Biopersistent Granular Dust and Chronic Obstructive Pulmonary Disease: A Systematic Review and Meta-Analysis

Objective Applying a systematic review to identify studies eligible for meta-analysis of the association between occupational exposure to inorganic dust and the development of chronic obstructive pulmonary disease (COPD), and conducting a meta-analysis. Data Sources Searches of PubMed and Embase for the time period 1970–2010 yielded 257 cross-sectional and longitudinal studies on people exposed to inorganic dust at the workplace with data on lung function. These studies were independently abstracted and evaluated by two authors; any disagreement was resolved by a third reviewer. Of 55 publications accepted for meta-analysis, 27 investigated the effects of occupational exposure to biopersistent granular dust (bg-dust). Methods A random effects meta-analysis allowed us to provide an estimate of the average exposure effect on spirometric parameters presented in forest plots. Between-study heterogeneity was assessed by using I2 statistics, with I2>25% indicating significant heterogeneity. Publication bias was investigated by visual inspection of funnel plots. The influence of individual studies was assessed by dropping the respective study before pooling study-specific estimates. Results The mean FEV1 of workers exposed to bg-dust was 160 ml lower or 5.7% less than predicted compared to workers with no/low exposure. The risk of an obstructive airway disease—defined as FEV1/FVC < 70%—increased by 7% per 1 mg· m-3 respirable bg-dust. Conclusion Occupational inhalative exposure to bg-dust was associated with a statistically significant decreased FEV1 and FEV1/FVC revealing airway obstruction consistent with COPD.

An official statement of the American Thoracic Society (ATS) concluded that an increased risk of chronic cough, lower FEV1 (forced expiratory volume in one second after full inspiration), and a lower FEV1/FVC (forced vital capacity) ratio was related to such occupational exposures [14,15]. The population attributable risk (PAR) of COPD from occupational exposure is estimated at 15-20% (16). But as these estimates are proportions, they depend on how causes other than vapour, dusts, gases, and fumes contribute to the development of COPD. The overall estimate of PAR of COPD due to occupational exposure may be misleading and a more quantitative approach seems preferable and is the objective of this meta-analysis.
Exposure to mineral dusts [16] especially in underground mining, such as gold- [17,18], coal- [19,20], and uranium mining [21] has been shown to contribute to the development of COPD, but not much is known about the impact of poorly soluble low-toxicity particles also referred to as biopersistent granular (bg) dust. We therefore conducted a systematic review and meta-analysis to quantitatively evaluate the association between occupational exposure to bg dust at the workplace and the development of COPD.

Materials and Methods
Following the PICOS criteria [22,23] Participants, Intervention, Comparison, Outcome, Study Design were defined in advance (see Checklist S1 in Appendix S1). We searched for epidemiological studies (cohort, case-control, and cross-sectional) of people exposed to bg dust at the workplace with measurements of exposure levels and spirometric measurements of lung function. As bg dust we considered: Portland cement, carbon black, soot, rubber, talcum, and occupational exposure during metal processing and mining (other than gold-, uranium-and coal mining). Not included as bg dust were environmental tobacco smoke and traffic related dust. We searched for studies in English and German between 1970 and 2010 in PubMed applying Medline (Medical Literature Analysis and Retrieval System Online) and in Embase (Excerpta Medica Database). The following MeSH-Terms were used: "occupational exposure" OR "air pollutants, occupational" AND "pulmonary disease, chronic obstructive" supplemented by the text fields "respiratory function tests" OR "respiratory function" OR "lung function" OR "pulmonary function". Two investigators in our team (IB, KH) independently reviewed articles and extracted the following data in duplicate: first author and year of publication; study region and industry; study type; time and duration of the study or duration of the follow-up in cohort studies; number of exposed/unexposed subjects or cases and controls; sex-and age distribution of the study population, response rate; exposure assessment (interview, Job-Exposure-Matrix (JEM), type of dust measurement and average exposure to inhalable or respirable dust) outcome assessment (symptoms/physician diagnosis, spirometry and applied procedure for lung function measurements). Based on the abstracted protocol the validity of the study was independently assed and decision made jointly by the two reviewers (IB, KH); any disagreements were resolved by a third reviewer (DN). Studies were included in the systematic review and meta-analysis, if they fulfilled the following validity criteria: (1) transparent procedure of selection of study participants, no indication of selection bias; (2) response rate > 70% and < 100%; no response rate, or a response rate of 100%, which was considered as probably a post-hoc definition of the study population was not acceptable; (3) internal comparison with no/low exposed controls from the same company, or controls from a another company without exposure; (4) individual present or cumulative exposure preferably based on dust measurements (JEM was considered acceptable; company or questionnaire information was accepted only, if duration of exposure was also available); (5) COPD diagnosis according to obstructive signs in spirometry or physician-diagnosed (questionnaire information of symptoms was not considered as sufficient); (6) standardized pulmonary function test according to ATS/ERS criteria valid at the time of the study.

Statistical Analysis
Assuming that the true effect of exposure to bg dust at the workplace differed from study to study, we conducted a random effects meta-analysis [24], which allowed us to provide an estimate of the average exposure effect. Between-study heterogeneity was assessed by using I 2 statistics, with I 2 >40% representing moderate to considerable heterogeneity. Publication bias was investigated by visual inspection of funnel plots. The results of the meta-analysis in regard to specific lung function parameters of the cross-sectional and longitudinal analysis are presented in forest plots. The influence of individual studies was assessed by dropping the respective study before pooling study-specific estimates.
The most important sign of airway obstruction -a reduced FEV1 and FEV1/FVC -was measured in most studies and used for the meta-analysis of cross-sectional analyses taking the difference of the mean FEV1 and FEV1/FVC among exposed study participants versus not/low exposed participants and also dependent on cumulative exposure to dust. FEV1 was measured in different units either in liters or in % predicted. FEV1% predicted is defined as FEV1 of the patient divided by the average FEV1 in the population for any person of similar age, height and sex. To combine both units [l] and [%] of FEV1 we calculated the standardized mean difference, which is the difference of the mean FEV1 of exposed and low/not exposed study participants divided by the common standard deviation. This measure is dimensionless. Different studies applied different methods to obtain the ratio FEV1/FVC. It was calculated as a ratio in liter, as a ratio in % predicted, and as a ratio of two predicted values. Only the standardized mean difference of the ratio was used as a common estimate of FEV1/FVC for meta-analysis. Studies either adjusted or stratified for smoking status. In the latter case the results for smokers and nonsmokers were integrated separately into the meta-analysis. For some studies it was possible to perform a meta-analysis of the risk of obstructive airway disease by FEV1 and/or FEV1/FVC per 1mg/m 3 of bg dust. And for some longitudinal studies the annual decline of FEV1 among exposed and unexposed study participants could be compared and integrated in the meta-analysis.

Results
2012 publications were identified in PubMed and as well as 3604 publications in Embase. Without duplicates 3792 publications were potentially eligible. Title and abstract were screened by two investigators (IB, DN). Two German publications were added manually, one was a recent publication [21] outside the defined time frame and one [25] was a large longitudinal investigation among construction workers, which was published in a journal not listed in Embase. For details of the selection process see Figure 1.
After excluding studies that did not fulfill the inclusion criteria 257 publications were reviewed with data abstraction, 55 fulfilled the validity criteria and were accepted for review and meta-analysis. 27 publications  thereof investigated the effects of occupational exposure to bg dust (see Appendix S1, Table 1). Not all of these 27 studies could actually be included in the meta-analysis, some because of choosing rarely used endpoints [40,41,[50][51][52], some because of not providing a standard error [28,37], for more details see see Appendix S1, Table 1.
Short summary of the review: Early studies from the US (26) and Jordan (27) showed no detrimental effect of dust exposure on lung function in cement workers. But, later studies with a higher dust exposure of cement workers from Tanzania (44) and Iran (45) showed a strong association with impaired lung function. Especially, a new large and prospectively designed study of the European Cement Association (47) with 4265 exposed participants was very informative. All studies related to carbon black/soot/rubber/talcum (33)(34)(35)(36)(37)43,46,52) showed adverse effects of dust exposure on lung function with stronger effects for early studies with high dust exposure compared to late studies with low exposure. The evidence from studies with metal workers in the aluminum, iron and steel industry was inconsistent (30,31,(38)(39)(40)(41)49,51). Negative or only weakly   (28) were only low exposed and showed no lung function abnormalities, there was a high dust exposure in manganese mining underground in Iran (29). Here, spirometry showed a significant decrease of FEV1 and FVC of 1·mg·m -³ exposed workers compared to non-exposed workers with a stronger effect in smokers (29). A longitudinal study (42) performed in potash mining underground in Germany showed also a significant decrease in lung function over time. In the meta-analysis of cross-sectional study results we combined studies with endpoints measured in the same units: FEV1 measured in liter (see Figure 2), and FEV1 measured in % predicted (see Figure 3). The mean FEV1 of workers exposed to bg dust was 160 ml (95% CI: 40-270 ml) less than compared to workers with no/low exposure (see Figure 2). Comparing the mean FEV1 not in absolute measures [liter], but in % predicted, it was 5.7% (95% CI: 2.71-8.62%) lower for workers exposed to bg dust (see Figure 3). There was also a decrease of FEV1, taking the standardized mean difference between exposed and no/low exposed workers into the metaanalysis (results not shown). Different studies applied different methods to obtain the ratio FEV1/FVC. Only the standardized mean difference of the ratio could be used as a common estimate of FEV1/FVC for meta-analysis. The mean difference of the ratio FEV1/FVC between study participants exposed to bg dust at the work place and low/no exposed participants was significantly decreased -0.25 (95% CI:-0.09 to -0.41)(see Figure 4).  In the meta-analysis of longitudinal study results two studies [30,53] showed a mean annual decline of FEV1 of 6.3 ml higher for bg dust exposed participants compared to low/no exposed participants (results not shown). Using studies [36,[42][43][44] with a cumulative measure for the decline of lung function related to bg dust exposure [mg·m -3 ·years], the metaanalysis showed a decline of FEV1 of 1.6 ml per 1 mg·m -3 ·years (meta-analysis 1.58 ml (95% CI: 1.24-1.93ml)) (see Figure 5). And finally, applying the GOLD criteria two studies reported an increased odds ratio for COPD (FEV1/FVC < 70%) of 1.06 [47] and 1.07 [49] related to the increase of 1 mg·m -3 bg dust. The visual inspection of funnel plots gave no indication of publication bias, as large studies tended to be near the average and there were also small studies that reported null findings (see Figures S1-4 in Appendix S1).

Discussion
The meta-analysis revealed a strong heterogeneity between the studies which had to be expected considering the variant exposure conditions at the workplace in different countries from Europe and abroad over such a long time span. The results from the analysis of highly exposed workers indicated a stronger effect than in all workers combined [44,46]. Nevertheless, dropping the respective studies before pooling study-specific estimates had only a minor impact on the results of the meta-analysis and no impact on the statistical significance.
Inhalation of mineral dust such as quartz and asbestos fibers will induce fibrotic changes of the lung parenchyma accompanied by restrictive spirometric changes, such as a reduced FVC. No such findings were reported for biopersistent granular dust. The meta-analysis of cross-sectional studies showed an association of bg dust only with obstructive symptoms in the spirometry. The mean FEV1 of workers exposed to bg dust was 160 ml lower or 5.7% less than predicted compared to workers with no/low exposure. Whatever measure for airway obstruction was used the reduction of FEV1 or FEV1/FVC was always statistically significant. However, this is probably an underestimate of the true effect of bg dust exposure, as subjects with impaired lung function are more likely to quit their jobs and will therefore not be available as study participants [54]. This selection bias will be even stronger, when investigating an actual obstructive limitation, such as FEV1/FVC < 70% according to the GOLD criteria. Workers with such an impairment plus respiratory symptoms will probably not stay in the workforce. Nevertheless, the risk of an obstructive airway diseasedefined as FEV1/FVC < 70% -increased by 7% per 1 mg· m -3 bg dust [47,49].
If the inhalation of bg dust causes COPD, the exposure should be associated with an accelerated decline in lung function, which cannot be detected in a cross-sectional study design [55]. A longitudinal design including repeated spirometries in each person during a period of several years is needed. Two studies [30,53] showed a mean annual decline of FEV1 of 6.3 ml higher in bg dust exposed participants compared to low/no exposed participants. The observed effect (adjusted for age and smoking) was quite similar to the 7-8 ml reported before [56,57] and appears to be rather small compared to the normal age-related reduction of FEV1 (15-25 ml/year) and the decrease due to smoking (60-80 ml/year) [58]. A few studies provided data for the decline of FEV1 related to a cumulative dust concentration at the workplace [35,[42][43][44].
As the loss of FEV1 per year is typically small, it tends to be hidden by measurement variability and will become obvious only in longer follow-up periods. Whereas Wang et al. [59] consider a decrease of FEV1 > 8% or 330ml per year at the workplace as probably pathological, other authors (Hnizdo et between study participants exposed to bg dust at the workplace and no/low exposed participants. doi: 10.1371/journal.pone.0080977.g002 Figure 3. Mean difference (MD) of FEV1 in % predicted between study participants exposed to bg dust at the workplace and no/low exposed participants. Aiming at a quantitative assessment of the association between occupational exposures to bg dust at the workplace and the development of obstructive symptoms in spirometry, the requirements for a study to be included in the metaanalysis were very specific and led to a remarkable drop between studies identified in the systematic review and those finally included in the meta-analysis. From this follows that the studies included in the meta-analysis cannot claim to be representative of all studies on the subject, but only for those with quantitative data on bg dust exposure at the workplace and lung function measurements. For a more general overview refer to [60][61][62][63][64].
At present, COPD as a compensable occupational disease is included in two international lists of occupational diseases, one proposed by the International Labour Organization (ILO) [65], Figure 4. Standardized mean difference of the ratio FEV1/FVC between study participants exposed to bg dust at the work place and low/no exposed participants.  and the other established by the European Commission [66]. Both are only recommendatory in character; most EU-member states have their own lists, which are comparable just in some parts [67]. In Germany, COPD or emphysema due to underground bituminous coal mining with a cumulative exposure to respirable dust exceeding 100 (mg/m 3 ) x years is already part of the list of recognized occupational diseases entitled to compensation. Other jobs with comparable exposure levels to mineral dust or bg dust are presently not covered by the definition.
In summary, this meta-analysis shows a consistent decline of FEV1 of about 1.6 ml in regard to a cumulative bg dust concentration at the workplace of 1mg·m -3 ·years . Occupational exposure to bg dust was associated with a statistically significantly decreased FEV1 and FEV1/FVC revealing an airway obstruction consistent with COPD. The observed limitation of pulmonary function in workers exposed to bg dust probably underestimated the true effect, since both crosssectional and longitudinal studies in the workforce are often limited to a ''survivor'' population because of the inability to monitor workers who leave their jobs.

Supporting Information
Appendix S1. Supporting files. Figure S1, Funnelplot for Figure 2 Mean difference (MD) of FEV1 [liter] between study participants exposed to bg dust at the workplace and no/low exposed participants. Figure S2, Funnelplot for Figure 3 Mean difference (MD) of FEV1 in % predicted between study participants exposed to bg dust at the workplace and no/low exposed participants. Figure S3, Funnelplot for Figure 4 Standardized mean difference of the ratio FEV1/FVC between study participants exposed to bg dust at the work place and low/no exposed participants. Figure S4, Funnelplot for Figure 5 Decrease of FEV1 (ml) in relation to the cumulative exposure to bg dust at the workplace (1 mg·m -3 ·years). Checklist S1, PRISMA 2009 Checklist. (DOC)