 |
 |
In a hypothetically perfect and unrealisable trial all patients would enter a trial like UKPDS on the same day and (assuming they were still alive) leave the trial on the same day. in reality subjects will be followed up for varying lengths of time so it is necessary to adjust for this. to say that the event rate is 40.9 per 1000 patient years is to say that it would take one year of follow up of 1000 patients before 40.9 suffered the event or 2 years if 500 were followed up. an alternative method that does not adjust for variable length of follow-up is to state the event rate over the average length of follow-up and that is how you get a different figure altogether as james mccormack has demonstrated. i have posted a fuller discussion of this with examples on: http://www.suffolk-maag.ac.uk/ebm/pt-yrs&NNTs.html -- Kev Hopayian, Seahills, Leiston Rd, Aldeburgh, Suffolk IP15 5PL, England ---------- From: "James McCormack" <[log in to unmask]> To: <[log in to unmask]> Subject: percentages versus patient years Date: Fri, Oct 29, 1999, 7:15 am Hi: I was wondering if anyone could give me a brief synopsis (or where to find it) of the advantages and disadvantages of presenting clinical trial data as either percentages (events/patients) versus events per 1000 patient years. In particular, in the UKPDS 33 trial any diabetes related endpoints were reported as 40.9/1000 patient years in the intensive group and 46.0/1000 patient years in the conventional group. However when one calculates the absolute percentages it comes out as 963/2729 = 35.3% in the intensive group versus 438/1138 = 38.5% in the conventional group. Is the difference because more than one event can occur in one patient? In contrast, when one calculates the absolute percentage differences between the groups for MI or deaths the absolute differences seem to be fairly similar. Thanks in advance! Dr James McCormack Associate Professor Faculty of Pharmaceutical Sciences University of British Columbia Vancouver, B.C. 604) 822-1710 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%