OUP user menu

The new Sheffield risk and benefit tables for the elderly

N.D. Weatherley , P.R. Jackson
DOI: http://dx.doi.org/10.1093/qjmed/hcq167 3-12 First published online: 17 September 2010


Several charts or tables are used to guide treatment in primary prevention of cardiovascular disease (CVD). These usually relate to patients up to 75 years of age, leaving older patients without guidance. Most also present this information as risk, leaving patients to estimate the benefit of treatment and decide whether it is worthwhile. We present tables to display both CVD risk and benefit from treatment in the elderly. A systematic review identified CVD risk functions for the elderly. The Dubbo study of older patients’ 5-year CVD risk equation was deemed most appropriate, due to the population studied, endpoints observed and risk factors recorded. By dichotomizing most risk factors, we produced a new risk table in the form of the original ‘Sheffield table’. Risk is calculated by selecting the appropriate table for gender and the appropriate cell from the rows and columns, representing age and risk factor contributors, respectively. Total cholesterol above a cell value corresponds to a 20 or 40% 10-year CVD risk. A simple risk scoring system was then derived from the Dubbo equation. Calculation of risk score requires knowledge of a patient's simple demographics, systolic blood pressure and total and high-density lipoprotein cholesterol. Positive integers corresponding to level of risk for each contributing factor are then added together to give a final risk score. A Markov chain model was produced based on the Dubbo derived risk and relative risk reductions from published meta-analyses of 3-hydroxy-3-methyl-glutaryl-CoA reductase inhibitors (statins) and anti-hypertensive treatment. Using this model, individual scores were mapped to likely benefit from treatment in terms of disease free years. Our risk table provides a simple means for calculating risk in the elderly, to two major thresholds, while the benefit table explores the concept of presenting benefit of taking CVD-preventing medication.


The Framingham equation publicized in 1991 investigated those risk factors associated with the development of cardiovascular disease (CVD).1 This has been used to predict the risk of developing coronary heart disease (CHD) or CVD over the next 10 years for individuals in middle age who were initially disease free, according to a number of demographic factors, medical history and simple clinical measurements. Several charts or tables based on the Framingham equation have been created to target treatment in the primary prevention of CHD and CVD to ensure that those at the highest risk were offered drug therapy.

One of the first of these tools to be produced and implemented in the UK was the Sheffield Table.2 There followed a number of other risk scoring tools, including the European-based SCORE (Systematic coronary risk evaluation) charts and the Joint British Societies charts.3,4 One major limiting factor of all these charts is their inability to predict risk in an elderly population (i.e. over the age of 75 years), a consequence of the age range of patients studied in the original Framingham series. This shortcoming leaves older patients and their medical practitioners without the type of guidance available to younger patients. Following the publication of the Dubbo study results in 2003,5 we present a new table that aims to fill this void.

However, without explicit knowledge of their likely benefit, patients have to invoke simple heuristics rather than rational analysis to decide whether taking drug treatment to prevent CVD is worthwhile. Predicting an individual patient's risk is only one step in illustrating to them their likely chance of benefiting from treatment. It is then necessary to approximate the potential benefit from treatment for a given patient at a given level of risk and the effects of competing disease risks cannot be accommodated. It would be better if this estimate of individual benefit could be provided to people directly based on their individual characteristics. We have derived a table to illustrate not the risk, but the likely average benefit from taking CVD preventative treatment for a person aged >60 years, according to their personal characteristics. We believe that this represents a novel method of illustrating the potential benefits of treatment from the patient's perspective, allowing a better informed and patient-centred decision to be made by both clinician and patient.


A systematic review was performed to identify prospective epidemiological studies in the elderly over the past two decades, from which CVD risk functions could be derived. The search was defined by three main characteristics. First, the population had to be elderly (>50% over the age of 65 years). In an attempt to identify risk factors, we linked high-sensitivity filters for aetiology and prognosis. Finally, the endpoints of the studies required were specified as the CHD or CVD endpoints of angina pectoris, myocardial infarction or stroke. From 4898 papers identified, 80 were chosen for further analysis.

These were reviewed independently by two reviewers. A paper surviving into the second stage of analysis needed both to satisfy the original criteria and to provide suitable evidence from which to derive a risk function through a risk equation. Their measured or ‘adjusted for’ risk factors had to include age, gender, total cholesterol, systolic blood pressure and a diagnosis of diabetes mellitus.

The Dubbo study was chosen as the most suitable source to provide a risk function in the elderly (Figure 1).5 This is a longitudinal cohort study of elderly Australians living in the semi-urban town of Dubbo in New South Wales. All non-institutionalized residents born before the year 1930 were eligible on entry into the study in 1988. Baseline tests performed included serum total and high-density lipoprotein (HDL) cholesterol, serum triglycerides and plasma glucose concentrations, in addition to blood pressure. Average follow-up was for 13 years and endpoints included CVD, CHD and stroke.

Figure 1.

The Dubbo Study equation for calculating cardiovascular risk in the elderly. Ten- and five-year risk calculations included, where age, systolic blood pressure, total and HDL cholesterol levels are continuous variables.

As with the original Sheffield Tables, columns in the new tables are split into two in order to represent two thresholds of risk.2 In the first Sheffield Tables, these indicate 15 and 30% levels of 10-year CHD risk, respectively. It was decided to use CVD rather than CHD endpoints for the new tables. This includes stroke, an endpoint with greater incidence and potentially a more devastating complication in the elderly. Ten year CHD risks of 15 and 30% are approximately equivalent to 20 and 40% 10-year risk of CVD, respectively.6 These values of 20 and 40% risk of CVD were maintained to replicate the levels of risk published in the original tables and to enable a range of risks to be communicated to the patient in order to assist them when making the decision of whether to initiate treatment.

In calculating the table values from the formula, important assumptions had to be made about risk variables. Below are definitions of the risk factors included, most being taken directly from those of the Dubbo Study.

Hypertension: A systolic blood pressure of ≥140 mm Hg is defined as hypertension (HT). On anti-hypertensive medication was found to be a significant risk factor in itself, which corresponds to previous findings.7

Diabetes (mellitus): A fasting plasma glucose level of ≥7.0 mmol/l, a self-reported history of diabetes mellitus, or currently taking diabetes therapy.

Smoking: Current or recent tobacco smokers of any amount are classified as smokers. Ex-smokers of >1 year are defined as non-smokers.

The table has two uses; to decide whether to start anti-hypertensive treatment and whether to start a 3-hydroxy-3-methyl-glutaryl-CoA reductase inhibitor (statin). Thus the risk for hypertensive patients on treatment (statin decision) and off treatment (for anti-hypertensive treatment decisions) are needed.

Both the Systolic Hypertension in the Elderly Program (SHEP) and hypertension in the very elderly trial (HYVET) trials provide data showing the benefit of lowering blood pressure to target in the elderly, although the trials had different blood pressure targets. Systolic blood pressure targets in these two studies were <160 mm Hg and <150 mm Hg for SHEP and HYVET, respectively.8,9 As the primary care, Quality and Outcomes Framework (QOF) target level has been set at a systolic pressure of <150 mm Hg (Hypertension management in primary care, 2005), the HYVET trial target and, therefore, its results are more appropriate in assessing efficacy of antihypertensive therapy in a UK elderly population.10

In this trial, systolic blood pressure for both treatment and placebo arms were identical. This was lowered to an average of 143.5 mm Hg after 2 years in the treatment group and thus, this value plus ‘on antihypertensive medication’ was used in calculating the thresholds under the hypertension = yes; treated = yes columns. Systolic pressure in the placebo group in this trial was 158.5 mm Hg after 2 years and this value is therefore attributed to untreated hypertensive elderly patients. For normotensive people, a UK average elderly systolic pressure of 134 mm Hg was used.11

While HDL cholesterol appears to be a significant predictor of CVD in the elderly, the Dubbo formula could not be manipulated to express risk in terms of total: HDL cholesterol ratio. Thus, average values of HDL cholesterol for the elderly male and female population (1.24 and 1.57 mmol/l, respectively) had to be used in constructing the new tables.12

Realistic population values of total cholesterol were derived from The 2003 Health Survey for England (HSE) online data.13 Minimum and maximum values were initially estimated as three standard deviations below and above the mean, respectively. These ranges are 2.2–8.4 mmol/l for men and 2.4–9.8 mmol/l for women. However, the heart protection study only provides evidence of benefit for people with a pre-treatment cholesterol value above 3.5 mmol/l and this has been taken as a more appropriate minimum value for total cholesterol in the new tables.14 Thus, the total cholesterol ranges considered in the table for men and women are 3.5–8.4 and 3.5–9.8 mmol/l, respectively.

Indirect external validation of the equation derived from the Dubbo study was undertaken. Data for the persons in the 60- to 75-year age range (who are covered by both Framingham and Dubbo equation data) in the HSE data for whom complete risk factor data is available were identified. For each of these individuals, a predicted CVD risk was calculated using both the Framingham and Dubbo risk functions. The resulting estimates were then investigated for correlation and agreement using scatter and Bland–Altman plots.

It was decided that for simplicity of use, risk should be simplified into a numeric scale. This enables the impact of multiple risk factors to be captured in a single integer score. Points for systolic blood pressure and total cholesterol increased as the values of these variables increased, while points decreased with increasing concentrations of HDL cholesterol. The Dubbo 5-year risk of CVD equation was used to calibrate limits for the scoring system. Minimum and maximum limits for blood pressure and cholesterol were then chosen using population values for the ‘≥75’ age group in the HSE document.

The risk of death from non-CVD causes for individuals in the model also needs to be accounted for. The Office for National Statistics database provides figures for deaths in England and Wales in 2004.15 The figures are broken down into age, sex and cause of death. Cardiovascular causes of death as described by Dubbo and as classified by their International classification of diseases (ICD)-10 codes (I20–I25, I63 and I64, respectively) were subtracted from this total, as they would already be accounted for in the CVD risk in the Dubbo equation. Thus, non-cardiovascular deaths could be accounted for in both arms of the model by producing curves for each gender, comparing non-CVD death rate to age.

Three meta-analyses of longitudinal trials in these fields were used to derive the average benefit of taking a statin or anti-hypertensive therapy.16–18 The combined reduction in risk of developing new CVD with statins is between 30.3% and 32%, while reduction of risk with anti-hypertensive therapy is 28.3%. For simplicity, we adopted an overall reduction in risk of developing CVD of 30% with either form of drug treatment used alone.

For each combination of risk factors, a simple Markov chain model was run, following a cohort of 1000 virtual individuals. Through repeated annual cycles, their likelihood of developing CVD in that epoch was estimated from the Dubbo 5-year risk estimate. At the end of each year or cycle, this proportion of the previous year's cohort was considered to have developed CVD. In addition, an age-related proportion was considered to die from non-CVD causes. Each model was then re-run, but with the proportion of those developing CVD reduced by 30% to account for the proportion avoiding CVD by either lipid lowering or anti-hypertensive medication. Cycles were continued until all patients developed CVD or died of another cause.

This model makes it possible to predict the potential gain from taking anti-hypertensive medication or a statin. The average increase in healthy life expectancy was derived from the cumulative sum of the number of people in the healthy state at the end of each year, divided by the initial population of 1000. The likely benefit from treatment is affected by gender, age at the start of the model and the level of risk, as determined by other standard risk factors. Two tables, one for each gender, were then generated by repeated uses of the Markov model for each combination of age and risk factor score.


For eligible persons from the HSE dataset, the average population 10-year risk of developing CVD was calculated as 20.4% and 26.8% by Dubbo and Framingham equations, respectively. The simple Pearson's correlation coefficient for the association between predicted risk from the Dubbo and Framingham equations is 0.847, showing a strong level of correlation. However, the mean difference in 10-year CVD risk is 7%, with the Dubbo equation predicting risks lower than those estimated by the Framingham equation. The negative gradient on the Bland–Altman plot (Figure 2) shows not only that the greater an individual's risk, the greater the difference in the risk between that calculated by Dubbo and Framingham, but also that the mean Dubbo risk is never greater than that of Framingham risk.

Figure 2.

Bland Altman plot for agreement between 10-year Dubbo and Framingham risks.

To use the risk table (Figure 3), select the row according to the age of the patient at their last birthday from the left hand column. For patients between those ages shown, choose the higher age (e.g. for a patient who was aged 75 years at last birthday, read from the row age 76). Then select the appropriate column according to the patient's combination of risk factors (hypertension, smoking and diabetes); differentiating between hypertensive patients on treatment (HT treated = yes) and not receiving treatment (HT treated = no). The intersection between the selected row and column indicates either a cholesterol concentration (mmol/l) above which the patient exceeds the risk threshold or is blank, in which case the patient is unlikely to exceed the cholesterol value required to reach that risk threshold. In the latter case, cholesterol testing is not mandatory. The two thresholds indicate a 20% (yellow) and 40% (red) risk of developing CVD in the subsequent 10 years. It is important to note that the table is intended to be used for primary prevention, and not to estimate risk in those with established CVD.

Figure 3.

The New Sheffield Tables for the Elderly.

Worked examples for risk scoring chart (Figure 4):

  • Example 1: a 62-year-old female, non-hypertensive, non-diabetic, smoker, with a systolic pressure of 125 mm Hg, total cholesterol of 3.9 mmol/l and HDL of 2.0 mmol/l would accrue a score of 0 + 0 + 5 + 2 + 1 + 1 = 9.

  • Example 2: a 76-year-old male, who is on anti-hypertensive medication and diabetic, but non-smoking, with a systolic pressure of 160 mm Hg, total cholesterol of 5.3 mmol/l and HDL of 1.3 mmol/l would accrue a score of 9 + 7 + 0 + 5 + 3 + 3 = 27.

For our benefit table (Figure 5), the cell at the intersection of the row representing a patient's age and the column associated with their risk indicates the value of starting treatment now in terms of the predicted average amount of healthy life gained from medication (in months). Each cell was calculated by subtracting mean life expectancy in the baseline model from that in the treatment model, incorporating age at the start of treatment and risk factors. The tables are colour coded, simply to emphasize the gradation in benefit; red represents an additional healthy life time gained of <6 months, yellow represents 6–12 months and green ≥12 months benefit (Figure 5).

  • Example 1: 62 year old woman with risk score = 9 would gain an average of 5.8 months of CVD-free life by taking treatment.

  • Example 2: 76 year old man with risk score = 27 would gain an average of 12.2 months of CVD-free life by taking treatment.

Figure 4.

Simple Risk Scoring Chart. To estimate likely benefit from treatment for an individual, total and HDL cholesterol concentrations and blood pressure need to be measured and the patient's risk score calculated according to these results and information from the patient's history. Worked examples are provided.

Figure 5.

The Sheffield Benefit Table for the Elderly. To use the table, select the row that relates to the age of the patient. If the patient is between the ages shown, choose the higher age (e.g. for a patient who was aged 75 years at last birthday, read from the row age 76). The column displaying the value closest to their risk score should then be selected. From the intersection of the appropriate row and column, likely benefit in terms of healthy life months gained can then be estimated. The earlier patient examples are continued and completed.


The elderly are, on average, at increased absolute risk of developing CVD and so have much to gain from preventative treatment. The HYVET trial showed that treating high-blood pressure in the elderly was effective in preventing CVD without excess adverse drug reactions. The prospective Pravastatin in the elderly at risk study supports the use of cholesterol therapy for those at high risk. However, when discussing drug treatment to prevent CVD with an individual patient, considering drug treatment to prevent CVD, a specific tool to illustrate individual risk would be helpful and a tool to suggest probable benefit from treatment even more useful.

Comparisons were made between the Dubbo and Framingham equations by the authors of the Dubbo study, which we chose to explore in greater depth in order to validate the New Sheffield Table. Validation had to be indirect in the absence of a prospective long-term cohort study of Western European elderly patients. This method was not ideal, due to the limited age overlap between the two risk functions Framingham (30–75 years; Dubbo ≥60 years). Nevertheless, it serves as some guide to the predictive ability of the Dubbo study in a UK population.

The results provide evidence that although Dubbo and Framingham show a strong positive correlation in their risk prediction scores, Framingham predicts a higher absolute 10-year CVD risk than Dubbo in those aged 60–75 years by 6.58% on average. The agreement between the two risk scores diminishes in patients at greater risk. However, given that on average the Framingham risk function overestimates risk in the UK population, particularly in high-risk males, this may be of limited concern.19 To gauge how CVD impacts on overall mortality in both Australian and UK populations, we compared the most recently available mortality data from the two countries using official government sources.15,20 In 2003, there were 68 330 deaths in Australian males and 63 962 deaths in Australian females. In 2004, there were 245 208 deaths in UK males and 269 042 deaths in UK females. The proportion of these deaths attributable to ischaemic heart disease or cerebrovascular disease is compared in Figure 6.

Figure 6.

CVD-related mortality as a proportion of the total number of deaths in UK and Australian populations.

Despite the recent literature suggesting ‘novel’ risk factors as predictors of vascular disease in elderly patients, such as heart rate variability, pulse pressure and C-reactive protein levels,21–23 the Dubbo equation features the ‘classical’ risk factors of smoking, hypertension, diabetes and cholesterol (total and HDL). Indeed, the only new risk factor included in the creation of the new table was that of ‘on anti-hypertensive medication’. The format we chose for our new tables was that of the earlier Sheffield table for ease of use. However, as with the original, several assumptions had to be made in constructing the table. Hypertensive status was dichotomized into average population and QOF target values. Clearly, this is less predictive than if blood pressure had been considered as a continuous variable.

One omission enforced due to the structure of the Dubbo function was that of HDL cholesterol as a risk factor. A method for incorporating a ratio of total: HDL cholesterol could not be found and to include it as a separate risk factor would have made the table unwieldy. Thus, HDL was omitted and replaced with a population average value. Clearly, this will further reduce the predictive ability of the table, although the effect is likely to be small.

Our second table has been designed to provide an indication of the likely benefit to an individual from taking one of these treatments. The results illustrate that the oldest patients gain least from preventative treatments, due to their lower life expectancy and increased risk of death from other causes. This reduction in benefit with age is marked, such that on average, no individual >88 years will gain ≥6 months of CVD-free life. Indeed, only men <78 years and women <80 years will gain ≥12 months from drug treatment.

Taking a high-risk person as an example [e.g. BP = 150 mm Hg + total cholesterol (=5.6 mmol/l) + HDL (=0.9 mmol/l) + diabetes; score 21], a 60-year-old male would have a predicted benefit of 19.0 months, a 70-year old would gain 13.9 months, an 80-year old 8.0 months and a 90-year old 3.7 months. As expected, for a given age, those at greater risk of CVD are predicted to gain more from treatment. While all elderly persons can be considered as being at a high 10-year risk of developing CVD (over 20%) if over 80 according to our risk table, they will not necessarily gain a great deal from taking medication. Even in those aged between 70 and 80 years, a risk score of 21 in men or 24 in women is required to produce an average healthy life gain of ≥12 months. Strategies suggesting a blanket approach to preventative therapy, where all patients over a given age are advised to take a statin have been suggested. However, the table shows a distinct lack of benefit, even for younger patients with a low CVD risk score.

Conversely, the table can give some guidance as to the possible effects of stopping preventative therapy beyond a certain age. The assumption that stopping treatment is the reverse of starting is strong and there is evidence that risk increases when preventative therapy is stopped.24

The major limitations to our approach are the provision of the mean gain rather than a personal gain from treatment and the omission of non-CVD morbidity from the model. Displaying a mean gain in life does not reflect how many additional months a specific individual will gain, rather the average months gained if many people of the same age and CVD risk were followed-up. In all probability, the distribution of benefit will be skewed and many will gain nothing from taking drug treatment, although a minority might gain several years. Some CVD risk factors (e.g. smoking) impact on the risk of death from other diseases and this is not accounted for in our model of the competing causes of death.

While anti-hypertensive drugs and statins are unlikely to have a major influence on non-CVD mortality, risk factors for CVD may well predict non-CVD mortality. Indeed, our model does not take into account non-CVD morbidity, which may adversely affect quality of life. Diabetes is an obvious example, as it is associated with multi-organ disease, which can affect quality of life severely. Thus, while we may prolong life by prescribing therapy, the patient's quality of life could still be poor or they may die prematurely from a cause associated with their CVD risk factor.

We present the new Sheffield Table as an easily accessible tool for calculating 10-year-CVD risk in the elderly. While most such people will be at moderate or high-CVD risk, this tool provides patients with the individual information they need to decide on whether or not to accept treatment. It should not be assumed that older people would automatically wish to be treated with anti-hypertensives or lipid-lowering therapy. Clearly, each patient will have a different threshold for the gain in healthy life they think worthwhile and a different attitude to taking preventative medication for the rest of their life. The likely positive effects of anti-hypertensive medication or statin treatment for an individual elderly patient are presented in an even more accessible form in the benefit table. Knowing the potential desirable effects of the medication should make it even easier for them to participate actively in any decision as to whether to take medication for the prevention of CVD.


We would like to acknowledge and thank the Joint Health Surveys Unit (Department of Health) for providing the Health Survey for England dataset.

Conflict of interest: None declared.


View Abstract