Bioequivalence and Bioavailability Forum

Dear Rajendra,

A weighting factor is to be used when you have a different variance across your range of concentrations (heteroscedasticity). This is generally the case in bioanalytical methods, where you will have a very wide range between LLOQ and ULOQ (typically 1-100 to 1-500). Practically, weighting will result in a better fit at the lower end of the curve (but the fit will not be as good in the higher end).

The most commonly used weighting factors are 1/x and 1/x². You can base your choice on complex maths formulas (which may result on recommending 1/x^1.892, or any such unpractical value), or make your choice based on experimental validation results in a more empiric way. What we used to do in a lab I worked in was to calculate all concentrations during the validation without weighting, with 1/x and 1/x². We made our choice in the end by looking at the back-calculated concentrations of the calibration curves, and the deviations from the nominal concentrations. Once the weighting factor was chosen, we stuck to it for the whole trial. Changing the weighting factor from run to run, based for instance on the result of the QC samples, would be a sure way to get into trouble (see also the http://www.fda.gov/cder/guidance/4252fnl.pdf FDA bioanalytical guideline).

According to the FDA guideline the weighting factor should be justified.

Regression model: usually linear for HPLC with UV or spectrofluorimetric detection, and many LC/MS/MS methods. Other fits are sometimes used in LC/MS/MS, mostly quadratic. According to the FDA guideline, use of a complex regression equation should be justified and "Response function is determined by appropriate statistical tests based on the actual standard points during each run in the validation. Changes in the response function relationship between prestudy validation and routine run validation indicate potential problems" (same pages as above).

Regards
Ohlbe

---

Edit: Link corrected for FDA’s new site. [Helmut]