Statistical Analysis

Statistical analysis of data from experimental and clinical studies performed as part of the research portfolio of CBI is connected with various fields of medical research.

Recently, the members of the CBI became involved in uncovering the links between genetic profile of a subject and his or her predisposition for the incidence of a primary cardiovascular event, namely a primary acute myocardial infarction or a stroke. One of the goals of this research is to formulate the set of genes which based on the results from gene expression profiling would prove predictive in terms of the incidence of (and/or mortality from) such primary cardiac events in the Czech population. First results of this research were presented at the 30th Annual Conference of the International Society for Clinical Biostatistics (ISCB), which took place 23-27 August 2009 in Prague.
Preventing allergy incidence in Czech infants is another example of interdisciplinary research performed at our centers. This research aimed at reducing the incidence of allergies in children of allergic mothers via colonizing the bowel flora of newborn children with E. coli bacteria O83. Statistical methods used in this research included the Kaplan-Meier survival analysis, complemented with Cox proportional hazards model, Aalen’s linear and Gray’s time-varying coefficients model.
Effective use of statistical methodology in analyzing medical data is of particular interest to medical statisticians practicing at our centers. As an example, we considered a problem of analyzing the association of pre-clinical atherosclerosis, defined on the basis of ultra-sound detected plaques in two femoral and six carotid arteries, with the risk factors for cardiovascular disease in perimenopausal women from Prague 4 district, a study discussed in the related paper. The study pointed at the advantage of employing proportional odds logistic regression model over ordinary logistic regression in situations when we are dealing with limited uncertainty in classifying clinical outcome as a binary variable.
Prognostic markers of subclinical atherosclerosis in middle-aged women and their relation to menopause was a subject of the population study conducted in cooperation with our Centre. Analysis of Covariance led to uncovering a modifying effect of reproductive status in women smokers. Consequently, perimenopause, also called “late menopausal transition”, was identified as the period when atherosclerotic processes in women smokers accelerate the most. Furthermore, a gene polymorphism of gap-junctional protein Connexin37 significantly modified the effect of hyperglycaemia. Women carrying TT polymorphism experienced highly significant protection from atherogenic effect of higher fasting glycaemia.
The breadth of scope of interdisciplinary research conducted by our research teams may be further illustrated with a study focusing on quality of life (QoL) in Crohn’s disease patients. Here we emphasized importance of domain (sub-scale) approach to analyzing the data and have shown advantages of employing multivariate analysis of the domain-specific QoL data.