Search

Due to the wealth of exposome data from longitudinal cohort studies that is currently available, the need for methods to adequately analyze these data is growing. We propose an approach in which machine learning is used to identify longitudinal exposome-related predictors of health, and illustrate its potential through an application. Our application involves studying the relation between exposome and self-perceived health based on the 30-year running Doetinchem Cohort Study. Random Forest (RF) was used to identify the strongest predictors due to its favorable prediction performance in prior research. The relation between predictors and outcome was visualized with partial dependence and accumulated local effects plots. To facilitate interpretation, exposures were summarized by expressing them as the average exposure and average trend over time. The RF model's ability to discriminate poor from good self-perceived health was acceptable (Area-Under-the-Curve = 0.707). Nine exposures from different exposome-related domains were largely responsible for the model's performance, while 87 exposures seemed to contribute little to the performance. Our approach demonstrates that ML can be interpreted more than widely believed, and can be applied to identify important longitudinal predictors of health over the life course in studies with repeated measures of exposure. The approach is context-independent and broadly applicable. [ABSTRACT FROM AUTHOR]

Determination of the level of exposure during and after a chemical incident is crucial for the assessment of public health risks and for appropriate medical treatment, as well as for subsequent health studies that may be part of disaster management. Immediately after such an incident, there is usually no opportunity to collect reliable quantitative information on personal exposures and environmental concentrations may fall below detectable levels shortly after the incident has passed. However, many substances persist longer in biological tissues and thus biological monitoring strategies may have the potential to support exposure assessment, as part of health studies, even after the acute phase of a chemical incident is over. Reported successful applications involve very persistent chemical substances such as protein adducts and include those rare cases in which biological tissues were collected within a few hours after an incident. The persistence of a biomarker in biological tissues, the mechanism of toxicity, and the sensitivity of the analysis of a biomarker were identified as the key parameters to support a decision on the feasibility and usefulness of biological monitoring to be applied after an incident involving the release of hazardous chemicals. These input parameters could be retrieved from published methods on applications of biomarkers. Methods for rapid decision making on the usefulness and feasibility of using biological monitoring are needed. In this contribution, a stepwise procedure for taking such a decision is proposed. The persistence of a biomarker in biological tissues, the mechanism of toxicity, and the sensitivity of the analysis of a biomarker were identified as the key parameters to support such a decision. The procedure proposed for decision making is illustrated by case studies based on two documented chemical incidents in the Netherlands. [ABSTRACT FROM AUTHOR]

Exposure assessment during and after acute chemical incidents and disasters is essential for health studies that may follow. During chemical incidents, the focus usually lies on risk assessment and afterward attention shifts toward possible (long-term) health effects. This may lead to insufficient available data on exposure to study the association between exposure and health outcome, and collection of additional exposure data is often required. Literature on health studies conducted after several chemical incidents was reviewed to obtain better insight on the needs of health studies. Four different types of scenarios were distinguished based on when exposure data were collected and the exposure data used for health studies. These four scenarios gave insight on exposure data needed for conclusive health studies and when different methods of exposure data collection should be used. Literature indicated that adequate and rapid exposure assessment during chemical incidents is vital for health studies, because data that are not collected during or directly after an incident may be irretrievably lost. Poor exposure assessment is not always the only problem in health studies. Problems in health studies including poor exposure assessment may be prevented when the general design and needs of health studies are taken into account when designing contingency plans. Together with measures that will help facilitate funding, design, and coordination of health studies, disaster management programs should, among others, prepare for methods that lead to a swift identification of released substances, determination of concentrations and dispersion of released substances, designing basic questionnaire outlines, and rapid evaluation of the usefulness and necessity of employing biological sampling.Journal of Exposure Science and Environmental Epidemiology (2008) 18, 341–359; doi:10.1038/jes.2008.23; published online 7 May 2008 [ABSTRACT FROM AUTHOR]

Epidemiological studies have established an association between outdoor levels of fine particles (PM2.5) and cardiovascular health. However, there is little information on the determinants of PM2.5 exposures among persons with cardiovascular disease, a potentially susceptible population group. Daily outdoor, indoor and personal PM2.5 and absorbance (proxy for elemental carbon) concentrations were measured among elderly subjects with cardiovascular disease in Amsterdam, the Netherlands, and Helsinki, Finland, during the winter and spring of 1998–1999 within the framework of the ULTRA study. There were 37 non-smoking subjects in Amsterdam and 47 in Helsinki. In Amsterdam, where there were enough exposure events for analyses, exposure to environmental tobacco smoke (ETS) indoors was a major source of between-subject variation in PM2.5 exposures, and a strong determinant of PM2.5 and absorbance exposures. When the days with ETS were excluded, within-subject variation accounted for 89% of the total variation in personal PM2.5 and 97% in absorbance in Amsterdam. The respective figures were 66% and 61% in Helsinki. In both cities, outdoor levels of PM2.5 and absorbance were major determinants of personal and indoor levels. Traffic was also an important determinant of absorbance: living near a major street increased exposure by 22%, and every hour spent in a motor vehicle by 13% in Amsterdam. The respective increases were 37% and 9% in Helsinki. Cooking was associated with increased levels of both absorbance and PM2.5. Our results demonstrate that by using questionnaires in connection with outdoor measurements, exposure estimation of PM2.5 and its combustion originating fraction can be improved among elderly persons with compromised health.Journal of Exposure Science and Environmental Epidemiology (2007) 17, 124–133. doi:10.1038/sj.jes.7500470; published online 25 January 2006 [ABSTRACT FROM AUTHOR]