Isn't the first point below just to do with the NNTs not being adequately
specified ie an NNT should always be stated in terms of 'number needed to do
what over what time'. 4S mortalities were 11.5% treatment, 8.2% placebo
after 5 years, NNT to prevent one death in 5 years = 100/3.3 = 30. As you
say, it's all about baseline risk, but (if you're willing to assume
treatment effects are linear) by varying the timeframe you can make the NNT
bigger or smaller at will (1 year NNTs are usually not very impressive, 10
year ones much more so). So your hypothetical primary prevention example
would give NNTs of 200 at 1 year, 40 at 5 years etc. If you choose a
different set of 'events' than death, you'll also get different numbers.
In terms of starting people on treatment at 15% risk, all these calculations
assume that you don't do anything else to them. What if their 15% risk is
partly because they are severely hypertensive, smoke and have a high
cholesterol. NNTs calculated on the basis of other risk factors remaining
the same may be too high because these other risk factors may also be
modified. It is of course too tedious to do the calculations properly, and
the advice I've had is that we shouldn't recalculate risk after treating one
risk factor.
I personally find the calculation of cardiac risk and its use in determining
treatment decisions very difficult. It's attractive, and easy as a one off
for the newly diagnosed hypertensive or hypercholesterolaemic person. But in
practice, it's difficult if you keep seeing the patient over time. Do you
recalculate using new numbers after a bit? How often? If they stop smoking,
would you stop their statin on the basis that they would now be below 15%?
Given the huge effect of age, is it sensible to take a 10% risk in a younger
person as 'OK' although you know that in a few years time it will be 20%
even if other risk factors are no different?
There is also some evidence that the Framingham equation overestimates risk
in UK populations (the British Regional Heart Survey BRHS population to be
precise) although I've only seen that as a conference presentation so far.
However, I think that is what we'd expect given that it is based on data
from the 70s and 80s (I think) and CVD has been declining since then even
without treatment changes. So our NNTs based on calculated risk are probably
optimistic.
I'd be very interested in how others find the practicalities of risk
calculation, and use of the numbers generated.
Bruce
Bruce Guthrie
Tayside Centre for General Practice
University of Dundee
This also puzzles me a bit, 4S was secondary prevention, accepting this,
if a starting 10 yr risk was say 30% then a 30% reduction would give a
risk of approx 20 % or a 10% ARR, which equates to a NNT of 10, doesn't
it?
In primary prevention say people are starting Statins at initial 10 yr
risks of 15% (as at least some are in the UK) then a 30% RRR would give
a 10 yr risk of 10% or an ARR of 5% which equates to a NNT of 20.
Am I reading this right?
|