Bioequivalence and Bioavailability Forum

Hi Ken,

❝ Method validation has been carried out on the drug where standard calibration curve was quantified without using weighting factor.

Bad idea. Unweighted regression is a valid model only for constant variances (homoscedasticity: variance of responses independent from concentration).

❝ We were told (later) that weighting factor was allowed.

In your posts you regularly use the phrase “We were told…” It would help to know who told you some­thing. A colleague, a regulator/inspector, the sponsor?

Correct – not only allowed, but mandatory in most cases. Do you remember this thread? In chroma­to­graphy (not necessarily in LBAs) we see more or less constant CVs – variances increase with con­cen­tration (heteroscedasticity). It is up to you to find a suitable weighting factor in method develop­ment/ validation (see Almeida et al. 2002).

1. You have to stick to what you stated in the protocol. You can recalculate samples as an additional analysis. No idea how regulators like that. In the future, go for a weighted calibration.
   
2. No. You could use the raw data of the MV and update the validation/SOP.

❝ We found that using weighting factor is easier to pass the standard calibration curve.

Sure, as expected – especially at the lower end.

---

PS: See this post.