jag009 ★★★ NJ, 2014-09-22 08:34 (3496 d 19:41 ago) Posting: # 13554 Views: 3,651 |
|
Hi everyone, A very simple question (may not be simple). Suppose you run a 4-way study with 3 formulations vs Reference (Yes Helmut, common variance), all formulations fail BE simply because there is an outlier in one of the treatment periods. If you remove that outlier and rerun stats, which leads to a reduction in the intrasubject CV, then one of the three formulations becomes BE, not borderline. Can you present an argument and remove the that particular treatment data from the overall concentration dataset and conclude BE on the good formula? How do you approach the hurdle? Thanks John |
ElMaestro ★★★ Denmark, 2014-09-22 11:38 (3496 d 16:37 ago) @ jag009 Posting: # 13555 Views: 2,863 |
|
Good morning John, ❝ (...) there is an outlier (...) ❝ Can you present an argument and remove the that particular treatment data from the overall concentration dataset and conclude BE on the good formula? Should be handled as per protocol and SAP. Technically you might prefer to call it a subjectively aberrant value at this point, but I guess semantics is not the real concern right now. If nothing is stated in the protocol/SAP I would audit the whole thing (audit means investing in a cheap plane ticket and staying three nights at the nearby Bedbug Inn & Suites, rather than to just ask the CRO to look into its own business – "We know them so well, have been doing 156 studies there since 2008 and never had any issues so no need to go there.") If you can find a deviation specific to the data point in question you might have a case for eliminating it; if nothing is found (= if it is plausible that all data have been treated more or less the same) and nothing was ever written about exclusions in the protocol/SAP then...well... — Pass or fail! ElMaestro |
jag009 ★★★ NJ, 2014-09-22 22:46 (3496 d 05:29 ago) @ ElMaestro Posting: # 13557 Views: 2,895 |
|
Hi ElMaestro, ❝ If you can find a deviation specific to the data point in question you might have a case for eliminating it; if nothing is found (= if it is plausible that all data have been treated more or less the same) and nothing was ever written about exclusions in the protocol/SAP then...well... Nothing on the protocol. The stat plan just states the comparisons: A vs D, B vs D, C vs D. But I am not eliminating him from the other 3 study periods though, just in the period that he showed very low (near BLQ) values. John |