Habits and heartbeats: learning from historical longitudinal data on primary prevention of atrial fibrillation

This editorial refers to ‘Hypothetical interventions and risk of atrial fibrillation by sex and education: application of the parametric g-formula in the Tromsø Study’, by L. Nilsen et al., https://doi.org/10.1093/eurjpc/zwad240.

Atrial fibrillation (AF), the most common cardiac arrhythmia with a global prevalence of 2–4%, represents a unique downstream cardiac pathology at the crossroads of genetic predisposition, environmental and lifestyle exposures, and pre-existing comorbidities.^1,2 Atrial fibrillation marks the electrophysiologic phenotype of atrial cardiomyopathy that, especially when diagnosed late, is hard to reverse and manage and is usually associated with increased medical care requirements and risk of morbidity and mortality.² The best-known modifiable risk factors for the development of AF are lifestyle related. Excessive alcohol consumption, sedentary lifestyle behaviour, smoking, obesity, and obesity-related metabolic complications, such as hypertension, diabetes, and obstructive sleep apnoea, are considered important risk factors for AF. In many circumstances, those factors may outweigh the influence of non-modifiable risk factors, such as sex and genetic predisposition.³

Atrial fibrillation incidence has steadily increased globally in the last decades, hand in hand with the increase in lifespan, obesity, and metabolic syndrome rates.⁴ The surge in the incidence of AF, especially when linked with obesity and its related metabolic complications, and its detrimental consequences on human health mark one of the most significant challenges in public health. Moreover, as people are variable in their exposure duration to different lifestyles, the independent effect of adopted behaviours remains unclear and hard to determine.⁵ Therefore, a significant need has arisen to define specific exposures and interventions that may potentially decrease the risk of AF, especially concerning cumulative exposure to modifiable metabolic and lifestyle risk factors. Given the typical long latency period between such exposures and AF occurrence, clinical intervention studies to address this gap in evidence are nearly implausible as they would require substantial sample sizes, follow-up duration, and financial resources. Hence, the potential and definition of early-life lifestyle and habitual intervention to prevent AF remains unknown and disputable. However, such studies require a unique statistical approach to deduce causal inferences and mitigate the risk of bias and interdependence of exposures and interventions.^6,7

Considering this knowledge gap, Nilsen et al.⁸ have utilized longitudinal data from the Tromsø population-based study and followed nearly 15 000 community-dwelling men and women over 25 years with at least two longitudinal baseline data points for AF incidence, which was determined by a review of health records. The investigators aimed to simulate separate or combined hypothetical interventions, including smoking cessation transforming from a sedentary to an active lifestyle (defined as ≥180 min/week or 90 min/week of moderate or vigorous physical activity, respectively), moderate alcohol consumption (1–2 units/week), body mass index (BMI) reduction to 25 kg/m², lowering systolic blood pressure to 130 mmHg, and lowering diastolic blood pressure to 80 mmHg. These interventions were examined across strata of sex and education level [as a surrogate of socioeconomic status (SES)] to address their possibly variable effect across these distinct non-modifiable risk factors. Hypothetical interventions were simulated using the parametric g-formula, a robust statistical method for simulating causal inference in situations where traditional randomized clinical studies may not be feasible or achievable.⁶ This analytic tool is specifically designed to provide causal inference while handling time-varying confounding variables that may affect and be affected by the investigated treatment or exposure, thus mitigating the risk of collider and confounding biases.⁹

A key point in addressing potential preventive measures is defining the appropriate population for intervention. The statistical plausibility and power to detect differences are influenced by sample size and, more importantly, baseline risk.¹⁰ In this study, with equal representation of sexes, there was a substantial imbalance in baseline risk factors between men and women. While men were more likely to consume alcohol excessively and had higher BMI and blood pressure and higher rates of diabetes and cardiovascular disease history, women were less physically active and, surprisingly, were more likely to smoke cigarettes. Of note, regardless of sex, the study population had a relatively low burden of obesity and cardiometabolic disease burden, which, by nature, mitigated the potential effect of lifestyle interventions targeting these risk factors. Interestingly, this study, representing a study case of an average community-dwelling population, reconfirmed the repeated known observation that men are roughly two times more likely to have AF compared with women, with an observed absolute risk of ∼7% in women and ∼14% in men during 22 years of follow-up.²

The results of the hypothetical interventions analyses revealed that among all individual hypothetical lifestyle and exposure interventions, only reduction of BMI yielded significant AF risk reduction, with 16% and 14% relative risk reductions among women and men, respectively. Sufficient physical activity intervention was, unexpectedly, associated with an 8% increased AF risk in men only. This intriguing finding might be related to the known J-shaped relationship between physical activity intensity and AF risk¹¹ and the probable higher rates of men engaging in high-intensity endurance exercise,¹² an important AF risk factor not disclosed in the current report. Hence, this counterintuitive finding should be perceived with doubt and might prompt further focused research.

A combination of all interventions was associated with a significant 41% relative risk reduction for AF in women but no significant hypothetical effect on AF risk in men. However, it might be argued that pooling of strongly interrelated interventions such as weight control, physical activity, and better blood pressure control might lead to overestimation of actual effects and be influenced by simulation of weight loss in metabolic responders, a response that is difficult to predict in real life. Also, the sex differences in their hypothetical response to interventions raise questions regarding the foreseeable hypothetical metabolic response to weight loss among men in this specific cohort or for a collider effect involving increased physical activity, weight loss, and AF among men. Thus, this alleged difference in response to intervention between sexes should be interpreted cautiously.

Obesity, particularly with central phenotype, is a well-based risk factor for AF.^13–15 The mechanism linking obesity to AF is multifactorial, encompassing changes in atrial size, inflammation, increased atrial fibrosis, and alterations in cardiac autonomic function, among others.¹⁴ Thus, the association between a body weight loss intervention and AF risk reduction aligns with existing literature and is somewhat unsurprising. However, the near-normal BMI of study participants might have diluted the hypothetical effect of weight loss on AF incidence. An approach to assess more representative hypothetical intervention effects could be to emulate a target trial where a population with predetermined baseline risk criteria, rather than a non-selective population, is targeted. Future studies in this field should focus on a target trial emulation framework resembling a randomized clinical trial design.

There were no differences between different interventions’ effects on AF risk and education-level strata. The relationship between SES and health outcomes, including those related to cardiovascular health, is well-documented in the epidemiological literature. Typically, lower SES is associated with worse health outcomes due to various factors, including limited access to healthcare, poor nutritional options, sedentary lifestyle, higher levels of stress, and increased exposure to environmental toxins, among others. A typical result of these exposures is obesity and obesity-related metabolic syndrome, which are substantial risk factors predisposing to AF, thus challenging the role of SES as an independent risk factor for AF.¹⁶ The lack of differential response to hypothetical intervention in this study reinforces the notion that SES is more of a risk factor that begets risk factors rather than an independent risk marker for AF.

In conclusion, as AF incidence and prevalence are expected to continue increasing in years to come, the study by Nilsen et al. provides novel insights into the potential of different lifestyle interventions to provide primary AF prevention. Despite several inherited methodological limitations related to relying on a non-selective population-based cohort, this study provides valuable and reassuring data on the importance of early intervention in keeping a healthy lifestyle and promoting weight control as effective preventive measures to reduce the risk of AF later. Further research is warranted to address possible multifactorial behavioural and exposure control to alter the risk of AF, support this study’s findings, and shed light on achievable tools that might be implemented in healthy people to prevent AF in later life.

Funding: The authors received no financial support for the research, authorship, and/or publication of this article.

Data availability

No new data were generated or analysed in support of this research.

References

Author notes