|Home | About | Journals | Submit | Contact Us | Français|
Y. Jin, J. A. Deddens. University of Cincinnati
ObjectivesOccupational studies often involve repeated measures of log‐normally distributed exposure data that contain values below the limit of detection (LOD). We will illustrate new methods for dealing with such data. Results will be contrasted with the traditional methods of substituting LOD/2 or LOD/sqrt(2) for values below the LOD. We will also illustrate new methods for computing the correlation of two variables both of which have values below the LOD. Simulations results will be given as well as examples from occupational studies (eg, pesticide exposure).
MethodsTraditionally when analysing exposure data with values below the LOD, one substitutes LOD/2 or LOD/sqrt(2) for values below the LOD and then uses standard methods for computing correlations, and uses usual mixed linear models for repeated measures exposure data. For data without repeated measures it is possible to obtain maximum likelihood estimates using PROC LIFEREG by treating the values below the LOD as being left censored. Recently Lyles et al (Biometrics 2001;57:1238–44) showed how to find the maximum likelihood of the correlation between two left censored variables. Theibaut et al (Comput Methods Programs Biomed 74;2004:255–260) has showed how to use PROC NLMIXED to perform maximum likelihood estimation of repeated left censored data with one random effect. These methods were developed for AIDS/HIV data and do not seem to be widely known in the occupational health literature. We have also developed SAS MACROs for performing Bayesian estimation of random effect linear models with values below the LOD.
ResultsThe NLMIXED method yields maximum likelihood estimates of parameters in linear mixed models. Simulations show that maximum likelihood methods provide excellent parameter estimates, correlation estimates, and tests of hypotheses. Bayesian methods also provide excellent estimates and tests of hypotheses. In contrast, the traditional methods of substituting LOD/2 provide very biased estimates of both geometric means and standard deviations, although the tests of hypotheses are not as bad.
ConclusionIt is now possible to obtain maximum likelihood estimates of parameters in linear mixed models with left censored data with one random effect using NLMIXED, and to obtain maximum likelihood estimates of correlations of left censored data.
Key wordsexposure; limit of detection; repeated measures
N. Saejiw1, N. Chaiear2, J. Ngoencharee3, S. Sadhra4. 1Institute of Allied Health Science and Public Health, Walailak University; 2Unit of Occupational Medicine, Faculty of Medicine, Khon Kaen University; 3Faculty of Medicine, Khon Kaen University; 4Institute of Occupational and Environmental Medicine, University of Birmingham
ObjectivesExposure assessment was conducted in a rubber wood sawmill to determine occupational exposure to particulates in terms of the concentrations and particle size in the rubber wood sawmills industry.
MethodsThis report was part of a cross‐sectional study on wood dust exposure and respiratory health effects in rubber wood sawmills. It was conducted in Nakhon Sri Thammarat which has one of the largest rubber wood sawmills in Thailand. All workers (n=340) from all jobs on a day shift were recruited for personal sampling for wood dust. Personal inhalable dust (n=742) and respirable dust (n=241) full shift samples were collected from 27 job titles. These data were used to classify workers into high, moderate and low exposure groups based on the concentration found in each job. Static samples were also collected to determine particle size distribution. Geometric mean (GM) was used to present the concentration of the rubber wood dust.
ResultsThe inhalable dust concentrations were clearly high with a range between <0.18 and 59.4 mg/m3 and a GM of 4.7 mg/m3. The GM of inhalable dust in each job title enabled us to classify the workers into three exposure groups: high exposure, >5 mg/m3; moderate exposure, 2–5 mg/m3; and low exposure, <2 mg/m3. Among the high exposure group, the highest GM inhalable dust concentrations were found in sawing green lumber (12.8 mg/m3) and cutting dry lumber (7.3 mg/m3). The respirable dust concentrations were low and in the range of <0.14 to 6.0 mg/m3 with GM 0.3 mg/m3. The largest percentage of dust in major operations was found in the thoracic fraction: 50% cut‐off diameter smaller than 9 µm. The significant determinants of personal inhalable dust exposure were job titles, type of machine used, wood processed and local exhausted ventilation.
ConclusionThe highest exposures likely occurred in sawing green lumber and cutting dry lumber. The size distribution of wood dust indicated a high proportion in the thoracic fraction. Job titles, types of machine used, wood processed and local exhausted ventilation were associated with personal inhalable dust exposure.
Key wordsinhalable dust; wood dust; sawmill
J. B. Coble1, R. C. H. Vermeulen2, B. T. Ji1, S. Xue1, M. Dosemeci1, W. Lu3, W. Zheng4, Y. T. Gao5, A. Blair1, W. H. Chow1, N. Rothman1. 1Occupational and Environmental Epidemiology Branch, National Cancer Institute; 2University of Utrecht; 3Shanghai CDC; 4Vanderbilt University; 5Shanghai Cancer Institute
ObjectivesOccupational exposure to benzene was assessed in a population‐based cohort study of 74942 women from Shanghai, China, based on industry and job titles obtained using work history questionnaires.
MethodsJobs with a potential for exposure to benzene were identified using two job exposure matrices (JEM), one by industry title and one by occupational title. A database containing over 70000 benzene measurements from factories in Shanghai compiled from historic monitoring was used to determine average benzene concentrations by JEM intensity rating and time period. A regression analysis revealed substantial reductions by time period in the benzene concentrations measured. The exposure estimates from the regression analysis were merged with the work histories to develop a set of quantitative exposure metrics for each subject that included lifetime duration of exposure, year of first exposure, lifetime average exposure, and lifetime cumulative exposure.
ResultsOver 99% of the women enrolled in this study provided work histories with a wide variety of occupations, of which 48% were in manufacturing industries, including textiles, electronics, metal fabrication and chemical production. The median duration of employment was 29 years, and over 90% began working prior to 1975. Various sets of exposure criteria based on the JEM ratings and the number of measurements were investigated to identify jobs with a moderate to high potential for exposure to benzene. When applying the least restrictive criteria to maximise sensitivity, the estimated prevalence of exposure was 22% with an average exposure level of 34.9 mg/m3. The use of more restrictive criteria to increase specificity resulted in an estimated prevalence of less than 2% with an average exposure level of 91.4 mg/m3.
ConclusionA tiered approach to estimate ranges in exposure prevalence and intensity reflects uncertainty inherent in the assignment of exposure levels based on industry and job titles, and provides flexibility to investigate disease risks associated with occupational exposure to benzene using both semi‐quantitative and quantitative estimates of exposure.
Key wordsbenzene; exposure; Shanghai
H. Kromhout1, W. Fransman1, F. de Vocht1, B. van Wendel de Joode2. 1IRAS, Utrecht University; 2IRET, UNA
ObjectivesDermal exposure to bitumen fume condensate was assessed semi‐quantitatively to facilitate development of a dermal exposure matrix to be applied in a nested case‐control study on lung cancer among European asphalt workers.
MethodsObservations were made with a recently developed observational method for dermal exposure assessment (DREAM). Two trained observers applied this method while observing outdoor pavers and indoor mastic workers from nine companies in four European countries. In total 18 jobs/tasks were evaluated for 76 individual workers. Linear mixed models were used to look for determinants of dermal exposure to bitumen condensate.
ResultsThe variability in DREAM scores was larger for pavers than for mastic workers. This was related to the fact that pavers showed more variability in use of (protective) clothing. Mastic workers had on average a 2‐fold higher actual dermal exposure at their hands than the highest exposed pavers (rakers and screed men). Within a paving crew, the maximum differences in actual dermal hand exposure were 6–7‐fold. Company (22%) and task (62%) together explained 84% of variability in actual hand exposure among pavers. For mastic workers, the differences in actual hand exposure were relatively small (maximum 4–5‐fold), and were smaller for actual exposure of the body (maximum 2–3‐fold). Differences between crews and circumstances were larger. Company and task together explained 50% of variability in actual hand exposure, with company alone accounting for 45%.
ConclusionThe results of the DREAM observations together with limited amounts of dermal measurement data available from the literature will at most be sufficient to create a semi‐quantitative dermal exposure matrix with a detailed job/task axis. A time axis that will allow for historical changes in dermal exposure will most likely not be feasible. On the other hand, incorporating modifiers based on these DREAM observations like glove use, hygienic behaviour and solvent use will most likely be feasible and will allow for assessment of inter‐individual differences in dermal exposure.
Key wordsasphalt workers; dermal exposure; bitumen
R. L. Neitzel1, W. E. Daniell1, H. W. Davies2, L. Sheppard3, N. S. Seixas1. 1Department of Environmental and Occupational Health Sciences, University of Washington; 2School of Occupational and Environmental Hygiene, University of British Columbia; 3Department of Biostatistics, University of Washington
ObjectivesIncorporation of subjective noise levels into exposure estimates may improve the accuracy of epidemiological exposure assessments, which remain challenging for workers with dynamic noise exposures. The current study explored the relationship between perceived and measured noise exposure.
MethodsTwenty workers at each of three worksites with different exposure profiles (continuous, intermittent and highly variable noise) were evaluated. Mean site noise levels were within 2.6 dBA of one another. Dosimetry measurements were made on each subject on 4 days over a 2‐week period, and subjects simultaneously reported perceived noise levels on a time‐line. Subjects also completed surveys on the first and last days of the study which included six alternative categorical questions on exposure intensity. Quantitative exposure levels were calculated for each subject based on their individual and trade averages. Perceived noise exposure was based on the categorical questionnaire responses, and combined with measurements or a priori assumptions to obtain survey‐based exposure estimates. Relationships between the survey‐ and measurement‐based levels were examined. Taking the individual‐based mean exposure as the true level, the bias and precision of the subjective levels were compared with those of the trade levels.
ResultsExposure levels increased with perceived noise level categories from dosimetry time‐line reporting. However, exposure levels did not increase with perceived noise level categories for most survey questions. When questions were collapsed into binary exposure categories, positive predictive values for identifying workers exposed above 85 dBA were high (0.74–0.81), while negative predictive values were lower (0.20–0.60). The bias and precision of survey‐based exposure estimates compared to the individual measured exposures were moderate, with bias ranging from 0.25 to −6.8 dBA and precision from 2.2 to 6.6 dBA, depending on the question. Compared to the trade mean exposures, bias and precision of the perception‐based estimates were generally worse by 0.5–3.5 dBA, though in some cases precision was better. One question assessing exposure relative to usual levels demonstrated the potential to be combined with trade‐specific averages to control measurement error.
ConclusionUse of subjective noise exposure questions has the potential to improve the accuracy of quantitative exposure estimates.
Key wordsexposure assessment; subjective exposure estimation; noise‐induced hearing loss
K. Teschke1, P. Johnson2, C. Trask1, Y. Chow1, J. Village1, M. Koehoorn1. 1University of British Columbia; 2University of Washington
ObjectivesThere have been numerous studies of the occupational causes of back disorders, but controversy remains about whether chronic exposures are aetiological, in part because of the difficulty of measuring potential risk factors in large numbers of subjects over long durations in a variety of working conditions. We conducted a study in five heavy industries in Western Canada to evaluate three methods that might facilitate exposure assessment for back injury studies. Here we report measurements of one risk factor, trunk posture.
MethodsWe used the following methods: a portable data‐logging inclinometer collecting posture data at 7.6 Hz for the full shift; observations of trunk angles once per minute by trained observers for the full shift; and end‐of‐shift reports by study subjects of the proportion of the shift spent in various postures. 126 subjects participated on 1 or 2 measurement days each, for a total of 223 person‐shifts. To compare methods, observations and self‐reports of risk factors were modelled as determinants of postures measured by the inclinometer.
ResultsAll three methods were successful in the demanding environments of 50 heavy industry worksites (data complete for >97% of shifts). Inclinometry provided measures of flexion–extension (mean 17°, SD 11.2°), lateral flexion (mean 8.5°, SD 2.6°), and trunk movement speed (mean 14.3°s−1, SD 4.9°s−1). The highest exposures were observed in the construction industry, with smaller differences between the remaining four (highest to lowest): warehousing; forestry; wood and wood products; and transportation. The observational data explained 61% of the variance in flexion–extension, 30% in lateral flexion, and 47% in trunk movement speed. The self‐reported data explained 40% of the variance in flexion–extension, 34% in lateral flexion, and 34% in trunk movement speed.
ConclusionThe inclinometer performed well in a wide variety of heavy industrial sites and did not require full‐time personnel after set‐up. Observations, while time intensive, had good predictive power for flexion–extension and trunk movement speed, but lateral flexion was not as well documented. The observation tool used had the advantage of simultaneously collecting data on other potential back injury risk factors, including manual materials handling and driving. Self‐reports were less accurate in predicting measured exposures.
Key wordsback disorders; posture; exposure assessment