Re: Fallacies in Predicting CVD events.
"Prediction is very difficult, especially about the future." - Niels Bohr
Reginald Marsh now appreciates using statistical methods, that betting on a horse is chancey, but I can safely predict that a horse ( and not a Donkey) will win the Derby.
Just as in quantum physics - individual predictions can be very imprecise, even though population aggregates are close to the mark. This is usually acheived by block use of 'population adjustments' such as Hippisley-Cox's geographically based deprivation correction, or Marsh's 'Maori' factor..
Tim Reynolds rightly points out that we use very blunt single point estimates of the known continuosly-variable Risk factors. The fact that our CVD risk comes close at all is remarkable, given that some 50% of causation remains unknown. We can go on tweaking the CVD score with race, waist measurement, exercise etc.. But to what avail?
I have computed Framingham CVD Risks for my own practice Population:-
Percentage of population at High Risk, by Gender and Age.
Decade _MALE ___FEMALE
We could compare two strategies:-
POPULATION-BASED: Offer every MALE over 50 a Statin ( after exercise, diet, ant-smoking advice, Aspirin, etc. ). Females lag 15 years behind Males at risk.
INDIVIDUAL-BASED: Assess each individuals numerous risks with the most accurate adjustments available, and then offer a Statin when risk exceeds a cost/benefit threshold.
NICE has assessed the costs of both strategies, I presume. NICE has clearly pre-empted Marsh's evidence that Risk Prediction rules out large numbers of people from risk, by continuing to recommend treatment only for those at high risk. Unless the costs of this additional assessment effort is large, then its efficiency as a strategy remains unassailed. The strategy, in essence, is not to 'waste' scarce and expensive medicine on those LEAST likely to benefit.
Competing interests: Time and Effort
Competing interests: No competing interests