Rapid Responses to:
|
|
Rapid Responses: Rapid Responses published:
-
![[Read Rapid Response]](/icons/misc/sectionDOWN.gif)
Family history of CHD should not be ignored in primary prevention
- Julian M Stribling (27 July 2002)
-
Default values should not be used
- Trefor J Roscoe, Peter Tyerman, Gill Tyerman (27 July 2002)
-
Statin' the obvious
- L S Lewis (28 July 2002)
-
What about indirect costs related to aspirin?
- Adam Jacobs (2 August 2002)
-
15% or 30%?
- dougal jeffries, Isles of Scilly TR21 0HE (5 August 2002)
-
RESPONSE TO CORRESPONDENTS AND REANALYSES
- Tom Marshall, B15 2TT (7 August 2002)
-
Simplified Risk Assessment Methods May Reduce Costs But Miss Necessary Treatments
- Jin Ling TANG, Yu JIANG, Jame XJ ZHANG, Jean WOO (20 August 2002)
|
|
|||
|
Julian M Stribling, Research Nurse Division of Cardiology, University of Leicester, Glenfield Hospital, Leicester, LE3 9QP
Send response to journal:
|
EDITOR- Marshall & Rouse1 make an interesting point about the potential health benefits and resource implications involved in developing primary prevention strategies for cardiovascular disease. However, there is one important point that the authors fail to highlight. All of the guidelines for the prevention of Coronary Heart Disease (CHD) (including those utilised for the National Service Framework2 (NSF)) use the notion of absolute risk, and largely ignore the importance of family history of CHD on the relative risk for an individual. A positive family history of CHD may be the single most important risk factor, and is at least as important as smoking. Having a first-degree relative affected by premature CHD increases an individual’s relative risk by 2 to 8-fold compared to individuals without a family history3. Furthermore, people with a positive family history of CHD are a readily identifiable group, even though they are often over-looked in primary prevention4. If we are to truly develop mechanisms and strategies for the primary prevention of CHD (which is implicit in Standard 4 of the NSF2), then a family history of CHD cannot be ignored. Indeed, perhaps a practical model of targeting individuals who may be at significant risk of developing CHD would include a systematic review of the relatives of individuals with established disease, rather than the largely opportunistic screening that occurs now. 1. Marshall T & Rouse A. Resource implications and health benefits of primary prevention strategies for cardiovascular disease in people aged 30 to 74: mathematical modelling study BMJ 2002;325:197 ( 27 July ) 2. National Service Framework for Coronary Heart Disease. Department of Health. (2000). London, HMSO. 3. Samani NJ & Singh RK. (2001). What is known about the genetics of acute coronary syndromes? In: de Bono D. & Sobel BE (Ed). Challenges in Acute Coronary Syndromes. Blackwell Scientific, London. 4. Hengstenberg C et al., (2001). Siblings of Myocardial Infarction patients are overlooked in primary prevention of cardiovascular disease. European Heart Journal. 22:11: 926-933. |
|||
|
|
|||
|
Trefor J Roscoe, GP Informatics Tutor North Trent Sothall Medical Centre, 24 Eckington Road, Sheffield S20 1HQ, Peter Tyerman, Gill Tyerman
Send response to journal:
|
We write to express our concern about several aspects of the paper by Marshall and Rouse. We feel that the assumptions they have made are too wide-ranging and lead to some erroneous conclusions. Using the Framingham equation with average values is only going to stratify the population by age sex and diabetic status, a too simplistic analysis on which to base screening and intervention decisions. Other aspects of the paper are misleading, confusing or not applicable to current practice. Stratifying the risk of the population by looking at only the age sex and diabetic status of the patient is only going to prove what is well known, diabetics and the older age groups are more at risk. By assuming that they are all non-smokers they are excluding the most important single weighting in the Framingham calculation. Most practices have a register of smoking status for the majority of their patients so this should have been included. By their own admission their method may not be able to screen all of the population so some people will loose out. Their method will not pick up the highest risk with the most to loose, for example the 50-year-old non-diabetic overweight smoker, as they will concentrate on the oldest and the diabetic first. Adding a few extra years to the life of a 70-year-old non-smoking normotensive diabetic may be laudable but just because it is easier and cheaper does not mean that they should be prioritised to be assessed first as this method seems to suggest. Many of the other assumptions they make demonstrate that they have not looked carefully enough at the day-to-day realities of General Practice. Very few GPs would be able to stabilise someone on antihypertensive treatment with only four prescriptions in a year and only two follow-up consultations. Government policy is to issue no more than 30 days supply and thus they underestimate dispensing costs. Not all of this work can and should be done by practice nurses. New diagnoses will be picked up that require medical assessment and many patients need more careful clinical assessment to decide on the correct treatment. Thus their economic assumption that two 20 minute consultations by a practice nurse is all that is needed to initiate treatment is in our view wrong. Their conclusion that statins and angiotensin converting enzyme inhibitors cost more than finding and treating new patients and thus should be avoided flies in the face of the evidence base. We cannot understand what they are proposing. It would be actionable to not do a cholesterol level on a high-risk patient or if a high level had been found, negligent not to treat. Perhaps they are suggesting that by applying their method you could ignore hypercholesterolaemia and not use the risk tables such as the Sheffield table to decide on treatment. If you screen the population for some risk factors and assume the average for the others you will be doing a great disservice to many at risk patients who could have had cardiac events or strokes prevented. Although the health of the whole population and the costs of improving it are the proper area of study for epidemiologists, they are not real world medical practice. We have a declaration of interest in this area as we have demonstrated a method of screening the whole population, and at the same time educating them to be involved in reducing their risks. Over 95% of a practice population has been screened in less than 5 years and significant reductions in risk achieved by intervention and rescreening. An observational paper on this method is being considered for publication. References for all of the points made in this letter can be had from the authors; we felt that we could not do justice to our arguments with the limit on references in a quick response. In conclusion, we feel Marshall and Rouse’s paper adds little to the subject other than confusing people into thinking there is a quick and cheap method of shortcutting the hard and expensive, but necessary work of finding those at risk of heart disease and saving their lives. We await their response with interest. |
|||
|
|
|||
|
L S Lewis, GP Surgery, Newport, Pembrokeshire, SA42 0TJ
Send response to journal:
|
Using a complex model of 'virtual patients' to model a real general practice, Marshall and Rouse have managed to prove the obvious - 1: A Rouse-Marshall (RM) strategy is cheaper than a Joint British Recommendation (JBR) strategy. But an RM strategy differs from a JBR strategy precisely by avoiding the cost of screening those n patients deemed at first assessment to be low-risk ( on the basis of known Age, sex, Diabetes, and substituted default values for other Framingham equation factors ) This cost is easily estimated ( one screening x n = approx £16000 ) and accounts for the fixed gap between RM and JB graphs in Fig.1 2: Depending on how much you spend on treatment, then more patients would benefit.. 3: By deploying the most cost-effective treatments first you will maximise that benefit. 1st line : Aspirin, Thiazide, Beta-blocker 2nd line : + Other anti-hypertensives 3rd line : + Statins RM3, RM2, and RM1 strategies (not the reverse notation) differ precisely by this progressive increment in treatment. Thus more lives can be saved, but at burgeoning but easily estimated cost. Strangely, they neglect a very cost-effective treatment , namely smoking cessation ! But how do GPs manage to know who is a diabetic, without seeing all patients at some point ? How many patients, and what harm, will accrue by not measuring every BP and every cholesterol ? Even a low-risk non-diabetic female might occasionally be found to have a severe hypercholesterolaemia .. What ethical problems flow from giving Statin to smokers, but not to non- smpkers with otherwise identical risks ? In practice, we GPs 'screen' patients by a combination of organised and opportunistic 'check-ups' according to a complex of patient's perceived need, professional peer-pressure, and national service 'at an initial risk assessment' - in other words - normal practice ! Yours sincerely Dr L S Lewis General Practitioner Surgery, Newport, Pembrokeshire |
|||
|
|
|||
|
Adam Jacobs, Director Dianthus Medical Limited, London SW19 3TZ
Send response to journal:
|
Marshall and Rouse's strategy requires many people to be treated with aspirin. I note that they have adjusted their results to take account of the costs associated with major bleeding events caused by aspirin, but I wonder how robust their results are to alternative assumptions about the indirect costs of aspirin use. Aspirin use could increase costs not only through major bleeding, but also through less major gastrointestinal adverse effects. Have the costs of these been considered? What proportion of patients might develop non- bleeding ulcers which may still required endoscopy and drug treatment? Would the costs of such treatment materially alter the conclusions about the cost effectiveness of the treatment strategy? I would have found the results of the paper more convincing if the uncertainties surrounding these estimates had been included in a sensitivity analysis. |
|||
|
|
|||
|
dougal jeffries, gp the health centre, st mary's., Isles of Scilly TR21 0HE
Send response to journal:
|
I have noticed with some disquiet that several recent papers on lowering CHD risk, including Marshall and Rouse, state that statins should be prescribed when an individual's 10-year risk is assessed as being 15% or higher. Until now I have been using the 30% threshold, and correspondence with the clinical lead for CHD in our local PCT confirms that this is his practice as well. Where and when did this halving of the threshold occur? Is this now the consensus? Has the DoH announced that it is affordable? For as long as I am being criticised for exceeding my annual prescribing budget, I cannot see how I can possibly justify changing to a lower intervention threshold. Given that more than twice as many people must have a 15% than a 30% 10 year risk, my statin prescribing costs would more than double. How our other practices interpreting the Joint Societies' recommendations? Dougal Jeffries |
|||
|
|
|||
|
Tom Marshall, Lecturer in Public Health Birmingham University, B15 2TT
Send response to journal:
|
SUPPORT FOR OVERALL CONCLUSIONS We note that there is no challenge to our principal conclusion: authors of guidelines should explicitly model the resource implications and health benefits of following their recommendations. We agree with Lewis that our main findings are mere common sense. Unfortunately it is not so common as to have informed expert recommendations on CVD prevention. INDIVIDUAL VERSUS POPULATION PREVENTION We concur with Roscoe et al that there is an unresolved - and hitherto unacknowledged -dilemma in CVD prevention. A clinician acting in the best interests of an individual patient prescribes a statin; a clinician acting in the best interests of the practice population does not prescribe a statin but instead finds and treats more patients. The former is consistent with the clinician's duty of care but makes for a poor prevention policy. The latter is consistent with utilitarian principles (the greatest good for the greatest number) implicit in government policy, but puts clinicians in the uncomfortable position of making explicit prioritisation decisions. A SIMPLIFIED PRE-SELECTION STRATEGY Readers who are interested in applying our pre-selection strategies may find the following suggestion helpful. A woman's risk of CVD is approximately equivalent to that of a man 12 years her junior. A good approximation of our pre-selection rankings is obtained by assigning a ranking number to every patient. For men it is their age, for women their age minus 12. The practice prioritise patients for CVD risk assessment by their ranking number (highest ranking number first). FAMILY HISTORY AND CVD RISK Stribling suggests that targeting of CVD prevention would be better served if we incorporated knowledge of a patient's family history of CVD to help refine our estimate of their risk. We agree. However this is not straightforward. Increased risk of heart disease associated with a positive family history partly reflect the tendency of smoking, raised cholesterol levels, high blood pressure and diabetes to cluster in families. These factors are already included in the Framingham risk equation. The effect of family history apart from these factors needs to be quantified and included in a risk equation. Simple addition of a "family history" factor to present risk equations could result in double counting of risk factors and overestimation of risk. We know of no risk equation which includes family history and has been validated in the UK population. SMOKING HISTORY AND CVD RISK Roscoe et al suggest that information on smoking should be included when calculating prior risk estimates. We agree. A prior risk estimate should be based on all available data: age, sex, diabetic status, smoking history, mean blood pressure and mean cholesterol levels. We modelled our approach applied in a practice with the minimum of electronically recorded data: age, sex and diabetic status (which can be inferred from prescribing records). Practices with electronic records of some or all their patients' blood pressures and smoking histories, can improve the accuracy of pre- selection. This serves to strengthen our finding that pre-selection strategies are more cost-effective. PRIMACY OF SCIENCE IN THE DEBATE In advancing their point Roscoe et al stress the importance of their "real world" clinical credentials. Arguments based on authority have no place in scientific debate. Our analysis suggests that clinical staff time may be more productively spent assessing a pre-selected population than assessing an entire practice. We acknowledge this is unlikely to persuade clinicians already committed to the latter strategy. However, readers interested in the scientific merits of our analysis may find the following response informative. ALTERING WORKLOAD ASSUMPTIONS Our findings remain robust in the face of almost any set of plausible assumptions. In keeping with the scientific tradition of trying to refute one's own findings we deliberately chose assumptions likely to weaken our case. Roscoe et al suggest GPs typically issue 12 prescriptions a year (rather than 4 as we suggest); they suggest CVD risk assessment and follow -up is carried out by GPs rather (than practice nurses as we suggest); and they suggest initial stabilisation on treatment requires more than two consultations a year. We present the results of an analysis using Roscoe et al's assumptions. GPs issue 12 prescriptions a year, all assessment and follow up is by doctors, patient assessment takes 1 hour of clinical time, initial stabilisation takes 1 hour of clinical time in the first year and follow-up in the subsequent 4 years takes 3 hours of clinical time. The cost of an hour of GP time is £118. [1] These assumptions strengthen our findings: pre-selecting patients for assessment is more cost-effective than assessing the entire patient population; strategies. This is principally because assessing 939 eligible patients now costs almost £120,000 over five years (939 x 1 hour x £118 + blood tests). It is also apparent that using a pre-selection strategy and not prescribing ACE inhibitors or statins, the practice can prevent 47% of preventable CVD at 17% of the cost of a traditional strategy using ACE and statins. [We cannot post the graph on the BMJ website, but it has been sent to Roscoe et al and is available to anyone who requests it by e-mail.] STATIN VERSUS IDENTIFYING AND TREATING A NEW PATIENT Even with these extreme assumptions it is easily verifiable that it costs less to identify and treat new patients than with aspirin, hydrochlorothiazide and atenolol than to prescribe statins to existing patients. In this simple example, costs are undiscounted. Simvastatin 10mg (5 years) = £1227 (simvastatin 20mg = £1987) Assessment and follow up (5 years) (1 hour + 4 hours) x £118 per hour = £590 Aspirin, atenolol, hydrochlorothiazide (5 years) = £246 Assessment, follow up and treatment with aspirin, atenolol and hydrochlorothiazide (5 years) (£590 + £246) = £836 The relative risk of CVD on aspirin in combination with antihypertensive treatment is approximately 0.6 (0.6 = 0.8 x 0.75) - a relative risk reduction of 40%. The relative risk of CVD on a statin is approximately 0.7 - a relative risk reduction of 30%. It is demonstrably more costly and demonstrably less effective to add simvastatin 10mg to the treatment of a known patient than to find (using our pre-selection strategy) assess and treat an additional patient with aspirin, hydrochlorothiazide and atenolol. ENALAPRIL VERSUS IDENTIFYING AND TREATING A NEW PATIENT With the cost assumptions outlined above, it is less costly to add enalapril 20mg to the treatment of a known patient (£466) than to assess, follow up and treat a new patient with aspirin, atenolol and hydrochlorothiazide (£836). However it is more cost-effective to assess, follow up and treat a new patient. As was explained in the discussion, further blood pressure lowering at best reduces risk of CVD by a further 20%. Treatment of a new patient reduces their risk by 40%. A paper exploring the cost effectiveness of preventive treatments in more detail has been submitted to the British Medical Journal. COST EFFECTIVENESS OF INTERVENTIONS TO PROMOTE SMOKING CESSATION Lewis correctly points out that we do not model the effects of smoking cessation, which may be one of the most cost-effective interventions. We removed a smoking cessation element from the paper during one of its revisions. The reason for this is that there are many uncertainties about the long-term effects of interventions to promote smoking cessation and uncertainties about the costs of interventions. This produces a wide range of possible cost-effectiveness ratios - difficult to explore in a single short paper. Smoking cessation also has a different cost structure to long-term medication, with costs occurring only in the first year. This means that a longer time horizon than 5 years is likely to enhance the cost effectiveness of smoking cessation. The effects of smoking cessation can be modelled by multiplying the increased quit rate attributable to health care intervention by the risk reduction attributable to quitting. Quit rates attributable to health care interventions vary from 2.5% at 12 months (for physician advice) to 20% at 6 months (for NRT and bupropion). There is little evidence whether quit rates achieved by health care interventions are sustained in the long term but one paper suggests that 50% of quitters relapse by 3.5 years. [2] Quitting smoking at best halves risk of CVD (50% relative risk after quitting). If increased quit rates at 6 to 12 months are maintained in the long term, the average effectiveness of interventions to promote smoking cessation: Physician advice: (2.5% x 50%) = 1.25% x Prior risk NRT & bupropion: (20% x 50%) = 10% x Prior risk If increased quit rates at 6 to 12 months are not maintained in the long term, the average effectiveness of interventions to promote smoking cessation is halved. If quitting reduces risk of CVD by 30% rather than 50%, this further attenuates the benefit. In a person at 15% five-year risk it is: Physician advice: (2.5% x 50% x 30%) = 0.38% x Prior risk NRT & bupropion: (20% x 50% x 30%) = 3.0% x Prior risk These assumptions therefore produce a 27-fold range of Relative Risk Reductions attributable to smoking cessation interventions (0.38% to 10% of prior risk) this rendered analysis somewhat complex. ADVERSE EFFECTS OF ASPIRIN Jacobs asks how robust our results are to alternative assumptions about the costs of aspirin use. We agree that adverse effects are an important consideration. We have carried out a reanalysis using the following additional assumption. Patients on aspirin suffer from adverse effects (in addition to major bleeds) at the rate of 5% per five years. Treatment of each adverse effect costs £2000. This has the effect of adding £100 per five years to the cost of aspirin treatment. This assumption changes the cost-effectiveness of treatment with aspirin, but makes no fundamental difference to our findings. The most cost-effective strategy is to pre-select patients and treat them with aspirin, hydrochlorothiazide and atenolol. It is apparent that using a pre -selection strategy, identifying the 200 highest ranked patients and not prescribing ACE inhibitors or statins, the practice can prevent 47% of preventable CVD at 15% of the cost of a traditional strategy using ACE and statins. [We cannot post the graph on the BMJ website, but it has been sent to Jacobs and is available to anyone who requests it by e-mail.] THRESHOLD FOR TREATMENT WITH A STATIN: 30% OR 15% TEN-YEAR RISK? Jeffries expresses uncertainty about the treatments thresholds for statins used in our analysis. This is an important point. The Joint British Recommendations state that at a minimum, patients at 30% ten-year coronary risk (15% five-year coronary risk) should be treated and that as resources permit, this should be extended to patients at 15% ten-year coronary risk (7.5% five-year coronary risk). [5] We analysed on the latter basis. We have carried out a reanalysis using a 30% ten-year coronary risk (15% five-year coronary risk). This reduces the total number of CVD events prevented with maximum treatment of all eligible patients, but it does not alter our findings. The most cost-effective strategy is to pre-select patients and treat them with aspirin, hydrochlorothiazide and atenolol. It is apparent that using a pre-selection strategy, identifying the 200 highest ranked patients and not prescribing ACE inhibitors or statins, the practice can prevent 59% of preventable CVD at 20% of the cost of a traditional strategy using ACE and statins. [We cannot post the graph on the BMJ website, but it has been sent to Jeffries and is available to anyone who requests it by e-mail.] LIFE YEARS GAINED BY PREVENTION AT AGE 50 AND AGE 70 Roscoe et al prefer to prioritise 50 over 70 year-olds for CVD prevention, apparently because of the latter's shorter life expectancy. Life expectancy at 50 is 25 years for a man and 29 for a woman. Life expectancy at 70 is about half this: 11 years for a man and 13 years for a woman. It may seem superficially plausible to suggest that the 50 year- olds therefore benefit more from prevention. However, CVD is less common and has a lower case fatality rate in 50 year-olds than 70 year-olds. (For example, case fatality after MI is 27% at age 50 and 53% at age 70.) [3] Because of this, CVD death is much rarer in 50 than 70 year-olds. Among 50 year-olds in England and Wales, there were 12 CVD deaths per 10,000 men and 4 per 10,000 women in the year 2000. Among 70 year-olds, there were 111 CVD deaths per 10,000 men and 56 per 10,000 women. [4] Per 10,000 50 year-old men and women, deaths due to CVD accounted for 312 and 124 lost years of life respectively. Per 10,000 70 year-old men and women: 1,243 and 739 lost years of life respectively. If our objective is to add years to life, we have respectively the possibility of 4 times more effect and 6 times more effect in 70 than 50 year-old men and women. This is not to say that Roscoe et al's preference for prevention in 50 year-olds is necessarily irrational. Economists use the concept of revealed preference to infer decision-makers' valuations. Roscoe et al's preference may simply indicate that they attach different values to a years of life gained by their middle-aged and elderly patients. By preferring to make 4 to 6 times the effort to add a year to the lives of their 50 year-old than their 70 year-old patients, they implicitly place 4 to 6 times the value on the life-years of 50 than 70 year-olds. REFERENCES 1. Netten A., Curtil L. Unit costs of health and social care 2000. Personal Social Services Research Unit. University of Kent at Canterbury, 2000. 2. John A Stapleton, Gay Sutherland, and Michael A H Russell How much does relapse after one year erode effectiveness of smoking cessation treatments? Long term follow up of randomised trial of nicotine nasal spray BMJ 1998; 316: 830-831. 3. R M Norris on behalf of the United Kingdom Heart Attack Study Collaborative Group Fatality outside hospital from acute coronary events in three British health districts, 1994-5 BMJ 1998; 316: 1065-1070. 4. ONS mortality data from England and Wales in 2000. 5. Wood D. Durrington P. Poulter N. McInnes G. Rees A. and Wray R. on behalf of the Societies. Joint British recommendations on prevention of coronary heart disease in clinical practice. Heart 1998;80:Supplement 2. PS - unfortunately we are having e-mail server problems here at Birmingham University following a lightning strike last weekend. I will try to send copies of the appropriate reanalyses to correspondents as soon as possible but they may arrive from a different e-mail address! |
|||
|
|
|||
|
Jin Ling TANG, Associate Professor in Epidemiology & Community Medicine The Chinese University of Hong Kong, Yu JIANG, Jame XJ ZHANG, Jean WOO
Send response to journal:
|
Simplified Risk Assessment Methods May Reduce Costs But Miss Necessary Treatments Editor---Drug treatment for primary prevention requires an assessment of cardiovascular risk.1 The widely used method for risk assessment is the Framingham risk equation and user friendly but simplified risk tables and figures derived from this equation.1,2 Simplified methods may be simple and reduce costs but are less accurate and could lead to unnecessary treatments and misses of high-risk patients who need the treatment.3 Marshall and Rouse proposed a new two-step risk assessment method, which in essence is a return to early simplified methods.4 In order to reduce costs, the risk assessment was simplified, in step 1 of the method, by replacing actual data with age and sex-specific average blood pressure and blood cholesterol and by assuming every person is a nonsmoker. Then, step 2 identifies, from potentially high-risk persons, truly high-risk persons by using the Framingham risk equation, the gold standard. Due to the approximation in the three major risk factors in step 1, the Marshall- Rouse method would inevitably miss some genuinely high risk persons. We estimate the percentage of people who would be missed by the method. We used the data on risk factors collected as baseline data in the Oxcheck trial in patients in British general practice.5 Framingham equation estimated the 5-year cardiovascular risk2 and those with a risk of 7.5% or above were considered to be high-risk persons. The percentage of high-risk persons missed by Marshall-Rouse method was estimated by comparing the two-step method with the Framingham equation. Among 8109 trial subjects, 2284 (28.2%) had a 5-year cardiovascular risk of 7.5% or greater. The Marshall-Rouse method missed 36.5%, 36.3% and 36.5% of high-risk persons respectively in strategies I, II and III. Many of them have a very high risk. The cost-effectiveness analysis presented by Marshall and Rouse failed to take into account these missed patients. If the sensitivity of the Marshall-Rouse method is improved to reduce the number of misses, say, by assuming every person is a smoker in step 1, the method would reduce the proportion of missed high-risk persons from 36.5% to below 4.0% in all three compared strategies. However, this would mean to assess 2179 more, or a total of 3821 (47% of the 8019 subjects), persons in step 2. Consequently, the advantage in cost-effectiveness of the Marshal-Rouse method over the Joint British recommendations may still remain but would be much smaller than that presented in Marshall and Rouse's paper. (Conflict of interest: none.) References 1. British Cardiac Society, British Hyperlipidaemia Association, British Hypertension Society, British Diabetic Association. Joint British recommendations on prevention of coronary heart disease in clinical practice: summary. BMJ 2000; 320: 705-8. 2. Anderson K, Odell PM, Wilson PWF, Kannel WB.. Cardiovascular disease risk profiles. Am Heart J 1990;121:293-8. 3. Tang JL, Dickinson JA, Liu JLY. Coronary risk assessment methods and cholesterol lowering. Lancet 1999; 353: 1095-96. 4. Marshall T, Rouse A. Resource implications and health benefits of primary prevention strategies for cardiovascular diseases in people aged 30-74: mathematical modelling study. BMJ 2002; 325: 197-202. 5. OXCHECK group. Prevalence of risk factors for heart disease in OXCHECK trial: implications for screening in primary care. Imperial Cancer Research Fund OXCHECK Study Group. BMJ 1991; 302: 1057-60. Jin Ling TANG MD PhD Associate Professor in Epidemiology and Community Medicine Yu JIANG MD MPhil Resaerch Associate James XJ ZHANG PhD Assistant Professor in Health Economy Jean WOO MD Professor of Medicine Department of Community and Family Medicine School of Public Health, Faculty of Medicine The Chinese University of Hong Kong Hong Kong SAR The People's Republic of China |
|||