 |
 |
This actually does occur relatively frequently (although I am not aware of the systematic review of the topic). An additional reason for the observed discrepancy is that the Cox analysis deals with the KNOWN variables, while crude rates are based on the (random)assigment to the treatment that may reflect both known and unknown variables. Furthermore, there are so many tips & tricks when in it comes to performing (and reporting) the Cox analyses- each of which may be a "culprit" for the observed differences. I think it is high-time that someone publish a paper on the proper way how the Cox analyses should be done and reported (we need something similar to the CONSORT statement). I hope out statistical colleagues will accept this challenge. ben Benjamin Djulbegovic, MD,PhD Professor of Medicine and Oncology -----Original Message----- From: Evidence based health (EBH) on behalf of Dan Mayer Sent: Fri 7/29/2005 1:29 PM To: [log in to unmask] Subject: Re: Question about Cox Regression model versus Crude Event Rates Hi guys, That is fascinating. Of course, the thing that immediately struck me was that the outcome was a composite of a biochemical marker (serum creatinine), a chronic disease state (ESRD), and death. All of these have different weights when it comes to patient values (I would hope). Therefore, before even doing the statistical test, the authors have created a 'void of reality' situation. Therefore we really shouldn't care about which test is being used. The use of the Cox test may be a legitimate of controlling for the differential times in teh study, but you have to ask why is that difference there in the first place. Was one group more likely to be in the study longer and could this bias the outcome? Cheers, Dan **************************************************************************** Dan Mayer, MD Professor of Emergency Medicine Albany Medical College 47 New Scotland Ave. Albany, NY, 12208 Ph; 518-262-6180 FAX; 518-262-5029 E-mail; [log in to unmask] **************************************************************************** >>> Mike/Linda Stuart <[log in to unmask]> 07/15 3:11 PM >>> We are evaluating the literature for managing chronic kidney disease and would very much appreciate opinions regarding the following issue in analyzing the results of an RCT comparing enalapril with placebo in diabetics. Ref: Brenner BM, Cooper ME, de Zeeuw D, et al., for the RENAAL Study Investigators. Effects of losartan on renal and cardiovascular outcomes in patients with type 2 diabetes and nephropathy. N Engl J Med. 2001 Sep 20;345:861-9. The patients were type 2 diabetes with nephropathy. Baseline characteristics, including disease severity, were similar in the two groups. Outcomes were - --the composite of a doubling of the baseline serum creatinine level, end-stage renal disease (ESRD), or death, --and the composite of cardiovascular morbidity or mortality. The authors used the Cox regression model and reported significant RRRs for all of the endpoints except death. The authors explain that there is often a difference between the risk reduction as determined on the basis of the Cox regression model and the risk reduction as determined on the basis of the crude rates of events. They state that the difference results in part from the fact that the Cox regression model accounts for the time at risk - i.e., the longer average follow-up in the losartan group than in the placebo group. In this study the mean follow-up was 3.4 years and actual follow-up ranged from 2.3-4.6 years. The authors presented the numbers of events per 100 patient-years of follow-up. The data is shown below: We recalculated the P value for the primary outcome, basing our calculations on crude event rates. We did not get statistically significant results (p=0.235). We are not statisticians but believe that the crude event rates should also be statistically significant for the study to be valid. We would be most grateful for opinions regarding this question and our belief that we should be able to use crude event rates for these outcomes. -- Thanks, Michael Stuart MD and Sheri Strite Delfini Group, LLC [log in to unmask] www.delfini.org <http://www.delfini.org/> ----------------------------------------- ########################################################################### ## This transmission may be confidential or protected from disclosure and is only for review and use by the intended recipient. Access by anyone else is unauthorized. Any unauthorized reader is hereby notified that any review, use, dissemination, disclosure or copying of this information, or any act or omission taken in reliance on it, is prohibited and may be unlawful. If you received this transmission in error, please notify the sender immediately. Thank you. ######################################### ####################################