At first I thought this might be a small number issue (7 vs 1 or 14 vs 2), but turns out the background rate of VTE (>25% on enox) is much higher than I had realized, so it's more likely this is not a statistical fluke.
The dose-response curve could be very different for this type of drug in a condition like this than for more typical drugs. So a non-linear dose response of some sort is not beyond the bounds of possibility. (Note we are dealing with two different dose response curves here - first the impact of the drug on Factor 11 levels, and second the impact of Factor 11 levels on VTE incidence. That makes a nonlinear overall dose response more plausible to me).