Annals of Occupational Hygiene Advance Access originally published online on February 10, 2005
Annals of Occupational Hygiene 2005 49(5):423-437; doi:10.1093/annhyg/meh113
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
© 2005 British Occupational Hygiene Society Published by Oxford University Press
Original Article |
Environmental Exposure Characterization of Fish Processing Workers
1 Occupational and Environmental Health Research Unit, School of Public Health and Family Medicine, University of Cape Town, Room 4.44, Fourth Level, Falmouth Building, Anzio Road, Observatary, 7925, South Africa; 2 Department of Environmental Health Sciences, University of Michigan, USA; 3 Department of Environmental and Occupational Health Sciences, University of Washington, USA; 4 Department of Health Sciences, Peninsula Technikon, South Africa; 5 Department of Medicine, Tulane University Medical Centre, USA; 6 Division of Immunology, IIDMM, Medical School, NHLS, University of Cape Town, South Africa
* Author to whom correspondence should be addressed. Tel: +27 21 4066309/6300; fax: +27 21 4066163; e-mail: mjeebhay{at}cormack.uct.ac.za
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
Background: Aerosolization of seafood and subsequent inhalation, during processing is a potential high-risk activity for allergic respiratory disease.
Objectives: To quantify total thoracic particulate, protein concentration and specific fish (pilchard, anchovy) antigen concentrations in fish processing plants; to determine the correlation between these exposure metrics; and to identify the major determinants of variability and the optimal grouping strategies for establishing dose–response relationships for fish antigen exposures.
Methods: Exposure assessments were conducted on randomly selected individuals within each of the identified ‘exposure groups’ (EGs) in two fish processing factories. Personal time-integrated sampling was conducted with a thoracic fraction sampler and analysed for particulate mass, total protein and specific fish antigens. Exposure metrics were developed on the basis of individually measured exposures and average levels of these personal samples within EGs. The main components of the exposure variability were determined using ANOVA techniques.
Results: A total of 198 full-shift personal aerosol samples were collected and analysed. Twenty-two percent of the samples were below the limit of detection (LOD) for pilchard and 23% for anchovy assays. Personal sampling revealed wide variations across EGs in arithmetic mean concentrations of thoracic particulate 0.61 mg m–3 (range: LOD–11.3), total protein 0.89 µg m–3 (LOD–11.5), pilchard antigen 150 ng m–3 (LOD–15 973) and anchovy antigen 552 ng m–3 (LOD–75 748) levels. The fishmeal loading and bagging sections of both plants showed consistently high thoracic particulate mass (0.811–2.714 mg m–3), total protein (0.185–1.855 µg m–3), pilchard antigen (538–3288 ng m–3) and anchovy antigen (1708–15 431 ng m–3). The a priori strategy that grouped workers according to EGs produced reasonably satisfactory summary exposure metric statistics. An alternative grouping strategy based on department revealed comparable elasticity (exposure contrast). While the correlation between the log-transformed thoracic particulate mass and fish antigen concentrations were generally modest (Pearson's r = 0.32–0.35, P < 0.001), a high correlation was found between pilchard and anchovy antigen concentrations (Pearson's r = 0.71, P < 0.001). Models using factory and department grouping strategies accounted for a significant portion of the variability (adjusted r2 = 0.18, P = 0.043) in pilchard antigen levels. Grouping strategies using a combination of factory and department yielded the highest degree of elasticity for thoracic particulate (0.38) and pilchard antigen (0.42) levels.
Conclusions: Workers involved in bony fish processing are at risk of inhaling aerosols containing pilchard and anchovy fish antigens. Antigen exposures are highest during fishmeal production and bagging. Grouping strategies based on department and factory may provide a more efficient approach than a priori classification of EGs for evaluating fish antigen exposures.
Keywords: bioaerosols • exposure assessment • fish antigens • fishmeal • fish processing • pilchard antigen
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
Occupational exposure to seafood allergens occurs mainly in the food and fishing industry (Jeebhay et al., 2001
). Workers in a number of these industries are exposed to seafood, especially those involved in either manual or automated processing of crabs, prawns, mussels, fish and fishmeal (Malo and Cartier, 1993). Seafood processing plants vary in technology levels, with some of the smaller workplaces relying entirely on manual handling of seafood while larger companies use highly automated modern processes. There is great variation in the processing procedures for the different types of seafood (Moody et al., 1993; Malagie et al., 1998).
Aerosolization of the seafood (meat, exoskeleton, blood, endolymph) during processing has been identified as a potential high-risk activity for sensitization through the respiratory route (Lehrer, 1990; Crespo et al., 1995; Douglas et al., 1995; Smith and Sechena, 1998; Weytjens et al., 1999; Aresery and Lehrer, 2002). Identified processes with high potential for aerosol exposure include: butchering/grinding, degilling, ‘cracking’ and boiling of crabs; ‘tailing’ of lobster; ‘blowing’ of prawns, washing/scrubbing of shellfish; degutting, heading and cooking/boiling of fish; mincing of seafood; cleaning of the processing line and storage tanks with high-pressured water (Jeebhay et al., 2001). Despite high levels of automation in large workplaces, workers employed in these facilities are often found to be at high risk due to inadequate and poorly designed local exhaust ventilation systems (Douglas et al., 1995). Furthermore, processes that generate dry aerosols (prawn blowing using compressed air) appear to generate higher particulate levels than wet processes (prawn blowing using water jets) (Gaddie and Friend, 1980). It has been suggested that because water provides general aerosol suppression, it may also influence the size, lifetime or other dynamics of small protein particles as water is a major feature of this work environment (Ortega and Berardinelli, 1999; Ortega et al., 2001).
A recent review that included evaluation of studies on exposure assessment in the seafood industry found that there is great variability of exposure within and among various jobs involved in seafood processing with reported seafood antigen concentrations ranging from 1 to 5061 ng m–3 (Jeebhay et al., 2001). Generally, much higher antigen concentrations have been found using personal sampling as compared with area sampling. The lack of standardized methods for environmental sample collection, extraction and analysis of high molecular weight sensitizers, such as seafood proteins, makes comparisons between various studies difficult (Heederik et al., 1999). Notably, there have been only two exposure characterization studies on bony fish reported in the literature. Douglas et al. found higher respirable particulate (2.71–3.57 mg m–3) in area samples of wet processes such as fish gutting and grading than in dry processes such as fish packing in stores (0.04–0.05 mg m–3) in a salmon processing plant. These particulate levels of 2.71–3.57 mg m–3 corresponded to salmon fish antigen concentrations of 100–1000 ng m–3 (Douglas et al., 1995). Taylor et al. demonstrated whiff, megrim and hake antigen levels of 2–25 ng m–3 obtained from area samples collected in a fish market (Taylor et al., 2000).
The main objectives of the current study were to quantify the total thoracic particulate, total protein and specific fish (pilchard, anchovy) antigen concentrations collected through personal aerosol sampling. The study also aimed to examine correlations among thoracic particulate, protein and specific fish antigen levels. Furthermore, the major determinants of variability of these exposure metrics and optimal grouping strategies for establishing dose–disease response relationships for fish antigen exposures were evaluated.
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
Exposure assessment was conducted using a combination of work history information and personal exposure time-integrated measurements to develop several exposure metrics for analysis.
Sampling strategy
An experienced industrial hygienist (E.R.) first performed a walk-through inspection of each department to identify work groups that were similar with respect to production processes and potential exposures. Job titles and descriptions for each worker in the plant were obtained from production managers and compared with information obtained from questionnaires completed during the health outcome measurement survey. A total of 99 different job titles were reported by workers from the different departments of both factories. The risk of exposure to fish aerosols associated with each job, as reported by the manager and worker, was compared with the estimated risk during the walk-through inspections conducted by the industrial hygienist. Exposure groups (EGs) were then identified to serve as the basis for the sampling strategy to be used during personal aerosol exposure assessment. Each EG consisted of a number of different job titles within a given department that were assessed as being similarly exposed based on the walk-through inspection together with the worker and supervisor descriptions. Job titles that were difficult to classify into any particular EG due to mobility, were classified as the ‘non-stationary exposure group’. Such group average exposure levels can be superior to individual exposure levels that examine exposure health outcome associations when limited data are available on the individual (Heederik et al., 1997).
The walk-through inspections revealed the existence of largely similar departments at the two seafood-processing plants, except for the rock lobster department, which was found only in one plant. Full descriptions of the department-specific processes in these plants have been presented previously (Jeebhay et al., 2004). Some structural processing differences existed between the two factories, which influenced the final compilation of the EGs. Pick-ups, labelling and full can store constituted different departments at Factory B, while they were all in one department in Factory A. The EGs for the respective factories were finally classified from the following departments/sections: jetty, canning, cannery pick-ups, fishmeal manufacturing, fishmeal warehouse (loading, bagging), boiler, rock lobster, labelling, workshop (carpentry, electrical, mechanical), stores, administration and laundry/cleaners. Of the total number of workers that formed part of the study population, 260 workers were employed in Factory A and 383 were employed in Factory B. This constituted 100 and 30% of the respective total workforce populations, with invited Factory B workers being a random sample of all workers stratified on department. Almost 65% of the workers in Factory A and 72% in Factory B were seasonal workers mainly employed in the canning department. The departments employing the largest number of workers were cannery (64% of the total of 643 participants), fishmeal manufacture/warehouse (8%), rock lobster processing (8%) and can labelling (8%). Almost half the workforce (48%) in both factories worked on one of the two night shifts.
Initial observations suggested a total of 36 identifiable EGs (14 EGs in Factory A and 22 in Factory B). In Factory A, this included: five EGs in canning (sorting table, steam exhaust box, retorts/autoclave, pick-ups/labelling, stores), three in fishmeal production (fish pits, manufacturing, loading) and one each in jetty, workshop, boiler room, laundry and administration section. In Factory B, this included seven EGs in canning (sorting table, steam exhaust box, retorts/autoclave, pick-ups, labelling, two stores), three in fishmeal production (maintenance, shovelling, loading/bagging) and one each in the jetty, workshop, boiler room and the administration section. There were six additional EGs in the lobster processing department (administration, sorting, steaming/washing/scrubbing, packing, weighing, mastering). All non-stationary workers in the cannery and lobster sections were assigned a separate EG in each factory. These EGs formed the basis for the sampling strategy. More detailed descriptions of the tasks in these EGs have been described previously (Jeebhay et al., 2004).
The fishing season for bony fish begins in October and continues until August of the following year. The rock lobster harvesting season starts later in December and ends in August. Seasonal workers work a maximum of 11 months per year and only on production days. In both factories, permanent workers worked mainly on the day shift (0800–1700). Production workers in Factory A worked two shifts (0600–1800, 1800–0600) and in Factory B on three shifts (0700–1500, 1500–2300, 2300–0700). Personal sampling for bony fish processing was done during November and December 2001 and lobster processing in February 2002. Sampling was done on all shifts during the entire work shift in both factories to obtain a full-shift time-weighted average exposure for each job. Ninety-two percent of the samples were collected during the day shift and 8% during the night shift (total number = 198).
All workers on a particular shift were briefed prior to the commencement of sampling by the study's industrial hygienist. Once workers commenced their work shift, they were approached by the industrial hygienist based on the job they performed within a particular EG. At least five personal aerosol exposure samples were obtained from various jobs in each EG in order to account for variability in exposure and to ensure representative exposure assessment. In EGs where there were more than five job titles, jobs entailing similar tasks were grouped together and workers were randomly sampled from each sub group within the EG. In EGs with less than five job titles all jobs were sampled. Morevoer, for these EGs with fewer job titles, a random selection of workers in that EG underwent repeat personal sampling on additional days. Most workers (89%) were sampled once, while 7% of workers in the same job were sampled twice and 4% were sampled three or more times. None of the workers refused to undergo personal sampling and all workers provided informed consent before sampling proceeded.
Aerosol sampling instrumentation and procedure
All personal sampling (particulate, protein and specific antigen) was conducted using full-shift-time weighted average samples (range of sampling period: 2.8–7.6 h) collected on each participating worker. A sampling train was set up for each worker by an occupational hygienist, with the flow rates of each pump calibrated according to the specific needs of that pump. The pumps were calibrated prior to the survey and immediately after sampling using a soap-bubble metre (primary standard) and an electronic stopwatch. A flow rate check was also conducted at the end of each sampling session. Due to the large number of samples being collected over a continuous period (night and day shifts) either an SKC Aircheck Sampler (Model 224-PCCXR) or a Gill Air or Du-Pont Alpha-1 battery operated air-sampling pumps running at an average flow rate of 3 l min–1 was used. A two-stage sampler, Personal Environment Monitor (PEM10) Model 200 PEM-2-10 (37 mm filter diameter) (manufactured by MSP Corp., Minneapolis, MN), consisting of a single impaction stage (preselector) and a backup filter was used at the end of the sampling train. The preselector has an upper median cut diameter of 10 µm, resulting in the collection of approximately the ‘thoracic’ fraction (i.e. reaching the tracheo-bronchial and alveolar regions of the respiratory tract) of particulate on the filter (Lippmann, 1989; Marple, 1989). The PEM10 was fixed to the lapel of the overall/apron, within the breathing zone of the worker. During the sampling period any unusual occurrences were noted and recorded on the sampling data sheet. It must be noted that the PEM10 used in this study is actually designed to operate at a flow rate of 2 l min–1 in order to obtain the desired 10 µm cut point of the thoracic fraction. However, in an attempt to improve the sampling efficiency for antigen collection the air-sampling pumps were set at 3 l min–1 during the collection of all the samples. Increasing the flow rate may have resulted in the actual cut point being reduced from 10 to 8.2 µm. This value is obtained from particle theory, i.e. the particle diameter collected by an impactor is proportional to the inverse square root of the flow rate (e.g. 10 µm x 
) (Willeke and Baron, 1993). Although the samples do not exactly fulfill the current thoracic fraction definition, since they were all collected in a similar manner, they are all comparable to each other. Furthermore, the cut point is still close enough to represent particles in the thoracic fraction range that penetrate to the bronchial tree.
The aerosols were collected on polytetrafluroethylene (PTFE) membrane laminated filters (0.3 µm pore size) (Quan-Tec-Air Inc.). These were weighed before and after sampling in a humidity and temperature-controlled environment using a microbalance, to determine particulate mass concentrations. The weighing process was repeated thrice with the average weight being recorded as the filter weight. At least 10% of the filters were submitted together with the field samples as field blanks in order to control for potential contamination during handling, storage, transportation and analysis of the filters. The filters were transported in a cooler box on ice to a freezer at the factory, where it was stored at a temperature of 4°C until transport to the main laboratory where filters were stored at –20°C until extraction.
Analysis of samples for protein and antigen concentration determinations
Preparation of seafood extracts
Extracts (1:10 w/v) of pilchard (Sardinops sagax) and Cape anchovy (Engraulis capensis) were prepared from fresh specimens obtained from the factory and extracted into phosphate-buffered saline (PBS) overnight at 4°C. The extracts were sterile filtered using 0.45 µm filters (Millipore). The protein concentration of each extract was determined by using bicinchoninic acid assay (BCA protein assay; Pierce) and aliquots were subsequently stored at –80°C. These extracts were used to develop the antisera for the immunoassays and also as internal standards in these assays for detecting and quantifying the protein antigens collected on filters housed in the personal sampling cassettes.
Preparation of filters for personal samples
For elution, the membrane backing was removed, filters were cut into pieces and placed into 0.5 ml PBS–0.05% Tween 20. The filter pieces were eluted overnight on a shaker at 4°C, the mixture was centrifuged at 2000 g for 2 min and the supernatant was stored at –80°C until further use. These procedures were conducted according to similar methods described previously for seafood and flour antigen extractions (Lillienberg et al., 2000; Griffin et al., 2001).
Determination of protein concentration of samples
A protein standard was prepared in duplicate using BSA (bovine serum albumin) ranging from 0.05 to 0.25 mg incubated at 60°C for 30 min. After incubation the plate was cooled to room temperature and the absorbance read at 540 nm. A standard curve was used to determine the protein concentration for each unknown protein sample using the BCA protein assay.
ELISA-inhibition assay for antigen determinations of samples
An ELISA-inhibition assay specifically designed for this study was used to identify antigen presence on personal sample filters (Lopata et al., personal communication). These assays were conducted in the Division of Immunology at Groote Schuur Hospital in South Africa. The ELISA method uses the following reagents: sodium carbonate/bicarbonate buffer (100 mM) pH 9.6; Tris-buffered saline (TBS) 10x pH 7.4; 2% blocking reagent (powder skim milk in TBS with 0.05% Tween 20); TBS–0.05% Tween 20 buffer; TBS–alkaline phosphatase buffer 5x pH 9.5; goat anti-rabbit IgG antibody (Southern Biotechnology Associates, Inc.); substrate (PNPP; Sigma) and polystyrene 96-well microtitre plates (Dynex flat bottom 2HB high binding plates).
The method is explained in detail elsewhere (Lopata et al., personal communication). In brief, polystyrene microtitre plates (Dynex flat bottom 2HB) were coated with optimized protein concentration for each different seafood extract, followed by blocking and incubation for 60 min at 37°C. One standard inhibition curve was used for each plate assay ranging from 0.2 to 200 µg ml–1. The eluant from personal samples was added to each well followed by rabbit antiserum to determine the degree of inhibition. A high degree of inhibition indicates larger amounts of antigen in the samples. Provision was made to demonstrate ‘total specific binding’ (TSB = 0% inhibition) and ‘non-specific binding’ (NSB=blank control).
Standard inhibition curves, expressed by a simple linear regression model (y = mx + a), were used to determine the limit of detection for each specific plate. As seven plates per antigen were used to analyse the large number of samples, the detection limit was presented as a range. The antigen concentration obtained in micrograms of antigen per millilitre was corrected for the volume of air flowing through the personal air-sampling pump to obtain a final concentration in terms of micrograms of antigen per cubic metre. The detection limit for the pilchard assay was generally lower (range: 0.010–0.235 µg m–3) than that of the anchovy assay (range: 0.069–0.225 µg m–3). In the statistical analyses, values below the detection limit were considered as being two-thirds of this limit (Helsel, 1990; Hornung and Reed, 1990; Houba et al., 1997).
Quality control of the analytical methods was assured by comparing the standard curves obtained for the pilchard and anchovy antigen assay used by the collaborating laboratories (University of Cape Town, South Africa, and Tulane Medical Centre, USA) using the same protocols. The method recommended by Bland and Altman was used to test the level of agreement of the standard curves (n = 10) of the two laboratories (Bland and Altman, 1986). The pilchard and anchovy standard curves both showed good agreement between the laboratories: the confidence interval of the level agreement of the mean log10 difference was 0.008–0.102 for the pilchard standard curves and 0.092–0.172 for the anchovy standard curves. Replicate analyses conducted on a 10% subset of the samples (n = 20) also found reasonable agreement between the two laboratories.
The intra-assay (within plate) coefficient of variation (CV = standard deviation/mean) obtained during the analysis of all standards in the main study (by the South African laboratory) varied from 2.2 to 5.1% for the pilchard standard curves and from 1.6 to 3.7% for the anchovy standard curves. The inter-assay variability, reflecting the ‘day-to-day’ variation between microplates was used as an indicator for the reproducibility of results within the ELISA method. This was assessed by calculating the CV using the mean and standard deviation of all the optical density readings for each standard concentration ranging from 0.20 to 200 µg ml–1 for each antigen. The average inter-assay variation (as assessed by the CV) for the pilchard and anchovy standard curves over 7 days was 12.6 and 8.7%, respectively. These results are considered to be satisfactory since the CV was less than 15% (Sasaki and Mitchell, 2002).
Statistical analysis
All statistical analyses were performed using Stata computer software and graphical illustrations used SPSS software package (SPSS Inc., 2001; Stata Corp, 2001). Descriptive univariate statistics were generated for the total sample distribution. A correlation matrix using Pearson's correlation coefficient (r) was computed on the log-transformed exposure metrics for the overall data and for each department. Correlational analysis using non-parametric tests (Spearman's rank correlation) gave very similar results. Linear regression models were developed to describe the determinants of variability of the various exposure metrics. Since the data showed a lognormal distribution, the natural logarithm (ln) of the measured exposure was used as the dependent variable. Independent variables considered for entry into the models included factory, department, EG, job type and shift sampled. Regression diagnostics were also performed to assess and deal with outlying and influential observations.
The effectiveness of the a priori EG definitions was evaluated using a random effect analysis of variance (ANOVA) and calculation of grouping statistics including a measure of exposure contrast (elasticity) (Kromhout et al., 1995). Groups, workers within these groups and their subsequent measurement periods were assumed to be selected at random during a shift, so random effect models were used (Houba et al., 1997). The structure of the exposure data was regarded as hierarchical (nested) (Samuels et al., 1985) with job type nested in the EG, which in turn was nested in the department of each factory. A two-way nested random effects ANOVA model was applied to estimate a between-group (
) and a within-group (
) variance component for different grouping schemes. The estimates of the variance components and were designated as 
and 
, respectively. Using these variance components, the elasticity was calculated as: 
(Kromhout and Heederik, 1995; Houba et al., 1997). This ratio will reach the value of 1 if each group is entirely distinct. At the other extreme, if all groups are similar then this ratio will approach the value 0. Since most workers were only sampled on one occasion, the ‘precision’ statistic could not be calculated from the variance components obtained.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
A total of 198 full-shift personal aerosol samples were collected from 145 workers for particulate mass, protein, pilchard and anchovy antigen concentration determinations. There were 20 samples on which total protein determinations could not be done since the amount of eluant obtained was insufficient. Descriptive statistics and results of the one-way ANOVA are presented in Tables 1 and 2. The thoracic particulate mass, protein concentrations as well as pilchard and anchovy antigen concentrations followed a lognormal distribution for most EGs; 23% of the samples were below the detection limits for pilchard and 21% for anchovy.
|
|
Summary exposure metrics per EG
This study demonstrated a wide range of mean exposures from least (administration) and most exposed (fishmeal loading/bagging) jobs for thoracic particulate (GM: 0.097 mg m–3 versus 0.889 mg m–3), protein (GM: 0.583 µg m–3 versus 1.964 µg m–3), pilchard antigen (GM: 51 ng m–3 versus 353 ng m–3) and anchovy antigen (GM: 65 ng m–3 versus 1120 ng m–3) (data quoted for Factory A, Tables 1 and 2). The mean exposures for the least and most highly exposed exposure groups in both factories combined varied by about 9-fold for thoracic particulate, 50-fold for protein, and 12- and 18-fold for pilchard and anchovy antigen, respectively. Overall, the mean exposures across EGs varied considerably indicating a reasonably robust exposure gradient for all parameters of interest.
Of interest are the notable differences in exposures between the two factories when comparing groups with similar job titles and departments. For thoracic particulate mass concentrations, higher concentrations were found in the boiler room and fishmeal loading in Factory A as compared with Factory B (Table 1). The wide range of values obtained in the general storeroom and workshop (Factory B) were due to a single outlier observation. High ambient protein concentrations were found in the workshop (Factory B), the cannery exhaust box/saucer area (Factory A) and fishmeal loading (Factory A) (Table 1). Personal exposures to pilchard and anchovy aerosol antigen concentrations were highest in the fish pits, fishmeal manufacture and loading/bagging of both factories, cannery retorts (both factories) and cannery fish sorting table (Factory B) (Table 2). Non-stationary workers in the canning departments also had high mean pilchard antigen exposures in both factories. Surprisingly high ambient anchovy (but not pilchard) antigen levels were found in certain lobster EGs that do not normally handle anchovy, reflecting possible cross-contamination by non-stationary workers.
In each factory, the proportion of thoracic particulate consisting of pilchard or anchovy antigen varied across the EGs, most likely due, in part, to the use of ingredients other than fish in the production process (Table 2). The ratio of the average amount of pilchard and anchovy antigen to thoracic particulate was the highest in the fishmeal manufacturing EGs of both factories (pilchard: 0.868 parts per thousand by weight; anchovy: 1.839 parts per thousand by weight), reflecting high fish antigen levels in the dust particulate produced in this phase of the production process. Other groups with above average ratios were the jetty and fish pits suggesting bioaerosol generation during handling of large quantities of fish. The isolated unexpectedly high values obtained for the administration EG (Factory B), laundry and certain lobster EGs may point to possible cross-contamination of work environments by non-stationary workers from other departments (Table 2).
Correlation between exposure metrics
There were statistically significant correlations between the log-transformed ambient thoracic particulate mass and protein concentrations (r = 0.32, P < 0.001, n = 169) (Table 3; Fig. 1). Department-specific correlations between these two measures were especially high in labelling (r = 0.89) and boiler room (r = 0.83). Both these parameters were also modestly correlated with fish (pilchard and anchovy) antigen concentrations, with slightly higher correlations seen between thoracic particulate concentrations and fish antigen concentrations (r: pilchard = 0.32; anchovy = 0.35, P < 0.001, N = 191) than with total protein (r: pilchard = 0.23; anchovy = 0.17, P < 0.05, n = 171) (Table 3; Figs 2 and 3). The correlations between particulate and fish antigen were significantly higher in the fishmeal warehouse (r: pilchard = 0.75; anchovy = 0.82). As expected, a significant correlation was found between ambient pilchard and anchovy antigen concentrations (r = 0.71, P < 0.001, n = 194) (Table 3; Fig. 4), again with even significantly higher correlation in the fishmeal warehouse (r = 0.97).
|
|
|
|
|
Determinants of exposure variability
In the bivariate analyses, the EG explained the greatest variability observed in the thoracic particulate (adjusted r2 = 0.31, P < 0.001) and protein (adjusted r2 = 0.28, P < 0.001) measurements, followed by job titles (adjusted r2 = 0.22, P = 0.011) for thoracic particulate levels (Table 4). The department, however, explained the greatest variability in fish antigen measurements of both pilchard (adjusted r2 = 0.15, P = 0.013) and anchovy (adjusted r2 = 0.11, P = 0.041) fish species. The work shift did not contribute significantly towards the variability in any of the measurements observed. The addition of other variables such as factory, department and job title to the main explanatory variable did not improve the overall fit of any of the models, except for a model predicting pilchard antigen levels that included factory with department. These variables explained the greatest variability (adjusted r2 = 0.18, P < 0.005) in the final exposure model (Table 5):
 |
|
|
Grouping schemes based on estimation of variance component
Random effect ANOVA and calculation of grouping statistics including a measure of contrast (elasticity) between factories, departments and EGs are presented in Table 6. Since the determinants for the variability in thoracic particulate and protein were very similar, the former exposure metric was considered to be representative of both for these comparisons. A similar approach was used for choosing the pilchard antigen levels as being representative for anchovy antigen exposure levels. Among the grouping schemes based on a single factor, the EG yielded the highest elasticity values for thoracic particulate dust (0.34). For thoracic pilchard antigen concentration, grouping workers on the basis of department yielded the highest single factor elasticity value (0.34). Slightly higher levels of elasticity were also obtained for grouping workers by factory and department for both thoracic particulate (0.38) and pilchard antigen levels (0.42).
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
This study used a combination of industrial hygiene, immunological and statistical techniques to characterize the exposure to ambient thoracic particulate, protein and fish (pilchard and anchovy species) antigens among workers in fish processing plants. While exposure variability has been characterized in great detail for crab processing activities (Griffin et al., 1994
; Ortega and Berardinelli, 1999; Weytjens et al., 1999; Beaudet et al., 2002), in previous studies, only respirable particulate concentrations for various work activities have been reported for bony fish, i.e. salmon processing workers (Douglas et al., 1995). Notable differences in exposure levels using various metrics were observed when comparing EGs having similar job titles in the two factories in this study. These differences, however, did not demonstrate a consistent pattern in one factory having higher exposures than the other. The observed differences are likely to be attributable to factors such as factory size (Factory A being much smaller), physical layout and ventilation, handling procedures of fish before and after arrival at factory, actual amount of fish processed during the working day, variations in certain work procedures, activities and equipment used in processing, and a host of other unmeasured covariates best summarized as a ‘plant effect’. Despite these variations, consistently high thoracic particulate mass (AM: 0.811–2.714 mg m–3), protein (AM: 1.503–1.855 µg m–3), pilchard fish (AM: 538–3288 ng m–3) and anchovy fish antigen (AM: 1708–15 431 ng m–3) concentrations were detected in the fishmeal loading and bagging sections of both plants. These antigen exposures generally appear much higher than those reported by Douglas et al. for respirable particulate levels (area samples) of gutting and grading activities (2.71–3.57 mg m–3), which corresponded to salmon fish antigen concentrations of 100–1000 ng m–3 (Douglas et al., 1995). Surprisingly, the upper limit in the range of exposures for anchovy fish antigen concentrations in the current study was up to seven times higher than for protein (75 748 ng m–3 versus 11 500 ng m–3) (Tables 1 and 2). Possible reasons for this could be that not all antigens were of protein origin (e.g. type 1 collagen such as gelatin) (Sakaguchi et al., 2000), thereby interfering with the performance of the protein assay (Luo et al., 2002). In any event, there is considerable evidence that the various exposure metrics examined were robust and valid in that both the pilchard and anchovy standard curves as well as the replicate analysis of a subset of samples showed good agreement between the two participating laboratories.
In this study, marked differences in mean exposures were observed on comparing the lowest (administration) and the highest exposed (fishmeal loading/bagging) jobs for all exposure metrics, with the mean exposures for the least and highly exposed EGs in both factories varying by about 9-fold for thoracic particulate, 50-fold for protein, 12- and 18-fold for pilchard and anchovy antigen, respectively (Tables 1 and 2). The fraction of antigen as a proportion of thoracic particulate was the highest in the fishmeal EGs suggesting that antigens derived from proteinaceous fish material (meat, exoskeleton, blood) are particularly likely to be aerosolized during these processing activities. However, for a number of EGs expected to have low antigen exposures (e.g. administration, laundry, lobster packers and masters) the fraction of thoracic particulate represented by antigen was notably high, perhaps because of the lack of local sources for generating thoracic particulate as well as cross-contamination of these work environments by fish processors transporting fish antigens that may adhere to their shoes and clothing. Unexpectedly, lower correlations were obtained between fish antigens and protein concentrations (r = 0.17–0.23, P < 0.05) than fish antigen–thoracic particulate correlations (r = 0.32–0.35, P < 0.001). This suggests that the measurement of total protein in ambient personal samples may be a poor surrogate for exposure to a specific fish antigen. This could be due to the inherent inaccuracies involved in wet weight measurement of filters exposed to aqueous aerosols of low protein content as has been demonstrated for crab exposures (Griffin et al., 1994). The inaccuracies due to wet weight measurement would, however, apply to any correlation between gravimetric analysis and protein, or antigenic, analysis. Since it is more likely that a protein assay has a lower sensitivity than an immunoassay, a protein assay would be more prone to inaccuracy at low concentrations of ambient fish protein levels. Similarly, the generally modest correlations (r = 0.32–0.35) between thoracic particulate mass and fish antigen exposures support the previous reports in the literature concerning the need to directly measure airborne high-molecular weight antigens rather than relying on the particulate mass of total dust as the main surrogate of exposure (Heederik et al., 1999).
A high correlation was found between ambient pilchard and anchovy antigen concentrations (r = 0.71, P < 0.001), suggesting possible cross-reactivity of antibodies to antigenic determinants of the two fish species. These findings support western blot studies described previously, in which allergens of different fish extracts demonstrated molecular weights below 15 kDa, probably representing the major fish allergen, Gad c 1, previously isolated by Elsayed (Elsayed and Aas, 1970; Lopata et al., 2004). The high levels of ‘anchovy’ antigen (AM: 270 ng m–3) detected in the canning departments (Table 2) that only process pilchard fish can probably be explained on the basis of such cross-reactivity observed.
Some studies have suggested that elevated endotoxin levels may be responsible for symptoms reported by exposed workers. Previous studies by Sherson et al. isolated Gram-negative bacteria (Klebsiella pneumoniae and Pseudomonas) and 1 µg endotoxin ml–1 (10 000 EU ml–1) in their investigation of contaminated water from the gutting machine thought to be responsible for the respiratory symptoms among trout processing workers (Sherson et al., 1989). A study among crab processing workers demonstrated relatively low levels of airborne endotoxin levels obtained through personal sampling despite large numbers of Gram-negative Mesophilic bacteria isolated through bulk sampling of plant processing tanks (Ortega and Berardinelli, 1999). The mean levels of endotoxin reported by Ortega et al. were 32.6 (total fraction) and 15.6 EU m–3 (respirable fraction). Preliminary studies indicate that fishmeal production operations in South Africa produce higher ambient endotoxin levels (mean: 136 EU m–3) than fish canning operations (mean: 49 EU m–3) (Jeebhay et al., 2004). Specific jobs such as autopacker/gutting machine operator (275 EU m–3) and cooker/press operator (195 EU m–3) in the cannery as well as scale operators (258 EU m–3) in fishmeal bagging had very high exposures. Although no health-based occupational exposure limit for airborne endotoxin levels currently exists, these levels are higher than the 50 EU m–3 (
5 ng m–3) based on inhalable dust exposure that has been proposed (DECOS, 1998). The findings of all these studies suggest a possible role for endotoxin exposures and adverse respiratory effects, including non-allergic asthma, observed among exposed working populations (Douwes et al., 2002). However, more detailed endotoxin exposure characterization and dose–response studies among fish processing workers are needed to assess the relative contribution of this exposure to symptoms experienced by this group.
Exposure characterization grouping techniques are generally based on the underlying assumption that the average exposure of an individual worker should be indistinguishable from the average exposure of the total group (Boleij et al., 1995). The a priori strategy used to group workers in this study produced reasonable summary group exposure metric statistics for most EGs. However, since repeated samples were only taken for a minority of individual workers (10%) the between (bwR0.95) and within-worker variability for most EGs could not be fully evaluated as has been possible in other epidemiological studies investigating dose–disease response relationships for high-molecular weight allergens (Rappaport, 1991; Houba et al., 1997). The estimation of the within-worker variance is not crucial in this context since the calculation of elasticity is based on the between-group and within-group variances. However, the within-group variance may be somewhat underestimated since the within-worker variance was not fully incorporated in the calculations.
The environmental exposure assessment strategy sampled workers according to EGs rather than by department. This was done so as to minimize the potential for exposure misclassification if workers were to be sampled at the department level. However, the ANOVA models indicated that factory and department explained a significant proportion (adjusted r2 = 0.18, P < 0.05) of the variability in pilchard fish antigen exposure levels, than the a priori EG (adjusted r2 = 0.12, P = 0.115) (Table 4). Similarly, using the grouping statistics reported previously in other occupational settings, the highest contrast in exposure to particulates (elasticity = 0.38) and fish antigens (elasticity: pilchard = 0.42) was also obtained when workers were grouped according to a combination of factory and department (Table 6). This provides clear evidence that in this study factory and department characterized exposures more efficiently than the a priori EG classification, because in addition to having higher contrast, there are fewer groups, suggesting a higher precision. However, without information on the within-worker variability, the exact precision cannot be calculated. These data also suggest that should information on EGs or jobs not be available for all workers, especially in resource scarce settings where large number of samples and sophisticated immunological laboratory expertise may not be easily available (Jeebhay et al., 2000), sampling strategies that group workers per factory and department may yield reasonably valid exposure–response coefficients, although they may lack a certain degree of precision (Tielemans et al., 1998).
Improved understanding of exposure–response relationships in occupational respiratory allergy requires measurement of inhaled particulate aeroallergen exposures using particle-size-specific sampling relevant to their site of biological action (Tarlo and Purdham, 2002). This study has employed a number of approaches for exposure characterization of fish processing workers that have not been previously utilized, which should help to advance understanding of exposure–response relationships. These approaches include random selection of workers for sampling in each EG; use of personal sampling as opposed to area sampling techniques; measurement of the thoracic fraction range (considered to be the more biologically relevant fraction for both airway and alveolar disease processes) versus the total particulate mass of aerosols generated; longer-term measurements (over a full 8 h shift); sampling collection that included day and night shifts in both factories; measuring protein and specific fish antigen concentrations in addition to particulate mass determinations that are conventionally used. Furthermore, it has been demonstrated that pilchard antigen levels can be used as a surrogate measure of exposure to at least some other fish antigens and that sufficient contrast of exposure required for assessing exposure–response relationships could also be obtained by grouping workers by a combination of factory and department. These findings offer the promise of a more viable strategy for personal sampling to be conducted in resource scare settings such as the large number of small workplaces in South Africa (Jeebhay et al., 2000). This is also the first study demonstrating elevated fish antigen and probably endotoxin exposures in fish processing operations in general and fishmeal operations in particular.
Despite the aforementioned strengths, there were certain logistical and resource limitations to the study. These included the inability to compute variance components for assessing homogeneity of EGS due to insufficient number of repeated samples collected, lack of extensive data collected addressing intra-individual variability that may arise due to daily or seasonal variations in work processes and lack of information obtained on peak exposures since real-time sampling instruments were not used. These factors need to be considered in any future studies of environmental exposure characterization in fish processing plants.
|
CONCLUSIONS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
In conclusion, this study has demonstrated that workers involved in typical bony fish processing are at risk of inhaling aerosols containing fish antigens. Antigen exposures appear particularly high during fishmeal production and bagging. Grouping strategies based on department and factory may provide a more efficient approach than a priori classification of EGs for evaluating fish antigen exposures. Furthermore, the study demonstrated that successful application of ANOVA enabled the development of a sufficiently robust grouping strategy required for generating valid exposure estimates for investigating exposure–response relationships between fish antigen exposures and adverse allergic respiratory health outcomes among workers in fish processing plants.
|
ACKNOWLEDGEMENTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
The contribution of Mr R Beukes, Mr S Mbuli and Mr S Hassan from the Department of Health Sciences at Peninsula Technikon in the collection of the personal samples is acknowledged. The assistance of Dr Jeroen Douwes and Dr Inga Wouters from the Institute of Risk Assessment (University of Utrecht, The Netherlands) with the endotoxin analysis is acknowledged as well as Mr J Fernandez from Tulane Medical Centre for his assistance with the validation study of the fish antigen analysis. We would like to thank Ms B Fenemore and Ms A Elliott from the Allergology Unit at the University of Cape Town for the laboratory technical support. The assistance of Prof. H Kromhout (Institute of Risk Assessment, University of Utrecht), Prof. J Myers and Mr R Sayed (Occupational and Environmental Health Research Unit, University of Cape Town) for statistical support is also acknowledged. This study was supported by research grants from the Medical Research Council of South Africa and R01 Grant No. F002304 from NIOSH, CDC, USA. The contents of this paper are solely the responsibility of the authors and do not necessarily reflect the official views of these agencies.
Received April 8, 2004; in final form December 24, 2004
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
Aresery M, Lehrer SB. (2002) Occupational reactions to foods. Curr Allergy Asthma Rep; 2: 78–86.[Medline]
Beaudet N, Brodkin CA, Stover B et al. (2002) Crab allergen exposures aboard five crab-processing vessels. AIHA J; 63: 605–9.
Bland JM, Altman DG. (1986) Statistical methods for assessing agreement between two methods of clinical measurement. Lancet; 1: 307–10.[CrossRef][Web of Science][Medline]
Boleij JSM, Buringh E, Heederik D et al. (1995) Occupational hygiene of chemical and biological agents. Amsterdam, The Netherlands: Elsivier.
Crespo JF, Pascual C, Dominguez C et al. (1995) Allergic reactions associated with airborne fish particles in IgE-mediated fish hypersensitive patients. Allergy; 50: 257–61.[Web of Science][Medline]
Douglas JD, McSharry C, Blaikie L et al. (1995) Occupational asthma caused by automated salmon processing. Lancet; 346: 737–40.[CrossRef][Web of Science][Medline]
Douwes J, Pearce N, Heederik D. (2002) Does environmental endotoxin exposure prevent asthma? Review Thorax; 57: 86–90.
Dutch Expert Committee on Occupational Standards (DECOS). (1998) Endotoxins: health-based recommended exposure limit. A report of the Health Council of the Netherlands. Publication no. 03WG. pp. 1–15.
Elsayed S, Aas K. (1970) Characterization of a major allergen (cod.) chemical composition and immunological properties. Int Arch Allergy Appl Immunol; 38: 536–48.[Web of Science][Medline]
Gaddie J, Friend JAR. (1980) Pulmonary hypersensitivity in prawn workers. Lancet; 2: 1350–53.[Web of Science][Medline]
Griffin P, Roberts FP, Topping MD. (1994) Measurement of airborne antigens in a crab processing factory. Ann Occup Hyg; 38(Suppl. 1): 923–6.
Griffin P, Allan L, Gibson M et al. (2001). Measurement of personal exposure to aerosols of Nephrops norvegicus (scampi) using a monoclonal-based assay. Clin Exp Allergy; 31: 928–33.[CrossRef][Web of Science][Medline]
Heederik D, Braum W, Kromhout H. (1997) The influence of random exposure estimation error on the exposure-response relationship when grouping into homogeneous exposure categories. Occup Hyg; 3: 229–41.
Heederik D, Doekes G, Nieuwenhuijsen MJ. (1999) Exposure assessment of high molecular weight sensitisers: contribution to occupational epidemiology and disease prevention. Occup Environ Med; 56: 735–41.
Helsel DR. (1990) Less than obvious statistical treatment of data below the detection limit. Environ Sci Technol; 24: 1766–74.[CrossRef]
Hornung RW, Reed LD. (1990) Estimation of average concentration in the presence of non-detectable values. Appl Occup Environ Hyg; 5: 46–51.
Houba R, Heederik D, Kromhout H. (1997) Grouping strategies for exposure to inhalable dust, wheat allergens and 
-amylase allergens in bakeries. Ann Occup Hyg; 41: 287–96.
Jeebhay MF, Lopata AL, Robins TG. (2000) Seafood processing in South Africa: a study of working practices, occupational health services and allergic health problems in the industry. Occup Med (Lond); 50: 406–13.
Jeebhay MF, Robins TG, Lehrer SB et al. (2001) Occupational seafood allergy: a review. Occup Environ Med; 58:553–62.
Jeebhay MF, Robins TG, Lopata AL. (2004) World at work: fish processing workers. Occup Environ Med; 61: 471–4.
Kromhout H, Heederik D. (1995) Occupational epidemiology in the rubber industry: implications of exposure variability. Am J Ind Med; 27: 171–85.[Web of Science][Medline]
Kromhout H, Loomis DP, Mihlan GJ et al. (1995) Assessment and grouping occupational magnetic field exposure in five electric utility companies. Scand J Work Envir Health; 21: 43–50.
Lehrer SB. (1990) Hypersensitivity reactions in seafood workers. Allergy Proc; 11: 67–8.[Web of Science][Medline]
Lillienberg L, Bauer X, Doekes G et al. (2000). Comparison of four methods to assess fungal alpha-amylase in flour dust. Ann Occup Hyg; 44: 427–33.
Lippmann M. (1989) Size-selective health hazard sampling. In Hering SV, editor. Air sampling instruments for evaluation of atmospheric contaminants. 7th edn. Cincinnati: American Conference Government and Industrial Hygienists. pp. 163–98.
Luo S, Takano M, Takasaki Y et al. (2002). Measurement of low level membrane proteins using bicinchoninic acid: modified procedures to eliminate interfering substances. Fukuoka Igaku Zasshi; 93: 236–46.[Medline]
Malagie M, Jensen G, Graham JC et al. (1998) Food industry processes. In Stellman JM, editor. Encyclopedia of occupational health and safety. Vol. 67, 4th edn. Geneva: International Labour Office. pp. 2–7.
Malo JL, Cartier A. (1993) Occupational reactions in the seafood industry. Clin Rev Allergy; 11: 223–40.[Web of Science][Medline]
Marple V. (1989) PEM development, fabrication, evaluation and calibration. Final report to MSP corporation. Subcontract no. 3-321U-3933.
Moody MW, Roberts KJ, Huner JV. (1993) Phylogeny of commercially important seafood and description of the seafood industry. Clinical Rev Allergy; 11: 159–81.
Ortega HG, Berardinelli S. (1999) Health Hazard Evaluation Report No. 98-0069. Unisea, Dutch Harbour, Alaska: National Institute for Occupational Safety and Health.
Ortega HG, Daroowalla F, Petsonk EL et al. (2001) Respiratory symptoms among crab processing workers in Alaska: epidemiological and environmental assessment. Am J Ind Med; 39: 598–607.[CrossRef][Web of Science][Medline]
Rappaport SM. (1991) Review. Assessment of long-term exposures to toxic substances in air. Ann Occup Hyg; 35: 61–121.
Sakaguchi M, Toda M, Ebihara T et al. (2000) IgE antibody to fish gelatin (type 1 collagen) in patients with fish allergy. J Allergy Clin Immunol; 106: 579–84.[CrossRef][Web of Science][Medline]
Samuels SJ, Lemaster GK, Carson A. (1985) Statistical methods for describing occupational exposure measurements. Am Ind Hyg Assoc J; 46: 427–33.[Web of Science][Medline]
Sasaki D, Mitchell RA. (2002) How to obtain reproducible quantitative ELISA results. Oxford Biomedical Research, Inc. Available from: http://oxfordbiomed.com/News&Reviews/invrev6.htm
Sherson D, Hansen I, Sigsgaard T. (1989) Occupationally related respiratory symptoms in trout-processing workers. Allergy; 44: 336–41.[Web of Science][Medline]
Smith DD, Sechena R. (1998) Arctic snow crab-related lung disease. Int J Circumpolar Health; 57(Suppl. 1): 601–8.
SPSS Inc. (2001) SPSS for Windows: Release 11.0.1. Chicago: SPSS Inc.
Stata Corp. (2001) Stata Statistical Software: Release 7.0. College Station, TX: Stata Corporation.
Tarlo SM, Purdham J. (2002) Air sampling in occupational asthma. J Allergy Clin Immunol; 109: 603–5.[CrossRef][Web of Science][Medline]
Taylor AV, Swanson MC, Jones RT et al. (2000) Detection and quantitation of raw fish aeroallergens from an open-air fish market. J Allergy Clin Immunol; 105: 166–9.[CrossRef][Web of Science][Medline]
Tielemans E, Kupper LL, Kromhout H et al. (1998) Individual-based and group-based occupational exposure assessment: some equations to evaluate different strategies. Ann Occup Hyg; 42: 115–9.
Weytjens K, Cartier A, Malo JL et al. (1999) Aerosolized snow-crab allergens in a processing facility. Allergy; 54: 892–3.[CrossRef][Web of Science][Medline]
Willeke K, Baron P. (1993) Aerosol measurement: principles, techniques and applications. New York: Van Nostrand Reinhold. p. 209.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||