PATIENTS AND METHODS

Data were collected from inception of the programme on 1 September 1999 to 31 November 2000. Patients were selected at the discretion of their own general practitioners. Information on age, sex, systolic blood pressure, and the presence or absence of smoking, diabetes mellitus, and LVH were collected from a self adhesive sticker fixed on the pathology request form. Serum total and HDL cholesterol were measured by routine automated kit methods (DiaSys Diagnostic Systems GmbH & Co, Holzheim, Germany and Bio-Stat Diagnostic Systems, Stockport, UK respectively). These data were used to generate each patient's CHD risk using a computer program written by MRH based on the full Framingham equation for the laboratory computer system (LMX, Bayer Diagnostics Ltd, Newbury, UK). The results were reported as a percentage 10 year CHD risk together with a comment on modifying risk if appropriate (for example, stopping smoking) and treatment with aspirin, antihypertensives, and statins based on previous recommendations.^{1, 8}

Two doctors were provided with each patient's relevant clinical details, serum cholesterol, HDL cholesterol, and total: HDL cholesterol ratio to predict CHD risk using the new Sheffield table and the modified JBS risk chart. They then recorded the 10 year CHD risk as either <15%, ≥15 to <30%, or ≥30%. Charts were evaluated at 15% and 30% risk as these values influence the use of antihypertensives and aspirin (15% risk) and antihyperlipidaemics (30% risk).^{1, 8}

Diagnostic sensitivity, specificity, and predictive values for the modified JBS chart and the new Sheffield table were calculated using the precise individualised laboratory based risk score calculated from the full Framingham equations as the reference method. The per cent confidence interval for each value was calculated using the exact Poisson method. For statistical analysis CHD risk scores of ≤14.9 were assigned the value of 1, those between 15% and 29.9% a value of 2, and those ≥30% a value of 3. Repeat measures analysis of variance with Tukey-Kramer post-test comparison was then used to assess the differences between the CHD risk scores from the three different CHD risk prediction methods.

RESULTS

During the study period, 1246 requests were received. The following requests were excluded: 51 due to inadequate clinical data, 10 because serum samples were erroneously discarded before lipid measurement, 10 because they were repeat requests, 67 on subjects aged >70 years because the new Sheffield table uses this as its upper age limit, and seven with LVH because LVH is not accommodated for in the JBS chart. Therefore, data from 1102 requests were evaluated (table 1).

The new Sheffield table CHD risk scores were higher (p<0.001) in comparison with the calculated CHD risk and modified JBS chart. The JBS chart risk scores were lower (p<0.05) than calculated CHD risk scores. The new Sheffield table and modified JBS chart performed equally well; both had high diagnostic sensitivity and specificity at 15% and 30% risk (table 2). The new Sheffield table identified all 38 subjects, with a 10 year CHD risk of ≥30% and serum total cholesterol ≥5.0 mmol/l, requiring statin treatment; whereas the JBS chart failed to identify three of these subjects. The new Sheffield table misclassified 17 patients, with a 10 year CHD risk <30% as requiring statin treatment (10 year CHD risk of 26.7%–29.8%); whereas the modified JBS chart did not misclassify any patients.

In comparison with the other methods of risk assessment, using the new Sheffield table as a screening tool to assess which subjects required measurement of serum lipids would have led to a 13.4% reduction in requests for serum lipid measurements (16.4% in women and 10.1% in men) without loss of diagnostic efficacy.

DISCUSSION

This study, the largest to date, confirms the high diagnostic sensitivity and specificity of the modified JBS chart and new Sheffield table^{5, 7} and therefore their suitability for use in primary care. The high and equable diagnostic performance of both charts suggests that the approximations regarding age and blood pressure do not significantly affect their accuracy. The results reported in this study are consistent with reports suggesting that tables and charts which incorporate individual HDL cholesterol concentrations have overcome the inaccuracies of tables and charts which used a population average HDL cholesterol value.^9–11

The National Service Framework recommends use of the JBS chart but states that new Sheffield table, European coronary risk chart, and New Zealand cardiovascular risk prediction charts are acceptable for assessing CHD risk.¹ We chose to evaluate the new Sheffield table and modified JBS chart as they are widely available, commonly used, and compatible with National Service Framework guidelines. The European coronary risk chart uses average HDL cholesterol values and is therefore inherently inaccurate⁷ and should not be used. The New Zealand cardiovascular risk prediction charts calculate cardiovascular risk rather than CHD risk and are therefore less compatible with National Service Framework and UK guidelines which offer recommendations for management of CHD risk.^{1, 6}

It is desirable that CHD risk tables and charts should have high sensitivity (no false negatives) to identify the maximum of high risk patients who would benefit most from drug treatment. It is also important that they have high specificity (no false positives) so that they do not falsely label individuals as being high risk and requiring drug treatment. The new Sheffield table was more sensitive (fewer false negatives) but less specific (more false positives) at 30% risk, and marginally less sensitive and specific at 15% risk in comparison with the modified JBS chart. Our results are similar but not identical to a recent smaller study reporting that the modified JBS chart should be method of choice for predicting CHD risk from tables.⁷ We, however, conclude that since the tables perform comparatively well the choice of chart lies in ease of use and their potential impact on healthcare costs.

We found the compact tabular layout of the Sheffield table easier to use but others have reported a preference for the graphic displays of the modified JBS chart.¹² An advantage of the graphic display of modified JBS chart is that it provides graded risk denoted as three differently coloured zones (<15%, ≥15% to <30%, and ≥30%), unlike the new Sheffield table that provides cut off values for risk at 15% and 30%. Therefore the modified JBS chart (unlike the new Sheffield table) can, for example, indicate the difference between 16% and 29% risk and this may have patient management implications.

In this study, the lower specificity at 30% risk of the new Sheffield tables in comparison with the modified JBS chart, necessarily increases “false positives” and therefore total prescribing statin costs. This, however, has to be offset against potential cardiovascular morbidity costs of the modified JBS chart in failing to identify and treat very risk high subjects. A not widely appreciated advantage of the new Sheffield table over all other assessment methods is its ability to screen which subjects do not require a blood test for measurement of total and HDL cholesterol concentrations.⁵ This could result in potential savings in healthcare resources by reducing laboratory costs, clinical costs (fewer revisits), and be seen as an improvement in patient care by reducing the number of unnecessary venepunctures and surgery visits. Even in this study targeted to subjects with potentially increased CHD risk, the new Sheffield table would have led to a 13.4% fewer lipid measurements in comparison with the modified JBS chart and computer program without loss of diagnostic efficacy.

An alternative to tables and charts is an individualised CHD risk score calculated on personal computers which give a more precise prediction of CHD risk.^{1, 6, 13} In offering more patient interaction by illustrating the benefits of modifying risk (for example, stopping smoking), computer programs may be preferable to tables in managing cardiac risk.¹³ A potential disadvantage of surgery based computer calculated CHD risk is that the practitioner is required to collate and enter clinical and biochemical data. An alternative is a laboratory based CHD risk calculation,^{7, 11, 14} which has the added advantages of easily disseminating and implementing locally and nationally revised guidelines. Although computer programs give a more precise prediction of risk than charts, other factors including familiarity, speed, and ease of use during consultation are also important and paper based charts may still have a useful role in assessing CHD risk.

Prediction of individualised CHD risk without resorting to a risk assessment tool is often erroneous with medical practitioners either under-estimating or over-estimating risk,^{15, 16} and this may lead to erroneous patient management decisions.¹⁷ Implementing systematic CHD risk assessment in primary care can be facilitated by cardiovascular disease registers to identify high risk individuals. The likely increase in workload may be a cause for concern,¹⁸ but CHD risk assessment is a requirement of the National Service Framework.¹

In conclusion, the new Sheffield table and modified JBS chart perform equally well in predicting CHD risk for primary care patients. Since assessing an individual's CHD risk is not intuitive we suggest that health practitioners should choose a risk assessment tool (the new Sheffield table or modified JBS chart or computer program) with which they are comfortable and then use it.