International Archives of Occupational and Environmental Health

, Volume 86, Issue 2, pp 177–187

Pesticide use and fatal injury among farmers in the Agricultural Health Study

Authors

  • Jenna K. Waggoner
    • Division of Respiratory Disease StudiesNIOSH, CDC, DHHS
    • Epidemiology BranchNIEHS, NIH, DHHS
  • Paul K. Henneberger
    • Division of Respiratory Disease StudiesNIOSH, CDC, DHHS
  • Greg J. Kullman
    • Division of Respiratory Disease StudiesNIOSH, CDC, DHHS
  • David M. Umbach
    • Biostatistics BranchNIEHS, NIH, DHHS
  • Freya Kamel
    • Epidemiology BranchNIEHS, NIH, DHHS
  • Laura E. Beane Freeman
    • Division of Cancer Epidemiology and GeneticsNCI, NIH, DHHS
  • Michael C. R. Alavanja
    • Division of Cancer Epidemiology and GeneticsNCI, NIH, DHHS
  • Dale P. Sandler
    • Epidemiology BranchNIEHS, NIH, DHHS
    • Epidemiology BranchNIEHS, NIH, DHHS
Original Article

DOI: 10.1007/s00420-012-0752-x

Cite this article as:
Waggoner, J.K., Henneberger, P.K., Kullman, G.J. et al. Int Arch Occup Environ Health (2013) 86: 177. doi:10.1007/s00420-012-0752-x

Abstract

Purpose

To assess whether pesticide use practices were associated with injury mortality among 51,035 male farmers from NC and IA enrolled in the Agricultural Health Study.

Methods

We used Cox proportional hazards models adjusted for age and state to estimate fatal injury risk associated with self-reported use of 49 specific pesticides, personal protective equipment, specific types of farm machinery, and other farm factors collected 1–15 years preceding death. Cause-specific mortality was obtained through linkage to mortality registries.

Results

We observed 338 injury fatalities over 727,543 person-years of follow-up (1993–2008). Fatal injuries increased with days/year of pesticide application, with the highest risk among those with 60+ days of pesticide application annually [hazard ratio (HR) = 1.87; 95% confidence interval (CI) = 1.10, 3.18]. Chemical-resistant glove use was associated with decreased risk (HR = 0.73; 95% CI = 0.58, 0.93), but adjusting for glove use did not substantially change estimates for individual pesticides or pesticide use overall. Herbicides were associated with fatal injury, even after adjusting for operating farm equipment, which was independently associated with fatal injury. Ever use of five of 18 herbicides (2,4,5-T, paraquat, alachlor, metribuzin, and butylate) were associated with elevated risk. In addition, 2,4-D and cyanazine were associated with fatal injury in exposure–response analyses. There was no evidence of confounding of these results by other herbicides.

Conclusion

The association between application of pesticides, particularly certain herbicides, and fatal injuries among farmers should be interpreted cautiously but deserves further evaluation, with particular focus on understanding timing of pesticide use and fatal injury.

Keywords

PesticidesMortalityWounds and injuries

Abbreviations

AHS

Agricultural Health Study

CI

Confidence interval

HR

Hazard ratio

Introduction

According to the Census for Fatal Occupational Injuries, farming remained one of the most hazardous occupations in the United States in 2008, with an estimated annual fatality rate of 40.3/100,000 workers (Bureau of Labor Statistics 2010). The complex farming environment may contribute to this increased risk. While exposure to animals (Solomon 2002) and machinery (Myers and Hendricks 2009) have been examined in regard to fatal injury, pesticide exposure has not. Pesticide use, particularly insecticide use, may contribute to fatal and non-fatal injury via neurotoxic mechanisms (London et al. 2012). Personal characteristics also play a role, with older farmers (Myers et al. 2009) and men (Myers and Hard 1995) at greater risk of fatal injury, and disabled farmers at increased risk of non-fatal injury (Sprince et al. 2002). Also, failure to use personal protective equipment when handling chemicals has been associated with increased risk of non-fatal injury (Day et al. 2009).

A recent all-cause mortality analysis of farmers enrolled in the Agricultural Health Study (AHS), a prospective cohort study of licensed pesticide applicators in IA and NC showed an increased risk of fatal injury against a backdrop of low all-cause mortality (Waggoner et al. 2011). While machinery is often the proximal mechanism of fatality, other intrinsic factors preceding the injury, including pesticide exposure, may contribute to increased risk. Medical examiner/death certificate-based studies have limited ability to identify these preceding factors; however, with its prospective data, the AHS cohort is uniquely suited to do so. Our objective was to identify whether pesticide use characteristics and other farm factors were associated with unintentional fatal injury.

Methods

Study population

The AHS is a large prospective cohort that includes 52,394 private pesticide applicators, mostly farmers, in NC and IA. Among eligible applicators, 82% enrolled in the study from 1993 to 1997 by completing a self-administered questionnaire at the time of pesticide licensing; 44% of the cohort completed a second take-home questionnaire within 4 months of enrollment. Institutional Review Boards of the National Institutes of Health and its contractors approved the study; return of the questionnaire implied consent. The present study is restricted to male private applicators (N = 51,035).

Outcome classification

Deaths were identified through annual linkage with death registries in NC and IA and the National Death Index. Injury deaths from enrollment through the end of follow-up (December 31, 2008) were defined by International Classification of Diseases (ICD) codes indicating a fatal injury ICD-9: E800–E999; ICD-10: V01-Y98. Both traffic and non-traffic motor vehicle accidents were included. Non-occupational fatal injuries were excluded, including homicide, suicide, forces of nature, medical complications, and railway, water, and air transportation accidents. No participants were lost to mortality follow-up. The comparison population consisted of farmers who did not experience a fatal injury during the study period, regardless of vital status. Individuals contributed person-years from enrollment through the end of follow-up or the date of death, whichever was earlier.

Exposure assessment

For all applicators, we had information from the enrollment questionnaire on factors that may be related to fatal injury, including occupational, behavioral, and demographic variables. The enrollment questionnaire obtained information on the use of 50 pesticides, with detailed information for 22 of these (including lifetime frequency and duration of use), general pesticide application information, use of personal protective equipment, farm characteristics, animal exposures, smoking and drinking history, medical history, and demographic factors. For personal protective equipment, we focused on the use of chemical-resistant gloves because these were shown in measurement studies of AHS applicators to be the single factor most associated with the greatest reduction in exposure among application methods studied (Hines et al. 2011; Thomas et al. 2010). The take-home questionnaire provided additional information for 28 pesticides not covered in detail on the enrollment questionnaire, as well as information on farm machinery use and additional medical characteristics. The take-home questionnaire also provided information about high pesticide exposure events, defined as “an incident or experience while using any type of pesticide which caused you unusually high personal exposure.” Differences between applicators who completed only the enrollment questionnaire and those who completed both questionnaires were small, with the most important being the slightly increased age of those completing both (Tarone et al. 1997). All AHS questionnaires are available on line at http://aghealth.nci.nih.gov/questionnaires.html.

Pesticides

Information was collected on average frequency of application of specific pesticides. To assess exposure–response for the frequency of pesticide application (days/year), we created 3- or 4-level variables depending on the number of users. Exposure–response was not evaluated for pesticides used by <1% of the cohort. For chemicals used by <5% of injured farmers, we split the distribution among users at the median (≤median or >median), with non-users as the referent group. For more frequently used pesticides, we split the distribution of users into tertiles, with non-users as the referent group. Using these criteria, 36 chemicals had sufficient data to assess exposure–response: 26 based on tertiles of use and ten chemicals were analyzed based on a median split. Due to low usage, we did not assess exposure–response models for fungicides and fumigants.

Farm practices

For those who completed the take-home questionnaire, we also assessed whether regular use patterns for large farm machinery, including driving combines, tractors, and trucks were associated with fatal injury. For tractors and trucks, frequency of use categories were daily, weekly, monthly, or less than monthly; for combine driving, a harvest-only activity, the categories were never, 1–10, 11–30, and ≥31 days/growing season. For tractors and trucks, we combined the lowest two groups to create the referent group. For combines, never users served as the referent group.

Other farm exposures evaluated included butchering animals (yes, no) and farm acreage, presented in the original categories of <5 acres (referent), 5–49, 50–199, 200–499, 500–999, and ≥1,000. We also examined collapsed categories of poultry and livestock production volumes: none (referent), <50, 50–499, and >500 animals.

Statistical analysis

We used Cox proportional hazards models to calculate hazard ratios for risk of fatal injury, using age as the timescale and adjusting for state as a covariate in all models. We considered other covariates (smoking, drinking, and acreage), but did not include them except as noted because they did not confound results, as determined by a 10% or greater change of the point estimate. Statistical significance was set at p = 0.05. Results are presented for exposures with 5 or more deaths. To examine exposure–response relationships, we created a “dose” variable by assigning consecutive integers to the response categories (starting with 0 for the referent group with no exposure and dividing the remaining exposed population into tertiles) and treated it as a continuous variable in model fitting.

Because farmers use multiple pesticides, we assessed potential confounding by examining pairwise correlations between chemicals that were significantly associated with fatal injury in single agent models. For chemicals with Spearman’s correlations ≥0.3, we constructed models including both chemicals, with one in dose categories and the other as ever used, running analysis with each chemical in each role. Using this process, we found no evidence of any inter-chemical confounding. These analyses, completed using SAS 9.2 (Cary, NC, USA), utilized AHS data release AHSREL0905.00.

Results

Among the 51,035 male farmers, 338 injury deaths occurred over the average follow-up period of 13.3 years, which provided 730,234 person-years, for a rate of 46.3 injury deaths/100,000 person-years. Farmers having a fatal injury were primarily white and 61% were from Iowa, representative of the cohort’s demographics (Table 1). Injury mortality increased with age, although decade-specific hazard ratios were not significantly increased. Decedents were similar to the entire cohort (at enrollment) with regard to body mass index. Individuals who were divorced or separated at enrollment were more likely to experience a fatal injury, as were decedents who reported experiencing difficulty with balance monthly or more frequently. Increased age, smoking, tremor, and depression were non-significantly associated with increased risk of fatal injury. Greater than high school education was associated with significantly decreased injury risk.
Table 1

Enrollment demographics and medical conditions for all male farmers and for male fatal injuries in the AHS, 1993–2008

Variable

Entire cohort

(n = 51,035)

Fatal injuries

(n = 338)

Age and state adj.

N

%

N

%

HRa

95% CIa

State

 NC

19,602

38

133

39

1.00

(Ref)

 IA

31,433

62

205

61

1.15b

(0.92, 1.45)b

Race

 White

48,277

95

314

93

1.00

(Ref)

 Other

2,758

5

24

7

1.38

(0.89, 2.12)

Age at enrollment

 <40

16,158

32

71

21

1.00

(Ref)

 40–49

13,668

27

67

20

1.02

(0.63, 1.64)

 50–59

10,773

21

69

20

1.45

(0.71, 2.94)

 60–69

7,494

15

76

22

1.67

(0.71, 3.92)

 70+

2,604

5

55

16

2.63

(0.98, 7.00)

BMI category (kg/m2)

 <25

36,834

72

246

73

1.00

(Ref)

 25–30

10,829

21

66

20

0.80

(0.61, 1.06)

 >30

3,372

7

26

8

1.15

(0.76, 1.72)

Marital status

 Married

42,647

84

277

82

1.00

(Ref)

 Divorced or separated

2,216

4

19

6

1.67

(1.05, 2.68)

 Widowed

508

1

9

3

1.64

(0.83, 3.23)

 Never

5,413

11

31

9

1.15

(0.75, 1.75)

Smoker status at enrollment

 Never

26,090

53

149

47

1.00

(Ref)

 Ever

23,111

47

167

53

1.22

(0.97, 1.53)

Alcohol consumption at enrollment

 Never

16,176

34

118

38

1.00

(Ref)

 Ever

31,095

66

191

62

1.11

(0.86, 1.43)

Education

 High school or less

30,893

61

239

71

1.00

(Ref)

 Greater than high school

20,027

39

99

29

0.73

(0.57, 0.93)

Ever farm machinery injuryc

 No

16,487

75

109

72

1.00

(Ref)

 Yes

5,580

25

43

28

1.16

(0.81, 1.67)

Difficulty with balancec

 Yearly or less

20,347

94

134

89

1.00

(Ref)

 Monthly or more

1,242

6

16

11

1.69

(1.00, 2.86)

Shaking or trembling of your handsc

 Yearly or less

20,088

93

132

90

1.00

(Ref)

 Monthly or more

1,416

7

15

10

1.53

(0.90, 2.62)

Depression

 Never

44,930

96

285

94

1.00

(Ref)

 Ever

1,703

4

19

6

1.55

(0.96, 2.51)

Previous injuryc

 No

16,378

75

109

72

1.00

(Ref)

 Yes

5,537

25

43

28

1.16

(0.81, 1.67)

High pesticide exposure eventc

 No

18,555

86

126

88

1.00

(Ref)

 Yes

3,110

14

18

13

0.95

(0.58, 1.57)

aHR hazard ratio, CI confidence interval

bHazards by state are adjusted for age only

cTake-home questionnaire variable; data available for approximately 40% of the cohort

Days/year of pesticide application but not years or cumulative days was significantly associated with fatal injury (60+ days/year application: hazard ratio (HR) = 1.87; 95% Confidence Interval (CI) = 1.10, 3.18; p trend = 0.02; Table 2). Adjusting for ever drinking slightly attenuated this estimate (HR = 1.67, 95% CI = 0.95, 2.94). Use of chemical-resistant gloves was associated with a significant reduction in risk (HR = 0.73; 95% CI = 0.58, 0.93), but adding glove use to models did not change risk estimates for the frequency of pesticide use (60+ days/year HR = 1.92, 95% CI = 1.13, 3.27; p trend = 0.04).
Table 2

General pesticide use characteristics and fatal injury among male farmers in the AHS, 1993–2008

Variable/risk factor

Entire cohort

(n = 51,035)

Fatal injuries

(n = 338)

Age and state adj.

p trend

N

%

N

%

HRa

95% CIa

Days/year mixed or applied pesticides

 <5 or 5–9

19,473

42

124

40

1.00

(Ref)

0.02

 10–19

14,080

30

92

30

1.16

(0.88, 1.53)

 20–39

9,087

19

58

19

1.24

(0.90, 1.71)

 40–59

2,384

5

17

6

1.54

(0.92, 2.59)

 60 or greater

1,877

4

16

5

1.87

(1.10, 3.18)

Years applied pesticides

 <1

4,803

9

36

11

1.00

(Ref)

0.79

 2–5

5,262

10

27

8

0.82

(0.49, 1.37)

 6–10

7,226

14

41

12

0.90

(0.57, 1.42)

 11–20

15,722

31

89

26

0.84

(0.56, 1.25)

 21–30

11,578

23

81

24

0.91

(0.60, 1.37)

 >30

6,444

13

64

19

0.97

(0.63, 1.48)

Cumulative days personally mixed/applied pesticides

 0–56

15,198

30

96

28

1.00

(Ref)

0.09

 57–200

10,931

21

67

20

0.87

(0.63, 1.21)

 201–396

11,910

23

69

20

0.92

(0.67, 1.27)

 ≥397

12,996

25

106

31

1.27

(0.96, 1.69)

Ever wear protective equipment when handling pesticides

 No

8,043

16

69

20

1.00

(Ref)

 

 Yes

42,985

84

269

80

0.78

(0.59, 1.03)

 

Chemically resistant gloves used when handling pesticides

 No

16,464

32

136

40

1.00

(Ref)

 

 Yes

34,564

68

202

60

0.73

(0.58, 0.93)

 

aHR hazard ratio, CI confidence interval

After adjustment for age and state, ever use of seven of 49 pesticides was associated with significantly increased risk of fatal injury; none was associated with decreased risk (Table 3). Significant HRs were seen for ever use of five of 18 herbicides: 2,4,5-T, alachlor, butylate, metribuzin, and paraquat. No insecticide was significantly associated with risk, although there was a suggestive association with ever use of coumaphos (HR = 1.39, 95% CI = 0.97, 1.99). Of four fumigants evaluated, risk was significantly increased for users of carbon tetrachloride/carbon disulfide. Ziram was the only fungicide of 6 evaluated that was significantly associated with fatal injury. Adjusting the model for use of chemical-resistant gloves did not change the findings with the exception of increased HRs for paraquat (HRparaquat|gloves = 1.56; 95% CI = 1.18, 2.05) and parathion (HRparathion|gloves = 1.36; 95% CI = 1.00, 1.84).
Table 3

Ever/never pesticide use and fatal injury among male farmers in the AHS, 1993–2008

Pesticide

Entire cohort (n = 51,035)

Fatal injuries (n = 338)

Age and state adj.

N

%

N

%

HRa

95% CIa

Herbicides

 2,4-D

37,218

76

259

80

1.24

(0.94, 1.64)

 2,4,5-T

9,782

22

99

35

1.41

(1.09, 1.83)

 2,4,5-TP

4,181

9

41

15

1.35

(0.96, 1.91)

 Alachlor

24,767

54

184

60

1.30

(1.03, 1.64)

 Atrazine

35,185

71

243

74

1.25

(0.97, 1.61)

 Butylate

14,472

32

107

37

1.29

(1.01, 1.65)

 Chlorimuron-ethyl

17,196

38

111

39

1.19

(0.93, 1.52)

 Cyanazine

19,129

42

138

45

1.19

(0.95, 1.50)

 Dicamba

22,952

50

148

49

1.02

(0.81, 1.28)

 EPTC

8,836

20

60

20

1.14

(0.85, 1.52)

 Glyphosate

37,555

76

238

73

0.91

(0.71, 1.16)

 Imazethapyr

19,705

43

125

41

1.12

(0.89, 1.42)

 Metolachlor

21,522

47

134

44

0.98

(0.78, 1.23)

 Metribuzin

20,566

46

144

50

1.29

(1.02, 1.64)

 Paraquat

11,161

25

91

32

1.35

(1.05, 1.74)

 Pendimethalin

20,937

46

131

46

1.07

(0.85, 1.36)

 Petroleum oil

21,311

48

144

50

1.14

(0.90, 1.44)

 Trifluralin

24,530

53

154

52

0.99

(0.79,1.25)

Insecticides

 Aldicarb

5,785

13

30

10

0.79

(0.53, 1.16)

 Aldrin

8,740

19

72

25

0.98

(0.73, 1.30)

 Carbaryl

25,949

56

170

57

0.88

(0.69, 1.11)

 Carbofuran

12,569

28

99

33

1.19

(0.93, 1.52)

 Chlordane

11,607

26

84

29

0.87

(0.66, 1.13)

 Chlorpyrifos

20,762

42

132

41

1.05

(0.84, 1.32)

 Coumaphos

3,822

9

35

12

1.39

(0.97, 1.99)

 DDT

11,835

26

106

37

1.00

(0.76, 1.33)

 Dichlorvos

4,527

10

29

10

1.02

(0.69, 1.50)

 Diazinon

14,326

32

91

32

0.94

(0.73, 1.21)

 Dieldrin

3,081

7

31

11

1.10

(0.75, 1.63)

 Fonofos

9,984

22

67

22

1.05

(0.79, 1.38)

 Heptachlor

7,018

16

68

24

1.17

(0.88, 1.57)

 Lindane

8,447

19

52

18

0.93

(0.69, 1.26)

 Malathion

32,513

70

210

71

0.96

(0.75, 1.24)

 Parathion

7,232

16

60

21

1.24

(0.93, 1.66)

 Permethrin on livestock

5,803

13

31

10

0.95

(0.65, 1.38)

 Permethrin on crops

6,242

14

28

9

0.69

(0.46, 1.03)

 Phorate

15,078

34

102

35

1.03

(0.81, 1.32)

 Terbufos

18,182

40

104

34

0.83

(0.65, 1.06)

 Toxaphene

6,695

15

50

18

0.94

(0.68, 1.29)

 Trichlorfon

316

1

2

1

Fumigants

 Aluminum phosphide

2,178

5

19

7

1.48

(0.91, 2.38)

 Carbon tetrachloride/carbon disulfide

2,482

6

31

11

1.54

(1.04, 2.28)

 Ethylene dibromide

1,708

4

11

4

0.96

(0.53, 1.76)

 Methyl bromide

8,048

16

54

17

0.92

(0.68, 1.25)

Fungicides

 Benomyl

4,840

11

36

12

1.04

(0.73, 1.48)

 Captan

4,940

11

30

10

0.96

(0.65, 1.40)

 Chlorothalonil

4,192

9

21

6

0.67

(0.42, 1.07)

 Maneb/mancozeb

4,531

10

34

12

1.07

(0.74, 1.53)

 Metalaxyl

11,174

24

76

26

1.08

(0.83, 1.40)

 Ziram

760

2

9

3

1.97

(1.01, 3.83)

aHR hazard ratio, CI confidence interval

In evaluating exposure–response for individual chemicals, we used frequency of use because this metric was associated with injury for the use of any pesticide (Table 2). For 17 of 18 herbicides, we evaluated exposure–response; 14 were based on four categories of frequency of use (Table 4), three by median split only. Significant trends were observed for 2,4-D, butylate, and cyanazine. For 2,4-D, alachlor, and cyanazine, the HR was significant in the highest tertile of use. For the three herbicides analyzed by median split, only chlorimuron-ethyl showed increased risk (HR>5 days/year = 1.59; 95% CI = 0.98, 2.60; p trend = 0.06). No insecticides showed an exposure–response relationship between days/year of application and fatal injury (data not shown). When we analyzed cumulative days of use for individual chemicals, only the trends for 2,4-D [HRhigh(>116 days) = 1.37; 95% CI = 1.01, 1.88; p trend = 0.04] and cyanazine persisted [HRhigh(>56 days) = 1.48; 95% CI = 1.05, 2.01; p trend = 0.05]; carbofuran was the sole insecticide that showed an exposure–response [HRhigh(>50.75 days) = 1.74; 95% CI = 1.22, 2.50; p trend = 0.01].
Table 4

Exposure–response associations for selected herbicides and fatal injury among male farmers in the AHS, 1993–2008

Herbicide (days/year)

Entire cohort

(n = 51,035)

Fatal injuries

(n = 338)

Age and state adj.

p trend

N

%

N

%

HRa

95% CIa

2,4-D

 0

12,025

25

65

21

1.00

(Ref)

0.01

 <5

15,015

31

92

29

1.02

(0.73, 1.41)

 5–9

11,114

23

71

22

1.08

(0.76, 1.53)

 ≥10

10,490

22

89

28

1.54

(1.11, 2.15)

Alachlor

 0

21,264

47

123

42

1.00

(Ref)

0.07

 <5

8,759

19

65

22

1.24

(0.91, 1.68)

 5–9

8,064

18

53

18

1.08

(0.77, 1.51)

 ≥10

7,075

16

53

18

1.41

(1.02, 1.96)

Atrazine

 0

14,255

29

84

26

1.00

(Ref)

0.18

 <5

13,581

28

91

28

1.14

(0.84, 1.56)

 5–9

11,284

23

82

26

1.24

(0.90, 1.72)

 ≥10

9,803

20

64

20

1.23

(0.88, 1.73)

Butylateb

 0

15,873

74

96

66

1.00

(Ref)

0.02

 <5

2,209

10

21

14

1.69

(1.05, 2.74)

 5–9

2,275

11

18

12

1.53

(0.92, 2.57)

 ≥10

1,102

5

10

7

1.66

(0.83, 3.32)

Cyanazine

 0

26,761

59

170

56

1.00

(Ref)

0.05

 <5

8,211

18

56

18

1.11

(0.80, 1.53)

 5–9

6,183

14

43

14

1.09

(0.76, 1.57)

 ≥10

4,343

10

36

12

1.54

(1.06, 2.24)

Dicamba

 0

22,673

50

157

52

1.00

(Ref)

0.27

 ≤2.5

11,566

26

67

22

0.85

(0.62, 1.16)

 >2.5–7

7,009

16

45

15

0.95

(0.66, 1.36)

 >7

3,819

8

33

11

1.42

(0.95, 2.12)

EPTC

 0

36,261

81

239

80

1.00

(Ref)

0.63

 <5

4,199

9

31

10

1.20

(0.82, 1.76)

 5–9

2,664

6

16

5

0.95

(0.56, 1.61)

 ≥10

1,733

4

11

4

1.16

(0.63, 2.14)

Glyphosate

 0

11,923

24

88

28

1.00

(Ref)

0.79

 <5

16,711

34

106

33

0.88

(0.66, 1.18)

 5–9

10,433

21

67

21

0.98

(0.71, 1.36)

 ≥10

9,868

20

57

18

0.92

(0.65, 1.30)

Imazethapyr

 0

25,706

57

177

59

1.00

(Ref)

0.35

 <5

8,985

20

57

19

1.03

(0.74, 1.44)

 5–9

6,853

15

43

14

1.10

(0.77, 1.59)

 ≥10

3,467

8

22

7

1.25

(0.78, 1.99)

Metolachlor

 0

24,381

54

170

57

1.00

(Ref)

0.89

 <5

7,566

17

49

16

0.95

(0.69, 1.32)

 5–9

7,251

16

43

14

0.91

(0.64, 1.28)

 ≥10

6,183

14

38

13

1.03

(0.72, 1.48)

Metribuzinb

 0

13,574

63

92

63

1.00

(Ref)

0.41

 ≤2.5

4,574

21

31

21

1.09

(0.71, 1.67)

 >2.5–7

2,490

12

15

10

1.04

(0.59, 1.83)

 >7

824

4

8

5

1.56

(0.71, 3.39)

Pendimethalinb

 0

13,585

63

89

61

1.00

(Ref)

0.46

 <5

3,961

18

32

22

1.35

(0.90, 2.03)

 5–9

2,603

12

15

10

1.06

(0.61, 1.84)

 ≥10

1,378

6

9

6

1.17

(0.56, 2.44)

Petroleum oilb

 0

16,991

79

112

77

1.00

(Ref)

0.14

 <5

2,073

10

13

9

0.96

(0.54, 1.70)

 5–9

1,219

6

8

6

1.16

(0.56, 2.39)

 ≥10

1,090

5

12

8

1.72

(0.92, 3.20)

Trifluralin

 0

21,467

47

145

49

1.00

(Ref)

0.49

 <5

8,601

19

49

17

0.82

(0.59, 1.15)

 5–9

8,890

20

52

18

0.89

(0.64, 1.24)

 ≥10

6,420

14

48

16

1.23

(0.88, 1.73)

aHR hazard ratio, CI confidence interval

bTake-home questionnaire variable; data available for approximately 40% of the cohort

Driving combines was associated with fatal injury, with an HR of 1.87 (95% CI = 1.07, 3.27) for 31 or more days per, p trend = 0.03 (Table 5). Adjusting for smoking and alcohol use did not substantially change results. Daily tractor driving was associated with increased risk, though not significantly so (HR = 1.41; 95% CI = 0.79, 2.52). While livestock production was not associated with fatal injury, butchering animals was (HR = 1.36; 95% CI = 1.01, 1.84). Poultry production showed a significant association with fatal injury, approximately doubling the risk of fatal injury regardless of flock size (HRs ranged from 1.78 to 2.26). These factors were unrelated to days/year of pesticide application and adjusting for driving combines, butchering animals, or poultry production did not alter pesticide findings (data not shown).
Table 5

Farm exposures and fatal injuries among male farmers in the AHS, 1993–2008

Variable/risk factor

Entire cohort

(n = 51,035)

Fatal injuries

(n = 338)

Age and state adj.

p trend

N

%

N

%

HRa

95% CIa

Drive combines (per growing season)b

 Never

4,368

20

28

19

1.00

(Ref)

0.03

 1–10 days

4,395

20

29

20

1.36

(0.79, 2.34)

 11–30 days

8,899

41

61

41

1.56

(0.93, 2.59)

 31 or more days

3,963

18

30

20

1.87

(1.07, 3.27)

Drive tractorsb

 Never/once/month

2,317

11

18

12

1.00

(Ref)

0.19

 Weekly

9,857

45

67

44

1.09

(0.63, 1.89)

 Daily

9,792

45

68

44

1.41

(0.79, 2.52)

Drive trucksb

 Never/once/month

10,575

49

64

42

1.00

(Ref)

0.30

 Weekly

5,855

27

51

34

1.40

(0.97, 2.03)

 Daily

5,277

24

37

24

1.18

(0.77, 1.82)

Butcher animals

 No

43,932

86

284

84

1.00

(Ref)

 

 Yes

7,103

14

54

16

1.36

(1.01, 1.84)

Acres planted last year

 None/did not work on a farm

3,510

8

24

8

1.00

(Ref)

0.30

 5–49

4,340

9

37

12

1.25

(0.74, 2.11)

 50–199

8,166

18

60

20

1.11

(0.67, 1.84)

 200–499

12,913

28

85

28

0.99

(0.59, 1.66)

 500–999

10,205

22

63

21

1.00

(0.59, 1.72)

 ≥1,000

6,742

15

32

11

0.84

(0.47, 1.51)

Poultry production

 None/did not work on a farm

40,246

90

242

85

1.00

(Ref)

0.0003

 <50

2,046

5

20

7

1.78

(1.13, 2.82)

 50–499

1,033

2

13

5

2.26

(1.29, 3.96)

 ≥500

1,169

3

11

4

1.97

(1.06, 3.66)

Livestock production

 None/did not work on a farm

21,193

47

145

50

1.00

(Ref)

0.76

 <50

3,750

8

26

9

1.00

(0.66, 1.53)

 50–499

13,874

31

86

29

0.97

(0.72, 1.31)

 ≥500

6,287

14

35

12

0.94

(0.63, 1.40)

aHR hazard ratio, CI confidence interval

bTake-home questionnaire variable; data available for approximately 40% of the cohort

Discussion

The rate of fatal injury within this cohort (46.3 injury deaths/100,000 person-years) was similar to that reported for farmers and ranchers nationally in 2008 (40.3 injury deaths/100,000 person-years) (Bureau of Labor Statistics 2010). Based on comparisons within the AHS cohort, we observed increased risk for 60+ days of pesticide application/year as well as increased risk associated with specific herbicides, in both ever use and exposure–response models. Conversely, we saw a decrease in risk with use of chemical-resistant gloves. We cannot assess whether this latter reduction is a result of an attenuation of pesticide exposures or differences in risk taking behavior. As expected, operation of machinery was associated with increased risk as shown by the association with combine use.

Risk for fatal injury increased with the number of days/year of any pesticide application. While herbicides were most strongly associated, the risk estimate for the highest category of applying any pesticide was higher than that of the highest category for any individual chemical. Typically, insecticides rather than herbicides are associated with neurotoxicity as shown previously (Gomes et al. 1998; Kamel et al. 2005; London et al. 1998; Pilkington et al. 2001), and neurotoxicity might predispose to higher rates of fatal injury. The increased risk of fatal injury associated with herbicides was unexpected—we know of no other reports of such associations. Rather than direct toxicity, this finding may instead suggest that some activity related to herbicide use is associated with fatal injury. Although this may be a chance finding, it was specific in that we found associations with several different herbicides, and the increased risk was not due to correlated use of other pesticides. Furthermore, not all herbicides, including some that are commonly used, were associated with fatal injury.

Observed associations with days/year rather than years or cumulative days of use, as well as the fact that for some chemicals risks were only seen in the highest exposure category, suggest that acute rather than chronic pathways may be operating. However, frequency of use may instead be a risk indicator: individuals in the highest category may also have more days of overall farm activities/year (e.g., tractor driving) or more overall complex work environments (e.g., use of multiple types of farm equipment) leading to more opportunities for injury; we did observe increased risk among people who drove combines routinely. While we recognize the complexity in assessing multiple simultaneous exposures, our pesticide results did not appear to be confounded by frequency of other farm activities.

Although they did not explain any of the pesticide associations, activities involving farm machinery and working with animals were also independently associated with fatal injury. Machinery is often responsible for fatal injury (Richardson et al. 1997; Solomon 2002; Voaklander et al. 1999). While tractors are often the primary cause for injury, we did not observe a significant association with reported patterns of tractor driving, perhaps due to limited variability in tractor usage within our cohort or lack of details about the specific injury (e.g., presence of a roll over protective structure). Also, several types of machinery used on farms were not enumerated at enrollment [e.g., ATVs (Thomas et al. 2010)], and it is not clear whether and to what extent these might be associated with fatal injury and pesticide use. We did, however, observe elevations in risk for use of combines and butchering animals. Neither association has, to our knowledge, been reported previously.

Use of chemical-resistant gloves was associated with decreased risk of fatal injury. This association was also seen in a study of non-fatal farm injuries conducted among 252 cases and 504 controls in Australia, where use of personal protective equipment was associated with a 21% decreased risk of non-fatal injury (Day et al. 2009). While this decreased risk may be the result of lower pesticide exposure, it could also identify individuals who are generally more cautious.

Fatal injuries may be unreported suicides (Kraus et al. 2005; Rockett et al. 2010). Although the rate of suicide in this cohort is lower than the general population, the rate of injuries is higher (Waggoner et al. 2011). Beard and colleagues saw no evidence of an association of pesticides with suicide in the AHS. Moreover, individual herbicides were generally inversely associated with suicide in the AHS, with a few significantly so (Beard et al. 2011).

Strengths of this study include its size and prospective nature; we followed more than 51,000 farmers over 13 years. This large study provides information complementary to other studies that are generally small and rely on medical examiner reports (Bernhardt and Langley 1999; Richardson et al. 1997). While we used death certificates for our outcome and thus did not have information about what participants were doing when the fatality occurred, this prospective cohort allowed identification of potential risk factors predating the injury that cannot be identified based on medical examiner reports. In contrast, studies based on medical examiner reports (which we did not have) tend to include more detailed data regarding the immediate circumstances of the fatality (for example, lack of a roll over protective structure on a tractor or blood alcohol concentration of the decedent). Lastly, given the number of analyses performed, there is a possibility of some chance findings.

We relied on self-reported pesticide use information provided by participants at the time of enrollment. Farmers (Blair and Zahm 1993), including those in the AHS (Blair et al. 2002; Hoppin et al. 2002), have been shown to provide accurate information regarding their pesticide use. Except for three applicators that died within a year following enrollment, we do not know which chemicals were used in the year immediately preceding death. We also relied on death certificates to determine our outcome, which are valid for broad categories of injury reporting (Moyer et al. 1989). We studied risk factors for all injuries grouped together. While it is possible that risk factors for subtypes of injury differ, we felt the intrinsic factors that put someone at a higher risk for injury would not be specific to one type over another and thus studied a general grouping. Similar groupings are seen in other injury mortality studies (Crandall et al. 1997; Cubbin et al. 2000; Richardson et al. 1997).

Risk factors for non-fatal injury have previously been studied in this cohort. Hearing problems were associated with significantly increased risk of non-fatal injuries from animals (Sprince et al. 2003a), machinery (Sprince et al. 2002), and falls (Sprince et al. 2003b). While we could not examine hearing problems in our present study, we did see increased risk associated with other age-related conditions, such as difficulty with balance and tremor.

In this large cohort of male farmers, pesticide use, in addition to farm machinery use, was associated with fatal injury. Given our lack of information about whether a pesticide was used 1 year or many years before fatal injury and the unexpected associations with herbicides, these findings should be considered cautiously.

Acknowledgments

The authors thank Stuart Long for assistance with data analysis. This work was supported by the Association of Schools of Public Health/Centers for Disease Control Fellowship program; the National Institute for Occupational Safety and Health and the Intramural Research Program of the National Institutes of Health, National Institute of Environmental Health Sciences (Z01-ES049030) and the National Institutes of Health National Cancer Institute (Z01-CP010119).

Conflict of interest

The authors declare that they have no conflict of interest.

Copyright information

© Springer-Verlag (outside the USA) 2012