ISO/TS 20914:2019 - a critical commentary

Area of Science:

• Clinical Laboratory Science
• Metrology
• Biostatistics

Background:

• ISO/TS 20914, published in July 2019, provides practical guidance for estimating measurement uncertainty in medical laboratories.
• The document claims to be based on the Evaluation of measurement data - Guide to the expression of uncertainty in measurement (GUM), JCGM 100:2008.
• Despite its detailed examples, the guide presents several potentially misleading or incorrect statistical procedures.

Purpose of the Study:

• To critically evaluate the statistical methodologies and statements within ISO/TS 2019.
• To identify and highlight major concerns regarding the estimation of measurement uncertainty in medical laboratories as presented in the ISO guide.
• To provide a detailed commentary on specific problematic areas within the ISO/TS 20914 document.

Main Methods:

• Comparative analysis of ISO/TS 2019 with established metrological principles (GUM).
• Identification and detailed examination of specific statistical procedures and statements within the ISO guide.
• Case study analysis of calculations for estimated glomerular filtration rate (eGFR) and international normalized ratio (INR).

Main Results:

• Concerns identified regarding the use and potential misuse of coefficient of variation.
• Oversimplification and potential misleading nature of pooled variance and measurement uncertainty across changing conditions.
• Incomplete inclusion of all known uncertainties in estimated glomerular filtration rate (eGFR) calculations.
• Incorrect international normalized ratio (INR) calculation identified.
• Problematic treatment of bias uncertainty and incomplete rules for evaluating combined uncertainty in functional relationships.

Conclusions:

• ISO/TS 2019, while offering practical guidance, contains significant statistical inaccuracies and potentially misleading information.
• Clarification and correction are needed for several statistical procedures, including coefficient of variation, pooled variance, bias uncertainty, and combined uncertainty calculations.
• The accuracy of eGFR and INR calculations as presented in the guide is questionable, requiring further review and revision to ensure reliable measurement uncertainty estimation in medical laboratories.