Atrial fibrillation: age at diagnosis, incident cardiovascular events, and mortality

Abstract

Background and Aims

Patients with atrial fibrillation (AF) are at increased risks of cardiovascular diseases and mortality, but risks according to age at diagnosis have not been reported. This study investigated age-specific risks of outcomes among patients with AF and the background population.

Methods

This nationwide population-based cohort study included patients with AF and controls without outcomes by the application of exposure density matching on the basis of sex, year of birth, and index date. The absolute risks and hazard rates were stratified by age groups and assessed using competing risk survival analyses and Cox regression models, respectively. The expected differences in residual life years among participants were estimated.

Results

The study included 216 579 AF patients from year 2000 to 2020 and 866 316 controls. The mean follow-up time was 7.9 years. Comparing AF patients with matched controls, the hazard ratios among individuals ≤50 years was 8.90 [95% confidence interval (CI), 7.17–11.0] for cardiomyopathy, 8.64 (95% CI, 7.74–9.64) for heart failure, 2.18 (95% CI, 1.89–2.52) for ischaemic stroke, and 2.74 (95% CI, 2.53–2.96) for mortality. The expected average loss of life years among individuals ≤50 years was 9.2 years (95% CI, 9.0–9.3) years. The estimates decreased with older age.

Conclusions

The findings show that earlier diagnosis of AF is associated with a higher hazard ratio of subsequent myocardial disease and shorter life expectancy. Further studies are needed to determine causality and whether AF could be used as a risk marker among particularly younger patients.

Structured Graphical Abstract

Methods and main findings in the study investigating associations of cardiovascular disease and mortality in accordance with age per decade among AF patients and matched controls from the background population. AF, atrial fibrillation; CI, confidence interval.

Open in new tabDownload slide

See the editorial comment for this article ‘Atrial fibrillation and long-term cardiovascular outcomes: bringing the whole picture into focus', by B.A. Steinberg, https://doi.org/10.1093/eurheartj/ehae347.

Introduction

Atrial fibrillation (AF) is the most common cardiac arrhythmia, and it is associated with increased risk of cardiomyopathy, heart failure (HF), ischaemic stroke, and premature mortality.¹ Atrial fibrillation incidence is increased by traditional clinical risk factors such as ischaemic heart disease (IHD), hypertension, diabetes, body weight, and age.² Additionally, studies have shown that AF has a complex genetic component, including genes typically involved in arrhythmia syndromes, cardiac structure, and cardiomyopathy [including dilated cardiomyopathy (DCM) and hypertrophic cardiomyopathy (HCM)], especially among individuals with onset of disease at ≤65 years.^3–8 Thus, it has been suggested that AF, among some younger patients, might be the initial indication of serious cardiac disease (e.g. risk marker).^6,9 Although AF has been extensively studied, its association with subsequent cardiovascular diseases and mortality according to age at diagnosis is not entirely understood. Previous observational studies have not been powered to investigate risks of subsequent outcomes across the range of age at diagnosis, and furthermore, there has not been broad attention concerning early-onset AF. Presently, the genetic knowledge of AF among especially younger but also older patients is steadily improving. Therefore, the objective of this study was to examine how time of AF debut influences the risk of cardiovascular disease, aiming to provide insight into the interpretation of age at diagnosis and AF as an early marker of subsequent morbidity and mortality.

This nationwide study investigated age-specific associations of selected cardiovascular diseases and all-cause mortality among AF patients using highly validated Danish registers. We examined the risks of any cardiomyopathy, DCM, HCM, HF, ischaemic stroke, and all-cause mortality. Finally, we estimated average life expectancy, including life years lost, a method evaluating age at diagnosis of disease, and the reduction in life expectancy.

Methods

The study was approved by the Capital Region of Denmark (approval number: P-2019-348). In Denmark, register-based studies on de-identified data do not require ethical approval.

Data sources

Every Danish resident is provided with a unique personal identification number, which can be used to crosslink various national registers.¹⁰ Information on sex, date of birth, time of death, and migration was obtained from the Danish National Population Registry. The Danish National Patient Registry was accessed to obtain data on diagnoses and surgical procedures made during hospital admission and emergency department visits since 1978 and outpatient visits from 1995 until 2021.¹¹ Diagnoses were classified according to the Danish modification of the International Classification of Diseases, 8th and 10th revisions (ICD-8 and ICD-10, respectively), and all surgical procedures have been registered according to the Nordic Medico-Statistical Committee (NOMESCO) classification (see Supplementary data online, Table S1). The Danish National Prescription Registry holds information on all prescribed drugs sold in pharmacies since 1994 in accordance with the Anatomical Therapeutic Chemical Classification System (ATC-code).¹²

Study design and population

We designed a nationwide retrospective population-based cohort study in Denmark including individuals diagnosed with AF through >20 years. All individuals between ≥20 and <90 years of age at AF diagnosis from 1 January 2000 until 30 November 2020 were identified (for flowchart, Supplementary data online, Figure S1). To reduce potential bias introduced by loss to follow-up, analyses were restricted to individuals residing in Denmark at the time of diagnosis and without multiple migrations leading to unaccounted locations. Individuals with HF, any cardiomyopathy, stroke (ischaemic and haemorrhagic), or death at the index date were excluded. Every individual was followed from the date of AF diagnosis (index date) until the outcome of interest, emigration, or end of study (31 December 2021), whichever occurred first.

Data were assessed in two cohorts: the ‘entire cohort’ included all AF patients, whereas the ‘low-morbidity cohort’ included AF patients without previous comorbidities and traditional cardiovascular risk factors, including those used in the CHARGE-AF score² (see Supplementary data online, Table S1). Using exposure density matching, we matched each AF patient with four controls from the background population by sex, year of birth, and index date. Controls in both cohorts were required to be alive and not diagnosed with AF nor outcomes at the index date. Controls in the low-morbidity cohort were furthermore required to be without the aforementioned risk factors or comorbidities at the index date. The index date for individuals in the control groups corresponded to the diagnosis date for each matched patient. Diseases were defined from diagnosis and procedure codes given before or on the index date. Previous studies have validated the diagnoses in the registers with high positive predictive values.^13–15 Medication was defined as a claimed prescription within 180 days prior to the index date.

Statistical analyses

Baseline characteristics are presented as descriptive data, shown as counts with proportions and medians with interquartile ranges. Crude cumulative incidences of cardiovascular diseases were estimated using the Aalen–Johansen method, accounting for death as a competing risk. The Kaplan–Meier estimator was used to evaluate all-cause mortality. The 95% confidence intervals (CIs) were calculated according to the Wald test. Applying Cox proportional hazard models, we estimated hazard ratios (HRs) of outcomes among AF patients compared with controls. The models were stratified by matched individuals and did not adjust for comorbidities but instead included a low-morbidity cohort in agreement with statistical advice. Time since index date was used as the time scale. Atrial fibrillation status was coded as different, pre-specified age groups (≤50, >50–60, >60–70, >70–80, >80 years) according to age at AF diagnosis for patients, or as controls. The likelihood ratio test was applied to assess the models for interactions with age and sex. The Cox proportional hazard assumption was violated for the exposure variable; hence, we performed sensitivity analyses with time splits.

We computed age-specific life expectancy among patients and controls, including differences between the groups (life years lost). The method has previously been applied and described in detail.^16–23 In summary, the conditioned survival analyses estimate the average remaining life expectancy at a specific age and before a maximum age (for this study: 100 years). Individuals in the exposure group are diagnosed with AF and alive at the specified ages, while individuals in non-exposure groups are all required to be alive without AF at the specified age. The 95% CIs were calculated using bootstrapping.

We did not adjust for multiplicity as the study investigated overall patterns in an exploratory manner. Data analysis was conducted using the R statistical software (version 4.2.1) and the survival (version 3.3.1), prodlim (version 2019.11.13), and lillies (version 0.2.9)¹⁹ packages.

Secondary analyses

Analyses stratified according to sex

Sex modified the associations of cardiomyopathy, HCM, HF, ischaemic stroke, and mortality but not of DCM (applying the likelihood ratio test, P-values = .001, <.001, <.001, .008, <.001, and .46, respectively). We therefore performed analyses stratified by sex.

Sensitivity analyses

Because of non-proportionality, we performed sensitivity analyses (i) in the entire cohort with time splits (0–1 years, >1–5 years, >5 years) based on Schoenfeld residuals and clinically relevant intervals. Second, we conducted sensitivity analyses (ii) with the application of a 90-day grace period from the date of AF diagnosis, excluding AF patients with outcomes and death within this period. New controls were matched for each sensitivity cohort (‘S’), named ‘Cohort SE’ and ‘Cohort SL’, respectively. Cox models corresponding to the main analyses were used. Third, we performed sensitivity analyses (iii) using the main models in the entire cohort with adjustment for comorbidities (hypertension, IHD, acute myocardial infarction, chronic obstructive pulmonary disease, diabetes, valvular heart disease, congenital heart disease, thyroid disease, peripheral artery disease, chronic kidney disease, and malignancy).

Results

Study population

The entire study population (‘entire cohort’) consisted of 216 579 patients diagnosed with AF between 20 and 90 years of age at diagnosis and 866 316 controls matched by sex, year of birth, and index date from the background population. Analyses were repeated in a cohort including 90 321 patients and 361 284 controls without comorbidities and traditional risk factors for AF (‘low-morbidity cohort’). In general, the patterns from the low-morbidity cohort were comparable with the entire cohort; hence, we will primarily present the findings from the entire cohort throughout the text, while observations in both cohorts are reported in figures and tables.

In the entire cohort, the median age at AF diagnosis was 72.2 years (histogram; Supplementary data online, Figure S2) and 55.5% were males. The mean follow-up time was 7.9 years and 356 818 died during follow-up. The baseline characteristics with complete data in both cohorts are shown in Table 1, and characteristics of the cohorts divided into age groups are available in Supplementary data online, Tables S2 and S3. Overall, when grouped by age at diagnosis, males were overrepresented in the four youngest age groups (≤50, >50–60, >60–70, and >70–80 years), whereas females represented the majority among the oldest patients (>80 years). The proportion of patients with male sex decreased from 70% males among the youngest patients to 42% males among the oldest patients. The prevalence of all comorbidities apart from congenital heart disease increased with age.

Incidences of cardiovascular diseases and mortality according to age at diagnosis

We investigated incidence rates and the absolute risks (cumulative incidences) of any cardiomyopathy, DCM, HCM, HF, ischaemic stroke, and all-cause mortality at 10 years, stratified into age groups according to the age at AF diagnosis (Figure 1). Plots displaying the cumulative incidences for the entire study period are provided in Supplementary data online, Figures S3–S6. Incidence rates and event counts for the entire study period are presented in Supplementary data online, Tables S4 and S5, while subtypes of cardiomyopathy and HF can be accessed in Supplementary data online, Tables S6 and S7.

The observations in both cohorts showed comparable patterns; with the estimates in the low-morbidity cohort being generally more attenuated. The absolute risks of cardiomyopathy, including HCM and DCM, tended to be higher among younger AF patients compared with older AF patients, whereas the opposite was observed among controls. Summarizing the results in the entire cohort, the risks of cardiomyopathy, DCM, and HCM, respectively, ranged from 1.81% (95% CI, 1.57%–2.05%), 1.02% (95% CI, 0.84%–1.20%), and 0.36% (95% CI, 0.26%–0.47%) among the youngest AF group (≤50 years) to 0.63% (95% CI, 0.55%–0.70%), 0.31% (95% CI, 0.26%–0.36%), and 0.19% (95% CI, 0.15%–0.23%) among the oldest AF group (>80 years). The absolute risks of HF, ischaemic stroke, and mortality showed a stepwise increase in risks among AF patients and controls from younger to older age at diagnosis. The risk of HF among AF patients increased from 6.69% (95% CI, 6.24%–7.14%) in the youngest group to 22.2% (95% CI, 21.8%–22.6%) in the oldest group, and ischaemic stroke risk among AF patients increased from 1.59% (95% CI, 1.36%–1.83%) in the youngest group to 6.92% (95% CI, 6.67%–7.17%) in the oldest group.

The study reports a stepwise, inverse relationship for cardiomyopathies with the highest cumulative incidence differences (excess risks) identified among younger patients and the lowest excess risks identified among older patients (Figure 2). The excess risk of HF and mortality generally increased from younger to older age. For ischaemic stroke, the lowest excess risk was observed in the youngest age group in both cohorts. The corresponding patterns in the two cohorts differed slightly; in the entire cohort, the highest excess risk was observed in the group aged >50–60 years, while the risks across the remaining age groups were comparable. The risks in the low-morbidity cohort showed a marginal stepwise risk increase.

Hazard ratios of cardiovascular diseases and mortality according to age at diagnosis

The estimated HRs of cardiovascular diseases and mortality among AF patients compared with controls in both cohorts grouped by age at diagnosis are presented in Figure 3. The associations of all investigated outcomes were modified by age (likelihood ratio test, all P-values < .001).

All outcomes showed increased rates among AF patients compared with controls regardless of age at diagnosis. We observed a stepwise change in HRs of any cardiomyopathy, DCM, HCM, and HF, with increasing ratios according to younger age at AF diagnosis compared with respective controls. In the entire cohort, comparing the youngest and the oldest age groups (≤50 and >80 years, respectively), the HRs of any cardiomyopathy ranged from 8.90 (95% CI, 7.17–11.0) to 2.90 (95% CI, 2.46–3.42), DCM ranged from 9.76 (95% CI, 7.25–13.1) to 3.77 (95% CI, 2.94–4.83), HCM ranged from 6.58 (95% CI, 4.31–10.1) to 2.40 (95% CI, 1.81–3.19), and HF ranged from 8.64 (95% CI, 7.74–9.64) to 4.11 (95% CI, 3.99–4.24). The findings were similar in both cohorts for all myocardial phenotypes, except for HCM, which showed a stepwise pattern in the entire cohort, but a marginal U-shaped relationship in the low-morbidity cohort. The HRs of ischaemic stroke and mortality among AF patients compared with controls were consistently increased across all age groups in both cohorts.

Average life expectancy according to age at diagnosis

Individuals diagnosed with AF displayed fewer expected remaining life years compared with controls (Figure 4). The average loss in life years was, naturally, largest among younger individuals and lowest among older individuals. Summarized, AF patients in the entire cohort were predicted to have an average life reduction of 11.3 years (95% CI, 11.0–11.6) when diagnosed at age ≤30 years, 7.4 years (95% CI, 7.3–7.5) when diagnosed at age ≤60 years, and 3.6 years (95% CI, 3.5–3.6) when diagnosed at age ≤80 years.

Secondary analyses

Absolute risks and hazard ratios stratified according to sex

Considering the results in the entire cohort stratified by sex, we present absolute risks (see Supplementary data online, Figure S7) and excess risks (see Supplementary data online, Figure S8) of all investigated outcomes among male and female patients compared with their respective controls. Absolute risks and excess risks of cardiomyopathy, HF, and mortality were generally higher among males, whereas the excess risk of ischaemic stroke was higher among females.

When studying HRs, we observed a modification by sex for associations of cardiomyopathy, HCM, HF, ischaemic stroke, and mortality but not for DCM (see Supplementary data online, Figure S9). However, the sex-specific HRs among male and female patients compared with controls were overall comparable. There was a general trend of higher HRs of cardiomyopathy and HCM among males, while females showed higher HRs of HF and ischaemic stroke. For mortality, the highest HRs were observed in females aged ≤70 years, whereas males showed higher HRs among individuals aged >80 years.

Sensitivity analyses

In sensitivity analyses I, we assessed HRs for the entire cohort in time periods because the exposure variable showed non-proportionality. In general, the rates of outcomes among patients compared with controls were increased across age groups in different time periods (1, >1–5, and >5 years), and proximity to AF diagnosis led to higher estimates, especially during the first year (see Supplementary data online, Figure S10). In sensitivity analyses II, we repeated our main analyses by performing 90-day landmark analyses (follow-up began 3 months after AF diagnosis). In general, our findings were replicated in these cohorts (named Cohorts SE and SL) with comparable patterns but attenuated estimates. The HRs and expected life years can be assessed in Supplementary data online, Figures S11 and S12, respectively. In sensitivity analyses III, the main analyses in the entire cohort were repeated with adjustment for comorbidities. The patterns were similar with reduced estimates (see Supplementary data online, Figure S13).

Discussion

This population-based study included Danish nationwide register data on more than 215 000 individuals with AF and more than 860 000 matched controls. Compared with controls, we observed that young age at the time of AF diagnosis was associated with a large reduction in expected average life years and increased rates of myocardial disease, especially cardiomyopathy, including DCM and HF, but also HCM (Structured Graphical Abstract).

Comparing AF patients with matched controls, the absolute risks, excess risks, and HRs of cardiomyopathies increased in a stepwise fashion as a function of younger age at diagnosis. The HRs of HF displayed a similar stepwise association, while the highest absolute risks and excess risks were observed among the older groups. Ischaemic stroke among AF patients compared with controls showed a modest U-shaped trend according to age at diagnosis, absolute risks increased with older age, and there was no clear pattern for excess risks. Finally, we showed increased HRs of mortality with no definite age trend and observed age-specific loss of expected remaining life years, which was considerable among younger patients (11.3 years when diagnosed with AF at ≤30 years of age). In general, the patterns were comparable in the comorbid and non-comorbid cohorts.

Atrial fibrillation has for many years been associated with HF and arrhythmia-induced cardiomyopathy.^24,25 More recently, several studies have examined the genetic component of AF, which includes the identification of deleterious variants in cardiomyopathy-associated genes among particularly younger patients.^3–6,26 Consequently, it seems plausible that early onset of AF may confer a relatively greater risk of cardiomyopathy compared with later onset of AF. The relationship between AF and HF is close, and the two conditions are known to precede each other²⁷; hence, to further investigate the hypothesis of AF as an early presentation of later serious cardiac disease, we performed sensitivity analyses by (i) splitting follow-up time and (ii) applying a grace period of 90 days as patients may be diagnosed with myocardial disease shortly after the presentation of AF. These observations showed that the risks of cardiomyopathy and HF indeed are highest in the first years (<5 years) after initial diagnosis, possible due to close medical examination and faster disease progression among some frail patients. The analyses also confirmed AF as an early sign of later (>5 years) ventricular disease.

Our findings of increased rates of cardiomyopathies, HF, ischaemic stroke, and mortality among particularly younger AF patients encourage continued research in this patient group. The complexity of atrial cardiomyopathy and its association with ventricular cardiomyopathy has received increased recognition, e.g. in cohort-based genetic studies.^9,26,28–30 It has even been demonstrated that among early-onset AF patients, rare pathogenic variants in genes associated with inherited cardiomyopathies and arrhythmias may increase mortality.³¹ Our observations provide novel information on AF as a potential precursor of ventricular cardiomyopathy. The fundamental knowledge on AF is continuously improving. Conventional AF prediction models have mostly been based on non-modifiable risk factors,^2,32,33 but the HARMS₂-AF model incorporating lifestyle factors was recently presented.³⁴ Furthermore, observations have shown that the combined effect of a genetic component and traditional clinical risk factors likewise is important in AF development, with a larger genetic contribution among especially younger patients.^35,36 Thus, as AF and cardiomyopathy/HF share genetic and non-genetic susceptibilities, the cause for disease progression among AF patients may be different. Cases of subsequent cardiomyopathy and HF can be introduced by the arrhythmia per se or because of shared underlying risk factors or diseases; however, other incidents may be a sign of a shared genetic component. The identified association with HCM, in which phenotypic presentation does not mimic arrhythmia-induced cardiomyopathy (a reversible type of DCM), strongly suggests that other factors than arrhythmic episodes explain our observations. A genetic component could be involved as a genetic overlap between AF and HCM has been reported.^6,37,38 The non-arrhythmic substrate is supported by a recent study, which reported increased risk of diastolic dysfunction among AF patients over time, not mitigated by rhythm control.³⁹

This study cannot provide a pathophysiological explanation of whether the cause of ventricular myocardial disease is due to atrial tachyarrhythmia, genetic predisposition, or underlying shared cardiovascular risk factors. Nonetheless, the results emphasize the necessity for developing better clinical risk prediction models and improved preventions, which could include modifiable risk factors reduction⁴⁰ and earlier rhythm control using drugs or catheter ablation treatment to reduce the risk of disease progression.^41–44 Current guidelines do not specifically focus on early-onset AF,^45,46 though the newly updated American guidelines (2023 edition) do have weak recommendation (class IIb, evidence level B-NR) for genetic testing among AF patients with onset before 45 years.⁴⁶

Ischaemic stroke in younger individuals is well-studied, and AF has been shown to be an important risk factor.^47–49 Other cardiovascular complications associated with AF include IHD, chronic kidney disease, vascular dementia, and falls.^1,50 Interestingly, a study has shown that the risk of developing vascular dementia among patients with AF is dependent on age at diagnosis with highest estimates in the younger groups.⁵¹ While prevention of AF complications typically revolves ventricular disease and ischaemic stroke, other vascular and neurological complications should not be disregarded.

We present sex-specific risks and HRs among AF patients compared with controls. The absolute risks and excess risks of any cardiomyopathy and HF were highest among males, while the excess risk of ischaemic stroke was highest among females as previously reported.^52–55 Overall, the HRs were conceptually identical when stratified by sex, though males generally showed higher HRs of any cardiomyopathy and HCM, while females showed higher HRs of HF, ischaemic stroke, and mortality, confirming previous reports.^53,56 The results should be interpreted with caution as other relevant variables, including migraine, pregnancy, preterm delivery, age at menopause, hormone replacement therapy, and anticoagulation therapy, have not been accounted for in our study.^55,57–62 We did not aim to explain the underlying pathophysiological mechanism; hence, we abstain from further elaborating on the possible causes. However, the reported differences in outcomes emphasize the continued need to investigate sex differences in diseases in order to improve risk stratification and treatment.

It is noteworthy that the absolute risks and excess risks of cardiomyopathy were more prominent among the younger age groups. However, these findings should be interpreted with the consideration that older AF patients are less likely to receive a cardiomyopathy diagnosis in general but also without being diagnosed with other diseases or experiencing a competing risk (death) compared with younger patients. Additionally, surveillance bias among patients is obvious and may be more pronounced among younger individuals; thus, carefulness should be applied. Furthermore, patients who have received an AF diagnosis are possibly more attentive and concerned regarding cardiac symptoms compared with controls; hence, they are more likely to contact a physician if experiencing cardiac symptoms caused by HF. In addition, patients might be closely followed, including more frequent echocardiographic examinations; thus, they can receive an HF diagnosis despite marginal symptoms. This could be more recognizable among younger AF patients compared with matched controls. The HRs are influenced by event rates among patients and controls (see Supplementary data online, Tables S4 and S5), and the relative difference in incidence rates between age groups was higher for controls compared with patients. It may therefore be considered that our findings of a stepwise increase in relative rates with age are mainly a consequence of low rates in controls, especially among the younger part of the population. While this could be true, the results imply that the rate of cardiomyopathy and HF is higher among both younger and older patients with AF compared with matched controls.

The study has several strengths but also noticeable limitations that must be considered. First, the nature of observational studies does not establish causal relationships and we acknowledge that reverse causation cannot be completely avoided. Second, as previously discussed, the possibility of surveillance bias is recognizable. Third, the analyses did not adjust for ablation and medication, including antiarrhythmic and anticoagulation drugs nor were time-dependent variables and clinical information (e.g. body weight and tobacco use) included; thus, there will be residual confounding. Fourth, the diagnoses in this specific study cannot be validated. This might especially be important concerning DCM and HCM, whose diagnostic criteria may sometimes not be completely fulfilled. Additionally, among younger and older patients presenting with ventricular dysfunction, younger patients are most likely to be diagnosed with cardiomyopathy, whereas older are probably more likely to receive an HF diagnosis. Fifth, genetic information was not available, which especially would have been relevant for the genetic forms of DCM and HCM. Sixth, diagnoses were limited to hospitals and did not include data from general practitioners.

In conclusion, this study suggests that age at diagnosis of AF is important in the assessment of subsequent risks of cardiovascular diseases and mortality. Younger age was particularly associated with shorter life expectancy and higher rates of cardiomyopathies among patients compared with the background population. Causality should be determined and it could be further explored whether AF is an appropriate risk marker for ventricular myocardial disease and mortality among these patients.

Supplementary data

Supplementary data are available at European Heart Journal online.

Declarations

Disclosure of Interest

C.T.-P. has received grants for studies from Bayer and Novo Nordisk not related to the current study. J.H.S. has received grants and speaker fee from Medtronic not related to the current study and is a member of an advisory board in Medtronic. L.K. reports speaker honoraria from Novo Nordisk, Novartis, AstraZeneca, and Boehringer Ingelheim not related to the current study. M.S.O. has received speaker fee from Johnson & Johnson Institute not related to the current study. S.Z.D. reports consultancy or speaker honoraria from Bristol Myers Squibb, Pfizer, Bayer, Cortrium, and Ascension and travel fees from Abbott and Boston Scientific not related to the current study. C.P.-M., E.L.F., L.A., L.M.M., N.K.S., and O.B.V. have nothing to report.

Data Availability

Underlying data cannot be shared publicly as data are only available through Statistics Denmark’s servers.

Funding

The work was supported by Rigshospitalet (the Research Foundation at Rigshospitalet and the Research Foundation of the Heart Center Rigshospitalet), Department of Clinical Medicine at the University of Copenhagen, John and Birthe Meyer Foundation, Villadsen Family Foundation, Arvid Nilsson Foundation, Skibsreder Per Henriksen R og Hustrus Fond, Novo Nordisk Fonden (NNF17OC0031204, NNF22OC0079592), and Sygeforsikringen Danmark. N.K.S. was funded by Horizon 2020 [the European Union’s Horizon 2020 Research and Innovation Programme (733381)].

Ethical Approval

The study was approved by the Capital Region of Denmark (approval number: P-2019-348). In Denmark, register-based studies on de-identified data do not require ethical approval.

Pre-registered Clinical Trial Number

None supplied.

References

Author notes