Dear Jean,
Most people are not followed for the whole 5 years.
Imagine a trial that takes 3 years to recruit then stops after 2 years
of followup.
The *first* patients will be followed for 5 years (3+2), but
The *last* patients will be followed for only 2 years
with most people spread between 2 and 5 years of followup.
So the survival curves are based on fewer and fewer folk toward the end
of the curves.
Hence the ends of survival curves are dangerous for reading of risk
differences as their CIs are so wide.
See:
http://www.graphpad.com/www/Book/survive.htm
particularly on Censored Survival Data.
Happy reading,
Paul Glasziou
Jean Levasseur wrote:
> I am trying to interpret the survival curves found in the JUPITER
> trial published in the NEJM last november and I am struggling with the
> interpretation of these curves.
>
> At the start of the trial there were 8901 patients in each group, the
> trial was stopped by the sponsor after 1,9 years for apparent large
> benefit for the treatment group, and there remained less than 200
> subjects at risk at the end of the 5 year follow-up. The hazard ratio
> is calculated at 5 years and shows a large benefit, but less than 2%
> remained at that time...
>
> I am no expert in survival curves, but isn't there something wrong
> with the interpretation ??
>
> Could someone help me with the interpretation of these curves?? and
> the decay of the number at risk and its effect on the interpretation
> of such a survival curve.
>
> The JUPITER trial slides can be found at :
> http://www.brighamandwomens.org/publicaffairs/Images/JUPITER%20Slides%20for%20BWH%20to%20Post%20November%209.pdf
>
>
> Many thanks !
>
> Jean Levasseur MD, MSc
>
--
Paul Glasziou
Director, Centre for Evidence-Based Medicine,
Department of Primary Health Care,
University of Oxford www.cebm.net
ph - +44-1865-289298 fax +44-1865-289287
|