Linear Cox regression = Dose dependent relationship?

1

Hi data enthusiasts!

I have a simple question. I have a continuous variable, which I assessed as a predictor for mortality using linear cox regression, yielding significant results. Can I simply infer that the association is dose-dependent ?

Thanks for your help!

2

yes, but a linear relationship is implausible

3

thanks! why is it implausible ?

4

in nature it’s always implausible i would think, eg what happens near the extremes of your covariate? the protective effect plateaus

5

ah ok you mean practically its never the case… yes, I agree! thank you!

6

i should say, the term ‘dose response’ is used when the factor is categorical, ordinal, i’ve not seen it used when the factor is continuous. The point is that it is reassuring to see the effect increases as the level of the factor increases, and we might refer to this as a dose response (metaphorically)

1 Like
7

True… it was just because the variable is usually used in a dichotomized manner, where the association with mortality was shown. And now I modeled it as continuous to see a „dose response“ relationship, as before it was just yes/no. That is the background to why I spoke of dose response :slight_smile: