Bioequivalence and Bioavailability Forum

Hi John,

❝ […] 3 consecutive day samples to assess steady state then it should be fine as long as you can demonstrate that steady state is achieved (regression analysis).

Been there, done that. Nowadays I don’t recommend it any more. Testing the slope for ≠ 0 (or assessing whether its CI does not contain 0) will lead to exclusion of ~5% of profiles (since we falsely conclude that we are not in steady state).
Small simulation: t_½,el 12 h, t_½,abs 3 h, τ 24 h, 5×t_½,el to reach “steady state”, profile on day 6, analytical variability 2.5–25% (excellent to lousy), concentrations lognormal, 50,000 profiles, regression of pre-dose concentrations (t 72, 96, 120 h), test for a significant p of the slope.

CV (%)  p <0.05 (%)
───────────────────
  2.5      5.10
  5        5.03
  7.5      5.05
 10        4.98
 12.5      4.93
 15        4.87
 17.5      4.89
 20        4.79
 25        4.75

Consequently the EMA requires only reporting pre-dose concentrations.
IIRC, in Taiwan one must sample beyond τ in order to estimate t_½,el. Duno whether subjects have to be excluded if “steady state” was not reached as planned.
There is a correlation (R² 0.9593) of p-values and the variability of the analytical method. Makes sense. If we design the study with 5×t_½,el, concentrations are still increasing (positive slopes). Good analytical methods will detect that. With not so good methods it will be hidden in random noise.

Same game with a longer saturation phase (7×t_½,el):

CV (%)  p <0.05 (%)
───────────────────
  2.5      4.83
  5        4.87
  7.5      4.83
 10        4.83
 12.5      4.90
 15        4.85
 17.5      4.80
 20        4.73
 25        4.71

Better (R² 0.5197) but you would still exclude profiles.