Treatment of an observed bias

Importance of the topic

The treatment of an observed significant bias is a core issue in quantitative analytical chemistry because systematic effects directly affect result accuracy, regulatory compliance and the reported measurement uncertainty (MU). Proper handling of bias influences method validation, routine decision-making (accept/reject results), comparability between laboratories and the defensibility of results in accredited testing and regulatory contexts. Understanding when to eliminate, correct or incorporate bias into uncertainty estimates is essential for robust quality assurance and traceability.

Objectives and overview of the guidance

This leaflet provides practical criteria for deciding whether to correct an observed significant bias and how such a decision impacts reported measurement uncertainty. Key decision drivers covered are: whether the cause of the bias is understood; whether its magnitude can be estimated reliably; whether the bias is consistent across the method scope (concentration range and matrices); and whether the bias behaves additively (constant offset) or multiplicatively (proportional to analyte level). The document aligns the decision process with the principles of the GUM and established recovery guidance.

Methodology and practical decision roadmap

• Initial assessment: quantify observed bias and assess significance relative to analytical goals and MU components. Small, negligible biases usually do not justify resource-intensive elimination or correction.
• Elimination: if a bias is non-negligible, the preferred approach is to remove its cause by modifying the procedure (method development, matrix cleanup, interference control).
• Correction: if elimination is impractical, consider correction only after checking regulatory constraints and three further conditions: whether correction is required by external rules, forbidden, or allowed; whether the bias magnitude can be reliably estimated; and whether the correction approach is applicable to all samples within the stated method scope.
• Reliability criterion: do not apply a correction if the bias estimate is unreliable or its cause is unknown. An unreliable correction can increase the overall MU rather than reduce it.
• Applicability across scope: a correction is only defensible if it is appropriate for all relevant concentration levels and matrices. The form of correction (additive vs multiplicative) must match the bias behaviour: additive for constant offsets, multiplicative for proportional effects.
• Measurement uncertainty consideration: according to the GUM, reported results should be corrected for recognised significant systematic effects when possible. However, correction should be applied only when the uncertainty contribution of the correction is smaller than the uncertainty introduced by leaving the bias uncorrected.
• Options when not correcting: possible practices include taking no action, reporting recovery/bias separately (with associated uncertainties), or inflating the MU to reflect the uncorrected bias. Recovery-based approaches and other strategies are discussed in the literature and by IUPAC.

Main results and discussion

The leaflet emphasises a risk-based, practical approach rather than a single prescriptive rule. Core conclusions:

• Understandability and reliable quantification of bias are the primary determinants for whether to correct.
• Correcting is justified only when it reduces the combined MU; otherwise, it can be counterproductive.
• Distinguish method bias from laboratory bias—empirical methods that define the measurand by the method have zero method bias by definition, but laboratory-specific bias must still be considered and addressed.
• Regulatory or contractual requirements may force a correction (or forbid it), overriding purely technical considerations.

The leaflet also provides a succinct operational roadmap: measure bias, evaluate options (eliminate, correct, or accept and account for it), verify reliability and applicability of any correction, and assess the net effect on MU before implementing changes to reporting practice.

Benefits and practical applications of the guidance

• Helps laboratories make defensible decisions about correcting bias, improving result accuracy and consistency across laboratories.
• Supports compliant reporting by grounding correction decisions in uncertainty analysis, reducing the risk of inappropriate adjustments that could mislead stakeholders.
• Provides a clear framework for method validation activities where bias assessment, correction strategy and MU estimation must be documented.
• Assists in accreditation workflows by aligning practice with GUM principles and relevant recovery guidance.

Future trends and opportunities for application

• Better statistical tools and modelling approaches for separating laboratory and method bias will improve the reliability of correction factors, especially across complex matrices.
• Standardised approaches for quantitative assessment of multiplicative vs additive bias will simplify decision-making and harmonise practice between laboratories and regulatory bodies.
• Integration of correction decision logic into laboratory information management systems (LIMS) and automated uncertainty calculators will streamline routine application and documentation.
• Expanded inter-laboratory studies and proficiency testing designed to probe bias behaviour across concentration ranges and matrices will supply the empirical basis needed for robust corrections.

Conclusion

Deciding whether to correct an observed significant bias must balance technical feasibility, the ability to quantify the bias reliably, consistency across the method scope and the impact on measurement uncertainty. The preferred route is elimination of bias during method development; where correction is necessary or allowed, it should only be applied when it is reliable, broadly applicable and demonstrably reduces the overall MU. If correction cannot be justified, laboratories should transparently report recovery/bias information and/or incorporate the uncorrected bias into the MU estimate.

Reference

1. JCGM 100:2008. Evaluation of measurement data — Guide to the expression of uncertainty in measurement (GUM). Bureau International des Poids et Mesures.
2. Harmonised guidelines for the use of recovery information in analytical measurement. Pure and Applied Chemistry, 1999, Vol. 71, No. 2, pp. 337–348.
3. Magnusson, B.; Ellison, S. L. R. Treatment of observed bias and related topics. Analytical and Bioanalytical Chemistry, 2008, 390, pp. 201–213.