Bland and Altman agreement method: to plot differences against means or differences against standard? An endless...

Area of Science:

• Biostatistics
• Measurement Science

Background:

• Controversy exists in Bland and Altman agreement analysis regarding plotting differences against the mean or the standard method.
• The choice of graphical representation implies an underlying regression model to assess biases.

Purpose of the Study:

• To resolve the controversy in Bland and Altman analysis by evaluating plotting methods.
• To identify conditions that minimize spurious proportional error in measurement agreement studies.

Main Methods:

• Reviewed papers that plotted differences against the standard method, identifying their statistical pitfalls.
• Analyzed conditions influencing the correlation coefficient and regression slope between differences and means/standards.

Main Results:

• Demonstrated scenarios where plotting differences against the mean or standard results in regression slopes and correlation coefficients closer to zero.
• Showcased how specific plotting choices minimize spurious proportional error between measurement methods.

Conclusions:

• Calculating expected values of correlation coefficients and regression slopes aids in selecting appropriate statistical models for agreement studies.
• Provided recommendations for evaluating the feasibility of agreement or calibration studies.