Keywords: bimodal distribution, diabetes mellitus, diagnostic guidelines, Gaussian distribution, incidence, prevalence, risk assessment, uncertainty, unimodal distribution, variability, disease risk, health risks, medical diagnosis, biochemical data, analytical error, biochemical measurements
The impact of variability in the risk of disease exemplified by diagnosing diabetes mellitus based on ADA and WHO criteria as gold standard
The diagnosis of clinically mute diabetes mellitus (DM) is based on a nominal concentration of fasting peripheral venous plasma glucose (f-vPG) at ≥7.0mmol/l, to be exceeded by first and confirmatory measurement. Aim: to identify error sources in the clinical use of biochemical data, to use the diagnostic concept for clinically mute DM. Materials and methods: error contributors were established. We estimated the probability of f-vPG >7.0 as a function of the biological set point, and the biological variation (CVwithinsubject=4.9%), the CLIA regulations and the ISO 15197 standard for total error (TE) (CLIA TE±10% and ISO 15197 TE±20%). Results: The probability of getting a single f-vPG measurement ≥7.0mmol/l as a function of the biological set-point, and the probability when introducing a 'confirmatory' test was investigated, based on algorithms. Half of those with a biological set-point of 7.0 mmol/l are measured ≥7.0mmol/l, and half of those are confirmed diabetic. The effects of analytical error types are different and need careful distinction.