Risk stratification of cardiac arrhythmias and sudden cardiac death in type 2 diabetes mellitus patients receiving insulin therapy: A population‐based cohort study

Abstract Introduction Metabolic abnormalities may exacerbate the risk of adverse outcomes in patients with type 2 diabetes mellitus. The present study aims to assess the predictive value of HbA1c and lipid variability on the risks of sudden cardiac death (SCD) and incident atrial fibrillation (AF). Methods The retrospective observational study consists of type 2 diabetic patients prescribed with insulin, who went to publicly funded clinics and hospitals in Hong Kong between January 1, 2009 and December 31, 2009. Variability in total cholesterol, low‐density lipoprotein‐cholesterol (LDL‐C), high‐density lipoprotein‐cholesterol (HDL‐C), triglyceride, and HbA1c were assessed through their SD and coefficient of variation. The primary outcomes were incident (1) ventricular tachycardia/ventricular fibrillation, actual or aborted SCD and (2) AF. Results A total of 23 329 patients (mean ± SD age: 64 ± 14 years old; 51% male; mean HbA1c 8.6 ± 1.3%) were included. On multivariable analysis, HbA1c, total cholesterol, LDL‐C and triglyceride variability were found to be predictors of SCD (p < .05). Conclusion HbA1c and lipid variability were predictive of SCD. Therefore, poor glucose control and variability in lipid parameters in diabetic patients are associated with aborted or actual SCD. These observations suggest the need to re‐evaluate the extent of glycemic control required for outcome optimization.

cholesterol, LDL-C and triglyceride variability were found to be predictors of SCD (p < .05).
Conclusion: HbA1c and lipid variability were predictive of SCD. Therefore, poor glucose control and variability in lipid parameters in diabetic patients are associated with aborted or actual SCD. These observations suggest the need to re-evaluate the extent of glycemic control required for outcome optimization. accounting for approximately 25% of deaths of cardiovascular origin. 1 By contrast, atrial fibrillation (AF), the most common sustained arrhythmia amongst adults, is increasingly prevalent around the world, particularly amongst developed countries. 2,3 Significant increases in morbidity and mortality amongst large-scale epidemiological studies have been demonstrated amongst AF patients, including an increase in SCD risk. [4][5][6][7] Furthermore, type 2 diabetes mellitus increases the risk of SCD, as demonstrated by a recent meta-analysis of populationbased prospective studies. 8 With the global shift towards a more personalized approach in the management of diabetes, there is an increasing interest in exploring the application of new parameters, such as HbA1c and lipid variability, to better monitor disease progression and evaluate the prognosis. Although the exact mechanisms remain unclear, increased long-term glycemic and lipid variability is hypothesized to lead to endothelial dysfunction via an increase in oxidative stress. [9][10][11] Since hemoglobin has an average lifespan of 100 days, HbA1c can reflect glycemic control in recent months. Therefore, HbA1c variability is not affected by short-term glycemic changes due to diet and medication changes, thus it is a better representation of long-term glycemic variation. However, existing studies have focused on risk prediction of allcause mortality and general cardiovascular adverse events, [12][13][14][15][16] with a limited number of studies exploring specifically arrhythmic risks amongst diabetics. Moreover, those type 2 diabetics who are partially or fully dependent on insulin are more likely to have severe disease and may be at higher risks of arrhythmias.
The aim of the present study was to assess the predictive value of HbA1c and lipid variability towards aborted or successful SCD, as well as incident AF in type 2 diabetic patients receiving insulin therapy.  [17][18][19][20] including diabetes mellitus, 21 in the past.

| Patient data
Clinical and biochemical data of the present cohort were extracted from CDARS. The outcomes of the present study are the occurrence of SCD and AF from January 1, 2009 to December 31, 2019. SCD is defined as episodes of VT, VF, or nonspecific cardiac arrest, which were diagnosed under clinical judgment with electrocardiographic or biochemical findings and subsequently coded into hospital records.
Both aborted and actual SCD events were included. VT/VF on electrophysiological study is not included. Hypoglycemia-induced cases of SCD or AF were defined as cases with dextrose infusion during the admission episode or had blood glucose measured ≤3.9 mg/mmol. Demographic details, including age and sex, were extracted. Patients were categorized into four groups based on their age: below age 55, between ages 55 and 64, between ages 65 and 74, above and include age 75. The number of baseline acute hospitalization episodes between January 1, 2004 and December 31, 2008 was also obtained.

| Statistical analysis
Statistical analysis was performed using R Studio, and statistical significance was defined as p value <.05. Kaplan-Meier survival curve was used to portray the difference in actual or aborted SCD and AF survival between patients of different age groups, with the statistical significance of the intergroup difference evaluated using the log-rank test. Temporal variability of HbA1c and lipid indices were evaluated using calculated parameters of SD and coefficient of variation (CV). CV was measured by multiplying 100-fold of the value calculated by the ratio SD and mean. SD and CV were used to measure variability since they were less affected by outliers. Whilst SD is independent of the mean, CV is independent of the scale thus more sensitive to small changes to the mean. To identify predictors for shorter time to aborted or actual SCD and AF occurrence, univariable Cox regression was first applied to clinical and biochemical parameters. Patients with missing data were excluded from the analysis. Furthermore, due to their limited prescription towards the study cohort, GLPA and meglitinide were not included. Subsequently, parameters with p value <0.10 were included in the multivariable Cox regression model. Only patients with no missing data for the selected parameters, and at least three measurements for the selected variability predictors, were included in the multivariate models. No data imputation was performed.
The inter-relations between HbA1c and lipid variability with intermittent hypoglycemia were evaluated using logistic and Poisson regression. Logistic regression was also used to examine the relationship between baseline hypoglycemia frequency and (1) occurrence of SCD/AF or (2) aborted or actual SCD/AF episodes that were associated with hypoglycemia. Odds ratios (ORs) were reported from logistic and Poisson regression, whereas hazard ratios were reported from Cox regression, along with the 95% confidence intervals.

| Predictors of SCD and AF
In the present cohort, 10.3% of patients suffered from at least one SCD event, and 7.7% suffered from AF. Amongst these cases, 25 and 10 cases of patients experiencing aborted or actual SCD and AF were associated with hypoglycemia on admission, respectively. Supplementary    Table 4, where several significant pre-

| DISCUSSION
The present study demonstrated the following major findings: (1)  variability is associated with hypoglycemia frequency. To the best of our knowledge, the present study is the first to report an association between increased variability in HbA1c and lipid markers with increased risk for VT/VF/SCD amongst diabetic patients.
The prognostic values of HbA1c and lipid variability have been increasingly explored over the past decades. However, prior studies have mostly focused on the prediction of all-cause mortality or cardiovascular events 12,22 with only a limited number of studies focusing on arrhythmic or SCD outcomes. 23 Although the underlying pathophysiology remains incompletely elucidated, there are several possible contributing factors towards the increased arrhythmic risk amongst patients with high glycemic and lipid variability. First, increased glycemic variability is found to be associated with QTc prolongation and increased QTc dispersion, which greatly elevates the risk of ventricular tachyarrhythmia. 24,25 There is evidence suggesting that QTc prolongation may be triggered by spontaneous hypoglycemia due to underlying coronary atherosclerosis or cardiac autonomic neuropathy. [26][27][28][29] Antidiabetic agent use may also play a role in the prognostic value of glycemic variability. Biguanide users are likely more stable or earlier in the disease course, thus have a lower cardiovascular disease burden. Sulphonylurea use, which was predictive of SCD in the present study, is known to have an increased risk of hypoglycemia. 30 In addition, it has been reported human ether-a-go-go-related gene (hERG) channel inhibitory effects of some sulphonylurea, which can lead to QT prolongation. 31 Amongst patients on insulin, who have more labile glucose control, the spontaneous glycemic fluctuations can induce the occurrence of arrhythmia. Unfortunately, continuous blood glucose monitoring was not available in the present study to demonstrate the association between spontaneous glycemic changes and arrhythmic episodes.
Furthermore, structural remodeling as a result of chronic hyperglycemia is also involved in the pathogenesis. [32][33][34][35][36][37] HbA1c variability has been associated with remodeling and fibrosis of the atria and ventricles, which could be arrhythmogenic. [38][39][40] In a recent nationwide observational study, an association between high lipid variability and increased risk of new-onset AF was reported, and statins protected against AF development via the reduction in adverse atrial remodeling. 41 Moreover, frequent intermittent hypoglycemia can induce the release of reactive oxygen species, 42 thereby leading to increased oxidative stress, chronic inflammation, and endothelial dysfunction. 43 Hypoglycemia itself is arrhythmogenic and can reduce the myocardial tolerance to ischemia and reperfusion injuries. 44,45 Similar to glycemic variability, the increase in oxidative stress with fluctuations in lipid levels due to atherogenic substance release from unstable plaque is hypothesized to underlie the increased arrhythmic risk. 12 Indeed, glycemic fluctuations were found to increase the formation of atherosclerotic plaques and thinning the fibrous cap, which suggests that intermittent hypoglycemia may contribute to lipid variability as well. 46 It should be noted that whilst triglyceride and HDL-C variability are dependent on glycemic control and other lifestyle factors, the use of statin plays a significant role in LDL-C variability. The significant interpersonal variability, as well as the varying effects between different types of statins on LDL-C variability, reflects the need for further research on the area. 47,48 The change in HbA1c and lipid variability from predictive of SCD, in univariable analysis, to being protective under multivariable analysis, can be attributed to several causes. The limitation of cohort size and multivariate analysis may have selected for patients of more advanced disease and undergone aggressive control. Given that a Jshaped association between adverse outcome and both glycemic and lipid indices have been described, patients with high variability that returned to the optimal glycemic and lipid range would have had a better prognosis. 16,49,50 Indeed, the duration of exposure to an optimal glycemic range is inversely associated with diabetic retinopathy progression, even after accounting for the effects of glycemic variability. 51 Additionally, Ceriello et al. reported that patients with elevations in both HbA1c and HDL-C variability were at higher risk for diabetic nephropathy than those with high variability in only one variable, highlighting the interacting effects between variability markers. 52 Therefore, the protective value may be a result of inevitable selection bias, the protective effects of pharmacotherapy, and the interactions between different indices. 53,54

| Strengths and limitations
There are four major strengths for the present study: (1)  However, several limitations should be recognized. Firstly, the cohort was limited to type 2 diabetic patients prescribed with insulin, which can limit the generalizability of the findings. Given that insulin is only prescribed for diabetic patients in later stages, an advanced disease state can be inferred for the selected patients. Secondly, the observational nature of the present study leads to inevitable errors from missing data, coding errors, and under-coding. A causational relationship cannot be established from the findings of the observational study, which can only demonstrate associations. Furthermore, unfortunately, ICD coding does not reflect the frequency of events, thus the frequency of SCD events was not evaluated. The coding also does not reflect whether the VT is sustained or associated with hemodynamic collapse. It is based on the assumption that coded VT is clinically significant as nonsustained VT would be clinically irrelevant thus not coded into the database. It should be emphasized that a diagnosis of VT is not the same as SCD, hence the effect of long-term glycemic variability on the risk of SCD should be interpreted with caution.
Novel therapies, such as GLPA (n = 9) and SGLT-2 transporter inhibitor (n = 0), were not assessed due to the fact that this retrospective study recruited patients in the year 2009, at which these agents were not yet developed. Finally, data on blood pressure, body mass index, echocardiogram, severity of HF and lifestyle were absent, which can affect the patients' cardiovascular health.

| CONCLUSION
Poor glucose control and variability in lipid parameters in diabetic patients are associated with SCD. These observations suggest the need to re-evaluate the extent of glycemic control required for outcome optimization. Further studies on the predictive value of variability in other glycemic measures, such as fasting and random blood glucose, in addition to other methods of measuring variability, should be performed to further examine the predictiveness of glycemic variability towards arrhythmias in diabetic patients.

CONFLICTS OF INTEREST
No conflicts of interest.

DATA AVAILABILITY STATEMENT
The deidentified dataset arising from this study is available upon request. ORCID