|
|
Hello
I have been performing some Kaplan Meier analyses over quintiles derived from continuous variables relating to clinical data. The survival patterns displayed on the corresponding curves approximate to what would be anticipated from a clinical perspective. On progressing to Cox regression analyses, I have used the corresponding continuous variables. Thus for example, by applying a suitable multiple (1/5) in the first instance, I have been able to examine the effect of a 5 unit change in the value of the continuous variable on the risk of mortality. I am rather perplexed, however, that the direction of the difference in risk has turned out to be the reverse of that which is clinically meaningful in such cases. I would therefore like to verify whether the relative risks calculated for continuous variables within a Cox regression analysis are in the direction
<risk following unit increase in variable/ risk before unit increase in variable>.
Alternatively, is there an inbuilt mechanism within SPSS which decides which may round this relative risk is calculated (so that for some variables I may need to take the reciprocal of the relative risk I have calculated)? I know that such a set-up arises with categorical variables, depending on the values of these variables on the top or bottom row of the spreadsheet and that it is a simple matter in such cases of checking what is in either of these (depending on which row has been used as the reference category). However, I am not sure what is happening with continuous variables and cannot understand why the continuous variable is providing a risk change, which is in the opposite direction to that anticipated for some variables but not for others (whilst this was not the case when I examined the quintiles).
If there is an underlying rule for determining direction in the calculation of relative risks with continuous variables within the SPSS Cox regression analysis procedure could you please kindly let me know. I don't wish to simply guess which relative risks to take the reciprocal of and which to leave unchanged. (I also have a the back of my mind that the anomalous results may simply reflect idiosyncrasies with the data.
I look forward to hearing from you.
Many thanks
Best wishes
Margaret
---------------------------------
New Yahoo! Mail is the ultimate force in competitive emailing. Find out more at the Yahoo! Mail Championships. Plus: play games and win prizes.
|
Locked
