• Cholesterol Risk Calculator Debacle – ctd.

    TIE readers are amazing. One writes me, but asks to remain anonymous:

    The cholesterol guidelines are part of a move towards risk-based prevention of CVD (as opposed to risk factor based) that’s been going on for about 10 years.

    If I understand the Times article correctly (which is difficult because the Lancet paper it’s based on isn’t available yet), it looks like what happened isn’t that the new scores have errors, it’s that they didn’t calibrate in another dataset. Roughly, this means that the average rate of events is different from predicted in one group of people to another. It has been seen in a lot of risk scores before. There are many reasons why this might happen, some that say important things about the score, but just as many that actually relate to sampling biases that are not real threats.

    One question is if this makes the transition from LDL to risk in the new guidelines a bad decision. I think it wouldn’t. A man who smokes has 5x the probability of having a heart attack or stroke prevented by a statin as a woman who doesn’t, but the old guidelines treated them identically. A 50% error in calibration does not change that. 

    The new guidelines are aggressive. This is a judgment call – what’s an appropriate number needed to treat for preventing a stroke or heart attack vs. the more common side effects of statins. They also could have made a guideline that would have been more amenable to shared decision-making. But this is being read as an attack on using risk instead of LDL, and I don’t think this seriously harms that advance.

    I look forward to more debate.


    • Aaron
      Is author (and you?) saying then calibration off, but discrimination of model adequate?

    • The irony of these guidelines is several fold. Foremost, the guidelines represent a step in the right direction by emphasizing total CVD risk rather than cholesterol alone as the basis for determining treatment. This was based on well done research that often met stiff resistance from the mainstream lipid crowd. Previous guidelines both overtreated and undertreated people, and this shift should prevent more events per person treated. The other part of the guidelines, to abandon treating to target goals, might actually reduce drug use as it will no longer be standard to keep increasing dose and adding drugs to hit some elusive threshold. However, when the guidelines were turned over to the AHA/ACC, the threshold for when to begin therapy was set at a level of risk that greatly increased the number of people being treated. As previous commenter noted, this is a judgment call and depends on your underlying assessment of how bad risks of statins are (or just philosophy about putting healthy people on drugs). FInally, the fiasco over the risk calculator just exacerbates the previous concern and plays into a narrative that this was a sloppy attempt to increase drug use (note: I dont believe this but others do). Maybe this is a like initial rollout of Obamacare — good thought, poorly executed. (please do not quote me by name)