A New Strategy to Determine Whose Cholesterol to Measure for Primary Prevention of Cardiovascular Disease -- A Modelling Study Using UK and Chinese Data

Objectives: Since the mid 1990s, most guidelines on primary prevention of cardiovascular disease (CVD) have recommended regular cholesterol measurement for all adults or those above a certain age (which is known as mass screening). Cholesterol measurement comprises a large cost of CVD prevention and is not necessarily required in those who do not need drug intervention. In order to reduce this cost, we have developed a new selective cholesterol screening model in order to determine whose cholesterol should be measured for drug prevention. The model was evaluated and compared with other widely adopted models in basic model performance as well as cost effectiveness.;Methods: The new model has two steps. In the first step, we purposely over-estimated the majority of respondents' CVD risk by substituting a sufficiently high hypothetical cholesterol value in the risk estimation. We then recommend cholesterol measurement only to those with the estimated CVD risk above a predetermined risk threshold for drug treatment. In the second step, the CVD risk is re-estimated based on the individual's real cholesterol consentration. Those with a risk above the treatment threshold are recommended for drug treatment.;We evaluated the performance of our two-step model with data from the Health Survey for England and re-evaluated it with data from the China Nutrition and Health Survey 2002. By varying the hypothetical cholesterol values and treatment thresholds in CVD risk, we assessed the sensitivity, specificity and proportion of the population who need to measure cholesterol and compared it with the US mass screening model and the UK NICE selective screening model. We further compared the costs and CVD events avoided in the compared screening programmes. We also examined how the age restriction should be set in cholesterol screening programmes.;Results: As compared to mass screening, our new model can achieve a high sensitivity and save some 80% the cost of cholesterol measurements. The sensitivity depends mainly on the hypothetical cholesterol level used and seems independent of population's CVD risk, treatment cut-off values and risk prediction model. The model performed well in almost all the conditions tested. When the hypothetical cholesterol was set at MEAN+2SD, the resulting sensitivity of our selective screening model was almost always above 95% and close to the expected 97.5%. The sensitivity was only compromised slightly if cholesterol is not measured at all for the Chinese population. Furthermore, in order to save more costs, cholesterol measurement could be better restricted to men aged 50-84 and women 60-84 years regardless of the screening model used. In CVD events prevented, mass screening is always the best but our model can prevent almost as many. In costs, mass screening is always the most expensive but our model can save all or most of the cost. The NICE selective model can perform as well as our model only when it is used in an appropriate manner and in certain circumstances.;Conclusion: Our new cholesterol screening model has a high sensitivity which is comparable to that of universal screening programs but can save most of the cost on cholesterol measurements. In where resources are particular sparse, our model can also perform well by applying it only to certain age groups, which will further save cholesterol measurement costs. Cholesterol measurement could even be completely avoided for the Chinese population if our findings can be re-confirmed correct with more updated data.