Changes in ventilatory function in grain processing and animal feed workers in relation to exposure to organic dust.

OBJECTIVES- The effects of organic dust exposure on the configuration of the maximum expiratory flow volume (MEFV) curve was examined multivariately. M~;TH()DS - Data from 390 male workers in the animal feed industry in The Netherlands were ana lyzed. A multivariate technique called nonlinear canonical correlation analyses was used to study the relationship between a set of organic dust exposure variables and a set of ventilatory function vari ables.RESULTS- The results indicate an almost independent effect of the overall mean organic dust expo sure and the number of years of organic dust exposure on ventilatory function. Increasing mean or ganic dust exposure was associated with a decrease in both forced expiratory volume in I s (FEV 10) and forced vital capacity (FVC) and decreased flows at high lung volumes only. Increasing number of years of dust exposure was associated with a decrease in FEV, 0 and a decrease in flow at all lung volumes, while the FVC seemed relatively constant. These two distinct patterns of reduction in venti latory function may represent two different pathological processes. Whereas workers with prolonged exposure showed reduced values for all of the MEFV curve variables, except the FVC, those with only a few years of exposure especially showed a decrease in FVC and peak expiratory flow. The effect of current organic dust exposure was more evident for nonsmokers than for ex-smokers and current smokers. CONCLUSION- The major finding of this study was an independent effect of overall mean organic dust exposure and the number of years of organic dust exposure on the MEFV curve.

TIELEMANS E, HEEDERIK 0 , van PELT W. Changes in ventilatory function in grain processing and animal feed workers in relation to exposure to organic dust. Scand J Work Environ Health 1994;20:435-43.
OBJECTIVES-The effects of organic dust exposure on the configuration of the maximum expiratory flow volume (MEFV) curve was examined multivariately. M~;TH()DS -Data from 390 male workers in the animal feed industry in The Netherlands were analyzed. A multivariate technique called nonlinear canonical correlation analyses was used to study the relationship between a set of organic dust exposure variables and a set of ventilatory function variables. RESULTS -The results indicate an almost independent effect of the overall mean organic dust exposure and the number of years of organic dust exposure on ventilatory function . Increasing mean organic dust exposure was associated with a decrease in both forced expiratory volume in I s (FEV10) and forced vital capacity (FVC) and decreased flows at high lung volumes only. Increasing number of years of dust exposure was associated with a decrease in FEV,0 and a decrease in flow at all lung volumes, while the FVC seemed relatively constant. These two distinct patterns of reduction in ventilatory function may represent two different pathological processes. Whereas workers with prolonged exposure showed reduced values for all of the MEFV curve variables, except the FVC, those with only a few years of exposure especially showed a decrease in FVC and peak expiratory flow. The effect of current organic dust exposure was more evident for nonsmokers than for ex-smokers and current smokers. CONCLUSION-The major finding of this study was an independent effec t of overall mean organic dust exposure and the number of years of organic dust exposure on the MEFV curve.
Studies of grain e lev ato r workers (1--4) and workers in th e co tto n ind ustry (5) have sho wn that occ upati onal e xposu re to organic du st is associated wi th the dev el opment of respiratory sy mptoms and airw ay ob struction. Thi s finding has recently bee n confirmed by severa l cross-s ectiona l st udies in the grain pr oce ssin g and animal fee d ind ustry (6,7). Th e etio logy of lun g di sease in re lation to exp osure to orga nic du st is hardl y under sto od . Th er e is some expe rimental evidence th at endotoxins m ay pl ay an important role in the eti ology of the acute lung function ch an ge s rel ated to organic du st exp osure (8). A rel at ion bet ween endotox in exposure and chro nic lun g disease was found in two epidemiolog ic stud-(chro nic) effects of organic du st ex posure on ve ntilator y function.
This pa per re po rts results fro m a cross-sectio na l study of 390 workers in the Dutch an imal feed indu stry. W e studied the relations hips bet ween ex posure to organic dust and ventilatory fun cti on usin g max imum ex piratory fl ow volume (M EFV) curves. In most epidemio log ic studies it is co mmo n to exami ne ve ntilato ry fun ct ion variables se par ate ly . A fe w stu d ies (9-10) have de scribed the use of the pri ncipal component anal ysis to investigate the main di rect ions of variability (pr inci pa l compo ne nts) of a whole se t of ventilatory functio n varia bles . However , th is multi variate techniqu e is hampered by th e fact tha t the principal co mpo ne nts are not necessa rily focused on the so urces of variability in the MEFV curv e . In the present study a mul tivariat e technique ca lle d nonlinear cano nical co rre lation ana lyses was used to study th e rel ation ship betwe en a set of orga nic du st ex posure variables and a set of ventila tory func tio n variab les . Th is technique ex trac ts info rmation fro m th e MEFV cur ve in relation to or ganic du st exp osure and sum ma rizes it in (two) newl y created independe nt va riables. Th is procedure a llows a parsimo nious description of patt ern s of abnorma lity in ven ti latory fun ct ion in rel ation to organ ic du st exposure (ie, the co nfig uration of the MEFV curv e) .
The multivariate analysis facilitated a more comprehen sive description of rel ationsh ips between several indices of org anic dust ex pos ure and ven tilatory fun cti on as a whole, and a compa rison of these effect s among non smokers, ex-smokers, and current smokers.

Subjects
We studie d a popul ation of Caucasian male workers in the Dutch anim al feed industry. Thr ee hundred and fiftee n wor kers ca me from a cross-sectional study ca rried out bet ween 1986 and 1988 in two animal feed eleva tors where grai n is processed. Details of this study have been re ported previously (6, II ). Addi tional data on 78 workers were obta ined from a follow-u p study carried out four years later in the same animal feed ele vators, the result bei ng a study popul ation of 393 animal feed workers. Of the 393 anima l feed wo rkers, 390 were used in the analyses becau se three had missi ng values for the exposure variab les, Th e characteristics of the study popu lation are give n in table 1. Lung f unction and expos ure Th e ventilatory fu nctio n was charac terized by maximum expirato ry flow volume (ME FV) curves that were obtained from all 393 subjec ts by normal spiro metry with a Vicatest 5 spirometer (Mijnhardt, Bunnik). Me asurements and proce dures inclu ding BTPS adjustm ent s for body temperature and pressure (saturated with water vapor) and procedures of data selectio n, were in acco rdance wit h the sta nda rds of the Europea n Com munity for Coal and Steel (12) ; they have bee n described in detail elsew here (6). Th e varia bles derived fro m the MEFV curve were forced vital capacity (FVC), forced expiratory volume in I s (FEV1.0); maximum mid-expiratory flow (MM EF); peak expiratory flow (PEF) ; and max imum expi ratory flow rates at 75, 50 and 25% of the FVC (MEF 75 , MEF 50 , and MEF 25 , respectively). For each animal fee d worker, age-and height-specific reference values (predicted values) of the MEFV curve vari ables were ca lculated from reference equ ations of the European Community for Coa l and Steel for adult males ( 12) (table 2). Th e resi duals were co mputed as fo llows: MEFV curve variables "measured" -MEFV curve variables "predic ted." They were then used in the multivari ate analys is. In eig ht rand oml y se lected faci lities , 8-h personal inspirable du st sam ples were take n amo ng the produ ction wor kers . Detailed infor mation on the exposure assessment strategy of orga nic dust has been des crib ed in earlier papers (6,13). For all group members a complete occ upational history was available. These histories allowed calc ulations of the followi ng exposure variables : cumulative org anic dust ex posure (CUMDUST) , years of orga nic dust exposure (EX PYEARS) , and overall long-term mean organic dust exposure (MEAN-DUST), whic h is CUMDUS TIEX PYEA R (ta ble 2). For co mpari ng the gro ups of wo rkers in the multiva riate analysis of var iance (see table 4 in the Re-  suits section) and in the graphic presentations in figures 2 and 4 (see the Results section for figures 2 and 4), the workers were divided into groups according to their exposure. First they were divided into two groups, one with present organic dust exposure levels <5 mg . rrr-' (N =230) and one with levels of >5 mg/m' (N = 160). The workers were also divided into the following five groups according to their years of organic dust exposure (EXPYEAR): unexposed workers (N =76), 0.5-5 years exposed (N =65),5.5-10 years exposed (N =88), 10.5-20 years exposed (N = 93), >20 years exposed (N = 68). A worker was categorized as a current smoker if he currently smoked or had stopped within one year prior to the study.

Statistical analysis
The nonlinear canonical correlation analysis was used to determine relationships between the set of age-and height-adjusted MEFV curve variables and the set of exposure variables. This technique extracts information on the adjusted MEFV curve variables in relation to organic dust exposure and summarizes this information in one or more newly created independent variables. Such variables are called canonical variables, and they are linear combinations of the original adjusted MEFV curve variables. In our analyses only two canonical variables were computed. The nonlinear canonical correlation analysis can, to a certain extent, cope with nonlinear relationships so that more information about the relationships between the two sets can be summarized with fewer canonical variables (14). Only the exposure variables were allowed to be nonlinearly transformed. (See figure 3 in the Results section.) The scaling method chosen allows for monotone, piecewise second-order spline transformations. The nonlinear transformations were chosen to be smooth except for a possible jump at the first knot, which separated the unexposed and exposed workers. The knot placement is given in table 2. The determination of both optimal transformations and the construction of canonical variables was done with SAS software (procedures TRANS REG and CANCORR) (15).
To present results of the nonlinear canonical correlation analysis graphically, vector diagrams were made in which the components of both sets of variables are depicted with respect to the canonical variables of the MEFV curve set. The canonical variables form a referential frame of perpendicular axes in which the variables are indicated by vectors. The length of a vector is a measure of the correlation between the variable and the canonical variables. The fraction of variance of a variable "explained" by its correlation with the two canonical variables is equal to the sum of the squares of these two correlations (ie, equal to the squared length of the corresponding vectors) (figure I and table 3).
The numerical value of a worker on the first or second canonical variable is called a (canonical) An interaction between exposure to organic dust and smoking habits was formally tested with a multivariate analysis of variance (MANOV A) with all age-and height-adjusted MEFV curve variables as dependent variables and smoking habits, present organic dust exposure, and years of organic dust exposure (EXPYEAR) as independent variables. (See table 4 in the Results section.) The independent variables were categorized as described earlier. The interaction terms between the current organic dust exposure and smoking habits, current organic dust exposure and EXPYEAR, and EXPYEAR and smoking habits were added to the model. Figure I shows the vector diagram depicting the relationship of the organic dust exposure variables with the variables defining the MEFV curve. The first and second canonical correlations are 0.41 and 0.31, respectively. This result indicates that organic dust exposure relates to a fraction of the total variability in lung function. Note that the normal variability of the MEFV-curve variables has been adjusted by age and height with the use of reference equations.

Results
The vector diagram in figure I shows that the vectors for the MEFV curve variables and EXPYEAR point in opposite directions. This finding indicates that all of the MEFV curve variables are negatively correlated (ie, decrease with increasing years of organic dust exposure). Both flows at high lung volumes (PEF, MEF 75 ) and flows at low lung volumes (MEF 50 , MEF z5 ' MMEF) decrease with increasing years of exposure. The FVC was hardly affected, as the vectors for the FVC and EXPYEAR are almost perpendicular and the length of the FVC vector is very small. The latter occurrence is important because the squared length of a vector is equal to its "explained" variance, which is the variance accounted for by the two canonical variables of the MEFVcurve set. As FEVI.O was negatively correlated with EXPYEAR and FVC was not, the FEVl.o:FVC ratio decreased with increasing years of organic dust exposure. Ingeneral one can say that when workers are   MEANDUST and EXPYEAR, occupied an intermediate position in the vector configuratio n between MEANDUST on one hand and EXPYEAR on the other. Figure 2 depicts the vectors of the MEFV curve variables and MEANDUST and EXPYEARS, as well as centroids of five EXPYEAR categor ies. The centroids suggest two different lung function effects with increasing years of exposure. This result can be evaluated further if the centroids are projected upon the vectors of the MEFV curve variables. When a line is drawn from the first category of EXP YEAR (0 years of exposure ) to the second category (0.5-5 years of exposure), the line is more or less parallel to the vectors of FVC and PEF. This result suggests that exposu re during the first years of employment in this industry is associated particularly with a decrease in FVC and PEF. Workers who are exposed for a longer time, however, show a reduction for all MEFV curve variables, except FVC, since the line drawn from the second and following categories of EXPYEAR is almost perpendicular to the vector of the FVC and, to a less degree, to the vector of the PEF. The line points more clearly in the same direction as that of the other MEFV curve variables.
We have also investigated EXPYEAR with respect to the MEFV curve variables by further dividi ng the five EXPYEAR groups into those for nonsm okers, curre nt smokers, and ex-smok ers (not show n). For these groups similar result s were found , as dep icted in figure 2. Th is vector diagram suggests an almost independent effe ct of MEANDUST and EXPYEAR on the ventilatory fun ction with respect to the ME FV-c urve co nfig uratio n. Th is finding is in accordance with the vector configuratio n in figure 1.
The tran sformations of the exposure variab les MEANDUST and EXPYEAR are given in figure 3. Th e vector diagr am in figure 4 is the sa me as shown in figure 2. In addition, centroids of the two categories of current low and high exposure to organic dust are projected for the nonsmokers, exsmokers , and current smokers onto the canonic al variables of the MEFV curve set. In all of the groups of smokers, flows at high lung vo lumes (PEF, MEF 75 ) and the FVC decreased with increasing current exposure to organic dust, while flows at low lung volumes (MEF so ' MMEF, MEF 2S ) and, to a less extent , the FEV 1.0 were not affected. The ce ntroids show that current organic dust ex posure is more strongly related to decreases in lung function among nonsmokers since the line which is drawn bet ween the centroi ds of current low and high expos ure is longer for the non sm okers than for curre nt or exsmokers . Such a possible interaction effect of current organic dust expos ure and smoking habit s with respect to the ventilatory functio n was tested in a mul tivariate analysis of variance (MANOVA) (table 4). Bec ause EXPYEAR was not equaIIy distributed among the three smoki ng categories , an interaction term bet ween curr ent organic dust exposure and EXPYE AR was added to the model. Onl y the intera ction term bet ween current organic dust expo- have been projected onto the canonical variables of the MEFV curve set (ie, CAN1 and CAN2, the first and second canonical variable, respectively) (number of subjects =390). (centroids: 0 =unexposed worke rs, 5 =0.5-5 years exposed , 10 = 5.5 -10 years exposed, 20 = 10.5-20 years exposed , >20 = more than 20 years exposed; exposure variables : MEAN- It can be dedu ced fro m figure 4 th at smoking has an adverse effec t on lung fun ct ion tha t more or less coincid es w ith th e effec t ca used by EX PYEAR. Th is smoking effec t was not confounded by the dis tri bution of EX PY EA R in the smo ki ng ca tegories, as the mean val ue for EXP YEAR was high est in the intermediate ex-smokers' ca tegory. T his findi ng was suppo rted by the MA NOV A res ults in ta ble 4, whic h shows sig nifica nt effects of b oth smoki ng and EXPYEA R, while th e inte rac tio n term betwee n smoking habits and E XP YEAR was not sig nificant.

Discussion
The main objective o f thi s study was to examine the ME FV curve rep resenting ve nti latory funct io n with res pec t to a set of d ifferent ex po sure variab les for or ganic dust. General knowledge about changes in the configuration of the MEFV curve in association wit h organic dust ex posure can add some insight into corresponding pa tho logical processes. An examina-  (EXPYEAR), independent of normal aging, are indicative of chro nic effects . An important finding is that MEANDUST and EXPYEAR are related more or less independently to the MEFY curve (figures I, 2, and 4). Evidence for two indepe nden t pathological mecha nisms lead ing to acut e and chro nic lung disease was also foun d by Kenned y et al (5) . Flows at the begi nning of the MEFY curve were negatively correlated (decrease) with MEANDU ST, while flows more toward s the end of the curve were not. Conseq uently FEY1. 0 was only slightly negatively correla ted with MEAN DUST , and MMEF did 44\ tion of the MEFY curve was accomplished with the nonlin ear cano nica l correlation analysis. This multivariate technique extracts those aspects of the MEFY curve which correlate maxima lly with organic dust exposure and summarizes the MEFY curve essent ials in two new independent variables. In other words, available information of the whol e MEFY curve (seven flow and volume variables) was extracted and redu ced to two new variables that are more strongly rela ted to lung pathology induced by exposure to organi c dust.
All of the expo sure variables analyze d were negative ly cor related with age-and height-adjusted MEFY curve variables. This findi ng confirms previou sly published evide nce that organic dust exposure in the grain processing and animal feed industries can induce airway obstruction. Changes in the MEFY curve in asso ciation with the mean organi c dust exposure (MEANDUST) probably refl ects, to a large extent , short-term lung effects, as these cha nges co incide almost comp letely with those assoc iated with current orga nic dust exposure (figure 4). On the other hand, changes in the MEFY curve in relation to years of orga nic dust exposure not correlate with it at all. This finding suggests that short-term effects related to MEANDUST probably involves a narrowing of central airways more than a narrowing of peripheral airways, while loss of elastic recoil of the parenchyma is not likely, as the shape at the end of the MEFV curve does not change.
The flows at all of the lung volumes, as well as the FEVl.o and MMEF, decreased with increasing EXPYEAR, while the FVC seemed hardly affected. These effects on the whole shape of the MEFV curve suggest obstruction involving both central and peripheral airways. Recent studies have shown that organic dust exposure affects mainly the airways and probably not the lung parenchyma. Chronic lung diseases such as chronic bronchitis and chronic obstructive airway diseases related to organic dust exposure have been described (16). The chronic effects are (partially) irreversible (17) in contrast with the more reversible acute effects (18)(19)(20). In two studies (5, 6) endotoxins have been shown to be associated with chronic obstructive airway disease. Although airway disease is the common effect in workers with prolonged exposure to organic dust, parenchymal lesions have been described a few times. Several animal studies describe emphysema in species exposed to organic dust (16). Hence an additional explanation for the decrease in flows at low lung volumes with increasing EXPYEAR may be, besides a narrowing of peripheral airways because of a thickening of airway walls through inflammation and plugging, a diminished tethering of the peripheral airways through loss of elastic recoil of the parenchyma. Loss of lung elastic recoil is also a possible explanation for the low negative correlation between FVC and EXPYEAR. According to two studies (21,22) of smokers, total lung capacity can increase due to a loss of parenchymal elasticity. On the other hand, an increased residual volume due to premature airway closure should be expected. A concurrent increase in total lung capacity and residual volume may result in an almost unaffected FVC. This possibility is in agreement with our finding that the MEFV curve changes that occurred in relation to EXPYEAR and smoking habits were very much alike (figure 4). Indeed, Cotton et al (4) showed that the effects of smoking and organic dust exposure are about equal in terms of lung function changes. They could not however separate the acute and chronic effects of organic dust exposure. Thurlbeck (23) and Nagai et al (24) showed that cigarette smoking can cause chronic bronchitis, emphysema, peripheral airway diseases, or any combination of these. As an analogue to studies about cigarette smoking (23,24), one can hypothesize that airway disease in workers with prolonged organic dust exposure is also multifactorial, involving both peripheral and central airways and possibly also lung parenchyma.
A striking finding is that the FVC and PEF seem to decrease, especially in the first years of exposure, 442 while workers with prolonged organic dust exposure show reduced values for all MEFV curve variables, except for the FVC (figure 2). Changes in the MEFV curve in the first exposure years may again be the short-term lung effects described for MEANDUST; these short-term effects develop a more chronic nature at a later stage. However, also health selection in the first years might have influenced the observed pattern in figure 2. Such selection in the first years was suggested by a longitudinal study of 164 workers in grain elevators (25).
Our results suggest an interaction effect between smoking and current organic dust exposure (figure 4). This interaction has formally been confirmed by a multivariate analysis of variance (table 4). Nonsmokers seem to be affected more by current organic dust exposure than ex-and current smokers. Such an effect has, to our knowledge, not been described previously. Only additive (26,27) and synergistic (3, 4) effects of smoking and organic dust exposure have been found with respect to ventilatory function. An explanation for our results might be a selection effect, meaning that ex-and current smokers would represent a survivor population that is less susceptible to present organic dust exposure than a population of nonsmokers. Another hypothesis is that the lungs of ex-and current smokers are already irritated, and therefore the additional acute effect of present organic dust exposure is small.
In conclusion, mean organic dust exposure (MEANDUST) and years of exposure (EXPYEAR) relate to almost independent aspects of ventilatory function. MEANDUST is mainly related to airway obstruction that originates from disease processes affecting central airways and, to a less extent, peripheral airways, and these processes seem to have a short-term nature. EXPYEAR relates to airway obstruction extending to both central and peripheral airways and perhaps also to the parenchyma, and this type of obstruction may have a more chronic nature. EXPYEAR relates to both short-term and long-term effects of organic dust exposure (ie, a decrease of the FVC and PEF in the first exposure years and, in following years, a reduction in all of the MEFV curve variables, except FVC. The long-term effects of organic dust exposure and smoking on ventilatory function are very much alike. In contrast to other studies the effect of current organic dust exposure is greater for nonsmokers than for ex-smokers and current smokers and has no synergistic or additive effect for smokers.