Bioequivalence and Bioavailability Forum

Dear Ketan!

❝ I am carrying out ligand bind assay where the linearity range for detection of analyte is 400 – 32000 ng/mL. where 400 is anchor point and second standard concentration (STD2) 500 is LLOQ. Regression method used is Log-Log.

Log/log is a bad model in LBA-calibration. It should only be used if nonlinear regression software is not available. Better models are 4- or 5-parameter logistic.^* See also this presentation (slides 38–42, especially slide 40 comparing the confidence intervals of log-log and 4-PL). If you are financially handicapped consider R’s function nls (free).

❝ 1. If calibration curve is within acceptance criteria should I include anchor point?

Are you talking about method validation or routine application of the method?
In logistic regression anchor points will result in more stable fits. Once you have established the model in validation it should be used in routine analyses as well.

❝ 2. If any calibration standard is not within acceptance criteria I should include anchor point.

❝ 3. If my LLOQ (500 ng/mL, STD 2) is not within acceptance criteria and if I include anchor point and then if it is within acceptance criteria then my LLOQ for that activity or batch will change from 500 to 400 ng/mL. Is it acceptable to regulatory?

Since I already failed to understand your #2, no recommendations here. The anchor point STD1 at 400 ng/mL is outside the calibration range. The LLOQ is still STD2 at 500 ng/mL – whether you include the anchor point or not. Regulators definitely don’t like any kind of “playing around” with data in order to get a batch valid.

---

• Findlay JWA and R Dillard
  Appropriate Calibration Curve Fitting in Ligand Binding Assays
  The AAPS Journal 9(2) Article 29, E260–E267 (2007)
  free online resource