Occupational and Environmental Health Research Group, University of Manchester, Room C4.13, Ellen Wilkinson Building, Oxford Road, Manchester M13 9PL, UK.
Occup Med (Lond). 2010 Mar;60(2):115-20. doi: 10.1093/occmed/kqp168. Epub 2009 Dec 2.
There is no agreed protocol for the prediction of low molecular weight (LMW) respiratory sensitizers. This creates challenges for occupational physicians responsible for the health of workforces using novel chemicals and respiratory physicians investigating cases of occupational asthma caused by novel asthmagens.
To iterate the external validation of a previously published quantitative structure-activity relationship (QSAR) model for the prediction of novel chemical respiratory sensitizers and to better characterize its predictive accuracy.
An external validation set of control chemicals was identified from the Australian Hazardous Substances Information System. An external validation set of asthmagenic chemicals was identified by a thorough search of the peer-reviewed literature from January 1995 onwards using the Medline database. The QSAR model was used to determine an 'asthma hazard index' (between 0 and 1) for each chemical.
A total of 28 external validation asthmagens and 129 control chemicals were identified. The area under the receiver operating characteristic (ROC) curve for the model's ability to distinguish asthmagens from controls was 0.87 (95% CI 0.76-0.97). Using a cut-off hazard index of 0.5 resulted in sensitivity of 79% and specificity of 93%. For prior probability ranging from 1:300 to 1:100, the negative predictive value (NPV) was 1 and positive predictive value (PPV) 0.04-0.1 while for prior probability ranging from 1:20 to 1:3, the NPV was 0.91-0.99 and PPV 0.39-0.85.
The ROC curve for this QSAR demonstrates good global predictive power for distinguishing asthmagenic from non-asthmagenic LMW organic compounds. Potential for utilization by occupational and respiratory physicians is evident from its predictive values.
目前尚无用于预测低分子重量（LMW）呼吸致敏剂的既定方案。这为负责使用新型化学物质的劳动力健康的职业医生和调查由新型变应原引起的职业性哮喘病例的呼吸内科医生带来了挑战。
迭代验证先前发表的用于预测新型化学呼吸致敏剂的定量构效关系（QSAR）模型，并更好地描述其预测准确性。
从澳大利亚危险物质信息系统中确定了一组控制化学品的外部验证集。通过在 Medline 数据库中从 1995 年 1 月开始全面搜索同行评议文献，确定了一组哮喘原性化学品的外部验证集。使用 QSAR 模型确定每种化学物质的“哮喘危害指数”（0 至 1 之间）。
共确定了 28 种外部验证哮喘原和 129 种对照化学品。该模型区分哮喘原和对照物的能力的接收器操作特征（ROC）曲线下面积为 0.87（95%CI 0.76-0.97）。使用危害指数的截止值为 0.5，敏感性为 79%，特异性为 93%。对于先验概率范围为 1:300 至 1:100，阴性预测值（NPV）为 1，阳性预测值（PPV）为 0.04-0.1，而对于先验概率范围为 1:20 至 1:3，NPV 为 0.91-0.99，PPV 为 0.39-0.85。
该 QSAR 的 ROC 曲线显示出良好的全局预测能力，可用于区分哮喘原性和非哮喘原性低分子有机化合物。从其预测值可以看出，它对职业和呼吸内科医生具有潜在的利用价值。
