To examine the association between microvascular disease (MVD) and risk of heart failure (HF) among individuals with type 1 diabetes mellitus (T1DM) or type 2 diabetes mellitus (T2DM).
We included 1,713 and 28,624 participants with T1DM and T2DM, respectively, from the UK Biobank who were free of HF during enrollment. MVD burden reflected by the presence of retinopathy, peripheral neuropathy, and chronic kidney disease (CKD) at baseline was prospectively evaluated for the association with incidence of HF. Hazard ratios (HRs) and 95% CIs of HF were estimated by Cox regression models adjusted for multiple traditional risk factors.
There were 145 and 2,515 incident cases of HF recorded among participants with T1DM and T2DM, respectively, during a median follow-up of 11.5 years. The association between the number of MVD and HF was stronger among participants with T1DM than among those with T2DM (P for interaction <0.001). Compared with participants with no MVD, those with all three MVD had an adjusted HR (95% CI) of 11.37 (5.62, 22.99) in T1DM and 3.66 (2.74, 4.88) in T2DM. In T1DM, HRs (CIs) were 2.69 (1.75, 4.14) for retinopathy, 2.11 (1.38, 3.23) for peripheral neuropathy, and 2.21 (1.53, 3.18) for CKD. The corresponding estimates in T2DM were 1.24 (1.13, 1.36), 1.63 (1.36, 1.96), and 1.73 (1.59, 1.89), respectively.
While a heavier burden of MVD was associated with excess risk of HF both in T1DM and T2DM, the association was evidently more pronounced in T1DM.
Introduction
In the U.K., heart failure (HF) currently affects ∼ 900,000 people, with up to 17% of patients dying within the first year of diagnosis (1). Diabetes and HF are frequent comorbid conditions (2), and it has been estimated that ∼ 40% of individuals with diabetes have HF (3). In fact, following peripheral arterial disease, HF is the second common initial presentation of macrovascular disease in the population with diabetes and is even more common than myocardial infarction or stroke (4).
Both type 1 diabetes mellitus (T1DM) and type 2 diabetes mellitus (T2DM) increased risk of HF, raising the risk by two- to fourfold (5–7). However, individuals with diabetes can have substantial heterogeneity in terms of clinical features—for example, differences in microvascular burden (8). Besides hyperglycemia, factors thought to explain the excess risk of HF in individuals with diabetes are the burden of microvascular diseases (MVD), manifested as retinopathy, peripheral neuropathy, and chronic kidney disease (CKD). For example, a previous study of the UK Clinical Practice Research Datalink found the cumulative burden of MVD to be associated with risk of incident HF among T2DM (9), and an analysis of the EMPA-REG OUTCOME trial also showed that among participants with T2DM, the coexistence of MVD in the setting of established macrovascular disease was linked to increased HF risk (10). These studies, however, have been limited by including only patients referred to a registry (9) or a post hoc analysis of a clinical trial (10). More importantly, T1DM and T2DM are two pathologically distinct conditions while these previous studies only included patients with T2DM. How MVD may be associated with risk of HF among individuals with T1DM remains open for investigation.
Using data from the UK Biobank, a large-scale population-based cohort study, we aimed to evaluate the associations of a group of MVD (i.e., retinopathy, peripheral neuropathy, and CKD), alone or in combination, with incident HF risk both among individuals with T1DM and those with T2DM.
Research Design and Methods
Study Population
The UK Biobank is a nationwide cohort study that recruited >500,000 participants aged 40 to 70 years at 22 assessment centers throughout the U.K. between 2006 and 2010 (11,12). At baseline, participants completed self-reported touchscreen questionnaires, interviews, and physical measurements, which collected information on lifestyle and other potentially health-related aspects. Biological samples were also collected for assays. All participants in the UK Biobank provided written informed consent. The UK Biobank’s study protocol was approved by the U.K. North West Multicenter Research Ethics Committee.
We included participants with DM in our analysis. DM was determined based on the following information: 1) self-reported history of DM; 2) self-reported use of DM medications; 3) a glycated hemoglobin (HbA1c) level ≥48 mmol/mol (6.5%); 4) random glucose ≥7.0 mmol/L if the interval between the last meal was ≥8 h or ≥11.1 mmol/L if <8 h; or 5) history of DM in the medical record (ICD-9 codes 250 and 6480; ICD-10 codes E10–E14) (13). T1DM was defined by medical records, a self-reported history of the disease, or by an age at onset of DM of <40 years with insulin treatment initiation within 1 year of diagnosis and T2DM otherwise (14,15). Of the initial 31,166 participants with DM, 678 had HF at baseline, and 151 withdrew from the study, leaving 1,713 participants with T1DM and 28,624 participants with T2DM (Supplementary Fig. 1).
Exposure Assessment
In the current study, DM-related MVD, including retinopathy, neuropathy, and CKD, were used as the exposures of interest. Prevalent MVD was defined according to self-reported information (noncancer illness code 20002) and hospital inpatient records. The date and diagnosis of hospital admissions were determined through record linkage to Hospital Episode Statistics in England and Wales and the Scottish Morbidity Records in Scotland. As such, participants were considered to have a history of MVD if the date of individuals’ hospital inpatient records was before the baseline date. The following ICD codes were used to define the prevalent cases of MVD (Supplementary Table 1): for retinopathy, ICD-9: 2504; and ICD-10: E10.3, E11.3, E12.3, E13.3, E14.3, H28.0, H33, H35.3, H36.0, and H40–H42; for neuropathy, ICD-9: 2505 and 3572; and ICD-10: E10.4, E11.4, E13.4, E14.4, G59.0, G62.9, G63.2, and G99.0; for CKD, ICD-9: 2503, 4039, 4401, 5829, 5845, and 5849; and ICD-10: E10.2, E11.2, E13.2, E14.2, N18–N19, N28.0, and I70.1. These ICD codes are generally consistent with those used in previous studies on MVD in the U.S. and U.K. (16,17). The estimated glomerular filtration rate (eGFR) was calculated by using the Chronic Kidney Disease Epidemiology Collaboration equation based on serum creatinine (Supplementary Table 2), and CKD was defined as an eGFR of <60 mL/min/1.73 m2 or microalbuminuria (urinary albumin-to-creatinine ratio [uACR] ≥3 but ≤30 mg/mmol) or macroalbuminuria (uACR >30 mg/mmol) (18).
HF Assessment
Prevalent and incident HF cases were defined based on self-reported diagnosis and hospital inpatient records. As mentioned above, the Hospital Episode Statistics in England and Wales and the Scottish Morbidity Records in Scotland were linked to the UK Biobank to determine the date and diagnosis for hospital admissions. Individuals without HF at baseline but who were subsequently found to have any hospital records indicative of HF based on the ICD codes were considered to have had the outcome of interest. The following ICD codes were used to define prevalent and incident HF: for ICD-9: 4280, 4281, and 4289; and for ICD-10: I11.0, I13.0, I13.2, I50.0, I50.1, and I50.9 (Supplementary Table 3). At the time of our analyses, hospital inpatient records were available up to 30 November 2020, and mortality data were updated to 18 December 2020 for England and Wales and 10 December 2020 for Scotland; therefore, these dates were used as the end of follow-up. Person-time was calculated from the date of baseline assessment to the date of diagnosis of the event, death, loss to follow-up, or end of follow-up, whichever occurred first.
Measurements of Covariates and Biomarkers
The UK Biobank study collected extensive information on the following factors: sociodemographic information (age, sex, and ethnicity), area-based social deprivation (Townsend deprivation index), lifestyle (smoking and sedentary time), and self-reported drug use (antihyperglycemic drugs and antihypertensive drugs, etc.) (Supplementary Table 4). Baseline coronary heart disease (CHD) and stroke were defined by self-reported and hospital diagnoses. BMI was calculated as measured weight in kilograms divided by measured height in meters squared. Baseline blood pressure (BP) measurements were obtained after the participant had been at rest for at least 5 min in a seated position using a digital BP monitor (Omron HEM-7015IT; OMRON Healthcare Europe B.V., Hoofddorp, the Netherlands) with a suitably sized cuff. Mean systolic BP (SBP) and diastolic BP values and heart rate were calculated from two automated or two manual measurements by trained nurses (19). Sedentary time (hours) was calculated by summing the times of three activities for the hours per day participants spent: 1) driving, 2) using a computer, and 3) watching television. DM duration was calculated as baseline age minus self-reported age at diagnosis (14,20). Participants without known DM but with an HbA1c level exceeding the threshold for DM (based on HbA1c or random glucose) at baseline were assigned a duration of 0 years.
Serum and urine creatinine were measured using an enzymatic (creatinase), isotope dilution mass spectrometry–traceable method on a Beckman Coulter AU5400 instrument. Urine microalbumin was measured by an immunoturbidimetric method using reagents and calibrators sourced from Randox Biosciences. Serum major lipids, CRP levels, as well as random glucose were quantified using standard procedures through the Beckman Coulter AU5800. HbA1c levels were measured by high-performance liquid chromatography analysis on a Bio-Rad Laboratories VARIANT II TURBO.
Statistical Analysis
Descriptive data were reported as median (interquartile range) for continuous variables and as percentages for categorical variables by type of DM as well as across the number of MVD. Categorical variables were compared by using the χ2 test, and continuous variables were compared by Kruskal-Wallis or unpaired Mann-Whitney U test, where appropriate.
Missing values accounted for ≤9.5% for all covariates, except for HDL cholesterol (HDL-C; 17.1%) (Supplementary Table 5). Participants with missing values were assigned to a separate “unknown” category for categorical variates. As for continuous variables, we conducted multiple imputations with chained equations to address missing data (21), and five data sets were imputed; all variables used in the analyses were included in the imputation model. Kaplan-Meier curves were generated for individual MVD and the number of MVD in relation to risk of HF, and log-rank tests were used to compare different groups. Cox proportional hazards models were used to estimate hazard ratios (HRs) and corresponding 95% CIs of HF after the assumptions of the proportionality of hazards were confirmed by examinations of Schoenfeld residuals. Two models were used. Model 1 was adjusted for age, sex, ethnicity (White or non-White), BMI, smoking status (never, previous or current), SBP, total cholesterol (TC), HDL-C, and LDL cholesterol (LDL-C). Model 2 was further adjusted for other potential confounders, including Townsend deprivation index, alcohol consumption, sedentary time, heart rate, history of CHD, history of stroke, CRP, HbA1c, antihypertensive drugs, antihyperglycemic drugs, statin use, and DM duration. The statistical significance of the interactions was assessed by adding a multiplicative term to the Cox models. We further examined the associations between the number of MVD and risk of incident HF in men and women separately. Sensitivity analysis was also performed by: 1) excluding participants with any missing data; 2) redefining T1DM by restricting the age of onset to <30 years; 3) using only the ICD codes to define MVD; and 4) excluding non-HF deaths in order to take competing risks into consideration.
All analyses were performed using STATA 14.0 (StataCorp LP, College Station, TX) or R software (version 4.0.1). A two-sided P value <0.05 was considered statistically significant.
Results
Baseline Characteristics
The number of prevalent cases of MVD among participants with T1DM and T2DM is presented in Supplementary Fig. 2. Briefly, 59 participants with T1DM and 130 with T2DM identified to have all of the 3 MVD baseline characteristics of the participants with T1DM and T2DM are shown in Table 1. Compared with participants with T2DM, those with T1DM had higher rates of all of the three MVD, and they were more likely to be women, White, and current smokers, had higher levels of HbA1c but lower BMI, SBP, LDL-C, and CRP levels, drank more alcohol, and were less likely to have CHD. Participants’ characteristics by the number of MVD among those with T1DM and T2DM are shown in Supplementary Tables 6 and 7, respectively. Generally, those with all of the three MVD were more likely to have a heavier burden of unfavorable risk factors compared with those without any MVD at baseline.
. | T1DM . | T2DM . | P value . |
---|---|---|---|
Number of MVD, n (%) | <0.001 | ||
0 | 806 (47.1) | 19,116 (66.8) | — |
1 | 623 (36.4) | 7,844 (27.4) | — |
2 | 225 (13.1) | 1,534 (5.4) | — |
3 | 59 (3.4) | 130 (0.5) | — |
History of retinopathy, n (%) | 757 (44.2) | 5,073 (17.7) | <0.001 |
History of neuropathy, n (%) | 141 (8.2) | 605 (2.1) | <0.001 |
History of CKD, n (%) | 352 (20.6) | 5,624 (19.7) | <0.001 |
HbA1c, mmol/L | 59.3 (50.7–68.3) | 50.7 (43.9–57.1) | <0.001 |
DM duration, years | 29.0 (18.0–38.0) | 4.0 (1.0–8.0) | <0.001 |
Age, years | 54.0 (47.0–61.0) | 61.0 (55.0–65.0) | <0.001 |
Women, n (%) | 793 (46.3) | 11,349 (39.7) | <0.001 |
White, n (%) | 1,586 (92.5) | 24,499 (85.6) | <0.001 |
Townsend deprivation index | −1.3 (−3.2 to 1.9) | −1.2 (−3.1 to 2.1) | 0.083 |
BMI, kg/m2 | 26.9 (24.2–30.6) | 30.6 (27.5–34.7) | <0.001 |
Current smokers, n (%) | 213 (12.4) | 3,280 (11.5) | <0.001 |
SBP, mmHg | 139.0 (127.0–152.0) | 143.0 (131.0–156.0) | <0.001 |
Heart rate, bpm | 73.0 (65.0–82.0) | 74.0 (65.0–83.0) | 0.099 |
Alcohol consumption, g/day | 7.1 (0.0–20.0) | 4.3 (0.0–17.1) | <0.001 |
Sedentary time, h/day | 5.0 (3.5–6.0) | 5.0 (4.0–7.0) | <0.001 |
TC, mg/dL | 4.6 (4.0–5.0) | 4.6 (3.9–5.3) | 0.046 |
HDL-C, mmol/L | 1.3 (1.2–1.7) | 1.2 (1.0–1.3) | <0.001 |
LDL-C, mmol/L | 2.6 (2.2–2.9) | 2.8 (2.3–3.3) | <0.001 |
CRP, mg/L | 1.8 (0.8–3.3) | 2.2 (1.0–3.9) | <0.001 |
eGFR, mL/min/1.73 m2 | 97.9 (85.0–106.2) | 93.9 (82.0–101.2) | <0.001 |
History of CHD, n (%) | 227 (13.2) | 4,872 (17.0) | <0.001 |
History of stroke, n (%) | 52 (3.0) | 955 (3.3) | 0.500 |
Antihyperglycemic drug use, n (%) | 1,698 (99.1) | 21,229 (74.2) | <0.001 |
Antihypertensive drug use, n (%) | 884 (51.6) | 16,685 (58.3) | <0.001 |
Statin use, n (%) | 1,095 (63.9) | 17,531 (61.3) | 0.027 |
. | T1DM . | T2DM . | P value . |
---|---|---|---|
Number of MVD, n (%) | <0.001 | ||
0 | 806 (47.1) | 19,116 (66.8) | — |
1 | 623 (36.4) | 7,844 (27.4) | — |
2 | 225 (13.1) | 1,534 (5.4) | — |
3 | 59 (3.4) | 130 (0.5) | — |
History of retinopathy, n (%) | 757 (44.2) | 5,073 (17.7) | <0.001 |
History of neuropathy, n (%) | 141 (8.2) | 605 (2.1) | <0.001 |
History of CKD, n (%) | 352 (20.6) | 5,624 (19.7) | <0.001 |
HbA1c, mmol/L | 59.3 (50.7–68.3) | 50.7 (43.9–57.1) | <0.001 |
DM duration, years | 29.0 (18.0–38.0) | 4.0 (1.0–8.0) | <0.001 |
Age, years | 54.0 (47.0–61.0) | 61.0 (55.0–65.0) | <0.001 |
Women, n (%) | 793 (46.3) | 11,349 (39.7) | <0.001 |
White, n (%) | 1,586 (92.5) | 24,499 (85.6) | <0.001 |
Townsend deprivation index | −1.3 (−3.2 to 1.9) | −1.2 (−3.1 to 2.1) | 0.083 |
BMI, kg/m2 | 26.9 (24.2–30.6) | 30.6 (27.5–34.7) | <0.001 |
Current smokers, n (%) | 213 (12.4) | 3,280 (11.5) | <0.001 |
SBP, mmHg | 139.0 (127.0–152.0) | 143.0 (131.0–156.0) | <0.001 |
Heart rate, bpm | 73.0 (65.0–82.0) | 74.0 (65.0–83.0) | 0.099 |
Alcohol consumption, g/day | 7.1 (0.0–20.0) | 4.3 (0.0–17.1) | <0.001 |
Sedentary time, h/day | 5.0 (3.5–6.0) | 5.0 (4.0–7.0) | <0.001 |
TC, mg/dL | 4.6 (4.0–5.0) | 4.6 (3.9–5.3) | 0.046 |
HDL-C, mmol/L | 1.3 (1.2–1.7) | 1.2 (1.0–1.3) | <0.001 |
LDL-C, mmol/L | 2.6 (2.2–2.9) | 2.8 (2.3–3.3) | <0.001 |
CRP, mg/L | 1.8 (0.8–3.3) | 2.2 (1.0–3.9) | <0.001 |
eGFR, mL/min/1.73 m2 | 97.9 (85.0–106.2) | 93.9 (82.0–101.2) | <0.001 |
History of CHD, n (%) | 227 (13.2) | 4,872 (17.0) | <0.001 |
History of stroke, n (%) | 52 (3.0) | 955 (3.3) | 0.500 |
Antihyperglycemic drug use, n (%) | 1,698 (99.1) | 21,229 (74.2) | <0.001 |
Antihypertensive drug use, n (%) | 884 (51.6) | 16,685 (58.3) | <0.001 |
Statin use, n (%) | 1,095 (63.9) | 17,531 (61.3) | 0.027 |
Data (continuous covariates) are reported as median (interquartile range) or n (%).
MVD and Risk of Incident HF
During a median follow-up of 11.5 years, 145 and 2,515 incident cases of HF were recorded among participants with T1DM and T2DM, respectively. Supplementary Figures 3 and 4 show the Kaplan-Meier curves for HF by number of MVD and individual MVD, respectively (all P for log-rank test <0.001). The excess risk of HF associated with increasing number of MVD was stronger in T1DM than in T2DM (P for interaction <0.001). Among participants with T1DM, the fully adjusted HRs (95% CIs) for HF were 2.01 (1.17, 3.44) for one MVD, 5.05 (2.90, 8.81) for two MVD, and 11.37 (5.62, 22.99) for three MVD, compared with no MVD; the corresponding figures for participants with T2DM were 1.48 (1.35, 1.61), 2.19 (1.92, 2.49), and 3.66 (2.74, 4.88), respectively (Table 2).
. | Number of MVD . | |||
---|---|---|---|---|
T1DM | 0 | 1 | 2 | 3 |
Events/person-years | 20/9,215 | 51/6,757 | 55/2,161 | 19/430 |
Model 1 | Reference | 2.52 (1.49, 4.25) | 7.18 (4.23, 12.18) | 17.6 (9.26, 33.50) |
Model 2 | Reference | 2.01 (1.17, 3.44) | 5.05 (2.90, 8.81) | 11.37 (5.62, 22.99) |
T2DM | 0 | 1 | 2 | 3 |
Events/person-years | 1,197/213,383 | 944/82,129 | 323/14,823 | 51/1,053 |
Model 1 | Reference | 1.74 (1.59, 1.90) | 3.11 (2.74, 3.52) | 6.45 (4.86, 8.55) |
Model 2 | Reference | 1.48 (1.35, 1.61) | 2.19 (1.92, 2.49) | 3.66 (2.74, 4.88) |
. | Number of MVD . | |||
---|---|---|---|---|
T1DM | 0 | 1 | 2 | 3 |
Events/person-years | 20/9,215 | 51/6,757 | 55/2,161 | 19/430 |
Model 1 | Reference | 2.52 (1.49, 4.25) | 7.18 (4.23, 12.18) | 17.6 (9.26, 33.50) |
Model 2 | Reference | 2.01 (1.17, 3.44) | 5.05 (2.90, 8.81) | 11.37 (5.62, 22.99) |
T2DM | 0 | 1 | 2 | 3 |
Events/person-years | 1,197/213,383 | 944/82,129 | 323/14,823 | 51/1,053 |
Model 1 | Reference | 1.74 (1.59, 1.90) | 3.11 (2.74, 3.52) | 6.45 (4.86, 8.55) |
Model 2 | Reference | 1.48 (1.35, 1.61) | 2.19 (1.92, 2.49) | 3.66 (2.74, 4.88) |
Model 1 was adjusted for age, sex, ethnicity, BMI, SBP, TC, HDL-C, LDL-C, and smoking status. Model 2 was adjusted for model 1 plus Townsend deprivation index, alcohol consumption, sedentary time, heart rate, history of CHD, history of stroke, CRP, HbA1c, antihypertensive drugs, antihyperglycemic drugs, statin use, and diabetes duration. P for interaction <0.001.
The interaction with type of DM on risk of HF was observed for retinopathy (P for interaction <0.001) and peripheral neuropathy (P for interaction = 0.019), but not CKD (P for interaction = 0.069). Among participants with T1DM, all of the three MVD were significantly associated with risk of HF, and the fully adjusted HRs (CIs) for retinopathy, peripheral neuropathy, and CKD were 2.69 (1.75, 4.14), 2.11 (1.38, 3.23), and 2.21 (1.53, 3.18), respectively. Among T2DM, the corresponding HRs (95% CIs) were 1.24 (1.13, 1.36), 1.63 (1.36, 1.96), and 1.73 (1.59, 1.89), respectively (Table 3). Furthermore, in the fully adjusted models, a single MVD seemed to be more strongly associated with risk of HF than the majority of other traditional risk factors, except for the history of CHD for T2DM (Fig. 1 and Supplementary Table 8).
. | Retinopathy* . | |
---|---|---|
P interaction <0.001 . | T1DM . | T2DM . |
Events/person-years | 113/7,760 | 685/52,193 |
Model 1 | 3.11 (2.07, 4.65) | 1.48 (1.35, 1.62) |
Model 2 | 2.69 (1.75, 4.14) | 1.24 (1.13, 1.36) |
Peripheral neuropathy* | ||
P interaction = 0.019 | T1DM | T2DM |
Events/person-years | 35/1,265 | 129/5,983 |
Model 1 | 2.33 (1.57, 3.46) | 1.88 (1.56, 2.25) |
Model 2 | 2.11 (1.38, 3.23) | 1.63 (1.36, 1.96) |
CKD* | ||
P interaction = 0.069 | T1DM | T2DM |
Events/person-years | 70/3,346 | 929/56,758 |
Model 1 | 2.63 (1.86, 3.72) | 2.10 (1.93, 2.28) |
Model 2 | 2.21 (1.53, 3.18) | 1.73 (1.59, 1.89) |
. | Retinopathy* . | |
---|---|---|
P interaction <0.001 . | T1DM . | T2DM . |
Events/person-years | 113/7,760 | 685/52,193 |
Model 1 | 3.11 (2.07, 4.65) | 1.48 (1.35, 1.62) |
Model 2 | 2.69 (1.75, 4.14) | 1.24 (1.13, 1.36) |
Peripheral neuropathy* | ||
P interaction = 0.019 | T1DM | T2DM |
Events/person-years | 35/1,265 | 129/5,983 |
Model 1 | 2.33 (1.57, 3.46) | 1.88 (1.56, 2.25) |
Model 2 | 2.11 (1.38, 3.23) | 1.63 (1.36, 1.96) |
CKD* | ||
P interaction = 0.069 | T1DM | T2DM |
Events/person-years | 70/3,346 | 929/56,758 |
Model 1 | 2.63 (1.86, 3.72) | 2.10 (1.93, 2.28) |
Model 2 | 2.21 (1.53, 3.18) | 1.73 (1.59, 1.89) |
Model 1 was adjusted for age, sex, ethnicity, BMI, SBP, TC, HDL-C, LDL-C, and smoking status. Model 2 was adjusted for model 1 plus Townsend deprivation index, alcohol consumption, sedentary time, heart rate, history of CHD, history of stroke, CRP, HbA1c, antihypertensive drugs, antihyperglycemic drugs, statin use, and diabetes duration.
All of the three MVD were mutually adjusted for.
Supplementary Analysis and Sensitivity Analysis
The association between the number of MVD and risk of HF was generally similar in men and women (Supplementary Table 9). The main results were not materially changed across a series of sensitivity analyses (e.g., defining MVD using ICD codes only) (Supplementary Table 10).
Conclusions
In this 11.5-year follow-up cohort study of middle-aged and older adults with T1DM or T2DM, we found that diabetes-related microvascular complications, including retinopathy, peripheral neuropathy, and CKD, were differentially associated with risk of HF among T1DM and T2DM; that is, the associations of MVD and HF risk seemed to be stronger among T1DM than among T2DM. Notably, these MVD–HF associations were independent of traditional risk factors and coexisting macrovascular disease.
Comparison With Other Studies
The impact of diabetes on HF risk may be rooted in the multiorgan injury that silently erupts long before any overt clinical outcomes (22,23). In this regard, MVD status, a proxy of silent complications in individuals with diabetes, may play an important role in the development of overt complications such as HF. Although some previous studies have reported a significant association between the presence of MVD and cardiovascular-related morbidity and mortality among patients with T1DM or T2DM, only a few studies focused on HF. Using a sample of 9,141 patients with T2DM from the Genetics of Diabetes Audit and Research in Tayside Scotland registry with a median follow-up of 9.3 years, researchers found the number of MVD states to be associated with HF in a dose-response fashion (24). Similarly, in a sample of 4,095 participants with T2DM from the Look AHEAD (Action of Health in Diabetes) study, Kaze et al. (25) found the burden of MVD to confer a 2.5-fold increased risk of incident HF over a median follow-up of 9.7 years. Among 2,707 patients who had HF with reduced ejection fraction, Kristensen et al. (26) found that individuals with both diabetes and MVD had a higher risk of cardiovascular death and HF hospitalization than those with diabetes but without MVD. These studies, however, were mainly on participants with T2DM only and of a relatively small sample size. Meanwhile, other studies that included participants with T1DM only assessed the risk associated with cardiovascular disease without delineating for HF (27,28).
Compared with previous studies, one novel finding of our study is that the association of MVD burden with HF risk was stronger among individuals with T1DM than among those with T2DM. In our study, the associations between each individual MVD and HF appeared to be more marked among participants with T1DM than those with T2DM. Thus, the additive effect of individual MVD on HF risk was not surprisingly stronger for T1DM than T2DM. A possible explanation for the stronger MVD–HF link among individuals with T1DM than T2DM may be the longer duration of diabetes among those with T1DM (median of 29 years) than those with T2DM (median of 4 years). As a proxy for cumulative exposure to chronic hyperglycemia, diabetes duration may better predict adverse health risks than other risk factors. Maple-Brown et al. (29) found that among patients with T1DM, the inclusion of the degree of hyperglycemia and its duration (i.e., chronic glycemic exposure) may improve the prediction of retinopathy beyond that obtained with models, including the mean updated HbA1c. Our previous studies also suggested that diabetes duration seemed to be a better predictor of cardiovascular disease and dementia than glycemic control was among individuals with diabetes (14,20). The exact mechanisms for prolonged diabetes duration driving the MVD–HF link are not completely understood. However, it has been pointed out that long-term exposure to a hyperglycemic environment may trigger oxidative stress; enhance the activation of the polyol pathway, protein kinase C, and hexosamine biosynthetic pathway; and promote the production of advanced glycation end products, all of which may disrupt the function of basement membrane in microvessels, including those of the retina, neuron, glomerulus, and heart muscles, leading to the development of diabetes-related microvascular complications (30,31).
Notably, diabetic microvascular complications may to some extent signify the existence of systemic MVD, including the dysfunction of coronary microvascular structures (32). In the current study, we found that CHD appeared to be the strongest determinant of HF in participants with T2DM. In fact, the presence of CHD may represent a severe stage of coronary microvascular dysfunction, which is suggested to precede the occurrence of heart disease such as HF and is also reported to be a prominent feature of diabetic cardiomyopathy (33). Previous studies have suggested that long-term exposure to hyperglycemia may lead to increased oxidative stress in coronary microvascular endothelial cells through enhanced production of reactive oxygen species (34), leading to endothelial dysfunction in individuals with diabetes and may ultimately contribute to the development of HF (33).
Potential Mechanisms
In the current study, our risk estimates for the association of the number of MVD with incident HF for both types of diabetes were attenuated in the fully adjusted model. This suggests that MVD may be in part on the pathways that link conventional HF risk factors to the development of HF. However, the persistence of the significant association of MVD burden with HF risk suggests that MVD burden per se serves as an important risk contributor for the development of HF. The exact pathophysiologic link between MVD and HF in diabetes has been described as being complex and not clear. It has been postulated that the presence of retinopathy may be indicative of systemic compromise in the microvasculature, which may impose increasing resistance on the heart and possibly lead to cardiac function compromise and HF (35). Although the link between neuropathy and HF is still obscure, it has been suggested that increased levels of advanced glycation end products, which have been found to be associated with the development of neuropathy, may also cause inflammation and atherosclerosis (36). This sequence of events may increase the risk of developing HF in individuals with diabetes. Besides, it also has been suggested that the presence of peripheral neuropathy may be an indicator for a severe form of sickness with a greater likelihood of other comorbidities, which may increase the risk of developing cardiovascular disease/HF (36). For the CKD–HF link, hyperactivation of the renin-angiotensin-aldosterone system in the presence of CKD has been suggested to negatively impact the cardiovascular system by increasing intravascular volume that may increase the risk of developing HF (37). Also, it has been postulated that uremic toxins as a result of compromised renal function could have a harmful effect on the heart and impair left ventricular relaxation (37).
Clinical Implications
While our study does not claim causality or establish a pathophysiological mechanism between MVD burden and HF, it holds important implications for assessing risk for HF among patients with diabetes. Firstly, during routine clinical practice, clinicians should pay critical attention to patients with one or more MVD and carry out a thorough assessment of cardiovascular function. Secondly, cardioprotective measures should be intensified for patients who have MVD, especially patients with T1DM given the poorer prognosis in this group of patients.
Strengths and Limitations
Our study has some strengths, which include the prospective study design, the uniquely large sample size of both patients with T1DM and T2DM, the long-term follow-up, the detailed collections of various traditional risk factors, and the comprehensive adjustment for potential confounders. However, our study also has potential limitations. First, the majority of participants in the UK Biobank are of Caucasian background and of higher socioeconomic status than the general U.K. population, which may limit the generalizability of our results (38). Second, although we adjusted for potential confounding factors in our analysis, the possibility of residual confounding still remains; additionally, the UK Biobank was not able to provide updated information for the whole cohort during follow-up, and thus, the models were not able to adjust for time-varying covariates such as glycemic control and drug compliance. Third, we did not examine the relationship of MVD with subtypes of HF, which have been observed to be differently associated with MVD (39); also, incident cases of HF were determined by ICD codes in hospital inpatient records, which may have missed some mild cases that did not require hospitalization. Fourth, the current study only included classic MVD, while the impacts of other forms of MVD, such as diabetic dermopathy and cardiac autonomic neuropathy, were not assessed. Fifth, the ascertainment of some of the MVD was based on self-reported diagnosis, which is vulnerable to inaccuracy, and misclassification might have occurred. Sixth, no detailed data on degree/severity of retinopathy were collected, thus precluding a detailed analysis. Last, as urine albumin and serum creatinine were obtained from one test only, the estimates of uACR and eGFR may not be accurate. Additional studies with repeated measurements are needed.
Conclusion
In conclusion, a heavier burden of MVD is associated with excess risk of HF both among individuals with T1DM and those with T2DM, while the increases in the risk of HF associated with MVD were evidently greater among those with T1DM.
See accompanying article, p. 2817.
This article contains supplementary material online at https://doi.org/10.2337/figshare.20360028.
F.-R.L. and D.N.H. contributed equally to this work.
Article Information
Acknowledgments. The authors thank the UK Biobank participants. This research has been conducted using the UK Biobank Resource under application number 60009.
Funding. This study was supported by the National Natural Science Foundation of China (82173607), the Guangdong Basic and Applied Basic Research Foundation (2021A1515011684), Open Project of the Guangdong Provincial Key Laboratory of Tropical Disease Research (2020B1212060042), and Guangzhou Science and Technology Project (202102080597).
Duality of Interest. No potential conflicts of interest relevant to this article were reported.
Author Contributions. F.-R.L. contributed to the literature research, study design, and manuscript writing. F.-R.L., and D.N.H. contributed to the data acquisition. F.-R.L. contributed to the data analyses. D.N.H., J.Y., H.-H.Y., and G.-C.C. contributed to the literature search. G.-C.C., and X.-B.W. reviewed the manuscript. X.-B.W. is the guarantor of this work and, as such, 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.