Eastman, Chelsea PhD, MPH; Schenker, Marc B. MD, MPH; Mitchell, Diane C. PhD; Tancredi, Daniel J. PhD; Bennett, Deborah H. PhD; Mitloehner, Frank M. PhD
Dairy work has been associated with an increased prevalence of respiratory symptoms, including chronic bronchitis, wheeze, allergies, hypersensitivity pneumonitis, and organic dust toxic syndrome.1–7 Dairy farming also has been associated with an accelerated decline in the ratio between forced expiratory volume in 1 second (FEV1) and forced vital capacity (FVC) and with moderate airway obstruction.1–3,5 Sufficient exposure to organic particles, specifically endotoxins, may result in dose-dependent organic dust toxic syndrome. The endotoxin content of respirable dust has been associated with decreased FEV1.8,9
Although previous studies have documented the association between dairy work and decrements in respiratory health, the situation in California is distinctive. In 2007, California's dairies had an average herd size of 952 cows.10 By contrast, the second and third highest-producing dairy states Wisconsin and New York had average herd sizes of 88 and 108 cows, respectively.10 European dairies had even smaller herds, ranging in size from an average of 35 cows (Belgium) to 78 cows (the United Kingdom).11 In addition to substantially larger herd sizes, California's dairy facilities have naturally ventilated barns or open corrals in contrast to enclosed dairy barns in Europe and the eastern United States.10,11 We aimed to determine whether the much larger herd size but open-air ventilation of the facilities in California resulted in lower or higher concentrations of air pollutants relative to previous studies, and whether there were any health concerns due to possibly different types of air pollutants that might be generated on these western US facilities. Given this unique arrangement, we studied whether dairy workers in California have lower baseline and greater cross-shift decrements in lung function than control employees.
MATERIALS AND METHODS
Study Subjects and Design
Male participants from 13 dairy facilities and a control facility in California completed questionnaires and spirometry before and after the work shifts from June to September 2008. The control facility (a vegetable-processing plant) was selected on the basis of a similar worker population (eg, ethnicity, income, and education) with a night shift. Study design, methods, and survey protocols have been described previously.12
All pulmonary function tests (PFTs) were performed by National Institute of Occupational Safety and Health–trained and certified spirometry technicians using EasyOne® spirometers (Medical Technologies, Inc, Andover, MA). Questionable spirograms were reviewed by a pulmonary physician with extensive experience in treating adverse health effects of occupational respiratory exposures.
Two hundred twenty-six dairy workers and 49 control employees completed all study protocols. Of these, 16 dairy workers (7.1%) and two control employees (4.1%) were excluded from PFT analyses because of incomplete or unusable spirograms. The remaining 257 participants were divided into categories on the basis of the variability of the two best maneuvers. The first category followed American Thoracic Society (ATS) recommendations (less than 5% variability).13 The second, more highly variable (5% to 8% variability category) was created to test whether we could include participants who had less reproducible spirograms, perhaps because they were sick or had less formal education. The spirometry values themselves are less reproducible, but it's also the participants performing spirometry and so in a way they are less reproducible themselves. Sicker patients have a harder time producing reproducible spirograms.14 Participants with more than 8% variability were excluded from analyses. Each PFT measure was considered separately (ie, a participant could be included in analyses for FEV1 but if his FVC measures had higher than 8% variability, then he was excluded from FVC analyses). Categorization of ratio measures from participants was based on whichever component in the ratio displayed a greater variability; for example, if a participant had more than 8% variability for FEV1 but less than 5% variability for FVC, the FEV1/FVC ratio would be categorized as having more than 8% variability (ie, unusable). To estimate baseline forced expiratory flow between 25% and 75% of FVC (FEF25%–75%), all the most highly variable spirograms were excluded using the categorization of FEV1/FVC. Baseline and cross-shift changes in FEV1, forced expiratory volume in 6 seconds (FEV6), FVC, FEF25%–75%, FEV1/FEV6, and FEV1/FVC were modeled separately as outcome measures in regression models. Cross-shift changes were calculated by subtracting the measure before the shift measure from the measure after the shift.
To account for possible clustering by facility, we used a mixed-effects multiple linear regression modeling strategy. To facilitate comparisons among outcomes, a common set of covariates was specified for baseline and cross-shift. Model selection for each set began with a core set of covariates. For baseline models, the core set included age, height, and smoking status. Smoking status was defined as a dichotomous variable (current vs former/never). A small number of candidate covariates also were evaluated for addition to the core model, including pack-years smoking, years spent in agriculture, years spent in the United States, number of days worked since the last day off, time of the test before the work shift, history of use of personal protective measures, history of time spent working in dust, income, and education level. In cross-shift models, in addition to the core variable of smoking status, the following candidate covariates were considered: whether the participant had smoked any cigarettes that day, time spent in dust that day, time of the work shift, and length of the work shift. Candidate covariates were screened individually with Akaike's information criterion to assess whether the models for most of the outcomes (at either baseline or cross-shift) were improved.15 Finalization of a parsimonious set of common covariates was achieved by removing those covariates that were highly collinear with other screened-in covariates.
In the cross-shift analyses, we dichotomized the start time of the work shift to adjust for the diurnal variation of lung function.16,17 For the dichotomous shift-time variable, 12 PM was used as the cut-point on the basis of the control workers' shift-start times. Because of varying work-shift lengths, we also considered whether the length of the work shift impacted lung function. The dichotomous start-time variable improved the cross-shift models, whereas the length of the work shift was not associated with PFT values and did not improve Akaike's information criterion. Regression models with only the core variable and those including the shift-time variable showed a similar effect; because the shift-time variable improved the model fit (based on Akaike's information criterion), it was included in the final cross-shift models. For cross-shift outcomes, the common set of covariates consisted of the core variable (smoking status) and shift time.
Additional analyses that were restricted to dairy workers explored whether mean levels of select cross-shift outcomes were associated with the presence of self-reported asthma symptoms. Finally, we performed a sensitivity analysis restricted to participants whose spirograms met ATS criteria13 to determine whether more stringent quality-control criteria would result in substantively different conclusions from analyses including participants with more variable outcomes (less than 8% variability between the two best maneuvers). All statistical analyses were performed using the Statistical Analysis System Software version 9.2 (SAS Institute Inc, Cary, NC).18
The study population was young, with an average age in the mid-30s for both dairy and control workers. Education levels among both groups were low; fewer than half of the dairy workers attended school past grade 6. Dairy workers earned significantly more than control employees (P < 0.001). The vast majority of the participants were Latino, with more than 90% having emigrated from Mexico. Current smoking levels were similar, and the mean of pack-years was low and not significantly different between the two populations (Table 1).
Baseline Pulmonary Function
For baseline outcomes, the covariate for days back at work since the last day off was included with the core baseline covariates (age, height, and smoking status). The final adjusted baseline model approached statistical significance and was suggestive of a chronic association between dairy work and FEV1 (P = 0.07) and FEV6 (P = 0.053) (Table 2). Nevertheless, there was not an association of dairy work and the baseline ratios FEV1/FEV6 or FEV1/FVC (P = 0.85 and 0.64, respectively).
Cross-Shift Change in Pulmonary Function
Dairy workers experienced a greater cross-shift decrement than controls in both markers of obstruction (FEV1) and restriction (FEV6 and FVC) (P = 0.02, 0.002, and 0.002, respectively) after adjusting for smoking status and shift time. The difference between dairy workers and control employees was not statistically significant for FEV1/FEV6 or FEV1/FVC (P = 0.61 and 0.42, respectively) (Table 3).
Asthma and Asthma-Like Symptoms
To test whether symptomatic workers were more likely to experience a cross-shift decrement, we modeled the change in FEV1 and FVC in a regression with the presence of asthma-like symptoms as a dichotomous predictor. Asthma-like symptoms include coughing, wheezing, feeling tightness in the chest, or feeling short of breath when exercising, working, or being exposed to cold air, animals, cows/livestock, dust, pollen, cigarette smoke, grain dust, or dairy chemicals. Only dairy workers were included in these models because complete information about asthma-like symptoms was not collected from among the control employees. Among dairy workers, the presence of asthma-like symptoms was associated with a mean difference in cross-shift FEV1 of −93.2 mL (95% confidence interval, −130.4 to −35.9 mL; P < 0.001) compared with no asthma-like symptoms, after adjusting for smoking status and shift time (Fig. 1). For cross-shift FVC, the presence of asthma-like symptoms was associated with a mean difference of −81.3 mL (95% confidence interval: −140.2 to −22.5 mL; P = 0.007) after adjusting for smoking status and shift time (Fig. 1).
Sensitivity Analysis Comparing Results Using ATS Criteria With Main Results
The analyses reported earlier include outcomes from participants with more variable maneuvers than allowed by the strict application of ATS criteria. When the models were restricted to the sample meeting ATS criteria, the estimates varied slightly from the main analyses for all measures, but the direction and statistical significance were identical between the group with only the “best” spirograms and the group with both the “best” spirograms and the more variable participants (results not shown).
Acute Change in Pulmonary Function
The cross-shift decline in FEV1 (P = 0.02) suggests an association of exposure with airway obstruction among dairy workers. In addition, with a significant cross-shift decline in FEV6 (P = 0.002) and FVC (P = 0.002), there seems to be a mild acute restrictive function change among dairy workers. Both percent predicted FEV1 and percent predicted FVC are key indicators of occupational asthma.19 Additional markers of airway obstruction (ie, the ratios of FEV1/FVC and FEV1/FEV6) were not significantly associated with dairy work, indicating that FEV1 did not show a greater cross-shift decrease than FVC (or FEV6) and pointing to a mixed airways effect.
Industrial bronchitis (also known as acute reversible airway inflammation) has been defined as cough, phlegm production, and mild airflow obstruction associated with occupational exposure to dust or fumes.20 About 20% of dairy (45 of 210; 21.3%) and control (9 of 49; 19.2%) workers experienced either cough or phlegm production during their work shifts. In contrast, few dairy (7 of 210; 3.3%) and control (1 of 49; 2.1%) participants experienced both cough and phlegm during their work shifts. Forced vital capacity can be reduced in industrial bronchitis,20 in keeping with our findings of reduced FVC and FEV1. Furthermore, industrial bronchitis primarily affects the large airways, leading to reversible obstruction.20 Although industrial bronchitis does not exactly describe our findings, the inconclusive findings of baseline FEV1 and FVC with respect to dairy work as well as mild mixed effects of acute restriction and acute obstruction point to a condition similar to industrial bronchitis and merit further investigation. Specifically, it would be useful to measure inflammatory toxicity of dairy dust, as we have done for other agricultural dusts,21 and to study acute markers of inflammation among workers.
Given that one fifth of the workers in this study had never accessed medical care in the United States,12 we used a more sensitive indicator of asthma, “asthma-like symptoms,” instead of doctor-diagnosed asthma. Dairy workers with asthma-like symptoms had significantly greater cross-shift decrements in FEV1 than dairy workers without asthma-like symptoms, indicating that they are a susceptible population. Because they represent about 40% (80 of 210) of the dairy participants, intervention efforts should target this susceptible population. Nevertheless, asthma-like symptoms include several nonspecific symptoms (including shortness of breath and coughing) that could be due to multiple causes.22 Unfortunately, we cannot assess causality of asthma-like symptoms and pulmonary function in a cross-sectional study. Either way, dairy workers experiencing asthma-like symptoms merit particular attention in further investigations to disentangle the lung function and symptoms.
Association of Acute and Chronic Pulmonary Function in Dairy Workers
Because baseline values of FEV1 and FEV6 are marginally associated with dairy work (P = 0.07 and 0.053, respectively), the association of dairy work and chronic measures of pulmonary function remains unclear. Longitudinal studies in France have shown mixed long-term effects of dairy work; it is associated with a moderate decrease in FEV1 that is heightened with more time spent on dairies and among older male subjects.3,5 Studies among swine farmers, coal miners, and cotton textile workers have shown that an acute change in pulmonary function leads to a long-term decrement.23–25 A 3% decrement in cross-shift lung function has been used previously as epidemiologically relevant to predict long-term effects and recommend exposure limits.9,26 Although the unadjusted mean levels of cross-shift change in FEV1, FEV6, and FVC among dairy workers in this study were −1.3%, −1.3%, and −0.9%, respectively, almost one third (62 of 210; 30%) of dairy workers experienced a 3% or more decrement in FEV1, indicating that they are at risk of developing a chronic deficit in FEV1 compared with 15% (7 of 49) of control employees (Fig. 2). Almost one third (61 of 210; 29%) of dairy workers and 13% (6 of 49) of control employees experienced a decrement of 3% or more in FVC (Fig. 3).
Studies among swine farmers indicate that long-term effects become apparent after 6 to 10 years of exposure.8,9,27 Because half of the dairy workers in this study have spent 5 years or fewer on dairies, they might not have been exposed long enough to show chronic conditions. Nevertheless, a study of Danish swine farmers showed an accelerated decline in FEV1 compared with dairy farmers; the dairy farmers' decline in FEV1 was close to the expected values,28 suggesting that there might not be a long-term effect of dairy work on FEV1. In contrast, a longitudinal French study found an association between dairy farming and moderate bronchial airway obstruction.1 We were unable to include both years working on a dairy and age in the same model because they are collinear, making it difficult to draw conclusions concerning the long-term effect of dairy work. Taking a subsample of only nonsmoking participants older than the median age (32 years) to assess long-term respiratory effects yielded results similar to those of the whole population (results not shown). Dairy workers in California likely do not show clear chronic respiratory effects because the population is too young to display a cumulative impact of working on dairies. In addition, previous studies have been conducted in Europe, where animals are confined in an enclosed building, unlike in California, where there are open, outdoor facilities, which results in lower exposures.29
The young Latino study subjects were light and infrequent smokers, consistent with published reports.30,31 Among current smokers, the median number of cigarettes per day was three and one among dairy and control workers, respectively. We explored several options of smoking variables to be included in the final models, including a five-level variable that was a combination of smoking status and pack-years. The final models use a dichotomous smoking variable (current vs former/never) because none of the explored options provided additional explanatory power. None of the smoking variables explored (including the final dichotomous variable) were significant in either the baseline or cross-shift models. It is likely that this young population simply has not smoked enough cigarettes long enough to show a significant respiratory effect.
Differences between dairy workers and control employees in income and smoking prevalence could introduce bias when making comparisons. Nevertheless, the difference would likely not be sufficient to create a meaningful difference in socioeconomic levels that might affect health practices, particularly because ethnicity and education levels were similar between dairy and control workers. In addition, smoking was not a significant predictor of lung function independent of employment as discussed previously.
Selection bias might be introduced by the healthy worker effect. Individuals who are unduly affected by dairy work could self-select out of the profession and thus would not be included in our sample, biasing our estimates toward the null. Nevertheless, because the dairy and control workers are similar in age and ethnicity, it is less likely that the healthy worker effect affects our results.
Additional studies of pulmonary function are necessary to identify the long-term respiratory health consequences of working in large dairies in California. Asthma and asthma-like symptoms merit further attention because almost half of the dairy workers (40%) in this study reported experiencing asthma-like symptoms. In this population with low education and limited access to health care, doctor-diagnosed asthma has questionable value. For this reason, we used “asthma-like symptoms” to encompass a broad range of nonspecific symptoms that could result from various exposures. A larger sample size would allow researchers to more completely characterize work tasks around a Californian dairy. Furthermore, a longitudinal study would provide the information needed to determine temporality of respiratory symptoms, including asthma.
Dairy work was associated with a mixed airways effect across the work shift, with reduction in measures of both airflow and the vital capacity of the lung. Because baseline values were marginally significant, the effect seems to be reversible in this population. Nevertheless, because acute changes in pulmonary function may predict long-term effects, the chronic effects of dairy work in California remain unclear. Given their nonmigratory nature, a longitudinal study of dairy workers in California is feasible and would elucidate the chronic effects of dairy work on lung health.
Asthma and asthma-like symptoms should be further researched among dairy workers in California because a large proportion of our dairy participants (40%) reported asthma-like symptoms. Future investigations should expand and separate asthma-like symptoms because the current definition includes several nonspecific symptoms.
Tracey Armitage was an invaluable resource for data cleaning and SAS coding. This research would not have been possible without the cooperation of the dairy owners as well as the dairy workers who agreed to participate. The authors thank the field staff (Johnny Garcia, Rona Silva, Gloria Andrade, Rebeca Gallo, Stephanie Harrington, Victor Hernandez, and Francesca Perrone) who worked tirelessly to collect data.
1. Chaudemanche H, Monnet E, Westeel V, et al. Respiratory status in dairy farmers in France: cross sectional and longitudinal analyses. Occup Environ Med. 2003;60:858–863.
2. Dalphin JC, Dubiez A, Monnet E, et al. Prevalence of asthma and respiratory symptoms in dairy farmers in the French province of the Doubs. Am J Respir Crit Care Med. 1998;158:1493–1498.
3. Dalphin JC, Maheu MF, Dussaucy A, et al. Six year longitudinal study of respiratory function in dairy farmers in the Doubs province. Eur Respir J. 1998;11:1287–1293.
4. do Pico G. Lung (agricultural/rural). Otolaryngol Head Neck Surg. 1996;114:212–216.
5. Gainet M, Thaon I, Westeel V, et al. Twelve-year longitudinal study of respiratory status in dairy farmers. Eur Respir J. 2007;30:97–103.
6. Gomez MI, Hwang SA, Lin S, Stark AD, May JJ, Hallman EM. Prevalence and predictors of respiratory symptoms among New York farmers and farm residents. Am J Ind Med. 2004;46:42–54.
7. Kronqvist M, Johansson E, Pershagen G, Johansson SG, van Hage-Hamsten M. Risk factors associated with asthma and rhinoconjunctivitis among Swedish farmers. Allergy. 1999;54:1142–1149.
8. Donham K, Haglind P, Peterson Y, Rylander R, Belin L. Environmental and health studies of farm workers in Swedish swine confinement buildings. Br J Ind Med. 1989;46:31–37.
9. Donham KJ, Reynolds SJ, Whitten P, Merchant JA, Burmeister L, Popendorf WJ. Respiratory dysfunction in swine production facility workers: dose–response relationships of environmental exposures and pulmonary function. Am J Ind Med. 1995;27:405–418.
10. California Department of Food and Agriculture. California Dairy Statistics Annual Review 2007. Sacramento, CA: California Department of Food and Agriculture; 2008.
11. van Berkum S, Helming J. European Dairy Policy in the Years to Come: Impact of Quota Abolition on the Dairy Sector. The Hague, the Netherlands: Agricultural Economics Research Institute; 2006.
12. Eastman C, Mitchell DC, Bennett DH, Tancredi DJ, Mitloehner FM, Schenker MB. Respiratory symptoms of California's dairy workers. Field Actions Sci Rep. 2010;2(Spec Issue):1–6.
13. American Thoracic Society. Standardization of spirometry, 1994 update. Am J Respir Crit Care Med. 1995;152:1107–1136.
14. Enright P, Vollmer W, Lamprecht B, et al. Quality of spirometry tests performed by 9893 adults in 14 countries: the BOLD study. Respir Med. 2011;105:1507–1515.
15. Burnham KP, Anderson DR. Model Selection and Multimodel Inference: A Practical Information-Theoretic Approach. New York, NY: Springer-Verlag; 2002.
16. Borsboom GJ, van Pelt W, van Houwelingen HC, van Vianen BG, Schouten JP, Quanjer PH. Diurnal variation in lung function in subgroups from two Dutch populations: consequences for longitudinal analysis. Am J Respir Crit Care Med. 1999;159:1163–1171.
17. Guberan E, Williams MK, Walford J, Smith MM. Circadian variation of F.E.V. in shift workers. Br J Ind Med. 1969;26:121–125.
18. SAS Institute Inc. SAS 9.2 Enhanced Logging Facilities. Cary, NC: SAS Institute Inc; 2008.
19. Francis H, Prys-Picard C, Fishwick D, et al. Defining and investigating occupational asthma: a consensus approach. Occup Environ Med. 2007;64:361–365.
20. Morgan WK. Industrial bronchitis. Br J Ind Med. 1978;35:285–291.
21. Vallyathan V, Pack D, Leonard S, Lawson R, Schenker M, Castranova V. Comparative in vitro toxicity of grape- and citrus-farm dusts. J Toxicol Environ Health. 2007;70:95–106.
22. Schenker M. Respiratory health hazards in agriculture. Am J Respir Crit Care Med. 1998;158:S1–S76.
23. Kirychuk SP, Senthilselvan A, Dosman JA, et al. Predictors of longitudinal changes in pulmonary function among swine confinement workers. Can Respir J. 1998;5:472–478.
24. Wang ML, Avashia BH, Petsonk EL. Interpreting periodic lung function tests in individuals: the relationship between 1- to 5-year and long-term FEV1 changes. Chest. 2006;130:493–499.
25. Wang X, Zhang HX, Sun BX, et al. Cross-shift airway responses and long-term decline in FEV1 in cotton textile workers. Am J Respir Crit Care Med. 2008;177:316–320.
26. Donham KJ, Cumro D, Reynolds SJ, Merchant JA. Dose–response relationships between occupational aerosol exposures and cross-shift declines of lung function in poultry workers: recommendations for exposure limits. J Occup Environ Med. 2000;42:260–269.
27. Reynolds SJ, Donham KJ, Whitten P, Merchant JA, Burmeister LF, Popendorf WJ. Longitudinal evaluation of dose–response relationships for environmental exposures and pulmonary function in swine production workers. Am J Ind Med. 1996;29:33–40.
28. Iversen M, Dahl R. Working in swine-confinement buildings causes an accelerated decline in FEV1: a 7-yr follow-up of Danish farmers. Eur Respir J. 2000;16:404–408.
29. Garcia JG, Bennett DH, Tancredi DJ, et al. Occupational exposure to particulate matter and endotoxin for California dairy workers. Int J Hyg Environ Health. 2012, May 12. [Epub ahead of print]
30. Perez-Stable EJ, Ramirez A, Villareal R, et al. Cigarette smoking behavior among US Latino men and women from different countries of origin. Am J Public Health. 2001;91:1424–1430.
31. Centers for Disease Control and Prevention. Hispanic/Latino Adult Tobacco Survey Guide. Atlanta, GA: Centers for Disease Control and Prevention; 2010.