Method Comparison Analysis: Assess Agreement Between Two Measurement Methods
When to Compare Methods
Perform a method comparison when a new method, instrument or laboratory must be validated against a reference (for example, IVD kit clinical evaluation, method replacement, or inter-instrument comparison). Measure the same set of samples with both methods and assess whether the results agree closely enough to be used interchangeably. A typical study uses 40 to 100 samples covering the clinical decision range (low, mid and high values). The result decides whether the new method can replace the old one or whether two laboratories can accept each other's results.
How to Choose a Statistical Method
Use Deming regression when the data are approximately normal and both methods have comparable error, because it accounts for measurement error in both variables. Use Passing-Bablok regression for skewed data or data with outliers, as it is non-parametric and robust. The Bland-Altman plot places the mean of the two methods on the x-axis and their difference on the y-axis, showing the limits of agreement (95% LoA = mean difference +- 1.96 x SD). The tool outputs the regression equation, 95% confidence intervals and agreement limits automatically.
How to Interpret the Results
A regression intercept near 0 and a slope near 1 indicate systematic agreement between the methods. The methods are considered comparable when the Bland-Altman limits of agreement fall within the clinically or operationally acceptable range, judged against medical decision levels. The tool also reports the correlation coefficient, but remember that strong correlation does not mean good agreement; agreement depends on the difference distribution and regression parameters. If the agreement limits exceed the acceptable range, investigate the source of the discrepancy (calibration, matrix effects, or differences in method principle) before deciding whether to accept the method.
How to Use It (Step by Step)
Enter paired data as two columns (method A and method B), select a regression method, then review the Bland-Altman plot and its statistics before drawing a conclusion against your acceptance criteria. Verify outliers against the original records rather than deleting them arbitrarily, and be cautious when the sample size is small. Report the sample range and statistical methods used, and do not extrapolate conclusions beyond the studied range.