Friesen M C, Davies H W, Teschke K, Marion S, Demers P A
University of British Columbia, School of Occupational & Environmental Hygiene, Vancouver, Canada.
J Occup Environ Hyg. 2005 Dec;2(12):650-8. doi: 10.1080/15459620500391676.
Nonspecific dust measurements are used as a surrogate for wood dust exposure in sawmills. However, the wood dust component of dust has been found to vary by job and work area. Thus, the use of nonspecific dust exposure levels in epidemiologic studies may introduce exposure misclassification when assessing wood-related health effects. To improve exposure assessment in a retrospective cohort of 28,000 sawmill workers, we developed and evaluated the validity of two empirical models of exposure: one for nonspecific dust and one for wood dust. The dust model was constructed using 1,395 dust measurements collected in 13 sawmills for research or regulatory purposes from 1981 to 1997. Inter-sampler conversion factors were used to obtain inhalable dust equivalents if necessary. The wood dust model was constructed after applying adjustment factors to subtract nonwood components of the dust from the original measurements. The validity of the two models was tested against measurements (n = 213) from a technologically similar mill that was not part of the cohort study. The proportions of variability explained by the dust and wood dust models were 35% and 54%, respectively. When tested against the validation mill, the biases in the dust model were -33% for outdoor jobs and 2% for indoor jobs. The biases in the wood dust model were 2% for outdoor jobs and -3% for indoor jobs. Strong correlations were observed between the predicted and observed geometric means of jobs (0.79 and 0.70 for the dust model and wood dust model, respectively). Testing the validity of predictive models examines the generalizability of the models. The low overall bias, especially in the wood-specific model, increases our confidence in the use of these models for all sawmills to assess both nonspecific particulate and wood-related health effects in the historical cohort study.
非特异性粉尘测量被用作锯木厂木材粉尘暴露的替代指标。然而，已发现粉尘中的木材粉尘成分会因工作岗位和工作区域而有所不同。因此，在评估与木材相关的健康影响时，在流行病学研究中使用非特异性粉尘暴露水平可能会导致暴露分类错误。为了改进对一个由28000名锯木厂工人组成的回顾性队列的暴露评估，我们开发并评估了两种暴露经验模型的有效性：一种用于非特异性粉尘，另一种用于木材粉尘。粉尘模型是使用1981年至1997年期间在13家锯木厂为研究或监管目的收集的1395次粉尘测量数据构建的。如有必要，使用采样器间转换因子来获得可吸入粉尘当量。木材粉尘模型是在应用调整因子从原始测量值中减去粉尘中的非木材成分后构建的。这两种模型的有效性是根据来自一个技术上类似但不属于队列研究一部分的工厂的测量数据（n = 213）进行测试的。粉尘模型和木材粉尘模型解释的变异比例分别为35%和54%。在与验证工厂进行测试时，粉尘模型在户外工作中的偏差为 -33%，在室内工作中的偏差为2%。木材粉尘模型在户外工作中的偏差为2%，在室内工作中的偏差为 -3%。在预测的和观察到的工作几何均值之间观察到很强的相关性（粉尘模型和木材粉尘模型分别为0.79和0.70）。测试预测模型的有效性可检验模型的可推广性。总体偏差较低，尤其是在木材特异性模型中，这增加了我们在历史队列研究中使用这些模型对所有锯木厂评估非特异性颗粒物和与木材相关的健康影响的信心。
