occupational exposure in several other industries were between 0.00 and 0.61 (job) and between 0.21 and 0.84 (job and plant combination),21 making the contrasts in our study an “average” contrast in occupational settings.

Of the total variance and between-worker variance respectively 26% and 17% was accounted for using determinants of exposure, which is in accordance with results from a subpopulation of the study group.33 The explainable part of the variation is lower than stated in earlier studies modelling wood dust exposure using conventional linear regression models30,34 or mixed models including within-worker variation,35 and also when compared to other studies related to airborne particular contaminants.36 Personal exposure to wood dust contains contributions from many sources in the production facilities, for example, the machine present, other machines in the vicinity, dust carried by recirculated air, or dust resuspended by compressed air or by sweeping. The relatively low overall GM and GSD in this study suggests that indirect sources are important for exposure, and determinants for indirect exposure were hardly included in our model. Scheeper and colleagues30 documented that differences between background exposures and near source exposures decrease with decreasing levels of exposure.

In conclusion, there seems to be an association between dust exposure and cross-shift changes in FEV1 among non-smoking woodworkers, especially among pine workers. Studies in the dry wood industry including personal samples on both terpenes and dust exposure are needed in order to explore the combined impact of wood dust and terpenes on temporal changes in lung function.

This study confirms the consequences of using different exposure assessment strategies studying exposure-response relations. Our results suggest that it is possible to combine information from an individual and a group based exposure assessment approach, in order to maximise the validity of exposure estimates in terms of both attenuation and precision.

## Acknowledgments

The study was supported by The Danish Work Environment Foundation; Viborg County; The Danish Medical Research Council; The Wood Industry and Building Workers Union in Denmark; The Danish Lung Association; The Asthma and Allergy Association; and The Health Insurance Fund.

## REFERENCES

View Abstract

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.