Job-exposure matrices addressing lifestyle factors.
Job-exposure matrices addressing lifestyle factors.
复制标题
解决生活方式因素的工作暴露矩阵。
DOI：
10.1136/oemed-2018-105425
复制
发表时间：
2018
影响因子：
4.9
通讯作者：
Friesen,MelissaC
中科院分区：
文献类型：
--
作者：
Friesen,MelissaC
本期《职业与环境医学》描述了 Bondo Petersen 等人开发的一种新颖的工作暴露矩阵 (JEM)，该矩阵旨在描述生活方式风险因素中特定工作的差异，1 旨在在丹麦癌症登记研究的病因学分析中提供生活方式调整。这些生活方式特征包括吸烟、闲暇时体力活动、酒精饮料摄入量、体重指数（BMI）以及水果和蔬菜摄入量。这些数据驱动的生活方式 JEM 是使用三十年来收集的超过 25 万丹麦工人生活方式特征的高质量个人调查数据汇总而开发的。据我所知，仅存在一套类似的生活方式 JEM：2005 年，芬兰工作暴露矩阵 (FINJEM) 根据 1993 年至 1999 年期间对芬兰成人健康行为进行的调查，添加了 1995 年至 1997 年期间针对类似生活方式特征的职业和特定性别 JEM 估计值。2 丹麦生活方式 JEM 是第一个提供多个时间段估计值的国家，采用了超过 30 年的 5 年时间窗口。此外，使用混合效应统计模型对丹麦调查数据进行了系统分析，该模型将职业分类代码作为随机效应，以提供特定于工作组的估计值，并将年龄、时间段、性别和数据源作为固定效应，以解释特定年龄、特定年份和特定性别的差异。该建模方法类似于根据工作场所暴露测量开发职业暴露 JEM 的框架。 3 4 将工作组纳入随机效应，采用高效且保守的估计程序，当数据稀疏和/或异质时，将工作组预测缩小到总体均值；而当数据更丰富且变量较少时，将工作组预测缩小到工作组均值。尽管这种生活方式的 JEM 是使用已建立的 JEM 建模框架建立在高质量数据的基础上，但它与为任何职业暴露而开发的 JEM 具有相同的重要局限性。 5 最值得注意的是无法解释工作中发生的暴露变化。丹麦生活方式 JEM 的组内相关性非常低 (0.3%–7%)，这一点就很明显。 Bondo Petersen 等人还发现工作组之间的对比度较低，四分位间对比度（第 75 个百分位数除以第 25 个百分位数）范围为 1.1 至 1.6。第 95 个百分位数与第 5 个百分位数（1.1 至 2.8）的比较以及最大值与最小值（1.2 至 6.9）的比较的对比度有所增加。工作组之间的最大对比是在工人吸烟比例和吸烟量上观察到的，酒精饮料摄入量和休闲时间体力活动的对比为低到中等，BMI和水果和蔬菜消耗的对比非常低。工作组之间的对比与工作组内变异性相比较低，表明使用工作组估计值与个人估计值相比存在很大的测量误差，并且生活方式JEM可能不足以完全控制混杂因素。然而，在缺乏个人信息的基于登记的研究的病因学分析中，使用这些生活方式JEM来控制与生活方式相关的混杂因素，与未经调整的分析或使用来自研究人群子集或其他研究的吸烟和其他混杂因素的估计接触层特定患病率的间接调整相比，具有潜在的改进。作者证明了吸烟 JEM 的内部有效性，具有单调暴露……
This issue of Occupational and Environmental Medicine includes the description of a novel job-exposure matrix (JEM) designed to characterise job-specific differences in lifestyle risk factors developed by Bondo Petersen et al, 1 with the aim to provide lifestyle adjustment in aetiological analyses in Danish cancer registry-based studies. These lifestyle characteristics included smoking, leisure time physical activity, alcoholic beverage intake, body mass index (BMI) and fruit and vegetable consumption. These data-driven lifestyle JEMs were developed using pooled high-quality individual survey data collected over three decades on lifestyle characteristics from more than a quarter-million Danish workers. Only one similar set of lifestyle JEMs exist to my knowledge: in 2005, the Finish job-exposure matrix (FINJEM) added occupation and gender-specific JEM estimates for the time period 1995–1997 for similar lifestyle characteristics based on surveys on Finnish adult health behaviours conducted between 1993 and 1999. 2 The Danish lifestyle JEMs are the first to provide estimates for multiple time periods, employing 5-year time windows over three decades. Additionally, the Danish survey data were systematically analysed using mixed-effects statistical models that incorporated occupation classification code as a random effect to provide job group-specific estimates and incorporated age, time period, gender and data source as fixed effects to account for age-specific, year-specific and gender-specific differences. The modelling approach was similar to the framework used to developed occupational exposure JEMs from workplace exposure measurements. 3 4 Incorporating job group as a random effect employs an efficient and conservative estimation procedure that shrinks the job group prediction towards the population mean when data were sparse and/or heterogeneous and towards the job group mean when the data were more plentiful and less variable. Although built on high-quality data using an established modelling framework for JEMs, this lifestyle JEM has the same important limitations of JEMs developed for any occupational exposure. 5 The most notable is the inability to account for variability in exposures that occur within a job. This is evident by the very low intraclass correlations (0.3%–7%) observed for the Danish lifestyle JEMs. Bondo Petersen et al also found low between job group contrast, with an interquartile contrast (75th percentile divided by 25th percentile) ranging from 1.1 to 1.6. Contrasts increased for comparisons of the 95th to fifth percentile (1.1 to 2.8) and comparisons of the maximum to minimum (1.2 to 6.9). The greatest between job group contrast was observed for both the proportion of workers smoking and the amount of smoking, with low to moderate contrast for alcoholic beverage intake and leisure time physical activity and very low contrast for BMI and fruit and vegetable consumption.The low between job group contrast compared with within job group variability suggests substantial measurement error from use of job group estimates over individual estimates and that the lifestyle JEMs may be insufficient to fully control for confounding. However, the use of these lifestyle JEMs to control for lifestyle-related confounding in aetiological analyses of register-based studies where individual information is lacking represents a potential improvement over unadjusted analyses or indirect adjustment using estimated exposure strata-specific prevalence of smoking and other confounders from subsets of a study population or from other studies. The authors demonstrated internal validity for the smoking JEM, with a monotonic exposure …