Results of several controlled trials have shown that lifestyle interventions can reduce the incidence of type 2 diabetes among people at high risk [7
]. Questions remain, however, about the impact of these targeted interventions, as well as those of population-wide interventions, on future diabetes incidence and prevalence rates. Although drawing definitive conclusions is difficult, due to parameter uncertainty, our model suggested that the greatest reduction in the number of diabetes cases would be achieved through implementation of a multitiered strategy involving a structured lifestyle intervention for adults with IFG in conjunction with risk-reduction policies aimed at the entire population. We projected that such an approach would result in 4.6 million fewer diabetes cases over 20 years and 3.6 million (14%) fewer Americans with diabetes in 2030 than corresponding projections based on the assumption of no national intervention. These projected reductions are substantial given the high lifetime risk of amputation, kidney disease, and CVD and the extensive lifetime health care costs among people with diagnosed diabetes [33
]. We projected that the next most effective strategy—and possibly the most plausible one under a setting of limited resources—would be the moderate-risk strategy (a lifestyle intervention for about 27% of the population), which we projected would result in a 12.2% attenuation in what we projected the increase in diabetes prevalence would be in the absence of any additional diabetes prevention intervention. We projected that the high-risk approach would produce a prevalence reduction only about one-third as large as the combined approach and that the population-wide approach would produce a prevalence reduction only about one-fourth as large; for both of these scenarios, our error propagation analysis reminds us that even these gains are somewhat uncertain. Our projection that the population-wide approach, by itself, would be relatively ineffective was in part because our assumption about the risk reduction to be derived was lower, but also because most people receiving the intervention would not have developed diabetes anyway.
Despite the benefits that can be achieved with preventive interventions, particularly the combined and moderate-risk approaches, our projections indicate a need to develop new options for diabetes prevention and/or increase the reach of both primary and secondary prevention efforts. We projected that the diabetes prevalence rate in the United States will surpass 20% by 2030 even if the combined strategy is implemented; a large increase in prevalence will occur even at the upper credibility bound for effectiveness of this intervention, the most optimistic plausible case. This projected increase was based on data showing a steady increase in the US diabetes incidence rate over the past 20 years [35
], a decline in diabetes mortality rates as diabetes has become increasingly treatable while remaining largely incurable [4
], and transition of the baby boom generation into the age range corresponding to peak diabetes incidence [38
Our projection that the US diabetes prevalence rate will increase substantially even under the combined intervention scenario indicates that the success of currently known primary prevention efforts will not diminish the need for more effective programs to help people manage their diabetes and prevent its complications. Diabetes has diffuse effects, leading to multiple microvascular and macrovascular complications and an increased risk for aging-related disability. Because numerous clinical, screening, community, and educational interventions reduce the incidence of complications [39
], many diabetes experts have argued that delivery of proven interventions is more important for the diabetic population than the development of new therapies [41
]. Our findings similarly indicate the importance of timely and efficient screening and treatment for hypertension, hyperlipidemia, chronic kidney disease, and diabetic eye and foot disease among people with diagnosed diabetes, especially as people with diabetes live longer with the condition. We should remember, however, that such success in the reduction of diabetes complications will likely lead to lower diabetes mortality rates and thus can contribute to an increase in diabetes prevalence.
Our projections are limited by our assumptions. If the US diabetes incidence rate continues to increase at its recent pace (rather than stabilizing at the 2007 rate, as we assumed in our primary analyses) and all other assumptions are unchanged, the prevalence of diabetes in 2030 would be three to four percentage points higher than we projected [13
]. If mortality rates associated with diabetes decline over time, as suggested by the results of some studies [4
], and all other assumptions are unchanged, the prevalence of diabetes in 2030 would also be higher than we projected. Our conclusions about the relative impact of various intervention policies on future diabetes prevalence rates would be relatively unaffected by changes in our assumptions about future diabetes incidence rates. However, our model also reflected our assumption that the interventions would have no effect on diabetes mortality rates other than their effect on diabetes risk. Any additional reductive effect on mortality rates associated with the interventions would actually result in smaller differences in diabetes prevalence projections among the intervention scenarios than those we reported. Perhaps the biggest practical limitation of our model is that it did not allow us to project diabetes prevalence by age, sex, or racial/ethnic group. Because the absolute prevalence of diabetes is much higher and more influenced by mortality rates among older adults than among younger adults, interventions could be more effective in reducing future diabetes prevalence rates among younger adults than our projections indicated they would be for the overall US adult population. Similarly, the prevalence rates among younger adults could stabilize or even decrease even though the prevalence rate among all US adults continues to rise.
As indicated by both our one-variable-at-a-time and error propagation analyses, our findings are sensitive to our assumptions about the effectiveness of interventions. Results of several randomized controlled trials among people with prediabetes have found a 50% or more reduction in incidence among people with prediabetes, with some finding extended effects up to 20 years later [9
]. Similar levels of weight loss among people with prediabetes as those seen in the diabetes prevention trials have been observed in interventions provided in community settings with economically sustainable staff and facilities [44
]. In the absence of data that would support a more complicated assumption, we made the simple assumption that the relative risk reduction associated with lifestyle intervention was the same in people with IFG as those with both IFG and IGT. However, our modeling did account for differences in the absolute risk between risk strata that would, in turn, lead to different levels of absolute risk reduction in response to an intervention. The accuracy of our projections was also dependent on the extent to which the “rule of halves” that has been reported in hypertension studies, wherein about half of the target population would be identified and referred to a lifestyle program and about half of those referred would actually initiate participation in a lifestyle program, is also applicable to diabetes lifestyle interventions [14
]. However, even if these assumptions are correct, the moderate- and high-risk approaches would require a strong system of reimbursement for providers of community lifestyle programs that does not yet exist.
The impact of a population-wide approach to diabetes prevention is particularly hard to predict because there have been few estimates of the health impact of such an approach; this is reflected in the large uncertainty in this term in our error propagation analysis. We based our estimates that population-wide strategies would lead to a 1 to 4% reduction in the relative incidence of diabetes on the results of studies concerning community-based efforts to reduce rates of CVD [17
]. Given the strong relationship between obesity and diabetes, the success of a population-wide approach to diabetes prevention will ultimately depend on the extent to which such an approach is successful in reducing obesity rates [48
]. Proposed approaches to doing this include taxing sugared beverages or sugar itself, mandating wide-scale menu labeling, providing incentives to increase the availability of healthy foods, encouraging urban designs that promote physical activity, and enhancing awareness of and education about risk factors and prevention behaviors [49
]. Programs to enhance employer-based health-promotion programs also may help reduce diabetes rates, particularly if they can effectively stratify the population by levels of diabetes risk and the intensity of the intervention appropriate for that risk level [17
]. Since these interventions could have important health effects on additional chronic conditions such as hypertension, cardiovascular disease, and disability, the pure focus of our analysis on prevention of diabetes cases may not reflect the value or the differential impact of the hypothetical interventions. Finally, the feasibility of any intervention depends on both effectiveness and cost. Real interventions similar to the hypothetical ones described here could have vastly different costs. However, due to the structure of our model, the scarcity of the literature, and the lack of details as to how national interventions like the ones we describe might be implemented, it was not feasible for us to consider cost in our estimates.
Although our projections of future diabetes prevalence rates among US adults are based on numerous assumptions, they are likely more accurate than previous projections because they are based on more recent data, including updated projections of future diabetes incidence rates and updated population projections. Most importantly, because the model we used estimated diabetes incidence and mortality rates for people at different stages of diabetes risk and in different diagnostic categories, it allowed us to more accurately project the impact of various intervention strategies. However, our projections, and others like these, need to be continuously updated as new intervention effectiveness and epidemiologic data become available. We hope that more rigorous controlled studies or perhaps natural experiments of population-targeted approaches for diabetes prevention will emerge in the future to permit more confident modeling of the impact on future prevalence.
The results of our analyses indicate a need for the provision of effective lifestyle interventions to the large number of adults with prediabetes as well as support research aimed at improving the effectiveness of these interventions. They also suggest a need for research designed to improve the effectiveness of population-wide approaches to diabetes prevention. Unless both strategies are much more effective than we assumed they will be in our analyses, the prevalence of diabetes among US adults will continue to increase even if the US health care system adopts something like the combined diabetes prevention strategy that we described. This continued increase in diabetes prevalence will result in an increase in demands for diabetes management and treatment, including services related to the prevention of diabetes complications and diabetes-related disability, and these services will need to be strengthened along with primary diabetes prevention efforts.
The findings and conclusions in this report are those of the authors and do not necessarily represent the official positions of the Centers for Disease Control and Prevention.