How Cost-Effective Is the Treatment of Dyslipidemia in Patients With Diabetes but Without Cardiovascular Disease?

The costs and benefits associated with the treatment of dyslipidemia were estimated using the Cardiovascular Disease Life Expectancy Model(4,16). With use of cost data from the U.S. and six other countries, the incremental cost effectiveness ratios included the direct medical costs associated with treatment and the cost savings associated with a reduction in CVD events. The economic perspective adopted in the present analysis is that of a third-party payer providing comprehensive coverage of all health care services.

The Cardiovascular Disease Life Expectancy Model estimates the annual probability of fatal and nonfatal CVD events(4). This Markov model can be used in primary prevention among patients free of diagnosed CVD or in secondary prevention among patients with prior coronary disease or stroke. The annual probability of fatal and nonfatal CVD events over 12 years of follow-up is based on a multivariate logistic regression model using the 15% random sample of the Lipid Research Clinic Follow-up cohort(17,18). Independent risk factors include age, sex, blood pressure, LDL cholesterol,HDL cholesterol, the presence of cigarette smoking, diabetes, and diagnosed CVD at baseline. The clinical criteria for coronary and cerebrovascular disease and the odds ratios for independent risk factors included in the final model have been reported elsewhere(4,19).

Briefly, a cohort of patients with or without CVD is entered into the model at a given age with specified levels of risk factors. Each year, subjects can die of coronary disease, cerebrovascular disease, or other causes. Survivors may have developed nonfatal coronary or cerebrovascular disease or remain disease free. Surviving subjects age 1 year and then reenter the model for the following year. This process continues until all subjects die. Mean life expectancy can be calculated by summing across the total person-years of life experienced by the cohort and dividing by the subjects at risk at entry into the model.

The model has been previously described in detail and shown to reasonably predict fatal events in nine clinical trials(4). These trials include treatments of dyslipidemia and hypertension. The model has also been validated on a published subgroup analysis of the diabetic subjects in the 4S(20).

We compared costs and benefits of simvastatin (versus no treatment) among diabetic patients without CVD (primary prevention) versus patients with CVD but no known diabetes (secondary prevention). cost effectiveness ratios represent the cost per year of life saved (YOLS) associated with simvastatin therapy where YOLS equals life expectancy with treatment minus life expectancy without treatment. Health care cost inputs into the model have previously been reported in detail and include the costs of hospitalizations, physician fees,outpatient care, emergency services, and drug prescriptions(16).

American health care costs were derived from published reports, including the following: hospital costs from the national Medicare Provider Analysis and Review data and the national sample of the Healthcare Cost and Utilization Project, laboratory tests and physician fees based on the Medicare Resource-Based Relative Value Scale, and annual medication costs from the 1998 Redbook(21,22,23,24).

The Canadian costs have previously been described in detail and are based on data from provincial fee schedules, the Canadian Institute for Health Information, and IMS Canada(16). Public prices for simvastatin and hospital costs for the five European countries were based on independent local costing studies conducted in each country and were provided by Merck. Other treatment costs were derived using Canadian costs adjusted for differences in hospital costs between Canada and the five European countries.

The distribution of simvastatin dose was taken from the final dosages of subjects in the 4S: 61.6% were given 20 mg simvastatin daily, 31.6% were given 40 mg daily, 0.1% were given 10 mg daily, and 6.7% discontinued the medication. The lipid modification achieved with simvastatin included a 25 and 35% reduction in total cholesterol and LDL cholesterol, respectively, and an 8% increase in HDL cholesterol. For the purposes of this analysis, we assumed that all patients had lipid profiles similar to the mean values observed in the simvastatin group of the 4S, including total cholesterol levels of 261 mg/dl (6.74 mmol/l), LDL cholesterol of 188 mg/dl (4.87 mmol/l), HDL cholesterol of 46 mg/dl (1.18 mmol/l), and triglycerides of 132 mg/dl (1.49 mmol/l)(19,20). Blood pressure was assumed to be the mean of subjects in the study or 138.5/83.2 mmHg(19,20). For the purposes of these analyses, we assumed that all individuals were nonsmokers. Risk factor levels and the impact of treatment on lipid levels were assumed to be uniform across age-groups.

We also estimated the annual outpatient costs of treating diabetes. On the basis of a detailed literature review, we assumed that a diabetic patient with no prior CVD would have two physician visits, two routine biochemical panels,and two HbA_1c tests per year(25,26,27). Each patient would also have a lipid profile, a urinalysis, and an annual consultation with an ophthalmologist. For those patients who subsequently developed CVD, the marginal costs of managing diabetes included only the additional costs of the ophthalmologist visit, the HbA_1c test, and the urinalysis. The annual costs of drug therapy for diabetes were based on the treatment identified in the 4S and health resource utilization data from an American health maintenance organization(1,28). We also estimated the costs of home glucose monitoring and assumed that each patient would have a half-hour consultation with a dietitian annually(28).

All costs data were expressed in 1998 local currencies and translated into U.S. dollars using 1998 annual exchange rates. Since the costs and health outcomes associated with disease prevention occurred at different times in the future, we discounted both by 3% annually. Sensitivity analyses were performed by varying patient age, sex, and health care costs across seven countries.

The benefits of primary prevention with simvastatin for diabetic individuals were estimated to range from 0.81 to 5.36 YOLSs(undiscounted) for men and from 0.54 to 2.83 for women(Table 1). Among men with CVD,secondary prevention was associated with an increased life expectancy ranging from 0.78 to 3.9 YOLSs. In the presence of CVD, the benefits for women were only slightly less than those for men ranging from 0.63 to 2.72 YOLSs. Accordingly, the estimated benefits of lipid therapy among diabetic patients without heart disease are comparable to those among heart disease patients without diabetes.

The direct health care costs associated with treating CVD complications are summarized in Table 2. These unit costs varied approximately threefold across the seven countries studied, with the U.S. at the high end. For instance, the average cost of coronary artery bypass surgery was $30,395 in the U.S. versus $11,726 (17,416 Canadian dollars) in Canada.

After both costs and benefits over the lifetime of treated individuals were discounted, the cost effectiveness of primary prevention among diabetic patients was compared with secondary prevention among CVD patients (Figs. 1 and 2). cost effectiveness ratios in men using U.S. costs ranged from $5,063 to $14,156 per YOLS in diabetic patients without CVD. In men with CVD but without diabetes, cost effectiveness ratios ranged from $8,799 to $14,996 per YOLS. Across all age and sex cohorts studied, the cost effectiveness of simvastatin in diabetic patients without CVD was similar to that in patients with CVD without diabetes.

Regardless of the health care costs involved in each country analyzed,primary prevention among diabetic patients was as cost-effective as secondary prevention among CVD patients (Table 3). For instance, among Canadian diabetic men who started therapy at 40 years of age, the cost effectiveness of simvastatin therapy was $3,869(U.S. dollars) per YOLS versus $7,156 (U.S. dollars) for men with CVD. For 40-year-old diabetic women, the cost effectiveness of primary prevention($10,932 [U.S. dollars] per YOLS) was similar to secondary prevention without diabetes ($11,189 [U.S. dollars]). These results were consistent among young and older patients in all seven countries studied.

The treatment of dyslipidemia among diabetic patients without CVD in the U.S. is estimated to be as cost-effective as treatment among CVD patients without diabetes. These results are consistent across Canada and the five European health systems evaluated in this study.

We have previously demonstrated that the Cardiovascular Disease Life Expectancy Model can reasonably predict the absolute risk reduction observed after lipid therapy among patients with both diabetes and CVD in such studies as the 4S(4,20). In the current model, the relative risk associated with diabetes for a future coronary death is only slightly less than that predicted after symptomatic CVD. Accordingly, the absolute risk for coronary death is approximately equal in the presence of either condition. This prediction is supported by a recent study by Haffner et al. (5)showing that the 7-year incidence of MI in nondiabetic subjects (45-64 years of age) with prior MI at baseline was shown to be similar to that for diabetic subjects without prior MI (18.8 vs. 20.2%, respectively). Furthermore,Miettinen et al. (8) showed that the 1-year case fatality rate for first MI (from the onset of symptoms,thus including prehospitalization mortality) in the FINMONICA population was 44% in diabetic men and 37% in diabetic women. These case fatality rates were significantly higher than rates in nondiabetic men and women (33 and 20%,respectively). Of the diabetic subjects who died, 63% of men and 28% of women died before hospitalization. These individuals, by definition, could not benefit from secondary prevention strategies, indicating that aggressive management of CVD risk factors in diabetic subjects should precede the onset of clinical CVD.

One must recognize the absence of clinical trial data confirming that treating dyslipidemia among diabetic patients without CVD significantly reduces primary CVD events and improves overall survival. Accordingly, the economic analyses presented here are based on a disease simulation model in which the relative risk reduction associated with treating dyslipidemia is consistent across a broad range of patients and that the absolute benefit associated with therapy is a function of this relative risk reduction and the absolute risk of disease. We also recognize that the lipid profiles of patients with diabetes versus CVD may be different outside of a clinical trial. Further analyses are therefore required on real population data.

On average in the U.S., Medicare beneficiaries with diabetes are ∼50%more expensive than all Medicare beneficiaries(29). Direct medical expenditures attributable to diabetes in 1997 totaled $44.1 billion, and CVD represented the single most important diagnostic category, accounting for $7.6 billion (30). In the setting of managed care, health care expenditures for diabetic patients were 2.4 times those of matched nondiabetic subjects with CVD complications, accounting for>20% of excess health costs(31). Once CVD had developed,direct medical care costs among diabetic patients were also significantly higher than those for nondiabetic patients and approximated $3,000 more per person-year of observation(32).

The health care costs associated with diabetes have been evaluated in some countries and appear to be strongly associated with CVD. In Germany, the cost of prescription drugs was significantly increased among diabetic patients in comparison to that of nondiabetic patients. The relative costs were still 1.5 times higher after excluding medications specifically for diabetes and adjusting for age and sex differences. CVD drugs also represented the single most important drug class, accounting for 39% of pharmaceutical costs(33).

In the U.K., one study found a fourfold increased risk of cerebrovascular events among diabetic patients in comparison to nondiabetic people(34). Although the prevalence of known diabetes in this study population was <2%, these patients accounted for ∼14% of the total cost of treating cerebrovascular disease and 17% of the total cost associated with treating coronary heart disease(34,35). In Spain, the total lifetime cost of treating type 1 diabetes has been estimated to average 12.7 million pesos ($85,000 [U.S. dollars])(36). However, the annual cost associated with type 2 diabetes or CVD complications was not specified in this study.

Lipid therapy among diabetic patients with known CVD has been shown to be beneficial. We believe that the epidemiological data and the modeling results from this study strongly support the expectation that lipid therapy will reduce CVD complications among diabetic patients. Economic analyses surrounding these projections demonstrate that lipid therapy among diabetic patients is not only clinically justifiable, but also economically attractive compared with other currently accepted therapies(37).

Support for this study was provided by Merck.