I am trying to understand validation better.
Are real patient samples required to test % recovery and specificity?.
I was told that you would need this for electrochemistry because of oxidation/ reduction but HPLC-UV should be assumed to be selective enough to differentiate the analytes.
For the previous validation using electrochem analyte calibration curve was determined with slight concentrations of the other analytes.
For HPLC-UV I did individual standard biomarker calibration curve for a separate detection method that I developed.
how does one test for sensitivity?
when for LOD the concentration is decreased to minimal 3:1 visual detection. (10 blank runs LOD calculation vs peak area set concentration to three times peak area of blank peak, calculations gave different results hence visual detection was simpler to apply to start with lowest and go to lower analyte concentration.
https://www.fda.gov/downloads/drugs/guidances/ucm368107.pdf
for recovery this doc says to use biological matrix and add background analytes.
For sensitivity its the LLOQ of limit of quantification. Therefore how does one get LOQ and what is the lower limit is it 25% of the LOQ graph?.
If no samples are available during method development than a surrogate matrix can be used. As far as determining the LOD, I like using 40 CFR Appendix B, see link below. Take your visual estimate, then run replicates at that concentration to find your CV - that will determine your method detection limit, which can be used to find your LLOQ. Your LOQ is simply the boundaries of your dynamic range.
https://www.law.cornell.edu/cfr/text/40/appendix-B_to_part_136
You can use the surrogate matrix approach to develop an estimate of your recovery, specificity and LOD/LOQ, but you can't simulate interferences that will inevitably be present in real samples. It most assuredly will affect the specificity and the LOD.
LOQ is normally defined as 3 to 3.3 times the LOD. The LOD, as Dave points out, can be determined by various methods. I agree with Dave about the USEPA approach; I use it as well. Visually you can determine the concentration equivalent to 2-5 times the LOD; then, use the statistical approach to get a better estimate for the LOD. Again, this will change in real samples as compared to simulated matrices.
Dear Mel Gilbson,
It is a common pratice using surrogate matrix approach in order to avoid undesirable matrix effects.
Best regards,
Elcio Olveira
- Badges
- Science topic