Accuracy and Precision in Measurement

VIM relates accuracy to closeness to a true value and precision to agreement among repeated measurements under specified conditions. In that terminology, accuracy is not itself a quantity assigned a percentage. VIM: accuracy, VIM: precision.

When a figure is presented as accuracy, look for the operation that produced it: recovery, deviation from a reference or another comparison.

A C-peptide standardization study examined serum samples using different methods and laboratories, with repetitions distributed across days. Variability was not uniform: it depended on the system and level of comparison.

The design shows why several consecutive readings do not automatically describe everyday performance. When reading a coefficient of variation, identify what remained constant and what changed. An instrumental repetition, a new preparation and another working day do not necessarily represent the same sources of variation. 2008 study.

Another investigation compared C-peptide immunoassays with an isotope-dilution mass-spectrometry procedure. The methods showed good correlation but not sufficient agreement or equivalent trueness relative to the reference procedure.

This explains a practical distinction: two methods can rank samples similarly while assigning them different values. A graph whose points follow a line does not show that the line represents equality of results. The article evaluated precisely that difference. 2006 study.

In the 2008 comparison, normalization using serum samples linked to the reference method improved comparability between laboratories. This finding does not mean that every source of imprecision disappeared.

To interpret a calibration change, retain both questions: whether displacement between methods was corrected and whether repetitions remain sufficiently consistent. The mean and dispersion provide complementary information.

The 2006 study presented several indicators of the analytical procedure separately. That approach allows readers to examine which aspect supports each conclusion rather than treating a single figure as a universal score.

As a reading criterion, bring together the analyzed material, reference comparison, repetition conditions and variability measure. If one is missing, identify which part of performance remains undescribed. These examples concern serum C-peptide assays; their numerical findings should not be transferred to other analytes.