Statistically significant ≠ clinically relevant! [General Statistics]
Dear Hiren!
Statistically significant does not imply clinically relevant. In BE any [≥-20.00% & ≤+25.00%] difference (log-scale ±0.2231) generally* is considered irrelevant.
Since the CI is within the acceptance range. BE is defined as ARlo ≤CLlo and CLhi ≤ARhi; nothing else. If we increase the sample size (keeping the CV constant) sooner or later any (!) formulation will show a statistically significant difference – if the PE ≠ 100%. Have a look at this slide: The minimum sample size according to many guidelines is 12. With T/R 0.95 and a CV of 15% we expect already a power of 83% (the CI will be 85.07 – 106.09%). With 48 subjects the upper CL drops below 100% and we will get a statistical significant difference (CI 90.27 – 99.98%).
Or, if we keep the sample size at 12 and our CV is even lower, the power will increase – and therefore, also the chance to get a significant difference (exemplified by the light blue curve in the linked presentation). With a CV% of 10% power will be 98.8% and with 5% 99.99999995%.
May I ask you for the sample size in your study?
❝ But I am not geting how can we claim two formulation bioequivalent if there is significant formulation effect????
Statistically significant does not imply clinically relevant. In BE any [≥-20.00% & ≤+25.00%] difference (log-scale ±0.2231) generally* is considered irrelevant.
❝ Just on the basis that the variability of such difference will be less (narrow CI) how can we claim BE???
Since the CI is within the acceptance range. BE is defined as ARlo ≤CLlo and CLhi ≤ARhi; nothing else. If we increase the sample size (keeping the CV constant) sooner or later any (!) formulation will show a statistically significant difference – if the PE ≠ 100%. Have a look at this slide: The minimum sample size according to many guidelines is 12. With T/R 0.95 and a CV of 15% we expect already a power of 83% (the CI will be 85.07 – 106.09%). With 48 subjects the upper CL drops below 100% and we will get a statistical significant difference (CI 90.27 – 99.98%).
Or, if we keep the sample size at 12 and our CV is even lower, the power will increase – and therefore, also the chance to get a significant difference (exemplified by the light blue curve in the linked presentation). With a CV% of 10% power will be 98.8% and with 5% 99.99999995%.
May I ask you for the sample size in your study?
- Except some NTIDs ([≥-10.00% & ≤+11.11%] = log-scale ±0.1054) or HVDs/HVDPs where the AR may be wider.
—
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
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:
- Formulation Effect hiren379 2012-07-28 12:56 [General Statistics]
- Formulation Effect is irrelevant Helmut 2012-07-28 13:42
- Formulation Effect is irrelevant hiren379 2012-07-28 13:49
- Statistically significant ≠ clinically relevant!Helmut 2012-07-28 16:23
- Statistically significant ≠ clinically relevant! hiren379 2012-08-14 14:51
- Statistically significant ≠ clinically relevant!Helmut 2012-07-28 16:23
- Formulation Effect is irrelevant hiren379 2012-07-28 13:49
- Formulation Effect is irrelevant Helmut 2012-07-28 13:42