Section11:Method Validation (Pre-Study)

From Assay Guidance Wiki

It is important to note that the precision profile is based on just the calibration curve. Consequently only the calibration curve factors (quality and stability of reference standards, quality and stability of reagents, statistical validity of the calibration curve model) are taken into account for deriving these quantitation limits. Sample factors such as analyte (similar physicochemical substances), matrix (other substances that can affect analytical result) and operational factors can affect the performance of the assay/method as well. Thus the quantitation limits derived from the precision profile of a calibration curve is an optimistic assessment of method performance. If these limits are not satisfactory, we need to re-optimize the assay further.

If the quantitation limits from the precision profile are close to the limits desired for the method’s intended use, proceed to a full validation experiment as outlined below. This validation experiment is used to establish the method quantitation limits using the analysis of recovery data from validation samples (spiked standards). This experiment will take into account the three major sources of variation described above (calibration curve factors, sample factors and operational factors).

For the full validation experiment, generate the following data in at least three independent runs.

• Calibration curve in each run, preferably in triplicate.
• Validation/QC samples (independent set of samples spiked with known amount of standards) in each run at six concentrations with four replicates; three concentrations near the precision profile estimates of lower and upper quantification limits, and three more equally spaced between lower and upper quantification limits.
• Estimate the concentrations of the validation samples of each run using the respective calibration curves. Then compute the % recovery of these validation samples using the following formula:
  
  % Recovery = 100*(Estimated Concentration)/True Concentration
  
• Now compute the average and standard error of the % recovery data of the validation samples from all runs for each concentration. Then standard error should be based on a separate variance component analysis of the multiple runs of validation data, and it should include the sources of variability relevant during the use of the assay in production. At the minimum, it will include inter-run and intra-run variability. Some of the other sources to consider may be analyst, plate, equipment, etc.
• Plot the average % recovery values along with the standard error (as calculated above) versus the true concentrations. Note that the % recovery along with the standard error as determine above reflects the Total Error of the assay.
• The %recovery and the standard error limits must be within +/- X% of the nominal value. If X is 30 of the nominal value at each concentration (i.e., 70% to 130%). That is, this means that the Total Error of the assay must be within 30%. The value of X should be set based on the intended use of the assay. Recommendations on the acceptance criteria are discussed later in this chapter.
• If X is set at 30%, the lower quantification limit is the lowest concentration at which the % recovery is within 70% to 130%. The upper quantification limit is the highest concentration at which the % recovery is within 70% to 130%.