Extreme Heat and Arrhythmia in Patients With Implanted Devices

Abstract

Background

Arrhythmia, particularly atrial fibrillation, is on the rise globally. While increased surveillance, aging populations, and other risk factors partially explain this rise, emerging risks such as extreme heat from climate change are underexplored.

Methods

We used Product Surveillance Registry by Medtronic (2016–2023) in the United States to examine patients with implantable cardioverter‐defibrillator or cardiac resynchronization therapy defibrillator devices. We identified first arrhythmic episodes (atrial tachycardia/fibrillation [AT/AF] or ventricular tachycardia/fibrillation) recorded by devices. Events were matched to ambient temperatures from local weather stations. We used a case‐crossover design where each arrhythmia event hour was matched with multiple control hours from the same patient, within the same day of week, month, and year. Conditional logistic models were then pooled across all climate zones in the United States to obtain nonlinear dose–response between temperature and arrhythmia.

Results

Our analysis included 3079 individuals across 103 cities. Fewer AT/AF and ventricular tachycardia/fibrillation events occurred during early morning hours (1 am–8 am) compared with daytime hours (9 am–4 pm) (P <0.001). AT/AF episodes were also fewer on weekends compared with weekdays (P <0.001). We found a significant dose–response relationship between high outdoor temperatures and AT/AF events. Relative to an optimal temperature (19 °C), odds ratios for AT/AF events increased substantially at temperatures of 39 °C (odds ratio, 2.41 [95% CI, 1.37–4.25]), 40 °C (odds ratio, 2.60 [95% CI, 1.38–4.89]), and 41 °C (odds ratio, 2.81 [95% CI, 1.40–5.63]). No significant association between temperature and ventricular tachycardia/ ventricular fibrillation was observed. Lagged temperature exposures did not change the results.

Conclusions

Extreme hot outdoor temperatures were significantly associated with higher odds of AF episodes.

Nonstandard Abbreviations and Acronyms

AT

atrial tachycardia

PSR

product surveillance registry

Research Perspective.

What Is New?

• This study is the largest investigation to date of hourly passive arrhythmic events in relation to temperature, expanding the existing evidence data set to different geographical regions to 103 different US cities.

• We show a doubling to tripling increase in odds of atrial fibrillation at extreme hot temperatures (≥39 °C).

What Question Should Be Addressed Next?

• Future research should focus on collecting personalized exposure data to improve individual‐level heat assessment and determine how to timely intervene on personal thermal environments to prevent arrhythmia risk.

Arrhythmia, and particularly atrial fibrillation (AF), is on the rise worldwide. ¹ , ² In the United States alone, between 3 and 6 million people currently have AF, with projections suggesting this number could rise to between 6 and 16 million by 2050. ³ , ⁴ From 2006 to 2018, AF incidence rates grew by 3% annually, ⁵ and over the past 50 years, its prevalence has tripled. ⁶ This alarming increase is partly due to improved surveillance and diagnostic capabilities. Arrhythmias are often asymptomatic or paroxysmal, which makes it challenging to estimate past prevalence. The rise in arrythmias could also be due to other contributing factors such as an aging population and the rise in traditional risk factors such as obesity, ⁷ hypertension, ⁸ diabetes, ⁹ and others.

Simultaneously, climate change is driving global temperature extremes, resulting in more frequent and severe heat waves. ¹⁰ Emerging evidence now suggests that climate change may exacerbate cardiovascular diseases and arrhythmias. ¹¹ , ¹² , ¹³ , ¹⁴ , ¹⁵ However, despite numerous studies, our understanding of the relationship between temperature and arrhythmias remains incomplete and inconclusive. Some studies point to adverse associations between extreme hot temperatures and arrhythmias, ¹⁶ , ¹⁷ , ¹⁸ while other studies find no significant link. ¹⁹ , ²⁰

To address this knowledge gap, high temporal resolution of rhythm data from cardiovascular implantable electronic devices is needed. Previous studies using implantable cardioverter‐defibrillators (ICDs) to examine the relationship between temperature and arrhythmias have been limited by small, geographically confined samples. ²¹ , ²² In this analysis, we used the Product Surveillance Registry (PSR) by Medtronic, with a large reach across multiple states in the United States, to investigate the impact of ambient extreme heat on atrial and ventricular arrhythmias in patients implanted with either an ICD or cardiac resynchronization therapy defibrillator. We hypothesized that extreme heat during specific hours and days would increase the incidence of arrhythmias detected by implanted devices.

METHODS

Patient Population

We extracted hourly events for patients who underwent cardiovascular implantable electronic device implants between January 1, 2016, and April 20, 2023, and were followed in Medtronic's PSR. The PSR is a prospective, multicenter registry designed for continuing evaluation and periodic reporting of safety and effectiveness of Medtronic market‐released products and currently has patients enrolled in hundreds of centers across the United States and worldwide. ²³ According to the PSR protocol, patient follow‐up was aligned with routine clinical care practices, and each patient was intended to have at least 1 status update reported annually until patient exit or the cardiovascular implantable electronic device was no longer followed. The PSR data were registered by Medtronic at clinicaltrials.gov (NCT01524276). This analysis was approved by the Institutional Review Board at Harvard T.H. Chan School of Public Health (IRB24‐0864). The use of product surveillance data does not require individual informed consent. Access to the data set is restricted and cannot be shared with external researchers.

Arrhythmia Ascertainment

All patients who received a Medtronic market‐released ICD or cardiac resynchronization therapy defibrillator implant and were followed in the PSR registry by a US site were eligible for inclusion in this analysis. We included patients if they had an atrial or ventricular arrhythmic event. We defined atrial arrhythmic events as atrial tachycardia (AT) or atrial fibrillation (AF) recorded in the device transmission data. Similarly, we defined ventricular arrhythmic events as ventricular tachycardia (VT) or ventricular fibrillation (VF). All PSR‐reportable arrhythmic events were monitored from the time of implant and reported to the exact hour it happened. For AF events, we conducted an additional analysis in which we limited to events that lasted >30 minutes. Although patient demographics (eg, age/sex) and baseline cardiac comorbidities (eg, previous history of arrythmia) were collected according to the standard requirements of the PSR, this information was not available for analysis due to data anonymization.

Exposure Assessment

For each patient, we used the centroid of their city of residence and linked it to the nearest available temperature station. Temperature and humidity data were obtained from the Integrated Surface Database, which is a comprehensive, global database managed by the National Centers for Environmental Information and the National Oceanic and Atmospheric Administration. The Integrated Surface Database collects hourly temperature data from thousands of weather stations across the United States and globally, providing a reliable source for historical weather information. ²⁴ In cases where the nearest temperature station data were unavailable, we used the second nearest station to impute the temperature value for that given hour. We used the Köppen‐Geiger climate classification ²⁵ to categorize US cities into broad climate zones: arid (class B), temperate (class C), and continental (class D). Only 1 city, Davie, FL, fell into the tropical (class A) climate group, and none were in the polar (class E) climate zone.

Analysis Methodology

We used a case‐crossover design where each subject with an implantable device serves as their own control ²⁶ (Supplementary Figure S1). This is an extension of conventional case–control that uses self‐matching to eliminate confounding at the individual level (eg, by age, sex, body mass index, past medical history, etc.), and it enables a comparison of the distribution of the exposure across control times. ²⁷ We compared temperature during the hour, day, month, and year of recorded arrhythmic event (case) to the temperature of the same hour and day of week within the same month and the same year where the same individual did not have an arrhythmic event (self‐control). That way, individual characteristics cannot be confounders by design. To avoid the potential interference from postevent medical intervention (eg, medication/ablation following an event), we limited the cases to only first events and created control copies for the first event only.

Statistical Analysis

In descriptive analyses, we examined temporal differences across hour, day, month, and year. In regression analyses, we fitted conditional logistic regression models for atrial and ventricular events separately. To allow for nonlinearity (ie, J‐ or U‐shape relationship), temperature was fitted with natural splines and 4 degrees of freedom. The best‐fitting functional form of temperature was determined from lowest Akaike Information Criteria. All models were adjusted for relative humidity. To understand the influence of humidity on the temperature–arrhythmia relationship, we fitted models with and without humidity adjustment as well as including a “heat index” term (a combined temperature–humidity metric). Odds ratios were reported by comparing extreme temperatures to optimal temperatures (chosen to be 19 °C). We explored delayed effect (lag) of temperatures by using moving averages of temperature for 24, 48, 72, and 96 hours before the event. All analyses were conducted using R (version 4.2.1). A P value <0.05 was considered statistically significant.

RESULTS

We analyzed a total of 3079 individuals with implanted devices who had at least 1 arrhythmic event. Within this population, we examined 2536 AT/AF events and 1348 VT/VF events, resulting in a total of 3884 events across the 3079 individuals. The population was spread across 103 US cities, with a fair distribution throughout the United States but less representation in the northwest (Figure 1). Approximately two‐thirds of our population resided in the Midwest and South regions (Supplementary Table S1). Most US cities were in the continental climate zone, followed by the temperate climate zone.

Figure 1. Distribution of geographical locations of patients and average temperature.

CRT‐D indicates cardiac resynchronization therapy defibrillator; and ICD, implantable cardioverter‐defibrillators.

Although we were unable to analyze individual characteristic data, we reported summary statistics for a representative subset of the data (Supplementary Table S2). The majority of the population was male (77.9%), with an average age of 70.6 years and an average body mass index of 30.2 kg/m².

In a temporal analysis of arrhythmic events, significant variations were observed across different times (Figure 2). We found higher AT/AF events during usual working hours (9 am–4 pm: 42.9%) compared with evening hours (5 pm–12 am: 32.0%) and early morning hours (1 am–8 am: 25.0%) (P <0.001). VT/VF episodes occurred in 38.6% between 9 am and 4 pm, in 37.1% between 5 pm and 12 am, and in 24.3% between 1 am and 8 am (P <0.001). Additionally, there were fewer AT/AF episodes on Saturday and Sunday compared with the rest of the week (P <0.001), whereas VT/VF incidents did not show significant variation across days of the week (P=0.326).

Figure 2. Counts of arrhythmic events by hour of the day, day of the week, and month of the year.

In pooled dose–response relationships across patients with implantable devices, there was a doubling to tripling of odds ratios (ORs) for AT/AF with extreme hot temperatures (Figure 3). Compared with a preselected optimal temperature of 19 °C, the ORs of AF events at outdoor temperatures of 39 °C, 40 °C, and 41 °C were (OR, 2.41 [95% CI, 1.37–4.25]), (OR, 2.60 [95% CI, 1.38–4.89]), and (OR, 2.81 [95% CI, 1.40–5.63]), respectively (Table1). At the maximum temperature of 42 °C, the odds ratio of AF was (OR, 3.03 [95% CI, 1.41–6.49]). Cold temperatures did not show significant associations with AF episodes.

Figure 3. Pooled dose–response relationship of ambient temperature and odds ratio of atrial and ventricular arrhythmias among patients with implanted devices.

AT/AF indicates atrial tachycardia/atrial fibrillation; and VT/VF, ventricular tachycardia/ventricular fibrillation.

Table 1.

Odds Ratio of Extreme Hot and Cold Temperatures and the Odds of Atrial Fibrillation

Cold temperatures (°C)	Odds ratio	95% CI	Hot temperatures (°C)	Odds ratio	95% CI
Reference=19 °C			Reference=19 °C
−6	0.86	0.763–0.963	30	1.327	1.177–1.496
−7	0.86	0.759–0.981	31	1.402	1.206–1.629
−8	0.87	0.754–1.001	32	1.487	1.233–1.793
−9	0.88	0.748–1.025	33	1.582	1.258–1.991
−10	0.88	0.741–1.053	34	1.689	1.280–2.228
−11	0.89	0.734–1.085	35	1.807	1.301–2.511
−12	0.9	0.726–1.120	36	1.938	1.320–2.846
−13	0.91	0.718–1.160	37	2.082	1.337–3.242
−14	0.92	0.708–1.205	38	2.24	1.354–3.708
−15	0.94	0.699–1.254	39	2.413	1.369–4.254
−16	0.95	0.689–1.307	40	2.602	1.384–4.891
−17	0.96	0.678–1.367	41	2.806	1.398–5.630
−18	0.98	0.667–1.431	42	3.026	1.412–6.486

No significant association was observed between extreme heat and VT/VF risk. For VT/VF events, the ORs at outdoor temperatures of 39 °C, 40 °C, and 41 °C were (OR, 0.72 [95% CI, 0.333–1.542]), (OR, 0.70 [95% CI, 0.297–1.643]), and (OR, 0.68 [95% CI, 0.263), 1.754), respectively, compared with the reference temperature of 19 °C. In contrast, colder temperatures showed a suggestive increase in VT/VF incidence, though CIs overlapped with the null (Figure 3).

Lagging temperature exposure did not improve the fit of the models, nor did it change the results (Supplementary Figure S2). Restricting the analysis to episodes longer than 30 minutes also yielded similar results (analysis not included). The dose–response relationship across climate zones suggested that arid climates may have higher odds of AF at hot temperatures, although the CIs overlapped (Figure 4). Regional (Supplementary Figure S3) and subregional (analysis not included) stratification did not reveal significantly different patterns. Removing relative humidity from the model did not change the results meaningfully (Supplementary Figures S4 and S5).

Figure 4. Dose–response relationship of temperature and atrial fibrillation across climate zones in the United States.

DISCUSSION

We conducted what we believe is the largest investigation of hourly arrhythmic events and temperatures among thousands of patients with passive rhythm monitoring through implanted devices. Our analysis shows a doubling to tripling of ORs for AF at extremely hot temperatures (≥39 °C). The wide geographical reach of the PSR enabled a novel examination of temperature–arrhythmia relationships across a range of temperatures and harsh climate zones. While our findings are still far from direct clinical implications, this multicity analysis in the United States supports consideration for precautionary measures for at‐risk.

In 2020, the American Heart Association Scientific Statement on lifestyle and risk factors of atrial fibrillation ²⁸ addressed several risk factors but did not mention climate change, environmental risk factors, or extreme heat. The 2023 Guideline for the Diagnosis and Management of Atrial Fibrillation by the American College of Cardiology and American Heart Association ²⁹ also did not provide any substantial information on environmental risk factors. A compendium on AF published on the epidemiology and future outlook of AF in the 21st century ⁴ identified ‘understudied external exposures’ and ‘environmental determinants of health,’ but these were limited to rural versus urban environments and neighborhood‐specific factors, such as pollution and socioeconomic status, essentially falling short from stating climate change and extreme heat as potential emerging risk factors. A search of all American Heart Association and European Society of Cardiology journals finds no official statements providing guidance for providers on preventing arrhythmia from extreme heat. Clinically, there is a need for more interventional studies at the individual level to understand how healthcare providers treating patients with arrhythmia can intervene to minimize the risks associated with extreme temperatures.

Reports on the relationship between extreme temperatures and all arrhythmias have yielded mixed findings. ¹⁶ , ¹⁷ , ²⁰ , ²¹ , ³⁰ , ³¹ Studies of arrhythmia in patients with implanted cardiovascular devices have been limited by small sample sizes, dated time periods, and limited geographical distribution. For example, a study of 49 patients enrolled in 1 hospital in Boston from 2006 to 2010 found cold temperatures, not hot, to be associated with AF. ³¹ Similarly, a study of 84 patients in Eastern Massachusetts from 1995 to 2002 showed an increase in ventricular arrhythmia with cold temperatures during the warm season. ²² The largest study, conducted in London, analyzed >600 patients from 1995 to 2003 and found similar results. ²¹ Newer studies, however, are revealing more nuanced relationships between environmental conditions and arrhythmias. A study in Taiwan found that apparent temperatures, but not ambient temperatures, were associated with increased ventricular tachyarrhythmias in patients with ICD, with prolonged susceptibility in those with a history of electrical storms. ³² A large 2024 study in North Carolina found that humidity, rather than hot temperatures alone, was a key driver of ventricular arrhythmias. ³³ Our analysis expands on these findings by including multiple locations across different climate zones in the United States, showing that very extreme hot ambient temperatures may now increase arrhythmia. The temperature ranges observed in our analysis were unlikely to have been encountered in these earlier studies.

A recent multicountry study found a sharp increase in ischemic stroke deaths at extremely hot temperatures. ³⁴ Here, we provide an additional piece of evidence where hot conditions could trigger AF, a precursor to ischemic stroke. This may also help contextualize the well‐documented excess stroke mortality in the “Stroke Belt” region of the United States, ³⁵ where extreme heat could be an overlooked contributor to cardiovascular risk. However, this analysis and the multicountry study should be interpreted with caution until mechanistic pathways are proven. Leading hypotheses for the link between extreme heat and arrhythmias include activation of the sympathetic nervous system during thermoregulation and electrolyte imbalance from dehydration. Hyperthermia can trigger disturbances in potassium and magnesium levels, increasing the risk of arrhythmias. ³⁶ , ³⁷ Extreme temperatures may also affect His‐Purkinje transmembrane potential and conduction velocities. ³⁸ Elevated core body temperature can lead to volume depletion, sympathetic activation, and tachycardia. ³⁹ Additionally, heart rate can increase by an average of 8.5 bpm for every 1 °C rise in core body temperature. ⁴⁰ Contrary to previous studies, we did not see an effect of cold temperatures on arrhythmia. However, we found some suggestive evidence that ventricular arrhythmias were associated with extreme cold temperatures, though the CIs overlapped with the null. This discrepancy with previous literature may be due to (1) the smaller sample size for serious ventricular events, and (2) the possibility of milder winters in our more recent analysis compared with older studies. Previous mechanistic physiology research indicated that hypothermia could lead to bradycardia. For instance, an animal study showed a 50% decrease in the heart rate of dogs with a core body temperature of 28 °C compared with those with a core temperature of 37 °C. ⁴¹

An incidental finding from this analysis was that some arrhythmic events occurred more frequently during working hours (9 am–4 pm) and weekdays. A previous study pooling multiple cohorts of patients with ICDs found higher occurrences of ventricular arrhythmia during daytime hours (8 am–10 pm) compared with midnight (12 am–1 am), ⁴² suggesting a potential circadian influence on arrhythmic events. Similar to our findings, that study did not observe differences in ventricular arrhythmias across days of the week. However, we found that atrial arrhythmias were more frequent on weekdays compared with weekends, a pattern that may be driven by variations in autonomic tone, adrenergic stimulation, and daily stressors associated with work and routine activities. Prior research has demonstrated a circadian variation in paroxysmal AF, with episodes peaking at specific times of the day. ⁴³ Additionally, behavioral patterns such as medication adherence, physical activity, and sleep quality may vary between weekdays and weekends, influencing arrhythmia onset. Further research is needed to explore these findings before drawing conclusions or implementing interventions, such as device programming that could reduce unnecessary therapies.

This analysis has 3 major limitations. First, we were unable to analyze individual demographic and comorbidity data due to logistical and data restrictions related to privacy concerns. Despite this, we designed the analysis using a self‐matched case–control approach, which removes confounding by age, sex, or body mass index, because these variables do not change within the same individual from 3 pm this Thursday to 3 pm last Thursday (Supplementary Figure S1). While we could not assess the effect modification by demographics (eg, is the relationship different between men versus women?), we provided summary statistics for a representative subset of the sample (Supplementary Table S2).

Second, we relied on outdoor temperatures from the nearest meteorology station to assign exposure. These temperatures may differ significantly from microenvironmental temperatures in indoor spaces, where heating and air conditioning can affect thermal comfort. We also did not have access to geolocation data or information on individual mitigation measures (eg, HVAC availability), limiting our ability to assess personal exposure with greater precision. However, this approach may have an advantage: while personal temperature measurements might improve exposure assessment, they could also introduce confounding by individual characteristics such as body mass index, which influences thermoregulation and arrhythmia risk. By using weather station data, we minimize this potential bias, because outdoor temperature measurements are independent of individual physiological factors.

Third, we relied on device algorithms to detect arrhythmic events. Due to the high number of events, adjudicating each strip was impractical. However, we adjudicated events for a small subset of individuals (n=10) in 1 institution. We found minimal differences between electrophysiologist readings and device readings; however, the sample was too small to draw definitive conclusions, and outcome misclassification cannot be ruled out.

CONCLUSIONS

Our analysis identified a dose–response relationship between rising hot outdoor temperatures and increased AF events, with the odds doubling to tripling as temperatures reached extreme levels. Emerging environmental factors, particularly those related to climate change–induced heat, are becoming increasingly hard to ignore for their potential to affect management and prevention strategies for AF. As global temperatures continue to rise, this analysis calls for heightened caution and proactive measures to mitigate the risks associated with extreme heat.

Looking ahead, future studies should improve exposure assessment by integrating personalized wearable sensors, which offer a cost‐effective way to capture real‐time heat exposure. Notably, device companies are already incorporating temperature sensors into their technologies, and that can be used towards collecting physiological data in response to heat. Second, expanding research to high‐risk regions in the United States, such as the Stroke Belt, will help identify population‐specific vulnerabilities. Third, mechanistic studies are needed to distinguish how extreme temperatures differentially impact VT/VF versus AT/AF. Finally, interventional studies should explore strategies to mitigate climate‐related arrhythmic risk, informing targeted prevention and adaptation measures in a warming world.

Sources of Funding

Parts of this work have been funded by the Harvard‐Cyprus Endowment Program for the Environment and Public Health, and by the Kuwait Institute for Scientific Research (KISR) grant to the Harvard T.H. Chan School of Public Health.

Disclosures

T. Oyekunle is a Medtronic employee. C. Strohmeyer is a Medtronic employee supporting the Medtronic‐funded PSR, NCT00271180, and a Medtronic shareholder T. Mela is a PSR prinicipal investigator for the Massachusetts General site, and has received honoraria from Medtronic, Biotronik, and Abbott. The remaining authors have no disclosures to report.

Supporting information

Tables S1–S2

Figures S1–S5