Trends in Prevalence of Diabetes and Control of Risk Factors in Diabetes Among US Adults, 1999-2018

Key Points

Question  What were the trends in prevalence of diabetes and control of risk factors in diabetes among adults in the US from 1999-2000 to 2017-2018?

Findings  In this serial, cross-sectional study of nationally representative data from 28 143 participants in the National Health and Nutrition Examination Survey (NHANES), the estimated age-standardized prevalence of diabetes increased significantly, from 9.8% in 1999-2000 to 14.3% in 2017-2018. Only 21.2% of adults with diagnosed diabetes achieved all 3 risk factor control goals in 2015-2018, including individualized hemoglobin A_1c targets, blood pressure less than 130/80 mm Hg, and low-density lipoprotein cholesterol level less than 100 mg/dL.

Meaning  Based on NHANES data from US adults, the estimated prevalence of diabetes increased significantly between 1999-2000 and 2017-2018, and only an estimated 21% of adults with diagnosed diabetes achieved all 3 risk factor control goals in 2015-2018.

Importance  Understanding population-wide trends in prevalence and control of diabetes is critical to planning public health approaches for prevention and management of the disease.

Objective  To determine trends in prevalence of diabetes and control of risk factors in diabetes among US adults between 1999-2000 and 2017-2018.

Design, Setting, and Participants  Ten cycles of cross-sectional National Health and Nutrition Examination Survey (NHANES) data between 1999-2000 and 2017-2018 were included. The study samples were weighted to be representative of the noninstitutionalized civilian resident US population. Adults aged 18 years or older were included, except pregnant women.

Exposures  Survey cycle.

Main Outcomes and Measures  Diabetes was defined by self-report of diabetes diagnosis, fasting plasma glucose level of 126 mg/dL or more, or hemoglobin A_1c (HbA_1c) level of 6.5% or more. Three risk factor control goals were individualized HbA_1c targets, blood pressure less than 130/80 mm Hg, and low-density lipoprotein cholesterol level less than 100 mg/dL. Prevalence of diabetes and proportion of adults with diagnosed diabetes who achieved risk factor control goals, overall and by sociodemographic variables, were estimated.

Results  Among the 28 143 participants included (weighted mean age, 48.2 years; 49.3% men), the estimated age-standardized prevalence of diabetes increased significantly from 9.8% (95% CI, 8.6%-11.1%) in 1999-2000 to 14.3% (95% CI, 12.9%-15.8%) in 2017-2018 (P for trend < .001). From 1999-2002 to 2015-2018, the estimated age-standardized proportion of adults with diagnosed diabetes who achieved blood pressure less than 130/80 mm Hg (P for trend = .007) and low-density lipoprotein cholesterol level less than 100 mg/dL (P for trend < .001) increased significantly, but not individualized HbA_1c targets (P for trend = .51). In 2015-2018, 66.8% (95% CI, 63.2%-70.4%), 48.2% (95% CI, 44.6%-51.8%), and 59.7% (95% CI, 54.2%-65.2%) of adults with diagnosed diabetes achieved individualized HbA_1c targets, blood pressure less than 130/80 mm Hg, and low-density lipoprotein cholesterol level less than 100 mg/dL, respectively. Only 21.2% of these adults (95% CI, 15.5%-26.8%) achieved all 3. During the entire study period, these 3 goals were significantly less likely to be achieved among young adults aged 18 to 44 years (vs older adults ≥65 years: estimated proportion, 7.4% vs 21.7%; adjusted odds ratio, 0.32 [95% CI, 0.16-0.63]), non-Hispanic Black adults (vs non-Hispanic White adults: estimated age-standardized proportion, 12.5% vs 20.6%; adjusted odds ratio, 0.60 [95% CI, 0.40-0.90]), and Mexican American adults (vs non-Hispanic White adults: estimated age-standardized proportion, 10.9% vs 20.6%; adjusted odds ratio, 0.48 [95% CI, 0.31-0.77]).

Conclusions and Relevance  Based on NHANES data from US adults, the estimated prevalence of diabetes increased significantly between 1999-2000 and 2017-2018. Only an estimated 21% of adults with diagnosed diabetes achieved all 3 risk factor control goals in 2015-2018.

Diabetes is a leading cause of disability and mortality.^1,2 In the US, its total economic cost was $327 billion in 2017 and care for people with diabetes accounted for 24% of all health care dollars.³ The estimated prevalence of diabetes among US adults increased from 5.3% in 1976-1980 to 11.5% in 2011-2014, increasing at a rate faster than the global increase during the same period.^2,4,5 The prevalence of diabetes varies substantially by sociodemographic variables and weight status.^6-8 Also, it has been associated with increased risk of cardiovascular disease.⁹Quiz Ref ID Cardiovascular disease prevention in diabetes requires appropriate management of well-established risk factors such as hemoglobin A_1c (HbA_1c) level, blood pressure (BP), and serum cholesterol level.^10,11 An improvement in the control of the 3 risk factors (HbA_1c <7.0% or individualized HbA_1c targets, BP <130/80 mm Hg, and low-density lipoprotein cholesterol [LDL-C] level <100 mg/dL) was observed between 1988 and 2010, but only an estimated 18.8% of US adults achieved all 3 goals in 2007-2010.^12,13 It is unclear whether trends in control of risk factors in diabetes have changed during the past decade. This information is critical to forecasting population-level complications and guiding prevention efforts.

The primary objective of this study, which used recently released data from the National Health and Nutrition Examination Survey (NHANES), was to provide updated national estimates to evaluate trends in prevalence of diabetes and control of risk factors in diabetes among US adults between 1999-2000 and 2017-2018.

Beginning in 1999, NHANES has been a continuous, multistage, nationally representative survey of the noninstitutionalized civilian resident US population. Data collected through in-home interviews and study visits at mobile examination centers have been released in 2-year cycles. This study included 10 cycles between 1999-2000 and 2017-2018. The overall response rates ranged from 52% to 84% for the interview component and from 49% to 80% for the examination component. Participants aged 18 years or older were included. Pregnant women were excluded. The National Center for Health Statistics Research Ethics Review Board approved NHANES. Written informed consent was obtained from all adult participants. Shanghai Jiao Tong University School of Medicine Public Health and Nursing Research Ethics Review Committee approved this study.

Information on age, sex, race, ethnicity, statin use, and medical conditions was collected during household interview. Race and ethnicity were self-reported according to fixed-category questions and included because of the known racial and ethnic differences in prevalence of diabetes. Weight, height, waist circumference, and BP were measured in mobile examination centers with standard protocols. Body mass index was computed by dividing weight in kilograms by height in meters squared. The mean of all available BP measurements was used to calculate systolic and diastolic BP. HbA_1c was measured and standardized to the Diabetes Control and Complications Trial method. A random subset of the participants was sampled to attend the morning session, during which fasting plasma glucose and LDL-C levels were measured among those who fasted for 8 to less than 24 hours. The Centers for Disease Control and Prevention’s Lipid Standardization Program was applied to ensure accuracy and precision of measurements between laboratories and over time. An oral glucose tolerance test using a 75-g glucose challenge was administered and 2-hour plasma glucose level was measured. Urine albumin and creatinine levels were measured with a fluorescent immunoassay and Jaffe rate reaction method, respectively. Serum creatinine level was measured with a Jaffe kinetic rate method. Glucose data between 2005-2006 and 2017-2018 were calibrated according to the recommended method by the National Center for Health Statistics to account for changes to the laboratory method, equipment, or site.¹⁴ Urine and serum creatinine levels were also calibrated. Calibrating HbA_1c was not required. Estimated glomerular filtration rate was computed according to the Chronic Kidney Disease Epidemiology Collaboration equation.¹⁵

Diagnosed diabetes was defined as self-report of diabetes diagnosis by a physician or other health professional. Undiagnosed diabetes was defined as having a fasting plasma glucose level of 126 mg/dL or more or HbA_1c level of 6.5% or more among individuals without diagnosed diabetes. Diabetes included both diagnosed and undiagnosed diabetes.

Clinical risk factors commonly targeted to decrease cardiovascular disease risk in diabetes are HbA_1c level, BP, and serum cholesterol level.^10-12 Low-density lipoprotein cholesterol was selected as the cholesterol control target because statins are the first-line treatment for lipid lowering.¹¹ Individualized HbA_1c targets have been emphasized for glycemic control,¹⁰ and were specified as follows: less than 6.5% for young adults aged 18 to 44 years without complications, less than 7.0% for both young adults with complications and middle-aged adults aged 45 to 64 years without complications, less than 8.0% for both middle-aged adults and older adults aged 65 years or older with complications, and less than 7.5% for older adults without complications.^10,16 Complications were defined as any of the following: having self-reported cardiovascular disease (congestive heart failure, coronary heart disease, heart attack, or stroke), retinopathy, urine albumin to creatinine ratio of 30 mg/g or higher, or estimated glomerular filtration rate less than 60 mL/min/1.73 m².¹³ Guidelines recommend less than either 130/80 or 140/90 mm Hg for BP control in diabetes.^11,17 Achieved systolic BP of 130 mm Hg or lower has been associated with a lower risk of cardiovascular disease.¹⁸Quiz Ref ID Lipid management in diabetes has shifted from controlling LDL-C level to less than 100 mg/dL to prescribing statins with different intensity according to age and presence of atherosclerotic cardiovascular disease and risk factors without a specific LDL-C target.^11,19 However, intensity of statin treatment was not collected in NHANES. Therefore, LDL-C level less than 100 mg/dL was used as the primary target.

Risk factor control analysis was conducted among adults with diagnosed diabetes. Primary outcomes included prevalence of diabetes and proportion of adults with diagnosed diabetes who achieved 3 risk factor control goals individually and collectively: individualized HbA_1c targets, BP less than 130/80 mm Hg, and LDL-C level less than 100 mg/dL. Secondary outcomes included prevalence of diagnosed and undiagnosed diabetes, percentage of diabetes that was undiagnosed, and proportion of adults with diagnosed diabetes who achieved HbA_1c level less than 7% or less than 8%, BP less than 140/90 mm Hg, and LDL-C level less than 70 mg/dL and who took statins. Factors associated with achieving risk factor control goals were also assessed.

The NHANES 2017-2018 cycle was used to estimate prevalence of diagnosed diabetes, undiagnosed diabetes and diabetes, and percentage of diabetes that was undiagnosed, overall and by age (18-44, 45-64, and ≥65 years), sex (men and women), race and ethnicity (non-Hispanic White, non-Hispanic Black, Hispanic overall, Mexican American as a separate Hispanic subgroup, non-Hispanic Asian, and other), education (less than high school, high school graduate, some college, and college graduate or more), body mass index level (18.5-24.9, 25-29.9, 30-34.9, 35-39.9, and ≥40), abdominal obesity (yes/no), and insurance status (uninsured and insured). Mexican American persons were oversampled before 2007 and all Hispanic persons were oversampled from 2007 onward in NHANES. The National Center for Health Statistics recommends not calculating estimates for all Hispanic persons for survey cycles before 2007 and for any Hispanic subgroup other than Mexican American in any survey cycle through 2018.¹⁴ Thus, for analyses including data before 2007, results for Mexican American adults instead of all Hispanic adults were reported; otherwise, results for all Hispanic adults and Mexican American adults were reported. Similarly, non-Hispanic Asian subgroup was not available before 2011 due to the survey design. Abdominal obesity was defined as having waist circumference greater than 102 or 88 cm in men or women, respectively. Participants with body mass index less than 18.5 were included in the analyses in the total sample and stratified by other variables. Estimates were age standardized to the 2017-2018 NHANES nonpregnant adult population, using the age groups 18 to 44 years, 45 to 64 years, and 65 years or older.

Joinpoint regressions with heteroscedastic and uncorrelated error were used to determine trends in log-transformed age-standardized prevalence, allowing 1 joinpoint.²⁰ The joinpoint location, if it existed, was identified with a grid search. The best-fitting model was selected by conducting 4499 permutation tests based on a Monte Carlo method, adjusting for multiple tests. Parameters were estimated with weighted least squares, with weights proportional to the inverse of the variance of prevalence at each cycle. Under a log-linear model, prevalence changed at a constant relative percentage per cycle, facilitating comparisons across strata with varying prevalence. Relative percentage change per 2-year cycle in prevalence and its 95% CI were obtained.

Two adjacent NHANES cycles were combined to estimate prevalence of achieving risk factor control goals. Estimates were age standardized to all adults with diagnosed diabetes in 1999-2018. Distribution of HbA_1c levels, BP, LDL-C levels, and statin use was displayed and differences over time were assessed with Rao-Scott χ² tests. Stratified analyses according to previously described subgroups and presence of complications (yes/no) were conducted for each goal. Changes over time were determined by comparing each of the estimates from earlier years with the estimate from 2015-2018, using t tests. An overall trend during the entire period was assessed with F tests. Proportion of adults achieving all 3 risk factor control goals by subgroup was assessed among the total study sample of diagnosed diabetes combined from all years to increase robustness of results. Factors associated with achieving risk factor control goals were identified with logistic models, with all of the aforementioned categorical variables included as covariates. Odds ratios (ORs) and 95% CIs were obtained. Key assumptions for logistic models included binary outcome structure, independence of errors, absence of high multicollinearity, linearity between covariates and log odds, lack of influential outliers, and having at least 10 events per covariate.

A sensitivity analysis was conducted by additionally including undiagnosed diabetes cases based on a 2-hour plasma glucose level of 200 mg/dL or more, available between 2005 and 2016 at this analysis.

Weights for the interview sample, examination sample, fasting subsample, and oral glucose tolerance test subsample were used appropriately to ensure the estimates were representative of the total civilian noninstitutionalized US population. Weights were adjusted for nonresponse, noncoverage, and unequal probabilities of selection. Weights and design variables were included to obtain unbiased estimates and SEs. Complete case analysis was applied if missing data level for primary analyses was 10% or less. Data were analyzed with SAS version 9.4 and Joinpoint Regression Program version 4.8.0.1.²⁰ A 2-sided P < .05 was used to determine statistical significance. Because of the potential for type I error due to multiple comparisons, findings for secondary analyses and secondary outcomes should be interpreted as exploratory.

Among the 28 143 participants included, the weighted mean age was 48.2 years (SE, 0.2), and 49.3% were men, 68.0% non-Hispanic White, 11.3% non-Hispanic Black, and 8.0% Mexican American; all were weighted proportions. For analyzing trends in prevalence of diabetes, included were 28 143 participants from the interview sample and 27 837 from the examination sample; 27 508 had data for HbA_1c level, 23 622 for fasting plasma glucose level, and 11 526 for 2-hour plasma glucose level. Missing data were found for education (n = 1), insurance (n = 127 [0.5%]), body mass index (n = 537 [1.9%]), and waist circumference (n = 1276 [4.6%]). For risk factor control analysis, 6678, 6372, and 2632 participants with diagnosed diabetes were included from the interview sample, examination sample, and fasting subsample, respectively; missing data were found for insurance (n = 27 [0.4%]), HbA_1c (n = 320 [5.0%]), BP (n = 295 [4.6%]), urine albumin to creatinine ratio and estimated glomerular filtration rate (n = 99 [1.6%]), body mass index (n = 228 [3.6%]), waist circumference (n = 574 [9.0%]), and LDL-C (n = 181 [6.9%]). The estimated proportion of uninsured adults increased significantly from 17.3% in 1999-2000 to 19.4% in 2011-2012 (P for trend = .02) and decreased significantly to 13.9% in 2017-2018 (P for trend = .04).

The estimated unadjusted prevalence was 11.2% for diagnosed diabetes (95% CI, 9.8%-12.5%), 3.4% for undiagnosed diabetes (95% CI, 2.5%-4.3%), and 14.6% for diabetes (95% CI, 12.8%-16.3%) (Table 1). Overall, 23.3% of adults with diabetes (95% CI, 18.6%-28.1%) were undiagnosed. The estimated age-standardized prevalence of diagnosed diabetes, undiagnosed diabetes, and diabetes was significantly higher in older than younger adults, adults identifying as a member of a racial or ethnic minority group than non-Hispanic White adults, people with lower than higher education level (except for undiagnosed diabetes), and people with greater than lower body mass index and waist circumference range. No significant difference by sex and insurance status was identified. The estimated percentage of diabetes that was undiagnosed was significantly higher in younger than older adults. Among young adults, 40.0% (95% CI, 28.4%-51.5%) of participants with diabetes were undiagnosed.

The estimated age-standardized prevalence of diabetes increased significantly from 9.8% (95% CI, 8.6%-11.1%) in 1999-2000 to 14.3% (95% CI, 12.9%-15.8%) in 2017-2018, with a 3.3% relative increase (95% CI, 2.2%-4.5%) per 2-year cycle (Table 2; Figure, A). A significant increase in the estimated age-standardized prevalence of diabetes was observed in young and middle-aged adults, men, women, non-Hispanic White adults, Mexican American adults, adults with overweight or abdominal obesity, and insured adults, as well as among all education levels (all P for trend < .05). The estimated age-standardized prevalence of diagnosed diabetes increased significantly from 6.8% (95% CI, 5.7%-7.9%) in 1999-2000 to 11.0% (95% CI, 9.8%-12.1%) in 2017-2018, with a 5.2% relative increase (95% CI, 4.3%-6.0%) per cycle (eTable 1 in the Supplement). The estimated age-standardized prevalence of undiagnosed diabetes was not significantly different across cycles (P for trend = .32) (eTable 2 in the Supplement). The estimated age-standardized percentage of diabetes that was undiagnosed decreased significantly from 31.0% (95% CI, 26.1%-36.0%) in 1999-2000 to 23.3% (95% CI, 18.9%-27.8%) in 2017-2018, with a –4.6% relative decrease (95% CI, –6.8% to –2.3%) per cycle (eTable 3 in the Supplement; Figure, B), but this significant decrease was not observed in several subgroups, including young adults and minority adults (all P for trend >.05). Detailed subgroup results are shown in eTables 1 to 3 in the Supplement.

Among adults with diagnosed diabetes, the estimated age-standardized distribution of HbA_1c, BP, and LDL-C levels improved significantly, with major improvement occurring between 1999-2002 and 2003-2006 (all P < .05) (eFigure, A-C in the Supplement). The estimated age-standardized prevalence of statin use increased significantly (P < .001) (eFigure, D in the Supplement). The estimated age-standardized proportion of adults with diagnosed diabetes who achieved individualized HbA_1c targets did not increase significantly overall and among all subgroups (all P for trend > .05) except young adults without complications (P for trend = .03) (Table 3). The estimated age-standardized proportion of adults with diagnosed diabetes who achieved BP less than 130/80 mm Hg and LDL-C level less than 100 mg/dL increased significantly (both P for trend < .05). There was a significantly higher proportion of adults with diagnosed diabetes who achieved individualized HbA_1c targets in 2015-2018 than 1999-2002 (66.8% vs 58.9% [95% CI, 63.2%-70.4% vs 54.4%-63.3%]), but no significant difference was found for estimates between 2003 and 2018 (all P > .05). There was a significantly higher proportion of individuals achieving BP less than 130/80 mm Hg in 2015-2018 than 1999-2002 (48.2% vs 38.5% [95% CI, 44.6%-51.8% vs 33.6%-43.5%]), but no significant difference was found for estimates between 2003 and 2018 (all P > .05). There was a significantly higher proportion of individuals achieving LDL-C level less than 100 mg/dL in 2015-2018 than 1999-2002 and 2003-2006 (59.7% vs 35.4% and 46.9% [95% CI, 54.2%-65.2% vs 27.2%-43.6% and 40.8%-52.9%], respectively), but no significant difference was found for estimates between 2007 and 2018 (all P > .05). There was a significantly higher proportion of individuals achieving all 3 goals in 2015-2018 than 1999-2002 (21.2% vs 9.0% [95% CI, 15.5%-26.8% vs 5.0%-13.1%]), but no significant difference was found for estimates between 2003 and 2018 (all P > .05). Subgroup estimates for primary and secondary goals are shown in eTables 4 to 10 in the Supplement.

All model assumptions were met and logistic models converged successfully. During the entire study period, compared with older adults, young adults were significantly less likely to achieve individualized HbA_1c targets (43.5% vs 79.8%; adjusted OR, 0.25 [95% CI, 0.19-0.33]), LDL-C level less than 100 mg/dL (41.3% vs 63.5%; adjusted OR, 0.40 [95% CI, 0.27-0.58]), and all 3 goals (7.4% vs 21.7%; adjusted OR, 0.32 [95% CI, 0.16-0.63]), but significantly more likely to achieve BP less than 130/80 mm Hg (56.5% vs 37.4%; adjusted OR, 1.99 [95% CI, 1.52-2.60]) (Table 4, eTable 11 in the Supplement). Compared with non-Hispanic White adults, non-Hispanic Black adults were significantly less likely to achieve individualized HbA_1c targets (60.4% vs 68.3%; adjusted OR, 0.64 [95% CI, 0.53-0.77]), BP less than 130/80 mm Hg (38.7% vs 48.5%; adjusted OR, 0.65 [95% CI, 0.55-0.78]), and all 3 goals (12.5% vs 20.6%; adjusted OR, 0.60 [95% CI, 0.40-0.90]). Compared with non-Hispanic White adults, Mexican American adults were significantly less likely to achieve individualized HbA_1c targets (55.7% vs 68.3%; adjusted OR, 0.59 [95% CI, 0.47-0.73]), LDL-C level less than 100 mg/dL (43.4% vs 56.6%; adjusted OR, 0.64 [95% CI, 0.47-0.87]), and all 3 goals (10.9% vs 20.6%; adjusted OR, 0.48 [95% CI, 0.31-0.77]). Other subgroup estimates for primary goals and results for secondary goals are shown in Table 4 and eTable 11 in the Supplement.

The estimated age-standardized prevalence of diabetes based on the definition including 2-hour plasma glucose level did not increase significantly between 2005-2006 and 2015-2016 (relative percentage change per 2-year cycle, 1.1% [95% CI, –2.3% to 4.6%]; P for trend = .44) (eTable 12 in the Supplement). If it was restricted to the same period without including 2-hour plasma glucose level, there was a significant increase (relative percentage change per 2-year cycle, 3.1% [95% CI, 0.02%-6.3%]; P for trend = .049).

Quiz Ref IDThe estimated prevalence of diabetes among US adults increased significantly between 1999-2000 and 2017-2018. The estimated proportion of adults with diagnosed diabetes who achieved individualized HbA_1c targets and BP less than 130/80 mm Hg in 2015-2018 was significantly higher than that in 1999-2002, but not in 2003-2014. The estimated proportion of adults with diagnosed diabetes who achieved LDL-C level less than 100 mg/dL in 2015-2018 was significantly higher than that in 1999-2006, but not in 2007-2014. Only an estimated 21% of adults with diagnosed diabetes achieved all 3 risk factor control goals in 2015-2018.

Quiz Ref IDThis work extends prior findings by providing the most updated estimates and characterizing previously unreported subgroups (eg, adults with abdominal obesity).^6-8,12,13 The significant increasing trends in the estimated prevalence of diabetes may in part be a collective product of improved survival in diabetes²¹; increasing burden of diabetes among children and young adults²²; more widespread screening for diabetes, particularly after the implementation of the Patient Protection and Affordable Care Act²³; increasing body mass index and waist circumference²⁴; and decreasing incidence of diagnosed diabetes among US adults.⁵ The estimated prevalence of diabetes continued to increase significantly among subgroups disproportionately affected by diabetes, including Mexican American adults and those with abdominal obesity. Underdiagnosis was common and the estimated prevalence of undiagnosed diabetes did not decrease significantly over time. The estimated percentage of diabetes that was undiagnosed decreased significantly over time, which may be due to better screening and survival. However, among all racial/ethnic subgroups, this significant decrease in the estimated percentage of diabetes that was undiagnosed was observed only in non-Hispanic White adults, which may be attributed in part to higher insurance coverage and more preventive services compared with that for minority adults.^25,26

Quiz Ref IDSimilar to that for young adults in many other parts of the world,²² the burden of diabetes among US young adults has been increasing. Compared with later-onset diabetes, young-onset diabetes appeared to be associated with worse glycemic control, progressed to adverse cardiometabolic risk profiles more rapidly, and had greater lifetime risk of vascular and nonvascular complications.²² In this study, young adults were significantly less likely than older ones to achieve individualized HbA_1c targets, LDL-C level less than 100 mg/dL, and all 3 goals combined. Accordingly, early detection and management of diabetes among young adults is critical, but the estimated percentage of diabetes that was undiagnosed remained high and unchanged during the previous 2 decades.

The improvement in risk factor control reported before 2010 did not continue despite extensive public health investments, as well as advances in therapeutic management of diabetes in the past 2 decades.^12,13 Similar to that in previous reports, greater improvement was observed for cholesterol control than glycemic and BP control. Only a small estimated proportion of adults with diagnosed diabetes achieved all 3 risk factor control goals, and stagnation in risk factor control occurred in 2003-2018, although the treatment goals are theoretically achievable via pharmacologic and lifestyle therapies for most people and cardiovascular risk factor control has been emphasized in guidelines.^10,11,27 Significant racial and ethnic differences in risk factor control existed. Reasons abound for poor risk factor control, but challenges lie in designing effective tailored approaches for improving adherence to medications and healthy lifestyle behaviors, as well as providing necessary health care access and resources, education, and self-management support for improving adherence and maintaining achieved adherence.^12,28

This study has several limitations. First, misclassification of diabetes was possible because of the use of self-reported diagnosis and reliance on single-occasion laboratory measurement. Repeating the same laboratory test on another day or performing a different test without delay with a new blood sample has been recommended to confirm a diabetes diagnosis unless there is a clear clinical diagnosis.²⁹ Second, oral glucose tolerance test data were available only between 2005 and 2016. Furthermore, intraindividual variability of 2-hour plasma glucose level (16.7%) was higher than that of fasting plasma glucose level (5.7%) and HbA_1c level (3.6%).³⁰ Thus, the primary definition of diabetes did not include 2-hour plasma glucose level. Third, a small shift in the distribution of HbA_1c data in 2007-2010 was identified by the National Center for Health Statistics, but the reason was unclear despite intensive investigations. Fourth, results from the analysis of risk factor control did not necessarily apply to all adults with diabetes. Individualized cholesterol control goals were not studied because intensity of statin therapy was not collected in NHANES. However, a number of risk factor control goals were considered to facilitate the understanding of the general risk factor control among US adults with diagnosed diabetes.

Based on NHANES data from US adults, the estimated prevalence of diabetes increased significantly between 1999-2000 and 2017-2018. Only an estimated 21% of adults with diagnosed diabetes achieved all 3 risk factor control goals in 2015-2018.

Corresponding Authors: Victor W. Zhong, PhD, School of Public Health, Shanghai Jiao Tong University School of Medicine, 415 East No. 1 Bldg, 227 South Chongqing Rd, Shanghai 200025, China (wenze.zhong@shsmu.edu.cn); Hui Wang, MD, PhD, Shanghai Jiao Tong University School of Medicine, 227 South Chongqing Rd, Shanghai 200025, China (huiwang@shsmu.edu.cn).

Accepted for Publication: May 31, 2021.

Published Online: June 25, 2021. doi:10.1001/jama.2021.9883

Author Contributions: Dr Zhong had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: L. Wang, Li, Carnethon, Greenland, H. Wang, Zhong.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: L. Wang, Zhong.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: L. Wang, Zhong.

Administrative, technical, or material support: Li, Carnethon, H. Wang, Zhong.

Supervision: Greenland, H. Wang, Zhong.

Discussion with the lead author (and study team) on what should be the ultimate focus of the article and conclusions: Bancks.

Conflict of Interest Disclosures: Dr Bancks reported receiving grants from the National Institutes of Health (NIH) as a coinvestigator with multiple unrelated epidemiologic studies and clinical trials and receiving funding from the NIH for his work with these cohorts, outside the submitted work. Dr Greenland reported receiving grants from NIH and the American Heart Association, outside the submitted work. No other disclosures were reported.

Funding/Support: This study was supported by the Program for Young Eastern Scholar at Shanghai Institutions of Higher Education (QD2020027), the National Key R&D Program of China (2018YFC2000700), the National Natural Science Foundation of China (82030099, 81630086), and the Shanghai Public Health System Construction Three-Year Action Plan (GWV-10.1-XK15).

Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Disclaimer: Dr Greenland, a JAMA senior editor, was not involved in the review of or decision to publish this article.

Additional Contributions: We thank Danita D. Byrd-Clark, BBA, Social & Scientific Systems, for assisting with the SAS program development for specific analyses related to definition of diabetes and calibration of laboratory measurements across NHANES cycles. Ms Byrd-Clark did not receive financial compensation for her contribution.