PharmPK Discussion - Sample size in exploratory studies

PharmPK Discussion List Archive Index page
• On 12 Nov 2008 at 15:58:59, "Charlie Brindley " (charlie.brindley.-at-.kinetassist.com) sent the message
`JUSTIFICATION OF SMALL SAMPLE NUMBERS IN EXPLORATORY STUDIESWhat is an appropriate number of subjects per dose group to include inan exploratory clinical study (e.g. first-in-man)?  In such studies,we have limited prior information and wish to collect informationgenerally on safety, tolerability, PK and possible potential PDeffects.  There is not a specific primary endpoint on which we canbase a sample size calculation but data collected will be importantfor the design of the next study.We have a study proposed with n of 6 active + 2 placebo in each of 3escalating dose groups.  Is n of 6 sufficient?  Alternatively, couldwe drop down to n of 4, 3, or even 2?Descriptive statistics will be used to summarise the data we collectand for quantitative variables, we will calculate the 95 % confidenceinterval (CI) for the mean using the standard formula:Mean +/- t(n-1).SD//nwhere t(n-1) is the value from the t-distribution on n-1 degrees offreedom with 2.5% of the distribution lying above.The value of t(n-1) increases dramatically as n decreases to 4 andbelow.  Thus the impact of increasing or decreasing the sample sizefrom 6 on the precision on any estimate can be calculated byconsidering the ratio of t(n-1) values divided by the ratio of /n .The following table shows the fold increase in the width of a 95 %confidence interval compared to n of 6.             n          fold increase in 95 % CI (compared to n of 6)                         [t(n-1)/t(6-1)] / [sqrt(n)/sqrt(6)]             2           8.56             3           2.37             4           1.52             5           1.18             6           1.00             7           0.88             8           0.80There is quite a marked deterioration in precision as n decreases fromn of 5.  Thus n of 4 and below may not be a good idea and n of 5 isrisky once you allow for drop outs.  Similarly, the improvement ofincluding 7 or 8 subjects does not increase precision to a great extent.Could this consideration be used as a rationale to justify the choiceof the number of subjects to include, for example, in a first-in-manstudy?Does anyone know of any other justification of subject numbers inearly clinical trials where there is no specific primary endpoint withassociated estimate of variability?Charlie Brindley (KinetAssist) and Sue McKendrick (AAIPharma, Edinburgh)`
Back to the Top

 Want to post a follow-up message on this topic? If this link does not work with your browser send a follow-up message to PharmPK@boomer.org with "Sample size in exploratory studies" as the subject Support PharmPK by using thelink to buy books etc.from Amazon.com

Copyright 1995-2011 David W. A. Bourne (david@boomer.org)