Glycated hemoglobin variability, mean HbA1c and heart failure in patients with type 2 diabetes

Background: Heart failure (HF) is a common presentation in patients with type 2 diabetes mellitus (T2DM). Previous studies revealed that HbA1c level is signicantly associated with HF. However, little is known about the association between HbA1c variability and HF. We aimed to evaluate the association of HbA1c variability and mean HbA1c with HF in patients with T2DM. Methods: Using Diabetes Share Care Program data, patients with T2DM had mean HbA1c (HbA1c-Mean), HbA1c variability (tertiles of HbA1c-SD and HbA1c-adjSD) within 12-24 months during 2001-2008 were included. The cutoffs of HbA1c-Mean were set at <7%, 7-7.9%, and ≥ 8%. Hazard ratios (HRs) for HF during 2008-2018 were estimated using Cox proportional hazards models. Results: A total of 3824 patients were included, of whom 315 patients developed HF during the 11.72 years observation period. The associated risk of HF increased with tertiles of HbA1c variability and cutoffs of HbA1c-Mean. In mutually adjusted models, HbA1c-Mean showed a consistent dose-response association with HF, while the association of HbA1c variability with HF disappeared. Among patients with HbA1c-Mean < 7%, the associated risk of HF in patients with HbA1c variability in tertile 3 was comparable to patients with HbA1c-Mean ≥ 8%. Conclusions: Mean HbA1c was an independent predictor of HF and not explained by HbA1c variability. In addition to absolute HbA1c level, targeting on stability of HbA1c in patients with well glycemic control was also important for the development of HF in patients with T2DM.


Background
Heart failure (HF) is one of the most common initial manifestations of cardiovascular disease in patients with type 2 diabetes mellitus (T2DM) [1]. A previous study showed that the mortality rate of patients with diabetes and HF is 4-to 6-fold increase than patients with diabetes who do not develop HF [2]. A recent study also reported that the development of HF is associated with the highest 5-year risk of death compared with the development of other cardiovascular diseases in patients newly diagnosed T2DM [3].
An epidemiological study revealed that the incidence of HF is approximately 2 times higher in patients with diabetes than patients without diabetes [4]. Clinically, hypertension and coronary heart disease are well known causes of HF. These two risk factors are common comorbidities in patients with T2DM [5] which may partly explain the higher incidence of HF in patients with diabetes. However, previous studies showed that diabetes itself is an independent risk factor for the development of HF irrespective of blood pressure and coronary heart disease [6,7] which implies that there is a residual risk(s) for the development of HF in patients with T2DM. Several studies revealed that glycated hemoglobin (HbA1c) level is positively associated with HF in patients with T2DM [8][9][10][11][12]. In addition, a cohort study reported that an 1% reduction in mean HbA1c is associated with a 16% risk reduction in HF [13]. Recently, many studies showed that HbA1c variability is positively associated with macrovascular complications [14][15][16] and allcause mortality [17] in patients with T2DM. However, little is known about the association of HbA1c variability with HF. In this study, we aimed to examine the associated risk of HbA1c variability and mean HbA1c with HF in patients with T2DM.

Data source and design
In 2001, the Taiwan National Health Insurance Bureau established a Diabetes Shared Care Program to promote diabetes care. In this program, certi ed diabetes educators used a standardized electronic questionnaire to record basic data, personal habits, current and past diseases, and medications. Patients with a clinical diagnosis of diabetes were enrolled to the program at the outpatient clinic of China Medical University hospital, Taichung, Taiwan. In this study, we used data of patients enrolled between January 2001 and April 2008 for study analyses. The exclusion criteria included patients with a history of HF before enrollment, development of HF within 1 year after enrollment (for avoiding reverse causality), aged >80 years, age of diabetes onset <30 years, type 1 DM, cardiac dysrhythmia, congenital heart disease, valvular heart disease, estimated glomerular ltration rate (eGFR) <30 mL/min/1.73m 2 , and HbA1c measurements fewer than 3 times within 12-24 months. We linked each patient's personal identi cation number to the annual inpatient and outpatient claim database of the Taiwan National Health Insurance, provided by the Taiwan Health and Welfare Data Science Center, to verify the diagnoses of primary outcome (HF) and cardiovascular risk factors (hypertension, coronary heart disease, valvular heart disease, congenital heart disease, cardiac dysrhythmia, and stroke). Each patient was followed from the identi ed date to the development of HF, death, or December 31, 2018. The International Classi cation of Diseases, Ninth Revision, Clinical Modi cation (ICD-9-CM) and ICD-10-CM codes used in this study were presented in supplementary Table S1. This study was approved by the Ethical Review Board of China Medical University Hospital in Taiwan (CMUH107-REC2-163).

Statistical analyses
The mean and standard deviation (SD) of continuous variables and number and percentage for categorical variables were used to describe the distributions of the patients. The Student's t test and the chi-square test (or the Fisher's exact test) were used to compare continuous and categorical variables between patients with HF and patients without HF, respectively. HbA1c variability was presented with SD (HbA1c-SD). Because the number of HbA1c measurements could in uence the SD [18], the interindividual difference in the numbers of HbA1c measurements was adjusted according to the formula: SD/√[n/(n-1)] (HbA1c-adjSD) (18). For comparison, individuals' HbA1c-SD and HbA1c-adjSD were divided by tertiles; for mean HbA1c (HbA1c-Mean), the cutoffs were set at <7%, 7-7.9%, and ≥8%. A Cox proportional hazards model was used to estimate the hazard ratios (HRs) and 95% con dence intervals.
Multiple confounders in the adjusted model included sex, age, diabetes duration, body mass index, systolic blood pressure, total cholesterol, triglyceride, High-density lipoprotein cholesterol, Low-density lipoprotein cholesterol, eGFR, coronary heart disease, hypertension, stroke, the use of sulfonylureas, metformin, thiazolinediones, insulin, statin, antiplatelet agents, warfarin, angiotensin converting enzyme inhibitors, angiotensin II receptor blockers, beta-blockers, calcium channel blockers, diuretics, and alfablockers. In Table 2, model 1 was adjusted for multiple confounders; model 2 was a mutually (HbA1c-SD or HbA1c-adjSD vs HbA1c-Mean) adjusted model, adjusted for the same confounders as model 1 plus HbA1c-Mean in HbA1c-SD and HbA1c-adjSD; in HbA1c-Mean, multiple confounders plus either HbA1c-SD or HbA1c-adjSD (shown with hashtag). Data management and analysis were performed using SAS 9.4 software (SAS Institute, Cary, NC, USA). The signi cance level was set at a P-value <0.05 for two-sided testing.

Results
As shown in Figure 1, a total of 8636 patients were enrolled in the program. We excluded 285 patients with a history of HF before enrollment, 44 patients who developed HF within 12 months after enrollment, 195 patients with age at enrollment of >80 years, 516 patients with age at diabetes onset of <30 years, 170 patients with type 1 DM, 246 patients with cardiac dysrhythmia, 4 patients with congenital heart disease, 55 patients with valvular heart disease, 173 patients with eGFR <30 mL/min/1.73m 2 , and 3124 patients who underwent HbA1c measurement fewer than three times. In total, 3824 patients were identi ed for analyses in this study; among these, 315 patients who developed HF during the 11.72 years observation period.

Discussion
This study shows that raised mean HbA1c level was associated with a higher risk to develop HF. The effect of HbA1c variability on HF was crucial in patients with mean HbA1c <7%. The associated risk of HF in patients with mean HbA1c <7% and a greater HbA1c variability was comparable to patients with HbA1c ≥8% irrespective of HbA1c variability.
Many previous studies revealed that HbA1c level is positively associated with HF in patients with T2DM [8][9][10][11][12]. However, these studies did not take HbA1c variability into consideration for the adjustment. So far, there were a few studies to evaluate the association of HbA1c variability with HF. A longitudinal study with small patient numbers (201 subjects) and limited HF events (18 events) showed a positive association between HbA1c variability and new-onset HF [19]. Another study focused on the prediction of incident HF by mean HbA1c, additionally reported that less HbA1c variability has a less incidence of HF [20]. Again, this study did not adjust for mean HbA1c in the analyses of the association between HbA1c variability and HF. In our study, the adjusted model showed that both HbA1c variability and mean HbA1c level independently predict HF. In mutually adjusted models, however, only mean HbA1c revealed a consistent dose-response association, while the association of HbA1c variability with HF disappeared, which indicated that the association of HbA1c variability was largely explained by mean HbA1c.
Nevertheless, HbA1c variability is still important among patients with well glycemic control. The risk of HF in well glycemic control patients with greater HbA1c variability was comparable to patients with poor glycemic control. Our study supported that further targeting on HbA1c stability after well glycemic control is crucial to reduce the risk of HF.
The underlying mechanism(s) to explain why the effect of absolute HbA1c level on HF was greater than HbA1c variability remains unclear. An excess production of reactive oxygen species has been reported to be an important pathophysiology to develop HF [21]. A previous study showed that acute (24 hours) glucose uctuations activate more oxidative stress than chronic hyperglycemia (by using one baseline HbA1c level) [22]. Another study also revealed that acute (24 hours) glucose oscillations produce more oxidative stress than 24 hours mean blood glucose level at 180 mg/dL [23]. However, these two studies were unable to explain our results because the glucose exposure period was rather short in comparison with our study. The possible mechanism(s) needs further investigation in the future.

Study Strengths and Limitations
The strength of this study included a large sample size with long-term follow-up of real-world data. Nevertheless, our study had some limitations. First, the patients in this study were from a tertiary referral outpatient clinic. They are usually more complicated. Therefore, our ndings may not be generalizable to all patients with diabetes. Second, most of our patients were on sulfonylureas and used metformin and statins less; such medication regimens are different from current managements. This difference should be considered before generalizing our results to patients with current medication patterns. Finally, as the inherent limitation of a database, we cannot categorize HF into HF with preserved ejection fraction and HF with reduced ejection fraction; echocardiography and laboratory data are also lack to con rm the diagnosis.

Conclusions
Our study shows that higher mean HbA1c was independently associated with increased risk of HF regardless of HbA1c variability. In patients with well glycemic control, less HbA1c variability was pivotal for the development of HF.

Declarations Acknowledgments
We thank the China Medical University branch center of Taiwan Health and Welfare Data Science Center to organize the use of the database of Taiwan National Health Insurance.
Author contributions YTL, WLH, and CCC designed the study. YTL, WLH, HPW, and CCC analyzed and interpreted the data.
MPC and CCC drafted the manuscript, which was critically revised by CCC; CCC approved the manuscript and is the guarantor of this work, and takes responsibility for the integrity of the data and the accuracy of the data analyses. All authors read and approved the nal manuscript.

Funding
This study is supported by a grant from the China Medical University Hospital, Taiwan (DMR-109-030).

Availability of data and materials
This study use the database of the Taiwan National Health Insurance (https://dep.mohw.gov.tw/DOS/lp-2501-113.html) which are available following application for Taiwanese researchers.

Ethics approval and consent to participate
The Ethical Review Board of China Medical University Hospital in Taiwan (CMUH107-REC2-163) approved the study before commencement.

Consent for publication
Not applicable.

Competing interests
The authors declare that they have no competing interests.