Not all biases are created equal: how to deal with bias on laboratory measurements.

Although the concept of bias appears consolidated in laboratory science, some important changes in its definition and management have occurred since the introduction of metrological traceability theory in laboratory medicine. In the traceability era, medical laboratories should rely on manufacturers who must ensure traceability of their diagnostic medical devices (IVD-MD) to the highest available references, providing bias correction during the trueness transfer process to calibrators before they are marketed. However, sometimes some bias can be observed arising from an insufficient correction during the traceability implementation. This source of bias can be discovered by the IVD-MD surveillance by traceability-based external quality assessment and confirmed by validation experiments. The assessment of significance should be based on its impact on measurement uncertainty (MU) of results. The IVD manufacturer, appropriately warned, is responsible to take an immediate investigation and eventually fix the problem with a corrective action. Even if IVD-MD is correctly aligned in the validation steps and bias components are eliminated, during ordinary use the system may undergo systematic variations such as those caused by recalibrations and lot changes. These sources of randomly occurring bias are incorporated in the estimate of intermediate reproducibility of IVD-MD through internal quality control and can be tolerated until the estimated MU on clinical samples fulfils the predefined specifications. A readjustment of the IVD-MD by the end-user must be undertaken to try to correct the bias becoming significant. If the bias remains, the IVD manufacturer should be requested to rectify the problem.