Treatment of an observed bias

Importance of the topic

Observed systematic bias in analytical measurements directly affects the trueness of reported results and therefore their fitness for purpose in regulatory compliance, quality control and research. Clear, pragmatic rules for deciding when to remove, correct or account for bias are essential to ensure comparability of results between laboratories, to avoid misleading conclusions, and to provide defensible estimates of measurement uncertainty (MU). The guidance reconciles the practical constraints of method development and routine testing with the principles of the GUM and good laboratory practice.

Objectives and overview of the leaflet

The leaflet aims to provide a decision framework for handling an observed statistically significant bias in an analytical method. It clarifies: when bias should be eliminated, when a correction may be applied, when correction is inappropriate or forbidden, and how the choice interacts with the measurement uncertainty. The leaflet highlights the practical tests and criteria required to justify correction and outlines conservative alternatives if correction is not applied.

Methodology and key considerations

The leaflet frames the problem around four essential questions:

• Do we understand the cause of the bias?
• Can the bias magnitude be reliably determined?
• Is the bias consistent across all test samples within the method scope?
• Should any correction be additive or multiplicative (i.e., constant vs concentration-dependent)?

Key procedural points:

• During method development, all known systematic effects should be investigated and minimised where possible in accordance with the GUM expectation that results be corrected for recognised significant systematic effects.
• Validation should include reliable determination of any remaining bias across relevant concentration ranges and matrices.
• Observed laboratory bias may arise from the laboratory processes or from the method itself; empirical (operationally defined) methods have zero method bias by definition but still require consideration of laboratory bias.

The leaflet provides a practical roadmap: first quantify bias and determine its practical significance; if negligible, do not expend resources on elimination or correction; if non-negligible, attempt elimination; if elimination is impractical, consider correction subject to regulatory constraints and reliability of the bias estimate.

Main results and discussion

Principal conclusions and rules of thumb from the guidance:

• Correction should only be applied when the bias can be reliably estimated, the correction is applicable to all samples in the declared method scope, and the uncertainty introduced by the correction is acceptably small compared with the uncertainty that would result from leaving the bias uncorrected.
• If the cause of bias is unknown, a general recommendation to correct is not appropriate because an unrecognised or sample-dependent cause may invalidate the correction for some samples.
• Correction may be required by regulation, explicitly forbidden, or allowed; each case leads to different actions and must be treated accordingly.
• A correction is meaningful only if it reduces overall measurement uncertainty: the uncertainty contribution of the correction must be less than the uncertainty component arising from not correcting.

If bias is not corrected, the leaflet summarises pragmatic options (drawing on IUPAC recovery guidance):

• Take no action;
• Report recovery or bias separately alongside the primary result and provide uncertainties for both;
• Incorporate the observed bias into the uncertainty budget for reported results.

The leaflet emphasises that these approaches apply beyond recovery experiments and can be generalized to other bias types. It warns that applying a correction based on an unreliable estimate can increase result uncertainty and lead to worse overall accuracy.

Benefits and practical applications of the guidance

This roadmap supports laboratories and method developers by:

• Providing a structured decision process to justify elimination or correction of bias in validation and routine testing.
• Helping decide whether investment in additional method development is warranted versus documenting and compensating for bias through uncertainty assessment.
• Ensuring compliance with GUM principles and regulatory constraints when reporting corrected results.
• Offering practical alternatives when correction is not feasible, thereby improving transparency of reported data (e.g., reporting recovery and its uncertainty).

Applied examples include trace-level quantification where small systematic offsets may be critical, method transfer between laboratories, and regulated testing where mandated corrections exist.

Future trends and potential uses

Expected developments and emerging opportunities include:

• Improved statistical tools and software to quantify and propagate correction uncertainty rigorously, enabling more routine, defensible corrections when appropriate.
• Broader adoption of measurement-uncertainty-driven decision-making in accreditation and regulation, clarifying when corrections are required versus when uncertainty reporting suffices.
• Increased emphasis on characterising matrix-dependent and sample-specific biases, particularly for complex biological and environmental matrices, to support scalable multiplicative or additive correction models.
• Development of standardized protocols for bias assessment during method validation and inter-laboratory studies to harmonize practice and facilitate acceptance of corrections across laboratories.

Conclusion

Correcting an observed bias is not automatic: it requires a reliable estimate of the bias, applicability of the correction across the method scope, and an assessment that the correction reduces overall measurement uncertainty. When elimination of bias is feasible it should be pursued first. Where correction is impractical, laboratories should adopt transparent alternatives (reporting recovery, including bias in the uncertainty budget, or explicitly documenting why no action was taken). Regulatory requirements and prohibitions must be observed. The overarching principle is that any action should be evidence-based and oriented to improving the trueness and the uncertainty characterisation of reported results.

Reference

1. Joint Committee for Guides in Metrology. JCGM 100:2008. Evaluation of measurement data — Guide to the expression of uncertainty in measurement (GUM).
2. Harmonised guidelines for the use of recovery information in analytical measurement. Pure and Applied Chemistry, Vol. 71, No. 2, 1999, pp. 337–348.
3. Magnusson B., Ellison S. L. R. Analytica Chimica Acta / Analytical and Bioanalytical Chemistry, 2008, 390, pp. 201–213.