Consistent with previous mortality findings from this cohort (1
), we observed lower overall mortality rates in participants in the AHS than in the general populations of North Carolina and Iowa. The AHS is larger than most agricultural cohorts, and this analysis not only provides new information on rare causes of death but is also more recent than most of the previously published studies. Similar to what was seen in other studies (2
), applicators and spouses in our study showed a mortality experience reflective of a healthy lifestyle. Despite this, applicators had increased rates of death due to unintentional fatal injuries compared with the general population. Additionally, after adjusting for the lower overall mortality rate of the cohort, applicators experienced higher rates of mortality from lymphohematopoietic cancers, melanoma, and malignancies of the digestive system, prostate, kidney, brain, thyroid, eye, and ovary, as well as from cardiomyopathy. Spouses experienced higher rates of death from lymphohematopoietic cancers and malignancies of the digestive system, breast, ovary, and brain after adjustment for the lower overall mortality rate.
The increased risk of unintentional fatal injuries among applicators is consistent with what has been seen in other studies (7
) but was not observed previously in this cohort (1
), possibly because there were fewer deaths. Farmers face significant occupational hazards from machines, as highlighted by the observed SMRs, which exceeded 2 for machine injury, motor vehicle nontraffic accidents, and collisions with objects. Deaths from machine accidents and motor vehicle nontraffic accidents may be related, as tractor deaths can be classified as either, depending on whether they occur on or off the roadway (15
). Combining deaths from these 2 categories, we observed an annual rate of 6.9 machine-related deaths per 100,000 person-years (which includes deaths from tractor rollovers), which is approximately 25% higher than the 2007 US annual average of 5.5 tractor-overturn deaths per 100,000 person-years in the agricultural population, as calculated from a report by the National Institute for Occupational Safety and Health (22
). Animals are a known risk factor for fatal and nonfatal farm injury (23
); however, we did not see an increased rate of animal-related fatalities. Although Iowa is ranked seventh in the United States for cattle and calf production, poultry and hogs are the most common livestock raised in Iowa and North Carolina (25
). Therefore, our failure to observe an elevated animal-related fatality rate may be due to a lower level of large animal production than in other agricultural regions.
Consistent with our previous mortality analysis (1
), we continued to observe lower rates for several major causes of death, including cardiovascular disease, all cancers, lung cancer, and diabetes. The all-cause mortality rate was approximately half that expected, similar to the previous mortality study and other studies of farmers (1
). The higher physical activity level inherent to farming likely plays a significant protective role in these deficits. The deficit in smoking-related deaths (e.g., chronic obstructive pulmonary disease and lung cancer) is reflective of a higher prevalence of never smokers (53% of applicators and 72% of spouses, versus 46% of US adults ≥45 years of age in 1995 (26
)). Although these deficits may reflect a healthy lifestyle, they are also consistent with the healthy-worker effect that is typically observed in working populations.
To account for the lower overall mortality rate in the cohort, we used an rSMR analysis that compared the SMR for each cause with the SMRs for all other causes. These rSMRs were meant to be exploratory and to be interpreted in relation to the mortality rates from all other causes of death in the cohort. In doing so, we identified an unusually high relative mortality rate from lymphohematopoietic cancers, digestive cancers, and cancers of the prostate, breast, brain, and ovary that were not apparent in the SMR analysis. We did not observe any increased rSMRs for noncancer internal causes, with the exception of cardiomyopathy among applicators.
We observed increased rSMRs for lymphohematopoietic cancers for applicators and spouses, consistent with meta-analyses among farmers (27
). Svec et al. (30
) noted that an increased risk of lymphohematopoietic cancers was associated with occupational exposure to animals, which occurred primarily in agricultural settings, though this effect could have been confounded by pesticides. Previous findings from the AHS have implicated several individual pesticides as being associated with these cancers (31
Farmers are at a higher risk of prostate cancer (32
), which is possibly associated with specific pesticides (33
). Chlorinated pesticides and methyl bromide were significantly associated with increased risk among applicators >50 years of age (35
). Increased prostate cancer mortality rates have been seen in some (5
) but not all (7
) other farming populations. We observed an elevated mortality ratio for prostate cancer only in the rSMR analysis.
An analysis of cancer incidence in the AHS through 2002 identified an increased incidence of ovarian cancer among the 1,563 female applicators but not among the 32,127 female spouses (37
). We observed increased rates of ovarian cancer death among applicators and spouses relative to all other causes. The larger rSMR for applicators compared with spouses is likely a chance occurrence, reflecting the small number of female applicators in the cohort. We also observed increased breast cancer relative mortality among spouses in the cohort. Although risk was not significantly increased among applicators, rSMRs were similar (spouses, 1.58; applicators, 1.76) and few applicators were female.
We conducted our analyses separately for applicators and spouses, as applicators are expected to have greater occupational exposures to pesticides and other farm risk factors than their spouses. Applicators are licensed to apply restricted-use pesticides; however, 58% of spouses in Iowa and 45% in North Carolina reported having applied at least 1 pesticide at enrollment (38
). Both men and women were represented among the applicators, with 1,563 female applicators contributing 20,886 person-years to the analysis, and thus providing a unique occupationally exposed group in which we could examine female-related cancers.
The cohort includes both private and commercial pesticide applicators; private applicators (mainly farmers) dominate the cohort, but both groups use similar pesticides. Commercial applicators accounted for only 205 of the 4,880 deaths observed for applicators; thus, their inclusion in the study population had little influence on the overall findings. Additionally, when analyzed separately, the mortality rates for both groups were similar: We observed that both farmers and commercial applicators experienced significantly fewer deaths than expected overall and had elevated cancer risk in the rSMR analyses. Although an Australian study (39
) focusing on cancer mortality and incidence rates among pest-control operators found rates similar to those in the general population, Fleming et al. (5
) noted SMRs <1.0 for most causes of death among commercial applicators in Florida.
SMR analyses are useful to assess the disease experience of a population relative to a general population; however, this strategy has some inherent limitations. Although we gain statistical stability by comparing data from our cohort with state and national rates, the ability to control for confounding is limited to factors reported on the death certificates. Those factors do not address the healthy-worker effect. We used the rSMR analysis, which allowed us to adjust for the overall disease experience in our study population. Both the SMR and rSMR analyses have limited ability to evaluate exposures that may contribute to elevated or lowered mortality rates. We relied on death certificates for our outcome measure. The overall validity of death certificates tends to be fairly high and is expected to be comparable across underlying causes of deaths occurring within a state (40
). By comparing the rates within states, we have comparable reporting for the cohort and population rates. We used the Life Table Analysis System.net
program to calculate SMRs; this program groups deaths to reflect occupational hazards, as illustrated by the fine strata for falls and the peculiar combinations of cancers (e.g., colon and small intestine grouped as “intestine”). All SMR analyses face limitations with regard to rare diseases. Our study benefited from its large sample size and high number of person-years of follow-up, which have more than doubled since the previous analysis (1
In summary, our analysis of 6,419 deaths occurring from 1993 to 2007 among 89,656 pesticide applicators and spouses showed that applicators were at an elevated risk of death from machine injury. The cohort experienced a lower mortality rate overall when compared with the general population. After adjusting for the lower mortality of the cohort, we observed relatively higher rates of death among applicators from lymphohematopoietic cancers, melanoma, and malignancies of the digestive system, prostate, kidney, brain, thyroid, eye, and ovary. Among spouses, we observed higher adjusted mortality rates for lymphohematopoietic cancers and malignancies of the digestive system, breast, ovary, and brain. Extended follow-up of this cohort will provide valuable information through the accumulation of deaths from rare diseases as the cohort ages.