[Reference standards in diagnostic research: problems and solutions].

The accuracy of diagnostic tests is of utmost importance as biased test results may lead to wrong decisions in clinical practice. In diagnostic accuracy research the results of a diagnostic test, model or strategy are compared to those of the reference standard, i.e. the best available method to determine whether a certain condition or disease is present or absent. Problems with the reference standard lead to biased test results. The umbrella term for this is 'verification bias'. Verification bias arises if the reference standard cannot be applied to all patients, if investigators use different reference standards or simply because there is no reference standard. Correction of these problems is often possible, and, if it is applied in a transparent and reproducible fashion it will deliver useful diagnostic information. Clinicians who use a diagnostic test should take possible verification bias into account.