cox proportional model - interpretation

[riyaaora275]

I have applied a cox proportional model on discontinuation data . There are two groups and OD and OW. 

If the hazard ratio from cox proportional model is such-

Class	HR
OD	1.456

Then do can we interpret it as OW are 45.6%more likely to stay on treatment as compared to OD? Or should we say OD are 45.6% more likely to discontinue? Or do they mean the same?

[vellad]

If the reference group here is OW i.e. Discontinuation rate HR (OD vs OW) = 1.456, then "OD is 45.6% more likely to discontinue" seems right.

If you need to use the language "OW are X% more likely to stay on treatment as compared to OD", you need OD as the control group so the HR becomes 1/1.456=0.687. Thus "OW are 31.3% less likely to discontinue as compared to OD" or "OW are 31.3% more likely to stay on treatment as compared to OD"