Association of Cardiometabolic Multimorbidity and Depression With Cardiovascular Events in Early-Onset Adult Type 2 Diabetes: A Multiethnic Study in the U.S.

Recent evidence from epidemiological studies has shown an increase in the incidence and prevalence of type 2 diabetes diagnosed at an earlier age, commonly defined as early-onset type 2 diabetes (1,2). Although definitions vary, these studies have indicated differences in the pathophysiological and phenotypical characteristics of patients with early-onset type 2 diabetes and a different risk profile of complications and mortality risk (2,3). In particular, early-onset diabetes is associated with a faster decline over time of β-cell function, exacerbated by higher rates of obesity and associated comorbidities, and suboptimal self-care behaviors, resulting in a quicker deterioration of glucose control compared with individuals diagnosed at an older age (4,5).

Among the possible risk factors for an earlier diagnosis, ethnicity has been suggested to play a role. Certain ethnicities, including African Americans (AAs), are at greater risk of type 2 diabetes than White Caucasians (WCs). According to the most recent U.S. Centers for Disease Control and Prevention report (data from 2013 to 2016), the age-adjusted prevalence (95% CI) of diabetes among adult non-Hispanic Blacks (16.8 [15.4–18.1]) was significantly higher compared with White counterparts (10.0 [9.2–11.0]) (6). However, these individuals are also further disproportionately represented in early-onset adult type 2 diabetes. While the causes of the racial/ethnic differences in incidence of type 2 diabetes as well as further disparities by age are not well understood, a nontraditional risk factors, depression, has been identified as a possible risk factor, although its role in the risk of complications, such as cardiovascular disease (CVD), is not clear (7–10).

A recent U.K. primary care electronic medical record (EMR)–based study reported a higher risk of cardiovascular events and mortality in people diagnosed with type 2 diabetes at <50 years of age compared with others with type 2 diabetes diagnosed later in life, irrespective of their cardiometabolic risk factor level at diagnosis (2). This study also reported that the proportion of people diagnosed with type 2 diabetes at age <50 years remained similar in the U.K. between 2011 and 2017. However, we are not aware of any study that has evaluated ethnicity-specific temporal trends of cardiometabolic multimorbidity (CM) (at least two events of atherosclerotic CVD [ASCVD], microvascular disease, cancer, obesity grade 2+ [BMI ≥35 kg/m²]), or depression at diagnosis of diabetes or how ASCVD and mortality risk differ at a population level among different ethnic groups by age of type 2 diabetes diagnosis. Therefore, using a nationally representative EMR database from the U.S., we investigated 1) the trends in early-onset type 2 diabetes diagnosis in AAs and WCs between 2000 and 2018, 2) the trends in CM and depression and the cardiometabolic risk factor distribution at diabetes diagnosis by age-group and ethnicity, and 3) the risk of ASCVD and three-point major adverse cardiovascular events (MACE-3) (heart failure [HF], myocardial infarction, or stroke) by ethnicities and age-groups at type 2 diabetes diagnosis.

This study was conducted following the Reporting of Studies Conducted Using Observational Routinely-Collected Data (RECORD) guidelines.

The Centricity EMR (CEMR) incorporates patient-level data from independent physician practices, academic medical centers, hospitals, and large integrated delivery networks in the U.S. CEMR partners contribute deidentified patient-level data to enable quality improvement, benchmarking, and population-based medical research. The CEMR database covers >40,000 health care providers from all U.S. states, where ∼70% are primary care providers. The similarity of the general population characteristics and cardiometabolic risk factors in the CEMR database with those reported in U.S. national health surveys has been reported previously (11,12).

The CEMR database has been used extensively for academic research (11,13,14). Longitudinal EMRs were available for >46 million individuals from 1995 until September 2018, with comprehensive patient-level information on demographics, anthropometric measures, disease events, medications, and clinical and laboratory measures.

The study cohort was identified with the following conditions: 1) data available on age and sex, 2) WC or AA ethnicity, 3) aged 18–70 years at the time of type 2 diabetes diagnosis, 3) diagnosed on or after 1 January 2000 to 30 September 2018, and 4) at least 1 year of available data in the EMRs before diabetes diagnosis to reduce the bias in identifying patients with incident diabetes. The clinically driven machine learning–based algorithms to identify patients with type 2 diabetes from EMRs have been described previously (15,16).

Ethnicity in CEMR is coded according to the U.S. Census Bureau categorization (17). HbA_1c measures at baseline were obtained as the nearest measure within 3 months either side of the diagnosis. Baseline body weight, systolic blood pressure, blood lipid levels, and estimated glomerular filtration rate (eGFR) were calculated as the average of available measures within 3 months of diagnosis.

A robust methodology for extraction and assessment of longitudinal patient-level medication data from the CEMR has previously been described (18). A detailed account of glucose-lowering drug use in the U.S. population, on the basis of the CEMR, has also been reported (19). Antihypertensive drugs included all U.S. Food and Drug Administration–approved diuretics, peripheral vasodilators, β-blockers, calcium channel blockers, and agents acting on the renin-angiotensin system. Lipid-lowering drugs included statins, bile acid sequestrants, fibrates, nicotinic acid, proprotein convertase subtilisin/kexin type 9 inhibitors, and potent (≥1 g) forms of omega-3, fish, or krill oil.

ASCVD was defined by the presence of a clinical diagnosis of ischemic heart disease (myocardial infarction, unstable angina, or coronary revascularization, excluding stable angina), cerebrovascular disease (ischemic/hemorrhagic stroke, transient ischemic attack, or carotid revascularization), or peripheral artery/vascular disease. CVD included ASCVD and HF. Microvascular disease was defined by a clinical diagnosis of neuropathy, retinopathy, or chronic kidney disease (CKD). The CKD definition included diagnostic codes (CKD stages 3–5, end-stage renal disease, dialysis, transplant, nephropathy, proteinuria, albuminuria, nephrotic syndrome, nephritis, eGFR <60 mL/min/1.73 m², or urine albumin-to-creatinine ratio >30 mg/g). Cancer was defined as any malignant neoplasm, excluding melanoma. Hypertension and dyslipidemia were defined by the presence of a clinical diagnosis or use of antihypertensive/lipid-lowering drugs before diabetes diagnosis. Depression was defined using a clinically guided machine learning algorithm (20–22). The definition included those with a diagnostic code or at least two prescriptions for antidepressants (within a 6-month window) used for treating depression; antidepressant medications were limited to those commonly prescribed for depression (23,24) (Supplementary Table 1). The algorithm accounted for other mental illnesses specified within the DSM-5, excluding developmental and substance use disorders (25). A disease was considered as prevalent if its first available diagnostic date was on or before the type 2 diabetes diagnosis (index date).

Baseline characteristics were summarized separately by age-groups for AAs and WCs as number (%), mean (SD), or median (first quartile, third quartile), as appropriate. Age-groups at type 2 diabetes diagnosis were 18–39, 40–49, 50–59, and 60–70 years; early-onset type 2 diabetes was defined by the age cut point of <50 years at diagnosis.

We first estimated temporal trends of early-onset type 2 diabetes diagnosis as a proportion of total cases of diabetes and the prevalence of depression and CM by ethnicity and calendar year of type 2 diabetes diagnosis. Second, the risks of ASCVD and MACE-3 were evaluated by ethnicity and age-group using flexible parametric survival models with time from index date (diabetes diagnosis) to the event or censoring (30 September 2018). Models were adjusted for age, sex, and smoking status and accounted for the use of insulin, hypertension, dyslipidemia, depression, and other comorbidities before the end of follow-up as time-dependent covariates (26).

Other comorbidities included microvascular diseases, cancer, HF, and grade 2 obesity before an ASCVD event for ASCVD risk assessment and any CVD (except myocardial infarction, HF, or stroke), microvascular disease, cancer, and grade 2 obesity before a MACE-3 event for MACE-3 risk assessment. The use of noninsulin antidiabetic drugs as a potential confounder was excluded from the model on the basis of statistical information criteria. Age-group–specific separate regression models were fitted to evaluate associations of depression and other comorbidities with ASCVD and MACE-3 risk in AAs and WCs. The multivariable model allowed the estimation of hazard ratios (HRs) and restricted mean survival time to ASCVD and MACE-3. The modifiable effect of depression in the risk of ASCVD and MACE-3 was evaluated using interaction modeling by ethnicity and depression status. Separate models were also fitted without depression as a confounder to evaluate the risk differentiation between ethnicities. Statistical significance level was set at 5%, with all CIs set at 95%.

Supplementary Fig. 1 shows the flowchart of people with incident type 2 diabetes identified in the study. The proportion of AAs diagnosed with diabetes within the age-groups 18–39 and 40–49 years was 13% and 21%, respectively, and was consistently higher by ∼4% over the past two decades compared with their WC counterparts (10% and 17%, respectively, all P < 0.01). The proportion of AAs and WCs diagnosed at <50 years between 2012 and 2018 increased from 34 to 38% and from 26 to 29%, respectively (Fig. 1A) (all P < 0.05).

The baseline characteristics along with missing data in relevant risk factors are presented in Table 1. In the cohorts of 101,104 AAs and 505,336 WCs, AAs had significantly higher mean HbA_1c at diagnosis across all age-groups compared with WCs (all P < 0.05), while both ethnic groups had the highest HbA_1c level in the youngest age-group compared with the older age-groups (Table 1). AAs had significantly higher BMI (83% obese) than WCs (77%) in the 18–39-year age-group (P = 0.02), while the BMI distributions (including grade 2 obesity) at diagnosis were similar between the ethnicities in all age-groups ≥40 years.

In the overall cohort, proportions with ASCVD, MACE-3, microvascular diseases, and depression at diagnosis among AAs and WCs were 13% and 16%, 6% and 5%, 27% and 26%, and 19% and 31%, respectively. While the prevalence of ASCVD and MACE-3 in early-onset type 2 diabetes was similar between ethnicities (all P > 0.05), in older people with type 2 diabetes, it was higher for ASCVD (P = 0.031) and lower for MACE-3 among WCs (P = 0.044) (Table 2). Although the proportions with microvascular disease were similar between ethnicities aged <50 years (P = 0.84), AAs had a higher proportion with CKD (9%) than WCs (7%) (P = 0.041).

The prevalence of depression at the time of type 2 diabetes diagnosis has been significantly increasing across all age-groups and both ethnicities, being significantly higher for WCs than for AAs (Fig. 1B). Among AAs and WCs, the overall prevalence of depression increased from 15 and 20% in 2000 to 23 and 34% in 2017, respectively (Fig. 1B). These proportions were similar among people with early- and usual-onset type 2 diabetes (20% in AAs and 34% in WCs). The increasing trend in the prevalence of CM post-2010 has been similar in both ethnicities aged <60 years (Fig. 1C–E).

The mean (median) follow-up time for ASCVD among AAs and WCs was 4.8 (4.1) years and 5.4 (4.7) years, respectively. The crude event rates per 1,000 person-years are presented in Supplementary Table 2.

AAs had a 17% (HR 1.17 [95% CI 1.05, 1.31]) higher adjusted ASCVD risk compared with WCs only in the 18–39-year age-group, while there was no difference in the ASCVD risk between the two ethnic groups in the older age-groups (Table 3 and Fig. 2). In the subgroup of people with depression, the observed higher ASCVD risk in AAs aged 18–39 years disappeared (HR 95% CI 0.96, 1.35).

AAs had significantly higher adjusted MACE-3 risk compared with WCs across all age-groups. However, the higher MACE-3 risk among AAs was lowest in the oldest age-group (60–70 years: HR 1.11 [95% CI 1.06, 1.15]) and highest in the youngest age-group (1.63 [1.42, 1.88]). Higher MACE-3 risk in AAs was consistent in those with and without depression (Table 3).

The restricted mean years to ASCVD post–type 2 diabetes diagnosis in AAs and WCs in the 50–60-year age-group were 15.2 and 15.1, ∼2.5 years earlier than the expected time to event in the 18–39-year age-group (17.4 and 17.7 in AAs and WCs, respectively) in both ethnic groups. The difference in time to MACE-3 was also ∼2.5 years in both AAs and WCs.

The depression state at diagnosis or during follow-up was independently associated with a 1–36% and 1–28% increased risk of ASCVD and MACE-3, respectively, a modest effect size consistently similar across age-groups and ethnicity (all P < 0.05) (Table 3). Other comorbidities were independently associated with the ASCVD and MACE-3 risk only among WCs. Hypertension was associated with a 2–50% and 2–39% significantly increased risk of ASCVD and MACE-3, respectively, with the effect size being similar across all age-groups and ethnicity.

In the absence of any comparative U.S. nationally representative ethnicity-based data on the prevalence of comorbidities, including depression, in early-onset type 2 diabetes, this study provides unique and comprehensive information on the temporal trend in the prevalence of CM and depression, with heterogeneity in early- and usual-onset type 2 diabetes. Another novelty of this study is the evaluation of ethnic differences in the risk of ASCVD and MACE-3 by age-group at diabetes diagnosis and the interplay of depression and other comorbidities in such differences in a large cohort of people from a representative U.S. EMR (12).

We have observed a consistently increasing trend in the prevalence of depression at type 2 diabetes diagnosis. The difference in the prevalence estimates between the ethnic groups has remained similar over the past decade, with prevalence estimates in AAs and WCs aged <40 years being 18% and 32%, respectively. One of the reasons for the significantly lower prevalence estimate of depression at type 2 diabetes diagnosis in AAs than in WCs is the discrepancies in the presentation of depression among AAs compared with WCs. Depression in AAs has a much higher likelihood of being underreported, underdiagnosed, or misdiagnosed (27,28). Studies have suggested that AAs with depression, particularly men, are more reticent to seek help from mental health services than non-AAs (29). While the observed overall proportion with comorbid depression at type 2 diabetes diagnosis is similar to that reported in the recent meta-analysis of observational studies (30), we are not aware of any ethnicity-specific comparable study evaluating the temporal trend in the prevalence of depression in incident diabetes.

Some studies that were based on observational data from the U.S. have discussed the cardiometabolic risk factor and biomarker differences between AAs and WCs in the general population (31,32). However, such ethnicity-based data in people with incident type 2 diabetes are scarce. While we are not aware of any observational study from the U.S. reporting prevalence of hypertension, dyslipidemia, and other comorbidities by age-group at diagnosis of diabetes, the observed prevalence of hypertension and dyslipidemia at type 2 diabetes diagnosis on the basis of the CEMR is comparable to those reported from survey data and observational studies from the U.S. (33). It has been suggested that differences in the risk of CVD complications comparing type 2 diabetes diagnosed at a younger versus an older age may be related to the differences in the control of risk factors. While this may be a possible explanation, we note that this was not formally tested in our study (i.e., accounting for levels of risk factors), yet CM was associated with a higher relative risk of ASCVD in those diagnosed at younger versus older ages.

The role of ethnicity in the higher risk of ASCVD events in the general population has been investigated in previous studies and U.S. surveys (34–36). However, the extent to which differences by age-groups and multimorbidity interplay with CVD outcomes has not been investigated to our knowledge. We showed consistently higher MACE-3 risk in AAs than in WCs in early-onset type 2 diabetes, although this difference tended to be lower for those diagnosed at an older age. Importantly, these differences were present regardless of the presence of depression, which was independently associated with a greater risk of ASCVD and MACE-3.

This EMR-based study has several limitations. Coding of conditions is a common limitation when using EMRs. However, we used machine learning methods to identify both the population and the depression at diagnosis. The U.S. Centers for Disease Control and Prevention 2015 survey reported 20% prevalence of any mental illness, similar to the U.S. CEMR during that period (23%) (37,38). It is important to mention here that the CEMR database does not directly link to hospitalization data. Thus, the MACE-3 event estimates (especially HF) may be underrepresented. Other limitations include unavoidable indication bias and residual confounding that remains as a common problem in any EMR-based outcome studies and lack of data on socioeconomic characteristics, physical activity, the nature of insurance, education, income, and other cultural drivers. Furthermore, while reliable information on medication adherence is a common problem in all clinical studies, detailed validation studies of U.S. EMRs suggested a high level of agreement between EMR prescription data and pharmacy claims data, especially in chronic diseases (39).

Nevertheless, our study highlights the increasing burden of early-onset type 2 diabetes and in AAs, the increased risk of ASCVD in this early-onset group. The increased risk of depression in both ethnic groups regardless of age highlights that early detection and management of depression in diabetes may be an important strategy in reducing cardiovascular risk.

Acknowledgments. Melbourne EpiCentre gratefully acknowledges support from the National Health and Medical Research Council and the Australian Government’s National Collaborative Research Infrastructure Strategy initiative through Therapeutic Innovation Australia. The authors acknowledge Dr. Digsu Koye from the University of Melbourne for supporting this research.

Duality of Interest. F.Z. has received speaker fees from Boehringer Ingelheim and Napp Pharmaceuticals. J.A.S. has received funding from AstraZeneca UK in the form of an investigator-initiated trial. M.J.D. has acted as consultant, advisory board member, and speaker for Novo Nordisk, Sanofi, Eli Lilly, Merck Sharp & Dohme, Boehringer Ingelheim, AstraZeneca, and Janssen; as advisory board member for SERVIER and Gilead Sciences; and as speaker for NAPP, Mitsubishi Tanabe Pharma Corporation, and Takeda Pharmaceuticals International. She has received grants in support of investigator and investigator-initiated trials from Novo Nordisk, Sanofi, Eli Lilly, Boehringer Ingelheim, AstraZeneca, and Janssen. K.K. has served as a consultant for and received speakers’ fees from Amgen, AstraZeneca, Bayer, Berlin-Chemie AG/Menarini Group, Boehringer Ingelheim, Eli Lilly, Merck Sharp & Dohme, NAPP, Novartis, Novo Nordisk, Roche, Sanofi, and SERVIER; has served on an advisory board for AstraZeneca, Eli Lilly, Merck Sharp & Dohme, Novo Nordisk, and Sanofi; and has received grants in support of investigator and investigator-initiated trials from AstraZeneca, Boehringer Ingelheim, Eli Lilly, Merck Sharp & Dohme, Novartis, Novo Nordisk, Pfizer, Sanofi, and SERVIER. S.K.P. has acted as a consultant and/or speaker for Novartis, Sanofi Aventis, GI Dynamics, Roche, AstraZeneca, Guangzhou Zhongyi Pharmaceutical, and Amylin Pharmaceuticals. He has received grants in support of investigator and investigator-initiated clinical studies from Merck, Novo Nordisk, AstraZeneca, Hospira, Amylin Pharmaceuticals, Sanofi, and Pfizer. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. J.E.D. and O.M. conducted the data extraction. J.E.D., F.Z., J.A.S., M.J.D., and K.K. contributed in the interpretation of results and finalization of the manuscript. J.E.D., O.M., and S.K.P. jointly conducted the statistical analyses. O.M. and S.K.P. conceptualized and designed the study and developed the first draft of the manuscript. O.M. and S.K.P. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented orally at the 56th European Association for the Study of Diabetes Annual Meeting, Virtual, 21–25 September 2020.