Allowable bias derived from the NOBIDA reference values.

Area of Science:

• Clinical Chemistry
• Biostatistics
• Laboratory Medicine

Background:

• Allowable bias figures guide quality control, reference value assessment, and material stability.
• Traditional allowable bias is often defined parametrically (0.25 * total biological standard deviation), based on limited sample sizes.
• Published reference values frequently originate from populations smaller than the typical 120 individuals used in parametric calculations.

Purpose of the Study:

• To non-parametrically estimate allowable bias using large-scale reference interval data.
• To compare non-parametric estimation methods with traditional parametric approaches.
• To evaluate the implications of non-parametric reference limit estimation on allowable bias figures.

Main Methods:

• Non-parametric estimation of allowable bias using percentile differences from the Nordic reference interval project biobank and database (NOBIDA).
• Estimation of allowable bias via resampling 120 reference values from NOBIDA datasets to derive non-parametric reference limits.
• Comparison of allowable bias figures derived from percentile differences versus resampling methods.

Main Results:

• Non-parametric estimation using percentile differences yielded allowable bias figures comparable to traditional parametric estimates.
• Resampling methods to estimate non-parametric reference limits resulted in significantly larger allowable bias figures.
• The findings indicate that non-parametric reference limit estimation implies a greater allowable bias than the conventional 0.25 * biological standard deviation.

Conclusions:

• Non-parametric estimation of allowable bias from large datasets provides a more robust assessment.
• Current methods for estimating reference limits non-parametrically may necessitate adjustments to allowable bias specifications.
• The clinical utility of meeting these revised bias specifications for ensuring measurement quality remains to be determined.