## –0.2025 < –0.2, right? [Study As­sess­ment]

Hi PKS,

I reordered your quotes for clarity.

» My protocol says within (-0.20 and 0.20)
» » » Should I conclude bio-equivalence with the values of [-0.2025, 0.076] in clinical end point study.

No!
{–0.2025, +0.076} {–0.20, +0.20} ∎

Furthermore, ±0.20 of what? Since you are working with untransformed data, possibly  = ±20% of the arithmetic mean of the reference. This was used in the dark ages of BE as well. Hence, the limits were {1 – , 1 + } or {0.8000, 1.2000}. Do you think that {0.7975, 1.0760} would pass?
Or do you want to want to deal directly with (as your protocol suggests) and round the CI to only 2–3* digits in order to pass the limits stated in your own protocol?

» Can I assume it as within the interval of (-0.20, 0.20)

You can assume whatever you like. If you give us the design, sizes (per sequence in a crossover, of groups in a parallel), and the CV we can calculate the α (probability of type I error, aka patient’s risk). However, α will be >0.05 and hence, the chances that authorities will accept the study as proof of equivalence are extremely low.

• IEEE 754  : round(–0.2025, 2) = –0.2 (unbiased from zero)
commercial: round(–0.2025, 3) = –0.2 (biased from zero)

Dif-tor heh smusma 🖖
Helmut Schütz

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