If the proposed intervention is giving a pill, it seems fair to test that instead of a proxy. The study had hundreds of people; if the effect were big enough to be clinically significant they hopefully would have detected it, even with some patients failing to take the pills.
I'm talking about assaying the subjects' blood for prior and subsequent levels of the vitamin under test. Inasmuch as we have no way to tell whether they took the pills, we don't know whether the groups were properly randomized for their pre-intervention levels, diets, or effectiveness at metabolizing the supplement. It may even be the case that people with high CAC scores in the first place (the test population) already suffers from one or more of these problems across the board.
It isn't that hard to test people's blood, and yet the researchers prefer us to take it on faith that their sorting of control and intervention groups ironed all of this out? If I sound unreasonable, I don't apologize for it: public health decisions are important.
The question in this trial was: Will people in this at-risk population have less calcium buildup in the heart if we give them vitamin K2? The answer was, on average, no. That's the key piece of information for public health.
Blood tests would tell us if the randomization ended with the two groups being very different, but the prior probability of that is small. Blood tests would also tell us if there is some subgroup that benefits more from the treatment, but with an increased risk of getting the answer wrong due to smaller sample sizes, which would lead to worse public health decisions.
