Randomized controlled trials have demonstrated the efficacy of lifestyle interventions aimed at preventing or delaying onset of type 2 diabetes (1-4). The Diabetes Prevention Program (DPP) found that either medication or intensive lifestyle interventions could prevent progression from impaired glucose tolerance to diabetes (1). This randomized controlled trial provided study participants with individualized, resource-intensive management and oversight (5). During the 3 years of the DPP, diet and exercise reduced the risk that patients with impaired glucose tolerance would develop diabetes by 58% (1). These outcomes appear sustainable given follow-up data from the Da Qing Diabetes Prevention Study (6) and the Finnish Diabetes Prevention Study (7). Unfortunately, such programs can be expensive to implement and can exclude people with some comorbidities, who may benefit from even modest improvements in diet and physical activity.

Community-based lifestyle interventions adapted from the DPP (1) demonstrate effectiveness for improving the risk factors for diabetes in community settings (8-10), but their costs are largely unexplored. US cost-effectiveness studies of intensive lifestyle interventions to prevent diabetes have differed in their perspectives, time frames, and inclusion of prediabetes screening costs, producing cost-effectiveness ratios that range from $1,100 to $143,000 per quality-adjusted life-year (QALY) (11-13). However, intervention efficacy and cost data for these studies were largely based on the intensive strategies and resources used by the original DPP (14,15).

The Diabetes Prevention Support Center of the University of Pittsburgh Diabetes Institute developed a modified version of the DPP lifestyle intervention (mDPP) and tested its effectiveness in the community and local medical practice settings with patients at increased risk of diabetes or cardiovascular disease (CVD) (8,16). We assessed cost-effectiveness of this mDPP.

By the 12-month point in the model, the mDPP intervention reduced metabolic syndrome risk at 1 year by 16.2% (compared with usual care, which reduced the risk of metabolic syndrome by 12.1%). During the 3-year time frame of the model, both costs and effectiveness of the mDPP were slightly higher than usual care. The mDPP costs totaled $2,528 (compared with $2,493 for usual care) and the effectiveness of the mDPP equaled 2.40 QALYs (compared with 2.39 QALYs for usual care). Taken together, the mDPP gained 0.01 QALYs (approximately 3.67 days) at an incremental cost of $34.50, equal to an incremental cost-effectiveness ratio of $3,420 per QALY. These results were due mainly to decreases in diabetes incidence with the mDPP. Without the mDPP, 9.6% of the cohort developed diabetes over 3 years; with the mDPP, 7.7% did. Over this period, little difference between groups was seen (1.1% vs 0.9%) in diabetes complication incidence, as broadly defined by the DPP.

In 1-way sensitivity analyses, results were most sensitive to changes in risk-factor reduction with or without the mDPP, intervention rates, risk-factor–positive screening rates, and diabetes incidence rates in risk-factor–positive people with or without an intervention. Results were also sensitive to utilities for risk-factor–positive patients (Figure 2).

When base case values were used for all parameters, an mDPP intervention cost $3,420 per QALY. When parameters were varied to the extremes of the ranges shown in the Table, the cost-effectiveness ratio remained less than $20,000 per QALY. The cost-effectiveness ratio rose highest, $18,600 per QALY, when the probability of reducing risk factors in the absence of an mDPP intervention was increased from 12.1% in the base case to 25.9% (Figure 2). In addition, the cost-effectiveness of the mDPP intervention remained less than $17,000 per QALY when the proportion of patients who screen positive for risk factors increased to 63.5% (base 31%), when the proportion of risk-factor–positive patients who enroll in the intervention increased to 86.7% (base 47.0%), when the probability of risk-factor reduction increased to 34.4% (base 16.2%), and when the relative risk of diabetes given risk factor positivity with or without an mDPP decreased by two-thirds. When all other parameters listed in the Table were varied in 1-way sensitivity analyses across their specified ranges, the cost-effectiveness ratio did not exceed $7,000 per QALY. Thus, results favoring the mDPP were stable to variations in risk-factor reduction and diabetes risks that might be seen in differing populations. Varying the discount rate for costs and effectiveness (base 3%) from 0% to 5% changed model results by less than $400 per QALY gained because of the brief 3-year span of the model.

In a separate sensitivity analysis examining differing assumptions for metabolic syndrome reduction, if we assume no reduction in metabolic syndrome for enrolled or nonenrolled patients after model year 1, the cost-effectiveness ratio of the intervention increased slightly, to $3,400 per QALY gained. If enrolled patients have no further reduction in metabolic syndrome after the first year but nonenrolled patients continue metabolic syndrome reduction through the 3 years at rates seen in the DPP placebo arm, the intervention would cost $7,270 per QALY.

In a probabilistic sensitivity analysis, when all parameters were varied simultaneously across their ranges, the mDPP intervention cost less than $20,000 per QALY gained in approximately 78% of model iterations and less than $50,000 per QALY in approximately 86% of iterations (Figure 3). This analysis assumed independence among parameter values and did not account for covariance between parameters, which tends to broaden result ranges. Thus, we may have underestimated the probabilities.

When the utility weight for risk-factor–positive patients not enrolled in an intervention (ie, receiving usual care) was set equal to the utility weight for patients enrolled in an mDPP, the mDPP intervention cost $8,300 per QALY and yielded life expectancy gains of 0.004 QALYs (approximately 1.63 days) for enrollees. We also examined scenarios using sets of parameter values unfavorable to the mDPP intervention. If we simultaneously assume no utility differences with risk factors between intervention groups (as above), a 9.8% diabetes risk on the mDPP (so that the risk is only slightly better than the 10.8% risk of no mDPP), and a utility of 0.77 with no risk factors (instead of a low value of 0.84), then the mDPP cost $56,200 per QALY gained. If we add to this scenario a decrease in mDPP-related benefit (from 16.2% to 14.1%, a 2 percentage point difference compared with no mDPP), then the mDPP cost $95,400 per QALY.

The mDPP was designed to teach groups of people how to change their diet and lifestyle to reduce their risk for diabetes and CVD. When we examined the costs and effects of implementing the mDPP intervention in a community setting, we found that at 12-month follow-up the mDPP reduced the relative risk of metabolic syndrome by 16.2% and yielded a life-expectancy gain of 3.67 days at a cost of $3,420 per QALY gained. Even when we varied parameter values in sensitivity analyses, the cost-effectiveness ratio remained less than $20,000 per QALY. Ratios less than $20,000 are generally considered to provide strong evidence in favor of adopting an intervention (31).

In analyzing the mDPP, we used several conservative practices and assumptions that would be expected to negatively bias our findings. First, we used outcome parameters based on data from real-world mDPPs in which participation ranged from attending 1 session to attending 12 sessions (mean, 8.9 sessions; median, 10 sessions). In contrast to usual study designs, in which attrition and dropout rates are final states (ie, patients do not return after dropping out), our study design included patients who skipped sessions throughout the 12-week course. Second, our design was based on an intent-to-treat analysis in which all patients who attended at least 1 session were included and any patient who did not return for the 12-month follow-up was assumed to be risk-factor–positive even if the 3-month data indicated resolution in the patient's metabolic syndrome. Third, we modeled the effects of the intervention over a 3-year time frame. If the effects are actually sustained beyond 3 years, as other studies suggest (6,12), then our modeling approach underestimates the cost-effectiveness of the intervention. Given the short time frame of our analysis, the factors that contributed the most to intervention effectiveness were the changes in quality of life that are due to avoidance of diabetes and its complications. Over longer time frames, other factors could have a greater impact.

Our study has 2 limitations that deserve mention. First, we used costs in 2000 US dollars. This decision might underestimate the costs of providing the mDPP intervention and usual care today, but it would not be expected to greatly underestimate the costs of the mDPP intervention relative to the costs of usual care in any particular year. Moreover, the choice of base year would not be expected to influence the criteria for cost-effectiveness (31). Second, as with any modeling exercise, we imposed several simplifications. For example, although the probabilities of acquiring diabetes, progressing to complicated diabetes, and dying of diabetes or other causes depend on a large number of covariates, we included only the most common covariates. Addressing these limitations would require the performance of a large-scale trial to compare the mDPP intervention with usual care for a long follow-up period. In the absence of such a trial, we believe that using the control arm of the original DPP as the comparison cohort for the mDPP intervention was a reasonable alternative.

Both the original DPP and the mDPP provide study participants with instructions about diet and physical activity. The main difference is that the DPP provides individualized instruction to participants with specific types of comorbidities, whereas the mDPP provides group instruction and can be applied to participants with a larger range of comorbidities. Although the DPP was found to be economically reasonable (12), it is more expensive than the mDPP and its complexity and rigor make it more difficult for medical practitioners to fully implement. The mDPP is simpler and less time-consuming to implement. When delivered in community and primary care settings, it is cost-effective and appears to be a sound investment.