Bioequivalence and Bioavailability Forum

❝ Dear Sir,

     ^^^ Not interested in Madams' opinions?

❝ Just suggest me the calculation for getting LLOQ concentration in any drug.

As a starting point see FDA's Guidance (2001), the Arlington III Whitepaper (2007), and EMA's Draft Guideline (2009). See also this post.
Essentially you have to set up and justify a calibration model first (linear vs. quadratic, weighting schemes). The justification of the model mostly is done by assessing the back-calculated calibrators (accuracy and precision). I would also recommend to have a look at the residuals and/or the minimum AIC. Forget about R². See the case study in one of my lectures (slides 34-37). Slide 36 shows a quadratic fit, weighting 1/x (left) and 1/x² (right). Although the correlation in the left panel is higher (0.9995 vs. 0.9994), the residuals show a funnel-type shape (a sign of heteroscedasticity) and the AIC is higher than in the right panel (149.5 vs. 109.9). Since regression requires homoscedasticity (constant variances), 1/x² is the correct weighting scheme. See also the next slide for the back-calculated concentrations, especially at the lowest level.
A calibration set consists of a blank (no internal standard), an IS-spiked blank, and spiked calibrators. The lowest calibrator should be at the LLOQ. Bias at the LLOQ ≤±20%, precision ≤20% and at all other levels ≤±15%, precision ≤15%. At least ⅔ of non-zero standards should meet these criteria, including at the LLOQ and the calibration standard at the highest concentration. Excluding individual standard points must not change the model used.
Once you have set up an appropriate calibration model, check precision of the method (QC samples; using a different stock than for calibrators recommended) at the LLOQ, at ≤3×LLOQ (low), intermediate (recommended in Arlington III WP: geometric mean of LLOQ and ULOQ), high; replicate (≥5 batches) of at least triplicates. Bias ≤±15% (except at LLOQ: ≤±20%), precision ≤15% (except at LLOQ: ≤20%). Both parameters within-batch and between-batch (repeatability). For EMA at least on batch must be of the size of an expected study batch (extracted plasma blanks between calibrators/QCs).
If you succeed, the LLOQ is established. If you get very nice values, you may consider repeating this cycle for a lower LLOQ.
You see – in bioanalytics the LLOQ is not calculated (like in environmental chemistry from the variances of the regression parameters), but assessed.

These rules are applicable to 'classical' (e.g., chromatographic) methods of small molecules. For ligand-binding assays see the documents and my lecture mentioned above.

❝ Also kindly suggest whether we will take five half from the Cmax or else from the ULOQ concentration.

You must be able to measure at ≤5% of the C_max (subjects with higher pre-dose values may be excluded from the PK evaluation). Talk to your PK-expert about the expected C_max! Don't rely on a mean value from the literature; you should be able to measure ≤5% of C_max of the subject with lowest concentrations. For some drugs the inter-subject variability is quite high. If the drug is subjected to polymorphism, ratios of 1:50 (lowest/highest) are possible!

❝ If any other criteria is there please do guide me.

If truncated AUC₇₂ is not the main metric in the study, the extrapolated AUC (from the last sampling time point to infinity) should be ≤20% of AUC_∞ for any subject. It is a good idea to be able to measure the last concentration in any given subject under all treatments. Otherwise the comparison of AUC_t is some kind of 'apples-and-oranges'-statistic. Again, talk to your PK-expert and/or biostatistician.