Validate Accuracy by Linearity

[Shanextan]

Could someone please point me to an example for following statement in USP <1225>? "Assessment of accuracy can be accomplished in a variety ways, ... or evaluating the linearity of the relationship between estimated and actual concentrations. ..."

[Jumpshooter]

They use verbiage that simply aims to convey the idea that: you can estimate the "accuracy" (how close your observed values are to the 'true' value or reference value) of your method by graphing dosage vs. response. In the case of chromatography---plot "known concentration of standard" [X variable] versus observed "peak area" [Y variable] for that standard. As the X value increases, the Y value should increase. Now, the key feature of this concomitant increase is "Rate of change" in Y due to X. This is called the "SLOPE". All of this can be computed by EXCEL.

[Jumpshooter]

And lastly, it would be good to have some "check sample" furnished by an independent lab/company that is purported to have a conc'n in the range of your calibration. When you shoot the "check sample", you will get a quant value--be sure to compare this "observed" quant in the check sample to the "stated level" that the manufacturer has inscribed on the vial. If your Standard Curve is worth it's weight in pixels, then when you plug in the peak area Y value for the check sample and divide it by the slope, then the resultant concentration will be in good agreement with the manufacturer's level. If not, then either your cal curve is not good, or there are matrix effects.

[DCL11]

Could someone please point me to an example for following statement in USP <1225>? "Assessment of accuracy can be accomplished in a variety ways, ... or evaluating the linearity of the relationship between estimated and actual concentrations. ..."

Accuracy is defined as the closeness of agreement between experimental value and theoretical value. When you are doing linearity test, you construct a linearity curve using your theoretical concentration and response (e.g. peak area). Using this calibration curve and the response (say peak area) of each solution, you will be able to calculate the experimental value of each solution. By comparing the experimental value against the theoretical value, you can calculate accuracy. Using this approach, you do not need bracketing standards.

[Shanextan]

Has anyone seen an example in a publication?

[shaun78]

USP and/or ICH directly states that accuracy can be inferred from linearity and precision data.

Shaun