Documentation of ill-health effects of occupational exposure to grain dust through sequential, coherent epidemiologic investigation.

S, BECKLAKEMR. Documentation of ill-health effects of occupational exposure to grain dust through sequential, coherent, epidemiologic investigation. Scand J Work Health 1994;20:13 21. This review describes the evolution in epidemiologic methods (study design, outcome, and expo sure measurements and target population) in relation to knowledge gained concerning the ill-health effects of exposure to grain dust in the workplace over the past three decades. The clinical conditions associated with grain dust exposure are briefly described. Thereafter the study approaches used to in vestigate grain-related conditions (clinical studies, workforce-based prevalence studies, prevalence studies includingunexposed workers, short-termresponseto exposure, longitudinalstudies,supporting evidence from other research arenas, and studies documenting dose-response relationships) are dis cussed. The objective is to illustrate the strength of sequential, coherent epidemiologic investigation into the ill effects of a particular work environmenton humanhealth.

usuall y foll ow ing exposure to high airb orne co ncentrat ions of du st a nd attributed to endotoxins or mycotoxi ns, (iii) extrinsic allergic alveol itis, due to expo sure to high levels of fungal spores, in particular Thermop olyspora polyspora. Among the chronic condition s attributed to grain du st exposure, chro nic bronchiti s and chro nic airflow limitati on are probably the most common and the most imp ort ant (2,3). Neither is distinct from the chronic bron chitis and chronic airflow limitation related to expo sures to other airborne pollutants, most notably smoking. Evidenc e that occupational expo sures to du sts of inorgani c or orga nic or igin are implica ted in the gene sis of both conditions has accumulated steadily over the past se veral decades (3)(4)(5), and the imp ortance of these conditions may have overtaken that of the traditional occup ational fibrotic lung diseases, such as the pneumoconiose s, in terms of numbers of individuals affected. In addition, in many work force s, the association between occupational exp osures and chronic obstruc tive airway disease is as str ong as the asso ciat ion with smoki ng (4), a finding which has emerged as a result of improved epidemiologic stud y des ign and more sophisticated multivariate analys is techniques to address the confounding effect of smoking with other airb orne exposures.
A worker who develop s an acute response to grain du st, in parti cul ar asthmatic-type re spon ses, is often unabl e to tolerate the work en vironment, and there is eviden ce that such individuals leave this type of employment within a short period aft er the y begin work. For instance a lower prevalence of atopy, as reflected by skin testing, has been sho wn for grain workers than for workers not exposed to grain dust 13 Scand J Work Env iron Health 1994. vol 20. no I (6)(7)(8)(9). This finding suggests that atopic individuals have selected them selve s out of the grain industry. An understanding of this phenomenon is important when the findings of research are interpreted, and it has bedeviled cross-sectional studies.

Research approaches used to investigate grain-related conditions
The distribution and determinants of respiratory condition s related to grain dust exposure and their significance has been examined over time with the use of clinic al and epidemiologic methods . The overall goal of such research has been to determine whether grain exposure is related to respiratory ill health, and, if so, to determine the nature , cause , and co-determinant s of adverse resp iratory effects in grain-exposed workers. In this paper we review the evolution in epidemiologic methods (study design, outcome, exposure measurements, and target population) in relation to knowledge gained concerning the ill-health effects of exposure to gra in dust in the workplace over the past three decades. The objective is to illustrate the strength of sequential, coherent epidemiologic investigation into the ill effect s of a particular work environment on human health . Many of the studies concern Canadian workers, Canada being one of the world's major wheat producing and exporting countries. Note that the papers we have cited , a selection of which is also listed in tables I through 3, have been chosen to illustrate the evolution of study methodology , and they do not con stitute a comprehensive review of all material , such as has been publish ed by Chari-Yeung et al (2).

Clinical studies
As is often the case, concern over health risks to local worker s is first expres sed in clinical report s of phy sicians serving communities in which a high proportion of the population is actively employed in a particular industry (l0). In the 1960s and 1970s such reports were published on Canadian prairie elevator worker s. The studies documented increa sed symptoms and impaired lung function and investi- Scand J Wark Environ Health J994, vol 20, no I Table2. Evolu tion of research approaches used to inve st ig at e the ill-health effects of g ra in dust exposure: cross-shift and crossw ee k changes. (F EV, .o = forced expiratory volume in 1 s)

Worksite
Subjec ts High prevalence of sym ptoms and abnormal lung funct ion in t hose lost to follow-up, demonstratio n of healt hy wor ker eff ect, increa sed abnormal lung function over the six years Annu al dec li ne greater in grain wo rkers than in unexposed subjects and reo lated to age, smoking and gated their relationship to immunologic mark~rs (10 ). They led to the involvement of the Canadian Thoracic Society in defining the existing state of knowledge and elaborating research priorities, including the importance of epidemiologic studies.

Cross-sectional studi es on work forces
Starting with the I940 s, cros s-sectio nal studies ha ve reported on the prevalence of respiratory symptoms and lung function level of exposed workers; these studies have been well summarized in a recent review (2). Since 1980, surveyors ha ve, for the most part , used the standardized respiratory sy mptom que stionnaire of the American Thoracic Society ( I I) and adh ered to the standards established by the various workshops on standardized lung function monitoring for screening purpose s (12,13). The imp ortance of the standardized methodology is that it allows comparisons to be made across time, within a study, and between studies on comparable work forces.
Thu s in a 1977 cross-sectional survey of 300 workers who handl ed grain in a North American lake port , 77 % of the workers rep orted eye symptoms , 64 % co mplained of nasal symptoms, and 88% had one or more respiratory symptoms (14 ). These symptoms were independent of age and length of empl oyment. Rate s tended to be higher among tho se with one or more positi ve skin reaction to common allergen s, and the differences were sig nificant for wheezin g. These findings lead to the hypoth esi s that the atopic worker is at risk of grain-related respirator y illness. Howev er in this cros s-sectio nal study no une xpo sed population was investigated . Subsequently, in 1980, the issue of the interaction of smoking and grain dust exposure was addressed in a study of prai rie elevator work ers (3), for who m a decrea se in lung function was found in smoking workers, and the decrease was greater than that in either smokers or grai n workers separate ly. It was co ncluded that gr ain workers who smoke are more likely to dev elop resp iratory insult than nonsmoking grain work er s or smokers not exp osed to grain dust. From the se and othe r studies, it wa s becoming clear that grain worker s were exper iencing high rates of resp irator y abn ormality. Howe ver the relation ship of these symptoms to work was not clear. Specificall y, the int eraction of atopy and smoki ng with grain ex pos ure in the generatio n of re spiratory symptoms and lung fun ction level needed to be addressed.

Cross-sectiona l studies including unexposed workers
Sin ce 1980 , most rep orted cro ss-sectional studies have included an unexposed co mpa riso n group (table I). Such co mpariso ns address the que stion of exposure-respon se relati on ships (ie, any ver su s no~x posure), a key element in establishing ca usality beyond associ ation . The se studies al.so all owed for the comparison of atopy pre valence In exposed and 16 unexposed work forc es. For insta nce, a study ( 15) of 103 grai n handlers in the port of Montreal and 39 unexposed dock workers showed no rel ationship between respiratory effects and atopic status in contrast to the findings of an earlier lake port study ( 16) . In ano ther study of lake port workers (9) a higher pre val ence of cou gh , sputu m, dyspne a, and whee~ ing was found in the grain-ex posed workers than 1U th e unexposed workers. In addition, althoug h the lung functi on level s and the prevalence of allergic reactions were higher among the unexposed than among the grain-ex pos ed workers, the re was a greater prevalence of nonspecific bronch ial reacti vity among the grain worke rs than among the une xpo sed population. Thi s finding sugg ested that bronchial hyperr eactivity in some grain workers may not be. allergic in etiology, and it became clear that respiratory chan ges wer e not confined to atopi c workers.
Another consequence of the introduction of an une xposed co mpariso n group as a feature of cro sssec tio nal studies on grain workers ca rried out si nce 1980 was to pro vide evid en ce in supp ort of the stro ngly held clini cal impression that active work forces are esse ntia lly surv ivor populations (7,8). For instance, it was note d that asthm atic subjects tended to withdraw from work in prairie grain studies and that turn over was rap id among workers in their fir st two years of employ ment (3,17). Similarl y atopy was found to be less fre quent in a study of west coast port elevator worke rs than in a nongrai n-ex pose d group of civic workers (7 ). By the mid-I 980s there was thu s substantial evidence for a "h ealth y worker effect" on work forces exposed to grain du st, and it became clear that the nature and frequency of the adv erse health effects of grain dust exposure had been under estimated by the cross -sectio nal design .
Given the dominant rol e of smoking as a determinant of respiratory health sta tus, an ea rly co nce rn of researcher s was to establis h whether grain du st exposure had adve rse res piratory effects indepen dent of tob acco exposure. One approach was to stratify by smoking statu s in the analysis. For instan.ce , in the 1980 port of Montreal study (15) , co mpan ng dock wo rke rs with grain handl er s, a high er prevalence of couzh and whe ezing was only fou nd for the grain hand ler s who smoked, but a decrease in lun g function was evi de nt whe ther or not they smo ked. Anot her app roach was to contro l for the co nfounding effects of smoking by selectin g for the study only those who did not hav e exposure to the co nfoundina fact or , in this case lifetime nonsmoker s. Thu s in b .
. . . a co mpa rison of lifet ime nonsm oking pram e gram wor kers with lifetime non smok ing subjects selected from the same co mmunities, a fiv e tim es greater pre va lence of chro nic bro nchitis and a three times greater prevalence of wheezi ng was found for the grain-exposed workers than for those not exposed ( 18). They also had a lower forced mid-expiratory flow rate and flow rate at 50 % of vital capaci ty (18) . These findings strengthened the evidence of the abil-ity of grain dust to cause respiratory changes in its own right. Other studies yielded evidence of a synergistic relationship between smoking and grain dust exposure (3). For instance, in one study (17) of workers exposed in prairie elevators, there was no difference in the prevalence of respiratory symptoms when these workers were compared with unexposed workers, but a difference in lung function was found for smoking grain workers , a finding suggesting some interaction between smoking and grain dust exposure in the pathogenesis of abnormal lung functions. In a second study of the same working population, matched groups of nonsmokers and smokers, grain exposed and unexposed, were compared (19). An increase in symptoms and decreased lung function were found in the grain workers and in the smokers, but the effects appeared to be additive rather than synergistic. Thus these study approaches allowed researchers to investigate the independent effects and interactive effects of smoking and grain exposure on respiratory health. These findings are in agreement with those of other studies that have demonstrated an effect of grain-dust exposure on respiratory health independently of smoking status. Subsequently, as multivariate methods of analysis evolved, they were used to explore these relationships. For instance, in a study of workers exposed to grain in lake port elevators and unexposed civic workers in the Lake Superior area; the prevalence of chronic bronchitis was significantly higher for grainexposed workers than for unexposed workers (20) . Dyspnea, chest tightness, and wheezing were also found to be more common among the grain-exposed workers. Chronic bronchitis and wheezing were related to both smoking history and grain dust exposure . The odds of grain exposed workers ' exhibiting the symptoms of chronic bronchitis and wheezing were 4.4 and 4.8, respectively, irrespective of smoking status, while the odds of smokers' exhibiting the symptoms of chronic bronchitis or wheezing was 2.9 and 1.9, respectively, irrespective of grain exposure. Thus multivariate analysis was able to provide a numerical description of the effect of grain exposure and smoking on respiratory symptoms . In this study no relationship between lung function level and length of employment was demonstrated, and therefore the researchers interpreted the result as evidence of a "healthy worker effect." Thus, from these cross-sectional studies , evidence was forthcoming to suggest that grain-dust exposure, independent of smoking, was related to respiratory symptoms and change in lung function.

Short-term responses to exposure
As part of the thrust to establish exposure-response relationships, acute changes in lung function over a week or a shift were studied in several work forces, often as part of cross-sectional prevalence studies, with or without unexposed comparison groups, com-Scand J Work Environ Health 1994. vol 20. no I ing either from within the same work force or from other work forces selected because they were not exposed to grain or, in some instances, any obvious workplace pollution (table 2).
For instance, in a study of three Vancouver work forces (port grain elevator workers, saw mill workers, and civic workers) (7) , one of the first studies to measure short-term changes in lung function, forced vital capacity (FVC) and forced expiratory volume in I s (PEV1.o) decreased over I d, as well as over one workweek, whereas in the other two work forces these lung function parameters increased over the same period. A subsequent study (21) on Canadian lake port workers who were examined on Monday, Wednesday, and Friday found that FVC and FEVl.o decreased on Wednesday when compared with the corresponding values on Monday , but there was no decrease in lung function in the afternoon compared with the morning measurement. The results were interpreted as providing some evidence for the reversibility of lung function changes related to grain exposure. Likewise studies in lake port grain workers in the United States (22) showed cross-shift decreases in lung function, whereas civic workers not exposed showed increases in lung function over the workshift , a finding independent of age or smoking status. In a subsequent study these cross-shift changes in lung function were shown to be dose-related to exposure levels when workers were stratified according to whether exposure levels were greater or less than 5 mg . m? (23). Studies on Canadian port elevator workers (24) also showed doserelated changes in lung function over a shift.
Changes in lung function over the workweek have also been reported for grain workers from other parts of the world. For instance, in a South African study of grain mill workers and factory workers at sea level, lung functions for both the grain-exposed and nongrain-exposed populations were similar (25). However, the grain workers demonstrated a decrease in FEV1.0and forced mid-expiratory flow rate, while the factory workers demonstrated an increase in these parameters over the workweek. A 10% decrease in FEVI.I) was seen in 23% of the grain-exposed population and in only 9% of the unexposed population. Neither atopic status nor smoking status was related to this change in lung function of the study population over the week. In another South African study (26), conducted in the Johannesburg area [approximately 6000 feet (1829 m) above sea level], a decline in lung function over the workweek was seen in those workers exposed to high concentrations of grain dust as assessed by the measurement of airborne grain dust in milligrams per cubic meter in comparison with a rise in function over the workweek in those in the low-exposure category. An Australian study (27) reported a decrease in lung function measured at the beginning of the employment season compared with the end of the season. The change was significantly associated with work-re-lated breathlessness but not with skin reactivity to allergens.
In summary, several studies in different work forces in different parts of the world have established that grain-dust exposure is associated with short-term decreases in lung function measured over the workshift, workweek, or work season. There is also evidence to suggest an exposure-response relationship between cross-shift or cross-week changes and subsequent chronic airflow limitation. In a review of occupational asthma (28) the measurement of crossshift change in lung function was considered good evidence of the work relatedness of this respiratory change . However the relationship of short-term lung function changes to the long-term health of grain workers can only be elucidated by longitudinal studies.

Longitudinal studies
The strength of longitudinal studies (table 3) of work forces lies in their ability to clarify time relationships; their weakness is the result of the difficulty of achieving satisfactory follow-up rates . One of the earlier longitudinal studies on grain workers was that of Warren & Manfreda (29), who reported a six-year follow-up of 68 Manitoba grain elevator workers. In this study , 19% of the original work force had changed jobs, and these workers were found to have a higher prevalence of symptoms, as well as abnormal lung function. In those workers who were retested after the six-year period, more were found to have abnormal lung function findings (FEV,/FVC = <70%). Although only a small group of workers was followed , the findings are nonetheless interesting and have subsequently been confirmed by other reports, in particular a coherent series of interrelated papers based on research with port elevator workers in the Vancouver area, conducted by a research team headed by Chan-Yeung et al (30-32). A two-year follow-up of grain-exposed workers and reference civic workers showed that respiratory symptoms increased over the two years in the grain-exposed workers, but not in the civic workers, and the annual decline in FEV 1.0 was also greater in the grain workers. In addition, the annual decline in FEV1.0 was found to be significantly related to the cross-shift and cross -week changes in lung function recorded in the earlier study. These findings suggest that acute changes in lung function predict or lead to irreversible airflow limitation. In 1984, a six-year followup of these workers (31) confirmed that the change in lung function over one week was a significant predictor of subsequent annual decline in FEV u)" Neither atopic status nor the initial lung function level was related to this annual decline in FEV 1.0' Nor, surprisingly, was smoking status. Methacholine challenge, used to define bronchial hyperreactivity, was also found to be a significant predictor of annual decline in FEV 1.0' In a further analysis of these data, using a nested case-referent design within the follow-18 up cohort (32), a rapid decline in FEV 1.0 (defined as a decrease of greater than 100 ml in FEV l. o per year) was found to be related to the level of grain-dust exposure, and a greater than 4% decrease in FEV10 over the workweek and bronchial hyperreactivity were significant predictors of accelerated annual lung function decline , These effects were demonstrable for exposure levels greater than 4 mg , m", In contrast to these findings, a four-year followup study of lake port workers (33) found no difference in the prevalence of symptoms between the two surveys and no difference in the lung function level. However not exactly the same subjects were retested. There was also a higher prevalence of eye irritation, shortness of breath , and cough, but no difference in the lung function level, in those who had left compared with those who had remained in the same employment. These findings were interpreted, correctly in our view , as evidence of a "healthy worker effect." The authors warned researchers that in longitudinal studies the "healthy worker effect" is likely to attenuate the exposure-response relationship in studies of grain workers. The fact that a selected healthy group of workers is available for long-term follow-up was recently underlined; the authors went on to question the usefulness of longitudinal studies, given their expense and difficulty, in assessing exposure-related health hazards since this relationship has been demonstrable with the use of cross-sectional data (34).
There is consistency in the findings of the followup studies indicating that acute cross-shift changes in lung function are of clinical importance and also in the evidence suggesting that cross-shift changes predict long-term irreversible obstructive airway disease (35).

Supporting eviden ce from other research arenas
In the nonoccupational setting, attention has also been given to early markers of chronic obstructive airway disease . For instance, in one analysis (36), a decrease in FEV 1.0 was the most important risk factor, other than age and cigarette smoking, for the future development of chronic obstructive airway disease . In addition, a reduction of FEV 1.0' as well as other measures of ventilatory function, has been shown to be associated with increa sed mortality and morbidity (37). Again, in one study (38) of the bronchodilator response of patients with chronic obstructive airway disease, the greater the degree of airway responsiveness to bronchodilators, the more rapid the decline in both FVC and FEVl.oover time. Other researchers (39) have reported that a low FEV,o predicts a more rapid decline in FEV, oand that this predictor is more reliable if the change in FEV,o is taken into account. Thus the finding that short-term change in FEV1.0 predicts long-term decline in lung function is not an isolated finding for workers exposed to grain dust, nor is it confined to one study design or one method of defining acute response to grain-dust expo sure. The demon stration of cross-shift changes in lung function become s more significant in lizht of the findings of the longitudinal studies on both grain workers and other populations in which early indicators of obstructive lung disease have been studied.
Stud ies docum enting exposure-response relationships Dose-response relation ships are a key element in establishing the harmful effects of exposure to an agent in the fields of pharmacology and toxicology. Dose in these circum stances is defined as the amount of agent delivered to, and remaining in con-t~ct with, the target organ for a sufficient length of time to evoke an effect. In occupational epidemiology, expo sure is the only practi cal substitute for dose , and expo sure-response relat ionships (with exposure characterized categorically as any versus none) are used to establish the occurrence of adverse he~lth effects, while exposure-response relationships (With exposure characterized as a continuous or orinal variable) provide information useful for publie he~lth pur~o ses , in particular for determining appropnate environmental control level s. The preceding discussion indicates the consistency with which adverse respiratory effects have been demon strated with the use of categorical exposure indices (ie, any versus no exposure) (3, 6-8, 15-23, 25). Several studies provided some evidence from quant itative environmental measurements of the exposure levels associated with detectable adverse respiratory effects in grain workers (23, 26, 32). Researcher s have also used duration of employment as a proxy for exposure (40). In addition, clear relationships between exposure, assessed subjectively by the worker, and both symptoms and lung function level have been demon strated in two studie s of workers in different South African grain mills; these findings point to the usefulness of qualitati ve exposure measurem ents for this purpose (25, 41). However, only recently has a s y~te mati c study of exposure-response relation ships, usmg quantitative industri al hygiene measurements of exposure, been reported (42). The study, incorporating inform ation from the extensive data on Va~cou ver grain elevator workers, used 781 personal air samples representing 20 job titles over the 15year study period to construct a job-time matrix for average dust exposure. This study provide s some of the strongest evidence to date of chronic adverse effect s of expo sure on lung function , exposure-associated effects having been demonstrable at levels as low as 4 mg . m", Confirmation in other work forces would furth er strengthen the evidence for a causl a s s~ciation . In addition, the first planned study to investigate survival in the industry from the time of recruitment was recently publi shed by Zejda et al (43). A.ft~r four years only one-third of the subjects who originally underwent an initial employment examination were still at work, and the short-term (an-Scand J Work Environ Health 1994, vol 20, no 1 nual) lung function loss was negatively related to the number of years a worker remained in the industry. Thus those who left after one year had an annual FEV 1 0 loss of function of 224 ml, whereas for those who left after three years the corresponding value was 71 ml.

Synthesis
An appre ciation of the relation ship of grain-dust exposure and respiratory health has developed through the ration al application of epidemiologic methods. Initially cross-sectional studie s indicated an apparent increase in respiratory symptoms and other health indicators among grain-exposed workers. This finding was confirmed by the use of more sophisticated study approaches in which exposed workers were compared with unexpo sed individuals, either from within the same work force or from other unexposed work forces, and exposure-response re!ationsh ips were established by contrasting those With any exposure versus those with no exposure. The demon stration of acute cross-shift changes in lun? function supported the evidence for grain-related Ill-heal th effects. Further evidence of expo sureresponse relationships was obtained in longitudinal studies which demonstrated a relationship between acute cross-shift or cross-week changes in FEV and long-t erm lung function loss. Supportive evi' denc~from nongrain~related research added weight to thiS argument. Finally, the docum entation of e xp o~ure-re sponse relationships for lung function de~clt, first using subjective (qualitative) exposure estimates and subsequently quantitative indu strial hygiene measurements, compl etes the documentation generall y required to establi sh associations as causal. T~dat e,~o use has been made of experimental studies, for Instance, in the form of a randomized control trial, to asse ss the effect of an acute intervention, such as decreasing expo sure in a given workplace over a short period, such as a week or a month. To achieve this situation, the cooperation of employers in conjunction with organized workers would be needed. While the evidence to date strongly suggests that efforts to control grain-dust exposure should be undertaken , a demon stration study of thẽ x~ected associated health benefits is now clearl y indicated , rather than further documentation of the problem. Moreover, positive results from a shortterm interv ention study would provide additional ammunition for the argument for control levels below 4 mg . m", the level suggested on the basis of the best available data to date (42). 19.