sury ★ India, 20190408 13:29 Posting: # 20130 Views: 1,874 

Hii all Hope all are doing well! I have recently got an bioequivalence study with clinical endpoints for some drug In OGD it states that “To establish bioequivalence, the 90% confidence interval of the test/reference ratio of the mean should be contained within [0.80, 1.25], using the per protocol (PP) population.” The scores of the clinical endpoint shall be as follows “Score the oral lesions and specific signs and symptoms of oropharyngeal candidiasis at each visit using the following two scoring systems:
As the above states that 20% difference is accepted, then in normal bioequivalence case (if untransformed data), it should be 80.00120.00? Correct me if I am wrong? Best regards 
Ohlbe ★★★ France, 20190408 15:18 @ sury Posting: # 20131 Views: 1,582 

Dear Sury, » I have recently got an bioequivalence study with clinical endpoints for some drug » In OGD it states that » » “To establish bioequivalence, the 90% confidence interval of the test/reference ratio of the mean should be contained within [0.80, 1.25], using the per protocol (PP) population.” For BE with PK endpoint, yes. Not with clinical endpoint. » The scores of the clinical endpoint shall be as follows [...] You're referring to the draft guidance on miconazole buccal tablets, right ? Would be easier for us if you could say so straight away and link the guidance... » For the above, do we need to convert the grading (0,1,2,3) to log transformed data and conduct the statistical analysis? » » As the above states that 20% difference is accepted, then in normal bioequivalence case (if untransformed data), it should be 80.00120.00? For the statistical analysis the draft guidance refers to another draft guidance, on adapalene + benzoyl peroxyde. Did you read it ? Did you see anywhere a mention of logtransformation ? And why are you changing acceptance limits of 0.20 to +0.20 into 80.00  120.00 ? — Regards Ohlbe 
sury ★ India, 20190409 06:19 @ Ohlbe Posting: # 20133 Views: 1,483 

» You're referring to the draft guidance on miconazole buccal tablets, right ? Would be easier for us if you could say so straight away and link the guidance... yes, i am actually referring to the miconazole buccal tablets drug » For the statistical analysis the draft guidance refers to another draft guidance, on adapalene + benzoyl peroxyde. Did you read it ? Did you see anywhere a mention of logtransformation ? And why are you changing acceptance limits of 0.20 to +0.20 into 80.00  120.00 ? There is no where in the draft guidance on what data (Untransformed and Log Transformed) to conduct the statistical Analysis. Edit: Full quote removed. Please delete everything from the text of the original poster which is not necessary in understanding your answer; see also this post! [Ohlbe] 
Helmut ★★★ Vienna, Austria, 20190408 15:38 @ sury Posting: # 20132 Views: 1,596 

Hi Sury, are you referring to buccal miconazole? » For the above, do we need to convert the grading (0,1,2,3) to log transformed data and conduct the statistical analysis? Grab you pocket calculator, fire up a spreadsheet or a statistical software. Which result do you get for log(0) ? Bonus question: Why?» As the above states that 20% difference is accepted, then in normal bioequivalence case (if untransformed data), it should be 80.00120.00? Why? You claimed to quote from the OGD’s guidance: » “To establish bioequivalence, the 90% confidence interval of the test/reference ratio of the mean should be contained within [0.80, 1.25], using the per protocol (PP) population.” If you were referring to buccal miconazole, sorry, such a statement is not given there. In Section 5 it is stated: The recommended primary endpoint of this study is the proportion of subjects with a clinical cure at the testofcure (TOC) visit on Day 21 (i.e., 7 days after completion of 14 days of treatment) ± 4 days in the Per Protocol analysis population. A clinical cure is defined as complete resolution of all signs and symptoms of oropharyngeal candidiasis (oral lesion score = 0, signs and symptoms score = 0). (my emphasis)I must confess that this guidance leaves me confused. At a first look this smells of testing for equivalence of the ratio of two means with normality on the original scale (i.e., Fieller’s ‘fiducial’ confidence interval and the Sasabuchi test)^{1,2} where the conventional limits are {0.80, 1.25}. But then I had some doubts. Before we perform statistics we have to consider the datagenerating process. Scores are of an ordinal scale at its best. Quite often they are just nominal. Suitable tests for the former are nonparametric or proportional odds and for the latter the χ². Throwing away any data with scores > 0 sounds strange to me. I didn’t run a test yet but consider these three cases:
— Diftor heh smusma 🖖 Helmut Schütz The quality of responses received is directly proportional to the quality of the question asked. 🚮 Science Quotes 