SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. REFERENCES
  9. Supporting Information

Objective

Farming and agricultural pesticide use has been associated with 2 autoimmune rheumatic diseases, rheumatoid arthritis (RA) and systemic lupus erythematosus (SLE). However, risk associated with other residential or work place insecticide use is unknown.

Methods

We analyzed data from the Women's Health Initiative Observational Study (n = 76,861 postmenopausal women, ages 50–79 years). Incident cases (n = 213: 178 for RA, 27 for SLE, and 8 for both) were identified based on self-report and use of disease-modifying antirheumatic drugs at year 3 of followup. We examined self-reported residential or work place insecticide use (personally mixing/applying by self and application by others) in relation to RA/SLE risk, overall and in relation to farm history. Hazard ratios (HRs) and 95% confidence intervals (95% CIs) were adjusted for age, race, region, education, occupation, smoking, reproductive factors, asthma, other autoimmune diseases, and comorbidities.

Results

Compared with never used, personal use of insecticides was associated with increased RA/SLE risk, with significant trends for greater frequency (HR 2.04, 95% CI 1.17–3.56 for ≥6 times/year) and duration (HR 1.97, 95% CI 1.20–3.23 for ≥20 years). Risk was also associated with long-term insecticide application by others (HR 1.85, 95% CI 1.07–3.20 for ≥20 years) and frequent application by others among women with a farm history (HR 2.73, 95% CI 1.10–6.78 for ≥6 times/year).

Conclusion

These results suggest residential and work place insecticide exposure is associated with the risk of autoimmune rheumatic diseases in postmenopausal women. Although these findings require replication in other populations, they support a role for environmental pesticide exposure in the development of autoimmune rheumatic diseases.


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. REFERENCES
  9. Supporting Information

Autoimmune rheumatic diseases (ARDs) such as rheumatoid arthritis (RA) and systemic lupus erythematosus (SLE) affect as many as 1.6 million adults in the US (1) and disproportionately impact women, the elderly (RA), and minorities (SLE) (2–4). Established risk factors include family history of autoimmune disease and genetic characteristics. Like most complex diseases, the etiology of ARD is also influenced by environmental exposures. However, knowledge of specific environmental risk factors is limited (5), with most evidence of associations pertaining to smoking and various occupational exposures (6, 7).

The occupation of farming has been associated with ARD in different populations and using different study designs, with several studies suggesting a role of pesticide exposure from farming (8–12). One recent study reported an association between farming occupation and death with systemic autoimmune disease (including RA and SLE), specifically for work in the crop industry (8). Using a job exposure matrix, the authors also described an association of occupational pesticide exposure and RA that was primarily related to farming and significant only in men. Other studies have shown an association of farming and RA primarily in men (10, 13–15), and a study of female spouses of licensed pesticide applicators enrolled in the Agricultural Health Study did not identify any association between RA and specific pesticides (16). On the other hand, a recent study described an association of elevated serum organochlorine pesticides and self-reported RA in a population-based sample of adults in the US (17). Another study reported an association of SLE with agricultural pesticide mixing (18). In population controls, pesticide mixing was also related to antinuclear antibodies, which can be a preclinical, although nonspecific, marker preceding the onset of autoimmune disease (19).

The low prevalence of professional farm work and occupational pesticide exposure, especially in women, limits the power of many studies to examine these associations with ARD. In contrast, pesticide exposure in other residential and work place settings is much more common, e.g., use of household insecticides with ingredients similar to those used for agriculture, although at lower concentrations. Using data from the Women's Health Initiative Observational Study (WHI-OS), we tested the hypothesis that insecticide use may increase ARD risk. We investigated whether self-reported residential and work place insecticide use (mixing/application by self and application by others) was associated with the development of RA and SLE in postmenopausal women. We also considered the role of farm history.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. REFERENCES
  9. Supporting Information

Sample and case classification.

The WHI-OS is a multicenter cohort of 93,676 postmenopausal women, ages 50–79 years, enrolled from 40 clinics throughout the US from 1993 to 1998 (20). The participants completed clinic visits at baseline, were contacted yearly to update their health status, and completed another in-person clinic visit at year 3. Questionnaires collected data on medical conditions and lifestyle factors. The participants were also asked to bring current medications to the clinic visits at baseline and at year 3. The institutional review boards of the participating institutions approved protocols and consent forms, which were signed by the women at enrollment.

An analysis sample included baseline participants with complete data who were at risk of developing RA or SLE (see Supplemental Figure 1, available in the online version of this article at http://onlinelibrary.wiley.com/journal/10.1002/(ISSN)2151-4658). We excluded prevalent cases (n = 815; case classification described below) under the assumption they could not become new cases, as well as those missing data on farm history, self-report of SLE, RA, or medication use at baseline or followup (n = 16,000), for a final sample of 76,861 participants at risk. The incident case group (n = 213) included 186 RA cases and 35 SLE cases, 8 of whom reported both RA and SLE.

A previous study of the WHI determined that self-report was a nonspecific indication of SLE or RA diagnosis compared with medical records, but specificity was improved in those participants who were currently taking disease-modifying antirheumatic drugs (DMARDs; specificity 95.4 for RA and 99.4 for SLE) (21) (see Supplemental Table 1, available in the online version of this article at http://onlinelibrary.wiley.com/journal/10.1002/(ISSN)2151-4658). Therefore, we classified prevalent cases based on self-reported RA or SLE and DMARD use (including prednisone) at baseline and classified incident cases based on newly self-reported RA or SLE at year 1, 2, or 3 plus DMARD use at year 3. We evaluated the impact of potential case misclassification in several ways. We first restricted the analysis sample to increase specificity of the case definition, limiting cases to those taking nonprednisone DMARDs, excluding 58 incident cases reporting prednisone only and including 151 prevalent cases taking prednisone only as noncases (total n = 76,799) (see Supplemental Figure 1, available in the online version of this article at http://onlinelibrary.wiley.com/journal/10.1002/(ISSN)2151-4658). Second, we restricted the analyses to the most likely cases and noncases, excluding those who reported either RA or SLE alone or nonprednisone DMARD use alone (n = 3,285 at baseline, 1,521 at year 3, and 725 at both time points; total n = 71,117). This sample was also limited to those participants most likely to be incident cases (n = 174), excluding 39 new cases who reported either an ARD or nonprednisone DMARD use at baseline. Finally, we addressed potential bias due to concurrent reporting of insecticide use and ARD status, excluding 43 DMARD-confirmed cases reported in the first-year followup.

Farm history and insecticide exposure.

A baseline questionnaire collected information on farm history, i.e., whether a woman had ever lived or worked on a farm and for how many years (<5, 5–9, 10–14, 15–19, or ≥20 years). The first-year followup questionnaire included questions on insecticide exposure since age 21 years. Women were asked if they or someone else had ever “poured, mixed, sprayed, or applied insecticides” in their immediate surroundings at home, leisure, or work. The examples given were bug or flea spray and garden, lawn, or crop insecticides. Women were instructed not to include insect repellents, weed killers, fungus/mildew killers, or flea, tick, or mite treatments applied to pets. The responses were further specified as: 1) at work only, 2) at home or leisure only, and 3) at both work and home/leisure. Women with positive responses were then asked if they personally mixed insecticides, personally applied them, a lawn service applied insecticides at their homes, a commercial service applied them at their home, or other. For personal use of insecticides (direct exposure through mixing or applying by self) and application by others (indirect exposure through commercial or other residential applications), women were asked the total duration (never or <1 year, 1–4, 5–9, 10–14, 15–19, or ≥20 years) and the frequency of use (never or less than once, 1–5, 6–12, 13–24, or ≥25 times per year). The responses for direct and indirect exposures were not mutually exclusive.

Covariates.

Covariate data were derived from the baseline questionnaire responses, including potential risk factors for ARD that might be related to farm history or insecticide use, i.e., age, race/ethnicity, education, occupation, pack-years of smoking, geographic region at screening, body mass index (BMI), pregnancy and breast feeding history, age at menarche and menopause, history of asthma, thyroid disease, other autoimmune disease (multiple sclerosis or Crohn's disease), and the Charlson Comorbidity Index of chronic diseases (Table 1). Other covariates considered were region at birth, income, employment and marital status, current smoking status, vitamin D (diet/supplemental), coffee, and alcohol use.

Table 1. Characteristics of the cohort and incident RA and SLE cases in the Women's Health Initiative Observational Study*
 Study cohort (n = 76,648)RA/SLE (n = 213)PRA cases (n = 186)PSLE cases (n = 35)P
  • *

    Values are the number (percentage). RA = rheumatoid arthritis; SLE = systemic lupus erythematosus.

  • Eight cases were both RA and SLE. Chi-square test comparing cases with noncases, excluding missing values.

  • Personal use = mixed or applied; others applied = commercial applications to home or garden or other. Four cases (1.9%) and 2,499 (3.3%) in the cohort did not complete the year 1 questionnaire on insecticide use (Form 48). Missing results are shown for participants completing a Form 48. Values include “don't know.”

  • §

    Other variables showed significant associations with RA/SLE, including “underactive thyroid” (associated with SLE; P = 0.029) and “overactive thyroid” (associated with RA; P = 0.037), but with more missing (4–6%).

Insecticide use and farm history       
 Insecticide exposure  0.30 0.45 0.047
  Never (reference)21,981 (29.6)48 (23.0) 43 (23.6) 5 (14.3) 
  At work only1,199 (1.6)4 (1.9) 2 (1.1) 2 (5.7) 
  At home or leisure only35,653 (48.1)107 (51.2) 96 (52.8) 17 (48.6) 
  Both work and home8,207 (11.1)25 (12.0) 22 (12.1) 4 (11.4) 
  Don't know6,305 (8.5)22 (10.5) 17 (9.3) 6 (17.1) 
  Missing804 (1.1)3 (1.4) 2 (1.1) 1 (2.9) 
 Personally used  0.010 0.020 0.07
  Ever28,137 (38.0)94 (45.0) 83 (45.6) 16 (45.7) 
  Missing7,509 (10.1)27 (12.9) 20 (11.0) 8 (22.9) 
 Others applied  0.87 0.86 0.84
  Ever23,426 (31.6)65 (31.1) 58 (31.9) 10 (28.6) 
  Missing7,509 (10.1)27 (12.9) 20 (11.0) 8 (22.9) 
 Lived/worked on farm  0.025 0.08 0.06
  Never56,642 (73.9)143 (66.2) 127 (68.3) 21 (60.0) 
  Ever20,006 (26.1)70 (33.8) 59 (31.7) 14 (40.0) 
Sociodemographic/environmental factors       
 Age, years  0.81 0.36 0.09
  50–5410,001 (13.0)26 (12.2) 19 (10.2) 8 (22.9) 
  55–5914,313 (18.7)36 (16.9) 31 (16.7) 6 (17.1) 
  60–6416,859 (22.0)47 (22.1) 38 (20.4) 12 (34.3) 
  65–6917,339 (22.6)46 (21.6) 44 (23.7) 2 (5.7) 
  70–7412,638 (16.5)38 (17.8) 34 (18.3) 5 (14.3) 
  75–795,498 (7.2)20 (9.4) 20 (10.8) 2 (5.7) 
 Race/ethnicity  0.39 0.38 0.77
  White65,171 (85.0)175 (82.2) 154 (82.8) 28 (80.0) 
  African American5,456 (7.1)21 (9.9) 18 (9.7) 4 (11.4) 
  Hispanic2,461 (3.2)10 (4.7) 9 (4.8) 1 (2.9) 
  American Indian295 (0.4)0 (0) 0 (0) 0 (0) 
  Asian/Pacific Islander2,248 (2.9)5 (2.3) 3 (1.6) 2 (5.7) 
  Unknown1,017 (1.3)2 (0.9) 2 (1.1) 0 (0) 
 Geographic residence  0.25 0.370 0.26
  Northeast17,466 (22.8)39 (18.3) 35 (18.8) 4 (11.4) 
  South19,381 (25.3)59 (27.7) 52 (28.0) 11 (31.4) 
  Midwest17,084 (22.3)56 (26.3) 48 (25.8) 11 (31.4) 
  West22,737 (29.7)59 (27.7) 51 (27.4) 9 (25.7) 
 Education  0.008 0.06 0.024
  0–8 years931 (1.2)2 (0.9) 2 (1.1) 0 (0) 
  Some high school2,293 (3.0)11 (5.2) 10 (5.4) 2 (5.7) 
  High school/GED12,023 (15.7)42 (19.7) 35 (18.8) 8 (22.9) 
  Post high school27,466 (35.8)89 (41.8) 75 (40.3) 19 (54.3) 
  College degree or higher33,354 (43.5)69 (32.4) 64 (34.4) 6 (17.1) 
  Missing581 (0.8)0 (0) 0 (0) 0 (0) 
 Occupation  < 0.001 < 0.001 0.24
  Managerial/professional32,831 (42.8)58 (27.2) 51 (27.4) 9 (25.7) 
  Technical/sales/administrative20,923 (27.3)75 (35.2) 66 (35.5) 12 (34.3) 
  Service/labor12,179 (15.9)42 (19.7) 37 (19.9) 8 (22.9) 
  Homemaker only7,652 (10.0)30 (14.1) 26 (14.0) 4 (11.4) 
  Missing3,063 (4.0)8 (3.8) 6 (3.2) 2 (5.7) 
Health history/reproductive factors       
 Asthma  < 0.001 0.002 0.039
  No70,643 (92.2)182 (85.4) 160 (86.0) 29 (82.9) 
  Yes5,965 (7.8)31 (14.6) 26 (14.0) 6 (17.1) 
  Missing40 (0.1)0 (0) 0 (0) 0 (0) 
 Other autoimmune  < 0.001 < 0.001 0.48
  No75,157 (98.1)203 (95.3) 177 (95.2) 34 (97.1) 
  Yes1,087 (1.4)9 (4.2) 8 (4.3) 1 (2.9) 
  Missing404 (0.5)1 (0.5) 1 (0.5) 0 (0) 
 Thyroid problems§  0.12 0.12 0.49
  No57,134 (74.5)147 (69.0) 128 (68.8) 23 (65.7) 
  Yes19,104 (24.9)62 (29.1) 55 (29.6) 10 (28.6) 
  Missing410 (0.5)4 (1.9) 3 (1.6) 2 (5.7) 
 Charlson Comorbidity Index (modified)  0.02 0.036 0.81
  050,505 (65.9)123 (57.7) 109 (58.6) 20 (57.1) 
  112,923 (16.9)52 (24.4) 46 (24.7) 7 (20.0) 
  28,965 (11.7)24 (11.3) 20 (10.8) 5 (14.3) 
  ≥33,298 (4.3)11 (5.2) 9 (4.8) 2 (5.7) 
  Missing957 (1.2)3 (1.4) 2 (1.1) 1 (2.9) 
 Smoking, pack-years  0.20 0.06 0.58
  Never smoked39,095 (51.0)105 (49.3) 89 (47.8) 18 (51.4) 
  <511,046 (14.4)23 (10.8) 18 (9.7) 8 (22.9) 
  5–<2010,581 (13.8)29 (13.6) 26 (14.0) 4 (11.4) 
  ≥2013,368 (17.4)47 (22.1) 44 (23.7) 5 (14.3) 
  Missing2,558 (3.3)9 (4.2) 9 (4.8) 0 
 Body mass index, kg/m2  0.030 0.007 0.59
  <2531,967 (41.7)69 (32.4) 56 (30.1) 15 (42.9) 
  25–<3025,819 (33.7)80 (37.6) 76 (40.9) 9 (25.7) 
  ≥3017,995 (23.5)59 (27.7) 50 (26.9) 10 (28.6) 
  Missing867 (1.1)5 (2.3) 4 (2.2) 1 (2.9) 
 Age at menarche, years  > 0.99 0.15 0.002
  ≤104,903 (6.4)17 (8.0) 14 (7.5) 3 (8.6) 
  1111,992 (5.6)37 (17.4) 34 (18.3) 4 (11.4) 
  1220,082 (26.2)47 (22.1) 43 (23.1) 7 (20.0) 
  1322,384 (29.2)56 (26.3) 45 (24.2) 12 (34.3) 
  1410,000 (13.0)32 (15.0) 31 (16.7) 3 (8.6) 
  154,115 (5.4)16 (7.5) 14 (7.5) 2 (5.7) 
  ≥163,001 (4.0)8 (3.8) 5 (2.7) 4 (11.5) 
  Missing171 (0.2)0 (0) 0 (0) 0 (0) 
 Age at menopause, years  0.28 0.58 0.09
  <406,536 (8.5)24 (11.3) 20 (10.8) 5 (14.3) 
  40–449,413 (12.3)28 (13.1) 22 (11.8) 8 (22.9) 
  45–4919,138 (25.0)54 (25.4) 48 (25.8) 8 (22.9) 
  50–5428,423 (37.1)74 (34.7) 65 (34.9) 11 (31.4) 
  ≥5510,404 (13.6)20 (9.4) 19 (10.2) 1 (2.9) 
  Missing2,734 (3.6)13 (6.1) 12 (6.5) 2 (5.7) 
 No. term pregnancies  0.06 0.10 0.55
  Never pregnant7,803 (10.2)12 (5.6) 11 (5.9) 2 (5.7) 
  Never term pregnancy2,093 (2.7)9 (4.2) 8 (4.3) 1 (2.9) 
  16,895 (9.0)21 (9.9) 20 (10.8) 1 (2.9) 
  220,158 (26.3)44 (20.7) 39 (21.0) 7 (20.0) 
  318,567 (24.2)56 (26.3) 46 (24.7) 12 (34.3) 
  411,021 (14.4)37 (17.4) 31 (16.7) 7 (20.0) 
  ≥59,752 (12.7)34 (16.0) 31 (16.7) 5 (14.3) 
  Missing359 (0.5)0 (0) 0 (0) 0 (0) 
 Months breastfed  0.65 0.75 0.007
  Never breastfed37,004 (48.3)96 (45.1) 88 (47.3) 11 (31.4) 
  1–619,508 (25.5)61 (28.6) 53 (28.5) 10 (28.6) 
  7–128,584 (11.2)21 (9.9) 19 (10.2) 3 (8.6) 
  13–236,578 (8.6)20 (9.4) 13 (7.0) 9 (25.7) 
  ≥244,013 (5.2)14 (6.6) 12 (6.5) 2 (5.7) 
  Missing961 (1.3)1 (0.5) 1 (0.5) 0 (0) 

Statistical analysis.

Analyses were conducted using SAS, version 9.1. Following bivariate analyses of RA/SLE and RA or SLE with covariate and insecticide-use variables, multivariate analyses were limited to those with complete data on all variables (excluding 4 [1.9%] incident cases and 2,499 [3.3%] in the cohort with missing data on the insecticide questionnaire). Cox regression was used to estimate hazard ratios (HRs) and 95% confidence intervals (95% CIs) in adjusted models. The model building process was established a priori, first including potential confounders (related to exposure and disease), then adding other potential risk factors, and finally considering farm history as a possible explanatory variable for insecticide-related effects. Models were constructed as follows: model 1 = exposure variable + age; model 2 = model 1 + environmental and sociodemographic covariates (pack-years smoking, education level, occupational class, geographic region); model 3 = model 2 + physiologic and medical covariates (including BMI, pregnancy and breast feeding history, age at menarche and menopause, history of asthma, thyroid disease, other autoimmune diseases, and Charlson Comorbidity Index of chronic diseases); and model 4 = model 3 + farm history. Models assessing frequency and duration of personal use of insecticides were adjusted for application by others, and vice versa. Other covariates were not included in the final models, either because they were not independently associated with disease risk or they did not contribute additional information to the models. Income and employment were marginally associated with RA/SLE status, but they were dropped due to having a larger proportion of missing data.

We modeled associations for each of the 2 types of exposure (personal use or application by others), both for duration and frequency separately and for cumulative measures. Trend test P values were obtained from models including categorical variables with different levels of duration/frequency, assuming a monotonic increase in effect across levels (significance P < 0.05). To reduce the influence of small cell sizes, some categories were combined for duration (none/<1, 1–4, 5–19, ≥20 years) and frequency (never/<1, 1–5, ≥6 times per year). Cumulative scores for “application-years” were constructed as the product of years and frequency of use. Each category was assigned the midpoint value (i.e., 7 for 5–9 years) or 50% above the highest category (i.e., ≥20 years was assigned a value of 30). If duration or frequency was missing, a value was assigned for use in cumulative estimates based on the most common level reported by others in the same duration or frequency stratum among noncases. Analyses were also run excluding participants with missing frequency or duration data (<1% cases and noncases: for personal use n = 3 for cases and 618 for noncases; for application by others n = 1 for cases and 590 for noncases). These cumulative scores were then regrouped for analyses, combining levels with sparse numbers of cases to derive a 3-level variable (0–2 = never/very low, 3–100 = low/moderate, >100 = high/very high).

Effect modification was considered in stratified analyses. To quantify apparent differences, we tested interactions by product terms included in a single multivariable model. Specifically, we examined whether associations varied by farm history, which might affect pesticide-use patterns and the distribution of unmeasured covariates.

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. REFERENCES
  9. Supporting Information

A new diagnosis of RA/SLE with DMARD use was reported by 213 women for a total of 0.28% of the cohort in 3 years. Insecticide use, farm history, and characteristics of cohort participants and cases are shown in Table 1. The majority (63%) of cases reported insecticide exposure at home, while fewer (14%) also reported exposure at work, and there was no significant difference in exposure venue compared with other cohort participants. Cases were more likely to report personal use of insecticides (mixed/applied by self) than noncases (45% versus 38%; P = 0.010), but there was no significant difference in application by others. Having lived or worked on a farm was more common in cases than noncases (34% versus 26%; P = 0.025). Cases also had less education, were less likely to report professional or managerial occupations, and were more likely to report asthma or other autoimmune diseases. Compared with noncases, RA cases reported higher smoking pack-years, higher BMI, and more comorbidities (higher Charlson Comorbidity Index), while SLE cases were younger, reported earlier menopause, differences in age at menarche, and were more likely to have breastfed.

Table 2 lists exposure-specific RA/SLE rates and adjusted HRs. Women reporting no insecticide exposure had the lowest 3-year risk (0.22%), and rates were similar for women with exposure at home (0.30%) or work (0.31%). RA/SLE risk was also higher in women who reported not knowing about their insecticide exposures (0.35%), higher for those reporting personally mixing (0.32%) and personally applying (0.34%) insecticides, and for other types of insecticide use (0.36%). Compared with women reporting no exposures, personal use of insecticides (mixing or applying) was significantly associated with increased risk (age-adjusted HR 1.57). Women reporting frequent personal use had the highest risk of ARD (0.50% for ≥6 times per year), with significant dose responses for duration, frequency, and cumulative use. Adding model 3 covariates slightly diminished these associations (10–20%). For application by others, frequency and duration trends were nonsignificant, but higher risk was seen for the longest duration category (age-adjusted HR 1.85 for ≥20 years). In models including farm history, associations persisted for both personal use of insecticides and long-term application by others. Sensitivity analyses showed little impact of varying case and cohort definitions (data not shown).

Table 2. Associations of farm exposure and insecticide use with risk of developing RA/SLE in the WHI-OS*
 3-year risk RA/SLE, %Cases, (%)OS cohort, (%)Hazard ratio (95% CI)
Age-adjustedFully-adjustedFarm-adjusted§
  • *

    RA = rheumatoid arthritis; SLE = systemic lupus erythematosus; WHI = Women's Health Initiative; OS = observational study; 95% CI = 95% confidence interval.

  • By Cox proportional hazards regression. Models analyzing frequency, duration, and cumulative dose of personal use were also adjusted for any commercial use. Similarly, models analyzing commercial exposure were also adjusted for any personal use.

  • Adjusted for age (5-year intervals), race/ethnicity, education, occupation, region, pack-years of smoking, body mass index, parity, months breastfed, age at menarche, age at menopause, history of asthma, thyroid disease and other autoimmune disease, and modified Charlson Comorbidity Index.

  • §

    Adjusted for covariates in fully-adjusted model plus farm history.

Insecticide use      
 Never0.22206 (23)73,551 (30)1.01.01.0
 Work only or home and work0.31206 (14)73,551 (13)1.52 (0.96–2.43)1.42 (0.85–2.38)1.39 (0.83–2.34)
 Home only0.30206 (52)73,551 (49)1.44 (1.02–2.02)1.35 (0.92–1.96)1.34 (0.92–1.95)
 Don't know0.35206 (11)73,551 (9)1.60 (0.97–2.66)1.70 (1.00–2.91)1.70 (1.00–2.91)
Type of insecticide use      
 Never or other method (reference)   1.01.01.0
 Personally mixed0.32206 (15)73,551 (13)1.28 (0.86–1.89)1.36 (0.89–2.08)1.34 (0.88–2.06)
 Personally applied0.34206 (46)73,551 (38)1.57 (1.18–2.11)1.51 (1.09–2.09)1.50 (1.08–2.07)
 Lawn service applied0.29206 (20)73,551 (19)1.09 (0.77–1.55)1.23 (0.84–1.79)1.24 (0.85–1.81)
 Commercial application0.22206 (15)73,551 (19)0.78 (0.53–1.15)0.73 (0.47–1.14)0.74 (0.47–1.15)
 Other0.36206 (12)73,551 (9)1.43 (0.93–2.20)1.29 (0.79–2.09)1.28 (0.79–2.07)
Personally mixed or applied      
 Years      
  Never used insecticides0.22181 (27)66,627 (33)1.01.01.0
  Never/<10.21181 (21)66,627 (27)1.05 (0.65–1.70)0.95 (0.56–1.62)0.94 (0.55–1.60)
  1–40.36181 (16)66,627 (12)1.82 (1.12–2.95)1.52 (0.88–2.64)1.50 (0.87–2.61)
  5–190.29181 (15)66,627 (14)1.45 (0.89–2.36)1.26 (0.72–2.19)1.24 (0.71–2.16)
  ≥200.43181 (21)66,627 (13)2.07 (1.31–3.25)1.97 (1.20–3.23)1.94 (1.18–3.19)
  P for trend   < 0.0010.0030.003
 Times/year      
  Never used insecticides0.22180 (27)66,508 (33)1.01.01.0
  Never/<10.24180 (24)66,508 (28)1.24 (0.78–1.98)0.98 (0.58–1.66)0.97 (0.57–1.65)
  1–50.30180 (34)66,508 (31)1.52 (1.02–2.29)1.49 (0.96–2.33)1.47 (0.94–2.30)
  ≥60.50180 (15)66,508 (8)2.47 (1.51–4.03)2.04 (1.17–3.56)2.01 (1.15–3.51)
  P for trend   < 0.0010.0030.003
 Cumulative use      
  Never/very low0.23182 (50)66,878 (60)1.01.01.0
  Low/moderate0.30182 (37)66,878 (34)1.35 (0.98–1.85)1.48 (1.04–2.09)1.46 (1.03–2.07)
  High/very high0.53182 (13)66,878 (6)2.36 (1.49–3.74)2.20 (1.31–3.71)2.17 (1.29–3.67)
  P for trend   < 0.0010.0010.002
Application by others      
 Years      
  Never used insecticides0.22182 (26)66,796 (33)1.01.01.0
  Never/<10.34182 (25)66,796 (20)1.26 (0.74–2.12)1.10 (0.62–1.97)1.10 (0.61–1.96)
  1–40.22182 (13)66,796 (15)0.89 (0.51–1.56)0.89 (0.48–1.64)0.89 (0.48–1.64)
  5–190.23182 (18)66,796 (21)0.95 (0.57–1.57)0.93 (0.53–1.63)0.93 (0.53–1.63)
  ≥200.45182 (18)66,796 (11)1.85 (1.13–3.04)1.85 (1.07–3.20)1.86 (1.07–3.21)
  P for trend   0.180.140.13
 Times/year      
  Never used insecticides0.22181 (27)66,543 (33)1.01.01.0
  Never/<10.36181 (27)66,543 (21)1.39 (0.83–2.31)1.20 (0.68–2.12)1.19 (0.67–2.11)
  1–50.29181 (37)66,543 (35)1.19 (0.77–1.85)1.19 (0.73–1.92)1.19 (0.73–1.93)
  ≥60.23181 (9)66,543 (11)0.98 (0.54–1.80)0.99 (0.51–1.91)0.99 (0.51–1.91)
  P for trend   0.850.900.89
Cumulative use      
  Never/very low0.27182 (51)66,965 (52)1.01.01.0
  Low/moderate0.29182 (43)66,965 (41)1.01 (0.74–1.37)1.06 (0.75–1.49)1.06 (0.75–1.50)
  High/very high0.23182 (5)66,965 (6)0.81 (0.42–1.56)0.89 (0.44–1.80)0.90 (0.44–1.82)
  P for trend   0.700.980.99

Disease-specific risks ranged from 0.02% for SLE and 0.20% for RA in those reporting no or low personal exposure-years to 0.12% for SLE and 0.44% for RA in those reporting high or very high cumulative personal use. Figure 1 shows associations for insecticide use in relation to RA or SLE, with similar findings for personal use as for RA/SLE combined. Application by others was nonsignificantly associated with SLE (P = 0.10 for trend). Other significant associations in multivariate models for RA included older age, higher BMI, nonprofessional/managerial occupation, asthma, and other reported autoimmune diseases, while SLE-specific models showed significant associations with lower education (data not shown).

thumbnail image

Figure 1. Disease-specific models of insecticide exposure and risk of rheumatoid arthritis or systemic lupus erythematosus. * = hazard ratios (HRs) and 95% confidence intervals (95% CIs) estimated from a Cox proportional hazards regression model, adjusted for age (5-year intervals), race/ethnicity, education, occupation, region, pack-years of smoking, body mass index, parity, months breastfed, age at menarche, age at menopause, history of asthma, history of other autoimmune disease, history of thyroid disease, modified Charlson Comorbidity Index, and farm exposure). Models analyzing personal use were also adjusted for any application by others, and vice versa.

Download figure to PowerPoint

RA/SLE risk was elevated in women with a longer duration of farm experience (0.48% for ≥20 years living or working on a farm), with an age-adjusted HR of 1.97 for ≥20 years (95% CI 1.14–3.42). This diminished after adjusting for model 3 covariates and was substantially attenuated in models accounting for cumulative personal use of insecticides and application by others (fully-adjusted HR 1.22, 95% CI 0.58–2.56). The association of RA/SLE with application by others differed significantly by farm history of the women (interaction P = 0.004) (Table 3). In women without a farm history, frequent application by others was not associated with increased RA/SLE risk (HRadj 0.34 for ≥6 times per year), whereas among women with a farm history, the HRadj was 2.95 for ≥6 times per year. The highest risk (0.81%) was seen in women with a farm history who reported higher cumulative personal use of insecticides, and the lowest risk (0.20%) was seen in women with no reported farm history or insecticide use. The risk associated with farm history plus higher insecticide use (age-adjusted HR 4.02) was somewhat reduced (HR 3.36), adjusting for model 3 covariates (Figure 2).

Table 3. Associations of RA/SLE with personal insecticide use and application by others, stratified by farm history*
 Never lived/worked on farmEver lived/worked on farm
3-year risk RA/SLECases, (%)OS cohort, (%)Adjusted HR (95% CI)3-year risk RA/SLECases, (%)OS cohort, (%)Adjusted HR (95% CI)
  • *

    RA = rheumatoid arthritis; SLE = systemic lupus erythematosus; OS = observational study; HR = hazard ratio; 95% CI = 95% confidence interval.

  • Estimated from a Cox proportional hazards regression model, adjusted for age (5-year intervals), race/ethnicity, education, occupation, region, pack-years of smoking, body mass index, parity, months breastfed, age at menarche, age at menopause, history of asthma, history of other autoimmune disease, history of thyroid disease, modified Charlson Comorbidity Index, and farm exposure). Models analyzing frequency, duration, and cumulative dose of personal use were also adjusted for any application by others, and vice versa.

Personally mixed or applied        
 Years        
  Never used insecticides0.20124 (26)49,255 (33)1.00.2857 (28)17,372 (33)1.0
  Never/<10.19124 (22)49,255 (28)1.01 (0.53–1.92)0.2657 (19)17,372 (24)0.80 (0.30–2.11)
  1–40.39124 (19)49,255 (12)1.92 (1.01–3.63)0.2857 (11)17,372 (12)0.74 (0.25–2.24)
  5–190.25124 (14)49,255 (14)1.31 (0.66–2.57)0.4157 (19)17,372 (16)1.09 (0.41–2.88)
  ≥200.41124 (12)49,255 (12)2.40 (1.33–4.33)0.4957 (23)17,372 (15)1.22 (0.49–3.05)
  P for trend   0.002   0.47
 Times per year        
  Never used insecticides0.20124 (26)49,175 (33)1.00.2856 (29)17,333 (33)1.0
  Never/<10.22124 (25)49,175 (29)1.00 (0.53–1.90)0.3156 (23)17,333 (24)0.89 (0.34–2.30)
  1–50.31124 (37)49,175 (30)1.91 (1.24–3.21)0.2656 (27)17,333 (33)0.72 (0.30–1.74)
  ≥60.40124 (12)49,175 (8)2.01 (0.99–4.08)0.7156 (23)17,333 (10)1.93 (0.76–4.90)
  P for trend   0.003   0.43
 Cumulative use        
  Never/very low0.21125 (50)49,447 (61)1.00.3057 (51)17,431 (56)1.0
  Low/moderate0.31125 (41)49,447 (33)1.85 (1.22–2.80)0.2757 (30)17,431 (37)0.80 (0.41–1.56)
  High/very high0.40125 (10)49,447 (6)2.15 (1.08–4.29)0.8157 (19)17,431 (8)2.18 (0.96–4.91)
  P for trend   0.002   0.28
Others applied        
 Years        
  Never used insecticides0.20124 (26)49,401 (33)1.00.2858 (28)17,395 (32)1.0
  Never/<10.38124 (29)49,401 (19)1.23 (0.61–2.46)0.2558 (17)17,395 (23)0.72 (0.24–2.21)
  1–40.19124 (12)49,401 (16)0.76 (0.31–1.61)0.3158 (14)17,395 (15)1.21 (0.42–3.49)
  5–190.17124 (14)49,401 (21)0.64 (0.31–1.34)0.4258 (24)17,395 (19)1.77 (0.71–4.40)
  ≥200.43124 (18)49,401 (11)1.79 (0.92–3.46)0.5458 (17)17,395 (11)2.05 (0.75–5.55)
  P for trend   0.73   0.028
 Times per year        
  Never used insecticides0.20124 (26)49,236 (33)1.00.2857 (28)17,307 (33)1.0
  Never/<10.40124 (31)49,236 (20)1.30 (0.66–2.58)0.2557 (18)17,307 (23)0.83 (0.27–2.51)
  1–50.28124 (40)49,236 (36)1.14 (0.63–2.05)0.3157 (32)17,307 (34)1.25 (0.52–3.01)
  ≥60.07124 (3)49,236 (11)0.34 (0.11–1.02)0.6957 (23)17,307 (11)2.95 (1.16–7.52)
  P for trend   0.14   0.019
Cumulative use        
  Never/very low0.27124 (55)49,523 (52)1.00.2758 (43)17,422 (54)1.0
  Low/moderate0.26124 (43)49,523 (42)0.56 (0.57–1.30)0.3758 (45)17,422 (40)1.75 (0.91–3.35)
  High/very high0.09124 (2)49,523 (7)0.35 (0.11–1.13)0.6358 (12)17,422 (6)2.93 (1.13–7.64)
  P for trend   0.10   0.017
thumbnail image

Figure 2. Joint effect of farm history and personal insecticide exposure on rheumatoid arthritis or systemic lupus erythematosus risk. * = hazard ratios (HRs) and 95% confidence intervals (95% CIs) estimated from a Cox proportional hazards regression model, adjusted for age (5-year intervals), race/ethnicity, education, occupation, region, pack-years of smoking, body mass index, parity, months breastfed, age at menarche, age at menopause, history of asthma, history of other autoimmune disease, history of thyroid disease, modified Charlson Comorbidity Index, and farm exposure. Models were also adjusted for any application by others.

Download figure to PowerPoint

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. REFERENCES
  9. Supporting Information

Our findings are consistent with the hypothesis that insecticide exposure may increase ARD risk in postmenopausal women. Our study extends previous evidence of an association of farming by showing an association of RA/SLE with self-reported personal use of insecticides, primarily in residential settings. These findings suggest that exposure to residential insecticides may play a role in RA/SLE etiology in the general population, as well as in those exposed occupationally. The plausibility of our findings is strengthened by the dose response seen for increasing frequency or duration of personal use, and similar findings for RA and SLE. The observed associations also did not appear to be explained by farm history or other risk factors examined.

In contrast, an apparent association of self-reported farm history (ever lived or worked) and RA/SLE risk was partly explained by covariate adjustment and substantially attenuated in models including insecticide exposure variables. However, risk was highest in women who reported both a farm history and higher personal use of insecticides or application by others. Women with a farm history may report insecticide use more accurately, have higher exposures within the same use categories, or apply different products. Compared with nonfarmers, measurement studies have shown higher exposures in farmers who applied pesticides and among family members living in a farm environment where pesticides were used (22, 23). Other farming practices or exposures (e.g., inorganic and organic dusts) also might impact ARD risk (24, 25).

Our study is subject to limitations inherent in its design and in the use of self-reported data in a cohort that is not optimized for either of these specific exposures or outcomes. Nevertheless, our case definition was likely to identify actively treated RA/SLE patients with a first diagnosis during followup. Although self-report of ARD is nonspecific, this definition was based on a previous validation in WHI participants showing high specificity of cases identified by self-reported RA or SLE plus current DMARD use compared with the gold standard of chart review (21). With this definition, the estimated occurrence of new RA/SLE in the cohort (0.28% in 3 years) was within an expected range for this population (e.g., incidence of 1 per 1,000 per year for RA) (26). DMARD use in self-reported cases varied by race and education status (data not shown); if such factors are related to disease diagnosis or DMARD treatment in true cases, our findings may not represent the full range of incident RA/SLE. Also, while we expected our case definition to be highly specific, some true cases were likely included in the baseline sample due to the lack of concurrent DMARD use. New cases not reporting DMARDs may have been missed by our case definition. Corticosteroid use (e.g., prednisone) was included as a DMARD in our case definition, but is widely used for other conditions and was the most common DMARD reported by noncases and incident cases at baseline. However, the influence of a small proportion of false-negative cases in the cohort is probably minimal, and sensitivity analyses showed no substantive differences in main effects based on various alternative definitions of case and noncase groups.

Relatively few studies have considered risk factors for RA and SLE in parallel and exclusively among postmenopausal women. SLE has a younger peak of age at onset, so we cannot know if these findings are generalizable to the majority of SLE patients. We saw some expected associations, primarily for RA. For example, older age was a significant risk factor for RA. Smoking was uncommon and only modestly associated with RA, whereas a fairly consistent association of smoking and RA has been observed in other studies (27). Other risk factors for RA were higher BMI and history of diagnosis with other autoimmune diseases and asthma. We also observed associations of potential indicators of lower socioeconomic status (SES) were associated with increased RA/SLE risk; specifically, nonprofessional or managerial occupation was associated with RA and lower education with SLE. Lower SES has been associated with RA in previous studies (28–30), although it is unclear whether such findings are influenced by biased case ascertainment methods or cohort participation. Individual measures of SES could plausibly be related to a variety of other risk factors or exposures on the pathway leading to autoimmunity, or to more broadly measured community-level characteristics, such as neighborhood poverty. A recent study described regional differences in RA incidence in a large cohort of nurses, with higher rates in the Northeast compared with the Western US (31). We saw no evidence of this pattern in the WHI; instead, RA and SLE rates were slightly, but nonsignificantly, higher in the South and Midwest compared with the Northeast. We are unaware of variations in diagnosis or treatment practices that might lead to such differences, although one study described clusters of SLE mortality in the South/Southwest partially attributed to geographic differences in poverty (32).

Our findings are primarily based on exposure data collected before the report of incident ARD, with limited potential for differential recall or reporting bias. Cases could have had symptoms when they reported their exposure, but we saw little impact of excluding cases reported in the first year of followup. Although recall of insecticide use is prone to nonspecific error, a recent study showed high reliability of self-reported residential insecticide use (33). Considering likely sources of error, we might expect the observed associations to be attenuated estimates of the underlying relationship. We had no information on the types of pests, application practices, or the chemicals used, so our results reflect average risks that combine likely differing risks for (unmeasured) specific exposures.

Associations with insecticides were not meaningfully changed in models adjusting for a range of demographic, socioeconomic, health, and behavioral covariates. However, some effect estimates were reduced by >10% in these full models, suggesting potential confounding by covariates examined. Although we adjusted for several factors associated with SES (education, occupation, health covariates), there may be residual confounding because SES is a complicated construct encompassing diverse individual and community-level characteristics. The observed associations could have been influenced by other unmeasured participant characteristics, such as living conditions that favor insects, the use of other household chemical products, or other personal, household, or environmental factors. To confound the observed associations, however, such factors would have to be strongly associated with both ARD risk and insecticide use. We are unaware of known risk factors that might explain the observed associations.

Since two SES measures (education and occupation) were related to RA/SLE risk, we further explored associations of RA/SLE with insecticides in stratified analyses (results not shown). Occupational class might be related to differences in occupational and environmental exposures, while education could impact awareness of insecticides as a potential hazard, use of measures to reduce personal exposures, or accuracy of reported insecticide use. Self-report of “not knowing” about insecticide use (8.5% of the cohort, 9.3% of RA, and 17% of SLE cases) was associated with RA/SLE in adjusted models. Both the associations of RA/SLE with personal use of insecticides and with “not knowing” were the least apparent in women with a high school education or less and in those working in technical/sales/administrative jobs. However, both associations were clearly seen in women at higher education levels or in those working in jobs classified as managerial/professional, service/labor, or homemakers. These apparent effect differences by SES highlight a need to further understand SES-related differences in RA/SLE risk in the cohort.

We conducted other exploratory analyses to better understand the observed associations (results not shown), speculating that insecticide use and associations might vary by region or age. Personal use of insecticides was mostly associated with RA/SLE in younger participants, while the association with longer-term application by others was primarily seen in older participants. This may reflect differences in recall, age at exposure, time since use, and temporal trends in products available for personal and professional use, as well as the potential for longer duration of insecticide use in the older participants. The association of RA/SLE risk with personal use of insecticides was more apparent in the West and Southeast, where higher insecticide use was more common in both the cohort and cases. Measurement studies show geographic differences in types and amounts of insecticides used and detected in environmental sampling (34), but it is premature to speculate on how these might relate to potential regional differences in association.

The mechanisms by which insecticides might lead to development of ARD are diverse, and few studies have been conducted on animal models of autoimmunity. Toxicology studies suggest a complex relationship given the diversity of pesticide effects on the immune system (35, 36): pesticides may impact differentiation and regulation of adaptive and innate immune responses, leading to acute and chronic immune suppression and decreased response to infections, inflammation, and autoimmunity. Exposures may directly impact disease risk, such as acceleration of autoimmunity in lupus-prone mice (37). Indirect effects could include modification of responses to other exposures, such as interactions of organophosphates and endotoxins, a common combination in agricultural settings (38).

The older age of the WHI-OS cohort is in some ways well suited to studies of environmental risk factors for ARD. Older age at onset and the declining female-to-male ratio in older age at onset of ARDs may reflect a larger component of risk due to nonintrinsic, nonhormonal, or environmental factors (39–42). Our findings are consistent with a hypothesis that personal insecticide use, primarily residential, may increase the risk of developing ARD in postmenopausal women. Work is needed to further characterize RA and SLE cases in the WHI and to conduct a detailed assessment of other environmental exposures. Our findings support the need for replication studies in other populations and for research to identify specific chemicals that increase susceptibility or promote ARD risk. Finally, our results provide new evidence of a potential role for a common environmental exposure in risk of developing autoimmune diseases.

AUTHOR CONTRIBUTIONS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. REFERENCES
  9. Supporting Information

All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be submitted for publication. Dr. Parks had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Parks, Walitt, Chen, Sarto, Howard.

Acquisition of data. Hunt, Sarto, Howard.

Analysis and interpretation of data. Parks, Walitt, Pettinger, Chen, De Roos, Sarto, Howard.

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. REFERENCES
  9. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. REFERENCES
  9. Supporting Information

Additional Supporting Information may be found in the online version of this article.

FilenameFormatSizeDescription
ACR_20335_sm_onlinesupplemental.doc63KSupplementary Data

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.