Log in
Register|
Imran ☆ Mumbai, 2006-12-30 06:36 (6704 d 01:50 ago) Posting: # 425 Views: 10,184 |
|
|
Dear All Can anybody suggest me the right formula/method to be used for sample size estimation in BA/BE study. what I understand that the Boltan suggest sample size estiamtion based on mean and SD. Some Statistician suggest to use the formula based on intrasubject CV only. However it is diffucult many time to get the intrasubjects CV unless u have pilot study data. In such case the sample size estimation using mean and SD will be questinable!!!. please comment. regards Dr.Imran Khan Sitec, India — Dr.Imran Khan |
|
Helmut ★★★   Vienna, Austria, 2006-12-31 19:38 (6702 d 12:48 ago) @ Imran Posting: # 427 Views: 8,266 |
|
|
Dear Imran! ❝ Can anybody suggest me the right formula/method to be used for sample size estimation in BA/BE study. ❝ what I understand that the Boltan suggest sample size estiamtion based on mean and SD. Are you talking about “Pharmaceutical Statistics” by Bolton and Bon? Have a look at their table 6.5 (p 169 of the 4th Ed., 2004), which uses the intrasubject CV, the expected point estimate (µt/µR), and the desired power. The original table* served many thousands (!) of BE studies in the last 15 years. The exact formula is a little bit tricky (see this post), but there are also approximations available (for references see this post). ❝ Some Statistician suggest to use the formula based on intrasubject CV only. At least for a cross-over study you definitely need the intrasubjct CV (whereas for a parallel study the total variability would suffice). ❝ However it is diffucult many time to get the intrasubjects CV unless u have pilot study data.In such case the sample size estimation using mean and SD will be questinable!!!. You also may use data from the literature. If the CV is not presented but the ANOVA-table, you may calculate it (from log-transformed data) by$$CV_{intra}(\%)=100\sqrt{e^{MSE}-1}$$If only the Confidence Interval is given, you have to do a little maths (see this post and followings).
— Dif-tor heh smusma 🖖🏼 Довге життя Україна! ![[image]](https://static.bebac.at/pics/Blue_and_yellow_ribbon_UA.png) Helmut Schütz ![[image]](https://static.bebac.at/img/CC by.png) The quality of responses received is directly proportional to the quality of the question asked. 🚮 Science Quotes |
|
Imran ☆ Mumbai, 2007-01-04 07:51 (6699 d 00:35 ago) @ Helmut Posting: # 430 Views: 8,194 |
|
|
Dear HS Thanks for reply. Yes I was talking about Boltan and Bon. Boltan gives a example to calculate sample size (page no. 157) where it estimate sample size using the mean and SD. Many time it is difficult to get the CV unless and v have pilot study data, however mean and SD is easily available. When I have a CV diffinately I follow the tablet of boltan on page no 169. so calcualtion of a sample size using this formula will be questionalbe!!!! regards Dr.Imran Khan Sitec, India — Dr.Imran Khan |
|
Helmut ★★★ Vienna, Austria, 2007-01-04 12:17 (6698 d 20:09 ago) @ Imran Posting: # 432 Views: 8,219 |
|
|
Dear Imran! ❝ Boltan gives a example to calculate sample size (page no. 157) where it estimate sample size using the mean and SD. Be careful! To quote Bolton/Bon: “Estimates of the standard deviations of the ratios of the values of each of the parameters […] were determined from a small pilot study.” If you don’t have the study’s data available, where to you get the SD of the individual ratios from?I used subjects 1-6 of my example data to get: SD of T/R-ratios: 0.19923581 applying Eq.(6.6) with a maximum deviation (Delta) to detect of 0.2 one gets N=9.71 which is in agreement with the common method based on CV%-intra, where N=10 is estimated for a point estimate of 0.95 (Delta 0.05). With a PE of 0.90 N is estimated with 18… Using subjects 7-12 (6.6) gets N=23.9, again in agreement with the CV%-intra method, which gets N=24. ❝ so calcualtion of a sample size using this formula will be questionalbe!!!! Hmm, I agree, but since you need to have access to raw data for Bolton/Bon’s formula also, why not using one of the ‘traditional’ sample size estimations based on the intra-subject CV anyway? It looks like that (6.6) gives the ‘correct’ sample size for a PE of 0.95; but other Deltas should be possible according to the Guidelines… — 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 |
Ing. Helmut Schütz