Bioequivalence and Bioavailability Forum

Dear Martin,

following our inspiring yesterday's discussion some points:

1. At least three ... In my experience four is most commonly employed.
   
2. Not necessarily doubled, but equally spaced in the logarithmic domain (modifying your example: 4 levels, lowest 0.5 mg, highest 8 mg: intermediates at 1.26 mg, 3.17 mg).

Have a look at two classical papers:

• Gough K, Byrom B, Ellis S, Lacey L, McKellar J, Hutchison M, and O Keene
  Assessment of Dose Proportionality: Report From the Statisticians in the Pharmaceutical Industry/Pharmacokinetics UK Joint Working Party
  Drug Information Journal 29(3), 1039–48 (1995)
  
• Smith BP, Vandenhende FR, DeSante KA, Farid NA, Welch PA, Callaghan JT, and ST Forgue
  Confidence Interval Criteria for Assessment of Dose Proportionality
  Pharm Res 17(10), 1278–83 (2000)

I came across some Fibonacci-stuff; too lazy to search.

It should be noted that the power model is a purely empirical one - i.e., is not based on physiological grounds. Unlike dose-corrected BE-assessments between two dose levels, any deviation of the exponent (beta>1: enzyme limited excretion, saturation; beta<1: enzyme induction) is not described in PK-models. Pharmacokineticists are not happy with the power model anyway. To quote Nick Holford (PKPD-List, 2006):

I am aware of the whacky power function used by non-pharmacological statistians to 'declare dose proportionality'.
Any rational attempt to understand kinetics in relation to dose would use PK theory not empirical ad hoc 'beat it to death with a P value' statistics.