INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Acute respiratory effects of airborne endotoxin exposure are well recognized. Endotoxin-related lung function decrease has been shown in several challenge studies (1, 2). Moreover, in a number of occupational settings associations between endotoxin exposure levels and lung function change across the work day have been observed (3). However, the influence of host factors such as smoking, atopy, and respiratory health status on these relationships has not yet been elucidated. In a few experimental studies the role of host factors has been addressed. Blaski and coworkers (6) conducted a challenge study among 20 healthy nonsmokers who inhaled corn dust extract with 30 µg of endotoxin. Atopy was defined as a skin test response to  2 environmental allergens. No significant differences in acute airway responses between atopics and nonatopics were observed. In another controlled study, Haglind and Rylander (7) observed a larger endotoxin-related decrease in FEV₁ among 17 smoking cotton workers compared with 11 nonsmoking cotton workers. Michel (8) found in a lipopolysaccharide challenge study that lung function decrease in 38 asthmatics was larger than in 13 normal subjects.

In addition to host factors, little is known about endotoxin-related respiratory effects during different types of work shifts. Airway caliber and, as a result, lung function are known to be subject to a circadian rhythm (9). Workers on other than day shifts will normally experience a phase shift of this circadian cycle to correspond with their own sleep-wake cycle. Several days are necessary to make this adjustment, and some individuals adapt more completely than others (12). Many studies have been conducted to assess lung function changes across the work day (3, 11, 13), but few considered different types of work shifts. However, already in 1966 the most comprehensive study on across-shift changes in FEV_0.75 among 473 shift workers employed in three cotton mills was reported (14). Different magnitudes of lung function change for early, late, and night shift were observed, being dependent on job title and presence of byssinosis. Later but smaller studies (13, 15, 16) confirmed the observation of different magnitudes of across-shift change for different work shifts, also for other lung function indices.

To our knowledge, only one study addressed acute lung function changes in relation to occupational endotoxin exposure for different types of work shifts. Milton and coworkers (5) studied peak flow changes across morning and night shifts in a fiberglass manufacturing facility. During the night shift a larger number of workers showed a peak expiratory flow (PEF) decrease  5% compared with the morning shift. During both shifts, an endotoxin-related exposure-response relationship with a PEF decrease  5% seemed present. Potentially modifying effects of type of work shift on the exposure-response relationship were not clearly evaluated, but it could be deduced from the data that the slope of these exposure-response relationships was similar for the morning and night shift.

As part of a larger study on respiratory health in potato processing workers, repeated peak flow measurements were performed among 97 workers. In this industry, a large range in endotoxin exposure across different plants and different departments has been observed (17), and endotoxin-related lung function changes across the afternoon shift have been described (18). This report focuses on across-shift peak flow changes and their relationships with endotoxin exposure in four potato processing plants. Special attention is given to the influence of type of work shift on these relationships. To identify workers who may be susceptible to endotoxin, the role of the host factors smoking and atopy was also investigated.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Population and Questionnaire

Production workers from all four plants of a large potato processing company were involved in a comprehensive study on occupational exposures and respiratory health in this branch of industry. The plants were in continuous operation from August to March. Three work shifts were distinguished: morning (7:00 A.M.-3:00 P.M.), afternoon (3:00 P.M.-11:00 P.M.), and night (11:00 P.M.-7:00 A.M.). In one of the four plants shifts changed 1 h earlier (6:00 A.M., 2:00 P.M., 10:00 P.M.). The sequence of shift rotation was afternoon  morning  night  afternoon, etc. A work period lasted alternately 3 or 4 d, with a subsequent leisure period of 2 d (after morning and afternoon shift) or 3 d (after night shift).

Workers completed a self-administered questionnaire with items on occupational history, smoking habits, and the Dutch version of a questionnaire on respiratory symptoms of the British Medical Research Council (19). Questions on work-related respiratory symptoms, determined as "occurring during work more frequently than normal" were also included. Cigarette-years, being the number of years smoked multiplied by the number of cigarettes per day, was calculated as a measure for cumulative smoking dose.

Peak Flow Measurements

All production workers were asked to participate in the peak flow survey. Peak flow meters and diaries were distributed among 128 workers; 103 workers returned a diary, of which six were not applicable because the number of days with more than one recorded measurement was too small. Thus, the study population comprised 97 male shift workers. This group was representative for the entire population regarding respiratory health and exposure characteristics.

PEF was measured using mini-Wright peak flow meters three times (leisure days) or four times (work days) daily for a 23-d period in August/September 1992. On work days target times were (1) just after rising, before work; (2) in the middle of the work shift; (3) directly after the work shift; and (4) before going to sleep. On leisure days target times were (1) just after rising; (2) in the middle of the day; and (3) before going to sleep. Exact time of measurement was also recorded. On each occasion, three maneuvers were performed (20, 21) and recorded in the diary. The coefficient of variation (CV) between the three maneuvers was calculated for each occasion and averaged per subject to evaluate measurement precision. Further analyses were performed using the highest of the three maneuvers. Peak flow-time graphs were made for each worker, and visually inspected to detect obvious data errors. To assess PEF variability within days, the ratio of amplitude to mean (Amp/Mean) (12, 22) was calculated for the different work shifts and leisure days.

Peak flow measurements closest to the start and to the end of the work shift were used to calculate across-shift changes in peak flow. Deviations of prescribed measurement times from start or end of the shift had to be within 2 h. Peak flow change on leisure days was calculated using times similar to those of the afternoon shift (3:00-11:00 or 2:00-10:00 P.M.).

Atopic Status

Serum samples were available from 93 workers; 85 workers were sampled both before and during the processing campaign. Total IgE as well as specific IgE to five common aeroallergens (house dust mite, grass pollen, birch pollen, cat dander, and dog dander) were determined in all 178 sera as described by Doekes and coworkers (26). Workers were considered atopic if elevated specific IgE to at least one allergen was detected, either before or during the campaign. A second definition of atopy required a total IgE concentration  100 U/ml, either before or during the campaign.

Endotoxin Exposure

Results of endotoxin exposure measurements have been described elsewhere (17). Briefly, 195 personal full-shift inhalable endotoxin exposure measurements were performed among 123 workers (among whom were 87 subjects included in the present study). Overall day-to-day variability was relatively low: the geometric standard deviation (GSD) within workers was 2.1 (range 1.1 to 3.2 across the 27 categories). Large differences in endotoxin exposure level between plants and between job categories were observed. A job exposure matrix (JEM) based on a categorization of job by plant with 27 categories was developed. The arithmetic mean endotoxin exposure for each category was used as exposure estimate for workers in each group. Using this approach, the between-group relative to within-group variance (contrast) was 0.38. Thus, for each of the 97 workers in this study endotoxin exposure was assessed on the basis of his plant and job category, and ranged from 53 to 8,167 endotoxin units per m³ (EU/m³). The overall geometric mean exposure of the 97 workers as estimated using the JEM was equal to 534 EU/m³ (GSD = 3.24).

Statistical Analyses

Data were analyzed using SAS version 6 (27). Associations between endotoxin exposure and respiratory symptoms were evaluated by calculating prevalence ratios (PR) (28) using the SAS PHREG procedure. Absolute across-shift peak flow changes were calculated by subtracting postshift from preshift values. Relative changes were computed by dividing the absolute change by the age- and height-specific predicted value (9), and multiplying by 100%. Advantage of the latter approach is that comparisons of relative decreases and increases across the different work shifts can be made. Relationships between endotoxin exposure and peak flow changes were analyzed using true repeated measures analyses (MIXED Procedure in SAS). Interaction terms in multiple models were used to test differences in exposure- response relationships between strata. Because the distribution of endotoxin exposure was skewed to the right, all analyses were performed using ln-transformed concentrations.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Respiratory Health Characteristics

Characteristics of the 97 potato processing workers are shown in Table 1. Ten percent of the workers indicated work-related obstructive symptoms. Respiratory health characteristics were comparable over the three categories, although symptoms tended to be less prevalent in never-smokers. The two definitions of atopy were not in concordance for 22 workers (24%). Analyses using data of 85 workers with sera both before and during the processing campaign revealed that no agreement in atopic status existed between "before" and "during" in seven and five workers for specific IgE and total IgE, respectively.

The PR of work-related respiratory symptoms on ln-transformed endotoxin concentration was 2.3 (95% confidence interval, 1.4 to 3.8). Adjustment for current smoking or atopic status in multiple models yielded a PR close to 2.3 as well. This illustrates that smoking and atopy did not act as confounding variables in the relationship between endotoxin exposure and respiratory symptom prevalence. Stratification for smoking and atopic status yielded similar point estimates for subgroups, indicating that these host factors did not modify the studied association. For all subgroups, the interquartile range in endotoxin exposure of 249 to 1,411 EU/m³ was associated with a PR for work-related symptoms of 4.2.

Peak Flow Measurements and Variability

All 97 workers performed peak flow measurements for at least 16 d; 71 of them completed all 23 d. There were no indications that peak flow patterns of the 26 workers who did not complete the entire study period were different from the 71 others. The mean CV between the three maneuvers was on average 2.9% (range, 0.9 to 9.1%). To investigate the presence of a learning effect, mean PEF was plotted versus measurement day. No increase of the mean PEF in the beginning of the study period could be observed.

In Table 2, Amp/Mean is shown for different time windows for different subgroups. Amp/Mean was highest during the morning shift, and lowest during leisure days. No differences between current smokers and nonsmokers were found. Amp/ Mean was also similar in ex- and never-smokers (results not shown). The number of cigarette-years was significantly associated with overall Amp/Mean: Pearson's r was 0.37. Atopics showed less peak flow variability than nonatopics, as indicated by the lower Amp/Mean values for all time windows.

Across-Shift Peak Flow Changes

Eight-hour change in peak flow was available for 376 morning shifts (93 workers), 367 afternoon shifts (93 workers), 319 night shifts (85 workers), and 555 leisure days (93 workers). In Table 3 the mean relative PEF change is shown for different time windows for different subgroups. The mean peak flow increased across the morning shift and decreased across both the afternoon and night shift. Across-shift change tended to be smaller in atopics compared with nonatopics. Across-shift PEF change was similar for ex- and never-smokers (results not shown).

Relationships between Endotoxin Exposure and Across-Shift Peak Flow Changes

Results of mixed regression analyses for relative across-shift PEF change on ln-transformed endotoxin exposure are presented in Table 4. p Values for interaction terms were only evaluated for models presented in the top row and the last column. Overall, negative coefficients indicate that endotoxin exposure was negatively associated with acute PEF changes. In the last column multiple regression models with adjustment for shift are presented. Endotoxin exposure-response relationships tended to be steeper for the first work day after a leisure period. Stratification for current smoking and atopy yielded similar point estimates for subgroups, indicating that smoking and atopy did not modify the relationship between endotoxin exposure and peak flow change. Point estimates for ex- and never-smokers were also similar (results not shown). Finally, adjustment for current smoking or atopic status in multiple analyses yielded similar coefficients as the crude estimates (results not shown). This shows that these host factors could not have confounded the relationships of interest.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

In this study, the prevalence of work-related respiratory symptoms and across-shift peak flow changes were associated with occupational endotoxin exposure among potato processing workers. Endotoxin-related peak flow change was most pronounced during the afternoon shift (3:00-11:00 P.M. or 2:00- 10:00 P.M.), and for all shifts larger on the first work day after a leisure period. No consistent differences in exposure-response relationships between smokers and nonsmokers, and between atopics and nonatopics were found.

In general, results of this study support the evidence for adverse respiratory effects of endotoxin exposure. In particular, this study focused on influences of shift work and host factors on these effects. As could be expected because of the circadian rhythm, peak flow increased during the morning shift and decreased during afternoon and night shifts (14, 15). The circadian rhythm has its top at about 4:00 P.M. and its minimum at about 4:00 A.M. (12). Even during consecutive days after rotating into the night shift, the peak flow followed this pattern (Table 3). This suggests that the circadian rhythm did not change to a large extent after change of work shift. This is supported by the fact that the applied shift system in this industry appeared to be favorable for workers because small adaptions of the circadian rhythm were noted in studies in which different shift systems were compared (29).

Amp/Mean was calculated to assess peak flow variability (Table 2). Neukirch and coworkers (25) found an average Amp/Mean of 9 to 12% in a study of 117 blue-collar workers, age 22 to 58 yr, which is higher than in our study. In a general population study among 265 Dutch males from 20 to 70 yr of age (24), Amp/Mean was 3 to 4%, which is somewhat lower than in our study. It has been suggested that a higher number of daily PEF readings increases the level of Amp/Mean (30). Other studies reported a median Amp/Mean of 5% (22) and 8.5% (23) among healthy adults. Amp/Mean was highest in current smokers, followed by ex- and never-smokers, respectively, which is in agreement with the quoted studies. Amp/ Mean tended to be lower in atopics in our study. A satisfactory explanation cannot be given. Boezen and coworkers (31) observed no differences in Amp/Mean between atopics and nonatopics, on the basis of both specific IgE and total IgE. Little is known about peak flow variability in shift workers. In our study, Amp/Mean was highest during the morning shift, and lowest during the afternoon shift. This can be explained by the first daily PEF measurement at 6:00-7:00 A.M., which is shortly after the minimum of the circadian rhythm, and this relatively low value results in a high variability (32), which is in agreement with results in Table 3. The influence of peak flow recording times for different work shifts on the Amp/Mean was too large to use this parameter as a valid health outcome variable in our study. Therefore, across-shift changes were computed and related to exposure.

To our knowledge, only in one study different work shifts were considered in the study of endotoxin exposure-response relationships. Milton and coworkers (5) studied peak flow changes both during morning and night shifts in a fiberglass manufacturing facility. For both shifts the same cutoff level of across-shift PEF change, 5%, was used in analyses. As expected because of the circadian rhythm, a larger number of workers showed a PEF decrease  5% during the night shift compared with the morning shift. The observed negative association between endotoxin exposure and peak flow change was similar for both types of work shifts. Results from our study support this finding. Very few other studies focused on the influence of type of work shift on respiratory effects of occupational exposures. Pasker and coworkers (13) showed that differences in across-shift lung function change between zinc oxide exposed and nonexposed workers were larger in the night shift, as compared with the day shift. More study is needed to evaluate the effect of type of work shift on exposure-related respiratory effects.

Endotoxin-related peak flow changes were largest on the first work day after a leisure period. This agrees with our previous study in this industry (18) using repeated spirometry across the afternoon shift, and is also consistent with findings that endotoxin-induced symptoms were most severe on Mondays and became milder the following days (33). The occurrence of a short-term adaption or "tolerance" has been reviewed previously (34), and might play a role in the explanation of this phenomenon.

The influence of the host factors smoking and atopy was also evaluated in this study. No consistent differences between atopics and nonatopics could be observed, which supports results from an experimental study (6). Consistent differences between smokers and nonsmokers could not be detected, either. A small controlled study (7) suggested that smokers showed a larger endotoxin-related response than nonsmokers. However, only 17 smokers were compared with 11 nonsmokers in the quoted study. It must be stressed that if our study had been limited to only one type of work shift, different conclusions with regard to interfering effects of the mentioned host factors could have been drawn (Table 4). A major implication of our findings is that results of studies among shift workers may be misinterpreted if not all work shifts are being taken into account.

Some aspects of this study require further discussion. First, repeated peak flow monitoring was used to measure acute respiratory effects. Workers performed the measurements themselves, hence no quality assurance was possible. One Canadian study (35) suggests that some workers investigated for occupational asthma had falsified their peak flow results. Visual inspection of peak flow-time graphs in our study showed no eye-catching patterns suggestive of falsification. Although falsification in our study cannot be excluded, it is not likely to play an important role because the purpose of our study was not the detection of occupational asthma, for which financial compensation is possible in Canada. In addition, in the quoted study workers had to record peak flow every 2 h for on average 36 d, which is more demanding than the effort required in our study. Moreover, results in Table 3 indicate that patterns in peak flow change were consistent with expectations on the basis of the circadian rhythm.

Another limitation of this study was that exposure monitoring was not performed concurrent to the peak flow measurements. However, endotoxin exposure was largely determined by plant and job category, and day-to-day variability in exposure was low (17). A major advantage was that all workers with peak flow data could be included in exposure-response relationships.

It cannot be excluded that other bacterial components like peptidoglycan (36, 37) play a part in observed peak flow changes because a previous study in this industry showed that endotoxin levels were related to bacterial counts (17). It could be argued that endotoxin is a general marker for the presence of microorganisms in organic dust in this industry. Therefore, generalizations of these findings to other occupational settings with different coexposures should be made with caution.

We conclude that endotoxin exposure in the potato processing industry is related to across-shift peak flow changes and the occurrence of work-related respiratory symptoms, and is therefore likely to play a role in acute work-related respiratory effects. The host factors smoking and atopy are not important confounding or effect-modifying factors in these relationships. In respiratory health studies among shift workers, it is important to investigate all work shifts before drawing definitive conclusions on exposure-response relationships. More study is needed to unravel the impact of possibly interfering factors in adverse effects of occupational endotoxin exposure.