Article Text
PDF
Abstract
Objective To examine the associations between specific dietary patterns and incident atrial fibrillation (AF).
Methods Using data from the UK Biobank, dietary intakes were calculated from 24-hour diet recall questionnaires. Indices representing adherence to dietary patterns (Mediterranean-style, Dietary Approaches to Stop Hypertension (DASH) and plant-based diets) were scored, and ultra-processed food consumption was studied as a percentage of total food mass consumed. Incident AF hospitalisations were assessed in Cox regression models.
Results A total of 121 300 individuals were included, with 4 579 incident AF cases over a median follow-up of 8.8 years. Adherence to Mediterranean-style or DASH diets was associated with a lower incidence of AF in minimally adjusted analyses (HR for highest vs lowest quintile 0.87 (95% CI 0.80–0.96) and HR 0.78 (95% CI 0.71–0.86), respectively). However, associations were no longer significant after accounting for lifestyle factors (HR 0.95 (95% CI 0.87–1.04) and 0.94 (95% CI 0.86–1.04) respectively), with adjustment for body mass index responsible for approximately three-quarters of the effect size attenuation. Plant-based diets were not associated with AF risk in any models. Greatest ultra-processed food consumption was associated with a significant increase in AF risk even in fully adjusted models (HR 1.13 (95% CI 1.02–1.24)), and a 10% increase in absolute intake of ultra-processed food was associated with a 5% increase in AF risk (HR 1.05 (95% CI 1.01–1.08)).
Conclusion With the possible exception of reducing ultra-processed food consumption, these findings suggest that attention to other modifiable risk factors, particularly obesity, may be more important than specific dietary patterns for the primary prevention of AF.
- epidemiology
- atrial fibrillation
- risk factors
Data availability statement
Data are available on reasonable request. Access to the UK Biobank Resources is available to all bona fide researchers for all types of health-related research that is in the public interest.
Statistics from Altmetric.com
WHAT IS ALREADY KNOWN ON THIS TOPIC
Lifestyle factors such as weight and exercise have a significant impact on atrial fibrillation (AF) risk; however, the role of diet remains unclear, with previous studies largely examining isolated dietary components.
WHAT THIS STUDY ADDS
Ultra-processed food consumption was associated with increased incident AF risk independent of other lifestyle factors including obesity, while adherence to Mediterranean-style or Dietary Approaches to Stop Hypertension (DASH) diet was not associated with reduced incident AF risk after accounting for weight.
HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY
Further investigation into the deleterious effects of ultra-processed food consumption is required. Attention to obesity itself may be more important than specific dietary patterns for the primary prevention of AF.
Introduction
Significant advances have been made in the primary and secondary prevention of atrial fibrillation (AF) in recent years, particularly in the management of cardiometabolic and lifestyle risk factors. However, beyond the recommendation of weight loss in those who are overweight or obese, little is known concerning the impact of diet on AF incidence and progression. At various points, attention has turned to a number of dietary components and their relationship with AF, including alcohol,1 caffeine,2 chocolate,3 fibre,4 nuts5 and extra virgin olive oil.6 Previous epidemiological studies have also investigated the intake of certain nutrients and their association with AF, including the role of carbohydrates,7 fibre,4 salt8 and omega-3 fatty acids.9 10
However, individuals do not consume foods in isolation, and so the characterisation of dietary patterns has become of increasing interest, which provides a more complete picture of what individuals eat and the cumulative and interactive effects of individual components. To this end, only two studies to our knowledge have investigated the role of the Mediterranean diet in the risk of developing AF, with potentially favourable outcomes seen.6 11 The significance of food processing in diet and the consumption of ultra-processed foods, which contain minimal whole foods and ingredients rarely employed in culinary use, is also increasingly implicated in non-communicable diseases. To our knowledge, however, no studies have investigated the association between ultra-processed food intake and AF risk.12
There is a paucity of evidence on the role of diet in AF, and we hypothesised that the beneficial associations of a ‘healthy diet’ seen to be associated with lower cardiovascular risk in observational studies may also be associated with lower AF risk. Specifically, we hypothesised that stronger adherence to Mediterranean-style, Dietary Approaches to Stop Hypertension (DASH) and plant-based diets and minimising ultra-processed food intake would be associated with a lower risk of incident AF.
Methods
We analysed data from the UK Biobank, a prospective cohort study of approximately 500 000 individuals aged 40–69 years across the UK.13 Individuals were identified from National Health Service (NHS) records and were invited by mail to attend one of 22 assessment centres between 2006 and 2010 to participate in the study. At enrolment, participants completed a questionnaire collecting information on sociodemographic, diet, lifestyle, reproductive and environmental factors. Anthropometric measurements were recorded, verbal interviews were undertaken to ascertain medical comorbidities and blood and urine samples were also taken. Since recruitment, participants have been followed for hospitalisations and mortality through linkage with NHS records and invited to complete a range of follow-up questionnaires and tests in-person and online.
Assessment of diet, other covariates and outcomes
Dietary information was collected using the Oxford WebQ, a self-administered web-based 24-hour diet recall questionnaire collecting information on the consumption of 238 common food and beverage items during the previous 24 hours.14 A subset of the UK Biobank participants in 2009 and 2010 completed the questionnaire during their baseline visit, and in 2011 and 2012, participants were invited on four separate occasions to complete the questionnaire online. The questionnaire has been validated against a traditional interviewer-administered multiple-pass 24-hour dietary recall with a mean Spearman correlation coefficient for macronutrients of 0.62.15 To capture the consistency of dietary intake, we only calculated mean values from participants who had completed two or more questionnaires, as single 24-hour dietary assessments may not adequately represent long-term dietary intake.14 This method achieves moderate reproducibility compared with the food frequency questionnaire administered at baseline which assesses longer-term dietary intake, with correlation coefficients ranging from 0.38 to 0.64.14 Questionnaires were not considered valid if the recorded total energy intake for the 24-hour period was less than 600 kcal or more than 3500 kcal for women, or less than 800 kcal or more than 4000 kcal for men.
We examined a range of exposures of interest as identified by a review of the literature (see online supplemental methods and table 1 for further details). In brief, intake of food and beverages in grams was summed to create food groups, which were used to calculate diet pattern indices previously described in the literature that represented adherence to a particular dietary pattern: the Alternate Mediterranean Diet score,16 17 the DASH diet score,17 18 the Plant-based Diet Index19 and the Healthful Plant-based Diet Index.19 We identified ultra-processed foods using the NOVA classification with further details provided in the online supplemental methods,20 and calculated the consumption of ultra-processed foods (in grams) as a percentage of total food mass consumed as previously described.21 Total energy (in kilojoules) and nutrient intakes (carbohydrate, fat, protein and their components) were estimated from the 24-hour dietary questionnaire, and nutrient intake was studied as nutrient densities, the percentages of total energy intake derived from these nutrients.
Supplemental material
Baseline comorbidities were self-identified at the initial interview or from previous hospital inpatient diagnosis, as previously described (see online supplemental methods and table 2).1 22 Incident AF, including atrial flutter events, was identified by the first occurrence of a hospital inpatient diagnosis, hospital ablation procedure or AF-related death (see online supplemental table 3).
Statistical analysis
Baseline characteristics are presented by quintiles of dietary pattern indices; categorical variables are reported as percentages, and continuous variables as mean±SD or median (Q1, Q3) as appropriate. Comparisons were assessed with Kruskal-Wallis, one-way analysis of variance and χ2 tests as appropriate.
In this analysis, we included participants who only completed at least two valid dietary questionnaires and excluded those with prevalent AF at baseline, who developed AF prior to completing their final questionnaire or were lost to follow-up prior to completing the final dietary questionnaire. Cox proportional hazard models were used to assess the association between dietary exposures and the first occurrence of AF. Individuals were considered at risk from the date of the most recently completed dietary questionnaire, until the date of incident AF, date of death, date lost to follow-up or end of available follow-up (for those residing in England, 30 September 2021; Scotland, 31 July 2021; Wales, 28 February 2018), whichever came first.
To assess the association of dietary patterns with AF, dietary pattern indices were included in the model as quintiles, with increasing quintiles representing increasing adherence. Ultra-processed food consumption was also included in the model as quintiles, with increasing quintiles representing increasing consumption. Additionally, we estimated the effect of a 10% increase in absolute consumption of ultra-processed food by weight. To test for linear trend across quintiles of dietary pattern indices, the median value of each quintile was assigned to the corresponding participants and included in the model as a continuous variable. A minimally adjusted model, model 1, was stratified by sex and assessment centre attended, and adjusted for age, Townsend deprivation index, ethnicity, education and total energy intake. Total energy intake was included in the models to examine changes in dietary pattern or nutrient intake independent of the energy intake consumed. Model 2 additionally adjusted for important lifestyle risk factors including body mass index (BMI), smoking status, total metabolic equivalent of task minutes/week, alcohol consumption status and alcohol consumption amount. The fully adjusted model, model 3, additionally adjusted for important comorbidities present at the date of the last completed dietary questionnaire. Further details regarding these covariates are described in the online supplemental methods.
To separately assess the association of nutrient intake with AF, nutrient densities were included in models 1–3 described above as restricted cubic splines with four knots to allow for potential non-linearity. Dietary pattern indices were not included in these models, and effect sizes were estimated by general contrasts of regression coefficients using the median values of covariates as the reference value. Components of macronutrients (sugar, starch and fibre; monounsaturated, polyunsaturated and saturated fats) were mutually adjusted for, and overall p values represent the result of a likelihood ratio test with the nested model without the nutrient of interest.
As sensitivity analyses, we excluded participants with events that occurred in the first 2 years of follow-up to mitigate any potential effect of reverse causality, and limited analyses to those completing three or more 24-hour dietary questionnaires.
The proportional hazards assumption was tested using Schoenfeld residuals and interaction with time. Sex and assessment centre attended at baseline did not satisfy the proportional hazards assumption and so models were stratified by these variables. A two-tailed p value was set at 0.05 for statistical significance. Analyses were performed using R V.4.0.2, and packages survival V.3.2.3 and rms V.6.0.0.
Results
The study population consisted of 121 300 UK Biobank participants as shown in figure 1. Of the included participants, the median follow-up time was 8.8 years (IQR 8.7–9.2), and a total of 4 579 incident AF events occurred over 1 042 272 person-years follow-up. Baseline characteristics for the study population are shown in table 1 by quintiles of Mediterranean Diet score. Participants with increasing adherence to a Mediterranean-style diet were older, had a lower BMI, reported greater physical activity and were less likely to be current smokers and to have comorbid hypertension and diabetes. Largely similar findings were present across the dietary pattern indices and as well as in those with lower ultra-processed food consumption (online supplemental tables 4–7).
Study population. The UK Biobank participants included in this analysis must have completed at least two 24-hour dietary questionnaires whereby energy estimates were considered plausible and were free from atrial fibrillation prior to the completion of the last questionnaire.
Baseline characteristics of the study population by Alternate Mediterranean Diet score
Dietary patterns and incident atrial fibrillation
In the minimally adjusted analyses (model 1), increasing Mediterranean and DASH Diet scores were associated with a lower risk of incident AF (table 2; p for trend 0.002 and <0.001, respectively). After adjustment for smoking, exercise, alcohol consumption and BMI in model 2, results were attenuated, such that the Mediterranean and DASH Diet scores were no longer significantly associated with the risk of AF (p for trend 0.25 and 0.29, respectively). Adjustment for BMI in model 2 was responsible for approximately three-quarters of the effect size attenuation (online supplemental table 8). These results were largely unchanged following the adjustment for comorbidities in model 3 (p for trend 0.50 and 0.60, respectively). No association was found between the Plant-based Diet Indices and the risk of AF in any of the models. Sensitivity analyses did not materially change the results (online supplemental table 9).
Association of dietary patterns and risk of incident atrial fibrillation
Ultra-processed food intake and incident atrial fibrillation
The median (IQR) consumption of ultra-processed food by percentage of total food mass was 11.5% (7.0%–17.6%), and the distribution of intake is shown in online supplemental figure 1. Increasing quintiles of ultra-processed food intake were associated with an increased risk of AF even following adjustment for lifestyle factors and comorbidities (table 3). In the fully adjusted model (model 3), intake in the highest quintile was associated with a 13% increase in the risk of AF compared with the lowest quintile (HR 1.13 (95% CI 1.02–1.24); p for trend 0.02). Sensitivity analyses did not materially change the results (online supplemental table 10). When studied as a continuous variable, a 10% increase in absolute intake of ultra-processed food by weight was associated with a 5% increase in the risk of AF (HR 1.05 (95% CI 1.01–1.08)), and there was no evidence of a non-linear association (p for non-linearity 0.66).
Association of ultra-processed food intake and risk of incident atrial fibrillation
Nutrient intake and incident atrial fibrillation
In the minimally adjusted model (model 1), low carbohydrate intake, in particular low starch intake, and high protein intake were significantly associated with the risk of incident AF (online supplemental figure 2). Results were attenuated after adjustment for lifestyle factors in model 2 (online supplemental figure 3) and after adjustment for comorbidities in model 3 (figure 2), such that no pattern of nutrient intake was significantly associated with the risk of incident AF. Sensitivity analyses did not materially change the results (online supplemental figures 4 and 5).
Association of nutrient intake and risk of incident atrial fibrillation. Shaded areas represent 95% CIs.
Discussion
In this analysis of the UK Biobank cohort study, we characterised associations of specific dietary patterns, ultra-processed food intake and nutrient intake with incident AF. In this cohort of 121 300 participants with a median follow-up time of 8.8 years, our principal findings are as follows:
Increasing consumption of ultra-processed foods was associated with an increased risk of AF, with a 10% increase in absolute intake of ultra-processed foods associated with a 5% increase in the risk of AF in our fully adjusted model.
Although adherence to a Mediterranean-style or DASH diet was associated with a lower incidence of AF in our minimally adjusted models accounting for demographic data, these associations were no longer significant after accounting for other lifestyle factors, with BMI responsible for approximately three-quarters of the attenuation.
Similarly, although low carbohydrate and high protein intake were associated with AF risk in our minimally adjusted models, we observed no significant association of these nutrients or others with AF risk after accounting for other lifestyle factors and comorbidities.
Ultra-processed food consumption and atrial fibrillation
In the current study, ultra-processed food consumption was associated with AF risk, although with a magnitude of effect smaller than that seen for other cardiovascular diseases.21 23 Notably, the amount of ultra-processed food consumed in this study appears similar to that of other cohorts.21 23 A recent controlled crossover trial demonstrated that an ultra-processed diet promotes increased caloric consumption and weight gain, compared with an unprocessed diet matched in nutritional parameters,24 and it may be that this predisposition to weight gain that mediates the increased risk of AF. Furthermore, a number of neoformed contaminants and contact materials from food packaging have been demonstrated in animal models to be arrhythmogenic, including acrolein,25 bisphenol A26 and titanium dioxide.27 However, further experimental and epidemiological studies are much needed to clarify the role of ultra-processed foods in cardiovascular disease.
Dietary patterns and atrial fibrillation
Previous studies have focused on individual dietary components3–5 or have been smaller in size and have not systematically examined a variety of dietary patterns of interest.6 11 While our minimally adjusted models demonstrated beneficial associations for adherence to a Mediterranean-style or DASH diet, our findings were attenuated and no longer significant following adjustment for BMI, which has a substantial impact on AF risk. We were unable to perform formal mediation analyses to elucidate whether diet could influence AF risk through changes in weight, as weight measurements preceded dietary evaluation and so the effect of diet cannot be assumed to have affected changes in weight. However, the beneficial effects of such dietary patterns on other cardiometabolic and lifestyle factors, particularly BMI and blood pressure, and other comorbidities cannot be understated.28 29 Notably, if BMI was indeed an effect mediator along a causal pathway between diet and AF, then adjusting for BMI would bias the true association toward the null. Even in our final model including BMI, the direction of the associations observed appeared consistent with the protective effect observed in other cardiovascular diseases,16 18 30 though tests for linear trend were not significant and effect sizes appear modest in this population, with an upper estimate of approximately 12% lower risk of AF with the highest level of adherence.
Nutrient intake and atrial fibrillation
To our knowledge, no studies have examined the relationships between nutrients other than carbohydrates with AF risk.7 Although low carbohydrate intake, in particular low starch intake, and high protein intake were associated with AF risk in our minimally adjusted models, we did not observe any significant associations between any macronutrients or their components and AF risk after adjusting for lifestyle factors and comorbidities, including no association with carbohydrate and polyunsaturated fatty acid intake. Differences in modelling may have contributed to contrasting findings regarding carbohydrate intake,7 and this requires further study in other populations. Notably, we did not observe any counteracting associations in the components of carbohydrates which may have contributed to the absence of association.
Limitations
A number of limitations warrant discussion. First, our analyses are observational in nature and we cannot exclude the possibility of unmeasured and residual confounding. Second, as we used hospitalisation and death data, asymptomatic AF episodes or outpatient diagnoses may not have been captured. Third, measurement error in the 24-hour diet recall questionnaire may have contributed to regression attenuation which may bias estimates toward the null. Dietary patterns vary over time, season and day of the week, so a single 24-hour dietary assessment may not adequately represent long-term dietary intake. However, in this study we averaged dietary intakes from at least two completed questionnaires, which has been recommended to minimise measurement error and better reflect long-term or ‘usual’ dietary intake in this population.14 The availability of additional questionnaires over time would further improve the assessment of ‘usual’ dietary intake, though we note no material differences in sensitivity analyses limiting analyses to those completing more than two questionnaires. Fourth, the representativeness of the UK Biobank to general populations is also limited by the ‘healthy volunteer’ phenomenon,13 which may lead to an underestimation of the strengths of association. Nonetheless, valid assessments of exposure–disease relationships do not require participants to be representative of the population at large.13 Finally, dietary patterns and other lifestyle characteristics vary by population; therefore, these findings require confirmation in different populations beyond our predominantly white British cohort.
Conclusion
In this observational study, increasing consumption of ultra-processed foods was associated with an increased risk of incident AF. Increasing adherence to a Mediterranean-style and DASH diets was only associated with reduced AF risk in minimally adjusted analyses, and associations were no longer significant after adjusting for BMI and other lifestyle factors. With the possible exception of reducing ultra-processed food consumption, these findings suggest that attention to other modifiable risk factors including weight may be more important than adherence to specific dietary or nutrient patterns for the primary prevention of AF.
Data availability statement
Data are available on reasonable request. Access to the UK Biobank Resources is available to all bona fide researchers for all types of health-related research that is in the public interest.
Ethics approval
This study involves human participants and was approved by the UK Biobank study. The UK Biobank study has research ethics approval from the North-West Multi-centre Research Ethics Committee as a Research Tissue Bank (RTB) approval. This approval means that researchers do not require separate ethical clearance and can operate under the RTB approval. This research has been conducted using the UK Biobank Resource under the application number 62306. Participants gave informed consent to participate in the study before taking part.
Acknowledgments
This research has been conducted using the UK Biobank Resource under application number 62306.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
Twitter @pitmabm, @J_Hendriks1, @PrashSanders
Contributors SJT, CG and CXW were responsible for the conception and design of the study. SJT and CXW were involved in data acquisition, conducted the statistical analysis and wrote the initial draft of the manuscript. SJT and CXW are the guarantors of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. All authors contributed to the interpretation of the study data, critical review and revision of the manuscript.
Funding SJT is supported by a Postgraduate Scholarship from the National Health and Medical Research Council of Australia. CG is supported by a Postdoctoral Fellowship from the University of Adelaide. AE and JH are supported by Future Leader Fellowships from the National Heart Foundation. MEM is supported by a Postdoctoral Fellowship from the University of Adelaide and by a Fellowship from the American Association for University Women. DHL is supported by a Mid-Career Fellowship from Hospital Research Foundation. PS is supported by Practitioner Fellowships from the National Health and Medical Research Council of Australia and by the National Heart Foundation of Australia. CXW is supported by a Mid-Career Fellowship from the Hospital Research Foundation and a Postdoctoral Fellowship from the National Heart Foundation of Australia.
Competing interests GMM has received research funding from Baylis Medical, is a consultant for Johnson and Johnson and InCarda and holds equity in InCarda. DHL reports that the University of Adelaide has received on his behalf lecture and/or consulting fees from Abbott Medical, Biotronik, Medtronic and MicroPort CRM. JH is an International Advisory Board member for BMJ Heart and reports that Flinders University has received on his behalf lecture and/or consulting feels from Biotronik. PS reports having served on the Advisory Board of Boston Scientific, CathRx, Medtronic, PaceMate and Abbott Medical. PS also reports that the University of Adelaide has received on his behalf lecture, consulting fees and/or research funding from Medtronic, Boston Scientific, Abbott Medical and LivaNova. CXW reports that the University of Adelaide has received on his behalf lecture, travel and/or research funding from Abbott Medical, Bayer, Boehringer Ingelheim, Medtronic, Novartis, Servier, St Jude Medical and Vifor Pharma.
Patient and public involvement Patients and/or the public were not involved in the design, conduct, reporting or dissemination plans of this research.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.