Recovery in validation

[zhengmin]

Hello, everyone!

now I want to ask a question about recovery of accuracy which is discribed in USP <1225 Validation of compendia procedures>, there is a sentence I can not understand:

Assessment of accuracy can be accomplished in a variety of ways, including evaluating the recovery of the analyte (percent recovery) across the range of the assay, or evaluating the linearity of the relationship between estimated and actual concentrations. The statistically
preferred criterion is that the confidence interval for the slope be contained in an interval around 1.0, or alternatively, that the slope be close to 1.0. In either case, the interval or the definition of closeness should be specified in the validation protocol. The acceptance criterion will depend on the assay and its variability and on the product. Setting an acceptance criterion based on the lac of statistical significance of the test of the null hypothesis that the slope is 1.0 is not an acceptabl approach.

I don't know what is meaning of "slope" in above sentences. may be it is the slope which is in the linearity of the relationship between estimated and actual concentrations? can you help me?

thanks a lot

zhengmin

[krickos]

Hi

Here is a link to a simpler explenation + some excel hints, but you always google it or look in wikipedia:
http://phoenix.phys.clemson.edu/tutoria ... ssion.html

AS for the confideninterval for the slope (k in y=k*x+m), you may have to search a bit harder.
Unsure if Excel has it as standard nowdays, we have an add-on at our company that calculates the confidenceinterval. There are some softwares that also graphicly displays the confidence interval around the curve (no software name comes to mind at the moment).

Cheers

[krickos]

Hi again
Sorry misread ´the question in part

As described in the definition in chapter 1225 the definition is: "The accuracy of an analytical procedure is the closeness of test results obtained that procedure to the true value".

To obtain a confidence interval around 1,0 in this case you must "plot" the relationship between known and obtained results, just like ju said.



Again sorry for the misunderstanding.

Cheers

[Terry]

Very helpful to me.

Thanks, Krickos.

[zhengmin]

I see. thank you very much.

[aceto_81]

You can do it by the slope and calculate if 1 is in the confidence interval of the slope, or you can use the mean %recovery and calculate if 100% is in the confidence interval of the mean.

It looks like you prefer the first way, so here is the way to go in excel:
Make a matrix with linest, with the extended statistics, so that you have the errors of the slope and intercept calculated.
Then use the t.inv function with 0.05 confidence and the degrees of freedom as given by the matrix.
Then your confidence interval= slope ± (error*t).

For example:
Make your amounts in column A, going from row 1 to 9.
Make your %Recovery in column B, going from row 1 to 9.
Now make a selection of E1 to F4.
Type in as function: =linest(B1:B9;A1:A9;TRUE;TRUE)
Instead of pushing the ENTER button, hold ctrl and shift, and then push ENTER.

This will give you the matrix.

Now type in cell E6: =T.INV(0.05; F4)
Now type in cell E7: =F1-(E6*F2)
Now type in cell F7: =F1+(E6*F2)

This is your confidence interval where 1 should inbetween.

Hope this helps

Ace