Bioequivalence and Bioavailability Forum

Hi Krishna,

❝ As per the recent guideline of FDA for warfarin sodium Tablets, they designed the BE acceptance criteria assuming all NTIs are low variable.

Likely they are. From a presentation by Lawrence Yu:¹

Residual variability (%CV) from ANDAs reviewed between 1996–2008
                              AUC0-t             Cmax
Drugs                    Mean     Range     Mean     Range
Warfarin (n=29)           5.7   3.3, 11.0   12.7   7.7, 20.1
Levothyroxine (n=9)       9.3   3.8, 15.5    9.6   5.2, 18.6
Carbamazepine (n=15)      8.0   4.4, 19.4    8.7   5.2, 17.6
Lithium Carbonate (n=16)  7.8   4.5, 14.0   13.5   6.4, 24.4
Digoxin (n=5)            21.7  13.1, 32.2   21.0  14.3, 26.1
Phenytoin (n=12)          9.2   4.1, 18.6   14.9   7.4, 20.0
Theophylline (n=3)       17.9  12.8, 24.2   18.2  11.8, 25.8

❝ I have gone thorough the guideline it is very stringent to rest of regulatory.

Agree. I think it is an attempt to avoid arbitrary limits ignoring potential differences in variances – important if formulations are switched under treatment. Reminds me somehow on IPE/PBE of the late 1980s.

❝ […] Suppose if the full replicated study results are like for Cmax: 86.00 (80.00 - 93.00) with %CV: 26

Hhm, \(\small{\sqrt{0.80\times 0.93}\neq 0.86}\). If I interpret your example right the study was a full replicate in 32 subjects.

❝ […] then the study passing with SABE, also with ABE and test variability is less than the Refe variability.

The study passes all criteria. What is the problem?

❝ Even, in the guideline they didn't mention about cut off for maximum sigmaR.

Scaling may result in an acceptance range wider than the conventional one² (if CV_WR >21.42%; scaled limits for your 26% would be 76.38–130.93%). Ways out:

1. A cutoff on s_WR (the estimate of σ_WR) and switch to conventional unscaled ABE. Donald Schuirmann explored values of 0.21179 (CV 21.42%) and 0.21 (CV 21.23%).
   
2. A “Must Pass Both” criterion: RSABE + ABE

Both methods preserve the patient’s risk with the same power. Empiric α was lower (more conservative) with method #2. I guess that’s the reason FDA implemented this method in their guidance.

❝ So, what is the advantage of above stringent criteria in this case?

The replicate design allows comparison of σ_WT with σ_WR. Arbitrary narrowing the AR does not take the actual variance into account.
Not an advantage but an obstacle: sample size estimation. Simulations are required and IMHO a pilot study is unavoidable.

---

1. Yu LX. Approaches to Demonstrate Bioequivalence of Critical Dose Drugs. Advisory Committee for Pharmaceutical Science and Clinical Pharmacology. 13 April, 2010.  webarchive.
   
2. Schuirmann DJ. Evaluation of Scaling Approaches to Demonstrate BE of NTI Drugs – OGD Simulation Efforts. Advisory Committee for Pharmaceutical Science and Clinical Pharmacology. 26 July, 2011.  webarchive.

---

P.S.: Nice to know:

• Simulations were carried out in the S-Plus, R [sic], or APL computer programming languages
  
• Each estimated probability based on one million (1,000,000) simulated studies