Interpreting, Integrating, and Individualizing Evidence about the Prevention of Diabetic Nephropathy

Policy-relevant research for the care of patients with diabetes mellitus is in high demand, given the prevalence, morbidity, and costs associated with this condition. Randomized trials, meta-analyses, and economic studies have evaluated myriad management strategies directed at the diagnosis, prevention, and treatment of the macrovascular and microvascular sequelae of diabetes. Accordingly, diabetes mellitus is an ideal condition for evidence-based disease management (1, 2). A decision analysis reported in this issue focuses on prevention of diabetic nephropathy [3]. Golan and colleagues address this question: In patients with newly diagnosed type 2 diabetes mellitus, what is the relative impact on quality-adjusted life expectancy and costs associated with 1) treating all patients with angiotensin-converting enzyme [ACE] inhibitors, 2) screening for microalbuminuria and treating with ACE inhibitors when it develops, or 3) screening and treating only for gross proteinuria? Where should we look for evidence to answer such questions? Increasingly, randomized trials are furnishing information on the consequences of different screening strategies for such conditions as hypertension, hypercholesterolemia, and breast cancer (4). A randomized trial in which patients are assigned to one of these three strategies would tell us what does happen under each alternative. If follow-up is complete, the trial is more likely to be unbiased; if the sample is large, its estimates will be more precise. If adherence of caregivers and patients with the protocol is high, the results will apply under ideal conditions. If adherence is low, the results may better reflect real world conditions. However, no randomized trial has addressed screening and prevention of diabetic nephropathy in patients with type 2 diabetes mellitus. The decision analysis by Golan and colleagues provides a model for what could happen under three strategies. The estimates of patient outcomes incorporate the rate of progression from normoalbuminuria to microalbuminuria to gross proteinuria to renal failure, modified by treatment and its side effects. To the extent that these estimates are accurate, the analysis paints a realistic picture of the relative impact of these management options. Using a lifetime horizon, Golan and colleagues apply Markov modeling to simulate progression of diabetic nephropathy with the treat all strategy compared with that under two different screen and treat approaches. Treating all diabetic patients with ACE inhibitors was associated with the lowest probability of death or end-stage renal disease (and thus the best quality-adjusted survival) for up to 25 years of follow-up. When the treat all approach was compared with screening for microalbuminuria, the marginal cost-effectiveness ratio was $7500 per quality-adjusted life-year. The lowest benefit and highest cost were associated with screening for gross proteinuria. Clinicians interested in appraising what they read will consider the rigor of this decision analysis (5, 6) and its application to their own setting (7, 8). The question driving this analysis is focused, the alternatives are well-defined, and the outcomes are important. However, the authors do not detail their methods of seeking and selecting evidence, which generates some uncertainty about the quality of the data incorporated in the model. With regard to the validity of the evidence presented, the authors acknowledge that the effectiveness of ACE inhibitors in slowing the development of microalbuminuria is based on only one randomized trial of 156 patients that showed an absolute risk reduction for microalbuminuria of 12.5% (95% CI, 2% to 23%) (P=0.042) over 6 years (9). The transition time from gross proteinuria to end-stage renal disease was extrapolated from data on patients with type 1 diabetes mellitus (10). Actual patient values rather than projections from health care providers were used to estimate quality of life; the authors reference the source documents from which they derived the utilities of diabetic patients and their weighting by end-stage renal disease. The limited journal space available for describing all databases, literature, and assumptions used for integrative articles such as this one is ineluctable. Nevertheless, transparent reporting is crucial (11) so that readers can determine whether relevant, valid, and current costs and consequences were included, measured, and valued (12). Although decision and cost-effectiveness analyses emphasize transparent reasoning, they also highlight uncertainty, showing how an approximate measure of the correct results is better than a precise measure of the incorrect ones. Golan and colleagues explore the influence of uncertainty through sensitivity analyses, which indicate that an increasing marginal cost-effectiveness ratio (that is, more cost to achieve a benefit) of the treat all approach relative to screening for microalbuminuria was associated with increasing patient age, decreasing effectiveness and increasing cost of ACE inhibitors, and impaired quality of life associated with this therapy. The marginal cost-effectiveness ratio decreased as the cost of end-stage renal disease increased. How might policymakers interpret Golan and colleagues' analysis? Taking a societal perspective, the investigators assumed that patient nonadherence with treatment is 25% at 3 months and an additional 2% per year thereafter. Screening nonadherence by clinicians is assumed to be 50%. Patients who continue and those who discontinue therapy with ACE inhibitors contribute to the overall estimates of cost and benefit. Most policymakers would acknowledge that recommendations to screen for microalbuminuria are weak and that alternative approaches (particularly the treat all approach) may be equally reasonable. However, policymakers in one setting might make different recommendations than policymakers in another setting on the basis of health care costs in their jurisdiction. How might clinicians interpret this analysis? For the patient with newly diagnosed type 2 diabetes mellitus who is 60 years of age and pledges to take medication religiously if it might help her, a clinician may recommend ACE inhibitor therapy; alternatively, the clinician may screen for microalbuminuria if she is unimpressed by the incremental health gains associated with the treat all approach. For the patient who is 80 years of age and abhors pills, the clinician may elect to screen for gross proteinuria; alternatively, she may neither screen nor treat if the patient's anticipated quality of life while receiving medication is poor and her compliance is implausible. In health systems in which the cost of medication is borne by the patient, the patient's financial circumstances may drive disease management. In summary, clinicians and patients with different perspectives and resources might come to different conclusions when presented with the same evidence about a screening procedure (13). In parallel, legislative, administrative, and industrial policymakers may extract different messages. Clinical and economic research evidence, political exigencies, patient and provider preferences, and many other concerns will ultimately determine how the prevention of diabetic nephropathy is played out at the micro, meso, and macro level (14). At issue, therefore, is not how this analysis will produce the right decision for a given situation but how different values will influence the way it is interpreted. Finally, new evidence about the prevention of nephropathy is emerging from the recently completed Heart Outcomes Prevention Evaluation trial. Among more than 9000 patients at risk for cardiovascular disease in 19 countries, ramipril was associated with fewer cardiac outcomes and lower rates of microalbuminuria than placebo, both in patients with and those without diabetes (Yusuf S. Personal communication, 1999). Interpreting, integrating, and individualizing evidence in this field will continue to evolve.