Permissible performance limits of regression analyses in method comparisons.

Area of Science:

• Analytical Chemistry
• Clinical Laboratory Science
• Biostatistics

Background:

• Method comparison studies are crucial for validating analytical procedures in laboratories.
• Laboratories frequently need to determine if a new method can replace an established one without compromising diagnostic accuracy.

Purpose of the Study:

• To propose a framework for deriving permissible equivalence limits based on false-positive error rates.
• To establish criteria for assessing method comparability and identifying potential interferences or discrepancies.

Main Methods:

• Analyzing split patient samples simultaneously using both the established (x) and new (y) methods.
• Graphical presentation of data pairs using scatter or difference plots.
• Defining equivalence based on data points scattering around the line of equality within permissible limits derived from error rates.

Main Results:

• If all data pairs are within the permissible equivalence limits, the methods demonstrate comparable false error rates.
• Discordance outside these limits indicates the x-method cannot be directly replaced by the y-method, necessitating further investigation.
• Deviations may stem from outliers, non-linearity, bias, or increased imprecision in one of the methods.

Conclusions:

• The proposed method provides a robust approach to validate analytical procedure replacement.
• Equivalence testing based on error rates ensures diagnostic reliability when comparing laboratory methods.
• Identifying and addressing discrepancies are essential for accurate method validation and reliable patient results.