The Cost-Effectiveness of Treating All Patients with Type 2 Diabetes with Angiotensin-Converting Enzyme Inhibitors FREE

Diabetes mellitus is the most common cause of end-stage renal disease (ESRD). In 1996, 42% of new cases of ESRD were caused by diabetes. Two thirds of these cases occurred in patients with type 2 diabetes (1 - 2). Although tight control of blood sugar and hypertension might slow the development and progression of nephropathy in type 2 diabetes (3), the most promising approach has been angiotensin-converting enzyme (ACE) inhibitor therapy (4 - 5).

Current recommendations emphasize identification and treatment of patients with incipient diabetic nephropathy by screening for microalbuminuria (6). Because this test is not uniformly available (and, until recently, samples had to be sent to specialized laboratories), some physicians screen only for gross proteinuria with a dipstick or urinalysis. Others undoubtedly do not screen at all, either because they are unaware of the recommendations or they forget to do so (7). Thus, some patients who might benefit from ACE inhibitors are not prescribed them (8). Treating all patients with diabetes might be a simpler strategy, but patients who take ACE inhibitors may experience significant side effects (9 - 10) and may not comply with treatment (11 - 14).

The opportunity to avoid [or delay] ESRD in a few patients must be weighed against the cost of treating many patients with ACE inhibitors. Unfortunately, this issue is unlikely to be settled in clinical trials because a large number of patients and a very long follow-up would be required. For this reason, we used a decision model to simulate costs and clinical outcomes associated with three strategies for preserving renal function in patients with newly diagnosed type 2 diabetes: 1) treating all patients with ACE inhibitors, 2) screening for microalbuminuria, and 3) screening for gross proteinuria.

We constructed a Markov model (by using DATA 3.0 [TreeAge Software, Inc., Williamstown, Massachusetts]) to evaluate the cost-effectiveness of three clinical strategies for managing type 2 diabetes. Input data on the effectiveness of ACE inhibitors were obtained from three randomized trials (15 - 17), each of which examined treatment in a different stage of proteinuria.

As shown in Figure 1, we used four health states to model the progression of diabetic nephropathy: normoalbuminuria (albumin excretion < 30 mg/d), microalbuminuria (excretion, 30 to 300 mg/d), gross proteinuria (excretion > 300 mg/d), and ESRD. The health states were further subdivided to reflect detection (whether or not patients have been screened) and treatment (whether or not patients are receiving ACE inhibitors).

With each “cycle” of the model, patients may progress to the next health state or to death, based on a specific transition probability. To simulate the typical screening interval, we used a cycle length of 1 year. We assumed that nephropathy does not regress and that progression is sequential (that is, no states are skipped). However, patients may make the transition to death at any time. By following a large cohort of identical patients through time until all are in the death state, the model calculates average lifetime cost, life expectancy, quality-adjusted life expectancy, and marginal cost-effectiveness (the additional cost per additional quality-adjusted life-year [QALY] compared with next best strategy) for each strategy.

We assessed three competing strategies. In the “treat-all” strategy, patients are not screened and all patients start receiving ACE inhibitor therapy at the time of diagnosis of diabetes mellitus. Screening adherence is not relevant to this strategy, but our modeling of side effects and noncompliance means that some patients discontinue treatment. In the “screen for microalbuminuria” strategy, patients are screened for microalbuminuria and treatment is attempted in all patients whose test result is positive; thus, both adherence to screening recommendations and treatment discontinuation are relevant to this strategy. In the “screen for gross proteinuria” strategy, patients are screened for gross proteinuria; these patients will also be affected by adherence to screening recommendations and treatment discontinuation.

The target population is patients with newly diagnosed type 2 diabetes mellitus who are not already receiving ACE inhibitors for other reasons (such as hypertension and heart disease). On the basis of the prevalence of protein excretion in patients with newly detected diabetes (18), the distribution of initial health states was established as follows: 79% of patients had normoalbuminuria, 18% had microalbuminuria, and 3% had gross proteinuria. The starting age for the base-case was 50 years, reflecting the average age at diagnosis of diabetes mellitus (19). We tested other age groups in sensitivity analyses.

Table 1 shows the transition probabilities with and without ACE inhibitor therapy. Transition rates from normoalbuminuria to microalbuminuria and from microalbuminuria to gross proteinuria were derived from data reported by two randomized trials in patients with type 2 diabetes (16 - 17). Because these data excluded patients in whom treatment failed, they represent “efficacy” data. Unfortunately, similar data on the subsequent transition from gross proteinuria to ESRD were not available for patients with type 2 diabetes. For this reason, we estimated this variable from data in a randomized trial in patients with type 1 diabetes (15). For each trial, we estimated the progression rate in the treatment and placebo groups by assuming that progression was constant over the study period. We then converted progression rates into annual probabilities.

We modeled mortality as a function of age. Our approach for the initial health states (normoalbuminuria, microalbuminuria, and gross proteinuria) was to use age-specific mortality rates in the general population (20) and multiply them by a standardized mortality ratio for persons with diabetes (the ratio of the all-cause mortality rate in persons with diabetes relative to that in the general population). We used a standardized mortality ratio of 2.0, derived from a long-term follow-up study of diabetic persons 45 to 64 years of age (21). Once a patient progressed to ESRD, we assumed that the mortality rate was age-independent. We used a mortality rate of 27% per year, obtained from the U.S. Renal Data System (1). Mortality rates were converted to probabilities.

We assumed that screening is performed once a year, as recommended by the National Kidney Foundation (6). The testing procedure for microalbuminuria is complex (for example, it stipulates that two of three samples must test positive within 3 months), and various conditions that could produce false-positive results must be avoided (such as heavy exercise, febrile illness, congestive heart failure, and urinary tract infection). Because physicians do not always adhere to screening recommendations, our model included a variable reflecting the proportion of patients actually screened each year. In our base case, we assumed a 50% screening adherence rate, based on a cross-sectional study of cholesterol screening in Medicare patients with diabetes (7 - 8). Although patients with proteinuria who missed screening in one year do not receive ACE inhibitors during that year, they may be screened and treated in the following year. All patients who receive ACE inhibitors or in whom this treatment fails are no longer screened.

Because noncompliance with medications may be high and side effects may be frequent, we allowed for discontinuation of ACE therapy in the model (8 - 9,11). We assumed that most side effects requiring discontinuation of therapy would appear during the first 3 months of treatment. In our base-case analysis, 25% of patients starting ACE inhibitor therapy discontinued this treatment in the first 3 months. To reflect the gradual decrease in compliance over time, we assumed that another 2% of patients discontinue treatment each year. We assumed that patients who discontinue ACE inhibitor therapy for any reason (side effects or noncompliance) do not restart treatment. We also assumed that patients who discontinue ACE inhibitor therapy during the first year of treatment do not receive any benefit from the medication (that is, they experience the same rate of disease progression as untreated patients do).

We conducted our analysis from the societal perspective. For our base-case analysis, we considered only health care costs that are associated with ACE inhibitor therapy, screening, or treatment of ESRD. We assumed that other health-related costs of diabetes are the same for the three strategies. The baseline cost estimates and the ranges that were tested in sensitivity analyses are shown in Table 2. We based the annual cost of ACE inhibitor therapy ($320) on the average wholesale price of lisinopril in 1998 (22). The costs of screening for microalbuminuria ($20) and gross proteinuria ($3) were based on the Medicare Clinical Diagnostic Fee Schedule for 1998 (23). Patients are first screened with urinalysis for gross proteinuria (Current Procedural Terminology code 81005) and, in the “screen for microalbuminuria” strategy, also for microalbuminuria (Current Procedural Terminology code 82044 for a dipstick and code 82043 for a quantitative assay of similar price). The annual cost of treating patients with ESRD was based on total Medicare payments in 1996, as estimated by the U.S. Renal Database System (1). This estimate is an average for patients treated with different modes of dialysis and renal transplantation. Because of considerable uncertainty about cost estimates, we varied these data widely in sensitivity analyses.

We measured three outcomes: average cost, life expectancy, and quality-adjusted life expectancy. We used a 3% discount rate for all costs and health benefits. We used utilities (quality-of-life adjustments) from the Beaver Dam Health Outcomes Study (24) to reflect the imperfect quality of life in patients with diabetes. Quality of life for patients with ESRD was further adjusted by using a weight obtained from a mixed sample of patients undergoing transplantation, hemodialysis, or peritoneal dialysis (26).

In the base-case analysis, the “treat all” strategy was associated with the lowest likelihood of ESRD (1.2%) or death (14.6%) and the highest likelihood of normoalbuminuria (57%) at 10 years. Figure 2 shows the distribution of health states at 10 years for all three strategies and for patients who were not treated with ACE inhibitors at all. As shown in Table 3, the finding that the proportion of patients in ESRD or death was the lowest in the “treat all” strategy persisted with various time horizons. However, the absolute difference in this outcome across strategies was small.

The “screen for gross proteinuria” strategy had the highest cost ($19 520) and the lowest benefit (15.39 life-years and 11.59 QALYs) and was thus dominated by the other strategies. Table 4 shows that the currently recommended strategy, “screen for microalbuminuria,” had the lowest cost ($14 940) and therefore served as our reference case. Compared with the “screen for microalbuminuria” strategy, the “treat all” strategy was more expensive ($15 240) but was associated with the highest life expectancy (15.63 and 15.59 life-years) and quality-adjusted life expectancy (11.82 and 11.78 QALYs). The marginal cost-effectiveness ratio (cost per additional QALY) of the “treat all” strategy compared with the “screen for microalbuminuria” strategy was $7500 per QALY gained.

The cost-effectiveness of the “treat all” strategy relative to the “screen for microalbuminuria” strategy was sensitive to age at diagnosis of diabetes, cost of ACE inhibitors, relative risk for progression to microalbuminuria, and quality-of-life adjustment for ACE inhibitors.

The marginal cost-effectiveness ratio increased with increasing age. It exceeded $20 000/QALY if the patient's age at diagnosis was 55 years or older (at the base-case cost of ACE inhibitors). “Treat all” became the dominant strategy (that is, the least expensive with the highest benefit) if patients were younger than 44 years at the time of diagnosis. Although the “screen for gross proteinuria” strategy was dominated across most age groups, it became dominant in very elderly patients (>85 years of age), for whom the competing risk for death was most pronounced.

As expected, the cost of ACE inhibitors influenced the cost-effectiveness significantly. The marginal cost-effectiveness ratio increased with increasing cost of this therapy and exceeded $20 000/QALY for patients 50 years of age if the annual cost of therapy was more than $420. As shown in Figure 3, the rate of increase was steeper for older patients. “Treat all” became the least expensive strategy if the annual cost of ACE inhibitors was less than $260. As expected, the marginal cost-effectiveness ratio of the “treat all” strategy decreased as the cost of ESRD increased; that is, the “treat all” strategy became more cost-effective. If the annual cost of ESRD exceeded $60 000, “treat all” became the least expensive strategy.

We varied the relative risk for progression to microalbuminuria with ACE inhibitor compared with no treatment. The benefit of treating all patients decreased and the marginal cost-effectiveness ratio increased when ACE inhibitors became less efficacious. As the relative risk exceeded 0.46 in patients 50 years of age and 0.6 in patients 40 years of age, the marginal cost-effectiveness ratio exceeded $20 000/QALY (Figure 3).

Not surprisingly, the magnitude of the adjustment for quality of life during ACE inhibitor therapy had a powerful influence on the marginal cost-effectiveness ratio for treating all patients. In our base case, we assumed that patients' quality of life was unaffected by the medication. In sensitivity analysis, treating all patients no longer provided the highest benefit if the utility (quality-of-life adjustment) for ACE inhibitors decreased to 0.99. Thus, individual patients who experience bothersome symptoms do not experience any gain in quality-adjusted life expectancy.

Our findings were not sensitive to screening adherence or treatment discontinuation. Even if screening adherence was perfect, the “treat all” strategy was still cost-effective (with a marginal cost-effectiveness ratio of $25 000). If patients missed half of their screenings a year (our base case), the marginal cost-effectiveness ratio decreased to $7500. Even if no patient discontinued treatment, the marginal cost-effectiveness ratio of the “treat all” strategy did not change significantly. Our conclusion was not affected by variations of the discount rate and other utilities.

Our findings suggest that a strategy of using ACE inhibitors in all patients with type 2 diabetes would slow the progression to ESRD at a relatively low cost. The currently recommended strategy of screening for microalbuminuria is the least expensive, but not the most effective one. Screening for gross proteinuria, a frequently used strategy, is the most expensive and the least effective.

Two factors favor the use of ACE inhibitor therapy in all diabetic patients at the time of diagnosis. First, a randomized trial proved that ACE inhibitors were efficacious in patients with normoalbuminuria because they slowed the rate of transition to microalbuminuria (17). Second, the low adherence to screening recommendations prevents many patients from receiving treatment. Kraft and colleagues (7) surveyed screening practices for microalbuminuria in primary care setting and found that only 17% of physicians reported screening for microalbuminuria (whereas 82% of physicians reported screening for gross proteinuria). Whether poor adherence is a result of insufficient knowledge of the screening recommendations or of their complexity is unknown. Regardless, treating all patients with diabetes would extend the benefits of ACE inhibitor therapy to more patients.

Our study has several limitations. We did not explicitly consider the sensitivity or specificity of the screening tests for microalbuminuria and gross proteinuria. Our assumption that the screening tests are perfectly sensitive favors the two screening strategies. The model indirectly accounts for their specificity—the outcomes for patients with false-positive test results are reflected in the progression rates observed in the randomized trials.

We also did not explicitly consider that a certain proportion of patients in the screening strategies would be prescribed ACE inhibitors for other reasons (such as hypertension or congestive heart failure). Functionally, this is equivalent to allowing patients to “cross over” from the screening strategies to the “treat all” strategy. Although modeling this crossover would move both the cost and outcome of the two screening strategies closer to that of the “treat all” strategy, it would not affect the marginal cost-effectiveness ratio.

Finally, our input data had limitations. First, only one study has tested the ability of ACE inhibitors to slow the transition from normoalbuminuria to microalbuminuria (17). If the point estimate of benefit obtained from this study is wrong, this could greatly affect the conclusion of our analysis. Second, it is important to emphasize that the effect of ACE inhibitors to slow or prevent ESRD in patients with type 2 diabetes and gross proteinuria represents an inference from a trial involving patients with type 1 diabetes (15). Finally, no single study has related the use of ACE inhibitors in patients with new-onset type 2 diabetes directly to the development of ESRD (27).

The absence of “ideal” data highlights one of the advantages of the decision analysis approach. Because it would require a large number of patients and a long follow-up, it is unlikely that we will soon see a randomized trial of ACE inhibitors in patients with new-onset type 2 diabetes that uses the outcome of ESRD or death. Our model combines the observations made at various stages of renal function and thus represents the next best alternative—a “virtual” trial. It also takes into account imperfect adherence to screening guidelines and thus may be more applicable to routine clinical practice than a clinical trial. The model allows us to vary the adherence to screening guidelines and treatment discontinuation, age at diagnosis, costs, different degrees of efficacy for ACE inhibitor therapy, mortality, and utilities and examine the robustness of the conclusion.

Not surprisingly, the cost of ACE inhibitors has a powerful effect on the cost-effectiveness ratio. If the annual cost of ACE inhibitors exceeds $420, the marginal cost-effectiveness ratio of the “treat all” strategy exceeds $20 000/QALY. On the other hand, if the cost decreases below $260, treating all patients becomes the least expensive strategy. Our base-case estimate ($320) was based on the average wholesale price of lisinopril and represents the mid-range of different medications and different doses we obtained from several local pharmacies. Some patients may be paying much more, whereas some health plans (as well as the Department of Veterans Affairs) may be paying much less. In the near future, the cost of ACE inhibitors is likely to decrease considerably because other agents will be off patent (only one, captopril, is currently off patent). Policymakers should adjust the cost-effectiveness ratio to reflect their local cost.

It is important to emphasize that ACE inhibitors, like all pharmaceutical medications, can cause complications. The most common serious complication of ACE inhibitor therapy is acute renal failure, which, were it to become permanent, would be reflected in the progression rates obtained in the randomized trials. However, no serious complications were observed in these trials. Angiotensin-converting enzyme inhibitors seem to be safe and, in addition to their benefit in diabetic nephropathy, may also benefit some patients with heart disease. These drugs are already indicated for patients with hypertension or congestive heart failure, after myocardial infarction, and in microalbuminuria or gross proteinuria due to diabetes. These patients are at greater risk for acute renal failure induced by ACE inhibitors. In our analysis, we evaluate expanding the recommendations to a relatively healthier population in which risk for acute renal failure is relatively lower. Nevertheless, judicious monitoring of potassium and renal function is indicated for all patients receiving ACE inhibitors.

Finally, we should be clear about to whom the “treat all” strategy applies. First, our analysis applies only to the patient population that was included in the randomized trials—those who meet the older diagnostic criteria for diabetes (fasting plasma glucose level ≥ 7.8 mmol/L [≥ 140 mg/dL]). The recommendation to “treat all” should not be extrapolated to include patients who now receive the label “diabetes” according to the lower diagnostic threshold recommended by the American Diabetes Association (which has undoubtedly increased the number of persons with diabetes and normoalbuminuria) (28). Second, because the benefit (in terms of QALYs) is very sensitive to patients' quality of life during ACE inhibitor therapy, the “treat all” strategy makes sense only for patients who are not bothered by treatment. Patients should not be expected (or encouraged) to accept bothersome side effects of ACE inhibitors in hopes of an expected future benefit.

Both the prevalence of diabetes mellitus and ESRD due to diabetic nephropathy are increasing substantially. End-stage renal disease is associated with a very high mortality rate, reduced quality of life, and high cost of treatment. Any strategy that could slow or prevent the development of ESRD is attractive from the perspective of both the patient and society. Our study suggests such a strategy: treating all persons with new-onset type 2 diabetes with ACE inhibitors. This strategy avoids a complex screening process and provides additional benefit at a modest additional cost.