Discussion
Following an almost two-decade steady increase in national prevalence and incidence of diagnosed diabetes, we report an 8-year period of stable prevalence and a decrease in incidence. The decrease in incidence of diagnosed cases brings the rate to the same levels as the year 2000, representing a 35% decrease since the peak year of 2009. Subgroup analyses indicate that all age, sex, and race/ethnic groups have similarly flat trends in prevalence. The overall decrease in incidence appears to be dominated by the non-Hispanic white population. These analyses build on a prior report through 2012 data showing a plateau in prevalence and incidence but no significant decrease.1
The NHIS is the nation’s largest nationally representative health survey to assess diabetes incidence and provides a compelling indication that there may be success in the efforts to stem the US diabetes epidemic. These findings are also generally consistent with estimates from the BRFSS, where many states have also observed flat prevalence and decreasing incidence.2 The specific causes of the large decrease cannot be determined from these analyses. Efforts continue to reduce type 2 diabetes through the National Diabetes Prevention Program lifestyle change intervention, which includes targeted screening, as well as population approaches to improve healthy food availability, diabetes awareness and education, and walkability of communities.6 Trends in several risk factors for type 2 diabetes, including intake of added sugar, sugared beverages, total calories, and physical inactivity, peaked in the mid-2000s and either plateaued or decreased thereafter,7 8 consistent with the slowing in diabetes incidence. However, although these diet and activity-level changes are encouraging, their association with diabetes incidence trends is only ecological and no causal inference can be made at present. Furthermore, obesity and severe obesity trends have generally increased over the past 10 years and pre-diabetes remains unchanged and high, affecting 84 million US adults, or 34% of the US adult population.3 9
Given the reliance on diagnosed cases for diabetes prevalence and incidence in our analyses and the persistence of type 2 diabetes risk factors, the potential impact of changes in screening, testing, and diagnostic thresholds are also important considerations. A change in the American Diabetes Association (ADA)-based definition of diabetes in 1997 lowered the fasting glucose level and may have contributed to part of the increase in the late 1990s. Similarly, the 2010 ADA-based recommendation to use HbA1c in the diagnosis of diabetes could decrease incidence because it is a less sensitive indicator than the fasting glucose threshold, or alternatively, could increase incidence if its practicality leads to increased awareness and detection. Unfortunately, practices of screening, detection, and diagnosis of diagnosed diabetes are variable and not monitored well by national surveillance systems.
The disassociation of prevalence and incidence trajectories observed here is consistent with the large reductions in cardiovascular and all-cause mortality recently reported among adults with diagnosed diabetes.10 Extended lifespans of persons after they are diagnosed with diabetes likely contribute to the lack of reduction in prevalence. Saturation of diagnosed diabetes is also an important consideration.11 An increased emphasis on screening may have decreased the susceptible population, driving down incidence. However, according to the Centers for Disease Control and Prevention’s 2017 National Diabetes Statistic Report, in 2015, twenty-four percent of persons with diabetes were undiagnosed.9 Although this proportion may be inflated due to error in self-report,12 it still suggests that saturation has not been reached.
In addition to the reliance on self-reported diagnosed diabetes, our report is limited by an inability to reliably distinguish type 1 from type 2 diabetes. Also, limited statistical power prevented reliable estimates of trends among young adults and minority racial/ethnic groups to determine whether increasing incidence of both type 1 and type 2 diabetes observed in the SEARCH for Diabetes in Youth study is also affecting young adults.13 We did find a significantly increasing trend in incidence among Hispanic adults during the study period. Similarly, our estimates may obscure differential trends in geography and social class. Incidence has been flat in both the less than high school and more than high school educated populations for over a decade, but the less than high school educated population still has roughly double the incidence of diabetes. Although our plotted trends are consistent with the scenario in which non-Hispanic whites are driving the reduction in incidence, other race/ethnic populations could be affected by differences in detection.
In summary, using these national data, we report the longest period of a sustained plateau in prevalence since the 1980s and longest period of declining incidence ever. Although an encouraging sign of success, due to the persistence of major risk factors such as obesity and pre-diabetes, we caution that trends are likely affected by changing awareness, detection, and diagnostic practices. Even in the event of true reductions in incidence, the high prevalence and declining mortality signifies a continued high overall burden of diabetes. For these reasons, we urge a continued emphasis on multilevel, multidisciplinary prevention to reduce both type 2 diabetes and diabetes complications, along with improved surveillance of trends in screening and detection.