Temporal trends in respirable dust and respirable quartz concentrations within the European industrial minerals sector over a 15-year period (2002–2016)

ABSTRACT

Objectives

Since 2000 the European Industrial Minerals Association’s Dust Monitoring Programme (IMA-DMP) has systematically collected respirable dust and respirable quartz measurements from 35 companies producing industrial minerals. The IMA-DMP initiative allowed for estimating overall temporal trends in exposure concentrations for the years 2002–2016 and for presenting these trends by type of mineral produced, by jobs performed and by time of enrolment into the DMP.

Methods

Approximately 32 000 personal exposure measurements were collected during 29 sampling campaigns during a 15-year period (2002–2016). Temporal trends in respirable dust and respirable quartz concentrations were studied by using linear mixed effects models.

Results

Concentrations varied widely (up to three to four orders of magnitude). However, overall decreases in exposure levels were shown for the European minerals industry over the 15-year period. Statistically significant overall downward temporal trends of −9.0% and −3.9% per year were observed for respirable dust and respirable quartz, respectively. When analyses were stratified by time period, no downward trends (and even slight increasing concentrations) were observed between 2008 and 2012, most likely attributable to the recent global economic crisis. After this time period, downward trends became visible again.

Conclusions

Consistent and statistically significant downward trends were found for both exposure to respirable dust and respirable quartz. These downward trends became less or even reversed during the years of the global economic crisis. To our knowledge, this is the first time that analyses of long-term temporal trends point at an effect of a global economic crisis on personal exposure concentrations of workers from sites across Europe.

MATERIALS AND METHODS

IMA-DMP database

The longitudinal prospective IMA-DMP exposure database contains personally measured respirable dust and respirable quartz concentrations. Measurements used for the analyses reported here were taken during 33 sampling campaigns between winter 2000/2001 and winter 2016/2017. The respirable dust and quartz exposure data originated from 163 sites owned by 35 companies located in 23 European countries. Companies in Belgium, Denmark, Finland, France, Germany, Greece, Italy, The Netherlands, Norway, Portugal, Spain, Sweden, Switzerland and the UK participated from the start (2002) and were defined as ‘original participants’. Companies or sites in Austria, Bosnia, Czech Republic, Hungary, Poland, Russia, Slovakia, Turkey and Ukraine joined the project later (from 2010) and are designated as ‘recent participants’. In our previously published paper, we have described in detail the organisational structure within IMA-DMP. Briefly, the company-specific exposure data are sent to IRAS-UU and, after meeting rigorous quality criteria, entered into the central IMA-DMP database. The data are not publicly available and are not shared between companies in order to maintain confidentiality. Yet, results of statistical analyses of pooled data are being discussed during biennial debriefing with Occupational Health and Safety representatives from the participating companies. More details of the IMA-DMP database (as it existed per 1 May 2015) can be found in the same paper. For the temporal trend in respirable dust and respirable quartz analyses, a selection was made of measurements present in the exposure database as per 1 July 2017 (31 760 respirable dust and 27 148 respirable quartz measurements).

Our a priori exclusion criteria are presented in figure 1. In short, measurements with extreme concentrations (>100 mg/m3 for respirable dust and >10 mg/m3 for respirable quartz), most likely due to measurement procedure errors, tampering with samples or extreme conditions were considered outliers and excluded. Also, measurements with a sampling duration shorter than 4 hours and longer than 10 hours were excluded. Exposure measurements with missing ‘worker ID’ data were excluded. In addition, the first three campaigns (campaigns 1–3) and the most recent campaign (campaign 33) were excluded because of limited amounts of data available in these campaigns, leaving data from 29 sampling campaigns for analyses. Consequently, the temporal trend analyses considered 25 539 respirable dust and 22 593 respirable quartz measurements collected between summer 2002 and summer 2016 during 29 campaigns, which amount to 80% and 83%, respectively, of the data present in the database per 1 July 2017

As described previously, a standardised common protocol was developed in 2000 and submitted to each of the participating companies. This protocol included all requirements related to how the measurements for the IMA-DMP should be performed. Briefly, personal measurements had to be collected per specific site-job-campaign combination. A unique worker code had to be registered to identify monitored workers. To avoid interlaboratory differences, internal and (accredited) external laboratories are supposed to participate in an interlaboratory round-robin exercises. To what extent this is common practice in the involved laboratories is unclear. Complementary information concerning type of minerals mined or used in the production processes was collected at enrolment and later when new companies and sites entered the programme. The sampling devices had to be either personal respirable dust cyclones with a filter or multifraction samplers containing a sized-selective foam matrix. All samplers used had to be in conformity with the European standard EN-481 (NEN-EN481, 1993).

Appropriate correction factors were applied to account for differences in sampling and analytical efficiency between samplers. The corrections were based on published ratios from different field and experimental studies. The derived correction factors are shown in online supplementary material 1. Quartz concentrations had to be determined with either X-ray diffraction or Fourier transform infrared spectroscopy.

Imputations

Three thousand two hundred six of 25 539 (13%) respirable dust measurements and 6306 of 22 593 (28%) respirable quartz measurements resulted in concentrations below the limit of detection (LoD). To prevent overestimation of measures of central tendency and underestimation of variability in exposure concentrations, we resorted to multiple imputation techniques for concentrations below the LoD following the method described by Jin et al. The imputation was repeated 20 times, creating 20 datasets in order to take into account uncertainties in imputed data.

Linear mixed models to describe temporal trends

Measured exposure concentrations appeared to be log-normally distributed. Therefore, all statistical analyses were carried out on log-transformed concentrations. Linear mixed effect models to estimate temporal trends were elaborated using the PROC MIXED procedure in SAS software V.9.4. In the mixed model, the hierarchical structure of the data was taken into account. Collected repeated measurements on individual workers were nested within job titles, which were consequently nested within a site. A mixed model was fitted with fixed effects for temporal trend (sampling campaign) and random effects of site, job title nested within a site and a random effect worker within job title within a given site. The model can be described as follows:

Ysjwd = β0 + β1Campaign + αs + αsj + µsjw + εsjwd

where Ysjwd=the log-transformed exposure concentration measured on day d=1, 2, …, d, for worker w=1, 2, …, w, in a job title j=1, 2, …, j, at site s=1, 2, …, s;

β0=true underlying mean of log-transformed exposure (intercept);

β1Campaign=fixed effect per sampling campaign (temporal trend);

αs=random effect of site s;

αsj=random effect of job title j within site s;

µsjw=random effect of worker within job title j within a site s; and

εsjwd=random day-to-day effect for day d for worker w within a job title j within site s.

The linear mixed modelling was repeated 20 times and the results were summarised by using SAS procedure MIANALYZE. The temporal trends obtained from the mixed models were expressed as percentages temporal change per sampling campaign (half-year period). Stratified analyses were carried out by job title, by type of mineral and by three time periods (2002–2008, 2008–2012 and 2012–2016).

In our attempts to analyse macroeconomic effect, especially the effect of the global economic crisis, we stratified the data in three time periods; before (2002–2008), during (2008–2012) and after (2012–2016) the global economic crisis.

Additionally, cubic spline models were used to visualise the temporal trends in a less restricted (non-linear) manner. Five knots at the campaigns 6 (2003), 13 (2006), 21 (2010), 26 (2013) and 31 (2015) were placed. The restricted cubic splines models were fitted in SAS using Frank Harrell’s macro %RCSPLINE to create the spline terms in the mixed model.
RESULTS

The estimated geometric means for a specific job at a specific site during a specific sampling campaign were plotted against sampling campaign, as shown in figure 2A,B. Figure 2A shows that geometric mean respirable dust concentrations per job per site per campaign varied over three orders of magnitude from 0.01 to 10 mg/m3. Geometric mean respirable quartz concentrations per job per site per campaign varied even more at four orders of magnitude from 0.1 µg/m3 to 1 mg/m3 (figure 2B).

An unadjusted spline was plotted through the median geometric means respirable dust and respirable quartz concentrations, showing the trends in exposure levels over time. We estimated two separate curves for the original and more recent participants in the IMA-DMP. The curves for respirable dust and quartz for the original participants in figure 2A,B show an overall downward temporal trend in median geometric mean concentration over the 15-year period. The median respirable dust concentrations decreased roughly fourfold from 0.52 mg/m3 in 2002 to 0.12 mg/m3 in 2016. The median respirable quartz concentration declined less steeply about twofold from 0.018 mg/m3 in 2002 to 0.009 mg/m3 in 2016.

The trend for respirable dust in figure 2A suggests a steep decline in concentration levels in the first few years of the project. However, this downward trend reversed in 2008 and even started to increase slightly until 2012, after which it started to decrease again. Figure 2B suggests the same phenomenon for respirable quartz where an overall decrease in exposure over the full period of 15 years is interrupted by a short period of increase in exposure between 2008 and 2012. Sites enrolled recently had clearly higher exposure levels in 2010, compared with original sites participating already for 10 years in the IMA-DMP. These differences in exposure levels between original and recent participating countries were no longer discernible in 2016.

Table 1A shows the number of sites and observations for all, original and recent participating countries. In addition, estimated geometric means in 2002, 2010 and 2016 and estimated temporal trends, obtained from a linear mixed model adjusted for site and job, are presented in table 1A. Overall, we found statistically significant, consistent downward temporal trends of −4.6% and −2.1% per sampling campaign (which translates to −9.0% and −3.9% per year) over the 15 year period for respirable dust and respirable quartz, respectively.

Similar statistically significant consistent downward temporal trends were found for the sites enrolled from the start of the dust monitoring programme from whom 95% of data in the IMA-DMP database originated. However, for site and job, adjusted temporal trends estimated for the recently enrolled sites (between 2010 and 2016) suggested an increase in exposure levels by +1.6% and +5.6% per sampling campaign for respirable dust and respirable quartz, respectively. Both of these trends, however, were not statistically significant due to the still limited amount of collected data and large variability.

Temporal trends and geometric means were estimated for each job and each mineral, as shown in table 1B,C. For respirable dust, statistically significant consistent downward temporal trends per campaign were found for all jobs (table 1B). These trends ranged between −2.9% for ‘crusher’ to −8.4% for ‘plastification worker’. Statistically significant temporal trends for respirable quartz were also observed for almost all jobs (except for the ‘crusher’). They ranged from −10.1% for ‘plastification worker’ to −1.3% for ‘quarry worker’. In general, the downward temporal trends for respirable dust were steeper than for respirable quartz for all jobs except for ‘plastification worker’.

For respirable dust, statistically significant consistent downward temporal trends were found for all minerals produced, ranging from −2.0% for ‘talc’ to −7.3% per campaign for ‘mixed minerals’, as can be seen in table 1C. Downward trends for respirable quartz were statistically significant except for ‘kaolin’, ranging from −0.9% for ‘silica’ and −4.7% for ‘clay’. Again, as noted for jobs, temporal trends estimated for respirable dust were steeper than trends for respirable quartz.

Estimated time trends per time period adjusted for site and job sampled are presented in table 2. For measurement sampled between 2002 and 2008, we found downward temporal trends of −7.4% and −4.0% per campaign for respirable dust and respirable quartz, respectively. During the period (2008–2012), only a slight downward temporal trend was visible for respirable dust (−1.8% per campaign). For respirable quartz, the temporal trend was actually reversed with increasing exposure concentrations during those years (+4.1% per campaign). For the most recent years (2012–2016), decreasing temporal trends were apparent again for both respirable dust and quartz of −7.1% and −8.7% per campaign, respectively. All these trends and differences in trends between the periods were statistically significant (online supplementary material 2) . These changes in trends appeared to be consistent when we performed these analyses stratified by region and size of workforce (data not shown).

Figure 3 shows both the unadjusted splines (as presented in figure 2A,B) and the adjusted (for site and job) splines for both respirable dust and respirable quartz. Over the 15 years of the project, the unadjusted and adjusted time trends look comparable; with an overall consistent decline in respirable dust and respirable quartz exposures. However, between 2008 and 2012, there is a slight difference between unadjusted and adjusted time trends especially for respirable dust. The adjusted time trends showed a levelling off trend where the unadjusted time trends showed an increase in respirable dust exposure levels. For respirable quartz, both unadjusted and adjusted time trends showed similar patterns.
