Data Analysis for Helmholtz Zentrum München

The goal of the project was to quantify the relative importance of different metabolic pathways for coronary heart disease and type 2 diabetes. We analyzed data from the KORA study for which more than 15000 people are repeatedly medically examined since 1984. The main features of the analysis were 47 different biomarkers (see the image above) that represent different metabolic pathways. Additionally we controlled for several personal characteristics like sex or age. The target features were the incidence of the disease and the time until the incidence. Before the analysis missing data was imputed. We modeled the data using a Cox proportional hazards model. The case-cohort study design (Barlow 1994) was taken into account by weighing the data according to Barlow et al. (1999) upon request of our project partners. Because this weighing method is not supported by the R package survival (Therneau 2015) we had to implement the routine by ourselves. The estimation of the relative contribution of different biomarkers to the risk of the disease was done as proposed by Montonen et al. (2011). Finally, estimations from multiple imputations had to be combined following Rubin’s rules (Rubin 1987).

A paper using the results from this project was published in March 2020 (Huth et al “Biomarker-defined pathways for incident type 2 diabetes and coronary heart disease”).