Combining Population Health and Baseline Risk Strategy by Determining an Age Cutoff for Initiating Statins in Patients With Diabetes: A population-based study

This was a cross-sectional cohort study using The Health Improvement Network (THIN) dataset, which contains anonymous patient data from 304 general practices throughout England and Wales (12). Information obtained from the dataset included patient demographics, medical history, laboratory results, prescription information, and lifestyle characteristics. THIN has been validated at both practice and dataset levels (13) by comparing its demographics, morbidity, mortality, prevalence, and geographical rates with various national data sources, including Department of Health–issued read codes for the Quality and Outcomes Framework 2001 Census and the National Statistics and Office for National Statistics (www.statistics.gov.uk).

We identified 60,258 patients with diabetes and reviewed their biochemical and demographic profiles available on 31 December 2005. Patients had to have been registered by their practices for the entire preceding 12 months to be included in the analysis. A total of 11,005 patients with diabetes aged between 30 and 74 years who were not prescribed lipid-lowering drug therapy and who were without arterial disease (no history of ischemic heart disease, cerebrovascular disease, and peripheral vascular disease) recorded in general practice databases were suitable for analysis. This large patient cohort reflects the clinical and biochemical parameters of patients before full implementation of the Joint British Societies’, NICE, and GMS contract guidelines. We utilized the above age criteria because decisions to initiate statins beyond this criteria should be based on individual risk basis rather than from a public health perspective. National estimates were calculated by the multiplication factor of 32 derived from the ratio of the number of people with diabetes in England and Wales (1,922,051 individuals) within the total cohort obtained from the THIN dataset (60,258). The study was approved by the Eastern Multi Centre Research Ethics Committee.

We used the Joint British Societies’ risk calculator derived from the Framingham risk algorithm, which utilizes eight risk factors (age, sex, systolic or diastolic blood pressure, smoking status, diabetes status, left ventricular hypertrophy, and total and HDL cholesterol) to calculate a CVD risk. We used a Framingham-based risk engine (14) rather than the UK Prospective Diabetes Study (UKPDS) (15) because our dataset did not include duration of diabetes and microalbuminuria status. Although Framingham risk calculation is thought to underestimate mean CVD risks of patients with diabetes (16), previous work has shown that such underestimation is only relevant when considering patients whose 10-year CHD risk is >20% (17). Thus, when determining CHD risk of <20% in the context of the threshold for prescribing statins, the difference between the two methods of risk assessment is negligible.

We considered four strategies to reduce CVD risks in patients with diabetes. For each strategy, we determined the number of patients with diabetes eligible for statin treatment (and the cost incurred). We calculated the number of CVD events potentially avoided (effectiveness) using the product of patients’ 5-year baseline risk and an estimate of 33% relative reduction of CVD events (based on the effect of 40 mg simvastatin in preventing CVD event over 5 years with a primary prevention cohort in the HPS trial) (3). The number needed to treat (efficiency) to prevent one CVD event over 5 years was estimated as the sum of number of treated divided by the number of CVD events prevented. We calculated population-based estimates for England and Wales (53,390,300 individuals) with an estimated diabetes prevalence of 3.6% (11). We chose 40 mg simvastatin instead of 10 mg atorvastatin because for every new patient treated with 40 mg generic simvastatin rather than 10 or 20 mg atorvastatin, the National Health Service (NHS) saves £921–1,352 over 5 years without compromising lipid-lowering efficacy (18). The cost of 40 mg simvastatin (based on the NHS reimbursement price) is £44.20 per patient per year.

In the first part of the analysis, we examined the relationship between age and baseline 10-year CVD risk according to age. We used a regression technique to plot the relationship between age and baseline CVD risk using a linear, exponential, or quadratic equation. We used the line of best fit between these two variables to establish the mean age at which men and women with diabetes moved from low risk (<10%) to moderate/high 10-year CVD risk (>10%). These thresholds were chosen on the basis of corresponding 10-year CVD risk estimates used by various clinical practice guidelines using the Framingham risk algorithm (19). In addition, a 10% 10-year CVD risk has been advocated by the U.S. Preventive Services Task Force as the threshold level above which the cardio-protective benefit of aspirin therapy will outweigh its risk of bleeding (20).

All analysis was performed using SPSS for Windows (version 14) and SAS (version 8.2). Normally distributed data were presented as means ± SD, skewed data as median (range), and categorical data as percentages. Student's t test, χ² test, and regression analysis were also used. Where there was clustering (patients in practices), the random-effects logistic regression technique was used. Multivariate regression models were used to assess the predictive power of variables in determining the need to initiate statins (based on a 10-year CHD risk >10%). Sensitivity and specificity values for age criteria to initiate statins were determined using the receiver operating characteristic plot.

The study population consisted of 60,258 patients with diabetes aged between 30 and 75 years. From this group, we identified 11,005 patients with complete datasets who were not taking any lipid-lowering agent, were free from any history of atherosclerotic arterial disease, and therefore were eligible for primary CVD prevention. Table 1 shows the baseline characteristics of patients. Mean age was 53.7 years (55.9 years for men and 53.4 years for women). The mean 10-year CVD risk for the total population was 20.7% (i.e., 21.3 and 17.8% for men and women, respectively).

For both men and women, baseline CVD risk increases with age (Fig. 1). We used the quadratic equation (best fit for our data), y = 13.99 + 0.388 (age) + 0.005 (age²) for men and y = 13.99 + 0.388 (age) + 0.005 (age²) for women. The transition from low to moderate/high baseline risk of developing CVD occurred at about age 40.6 years for men and 44.2 for women. If a high baseline risk is used as a threshold, the age transition from a low/moderate to high baseline risk of developing CVD took place at ages 52.4 years for men and 59.3 years for women (Table 2). We set the optimal cutoff for prescribing statin at a 10% 10-year CVD risk for both sexes. Using age 40.6 years as the cutoff for prescribing a statin in men gives a sensitivity and specificity of 92.2 and 84.4%, respectively. Using age 44.2 years for women gives a sensitivity and specificity of 90.3 and 81.3%, respectively.

The total numbers of patients eligible for statin therapy based on different primary CVD strategies extrapolated to the national estimate are 352,160 (treating all), 172,736 (treating high baseline risk), 264,608 (treating moderate/high baseline risk), 127,456 (treating cholesterol >5 mmol/l), and 278,800 (treating patients above the new age cutoff criteria). In financial terms, based on the cost of 40 mg/day simvastatin, these figures amount to an annual statin expenditure of £15,565,472, £7,634,931, £11,695,670, £5,633,555, and £12,322,960, respectively. The population health, baseline moderate/high risk, and the age cutoff strategies were the most effective, potentially avoiding the most cardiovascular events over 5 years (12,050, 11,214, and 11,094 events, respectively). Strategies to treat if cholesterol is >5 mmol/l based on the GMS contract or if baseline risk is high based on the previous NICE guideline criteria were the least effective. Using the moderate/high baseline risk and the age cutoff strategies are the most efficient and cost-effective, with the lowest number needed to treat, 23.6 and 25.1, while recommending statin treatment to a relatively low number of patients. Thus, utilizing the age cutoff strategy instead of the population health strategy will confer a cost saving of £16,212,560 over 5 years to the NHS, with a lower number needed to treat and without compromising effectiveness (Table 3).

Our findings highlight the scientific rationale for combining population and baseline risk strategy for preventing CVD in patients with diabetes without overt atherosclerotic disease. We showed that statin prescribing guidelines for primary CVD prevention varied in their effectiveness (potential to prevent CVD events) and efficiency (number needed to treat to prevent one CVD event). A desirable strategy is one that would potentially prevent the largest number of CVD events and would recommend treatment to the least number of patients. Age appears to be the most important determinant of an individual's baseline CVD risk. Baseline CVD risk in people with diabetes reached a threshold for moderate/high risk at ∼40 years for men and 45 years for women. This age cutoff confers high sensitivity and specificity for individuals with diabetes to have a moderate baseline risk of developing CVD. Because the number of patients with diabetes at risk of CVD are large and statins are highly effective in reducing CVD events, even small changes in guidelines have large consequences on the number of patients eligible for statin use, the potential for preventing CVD events, and the millions of pounds spent each year by the NHS.

Current U.K. guidelines for prescribing statins using age cutoff strategies are largely based on the minimum age of patients recruited into the HPS (3) and CARDS (4). This approach, however, does not take into account patients’ sex, their baseline CVD risk, or the effectiveness of statins in patients with low CVD risk in both intervention and placebo arms of these studies. Our findings concurred with findings from Booth et al. (21), which showed that the transition to a high CVD risk category occurred at the ages of 41.3 and 47.7 years for men and women with diabetes, respectively. Some, however, recommend the routine use of statins in all patients with diabetes irrespective of age (22). Whereas this concept of “coronary risk equivalent” (i.e., all patients with diabetes without atherosclerotic disease should be considered as already having a myocardial infarction or are at “high risk” defined as CVD risk >20% over 10 years) has gained significant momentum following the study by Haffner et al. (23), which utilizes a patient cohort from Finland; subsequent studies from the U.K., U.S. (24–25), and Canada (21) did not support this observation. Our data suggest that although the mean 10-year CVD risk of patients with diabetes in England and Wales was indeed ∼20%, baseline CVD risks varied considerably between individuals with diabetes, based on patients’ age and sex. Hence, the absolute risk reduction potentially conferred by the use of statin in these patients will also likely vary. Previous guidelines have also taken into account duration of diabetes—a surrogate for prior glycemic exposure—as an indication for statin use (26). Whereas our study was not designed to investigate the validity of this strategy, previous observational studies of CVD prediction in young people with diabetes showed that duration of diabetes did not appear to be an independent predictor (27). Given the various uncertainties in predicting CVD risk for younger people with diabetes, we would recommend that primary CVD prevention strategy be individualized in this group of patients.

When comparing how well different treatment strategies perform, a compromise needs to be achieved between effectiveness and efficiency. Thus, although a strategy to treat all patients with diabetes was shown to be highly effective, it would involve treating the largest number of patients to prevent one CVD event. Conversely, whereas a strategy to treat patients who have a 10-year CVD risk >20% is the most efficient (lowest number needed to treat), this strategy was not very effective (potentially preventing a relatively small number of CVD events). The GMS contract criteria was the least effective, whereas the two most effective and efficient strategies are those that advocate statin therapy for patients with diabetes whose 10-year CVD risk is >10% or for those whose age is above the previously mentioned cutoff values. From a public health approach, the latter strategy is perhaps the more attractive, given the practicalities of implementing such a simple strategy.

Several limitations of our study must be acknowledged. The presence of diabetes was dependent on patients visiting the general practitioner and would therefore not identify people with undiagnosed diabetes. The dataset did not allow us to determine diabetes duration or microalbuminuria status or distinguish between type 1 and type 2 diabetes. This, however, should not affect the outcome of this study because primary CVD prevention strategies in the U.K. do no distinguish between types of diabetes or take diabetes duration into account, whereas the presence of microalbuminuria is an indication for statins on the basis of secondary prevention. Another limitation is our use of a Framingham-based risk algorithm to estimate CVD risk in patients with diabetes. We believe, however, that our use of the Framingham algorithm is valid for two reasons. First, it has been validated in the U.K. population (28,29); and second, it has been shown to provide a CVD risk estimate that is equivalent to that of the UKPDS engine when assessing risk <20% over 10 years (17,29).

When implementing a policy for statin treatment in patients with diabetes, it is necessary to utilize a strategy that will give a high pick-up rate but will also identify patients who will benefit most. From this study, we advocate that in the absence of specific indication for statin therapy (e.g., microalbuminuria, strong family or personal history of CVD risk, etc.), statins should still be routinely prescribed to all men and women with diabetes aged >40 and 45 years, respectively, for primary CVD prevention. This strategy is highly effective and efficient to prevent CVD events from a public health perspective. Further studies, however, are required to clarify whether this treatment strategy can be extrapolated to diabetic patients outside the U.K., and longitudinal data are required to confirm the absolute risk reduction estimated using this strategy.