Search

Objective: To determine whether there are clinical and public health dilemmas resulting from the reproducibility of routine vitamin D assays. Methods: Blinded agreement studies were conducted in eight clinical laboratories using two commonly used assays to measure serum 25-hydroxyvitamin D (25(OH)D) levels in Australasia and Canada (DiaSorin Radioimmunoassay (RIA) and DiaSorin LIAISON® one). Results: Only one laboratory measured 25(OH)D with excellent precision. Replicate 25(OH)D measurements varied by up to 97% and 15% of paired results differed by more than 50%. Thirteen percent of subjects received one result indicating insufficiency [25-50 nmol/l] and another suggesting adequacy [>50 nmol/l]). Agreement ranged from poor to excellent for laboratories using the manual RIA, while the precision of the semi-automated Liaison® system was consistently poor. Conclusions: Recent interest in the relevance of vitamin D to human health has increased demand for 25(OH)D testing and associated costs. Our results suggest clinicians and public health authorities are making decisions about treatment or changes to public health policy based on imprecise data. Clinicians, researchers and policy makers should be made aware of the imprecision of current 25(OH)D testing so that they exercise caution when using these assays for clinical practice, and when interpreting the findings of epidemiological studies based on vitamin D levels measured using these assays. Development of a rapid, reproducible, accurate and robust assay should be a priority due to interest in populationbased screening programs and research to inform public health policy about the amount of sun exposure required for human health. In the interim, 25(OH)D results should routinely include a statement of measurement uncertainty.

Decision tree induction has been studied extensively in machine learning as a solution for classification problems. The way the linear decision trees partition the search space is found to be comprehensible and hence appealing to data modelers. Comprehensibility is an important aspect of models used in medical data mining as it determines model credibility and even acceptability. In the practical sense though, inordinately long decision trees compounded by replication problems detracts from comprehensibility. This demerit can be partially attributed to their rigid structure that is unable to handle complex non-linear or/and continuous data. To address this issue we introduce a novel hybrid multivariate decision tree composed of polynomial, fuzzy and decision tree structures. The polynomial nature of these multivariate trees enable them to perform well in non-linear territory while the fuzzy members are used to squash continuous variables. By trading-off comprehensibility and performance using a multi-objective genetic programming optimization algorithm, we can induce polynomial-fuzzy decision trees (PFDT) that are smaller, more compact and of better performance than their linear decision tree (LDT) counterparts. In this paper we discuss the structural differences between PFDT and LDT (C4.5) and compare the size and performance of their models using medical data