The Impact of Occupational Exposures on the Risk of Idiopathic Pulmonary Fibrosis: A Systematic Review and Meta-Analysis

Abstract

Rationale

Idiopathic pulmonary fibrosis (IPF) is a progressive fibrotic pulmonary disorder of unknown etiology that is characterized by a usual interstitial pneumonia pattern. Previous meta-analyses have reported associations between occupational exposures and IPF, but higher-quality studies have been published in recent years, doubling the number of studied patients.

Objectives

To provide a contemporary and comprehensive assessment of the relationship between occupational exposures and IPF.

Methods

We searched PubMed, Embase, and Web of Science through July 2023 to identify all publications on occupational exposure and IPF. We conducted a meta-analysis of the occupational burden, odds ratio (OR), and population attributable fraction (PAF) of exposures. Five exposure categories were analyzed: vapors, gas, dust, and fumes (VGDF); metal dust; wood dust; silica dust; and agricultural dust. A comprehensive bias assessment was performed. The study protocol was registered in the International Prospective Register of Systematic Reviews (identifier CRD42021267808).

Results

Our search identified 23,942 publications. Sixteen publications contained relative risks needed to calculate pooled ORs and PAFs, and 12 additional publications reported an occupational burden within a case series. The proportion of cases with occupational exposures to VGDF was 44% (95% confidence interval [CI], 36–53%), with a range of 8–17% within more specific exposure categories. The pooled OR was increased for VGDF at 1.8 (95% CI, 1.3–2.4), with a pooled PAF of 21% (95% CI, 15–28%). ORs and PAFs, respectively, were found to be 1.6 and 7% for metal dust, 1.6 and 3% for wood dust, 1.8 and 14% for agricultural dust, and 1.8 and 4% for silica dust. The pooled ORs and PAFs within specific exposure categories ranged from 1.6 to 1.8 and from 4% to 14%, respectively. We identified some publication bias, but it was not sufficient to diminish the association between occupational exposures and IPF based on sensitivity analysis and bias assessment.

Conclusions

Our findings indicate that 21% of IPF cases (or approximately one in five) could be prevented by removal of occupational exposure (alongside a pooled OR of 1.8). Additionally, 44% of patients with IPF report occupational exposure to VGDF. This meta-analysis suggests that a considerable number of cases of IPF are attributable to inhaled occupational exposures and warrant increased consideration in the clinical care of patients and future prevention efforts.

Idiopathic pulmonary fibrosis (IPF) is a chronic, progressive fibrotic disorder typically affecting adults older than age 50 years and is associated with a poor prognosis. Its incidence ranges from three to nine cases per 100,000 annually in North America and Europe (1). The hallmark of IPF is a radiological or pathological usual interstitial pneumonia (UIP) pattern with known causes of UIP excluded (e.g., collagen vascular disease, drug-induced pneumonitis, radiation fibrosis, or occupational exposures) through a focused history or additional diagnostic testing (2, 3).

Although IPF is considered a disease of unknown etiology, profibrotic risk factors have been described, including genetic factors (4), cigarette smoking (5), viral infections (6), occupational exposures (7), and gastroesophageal reflux disease (8). In particular, the association with occupational exposures warrants further investigation, given the number of studies informing potential relationships (2). Additionally, occupational exposures remain a highly preventable and modifiable inhalational hazard if associated with IPF.

Retrospective case-control and cohort studies have found associations between specific occupational exposures and the risk of developing IPF. A previous systematic review of IPF and occupational exposures published in 2017 identified 11 studies that reported associations between occupational exposures and IPF (7, 9–12). This review included 1,229 cases of IPF, with only limited analysis of bias, confounding, susceptibility, and study quality (7). This present study was undertaken in light of new studies with large numbers of patients to evaluate whether their findings have continued to demonstrate an association between occupational exposures and IPF.

The primary objective of the present study was to provide a contemporaneous and comprehensive assessment of relationships between occupational exposures and IPF to qualitatively characterize associations between occupational exposures and the risk of diagnosis. This meta-analysis intends to improve upon previous work with an additional 5 years of research, during which a large number of studies have been published, using updated diagnostic definitions relevant to contemporary practice while integrating an analysis with intensive evaluation for bias and quality assessment.

Methods

Data Sources and Searches

We performed a systematic review and meta-analysis according to a predetermined protocol per the Preferred Reporting Items for Systematic Reviews and Meta-Analysis guidelines (13, 14). We registered the study protocol to the International Prospective Register of Systematic Reviews on September 8, 2021 (identifier CRD42021267808).

The screening process and results are shown in Figure 1. We conducted an initial systematic electronic literature review using PubMed, Embase, and Web of Science on July 15, 2023. Searches were designed with the goal of identifying all publications that reported on occupational exposure in patients diagnosed with IPF. Search criteria are provided in the data supplement. We examined bibliographies to search for additional studies. No language or regional restrictions were applied. We excluded duplicates. We contacted authors for raw data or preprint papers for studies that had been published only as abstracts. If adequate data for analysis were available, abstracts were included.

Figure 1.

Article selection process for meta-analysis using Preferred Reporting Items for Systematic Reviews and Meta-Analysis guidelines. IPF = idiopathic pulmonary fibrosis; OR = odds ratio.

Study Selection

Two reviewers (S.A.G., B.M., and/or J.C.F.) independently screened all citations for eligibility using a sequential review of titles, abstracts, and full publications. At each stage, lists were compared, and disagreements were resolved through consensus (S.A.G., B.M., and/or J.C.F.). Per protocol, studies were included if the study population comprised adults (⩾18 years old) diagnosed with IPF (or equivalent criteria) based on consistent clinical, radiographic, or pathologic findings. The diagnosis was established according to the authors of the parent studies. Case-control, cohort, case-series, and cross-sectional studies were included, but studies with study populations of fewer than 10 were excluded. We included studies reporting prevalence or risk (odds ratio [OR], relative risk) of a diagnosis of IPF based on specific occupations or occupational exposures. In the event of multiple publications with overlapping recruitment periods at a single center, we included the study with the more detailed description of occupational exposure to prevent patients from being counted twice. Conference abstracts were included for pooled population attributable fraction (PAF) and OR calculation if the parent study authors provided raw data and methodological information for OR and PAF calculations and quality assessment.

Data Extraction and Quality Assessment

A standardized, prepiloted form was used to extract data from the included studies to assess study quality and evidence synthesis. Two review authors (S.A.G., B.M., and/or J.C.F.) extracted data independently, and discrepancies were identified and resolved through discussion (S.A.G., B.M., and/or J.C.F). Two authors (S.A.G., B.M., and/or J.C.F.) extracted data on study design, participant inclusion and exclusion criteria, sex, age, method of diagnosis, types of exposures, and effect measures (ORs, risk ratios, and standard errors). Aggregate participant data was used. For articles not in English, Google Translate was used for data extraction, which was verified by a native speaker (Google Translate version 6.34.0, 2022; Google LLC).

Inhalational exposure was defined by job or exposure history. In publications with sufficient data, analysis of occupational exposures in five exposure categories was performed: vapors, gas, dust, and fumes (VGDF); metal dust; wood dust; silica dust; and agricultural dust. Agricultural dust exposure was defined as exposure to farming, livestock, or organic dust. Silica dust exposure also included exposure to stonework or grinding if a discrete silica dust exposure category was not available. VGDF represents all other exposure categories combined, and, in selected studies, additional exposures.

Prespecified outcome measures included the proportion exposed and OR of diagnosis, with consideration of whether outcomes were adjusted for smoking history. The adjusted OR was extracted if unadjusted, and adjusted ORs are presented after adjustment for age, sex, or smoking. The methodological quality of studies, included in OR and PAF meta-analysis, was assessed independently by two coauthors (S.A.G. and C.J.R.) using the Risk of Bias in Studies estimating Prevalence of Exposure to Occupational risk factors (RoB-SPEO) tool (in eight categories) (15). Eight domains of bias were considered: selection, blinding, exposure misclassification, incomplete exposure data, selective reporting of exposures, conflict of interests, differences in numerator and denominator, and other biases. Risk of bias was rated in each domain: low, probably low, probably high, high, and no information, which were assigned scores of 1, 2, 3, 4, or 5, respectively. A cumulative score was calculated, with a lower score indicating a higher-quality study.

Statistical Methods and Meta-Analysis

Descriptive statistics were used to quantify the occupational burden of IPF (i.e., proportion of those with the disease who had occupational exposure). We pooled published or derived occupational burdens to obtain summary estimates using the metaproportion command in Stata software (version 14.2; StataCorp LP). For the occupational burden, we performed a subgroup analysis evaluating only full-text articles.

We estimated the occupationally related PAF reported by or derived from case-control studies. The PAF is the proportional reduction in population disease or mortality that would occur if exposure to a risk factor were eliminated. We calculated PAF using the OR and proportion of cases exposed as pc(OR − 1) / OR, where pc is the proportion of cases exposed (7, 16). We calculated pooled OR and pooled PAF for occupational exposures using a random-effects model with the metan command in Stata 14.2 (release 14; 2015). We selected a random-effects rather than a fixed-effects model because of between-study differences in design and populations. The pooled PAF relied on the ratio of attributable cases to all cases underlying each risk estimate. We also estimated statistical heterogeneity using I² statistics.

To assess for publication bias, we analyzed a funnel plot for asymmetry. Additionally, we conducted a series of sensitivity analyses of studies to evaluate the impacts of various factors. Subgroup analysis was performed for pooled OR evaluating only studies adjusted for smoking and studies with larger sample sizes (case number >100). To analyze high-quality studies, we excluded those ranking in the bottom tertile of the overall quality assessment, analyzing studies with RoB-SPEO scores no greater than 16. An analysis of studies with “healthier” controls was performed, excluding studies that recruited controls from respiratory clinics or inpatient wards, because differing comparative groups could increase the heterogeneity of studies. Because some studies relied on chest X-ray rather than computed tomography (CT) of the chest for diagnosis, a subgroup analysis was performed for studies that required CT scans, consistent with contemporaneous guidelines (2, 3, 17). Finally, because of concerns for funnel plot asymmetry, Egger’s and Begg’s tests were also performed (18, 19).

Results

Search Results

Our initial search identified 23,942 citations, and 21,475 abstracts were reviewed after the removal of duplicates (Figure 1). Of these, 99 studies underwent full-text review, and 71 were excluded from data extraction. One study was excluded because of the omission of the proportion of exposed cases to calculate PAF (20), and we were able to include one study by contacting the authors for this calculation (21). Two conference abstracts were included in the calculations for pooled OR and PAF after we were able to contact the authors for raw data and methodological information for calculation and quality assessment (22, 23). One abstract was excluded because we were unable to gather this information and the authors did not plan to publish the study in a peer-reviewed journal (24). One study was found via published abstract in 2021 (25), with a peer-reviewed manuscript published in September 2023, which we were able to include in the analysis (26).

Study Characteristics

We identified 28 papers that met the inclusion criteria. Of these, 16 papers contained a risk estimate to calculate a pooled occupational burden, OR, and PAF, and 12 additional papers reported a proportion of IPF cases with occupational exposure (i.e., occupational burden) within a case-series/cross-section study.

Table 1 describes 16 case-control studies that contained a risk estimate to calculate a pooled OR and PAF (Table 1). One study by Awadalla and coworkers included separate risk estimates by sex. Therefore, two separate risk estimates are included in the calculations (27). The case-control studies included 2,631 patients with IPF (Table 1). Studies originated from nine countries and were published between 1990 and 2023. Early studies included patients in the IPF case population without a diagnosis by CT of the chest, relying instead on chest X-rays (28–30). In one study, case and control definitions were unavailable at the time of this analysis (23). In creating control groups, five studies used the general population (21, 31–34), one used an employment registry (35), and the remainder used clinical practices at participating centers. Of these, four recruited control subjects from respiratory clinics or inpatient pulmonary wards consisting of patients without IPF (26, 27, 36, 37). Nine of 15 studies relied on exposure assessment based solely on a questionnaire or patient interview without a job-exposure matrix (27–30, 32, 33, 36–38). Two studies did not adjust the risk estimates for smoking (26, 30).

Table 1.

Overview of occupational idiopathic pulmonary fibrosis case-control studies

Study, Year, Location	Cases	Case Definition	Control Definition	Exposure Measure	Occupational Burden
					VGDF	Metal	Wood	Agri	Silica
Scott et al., 1990, UK (28)	40	Clinical assessment; CXR; PFTs	Matched on age and sex of cases using general practice register, ratio 1:4	Questionnaire	68%	15%	15%	13%	15%
Hubbard et al., 1996, UK (29)	218	Clinical assessment; CXR; CT; PFTs; biopsy	Matched on age and sex of cases using general practice register, ratio 1:4	Telephone interview; questionnaire	25%	14%	NA	NA	NA
Mullen et al., 1998, UK (30)	15	CXR or biopsy	Matched on age and sex of cases using orthopedic practice list, ratio 1:6	Telephone Interview	40%	NA	13%	NA	NA
Baumgartner et al., 2000, US (31)	248	Clinical assessment; CT; BAL; biopsy	Matched on age, sex, and geographic region of cases using random digit dialing, ratio 1:2	Telephone Interview; JEM	NA	10%	5%	25%	3%
Hubbard et al., 2000, UK (35)	22	Death certificate	Random sample of deceased Rolls Royce employees, ratio 1:10	JEM	NA	59%	NA	NA	NA
Miyake et al., 2005, Japan (36)	102	Clinical assessment; CT; biopsy	Respiratory department inpatients at 21 participating hospitals, unmatched, 2:1 ratio	Questionnaire	32%	12%	5%	NA	11%
Gustafson et al., 2007, Sweden (32)	140	Pulmonary fibrosis of unknown cause, requiring LTOT, identified from LTOT register	Random age-matched population sample, 1:5 ratio	Questionnaire	61%	5%	3%	NA	10%
García-Sancho et al., 2011, Mexico (33)	100	Clinical assessment; CT; biopsy	Matched on age, sex, and geographic region using neighborhood sampling ratio 1:1–1:3	Questionnaire	77%	NA	NA	NA	NA
Awadalla et al., 2012, Egypt (27)	106 women, 95 men	Clinical assessment; CT; PFTs	Matched on age, sex, residence and smoking, non-IPF respiratory inpatients 1:1	Clinical Interview	NA	NA	8%	21%	2%
					NA	18%	16%	21%	13%
Koo et al., 2017, S. Korea (37)	78	Clinical assessment; CT	Matched on age, sex, and area, ratio 1:1, recruited from respiratory inpatients and outpatients	Telephone Interview	55%	27%	8%	NA	27%
Kim et al., 2017, S. Korea (38)	70	Clinical assessment; CT; biopsy	Matched on age and sex from healthy clinic visits; 1:1 ratio	Telephone Interview; JEM	NA	13%	10%	23%	14%
Paolocci et al., 2018, Italy (21)	69	Clinical assessment; CT; biopsy	Matched on area but not age or sex from general population	Telephone Interview; JEM	87%	13%	9%	54%	NA
Reynolds et al., 2019, UK (22)	488	Clinical assessment; CT; biopsy	Matched on age and sex from outpatient clinics, 1:1 ratio, study recruited only men	Telephone Interview; JEM	28%	18%	10%	NA	5%
Abramson et al., 2020, Australia (34)	503	Clinical assessment; CT; biopsy	Matched on age, sex, and state of residence using random digit dialing, ratio 2:1	Telephone Interview; JEM	52%	11%	1%	9%	13%
Park et al., 2020, S. Korea (23)	206	Unknown	Without lung disease	Questionnaire, JEM	25%	2%	5%	11%	10%
Zubairi et al., 2023, Pakistan (25, 26)	131	Clinical assessment, PFTs, HRCT	Matched on age and sex, non-IPF respiratory patients inpatient and outpatient	Questionnaire	19%	NA	NA	NA	NA

Definition of abbreviations: Agri = agricultural dust; BAL = bronchoalveolar lavage; CT = computed tomography (thorax); CXR = chest X-ray; HRCT = high-resolution computed tomography; JEM = job exposure matrix; LTOT = long-term oxygen therapy; NA = not available; PFT = pulmonary function test; VGDF = vapors, gas, dust, or fumes, which represent all the exposure categories shown combined and, in selected studies, additional exposures.

An additional 12 studies met the inclusion criteria and reported proportions of cases with occupational exposure to calculate an occupational burden (Table 2). Of these, two studies were cross-sectional (39, 40) and eight were case series (41–47). A total of 21 studies were used to calculate the occupational burden of VGDF. One article was published in Turkish (44). Only one of these 11 studies used chest X-rays rather than CT scans for the case definition of IPF (39), and one study relied on pathological findings (45). One case-control study was not included in the occupational burden calculation as a result of likely overlap of patients within a larger cross-sectional study (23). Two studies assessed the addition of an occupational medicine provider to a multidisciplinary discussion (48, 49), but one of them did not contain a sufficient sample size to meet the inclusion criteria (49). Finally, one abstract did not provide the case definition (43, 46). Of the studies with known exposure methods, four studies used job-exposure matrices (40, 41, 44, 48) and five used questionnaires or interviews to determine exposure (39, 42, 46, 50).

Table 2.

Overview and occupational burden of idiopathic pulmonary fibrosis cross-sectional and descriptive studies

Study, Year, Location	Cases	Case Definition	Exposure Measure	Full Text	Occupational Burden
					VGDF	Metal	Wood	Agri	Silica
Johnston et al., 1997, UK (39)	588	Clinical assessment; CXR; PFT; biopsy	Physician-administered questionnaire	Yes	47%	NA	NA	19%	NA
Sengül et al., 2009, Turkey (44)	25	Clinical assessment; CT; biopsy	Clinical interview; JEM	Yes	48%	16%	NA	16%	NA
Lee et al., 2015, S. Korea (41)	1,311	Clinical assessment; CT; PFT	Questionnaire; JEM	Yes	13%	NA	NA	18%	NA
Alaa Rashad and Ibrahim, 2015, Egypt (50)	568	Clinical assessment; CT; PFT	Clinical Interview	Yes	NA	NA	NA	12%	NA
Walters et al., 2015, Australia (46)	389	Unknown	Questionnaire	No	50%	NA	NA	7%	12%
Wuyts et al., 2016, Belgium/Luxembourg (47)	175	Clinical assessment; CT; PFT	Unknown	No	43%	8%	NA	NA	NA
Liviero et al., 2017, Italy (43)	44	Unknown	Unknown	No	36%	NA	NA	NA	NA
Toyoshima et al., 2020, Japan (45)	195	Biopsy	Unknown	No	22%	NA	NA	NA	NA
Lee et al., 2021, US (42)	41	Clinical assessment; CT; biopsy	Questionnaire	Yes	56%	12%	NA	17%	7%
Sese et al., 2021, France (40)	200	CT; biopsy	JEM	Yes	26%	11%	3%	6%	2%
Jegal et al., 2022, S. Korea (64)	1,775	Clinical assessment; CT; biopsy	Clinical interview; questionnaire	Yes	30%	6%	5%	9%	10%
Carlier et al., 2022, France (48)	31	Clinical assessment, CT	Clinical interview; questionnaire	Yes	45%	42%	23%	NA	35%

Definition of abbreviations: Agri = agricultural dusts; BAL = bronchoalveolar lavage; CT = computed tomography (thorax); CXR = chest X-ray; JEM = job exposure matrix; NA = not available; PFT = pulmonary function test; VGDF = vapors, gas, dust, or fumes, which represent all the exposure categories shown combined and, in selected studies, additional exposures.

Meta-Analysis

The reported estimates of individual studies for occupational burden are listed in Table 2. Twenty-one studies were used to calculate the occupational burden of VGDF (10 case-control studies and 11 cross-sectional studies or case series). The occupational burden (i.e., proportion with occupational exposure) for VGDF among patients with IPF was 44% (95% confidence interval [CI], 35–56%; Figure 2). Table 3 demonstrates additional exposure categories, in which metal dusts contributed a burden of 16% (95% CI, 12–19%), wood dust 8% (95% CI, 6–11%), agricultural dusts 16% (95% CI, 13–20%), and silica dust 10% (95% CI, 7–13%). Subgroup analysis limited to published full-text studies demonstrated a higher occupational burden of 47% (95% CI, 37–58%) based on 16 studies. Forest plots for the pooled occupational burden of the subcategories are provided in the data supplement.

Figure 2.

Idiopathic pulmonary fibrosis (IPF) occupational burden. Forest plot of studies relevant to estimating the contribution of work exposures to IPF. The occupational prevalence of IPF, confidence interval (CI), and weighted contribution for each study are shown, as are the calculated pooled estimate (red dashed line) and 95% CI. The pooled proportion of occupational IPF among all IPF cases is 46% (95% CI, 37–55%). ES = effect size.

Table 3.

Pooled occupational burden (proportion) estimates for occupation and idiopathic pulmonary fibrosis

Exposure	Estimate	Occupational Burden (95% CI)
VGDF	21	44% (36–53%)
Metal dusts	18	16% (12–19%)
Wood dusts	16	8% (6–11%)
Agricultural dusts	15	16% (13–20%)
Silica/stone dust	16	10% (7–13%)

Definition of abbreviations: CI = confidence interval; VGDF = vapors, gas, dust, or fumes, which represent all the other exposure categories shown combined, and, in selected studies, additional exposures.

Awadalla and colleagues (27) stratified their study sample by male and female, contributing two risk estimates to the wood, agricultural, and silica subcategories.

The reported estimates of individual case-control studies are listed in Table 4. The pooled ORs and PAFs for each exposure category are provided in Table 5. The pooled OR for VGDF was 1.8 (95% CI, 1.3–2.4), with a pooled PAF of 21% (95% CI, 15–28%) based on 11 risk estimates (Figure 3). There was no relationship between publication year and OR. The ORs and PAFs, respectively, were found to be 1.6 and 7% (metal dust), 1.6 and 3% (wood dust), 1.8 and 14% (agricultural dust), and 1.8 and 4% (silica dust). The pooled ORs and PAFs for specific exposure categories are provided in Table 5, with forest plots provided in the data supplement.

Table 4.

Case-control studies of occupational risk factors for idiopathic pulmonary fibrosis

Study, Year, Location, Reference	Cases	OR (95% CI)					PAF
		VGDF	Metal	Wood	Agri	Silica	VGDF	Metal	Wood	Agri	Silica
Scott et al., 1990, UK (28)	40	1.3 (0.8–2.0)	11.0 (2.3–52.4)	2.9 (0.9–9.9)	10.9 (1.2–96.0)	1.6 (0.5–4.8)	17%	12%	10%	12%	5%
Hubbard et al., 1996, UK (29)	218	NA	1.7 (1.1–2.7)	1.7 (1.0–2.9)	NA	NA	NA	9%	6%	NA	NA
Mullen et al., 1998, UK (30)	15	2.4 (0.7–8.4)	NA	3.3 (0.4–25.8)	NA	11.0 (1.1–115.0)	20%	NA	7%	NA	20
Baumgartner et al., 2000, US (31)	248	NA	2.0 (1.0–4.0)	1.6 (0.8–3.3)	1.6 (1.0–2.5)	3.9 (1.2–12.7)	NA	5%	3%	9%	2%
Hubbard et al., 2000, UK (35)	22	NA	1.1 (0.4–2.7)	NA	NA	NA	NA	5%	NA	NA	NA
Miyake et al., 2005, Japan (36)	102	5.6 (2.1–17.9)	9.6 (1.7–181.1)	NA	NA	1.8 (0.5–7.0)	26%	11%	NA	NA	5%
Gustafson et al., 2007, Sweden (32)	140	1.1 (0.7–1.7)	0.9 (0.5–1.6)	1.2 (0.7–2.2)	NA	1.4 (0.7–2.7)	6%	NA	3%	NA	3%
García-Sancho et al., 2011, Mexico (33)	100	2.8 (1.5–5.5)	NA	NA	NA	NA	50%	NA	NA	NA	NA
Awadalla et al., 2012, Egypt (27)
Men	95	NA	1.6 (0.7–3.6)	2.7 (1.0–7.4)	1.0 (0.4–2.3)	1.1 (0.5–2.7)	NA	6%	9%	NA	1%
Women	106	NA	NA	4.3 (0.8–22.1)	3.3 (1.2–10.1)	NA	NA	NA	6%	14%	NA
Kim et al., 2017, S. Korea (38)	70	NA	0.8 (0.2–3.9)	1.2 (0.3–4.4)	4.5 (1.3–16.2)	8.8 (1.1–73.5)	NA	NA	1%	17%	13%
Koo et al., 2017, S. Korea (37)	78	2.7 (0.7–10.9)	5.0 (1.4–18.2)	2.5 (0.5–12.3)	NA	1.2 (0.4–3.8)	35%	22%	5%	NA	5%
Paolocci et al., 2018, Italy (21)	69	4.1 (2.3–7.5)	3.8 (1.2–12.2)	1.4 (0.5–4.0)	2.4 (1.3–4.3)	NA	67%	10%	3%	32%	NA
Reynolds et al., 2019, UK (22)	466	1.7 (1.2–2.3)	1.4 (0.9–2.0)	1.4 (0.9–2.3)	NA	2.9 (1.3–6.7)	12%	5%	3%	NA	3%
Abramson et al., 2020, Australia (34)	503	1.1 (0.9–1.4)	1.3 (0.9–1.9)	0.7 (0.3–1.8)	1.0 (0.7–1.5)	0.6 (0.4–0.8)	3	3	NA	NA	NA
Park et al., 2020, S. Korea (23)	206	2.1 (1.2–3.7)	0.8 (0.2–2.9)	2.1 (0.7–7.7)	1.8 (0.9–3.9)	2.5 (1.0–6.7)	18%	NA	2%	NA	6%
Zubairi et al., 2023, Pakistan (25, 26)	131	1.1 (0.7–1.7)	NA	NA	NA	NA	2%	NA	NA	NA	NA

Definition of abbreviations: Agri = agricultural dusts; CI = confidence interval; NA = not available; OR = odds ratio; PAF = population attributable fraction; VGDF = vapors, gas, dust, or fumes, which represent all the exposure categories shown combined and, in selected studies, additional exposures.

Table 5.

Pooled OR and PAF estimates for occupation and idiopathic pulmonary fibrosis

Exposure	Risk Estimates	Pooled OR (95% CI)	Pooled PAF (95% CI)
VGDF	11	1.8 (1.3-2.4)	21% (15–28%)
Metal dusts	13	1.6 (1.2-2.1)	7% (5–10%)
Wood dusts	13	1.6 (1.2-2.0)	3% (2–4%)
Agricultural dusts	8	1.8 (1.2-2.7)	14% (8–20%)
Silica	11	1.8 (1.1-3.0)	4% (2–5%)

Definition of abbreviations: CI = confidence interval; OR = odds ratio; PAF = population attributable fraction; VGDF = vapors, gas, dust, or fumes, which represent all the other exposure categories shown combined, and, in selected studies, additional exposures.

Awadalla and colleagues (27) stratified their study sample by male and female, contributing two risk estimates to the wood, agricultural, and silica subcategories.

Figure 3.

Idiopathic pulmonary fibrosis (IPF) odds ratio (OR) and population attributable fraction (PAF) for vapors, gas, dust, or fumes from case-control studies. Forest plot of articles relevant to estimating the occupational contribution to IPF of vapors, gas, dust, or fumes (combined categories of exposure considered in the studies included). The estimated OR (A) and PAF (B), confidence interval (CI), and weighted contribution for each study are shown, as are the calculated pooled estimate (red dashed line) and 95% CI. (A) The pooled OR is 1.9 (95% CI, 1.4–2.6). (B) The pooled PAF is 24% (95% CI, 16–32%). ES = effect size.

Heterogeneity and Bias Assessment

To investigate possible publication bias, we looked for funnel plot asymmetry for VGDF (Figure 4A), which visually appears to show a decreased number of studies with ORs of less than 1.0 among those studies with larger standard errors. A series of sensitivity analyses demonstrated a similar pooled OR for VGDF. Studies with larger sample sizes demonstrated a pooled OR for VGDF of 1.5 (95% CI, 1.1–2.0). The sensitivity analysis for VGDF among studies with OR adjusted for smoking was similar at 1.9 (95% CI, 1.4–2.7), with a pooled PAF of 24% (95% CI, 16–33%). Studies with healthier controls recruited from the general population or from populations outside of respiratory clinics demonstrated a pooled OR of 1.7 (95% CI, 1.3–2.4) and a PAF of 22% (95% CI, 13–31%). An additional sensitivity analysis was performed of studies that required CT scans for diagnosis based on the 2000 American Thoracic Society and European Respiratory Society International Consensus Statement (17). This analysis demonstrated an OR of 1.9 (95% CI, 1.3–2.6) and a PAF of 22% (95% CI, 14–29%).

Figure 4.

(A) Funnel plot for meta-analysis. To investigate possible publication bias, we looked for funnel plot asymmetry for vapors, gas, dust, or fumes, which visually appears to show a decreased number of studies with odds ratios of less than 1.0 among studies with larger standard errors (demonstrated by the hollow red square). (B) Egger’s test for publication bias. The Egger’s test is a linear regression of the effect estimates on their standard errors weighted by their inverse variance. Using the Egger’s test, we found some evidence of publication bias for vapors, gas, dust, or fumes (P = 0.02), agricultural dust (P = 0.03), and silica dust (P < 0.01), but not for metal dust (P = 0.10) or wood dust (P = 0.13). CI = confidence interval.

There was heterogeneity in case-control studies of occupational exposures in IPF. The I² statistics for the 11 studies used to calculate the pooled OR were 71.9% for VGDF and 71.0% for silica dust. They were lower for metal dust (42.9%) and agricultural dust (55.1%). Wood dust demonstrated low heterogeneity, with an I² statistic of 0%. Using Egger’s test (Figure 4B), we found some evidence of publication bias for VGDF (P = 0.02), agricultural dust (P = 0.03), and silica dust (P < 0.01), but not for metal dust (P = 0.10) or wood dust (P = 0.13). Begg’s test demonstrated a P value of 0.15 for VGDF, which does not indicate publication bias due to excessive influence of studies with small sample sizes.

Because of the possibilities of methodological issues, we tabulated study case and control definitions and exposure methods. We assessed the risk of bias using the RoB-SPEO tool (Table 6) (15). The median cumulative RoB-SPEO score was measured at 16. A high-quality subgroup analysis that focused on the seven studies with total scores no greater than 16 (21, 22, 28, 30, 33, 34, 37) found that the pooled OR for VGDF was similar to that of the overall analysis at 1.9 (95% CI, 1.3–2.8), with a similar I² statistic of 75.5%.

Table 6.

RoB-SPEO risk of bias scores for occupational idiopathic pulmonary fibrosis articles

Study, Year	S	B	E	I	SR	C	D	O	Total
Scott et al.,1990 (28)	3	4	2	2	3	1	1	1	16
Hubbard et al., 1996 (29)	3	2	2	2	3	1	1	1	14
Mullen et al., 1998 (30)	3	2	2	2	3	1	1	4	14
Baumgartner et al., 2000 (31)	3	2	2	2	3	1	1	1	14
Hubbard et al., 2000 (35)	4	4	2	3	3	2	1	2	19
Miyake et al., 2005 (36)	4	4	2	3	3	1	1	1	18
Gustafson et al., 2007 (32)	4	4	2	3	3	1	1	1	18
García-Sancho et al., 2011 (33)	3	3	2	3	3	1	1	1	16
Awadalla et al., 2012 (27)	4	4	2	3	3	1	1	1	18
Kim et al., 2017 (38)	4	4	2	2	2	1	1	1	16
Koo et al., 2017 (37)	4	2	2	2	3	1	1	1	15
Paolocci et al., 2018 (21)	3	2	3	3	2	1	1	1	15
Reynolds et al., 2019 (22)	1	1	3	3	3	1	1	1	13
Abramson et al., 2020 (34)	2	2	1	2	1	1	1	1	10
Park et al., 2020 (23)	5	2	3	3	3	1	1	1	18
Zubairi et al., 2023 (26)	4	4	2	3	3	1	1	1	19

Definition of abbreviations: B = blinding; C = conflict of interests; D = differences in the numerator and denominator; E = exposure misclassification; I = incomplete exposure data; O = other bias; RoB-SPEO = Risk of Bias in Studies estimating Prevalence of Exposure to Occupational risk factors; S = selection; SR = selective reporting of exposures.

Eight domains of bias were considered: selection, blinding, exposure misclassification, incomplete exposure data, selective reporting of exposures, conflict of interests, differences in the numerator and denominator, and other bias. Risk of bias was rated in each domain: 1 = low; 2 = probably low; 3 = probably high; 4 = high; and 5 = no information.

Discussion

This meta-analysis indicates that a considerable number of cases of IPF are attributable to inhaled occupational exposures. The proportion of cases with occupational exposures to VGDF was 44%, with a range of 8–16% within more specific exposure categories. The pooled OR was increased for VGDF at 1.8, with a pooled PAF of 21%. The ORs and PAFs, respectively, were found to be 1.6 and 7% (metal dust), 1.6 and 3% (wood dust), 1.8 and 14% (agricultural dust), and 1.8 and 4% (silica dust). Taken together, our findings indicate that 21% of IPF cases (or approximately one in five) could be prevented by removal of occupational exposure to VGDF.

These findings are important, given that IPF is otherwise considered a diagnosis without a known cause. Identifying potentially pathogenic occupational exposures is important because they are preventable through proper workplace engineering controls and personal protective equipment. Emphasis has been placed on clearly occupational fibrotic lung diseases such as asbestosis and silicosis, but a clear connection between occupational exposures and other fibrotic lung diseases such as IPF can motivate the introduction of further exposure controls. For example, the recognition of an association between agricultural dust and fibrotic lung disease could prompt further investigation into the pathophysiologic mechanism, as well as increased emphasis on lowering dust exposure in the workplace. Physicians often fail to document a thorough occupational exposure history despite the fact that occupational exposures are responsible for 860,000 illnesses annually (51, 52). Recognition of such associations could benefit patients with IPF and their families by affording them the opportunity to apply for fair compensation, and could benefit current and future workers through exposure and disease prevention.

Our analysis found increased risks of IPF with metal dust, wood dust, agricultural dust, and silica dust. The diagnosis of IPF based on radiographic and pathologic findings of UIP without clear etiology is fraught with danger for premature diagnosis without complete evaluation of etiologies. Occupational exposures, including organic, metal, and silica dust, can cause similar findings on clinical presentation. Many have postulated that work exposure may trigger a profibrotic response that is clinically indistinguishable from IPF. Organic and inorganic dust have demonstrated the development of a UIP pattern on imaging, and, without a thorough occupational exposure assessment, pulmonary fibrosis may be misclassified as idiopathic.

Agricultural and wood dusts fall under the category of organic dusts, and the pathogenesis of organic dust to lung fibrosis is not completely clear. However, the possibility of unrecognized end-stage fibrotic hypersensitivity pneumonitis (HP) cannot be excluded because this can have a UIP pattern (53). Fibrotic HP can be easily confused with IPF even by experienced physicians, especially if patients have negative precipitins or have been removed from exposure by the time of clinical evaluation. A study by Morell and colleagues found that 20 of 46 patients initially diagnosed with IPF later had their diagnosis changed to HP with additional history of exposure to animal antigens (54). There also must be the consideration of a fibrogenic process of organic dust causing pulmonary fibrosis that is yet to be described (55).

Studies that use mineralogical analysis techniques have found metal dusts (e.g., nickel, iron, and aluminum) as well as silica in the lymph node and lung tissue of patients with pulmonary fibrosis, even though a clear causation was not established (56, 57). The association between lung fibrosis and metal dust exposure can be explained by the transition of alveolar cells to mesenchymal cells caused by reactive oxygen species (ROS), and other metals can induce ROS (58, 59). These ROS can induce a fibrotic response that can appear radiographically as UIP. Silica dust exposure is associated with profibrotic response through similar means of induction of free radicals, leading to inflammation and eventual fibrosis (60). Silica particles found on biopsies of patients with IPF were associated with a more rapid decline of pulmonary function (61). A recent study demonstrated that 16–20% of silica-exposed Korean workers who had received compensation for pneumoconiosis had radiographic findings consistent with UIP (62). However, we cannot discount the possible coexposure of asbestos in the metalworking and construction industries, especially among those working in steel mills in the 20th century (63).

This meta-analysis is a contemporaneous evaluation of occupational exposures and IPF and strengthens previous findings with the inclusion of high-quality studies that have been published since prior reviews. A major strength of this meta-analysis is the inclusion of an additional six case-control studies that have been published on the topic since the last major review, more than doubling the number of patients with IPF analyzed at 2,631 cases. A review by Blanc and coworkers was conducted in 2019 and included 1,229 cases of IPF, with only limited analysis of bias, confounding, susceptibility, and study quality (7). Of the six new studies with risk estimates, four had RoB-SPEO scores of 16 or lower, indicating they were high in quality. These six studies used updated diagnostic definitions that are more relevant to today’s practice. Another interim review by Park and colleagues included two further studies that were published since the review of Blanc and coworkers, but this analysis did not include an overall VGDF assessment or assess occupational burden or PAF calculations, limiting the scope of their assessment (5). Additionally, they combined occupational and environmental exposures rather than discerning only occupational exposure. Uniquely, we included conference abstracts in our screening and analysis, provided they included adequate data, as well as papers not published in English, which have not consistently been included in past reviews. Our included articles represent a more comprehensive range of regions of the world, with many studies from Asia and Central/South America compared with other studies (including a study published in Turkish). Additionally, the inclusion of conference abstracts and a wide range of study types extended our search beyond those of other studies, ensuring a thorough literature review. We have ameliorated and updated the analysis with a high-quality literature search, extensive sensitivity analysis, and bias assessment to develop a more solid foundation of association.

A thorough assessment of publication bias was performed, and, in general, results were similar across analyses, demonstrating the robustness of our findings. Even though the I² statistic shows marked heterogeneity in study results, heterogeneity was expected because of the differences across exposures, study populations, and study designs. All ORs in the pooled OR calculation for VGDF were greater than 1.0, demonstrating homogeneity and strengthening our findings. To further evaluate for publication bias, we analyzed the pool OR of studies with larger sample sizes, demonstrating a pooled OR of 1.5, reinforcing our findings, in addition to consistent findings in only studies of higher quality (OR = 1.9), studies adjusted for smoking (OR = 1.9), and studies that required CT scans for IPF diagnosis (OR = 1.9). The Egger’s test demonstrated funnel plot asymmetry for VGDF, agricultural dust, and silica dust exposures, which may indicate concern for publication bias, with fewer papers published with an OR of less than 1.0. The Begg’s test did not demonstrate publication bias in VGDF, demonstrating that studies with small sample sizes did not excessively influence our findings. The tests for publication bias did demonstrate that some may be present, emphasizing the importance of publishing negative results; nevertheless, the impact of publication bias is not sufficient to diminish the association between occupational exposures and IPF.

This study has limitations, primarily based on the quality of the parent studies. Heterogeneity was high, especially for the VGDF calculations, even though all risk estimates demonstrated an OR greater than 1. A thorough sensitivity analysis was performed, but not all heterogeneity was explained. Variability could be present as a result of differential assessment of occupational exposure by clinicians conducting the case-control studies, whereby patients with occupational exposure were less likely to be diagnosed with IPF. There was additional variability present in control group definitions. Subgroup analysis for healthier controls did not explain the overall heterogeneity. The remaining heterogeneity should inform future study design, including the need for a standardized questionnaire and control groups to assess occupational exposures. Recall bias is likely because many studies relied on clinical interviews and questionnaires to gather occupational exposure history. Six of 16 studies used a job-exposure matrix with the intent to decrease differential recall of exposures, and, as noted, we considered these to be of higher quality. Only one study grouped environmental and occupational exposures to VGDF (38). Therefore, there is the possibility for a misclassified source of exposure, but this risk was believed to be low. There is a lack of industrial hygiene confirmation of exposures through objective exposure measurement (e.g., dust air sampling) or duration of exposure, which is a common limitation of many occupational exposure studies. Our overall bias assessment performed with the Rob-SPEO tool (15) showed that, collectively, these studies were at risk for bias arising from selection, lack of blinding, exposure misclassification, incomplete exposure data, and selective reporting of exposures. However, on average, the quality rating fell into the “probably low risk” category. These limitations again emphasize the importance of future studies conducted with high-quality exposure methods to further elucidate the relationship between occupational exposures and IPF.

In summary, this study reports associations between inhalational occupational exposures and the diagnosis of IPF, an otherwise idiopathic disease with a poor prognosis and increasing epidemiology. Despite limitations of heterogeneity and varied exposure assessment, a substantive portion of patients diagnosed with IPF have disease that is likely attributed to occupational exposures to VGDF, including agricultural, wood, metal, and silica dusts. Recognizing this association as a preventable form of pulmonary fibrosis is paramount to further preventive efforts within workspaces and the environment to improve lung health.