To investigate the association between primary systemic vasculitis (PSV) and environmental risk factors.
To investigate the association between primary systemic vasculitis (PSV) and environmental risk factors.
Seventy-five PSV cases and 273 controls (220 nonvasculitis, 19 secondary vasculitis, and 34 asthma controls) were interviewed using a structured questionnaire. Factors investigated were social class, occupational and residential history, smoking, pets, allergies, vaccinations, medications, hepatitis, tuberculosis, and farm exposure in the year before symptom onset (index year). The Standard Occupational Classification 2000 and job-exposure matrices were used to assess occupational silica, solvent, and metal exposure. Stepwise multiple logistic regression was used to calculate the odds ratio (OR) and 95% confidence interval (95% CI) adjusted for potential confounders. Total PSV, subgroups (47 Wegener's granulomatosis [WG], 12 microscopic polyangiitis, 16 Churg-Strauss syndrome [CSS]), and antineutrophil cytoplasmic antibody (ANCA)–positive cases were compared with control groups.
Farming in the index year was significantly associated with PSV (OR 2.3 [95% CI 1.2–4.6]), with WG (2.7 [1.2–5.8]), with MPA (6.3 [1.9–21.6]), and with perinuclear ANCA (pANCA) (4.3 [1.5–12.7]). Farming during working lifetime was associated with PSV (2.2 [1.2–3.8]) and with WG (2.7 [1.3–5.7]). Significant associations were found for high occupational silica exposure in the index year (with PSV 3.0 [1.0–8.4], with CSS 5.6 [1.3–23.5], and with ANCA 4.9 [1.3–18.6]), high occupational solvent exposure in the index year (with PSV 3.4 [0.9–12.5], with WG 4.8 [1.2–19.8], and with classic ANCA [cANCA] 3.9 [1.6–9.5]), high occupational solvent exposure during working lifetime (with PSV 2.7 [1.1–6.6], with WG 3.4 [1.3–8.9], and with cANCA 3.3 [1.0–10.8]), drug allergy (with PSV 3.6 [1.8–7.0], with WG 4.0 [1.8–8.7], and with cANCA 4.7 [1.9–11.7]), and allergy overall (with PSV 2.2 [1.2–3.9], with WG 2.7 [1.4–5.7]). No other significant associations were found.
A significant association between farming and PSV has been identified for the first time. Results also support previously reported associations with silica, solvents, and allergy.
The primary systemic vasculitides (PSV) are a group of relatively rare diseases of unknown etiology (1). A number of potential environmental risk factors have been reported. Systemic vasculitis has been associated with exposure to particulate silica (e.g., quartz, granite, sandstone, and grain dust) (2–5). There has been conflicting evidence regarding a link between occupational exposure to hydrocarbons (e.g., paints, glues) and Wegener's granulomatosis (WG) and microscopic polyangiitis (MPA), and the association with glomerulonephritis is stronger (6–8). A case–control study carried out at the National Institutes of Health (NIH) revealed an association with exposure to fumes or particulates and pesticides in patients with WG compared with healthy or rheumatic disease controls but not respiratory disease controls (9); however, exposure to metal and welding fumes has been reported to increase the risk of chronic renal failure, glomerulonephritis, WG, and pulmonary vasculitis (5, 10, 11). Allergies and a family history of atopy were associated with WG, Churg-Strauss syndrome (CSS), and polyarteritis nodosa in a case–control study (12). Reports also suggest that desensitization procedures, vaccinations (e.g., influenza and hepatitis), drugs (e.g., propylthiouracil and hydralazine), and withdrawal of steroids can precipitate PSV in some individuals (13, 14). Hepatitis B and hepatitis C are strongly associated with polyarteritis nodosa and cryoglobulinemic vasculitis, respectively, and infections (e.g., parvovirus B19 and chlamydia) might be important in other types of PSV (15–17).
We designed a case–control study to investigate the role of environmental factors in PSV. Norfolk, in eastern England, is favorable for epidemiologic study because of its geographic location, partially bordered by coastline. The Norfolk and Norwich University Hospital (NNUH) is the single central referral center for the Norwich Health Authority (NHA) population of ∼415,000 adults. This population has also been used for prospective studies of rheumatoid arthritis (by the Norfolk Arthritis Register) (18) and of the epidemiology of PSV (19). Patients living in the area surrounding the former NHA are also referred to the NNUH. Few patients are referred out of district without first attending this hospital.
All adult patients (>15 years old) diagnosed as having PSV at the NNUH between May 1988 and July 2000 were identified from a prospective vasculitis register and were invited to participate. The register has been maintained since January 1988 through direct referral by hospital physicians and review of the hospital histopathologic and inpatient discharge records. Details of the methods for patient identification have been reported previously (19). Case ascertainment is thought to be complete because general practitioners (the first point of contact for patients in the UK) serving the denominator population refer all PSV cases to the NNUH. The register benefits from good communication between primary and secondary care providers, between practitioners in different hospital specialties, and between adjacent health districts. Potentially, cases with mild disease (e.g., limited WG) could remain undetected and severe cases could have been undiagnosed at the time of death.
Case notes were reviewed for clinical and laboratory details including antineutrophil cytoplasmic antibody (ANCA) findings. Cases were classified as follows: WG according to the American College of Rheumatology (ACR) criteria (20) and the Chapel Hill Consensus Conference (CHCC) definition (21), MPA according to the CHCC definition (21), and CSS according to the ACR criteria (22) and the Hammersmith criteria (23). Several cases fulfilled definitions/criteria for both MPA and WG. For the purposes of the case–control study, those with ear, nose, and throat (ENT) involvement, nodules or cavitation seen on chest radiograph, or positive classic ANCA (cANCA) (and/or specificity for proteinase 3 [PR3]) were classified as having WG; remaining patients with predominantly renal vasculitis and no ENT symptoms were classified as having MPA. Patients with primary vasculitis who did not fulfill these criteria or who had secondary vasculitis (e.g., systemic lupus erythematosus, infection, cryoglobulinemia, or malignancy) were excluded.
One hundred three cases of PSV were identified, of whom 27 had died prior to the start of the study, 1 declined to participate, and 75 were recruited (98.7% response rate of available cases, 72.8% of incident cases). Forty-seven patients were classified as having WG, 12 MPA, and 16 CSS. Thirty PSV cases were cANCA and/or PR3 positive, 19 were perinuclear ANCA (pANCA) and/or myeloperoxidase (MPO) positive, 9 had positive ANCA by radioimmunoassay, 16 were ANCA negative, and in 1 case there was no recorded ANCA result. Fifty-eight patients had renal involvement (defined as hematuria, increased creatinine level, or renal biopsy findings consistent with vasculitis), and 45 had respiratory involvement (defined as hemoptysis, persistent cough or dyspnea, or chest radiography findings consistent with vasculitis).
Two hundred twenty hospital inpatients and outpatients of the orthopedic and rheumatology departments with noninflammatory musculoskeletal disease (e.g., osteoarthritis, fracture, osteomyelitis) and common symptoms (e.g., cellulitis, deep vein thrombosis, skin lesion excision) were recruited as nonvasculitis controls (3 controls per case). Fifteen additional patients declined to participate (93.6% response rate). All nonvasculitis controls were matched for sex and age at vasculitis onset (within 3 years) to the case. The first 83 controls were matched to the age that cases had been when they first experienced a symptom attributable to vasculitis (index date) and were asked to recall events in the year immediately preceding the interview. All other controls were matched to the age of the case at the time of interview and were asked to think back to events that occurred prior to the index date of the matched case (the change was made to minimize recall bias). Age-matched controls were not obtained for 1 17-year-old WG patient who presented at the end of the data collection period.
Patients were excluded if they had a personal history of any of the following: autoimmune disease including rheumatoid arthritis, thyroid disease, diabetes, vitiligo, or pernicious anemia; connective tissue disease including systemic lupus erythematosus, scleroderma, myositis, Sjögren's syndrome, or undifferentiated connective tissue disease, seronegative spondylarthropathy, or inflammatory bowel disease. The rationale for their exclusion was that many studies have linked autoimmune diseases to some of the environmental factors under investigation, so bias could have been introduced and real associations overlooked. In addition, patients with dementia, mental handicap, or cognitive impairment caused by acute illness or medication, as well as severely ill patients, were excluded due to a likely reduction in ability to complete the questionnaire accurately.
Nineteen of 20 surviving adults (>15 years old) who had been diagnosed in 1988 or later as having systemic rheumatoid vasculitis (SRV) and who resided in the NHA district were recruited to compare risk factors for primary and secondary vasculitis. The register was known to be complete for SRV but had not been verified for other types of secondary vasculitis, so patients with secondary vasculitis other than SRV were not recruited.
Thirty-four asthma controls, age- and sex-matched to PSV cases, were recruited consecutively from an asthma clinic to serve as specific controls for CSS, which is characterized by asthma. None of the patients recruited as asthma controls declined to participate.
There were no significant differences in the geographic area of residence between control groups and cases. Urban areas were defined according to postal codes using details provided by the Norfolk City Council Demographic Unit, and rural areas constituted the surrounding areas in the NHA district. The percentage of PSV cases who lived in rural areas was 61.3%, compared with 57.6% of nonvasculitis controls, 61.1% of SRV controls, and 37.5% of asthma controls (Table 1).
|Diagnosis*||Age, years||Rural residence, %†||Norfolk history, %‡||Manual occupation, %§|
|WG (n = 47)||58.9||59||17–89||61.7||46.8||34.0|
|CSS (n = 16)||59.4||60.5||31–84||62.5||62.5||20.0|
|MPA (n = 12)||66.3||67.5||42–76||58.3||58.3||41.7|
|Nonvasculitis (n = 220)||60.7||62||28–87||57.6||56.4||31.8|
|Asthma (n = 34)||57.9||58||21–84||37.5||52.9||32.4|
|SRV (n = 19)||55.8||56||22–74||61.1||52.6||21.1|
Ethics approval was obtained from the Norwich District Ethics Committee, and the study was registered with the research and development office prior to recruitment. PSV and SRV patients were invited, either in person at a clinic appointment or by letter, to participate in the study. Nonvasculitis and asthma controls were invited in person at the time of their clinic appointment or inpatient admission. All participants gave informed consent.
A structured questionnaire to measure factors of interest and potential confounders was designed to be administered by an interviewer. It was based on 2 previously used questionnaires (9, 12) and adapted for the Norfolk population. Additional questions were formulated after a literature review. A copy of the full questionnaire may be obtained from the authors on request. A single interviewer (SEL), not blinded to case/control status of interviewees or to the study hypothesis, conducted interviews with all PSV patients, SRV patients, and nonvasculitis controls, and a second interviewer (NJI) interviewed the asthma controls. NJI was trained to interview in the same manner as SEL, by observation of technique. The first part of the questionnaire was used to establish patient details, and in PSV cases the date of the first symptom attributable to vasculitis (the index date) was estimated. All subsequent questions were related only to exposures prior to this date.
Subhypotheses for particular environmental factors were formulated, and questionnaire data were used to investigate each individually. Definitions of each item investigated were as described below. Age at the index date and social class were analyzed. Social class was defined by coding the last occupation by the Standard Occupational Classification (SOC) 2000 and then deriving social class using the Office for National Statistics Occupational Support service guide 6.1 (24). Nonmanual occupations (social classes I, II, IIIN) were compared with manual occupations (social classes IIIM, IV, V). Urban/rural residence was defined by postal district as previously described, and “Norfolk history” was defined as birth and subsequent main area of residence in Norfolk or a history of generations of the family living in Norfolk. Occupational exposure to silica, solvents, and metals was investigated. Occupations were coded using the SOC 2000 and job exposure matrices used to define high, intermediate, and low exposure. Appendix A shows the actual occupations of the study participants as coded in the SOC 2000. They are categorized according to the job exposure matrices described below and in refs.25–27. For each of silica, solvents, and metals, “any exposure,” “high exposures,” “intermediate exposures,” and “low exposures” were investigated for both the index year and any time during the working life. In addition, the duration of occupational exposure to silica and solvents was analyzed using data categorized by 5-year periods.
Direct questions were asked to ascertain any history of specific jobs previously reported to be associated with silica exposure and vasculitis, i.e., coal worker (coal miner/delivery), sandblaster, mine/quarry worker, baker, dental worker, painting and decorating, working in the construction industry. The job exposure matrix of Parks et al (25) was used to define high silica exposure (including agriculture, roofing, automobile repair, cement and ceramics working, construction industry, foundry, iron, and steel working, mining, quarrying, shipbuilding, paint, rubber, plastics, and abrasive industries), intermediate silica exposure (including dental occupations, baking, sculpting, occupations ancillary to occupations with high exposures, or the construction trade, e.g., carpentry, painting/decorating, electrical work), and low exposure (all other occupations). Silica exposure was also divided into “agricultural silica exposure” (exposure to grain dust, including individuals participating in harvesting, those handling grain in other occupations, and those otherwise exposed, e.g., living within 30 feet of grain harvest) and “nonagricultural silica exposure” (exposure to other silica dust, e.g., quartz, sand, or flint in the construction industry, mining, or sandblasting).
The job exposure matrix published by Steenland et al was used to define occupational solvent exposures (26). A new matrix was designed to assess occupational exposures to metals. Occupations defined as having high metal exposures were aerospace/aircraft manufacture, metal workers including smelting, milling, automobile/heavy machinery and lead battery manufacture, metal mining, and welding. Intermediate metal exposures included canning, construction, cement and ceramic production, chemical, rubber, electronics, nuclear and leather industries, boat-building, dental work, soldering, paint spraying, explosives, fertilizer and pesticides (27). All other occupations were considered to have low exposure.
A detailed history about farming was obtained, and factors analyzed were “ever farm” (exposure to farm/agriculture of any type in the index year and/or any occupational exposure to farms in a working lifetime), index farm (exposure to a farm in the index year), crop exposure in the index year, livestock exposure in the index year, and exposure to both livestock and crops in the index year. The type of farming exposure (living, working, or visiting a farm) and frequency of exposure (daily, weekly, less frequently) were also included. Although the questionnaire was not initially designed to address this, the proximity to livestock was investigated by comparing subgroups who handled livestock directly (“direct”), those who worked at or were regularly present on a farm with livestock (“close”), and those who were only exposed within 30 feet of livestock (“distant”).
Allergy was investigated by comparing self-reported allergy at any time prior to the index date (any allergy, skin, drug, insect, food). Other factors investigated were allergic rhinitis (recurrent, seasonal, runny blocked nose before the index date), asthma (personal history prior to the index date), and family history of allergy or asthma in parents, siblings, or children. Other items studied included owning/regular contact with a domestic pet in the index year, any blood transfusion prior to the index date, a self-reported history of personal tuberculosis (TB) or contact with an individual with TB, infectious hepatitis (A, B, C), any vaccination within 6 months of the index date, steroid withdrawal within 6 months of the index date, smoking history (smoking ever, in the index year, and pack-years).
Conditional binary logistic regression was carried out using Minitab Release 13 software for each exposure of interest to obtain odds ratios (ORs) and 95% confidence intervals (95% CIs). Stepwise multiple logistic regression was used to calculate ORs and 95% CIs adjusted for potential confounding factors. Because there are no known definite risk factors for PSV, the model included all other factors that had shown a significant association on binary regression, rural residence (an earlier study suggested that CSS was more common in rural areas) (28), social class (a potential confounder for occupational data especially), and smoking (may confound in respiratory disease in particular). A stepwise approach was used rather than using a fixed model, because a fixed model may have resulted in the loss of excessive degrees of freedom, especially where numbers of cases were small.
In view of the unknown etiology and pathogenesis of the disease, the cohort was investigated in a number of ways. The whole cohort of PSV patients was compared with nonvasculitis and SRV controls. Disease subtypes (WG, MPA, and CSS) may have different triggering factors and were compared with nonvasculitis controls individually. CSS was also compared specifically with the asthma controls. Additionally, different factors may be important in triggering a particular pattern of disease (e.g., an airborne stimulant may be responsible for respiratory vasculitis and renal disease may be related to solvents), and different ANCA types may have distinct etiologies. Therefore, cases were also divided according to renal and respiratory involvement and ANCA specificity and compared with nonvasculitis controls. Cases were divided into 4 groups for the analysis of ANCA according to immunofluorescence and enzyme-linked immunosorbent assay results: cANCA- and/or PR3-positive cases, pANCA- and/or MPO-positive cases, ANCA-negative cases, and ANCA result unknown. It must be noted that the clinical use of ANCA results and laboratory methods for detection have changed over the period that cases recruited to the study were diagnosed. Therefore, unknown bias is introduced by the use of retrospective ANCA data, and results should be interpreted with this in mind.
There were no significant differences between any pairs of groups investigated, in terms of age, sex, social class, urban/rural residence, or a personal/family history of longstanding residence within Norfolk (Table 1).
PSV was significantly associated with farming exposures both during the working lifetime (ever farm) and within the index year (index farm) (Table 2). Significant ORs of index farm were found for WG (2.7 [95% CI 1.2–5.8]) and MPA (6.3 [1.9–21.6]) but not for CSS (2.1 [0.6–7.0]), and a significant OR of ever farm was found for WG (2.7 [(1.3–5.7]) but not for MPA (2.5 [0.7–9.0]) or CSS (0.8 [0.3–2.6]). This may suggest a specific association of farming with WG/MPA or may reflect the relatively small numbers of CSS and MPA cases. It was not possible to distinguish between exposures to crops, livestock, or individual animal species because most individuals were exposed to more than one type and there were no significant differences in associations between modes of exposure (Table 2) or frequencies of exposure. ORs were significant for exposure to cows (3.0 [95% CI 1.1–8.0]) and chickens (4.0 [1.5–10.9]).
|Item||PSV, no. (%)||Nonvasculitis controls, no. (%)||Adjusted OR (95% CI)|
|Farm ever†||46/75 (61.3)||82/220 (37.3)||2.2 (1.2–3.8)|
|Farming exposure during index year|
|Any farming exposure‡||25/75 (33.3)||30/219 (13.7)||2.3 (1.2–4.6)|
|Living near a farm§¶||7/58 (12.1)||12/202 (5.9)||2.2 (0.8–6.0)|
|Working on a farm§||3/54 (5.6)||6/196 (3.1)||2.5 (0.5–11.7)|
|Visiting a farm§||9/60 (15.0)||11/201 (5.5)||3.0 (1.1–7.9)|
|Living & working on farm§¶||4/55 (7.3)||0/190 (0)||–|
|Livestock only#||6/56 (10.7)||6/195 (3.1)||3.7 (1.1–12.4)|
|Crops only#||9/59 (15.3)||14/203 (6.9)||1.7 (0.7–4.6)|
|Livestock and crops#||10/60 (16.7)||10/199 (5.0)||4.4 (1.7–11.7)|
Working in occupations with high silica exposure in the index year was associated with significantly increased ORs for PSV (3.0 [95% CI 1.0–8.4]) (Table 3) and CSS (5.6 [1.3–23.5]), but not WG (2.5 [0.8–8.5]) or MPA (3.2 [0.8–8.5]). Working in these jobs at any time during a working lifetime did not have any positive associations. However, there was a significant association between MPA and occupations defined as having high or intermediate exposures throughout a working lifetime (OR 4.6 [95% CI 1.3–16.5]). It is likely that these differences between diagnoses are attributable to the small numbers of exposed cases. There were no significant findings for intermediate- or low-exposure jobs alone. There was no trend toward an increased risk of PSV with duration of silica exposure, although there was a significant relationship of PSV to jobs of >480 months (OR 2.63 [95% CI 1.03–6.73]). Although not a primary question in the questionnaire design, it was noted that PSV showed a weak association with agricultural silica (in the form of agricultural dust) (OR 4.4 [95% CI 1.1–18.1]) but not nonagricultural silica (e.g., brick dust) (1.2 [0.4–3.3]).
|Item||PSV, no. (%)||Nonvasculitis controls, no. (%)||Adjusted OR (95% CI)|
|High silica ever†||18/75 (24.0)||42/220 (19.1)||1.4 (0.7–2.7)|
|High silica index year‡||10/72 (13.9)||9/211 (4.3)||3.0 (1.0–8.4)|
|High solvent ever†||11/75 (14.7)||15/220 (6.8)||2.7 (1.1–6.6)|
|High solvent index year||7/75 (9.3)||5/220 (2.3)||3.4 (0.9–12.5)|
|High metal ever†||5/75 (6.7)||12/220 (5.5)||1.0 (0.3–3.6)|
|High metal index year||0/75 (0)||3/220 (1.4)||–|
A history of high solvent exposure at any time was associated with PSV (OR 2.7 [95% CI 1.1–6.6]) (Table 3) and WG (3.4 [1.3–8.9]). The association was greater for individuals working in these occupations in the index year (for PSV, OR 3.4 [0.9–12.5]; for WG, OR 4.8 [1.2–19.8]). No significant associations were found for MPA (any exposure 1.3 [0.1–12.9]; index year 2.7 [0.2–34.1]), for CSS (no cases exposed), or for jobs defined as having intermediate or low exposures or high/intermediate exposures combined.
There were no significant associations of occupational metal exposure, either throughout the working lifetime or in the index year, with PSV (Table 3) (no cases exposed in the index year) or any subgroup (data not shown). There were no significant associations for any of the individual job codes investigated. This may have been due to small numbers of individuals with each job.
A history of allergy was significantly associated with PSV overall (OR 2.2 [95% CI 1.2–3.9]) (Table 4) and with WG (2.7 [1.4–5.7]), but not with MPA (1.5 [0.4–5.7]) or CSS (2.4 [0.8–7.2]). The only subtype of allergy that showed a significant association was drug allergy (for PSV, OR 3.6 [95% CI 1.8–7.0]; for WG, OR 4.0 [95% CI 1.8–8.7]). There was no association between PSV and family history of allergy or asthma (Table 4) but, as may be anticipated, CSS was significantly associated with rhinitis, asthma, and steroid withdrawal when compared with nonvasculitis controls but not with asthma controls (data not shown).
|Item||PSV, no. (%)||Nonvasculitis controls, no. (%)||Adjusted OR (95% CI)|
|Allergy†‡||43/75 (57.3)||82/217 (37.8)||2.2 (1.2–3.9)|
|Drug allergy†‡§||25/73 (34.2)||29/217 (13.4)||3.6 (1.8–7.0)|
|Family history of atopy‡||27/75 (36.0)||67/217 (30.9)||1.1 (0.6–1.9)|
|Blood transfusion†||13/75 (17.3)||55/220 (25.0)||0.5 (0.3–1.1)|
|Vaccination¶||17/75 (22.7)||37/220 (16.8)||1.7 (0.8–3.4)|
|Smoke ever†||53/75 (70.7)||156/220 (70.9)||0.9 (0.5–1.6)|
|Smoke index year||15/75 (20.0)||55/220 (25.0)||0.7 (0.4–1.5)|
|Tuberculosis†#||21/74 (28.4)||53/220 (24.1)||1.4 (0.7–2.6)|
There were no positive associations with previous blood transfusion, recent vaccinations, smoking, or TB history (Table 4). No significant associations were found for having domestic pets, history of hepatitis, or gardening (data not shown).
Analysis by the presence or absence of renal and respiratory involvement and ANCA type showed little difference in associations, with few exceptions. There were significant associations of pANCA/MPO with farming in the index year (OR 4.3 [95% CI 1.5–12.7]), farming ever (3.3 [1.1–9.9]), and high occupational silica exposure in the index year (4.9 [1.3–18.6]), while cANCA/PR3 showed associations with farming ever (3.9 [1.6–9.5]), and with solvent exposures (3.3 [1.0–10.8] for ever exposed; 3.9 [1.6–9.5] for exposure in the index year) and allergy (3.1 [1.3–7.4]) (Table 5).
|Item||No. of cases||Farm index year||High silica index year||High solvent ever||Allergy|
|Renal involvement†||58||2.9 (1.4–5.8)||2.8 (0.9–8.7)||2.4 (0.9–6.6)||2.1 (1.1–3.8)|
|No renal involvement||17||3.4 (1.1–10.5)||6.5 (1.5–28.6)||2.1 (0.4–10.4)||3.0 (1.0–9.4)|
|Respiratory involvement‡||45||2.3 (1.1–5.1)||2.8 (0.9–8.7)||2.3 (0.8–6.8)||2.1 (1.1–4.2)|
|No respiratory involvement||30||4.4 (1.8–10.6)||4.1 (1.1–15.6)||3.8 (1.1–12.3)||2.8 (1.2–6.5)|
|cANCA/PR3||30||2.5 (0.9–6.3)||1.8 (0.4–8.5)||3.3 (1.0–10.8)||3.1 (1.3–7.4)|
|pANCA/MPO||19||4.3 (1.5–12.7)||4.9 (1.3–18.6)||0.6 (0.06–6.6)||1.6 (0.6–4.6)|
This study is the first to demonstrate an association between PSV and farming. Although it was not possible to distinguish between specific farming exposures (most individuals were exposed to more than one type), the association appeared to be stronger for exposure to livestock compared with crops, with significant ORs for work with cows and chickens. Results were not biased in terms of the distribution of cases or controls living in rural areas. This part of the questionnaire used direct questions and response was virtually complete, so interviewer and response bias was minimized. In a previous study from the NIH, Duna et al, using the same questionnaire (9), failed to detect a significant association between WG and farming. The proportion of cases exposed to farms during the index year was similar in the two studies (33.3% in our study, 35.6% in the NIH study), but a greater percentage of nonvasculitis controls were exposed to farms in the NIH study. The controls and cases in that study may have been drawn from differing denominator populations leading to bias, e.g., vasculitis cases are likely to have been referred for tertiary care from a wider population than controls (sequential age- and sex-matched patients attending a general medical clinic). The NIH study did not specifically exclude autoimmune disease among controls, and similar environmental factors may be important in the initiation of autoimmunity.
One explanation is that an infectious agent may play a role in the onset of PSV. Recent reports have linked a previously undescribed paramyxovirus affecting pigs with an outbreak of viral encephalitis with widespread vasculitis-induced thrombosis in Malaysian pig farmers (29). Pseudorabies virus causing vasculitis has also been noted to transfer between species, and another report suggested an association between close contact with pigs and Behçet's disease (30, 31). Alternatively, other factors in the care of livestock and their surroundings may be important, including animal feeds, antibiotics, and disinfectants. Duna et al also reported that exposure to pesticides and insecticides was associated with WG (9). This question was omitted from our study due to the retrospective design, because it is particularly subject to recall bias and overreporting by cases. Clearly, further investigation is needed to better establish the nature of the association with farming.
Our results support published evidence that silica is associated with PSV. Most previous studies indicate that the duration of silica exposure is less important than its intensity (e.g., ORs for sandblasting are much higher than for building trades) (32–34). We also found no increased risk with duration of exposure, but the questionnaire was not designed to assess levels of exposure. A difference was noted in the relationship of WG and CSS with silica versus the relationship of MPA with silica, in terms of timing of exposure (WG and CSS were associated with recent exposures). This may reflect a difference in the role of silica, e.g., acting as a trigger of disease in WG and CSS but as an autoimmune adjuvant in MPA. However, the difference could also be explained by classification of patients, the insidious onset of MPA making determination of the index date less accurate, or the relatively small number of MPA patients. Our results suggest that silica has a stronger association with pANCA/MPO-positive than with cANCA/PR3-positive PSV. This is also consistent with the literature since many reports address MPO-positive vasculitis; however, our data should be interpreted with caution because the ANCA results were collected retrospectively, and clinical use of ANCA results and laboratory methods for detection developed over the study period (35–37).
An association of PSV with solvents and hydrocarbons has been reported previously, but the specific association with WG has not. This may be due to differences in cases and classification between studies. The mean duration of exposure in our cases (23.7 years [range 3–50 years]) was similar to that in previous studies by Nuyts et al (>20 years) (5) and Steenland et al (∼20 years [range 11–28 years]) (26). This suggests that PSV is associated with prolonged exposure to solvents. A specific link with renal vasculitis or a previously proposed association with pulmonary hemorrhage (7) was not substantiated. We failed to detect any significant relationship with metal exposure. Positive associations have previously been described for very specific exposures, e.g., welding fumes (5), whereas our study included a wider range of exposures.
An association of drug allergy with WG, but not with CSS, in patients in London has previously been reported by Cuadrado et al (12), from a study using virtually the same questionnaire as the one we used. They also reported significant associations of WG and CSS with both a family history of atopy and other types of allergy, a finding we were not able to reproduce. This may be explained by their smaller number of cases (20 WG and 8 CSS) or differences between case or control groups. Because their study was conducted at a tertiary referral center, their cases are more prone to referral bias and likely to have more severe disease and to be younger than a population-based cohort such as ours. Their disease controls were selected from among rheumatology patients with common symptoms less likely to be referred for specialist consultation compared with PSV patients. The inclusion of rheumatoid arthritis patients in their control group may also have introduced a positive bias toward a hypothesis that allergy is associated with PSV, since a reduced frequency of atopy in rheumatoid arthritis has been described (38).
Reports have suggested a role for a Th2 (atopic) cytokine environment in the initiation of PSV (39), and the association of allergy with PSV may support this theory. However, drug allergies are associated with heterogeneous immune responses (40), so other immunologic mechanisms may be important. The majority of allergies described in our cohort were to antibiotics, so it is possible that drug allergy is a marker for previous exposure to infection, rendering the individual more susceptible to PSV.
We recognize that the present study has limitations, and results should be interpreted with these in mind. We are confident that case recruitment was as complete as possible, but bias may have been introduced by control selection. Resource constraints meant that we were unable to recruit a random sample from the “healthy” denominator population. Use of hospital controls could introduce referral bias, but the location of the population and the fact that the NNUH is the only referral center mean that a geographically representative sample is more likely, compared with most centers. We suggest that referral bias was minimal since there was no significant difference between cases and controls when comparing urban and rural residence defined by postal code, which reflects an increasing distance from the hospital in this population.
Some illnesses affecting the controls may have biased individual outcomes. For example, osteoarthritis patients may be less likely than PSV patients to have had manual jobs. This would not introduce a bias in many items of interest, e.g., allergy, and both cases and controls were required to describe events that occurred several years earlier, when their health status was likely to be different.
The exclusion of autoimmune disease controls may have excluded some potential triggering factors (e.g., propylthiouracil has been linked with MPO-positive PSV), but we were unable to examine this association since patients with thyroid disease were specifically excluded. Recall bias can be a problem in retrospective case–control studies but was unavoidable. It was minimized by ensuring that both cases and controls had to recall a similar period. Although it was not possible for the interviewer to be blinded to the subject or case/control status of the interviewee, observer bias was minimized by the use of direct questions. The major selection bias was the exclusion of deceased patients from the study, who may differ from survivors in terms of some of the variables investigated. However, review of the clinical features and vasculitis activity of survivors compared with deceased cases did not reveal significant differences.
In conclusion, our study has identified, for the first time, a plausible, significant association between farming and PSV, and the results support previously reported associations with silica, solvents, and allergy. Associations are sufficiently strong to warrant further investigation. It would be of interest to determine if the association with farming can be replicated in other populations. However, our study had a number of potential biases, and a prospective analysis of incident cases compared with community controls would be desirable to provide definitive results.
|Occupations of Study Participants*|
|Occupation||No. of subjects||Occupation||No. of subjects|
|Occupations with silica exposure||Intermediate solvent exposure|
|High silica exposure||Armed forces||35|
|Arable farm workers||26||Factory worker||3|
|Both farming and construction (+abrasive blasting in Lotus car manufacture)||7||Boatbuilder||6|
|Construction industry/bricklayer/roofing||21||Health and safety consultant||1|
|Road worker||2||Window cleaner||1|
|Cement industry worker||1||Cabinet/furniture maker||6|
|Abrasive blaster automobiles/maltings worker||1||Electronics engineer||1|
|Terrazzo floor polisher||1||Engineer||7|
|Total||68†||Solvents used in cleaning||5|
|Intermediate or probable silica exposure||Plastics industry||7|
|Dentist/dental technician/dental nurse||5||Manufacturing chemist/laboratory technician||6|
|Property maintenance||1||Petrol pump attendant||2|
|Painter and decorator||6||Diesel fumes/refrigerants||5|
|Coal delivery||1||Paint delivery||1|
|Land army||2||Dye manufacture||1|
|Fireman/railways/Navy stoker||2||Forestry worker||1|
|Transport of grain||1||Occupational exposure to metals|
|Boatbuilder||7||High metal exposure|
|Paint analyzer||1||Aluminum/steel plant worker||7|
|Poultry farm/grain||1||Grinding metals/abrasive work||7|
|Industrial boiler servicing||1||Iron molder||1|
|Occupations with solvent exposure||Intermediate metal exposure|
|High solvent exposure||Boatyard workers||6|
|Laundry worker||3||Cement manufacture||1|
|Aircraft fitter/aircraft fuel pumping||4||Chemical processor/analytical chemist||4|
|Painter and decorator||9||Dentist/dental nurse/dental technician||5|
|Building and roofing||1||Dye manufacture||1|
|Factory work exposed to glues||1||Electrical wiring manufacture||1|
|Car repair/paint sprayer/paint factory||2||Farm worker (possible fertilizer exposure)||35|
|Paint analyzer||1||Forestry worker||1|
|Total||29§||Gas fitter/heating engineer||1|
|Other factory worker||1|
|Painter and decorator/paint sprayer||8|