Bioequivalence and Bioavailability Forum • Rant

Rant [Regulatives / Guidelines]

posted by Helmut – Vienna, Austria, 2014-10-15 19:38 (3804 d 13:31 ago) – Posting: # 13718
Views: 8,860

<rant>

Slowly I’m getting sick of regulators introducing “methods” out of the blue (based on gut-feeling or what?) and leave it to us exploring whether these “recommendations” work at all. My private collection:

The Japanese Add-On design which inflates the patient’s risk as shown by Wonnemann et al. (2014).¹
The fixed effect sequence(stage) in EMA’s Q&A Rev.7. As shown by Karalis and Macheras (2013)² this modification is not relevant (as expected). So why was it introduced in the first place?
Some European regulators stating “Adapting the confidence intervals based upon power is not acceptable […] Confidence intervals should be selected a priori, without evaluation of the power” or even “Potvin is not valid in Europe”.
The Group Sequential Design as stated in by Health Canada’s TPD (Section 2.3.2.1). Pocock’s 0.0294 is recommended, but we read also:
“The first stage N₁ is generally based on the most likely intra-subject variance estimate with some added subjects to protect against drop-outs. The additional subjects required for the second stage N₂ is usually based on a worst-case scenario using a larger intrasubject variance estimate, such that N₁ plus N₂ is equal to the estimated sample size for the larger intra-subject variance. Usually the strategy with this design is to accept bioequivalence at the first stage and only go to the second stage when the intra-subject variance from the first stage is very large.”
Did Eric Ormsby perform simulations? From some preliminary ones I got the impression that inflation is not an issue,* but power might be extremely high if proceeding to the second stage due to a slightly larger than expected CV. Forced BE?

* Edit: Not correct and there is yet another problem. One has to assume a “worst-case” CV in order to calculate N₂ (rule of thumb: 50% higher CV ~doubles the sample size). In many cases N₁ will not be (N₁+N₂)/2 and Pocock’s 0.0294 might lead to inflation.³
Example: “Most likely” CV 15%, T/R 0.95, 80% power, expected drop-out rate 5% ⇒ N₁ 16. “Worst case” CV 30% ⇒ N₂ 28. Actual study: N₁ 15, N₂ 27
CV% alpha % power % studies % power stage 1 in stage 2 overall ─────────────────────────────────────── 15 0.05167 84.86 15.14 99.95 ‘neglible infl.’ (<0.052), sign. >0.05 21 0.05166 51.62 48.38 96.86 ‘neglible infl.’ (<0.052), sign. >0.05 30 0.04527 12.95 87.05 76.94
In other words Mr Pocock’s α doesn’t work that well. Although we tried to compensate for drop-outs, in the worst case scenario power <80%.
“[…] only go to the second stage when the intra-subject variance […] is very large.” Ahem; what is “very large”?
Indonesian regulators mandating a second stage even if a study passed already in stage 1. For details see Fuglsang (2014).⁴
… and now the descendants of Pancho Villa!

</rant>

Wonnemann M, Frömke C, Koch A. Inflation of the Type I Error: Investigations on Regulatory Recommendations for Bioequivalence of Highly Variable Drugs. Pharm Res. 31 (preprint published 18 July 2014). doi 10.1007/s11095-014-1450-z
Karalis V, Macheras P. On the statistical model of the two-stage designs in bioequivalence assessment. J Pharm Pharmacol. 2014;66(1):48–52. doi 10.1111/jphp.12164
Jennison C, Turnbull BW. Group Sequential Methods with Applications to Clinical Trials. Boca Raton: Chapman & Hall/CRC; 1999. p. 71–6.
Fuglsang A. A Sequential Bioequivalence Design with a Potential Ethical Advantage.
AAPS J. 2014;16(4):843–6. doi 10.1208/s12248-014-9622-7

—
Dif-tor heh smusma 🖖🏼 Довге життя Україна!
Helmut Schütz

The quality of responses received is directly proportional to the quality of the question asked. 🚮
Science Quotes

Complete thread:

Mexico BE – add-on design? nobody 2014-10-15 15:38 [Regulatives / Guidelines]
- Mexico BE – add-on design? Helmut 2014-10-15 15:50
  - Mexico BE – add-on design? nobody 2014-10-15 16:47
    - RantHelmut 2014-10-15 17:38
      - Rant nobody 2014-10-15 18:26
    - Mexico BE – add-on design? nobody 2014-10-16 10:22
      - Mexico BE – add-on design? Risherd 2014-10-24 00:26
        
        Mexico BE – add-on design? Helmut 2014-10-24 00:59
        Mexico BE – add-on design? nobody 2014-10-24 08:46
        
        Mexico BE – add-on design? Helmut 2014-10-24 19:05
        Mexico BE – add-on design? xtianbadillo 2016-04-07 18:54
        
        Mexico BE – add-on design? Mahesh M 2016-04-11 09:42
        
        No English translation Helmut 2016-04-11 13:29