On estimating biological variation in diagnostic tests: application to the oral glucose tolerance test.

Biological variation (BV) in diagnostic tests can be conveniently estimated by the equation, BV = magnitude of reference limit - reference median magnitude of - 2(SD)A, where "reference limit" refers to either the 2.5th or 97.5th percentile in the reference population, magnitude of indicates absolute value, and (SD)A is the standard deviation of random analytical variation at the reference median. The value of (SD)A is calculated from the equation, (SD)A = (CV)A (reference median)/100, where (CV)A, the coefficient of variation of random analytical variation, is obtained from routine stable quality control material. The BV was calculated for plasma glucose concentration at the time points in the oral glucose tolerance test in an asymptomatic reference population and found to vary in the order: fasting less than 3 hour less than 1/2 hour less than 1 hour less than 2 hour. We present correlation coefficients between subject age and plasma glucose concentration that suggest that BV at the fasting, 1/2, 1, and 2 hour points might be reduced by subdividing reference populations according to subject age.