Clarke Error Grid Analysis

Clarke EGA quantifies the accuracy of glucose estimates generated by meters as compared to reference

Description

The Clarke error grid approach is used to assess the clinical significance of differences between the glucose measurement technique under test and the venous blood glucose reference measurements. The method uses a Cartesian diagram, in which the values predicted by the technique under test are displayed on the y-axis, whereas the values received from the reference method are displayed on the x-axis. The diagonal represents the perfect agreement between the two, whereas the points below and above the line indicate, respectively, overestimation and underestimation of the actual values. Zone A (acceptable) represents the glucose values that deviate from the reference values by ±20% or are in the hypoglycemic range (<70 mg/dl), when the reference is also within the hypoglycemic range. The values within this range are clinically exact and are thus characterized by correct clinical treatment. Zone B (benign errors) is located above and below zone A; this zone represents those values that deviate from the reference values, which are incremented by 20. The values that fall within zones A and B are clinically acceptable, whereas the values included in areas C-E are potentially dangerous, and there is a possibility of making clinically significant mistakes. [1,2]

 Syntax:
 
 [total, percentage] = clarke(y,yp)
 
 Inputs:
 y = reference values (mg/dl)
 yp = predicted/estimtated values (mg/dl)
 
 Outputs:
 total = total points per zone:
 total(1) = zone A,
 total(2) = zone B, and so on

 percentage = percentage of data which fell in certain region:
 percentage(1) = zone A,
 percentage(2) = zone B, and so on.
 
 Example:
 load example_data.mat
 [tot, per] = clarke(y,yp)

 References:
 [1] A. Maran et al. "Continuous Subcutaneous Glucose Monitoring in Diabetic Patients"
     Diabetes Care, Volume 25, Number 2, February 2002
 [2] B.P. Kovatchev et al. "Evaluating the Accuracy of Continuous Glucose-Monitoring Sensors"
     Diabetes Care, Volume 27, Number 8, August 2004
 
 Edgar Guevara Codina
 [e-mail]
File Version 1.1
November 18 2008
Ver. 1.1 corrected upper B-C boundary, lower B-C boundary slope ok, thanks to Steven Keith from BD Technologies for the corrections!
 MATLAB ver. 6.5.0.180913a (R13)

• MATLAB release: MATLAB 6.5 (R13)

[Guillermo Quintas]

A really useful job!.

[Steven Keith]

Overall, the logic is excellent. There are 2 suggestions: for the lower C region, the slope of the boundary should be 7/5 rather than 6/5. For the upper B-C boundary, there are two schools of though displayed in error grids being published. I think Clarke still uses a line parallel to 45deg rather than a line parallel to the +20% line. Very nice code.

[Edgar Guevara Codina]

Steven,

First of all, thank you for the suggestions, I think I might have made some mistakes on the regions boundaries, since I measured all boundaries from a printed plot (on the referred articles) so nothing was very exact. I hope you could provide me with the original work from Clarke, to make the necessary corrections.

[Peter Jacobs]

Thank you very much for the useful code. Well done!

[Peter]

Thank you for the interesting code! However, I suspect that there is an inconsistency in the interpretation of zone A and B. The totals and the plots are not compatible e.g. zoom in on clarke(200,243) and clarke(200, 160).