After moderate research, including your links, I agree that the data is strong that PLA2 is independent of CRP - albeit not as much so for LDL or HDL.
But the real question is the total C-statistic (aka ROC AUC) of a combined Framingham, CRP and LPA2 model. In the overall population it comes close to the c-statistic of the calcium score (C=0.76 vs a little less than 0.80 for calcium score) - see below link (note that I just found this link after fairly extensive searching by searching under "c-statistic" instead of "ROC AUC"). But I'd be surprised if it could match the predictive value in the sicker population - i.e. the best thing to do probably is to use the non-radiative improved Framingham model (assuming it is published as a computable thing) to decide whether you where at moderate risk or above and then if at moderate risk or above get a calcium score test to decide how aggressively to treat - e.g. 0=low dose statins or nothing, <10 moderate dose statins, >100 high dose statins and niacin.
BTW - there seems to be a lot of "air play" about the fact that CRP and LPA2 are 'complementary' and the evidence for it is that if you take the highest quartile of CRP and the highest quartile of LPA2 the risk is >10x the low/low set. But this is misleading math because, since CRP and LPA2 are very independent, the top quarter of each equates to 1/16 of the total population being compared to the lowest risk 1/16th of the population. Of course it is going to be an astronomical risk.
Market Data 
Markets 
AI
