Dear all
I would be extremely grateful if someone could provide me with some advice on the following:
Quick version: How should one decide how large a pilot study should be, in order that it is a representative enough sample to serve as the basis for a sample size/power calculation (that will in turn determine how large the subsequent, substantive study needs to be)?
Background info:
I am currently working on a protocol (due for submission to ethics imminently) for a pilot study, which is to be used as the basis of a sample size calculation for a clinical trial looking at the effects of statins on carotid artery thickness in Lupus patients. They intend to run a two arm, placebo controlled, age and sex matched trial, all participants to be Lupus patients, half to receive the drug in question. Those running the trial were initially advised by the university's dedicated stats department that they would need 30 patients in each arm to produce a reliable estimate of the mean, but our clinical trials coordinator has just informed them that ethics will not generally approve a pilot study with more than 10 patients in each arm.
The problem in full:
The rate at which arterial thickness changes is slow. The substantive study is set to run for 2 years. Although there is existing longitudinal data on 'normal' patients (not lupus sufferers) showing a twenty percent reduction over a year with statins, no-one knows exactly how the drug will affect Lupus patients (they suspect the reduction will be greater as Lupus sufferers are generally worse off to start with). The researchers are naturally loathe to run a pilot for one or two years with 10 patients in each arm only to find that this is too small a number to produce a meaningful mean/sd etc on which to base subsequent sample size calculations for the substantive study. I am currently recommending that we compromise and see what happens if we follow 20 patients for a year, but I am still left plucking a sample size from the ether, as it were. This question is further compounded by the fact that the arterial thicknesses are small (mean in one lupus cohort 0.71 +/- 0.14), and the ultrasound machine used to measure them only goes to 2 decimal places.
Any suggestions you have as to how best to proceed would be much appreciated.
Thanks you for your time
Regards
Liz Hensor
Dr Elizabeth M. A. Hensor PhD BSc (Hons)
Data Analyst
Academic Unit of Musculoskeletal and Rehabilitation Medicine
36 Clarendon Road
Leeds
West Yorkshire
LS2 9NZ
(0113) 3434944
[log in to unmask]
|