Statistical methods for assessing agreement between two methods of clinical measurement

Area of Science:

• Clinical measurement
• Biostatistics
• Medical device evaluation

Background:

• Comparing new and established clinical measurement techniques is crucial for validating new methods.
• Current analysis methods, particularly correlation coefficients, are frequently misused and provide misleading results.
• Inappropriate statistical analysis can lead to incorrect conclusions about measurement agreement.

Purpose of the Study:

• To highlight the inappropriate use of correlation coefficients in clinical measurement comparison studies.
• To introduce and describe an alternative, appropriate analytical approach for assessing measurement agreement.
• To explain the relationship between the proposed analysis method and the assessment of measurement repeatability.

Main Methods:

• Description of a graphical technique for analyzing measurement agreement.
• Explanation of simple calculations to support the graphical analysis.
• Demonstration of the link between measurement agreement analysis and repeatability assessment.

Main Results:

• Correlation coefficients are misleading for assessing measurement agreement.
• The proposed graphical and calculation-based method provides a more appropriate analysis.
• The method facilitates a clear understanding of how well new measurements agree with established ones.

Conclusions:

• The use of correlation coefficients in clinical measurement comparison is inappropriate and misleading.
• A graphical and calculation-based approach offers a superior method for evaluating measurement agreement.
• This alternative analysis is essential for determining if a new measurement technique can replace an established one.