The Changing Tides of the Type 2 Diabetes Epidemic—Smooth Sailing or Troubled Waters Ahead? Kelly West Award Lecture 2016

Chronic disease epidemics have been described as tsunamis for the forceful way that societal and behavioral instigators carry disease and morbidity in a single direction before receding, only after decades of counter efforts by clinical medicine and public health (1). Type 2 diabetes is considered a classic chronic disease, as long-term cultural and behavioral changes have interacted with diverse susceptible genes to increase prevalence and extensive morbidity in the U.S. and the rest of world (2,3). While chronic disease epidemics can indeed be vast, multifaceted, and difficult to reverse, they can also be dynamic and full of paradoxes and contradictory trends that vary based on the specific population of concern, the time horizon, and the health metrics under consideration. The type 2 diabetes epidemic in the U.S. has encountered numerous phases and transitions in recent decades, including periods of rapid growth in prevalence and incidence, a peak and decline in the incidence of complications, the alarming growth among youth, and, most recently, a peak and apparent decrease in incidence of diagnosed cases in some populations. The most defining characteristics of the recent era of the global diabetes epidemic have been the success in reducing the rate of complications among adults with diagnosed diabetes in high-income countries, the large increases in overall prevalence, and the explosion of prevalence in lower- and middle-income countries (4). However, there are many currents under the surface, variations in the trends, and emerging concerns that are changing the character of the diabetes epidemic in ways that will affect the direction of public health efforts. This synthesis reviews the recent major trends and transitions in the diabetes epidemic in the U.S., describes their impact on the character of diabetes complications, and discusses the challenges and implications for prevention and epidemiology.

Most of the findings in this review are derived from published and ongoing analyses of the United States Diabetes Surveillance System (USDSS) (5). The earliest reports from the USDSS were in the 1990s, when data on diabetes-related complications were reported using hospitalization data, when prevalence was derived from the National Health Interview Survey (NHIS), and when the first comprehensive assembly of diabetes surveillance estimates was published in 1997 (6–8). The USDSS now assembles data from several national surveys and registries, including the National Health and Nutrition Examination Surveys (NHANES) as the core database for risk factor and disease prevalence and quality of care, NHIS for diabetes incidence and risk factors, the Behavior Risk Factor Surveillance System (BRFSS) for behaviors and state-level burden and care, the National Hospital Discharge Survey (NHDS) for diabetes-related complications, and registries for end-stage renal disease (ESRD) and diabetes in youth (5). USDSS monitors the diabetes problem in the U.S. from three dimensions—across levels of risk (e.g., prediabetes vs. diabetes) and severity of outcome; across geographic perspectives (national, state, and county); and across intervention target groups, including individual, provider, and system/policy levels. In addition, the system has fueled several lines of modeling efforts, including disease progression, lifetime risk, small area modeling, cost-effectiveness, and forecasting (9,10).

The mid-1990s were a special turning point in the diabetes epidemic, as several factors began to initiate changes. Population-based cohort studies had raised attention to the enormous burden of morbidity for people with diabetes, who were found to have microvascular complications 10–20 times and macrovascular complications 2–4 times that of people without diabetes (11–14). Major trials established the efficacy of glycemic, blood pressure, and lipid management to reduce risk of major macrovascular and microvascular complications (15). Surveillance systems brought attention to the growth of the diabetes epidemic (7,16) and revealed major gaps in care (17). These observations led diabetes to be labeled a public health problem (18), implying that the scope of the problem is beyond what a clinician and patient can manage on their own, and in turn stimulated efforts to improve health systems, increase national awareness and education, and improve the organization of care. A focus on improving the quality and organization of care became the primary vehicle of the public health response to diabetes in the late 1990s and early 2000s (19,20). By the mid-2000s, success in improving processes of care and reductions in national rates of selected diabetes-related complications were already apparent (21–25).

Reductions in a diverse spectrum of diabetes-related complications were captured in a 2014 study that assembled data from the NHDS, the United States Renal Disease Data System (USRDS), and the National Vital Statistics System (NVSS) (26). The study found substantial improvements across the full range of complications for adults with diagnosed diabetes, including acute myocardial infarction (AMI), stroke, lower-extremity amputation (LEA), ESRD, and death due to hyperglycemic crisis, which includes diabetic ketoacidosis and nonketotic hyperosmolar hyperglycemic syndrome (26) (Fig. 1). The greatest absolute and relative declines between 1995 and 2010 occurred for AMI, which declined 68% from 14.0 to 4.6 events per 1,000 per year, drawing even with rates of stroke. Rates of LEA also declined by about half, and the absolute magnitude of decline was smallest for ESRD, but still significant, at more than one-fourth. Deaths from hyperglycemic crisis declined by a relative magnitude similar to that of AMI but are much less common in absolute terms. More recent analyses of NHIS data also reveal similar magnitudes of reduction in all-cause mortality and specific causes of cardiovascular disease (CVD) death among the population with diagnosed diabetes (27–29) (Fig. 2). All-cause mortality rates declined over 40% in men and 25% in women with diabetes from 1990 to 2010, a rate of decline that was more than twice as steep as among those without diabetes. These reductions in complication rates in the diabetes population were so much greater than rates of decline in same-aged counterparts without diabetes that absolute differences between adults with and without diabetes declined by half and relative risks of LEA declined from 18.8 to 10.5, ESRD from 13.7 to 6.1, AMI from 3.8 to 1.8, and stroke from 3.1 to 1.5.

Despite these successes, many gaps in the understanding of macro-level trends in diabetes complications remain, including a lack of national data on retinopathy incidence and peripheral neuropathy and specific data on microvascular aspects of coronary disease, including contributors to congestive heart failure. These gaps limit understanding of whether improvements in complications are being driven primarily by better management of classic CVD risk factors or by glucose control, diabetes care, or early screening for complications. Furthermore, average rates can hide large variation between groups or obscure new problems. This is evident in the threefold variation in LEA rates across geographical areas and the 20 percentage point variation in quality of care across states in the U.S. (5,30).

Global trends in diabetes-related complications are much less well described. Existing studies suggest that there have been improvements in LEA (31) and CVD incidence and mortality (32–34), although such data come mostly from northern Europe, North America, Australia, and high-income Asian countries that collectively represent only a small fraction of the world. Lack of standardized data on diabetes-related complications makes it difficult to know global patterns and how much variation really exists. The few studies of diabetes-related complications from low- or middle-income countries, such as a recent report from Mexico showing relative risks considerably higher than those of high-income countries (35), raise the concern that successes in reducing diabetes complications are not shared by many parts of the world.

The large decline in complications in the U.S. were paralleled by large improvements in risk factor management and processes of care during the 2000s, as the proportion of adults meeting recommended targets improved by 9, 12, and 21 percentage points for glycemic control, blood pressure, and lipid control, respectively (36). Daily glucose monitoring improved by 13 percentage points, while levels of attention and adherence to lipid measurement, foot exams, and annual vaccination rates improved by 4 to 7 percentage points.

However, the bulk of long-term reductions in diabetes complications are likely the result of factors originating before the past decade, evolving from changes in multiple risk factors occurring in stages over several decades. These factors include the large reductions in smoking in the 1990s, the new emphasis on blood pressure and lipids in the 1980s, continued moderate improvements in lipids and blood pressure in the 1990s, and large improvements in lipids and to a lesser extent glycemic control in the 2000s (36–41).

Some of the decline in complications could be due to earlier identification of diabetes, leading to cohorts in later years who are at an earlier stage in the natural history of complications. The few analytic attempts to examine this bias suggest that this could lead to an overestimation of the decline in complications but that it does not explain the majority of this effect (26). In addition, the lack of change in age at diagnosis and the lack of data showing earlier age at diagnosis leave the magnitude of this bias unclear (5).

During the period of declining rates of diabetes complications, the prevalence and incidence of diagnosed diabetes increased in virtually all segments of the U.S. population (42). Although diabetes prevalence increased from the late 1950s to the mid-1990s, the most substantial increase in prevalence occurred between 1990 and 2008 (42), when it more than doubled in the general population and most subgroups (42). Absolute changes were greatest in the older population, while relative changes were greatest in the younger population. Although all racial/ethnic and socioeconomic groups were affected, the increases were most pronounced among people with lower levels of education, and there was a general widening of disparities between high-risk and low-risk counties (43). This period was also distinguished by an accentuation of a “diabetes belt,” characterized by broad clustering of counties of high prevalence in the Mississippi Valley, through the Deep South, across the Carolinas and the Appalachian region, and in American Indian reservations (44).

However, in 2014, the first plateau in rates of diabetes prevalence was noted nationally in a report using data from the NHIS (42). In the overall adult population, prevalence and incidence rates peaked in 2008 with no further increase in prevalence and an approximate 2 percentage point decline in incidence from about 9 to 7 cases per 1,000 adults in 2014 (Fig. 3). A subsequent data release and analysis 2 years later confirmed a continuation of the plateau in prevalence and the decrease in incidence (5). Although the decrease in incidence was not significant in all groups, the same general trends were observed in all major subsets of the population (Fig. 4).

Several potential explanations exist for the recent plateau in prevalence and downturn in incidence of diagnosed diabetes, including chance observation, biases related to detection, changes in diagnostic thresholds, or, hopefully, real changes in the risk and incidence for diabetes in the U.S. Several pieces of information argue against a chance observation. Most convincingly, analyses of state-specific trends from the BRFSS have revealed similar downturns in incidence in the vast majority of states and notable reductions in the number of counties experiencing increasing prevalence (45). This plateau has also been portrayed in geographic analyses showing that the number of counties experiencing increasing prevalence of diabetes, obesity, and physical inactivity has decreased while the number of counties with decreases in these indices is decreasing (Fig. 5) (45).

Ongoing public health efforts to raise awareness and education, scale diabetes prevention programs nationally, and reduce obesity, combined with encouraging trends in diabetes risk factors, provide hope that the trends in diabetes incidence are on a true downward trajectory (46,47). Several published reports suggest that these trends in incidence are indeed supported by changes in underlying risk factors. Levels of obesity prevalence in the U.S., which increased concurrently with diabetes over recent decades, reached a similar plateau in most age, sex, and racial/ethnic subgroups between 2008 and 2014 (48). Levels of total energy intake and intake of sugar-sweetened beverages reached a peak midway through the 2000s and declined thereafter (49–51). Similarly, between 2005 and 2012 there was a notable increase in the prevalence of U.S. adults meeting federal physical activity guidelines, before reaching another plateau between 2012 and 2015 (52). Despite these encouraging trends in risk factors, the levels of risk of the general population remain very high and there are many established risk factors, including qualitative aspects of the diet such as whole grains and fibers and healthy fats, for which trends in the at-risk population remain unclear.

Although the recent shifts in prevalence and incidence of diagnosed diabetes are robust, there are limitations and remaining questions about the degree to which these trends represent the true underlying trends in diabetes. First, the lack of population-level data about trends in case detection prevents quantification of the degree to which trends are affected by changes in awareness, screening, and diagnostic policies. The change in diabetes definition (adopting a reduced fasting plasma glucose [FPG] and dropping the oral glucose tolerance test as a required index) reduces apparent prevalence but could increase prevalence practically if the simplicity of FPG (vs. oral glucose tolerance test) stimulates more testing and diagnosis (53,54). Similarly, the 2010 adoption of HbA_1c to diagnose diabetes would decrease apparent prevalence if clinicians follow suit and use the HbA_1c in place of the FPG (55). During the period between 1997 and 2010, greater general awareness about diabetes may have increased testing and diagnosis; alternatively, greater awareness about the efficacy of lifestyle change for people with prediabetes could have conceivably led some clinicians to delay diagnosis in favor of encouraging lifestyle change. The epidemiologic data showing no change in average age at diagnosis further contribute to the ambiguity about trends in detection and underscore the need for development and application of methods to determine the underlying rate of disease (i.e., true diabetes) (56).

The epidemiologic profile of diabetes in the U.S, including multilevel impact of complications and care, along with encouraging recent trends in diabetes itself, raises the question of whether the overall burden of the diabetes epidemic is subsiding and whether future demands will lessen. However, macro-level trends can hide currents below the surface, including variation in risk factors, outcomes, population subgroups, and places that can have important implications for the character of the epidemic, future trends, and response that is needed.

Where improvements in diabetes complications have occurred, they have been most prominent among the older population. This is most apparent for hyperglycemic death and LEA, for which age-related differences have diminished dramatically; for AMI and stroke, age-related differences were reduced by half (26). Differential reduction by age means that, proportionally, middle-aged adults account for an increasing proportion of diabetes-related complications. For example, adults aged 45–64 years accounted for more than half of amputations in 2010 (vs. one-third in 1990), 44% of hyperglycemic deaths (vs. 32% in 1990), and almost one-third of stroke (vs. 20% in 1990). ESRD was the main exception to the change in age-related risk. Whether due to an increase in dialysis referral at a given level of renal failure or the reduction of CVD as a major competing risk, rates of ESRD decreased most in the young, leading to a reversal in the age relationship, such that ESRD rates are now higher in older adults.

A second undercurrent is the concerning trends in health among youth and young adults, including the Millennial generation (i.e., those born in 1980–2000). Recently published results from the SEARCH for Diabetes in Youth (SEARCH) study suggest that youth and adolescents aged 10–19 years have experienced increases in both type 1 and type 2 diabetes incidence from 2002 to 2012, with the most alarming increases seen in type 2 diabetes among American Indians and Hispanics (57). These trends in youth are ominous in light of the observation that youth with type 2 diabetes have about twice the prevalence of diabetes-related complications as same-aged youth with type 1 diabetes, as well as the implications for long-term risk given the development of diabetes so early in life (58). These findings are consistent with other concerning observations, including the continued increases in severe obesity and sedentary behavior in youth (59,60), the lack of improvements in blood pressure control, the high documented smoking rate, processes of care and risk factor improvements, and the lack of improvement in CVD mortality in young adults (36,39,40,61).

The third major current affecting diabetes complications is the relative persistence of microvascular complications. As described earlier, ESRD had weak improvement among people with diabetes in the U.S. and, when calculated as a rate per general population, doubled between 1995 and 2010 (26). National prevalence of stage 3 and 4 chronic kidney disease (CKD) has been flat since 2000, while the prevalence of any diabetes-related CKD has increased (62). These less encouraging trends could reflect several factors. First, they may be a function of the removal of competing risk, wherein the reduction in CVD mortality and concomitant improvements in medical management of CVD have permitted adults with type 2 diabetes to live long enough to develop CKD. The trends may also represent a weaker armamentarium, or less attention to the approaches that can reduce microvascular complications, including the fact that improvements in glucose levels occurred only long after improvements in other CVD risk factors. Data from Medicare beneficiaries revealed increases in both ESRD and retinopathy treatment between 1992 and 2012 (63,64). Similarly, the trends may signify a need to improve the care of people with CKD.

The encouraging reductions in diverse complications should not be interpreted as a complete win against the burden of diabetes-related complications in the U.S. In addition to the remaining 10-fold increased risk of amputation, 5-fold increased risk of ESRD, and 60% to 100% increased risk of AMI and stroke, the cumulative burden remains high for the average person with diabetes, as diabetes leads to substantial disability and a life span that is reduced by 5 years compared with the average person without diabetes (65,66). Moreover, the increase in diabetes prevalence that occurred between 1990 and 2008 continues to drive a large burden of diabetes-related complications. When rates of diabetes-related complications are calculated per 1,000 adults in the overall population (i.e., instead of per 1,000 adults with diabetes), the findings are not as encouraging; rates of AMI and hyperglycemic death declined by more than one-third from 1990 to 2010, but there was no change in rates of amputation or stroke, and rates of ESRD doubled (Fig. 6) (26). Examining trends from this perspective provides a picture of the diabetes trends in the U.S. population and is a reminder that reversing the public health burden of diabetes will require reducing the incidence of diabetes itself.

Long-term reduction in all-cause mortality occurring among adults with diagnosed diabetes represents a major clinical and public health success, yet the implications of this are likely to increase the public health burden in several ways for years to come (29). All-cause mortality rates declined over 40% in men and 25% in women with diabetes between 1990 and 2010. When combined with increased incidence during this time, this decline in mortality contributed to an increase in lifetime risk of diagnosed diabetes for both men and women, now estimated at 40%, along with a mean 3-year increase in years spent with diabetes for the average person diagnosed at age 60 years. When increases in years spent with diabetes were aggregated across the increased number of prevalent cases for a typical community, there was an approximate doubling of the cumulative number of years spent with diabetes.

The large increase in the number of years spent with diabetes will extend the opportunity for people with diabetes to develop classic complications, keeping the cumulative incidence and total burden high even if the yearly rate of complications per individual continues to decline. In addition, the increase in total years of diabetes, combined with the large reduction in CVD mortality, is likely to lead to a diversification in the complications of diabetes. This diversification has already been noted in the form of various cancers associated with diabetes, disability, cognitive decline, and geriatric syndromes (67,68). Recent analysis of nationally representative data from the Health and Retirement Study (HRS) indicates that diabetes leads to disability 7 years earlier than in the population without diabetes and that the impact on disability-free years is greater than the effect of diabetes on longevity (69). The diversification is also being seen in causes of mortality, as the large proportional reduction of CVD as a cause of mortality has been compensated for by an increase in all other causes of death, including those due to renal disease, infections, and accidents (28).

The recent trends and currents in the diabetes epidemic have led to expanded exposure and diversification of diabetes-related complications. This has several implications for descriptive epidemiology and surveillance, clinical effectiveness research, and translation research for public health action. The diversification of complications will expand the needed metrics for monitoring morbidity among patients with diabetes, including cancers, disability, and geriatric syndromes, as well as longitudinal assessment to monitor remission and recovery from morbidity.

The persistent problems of diabetes in youth in the U.S. population indicate a need for continued exploration of public health interventions to reduce incidence. Although trials of lifestyle interventions to prevent diabetes have led to successful translation in the form of community-based prevention programs, numerous modifiable risk factors for diabetes have been identified that remain unaddressed for diabetes prevention. Knowledge of such risk factors, including diverse qualitative aspects of the diet, physical activity, sleep, environmental toxins, depression, prenatal factors, and rapid weight gain early in life, has not been converted into practical, actionable interventions or used in effective health promotion. Overcoming this bottleneck will require new models of testing interventions, closer examination of population attributable risks, and the modeling of the health impact of diverse prevention approaches.

Finally, a stronger infrastructure for and use of natural experiments to test the health impact of population-wide policies and interventions on diabetes prevention and outcomes is needed to provide a more comprehensive, multitiered approach to confront the continuing problem of diabetes into the future (70).

This synthesis of epidemiologic findings related to the U.S. diabetes epidemic reveals several major trends and currents (Table 1). Rates of classic diabetes complications have declined in recent decades, representing a major success for people with diabetes as well as clinical and public health efforts that have driven these changes. Furthermore, after several decades of unabated increases, the prevalence of diagnosed diabetes in adults has reached a plateau and incidence has started to decline, indicating initial success in primary prevention efforts.

Despite these successes, several currents below the surface of the diabetes epidemic will present major challenges in the coming years. These currents include 1) the continued growth in the numbers of people with diabetes and continued increases in diabetes in youth, 2) a proportional shift in overall complications toward middle-aged adults, the relative persistence of microvascular disease, 3)diversification of diabetes-related morbidity following the reduction in CVD, 4) the continued high levels of diabetes complications in the overall population, and 5) the combined impact of increasing life span and high prevalence on the total years and cumulative morbidity burden in communities. These currents will all play major roles in the course of the diabetes epidemic in the future and will challenge all disciplines of epidemiology to continue to improve the monitoring and understanding of risk factors, risk stratification approaches, and the future prioritization of interventions.

Acknowledgments. The author thanks Linda Geiss, Mohammed K. Ali, and Elizabeth Luman (of the Centers for Disease Control and Prevention) for valuable comments on the manuscript.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.