Bioequivalence and Bioavailability Forum

Hi ElMaestro,

❝ […] I think this has the potential to be the kind of requirement that is associated with an unpalatable risk of failure, ultimately just scaring sponsors away from development of affordable alternatives.

❝ The intention is good, I have no doubt, but I am afraid the treatment is worse than the disease in this case.

I guess this is one of the cases where common sense nonsense hit: ‘Hey, we don’t even want the linear regression, which was standard in the last century. Let’s come up with sumfink easily doable with paper & pencil.’
In the meantime I performed some simulations. IIRC, the ANVISA recommends dosing for ten half lives. I played it safe and added yet another dose.

16 subjects, 2 periods (32 simulated profiles)
absorption half life :  6.00
elimination half life: 12.00
lag time             :  0.25
CV                   : 10.0%
dosing interval      : 24.00
doses                : 7
dosing at            : 0, 24, 48, 72, 96, 120, 144
last 3 pre-doses at  : 96, 120, 144
margins              : ±20% of the average of the last 3 pre-dose concentrations
 pd1 – pd2 within margins: 87.5%
 pd2 – pd3 within margins: 90.6%
 pd1 – pd3 within margins: 81.2%
 It cannot be concluded that steady state has been reached in 10 profiles.
 31% have to be excluded.

A study of a drug with low variability in 16 subjects (32 profiles in a 2×2×2 crossover) has high power. However, we would have to increase the sample size dramatically in order to compensate for the exclusions.

20 subjects, 4 periods (80 simulated profiles)
absorption half life :  6.00
elimination half life: 12.00
lag time             :  0.25
CV                   : 30.0%
dosing interval      : 24.00
doses                : 7
dosing at            : 0, 24, 48, 72, 96, 120, 144
last 3 pre-doses at  : 96, 120, 144
margins              : ±20% of the average of the last 3 pre-dose concentrations
 pd1 – pd2 within margins: 30.0%
 pd2 – pd3 within margins: 38.8%
 pd1 – pd3 within margins: 28.8%
 It cannot be concluded that steady state has been reached in 75 profiles.
 94% have to be excluded.

Ugly. For a drug with high variability this method does not work at all.

---

Edit: Meanwhile I assessed one of my studies. All European regulatory assessors were fine with what we did. With the ANVISA’s method we would get:

32 subjects, 2 treatments (64 profiles)
dosing interval      : 24
doses                : 13
dosing at            : -288, -264, -240, -216, -192, -168, -144, -120, -96, -72, -48, -24, 0
last 3 pre-doses at  : -48, -24, 0
margins              : ±20% of the average of the last 3 pre-dose concentrations
 pd1 – pd2 within margins: 59.4%
 pd2 – pd3 within margins: 65.6%
 pd1 – pd3 within margins: 54.7%
 It cannot be concluded that steady state has been reached in 37 profiles.
 58% have to be excluded. 10 subjects (31%) eligible for BE.

Although the study was not powered for the last pre-dose concentration, it passed the conventional limits (90% CI: 81.29 – 112.29%).
Applying this method would have meant the end of the story for the ANVISA. After excluding the majority of profiles the study would have been extremely underpowered for C_max and turned from a success to a failure (93.69 – 110.72% → 85.35 – 143.61%).