 |
 |
Dear all, Here's another forwarded e-mail -this time from Pat. Is anyone else having problems sending to PRIMSTAT? I'll try and sort out what's going on and let you know what the problem is. Sandra ----- Original Message ----- From: Pat Yudkin To: [log in to unmask] Sent: Friday, June 18, 2004 5:35 PM Subject: randomisation/minimisation Dear Sandra Another email for the PRIMSTAT list. It wouldn't accept my offering either! best wishes Pat Dear Robert (or is it Rob?) To answer your question sensibly, it would be helpful to know the intervention and the outcome, and why stratifying by clinic was thought to be important in the study design. In general, it's worth restricting randomisation only for an important prognostic factor, and severity of disease is an obvious choice. But it's difficult to see how GP clinic could be such a factor, when the number of subjects in each clinic is so small and (we assume) there is a large number of clinics. I'd endorse Toby's view that you consider whether there is another higher order factor that is strongly predictive of outcome and that may be related to whatever it is about the clinic that you are worried about! If so, you might want to stratify by that factor. Random allocation is a very powerful method, and unless you're very unlucky the different clinics should balance out well between the trial arms, even without any restriction. Best wishes Pat Dr Pat Yudkin Reader in Medical Statistics Department of Primary Health Care University of Oxford Old Road Campus Oxford OX3 7LF UK Tel: +44 (0)1865 226916 Fax: +44 (0)1865 227036