WHAT IS ALREADY KNOWN ON THIS TOPICHOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICYIntroduction

Organic dust consists of particles originating from microbes, plants and animals, including bacteria, fungi and pollen.1 Organic dust exposure can be encountered in many industries and occupations, including agriculture, woodworking and textile processing.2 3 Considerable exposure to organic dust originating from birds and other sources may also occur outside work.4 Endotoxin, which is part of the outer membrane of Gram-negative bacteria, is a well-studied organic dust component with strong inflammatory effects5 and is typically present at high levels in agriculture.2 6 Different components of organic dust may coexist, such as in wood dust, where both moulds and endotoxins may be found, primarily in fresh wood.3

Organic dust exposure has been linked with different types of interstitial lung disease (ILD). ILDs are a group of inflammatory and/or fibrotic lung diseases, often classified according to the presence or absence of a known cause,7 even though these groups share sometimes indistinguishable clinical and radiological features.8 9 In Denmark, the incidence of ILDs is approximately 20 per 100 000 per year.10 11 Hypersensitivity pneumonitis (HP) is a specific type of ILD with an annual incidence in Denmark of 1.16 per 100 000. It is associated with a higher mortality among younger people.12 Traditionally, the presence of a cause has been integrated into the diagnostic criteria of HP, and even though the new definition of HP introduced in 2020 no longer requires an identified cause,9 the previous criteria still challenge the examination of causal associations prior to this change. The onset of HP may be acute or insidious, with an associated exposure more often identified in cases with acute onset.9 Besides HP, associations between different organic dust constituents and ILDs without a known cause, such as idiopathic interstitial pneumonias and sarcoidosis, have also been suggested,13–18 though findings have been mixed.

For HP, the association with organic dust exposure is primarily based on case reports and series, as only a few epidemiological studies are available.4 19 This leaves a need for more epidemiological studies to verify the association. Epidemiological studies on idiopathic interstitial pneumonias and sarcoidosis do not include quantitative exposure information,13–18 leading to a lack of knowledge about exposure-response relations, which are central to the evaluation of causal associations.

The aim of this study was to examine the exposure-response relations between exposure to organic dust, endotoxin and wood dust and HP and other ILDs in a cohort study of the Danish general working population 1994–2015.

MethodsStudy population

The study population included past or present residents of Denmark with at least 1 year of gainful employment registered in the Danish Occupational Cohort (DOC*X) since 1976.20 Only residents born in 1956 or later were included. DOC*X contains annual information on occupation, which is harmonised and coded according to the 1988 International Standard Classification of Occupations (ISCO-88).21

From the study population, we also derived a subpopulation restricted to blue-collar workers (ISCO-88 major categories 6–9) at baseline for supplementary analyses.

Hypersensitivity pneumonitis and other ILDs

Within the study population, we identified incident cases as the first diagnosis of HP or other ILDs registered in the Danish National Patient Register,22 which covers all hospital contacts in Denmark. As HP could only be defined with certainty using the 10th version of the International Classification of Diseases (ICD-10) introduced in Denmark in 1994, we only identified cases from 1994 onwards.

The diagnosis of HP rests on typical radiological findings, bronchoalveolar lavage lymphocytosis, identification of relevant antigenic exposures, including measurement of precipitating antibodies, and, in some cases, a lung biopsy. HP was defined by ICD-10 code J67, which is used in Denmark for all cases of HP, independently of the underlying exposure.

We defined other ILDs by ICD-10 codes D86.0, D86.2, J60-65, J68.4A, J70.1, J70.3, J70.4, J82, J84, J99.0, J99.1, M05.1, M34.8C and M35.0B. In a supplementary analysis, we studied five specific types of ILD within the other ILDs group: pneumoconiosis, drug-related ILDs, connective tissue disease-related ILDs, idiopathic interstitial pneumonias and pulmonary sarcoidosis. Descriptions of the ICD-10 codes included are available in online supplemental table S1.

All prevalent cases of ILD registered in the National Patient Register 1977–1993 were excluded. They were defined by ICD-8 codes 515.0, 515.1, 515.2, 515.9, 516.0, 516.1, 516.2, 519 and 519.20.

For a sensitivity analysis of other ILDs, we excluded all cases with a subsequent diagnosis of HP in the same or following year as the diagnosis of other ILDs.

Exposure assessment

The exposure assessment was based on the work histories (coded according to ISCO-88) provided by DOC*X. If the ISCO-88 code was missing for a worker for a given year, we assigned the latest valid ISCO-88 code up to 5 years back, reducing work history years with missing ISCO-88 codes from 13% to 5%. Years missing an ISCO-88 code after this procedure were assigned an exposure intensity of 0.

Organic dust exposure, covering all exposure to particulate matter of organic origin, was assigned using the expert-based ALOHA+ job exposure matrix (JEM),23 which assigns three levels of exposure (0=none, 1=low, 2=high). The ALOHA+ JEM exposure intensity estimates were squared to estimate cumulative exposure.

Endotoxin exposure was estimated using a quantitative JEM24 based on 3384 personal endotoxin measurements from Western Europe and Canada. The JEM provides estimates of endotoxin exposure expressed as endotoxin units (EU) 1990–2009 with an annual trend of −2.9%. For years before and after this period, which are not covered by measurements, we assumed no trend in the JEM estimates.

Wood dust exposure was estimated using a quantitative JEM25 providing exposure estimates in mg/m3 based on 12 653 personal wood dust measurements from Europe. From 1978–2007, there is an annual trend of −7.8%, whereas for the years before and after this period, which were not covered by measurements, we assumed no trend in the JEM estimates.

Information on exposed ISCO-88 codes for all three JEMs can be found in online supplemental tables S2–S4.

For all three exposures, we assigned annual individual exposure intensity which was used to estimate individual (1) cumulative exposure as the sum of the annual exposure intensities, (2) duration of exposure in years and (3) recent exposure accrued during the previous year.

For a supplementary analysis, we grouped workers according to ISCO-88 major groups based on the first registered valid ISCO-88 code for each worker.

Statistical analyses

Exposure was assessed from the first registered employment year after 1976, which is the year of first registrations in DOC*X, and until the end of follow-up. We started follow-up in 1994 for those with a first year of employment before 1994. If the first year of employment was 1994 or later, follow-up started the year following the first year of employment, as we had no information on the month or day of employment. For the same reason, we also lagged all independent variables by 1 year. We followed workers until the year of first diagnosis (HP or other ILDs), death, emigration, disappearance or end of follow-up on 31 December 2015, whichever came first. No person could contribute more than one diagnosis, as the first ILD diagnosis led to the end of the follow-up. All cases of ILD diagnosed from 1977 to 1993 were excluded to avoid the inclusion of prevalent cases. After 1994, ILD diagnosed before a participant’s beginning of follow-up also led to exclusion.

For cumulative exposure and duration of exposure, person-years were first categorised as either exposed or non-exposed, and the exposed group was then divided into tertiles based on the distribution of person-years. For recent exposure, we classified exposure accrued outside the 1-year time window as zero and dichotomised exposure within the window by the median, except for organic dust exposure where the semi-quantitative JEM provided only two possible exposure values above zero within 1 year, and these were used to form the two exposed groups.

We used a discrete-time hazard model with person-years as a unit of analysis, yielding incidence rate ratios (IRR) with 95% CIs.26 Analyses were performed separately for each of the three exposures. The reference was the non-exposed group for the exposure in question. For the analysis based on occupations, ISCO-88 major groups 0 and 1 were used as reference, as these groups are not exposed to organic dust according to the ALOHA+ JEM. Analyses were adjusted for age, sex, calendar year of follow-up, education, JEM estimates of smoking, previously diagnosed connective tissue disease, fibrogenic medications and JEM estimates of cumulative asbestos and crystalline silica exposure. More information on the classification of and rationale for the covariates is available in the supplementary material (online supplemental appendix 1). All variables were treated as time-varying accounting for changes in status and increases in cumulative exposure over time.

We fitted restricted cubic splines with 95% CIs for the cumulative exposure metrics as continuous variables, placing the knots at the 5, 50 and 95 percentiles.27 The x-axis was cut-off at the 99 percentile to focus on the most relevant results.

Pearson pairwise correlation coefficients were calculated for all exposures.

We used Stata V.17 (StataCorp, College Station, Texas, USA) for all analyses.

The study was registered at the repository of the Central Denmark Region.

Results

The study population comprised 2 955 863 workers accumulating 49 228 874 person-years during follow-up. A flowchart of the establishment of the study population is shown in figure 1. We identified 411 incident cases of HP, 6724 cases of other ILDs, adding up to 7135 cases of all ILDs. The corresponding crude incidence rates per 100 000 person-years were 0.8, 13.7 and 14.5, respectively.

Figure 1

Establishment of the study population. ISCO-88, 1988 International Standard Classification of Occupations.

The distribution of person-years by cumulative exposure and worker characteristics showed that higher organic dust exposure was associated with previous exposure to fibrogenic medication; higher cumulative endotoxin exposure with male sex, being a skilled blue-collar worker and vocational or higher secondary education; and higher cumulative wood dust exposure with male sex and being unskilled blue-collar worker (online supplemental table S5).

Correlations between the three main exposures were 0.45 between cumulative organic dust and endotoxin exposure, 0.50 between cumulative organic dust and wood dust exposure and 0.02 between cumulative endotoxin and wood dust exposure. For cumulative silica and asbestos exposure all correlations with the three main exposures were below 0.40.

For organic dust exposure, we observed increasing IRRs of HP with increasing cumulative exposure with a fully adjusted IRR of 1.19 (95% CI 1.12 to 1.27) per 10 unit-years (table 1). Similarly, we observed increasing IRRs with increasing cumulative exposure for other ILDs and all ILDs but with lower risk estimates. The fully adjusted IRRs for other ILDs and all ILDs were 1.04 (95% CI 1.02 to 1.06) and 1.05 (95% CI 1.03 to 1.07) per 10 unit-years, respectively. IRR for the highest exposed tertile compared with the non-exposed group was 1.82 (95% CI 1.39 to 2.38) for HP and 1.18 (95% CI 1.10 to 1.26) for other ILDs. The IRRs also increased with increasing exposure duration for all three outcomes.

Table 1

Incidence rate ratios (IRR) of hypersensitivity pneumonitis (HP), other interstitial lung diseases (ILDs) and all ILDs following exposure to organic dust, endotoxin and wood dust among 2 955 863 workers born in 1956 or later, in Denmark

Increasing risk of HP was also found with increasing cumulative endotoxin exposure, with a fully adjusted IRR of 1.55 (95% CI 1.38 to 1.73) per 5000 EU/m3-years (table 1). We observed similar results for other ILDs and all ILDs with IRRs of 1.09 (95% CI 1.00 to 1.19) and 1.16 (95% CI 1.08 to 1.25) per 5000 EU/m3-years, respectively. For HP, IRR in the highest exposure tertile compared with the non-exposed group was 6.53 (95% CI 4.10 to 10.42), while it was 1.26 (95% CI 1.07 to 1.48) for other ILDs. We also observed increasing IRRs with increasing exposure duration for all three outcomes.

No increasing risks were observed for cumulative wood dust exposure (table 1). The fully adjusted IRR was 0.91 (95% CI 0.76 to 1.08) per 5 mg/m3-years for HP, 1.00 (95% CI 0.95 to 1.04) for other ILDs and 0.99 (95% CI 0.95 to 1.03) for all ILDs. Duration of exposure was also not associated with increased risk.

An expanded version of table 1 is shown in online supplemental table S6.

In spline analyses, risk of HP increased with increasing cumulative organic dust exposure, with the strongest increase up to around 20 unit-years (figure 2). Risk of other ILDs and all ILDs increased up to around 20 unit-years, after which it levelled off. For cumulative endotoxin exposure, increases for all three outcomes were seen up to around 5000 EU/m3-years, after which it reached a plateau. The increase was strongest for HP. For cumulative wood dust exposure, no clear pattern of increasing risk with increasing exposure was seen for any of the outcomes.

Figure 2

Restricted cubic spline fits of incidence rate ratios of hypersensitivity pneumonitis, other ILDs and all ILDs by cumulative organic dust exposure, cumulative endotoxin exposure and cumulative wood dust exposure. Dotted lines mark 95% CIs. EU, endotoxin units; ILDs, interstitial lung disease.

When excluding cases with subsequent HP diagnosis in the same or following year, the sensitivity analysis for other ILDs showed risk estimates similar to the main analysis (online supplemental table S7).

In the population restricted to blue-collar workers, we found similar risk estimates for cumulative organic dust exposure per 10 unit-years, but the IRR for the highest exposure tertile was only increased for HP (online supplemental table S8). The findings for cumulative endotoxin and wood dust exposure were similar to those of the total study population.

Increasing risk of HP with increasing recent exposure was observed for both organic dust (IRR per 10 units 15.58 (95% CI 6.44 to 37.68)) and endotoxin (IRR per 5000 EU/m3 24.30 (95% CI 8.64 to 68.35)) (table 2). Risk of other ILDs and all ILDs also increased with increasing recent exposure to organic dust and endotoxin. The number of cases with recent wood dust exposure was too low for meaningful analyses (results not shown). More details are available in online supplemental table S9.

Table 2

Incidence rate ratios (IRR) of hypersensitivity pneumonitis, other interstitial lung diseases (ILDs) and all ILDs following recent exposure to organic dust and endotoxins among 2 955 863 workers, in Denmark

Risk of HP was increased in ISCO-88 major group 6 ‘Skilled agricultural and fishery workers’ (online supplemental table S10). Less pronounced increases were found in ISCO-88 major groups 7–9 comprising other blue-collar workers. For other ILDs, risks were increased for ISCO-88 major groups 6–8, whereas no increase in risk was observed for white-collar workers in ISCO-88 major groups 2–4.

In analyses of specific types of ILD within the other ILDs group, we observed an increasing risk of pulmonary sarcoidosis with increasing cumulative organic dust exposure and risk also tended to increase with increasing cumulative endotoxin exposure (online supplemental table S11). We observed no clear exposure-response relations for the other types of ILD studied.

Discussion

In the present study, we observed an increasing risk of HP, other ILDs and all ILDs with increasing cumulative exposure to organic dust and endotoxin, but not to wood dust. Risk of all three outcomes also increased with increasing recent organic dust and endotoxin exposure. No increased risk was observed for wood dust exposure.

Our results are in accordance with organic dust being a known cause of HP.15 Furthermore, we found that increasing exposure entailed increasing risk of HP, and while an association between increasing antigen exposure and disease progression has previously been suggested, results have been inconsistent.28 29

HP is associated with farming, and endotoxin is abundant in livestock farming, pigeon coops and poultry houses.2 30 31 Our findings support a causal role of endotoxins in the development of HP. We cannot, however, rule out that endotoxin is a marker of other organic dust constituents that may contribute to the association observed.

An association between woodworking and HP has been suggested and may be caused by fungal and bacterial antigen contamination.32 33 Our findings did not confirm an association between wood dust exposure and HP. The JEM does, however, not distinguish between fresh and dry wood exposure or include information on the level of microorganisms present.

Results of previous studies on idiopathic interstitial pneumonias, which are mainly case–control studies, have been inconsistent with regard to organic dust exposure.13 14 16 34–37 However, studies of organic dust exposure and sarcoidosis have, though few, consistently reported an association.17 18 38 We observed an association between cumulative organic dust and endotoxin exposure and the entire group of other ILDs, attributable primarily to pulmonary sarcoidosis. Our findings for other ILDs are confirmed by the sensitivity analysis of other ILDs excluding cases with a subsequent HP diagnosis, which rules out that the findings for other ILDs can be attributed to cases of HP not diagnosed initially but only after further examination.

The observed association between cumulative organic dust and endotoxin exposure and other ILDs and all ILDs signal that organic dust exposure may not only be a risk factor for HP, but also for other types of ILD. This indicates that the importance of organic dust and other work-related exposures for the development of ILD will be underestimated, if only ILDs traditionally attributed to work, such as HP, are taken into account. The observed associations between all ILDs and the blue-collar trades suggest that these occupations in particular need to be targeted to decrease the future burden of ILD.

Integration of a cause in the diagnostic criteria of a disease, as has traditionally been the case with HP, challenges the investigation of the association with this cause, which was what this study aimed for. In time, the new definition of HP introduced in 2020 focusing on objective findings rather than identification of specific exposures will provide better opportunities for studying causal associations for HP. For now, however, the issue remains, and we chose to circumvent it by analysing other ILDs and all ILDs, as done previously in a study of ILD risk in pigeon breeders.4 Our confidence in the associations observed for other ILDs is increased by the similar findings in the main and sensitivity analyses. However, we consider the findings for all ILDs the main findings of our study, as these risk estimates should not be affected by the inclusion or exclusion of exposure in the disease definitions. Another potential way to circumvent disease definitions is studies of associations for disease phenotype, defined for instance by radiological findings, which to our knowledge has not been done for organic dust exposures yet.

Strengths and limitations

This study included the total Danish working population, and we retrieved diagnoses from a national health register with information on all hospital contacts, which means that selection bias is unlikely to have affected our results. We did not have access to individual clinical information, but in Denmark all suspected cases of ILD are referred to the four Danish national ILD centres. The extensive diagnostic programme carried out consists of a detailed medical history including occupational and avocational exposures, symptoms and findings related to autoimmune rheumatic diseases and other underlying diseases, as well as a clinical examination, including pulmonary function tests, a 6-minute walk test and blood tests. All patients undergo a high-resolution CT scan and are discussed at a multidisciplinary team conference. If necessary, lung biopsies are performed.

Because of clinical and radiological similarities between especially HP dominated by fibrosis and other ILDs such as idiopathic pulmonary fibrosis, diagnostic misclassification of HP as other ILDs may have contributed to the increased risk of other ILDs. However, this will not have affected the results for all ILDs combined, and we consider this way of addressing both diagnostic misclassification and the problem of integration of causes in disease definitions a major strength of this study.

Another strength is that the exposure assessment did not rely on self-reported exposure information that may be subject to recall bias, and the two quantitative JEMs used were based on large numbers of personal measurements. While the use of JEMs enables the study of an entire working population, it may lead to non-differential misclassification of exposure because the JEMs do not capture within-occupation exposure variation. However, this will mainly lead to Berkson-type error, resulting in increased uncertainty of risk estimates but an unbiased exposure-response association.39

We adjusted the risk estimates for a number of known causes of pulmonary fibrosis,7 but adding more confounders than age, sex and calendar year of follow-up only slightly attenuated the exposure-response relations. For HP, smoking is generally considered a protective factor,40 whereas it may be a risk factor for some types of other ILDs.41 We adjusted for smoking using a lifestyle JEM, which has previously predicted acute myocardial infarction in this population as expected.42 Individual smoking information would be preferable, but we do not expect the use of a JEM to be problematic when studying diseases where smoking is not a strong risk factor. We were not able to adjust for non-occupational sources of organic dust exposures associated with HP such as domestic birds or mould-contamination at home, and this may have resulted in residual confounding. Cumulative endotoxin and wood dust exposure were both moderately correlated with cumulative organic dust exposure, but we chose not to conduct mutual adjustment to avoid over-adjustment.

Conclusion

In this study, we observed strong exposure-response relations between organic dust and endotoxin exposure and HP. Exposure-response relations were also observed for other ILDs and all ILDs, but with lower risk estimates. No associations were observed for wood dust. We consider the estimates for all ILDs most representative of the true associations, as organic dust and other antigenic exposures have traditionally been integrated into the definition of HP, thereby predefining an association. Our findings indicate that organic dust exposure may contribute to the development of any ILD, underscoring the importance of preventive measures against organic dust exposure to decrease the future burden of ILD.

Data availability statement

Data may be obtained from a third party and are not publicly available. Data from Statistics Denmark were used for this study under license and are not publicly available, but may be accessed by permission from Statistics Denmark.

Ethics statementsPatient consent for publication

Not applicable.

Ethics approval

Register studies in Denmark without biological materials do not require approval from an Ethics Committee. Participants’ informed consent is not required for register studies in Denmark.

Acknowledgments

Professor Roel Vermeulen, Utrecht University, is acknowledged for his contribution to the elaboration of the ALOHA+ job exposure matrix.