Diabetes and atrial fibrillation: does the type of diabetes matter?

This editorial refers to ‘Prognostic impact of type 1 and type 2 diabetes mellitus in atrial fibrillation and the effect of severe hypoglycaemia: A nationwide cohort study’, by S. Karayiannides et al., https://doi.org/10.1093/eurjpc/zwac093. Atrial fibrillation (AF) and diabetes are common chronic diseases that frequently co-exist. The presence of both diseases substantially increases the risk of stroke, heart failure (HF), and all-cause and cardiovascular mortality, compared with the presence of either disease alone.1,2 The interaction between diabetes and AF has been best documented in the CHA2DS2-Vasc score, where diabetes is an important modifier of stroke risk in AF patients.3 Further, AF patients with diabetes often experience worse AF symptoms and lower quality of life and are more likely to have persistent and permanent AF, compared with AF patients without diabetes.4 Evidence suggests that diabetes is responsible for promoting atrial structural, electrical, electromechanical, and autonomic remodelling, all of which aggravate and maintain the AF substrate.1 Thus, diabetes is a well-established indicator of poor prognosis in AF patients. Diabetes is a heterogenous disease with different pathogeneses and risk factors between diabetes mellitus Type 1 (T1DM) and Type 2 (T2DM). Most evidence indicates a strong relationship between T2DM and AF. Type 2 diabetes mellitus is more prevalent among AF patients, in part due to similar risk factors for both diseases, particularly advancing age. Risk factors for stroke in T1DM and T2DM patients also differ. Reduced physical activity, elevated body mass index, and dyslipidaemia in T2DM patients are associated with an increased risk of stroke,1 while longer duration of time a patient lives with AF is a risk factor for stroke in T1DM.5 In general, there is limited evidence that Type 1 diabetes is a prognostic indicator for stroke or mortality in AF patients. In the current issue of the European Journal of Preventative Cardiology, Karayiannides et al.6 investigated potential effect modification by type of diabetes (T1DM or T2DM) on the association between diabetes and the incidence of all-cause mortality, myocardial infarction (MI), ischaemic stroke, HF diagnosis, and dementia diagnosis among AF patients. Swedish National Registries were used to create an AF population-level cohort (2013–17) and the exposure of T1DM or T2DM was determined at cohort entry (i.e. first AF diagnosis).6 Of 60 294 patients with AF and diabetes, only 2221 (3.7%) patients had T1DM. Baseline characteristics, age, and medications differed substantially between patients with T1DM and T2DM.6 Compared with the reference group of AF patients without diabetes, patients with T1DM had a higher risk of all-cause mortality [hazard ratio (HR) 1.87 (95% confidence interval (CI) 1.73–2.02) vs. HR 1.51 (95% CI 1.47–1.55)], MI [HR 2.49 (95% CI 2.17–2.85) vs. HR 1.70 (95% CI 1.59–1.81)], and dementia [HR 1.46 (95% CI 1.15–1.85) vs. HR 1.28 (95% CI 1.18–1.40)] than T2DM patients (T1DM vs. T2DM respectfully).6 A history of severe hypoglycaemia appeared to augment the risk of all-cause mortality and dementia in T2DM diabetes patients, however, the effect diminished when adjusted for prior MI.6 The authors should be commended for addressing an important, perhaps overlooked, knowledge gap when determining prognoses for patients with diabetes and AF. As the prevalence of AF and diabetes increase with the aging population, it is important to identify and quantify not only individual risk factors for major cardiovascular events and dementia, all of which add substantial burden to the healthcare system, but also potential interactions. The strength of their research design is the use of national administrative data sources, which allows a timely and cost-efficient answer to their research question. These databases are well-validated population-based registries with virtually complete follow-up and exact censoring information, thereby reducing the risk of selection bias.7 In addition, the size of these databases allows for adequate statistical power to investigate interactions that would otherwise be difficult to ascertain in smaller clinical datasets. As randomized trials cannot be performed, Karayiannides et al. is an important step to understand if a potential difference exists for the risks of cardiovascular events and dementia among AF patients with T1DM or T2DM. A note of caution, however, similar studies to Karayiannides et al. with an emphasis on study quality should be performed to reproduce and build on these findings to (i) limit selection bias, misclassification, and confounding and (ii) ensure accurate capture of incidence of diseases. In general, observational studies need to be concerned about selection bias, misclassification, and confounding. As mentioned previously, the current research design and data sources have limited the impact of selection bias. While misclassification of diabetic status and outcomes may be expected to be minimal, there is concern about misclassification of important time-varying confounders. Oral anticoagulation (OAC) is an important aspect of therapeutic stroke management for AF patients. Oral anticoagulation use was assessed at baseline and a patient was considered exposed or unexposed throughout the follow-up period. Patients who initiated OACs later during follow-up were always considered unexposed and patients who stopped OACs were always classified as exposed. The authors suggest that the utilization of OACs is low at baseline, which is plausible as patients entered the cohort and started follow-up at AF diagnosis. It is likely that many patients initiated OACs after cohort entry (i.e. date of AF diagnosis), especially since the diabetes increases the CHA2DS2-Vasc score, and that some patients stopped or switched OACs during follow-up. Oral anticoagulation use could have been addressed with the use of binary time-varying covariates for warfarin and direct acting OAC (DOAC) use. Accounting for time-varying use of OACs is important to determine the real-world risk of events. Diabetes increases the risk of stroke by 70% in AF patients.8 Compared with AF and diabetes patients prescribed warfarin, DOACs reduce strokes or systemic embolic events by 20%, intracranial bleeding by 43% and cardiovascular death by 17%.8 Therefore, accounting for AF management with time-varying OAC use would illustrate the real-world risk of cardiovascular events and dementia in AF patients with T1DM and T2DM. Similar to OACs, comorbidities that may alter the risk of the outcome, such as the CHA2DS2-vasc or HAS-Bled variables, may be captured during follow-up with time-varying covariates to determine the dynamic risks for the outcomes. The presence of diabetes and AF increases the likelihood of additional comorbidities, therefore residual confounding may exist for a longitudinal observational study. Special consideration is required when capturing the incidence or first diagnoses of diseases, such as dementia. Dementia develops over many years and may first appear as cognitive decline. As Karayiannides et al. follow patients for only 1–2 years after AF diagnosis, it is not biologically plausible that AF with diabetes caused the dementia. Patients diagnosed within 1–2 years after AF diagnosis likely had cognitive decline prior due to other pre-existing risk factors. A latency period of at least 1 year would be required to capture new diagnoses of dementia that may be associated with AF and diabetes. National registries and longitudinal community studies have been the source of evidence to assess the risk of dementia with AF and diabetes and follow-up for most of these studies ranged from >5 to >20 years.9 Thus, future studies with extended follow-up periods are warranted to assess the association between the type of diabetes and the incidence of dementia. To our knowledge, only a Danish National Registry study (N = 10 058) evaluated the potential heterogeneity of effect in AF patients with T1DM and T2DM, however, only the outcome of thromboembolism was assessed.10 Similar to the results in Karayiannides et al., the Danish study found no statistically significant difference in the risk of thromboembolism between patients with T1DM and T2DM [HR 1.15 (95% CI 0.91–1.44)]; however, analyses stratified by age indicated a statistically significantly increased risk of thromboembolism among patients <65 years old with T2DM compared with T1DM [HR 1.97 (95% CI 1.07–3.61)].10 It should be noted that unlike the Karayiannides et al., the Danish study excluded patients prescribed OACs.6 In conclusion, the results of the population-level study by Karayiannides et al. raise an important question about potential heterogeneity in AF prognosis by type of diabetes. Further high-quality evidence is warranted to build on these results to provide meaningful insights into the prognoses for AF patients with T1DM and T2DM. No new data were generated or analyzed in support of this research.