Radiotherapy-Induced Malfunction in Contemporary Cardiovascular Implantable Electronic Devices: Clinical Incidence and Predictors

Importance  Risk stratification and management paradigms for patients with cardiovascular implantable electronic devices (CIEDs) requiring radiotherapy (RT) vary widely and are based on limited clinical data.

Objective  To identify the incidence and predictors of CIED malfunction and describe associated clinical consequences in a large cohort of patients treated with photon- and electron-based RT.

Design, Setting, and Participants  Retrospective analysis of all patients with a functioning CIED who underwent RT between August 2005 and January 2014 with CIED interrogation data following RT at an academic cancer center. We identified 249 courses of photon- and electron-based RT in 215 patients (123 pacemakers [57%]; 92 implantable cardioverter-defibrillators [43%]). Substantial neutron production was generated in 71 courses (29%).

Exposure  Implantation of CIED with subsequent therapeutic radiation treatment (neutron producing with 15- or 18-MV photons and non–neutron producing with electrons, GammaKnife, or 6-MV photons).

Main Outcomes and Measures  Malfunction of CIED, characterized as single-event upset (data loss, parameter resets, unrecoverable resets), and delayed effects including signal interference, pacing threshold changes, and premature battery depletion.

Results  Malfunction of CIED attributable to RT occurred during 18 courses (7%), with 15 CIEDs experiencing single-event upsets, and 3, transient signal interference. All single-event upsets occurred during neutron-producing RT, at a rate of 21%, 10%, and 34% per neutron-producing course for CIEDs, pacemakers, and implantable cardioverter-defibrillators, respectively. No single-event upsets were found among 178 courses of non–neutron-producing RT. Incident CIED dose did not correlate with device malfunction. Patients treated to the abdomen and pelvis region were more likely to undergo a single-event upset (hazard ratio, 5.2 [95% CI, 1.2-22.6]; P = .03). Six patients with a CIED parameter reset developed clinical symptoms: 3 experienced hypotension and/or bradycardia, 2 experienced abnormal chest ticking consistent with pacemaker syndrome, and 1 developed congestive heart failure. The 3 episodes of signal interference did not result in clinical effects. No delayed malfunctions were directly attributed to RT.

Conclusions and Relevance  In a cohort of contemporary CIEDs, all cases of single-event upset malfunction occurred in the setting of notable neutron production, at a rate of 21% for neutron-producing RT and 0% for non–neutron-producing RT. Where clinically feasible, the use of non–neutron-producing RT is recommended. Given the lack of correlation between CIED malfunction and incident dose observed up to 5.4 Gy, invasive CIED relocation procedures in these settings can be minimized.

Cancer and cardiac disease represent the leading causes of mortality in the United States. The interaction of radiotherapy (RT) in patients who have a cardiac pacemaker (PM) or implanted cardioverter-defibrillator (ICD) raises unique concerns. Although RT interactions with these cardiovascular implantable electronic devices (CIEDs) have been described, determinants of this risk are not well characterized.

Current guidelines define risk classifications based on pacing dependency and the direct amount of radiation delivered to the CIED, with RT dose at least 2 Gy (to convert to rad, multiply by 100) typically proposed as a threshold for clinical concern.^1,2 Clinical practice varies widely with regard to monitoring and treatment planning.³ Additionally, the number of patients with a CIED who present for RT is expected to increase.⁴ In this study, we sought to identify the incidence and predictors of CIED malfunction and describe associated clinical consequences in patients with contemporary CIEDs treated with photon- and electron-based RT.

Box Section Ref ID

This retrospective review was approved by the University of Texas MD Anderson Cancer Center institutional review board, which granted a waiver of informed consent from study participants due to the retrospective nature of the study. The records of all patients with an in situ CIED who underwent RT between August 2005 and January 2014 were analyzed. Patients were added to a prospectively accumulating registry in the CIED clinic at the time of pre-RT interrogation or subsequent interrogation. Details regarding patients, treatment, and malfunction characteristics were reviewed retrospectively. Institutional practices for device evaluation and monitoring during treatment have been described previously.⁵ The CIED interrogations were typically performed prior to treatment, at treatment completion, and several months thereafter. Data were included from communication from outside cardiologists. Patients without CIED interrogations during or following RT were excluded. For patients who did not undergo an institutional interrogation within 30 days of RT completion, CIED records were reviewed for the presence of device data during the period of RT. Patients whose device was subject to interim interrogation at an outside institution resulting in the loss of all data during the RT course were excluded, unless a mid-RT interrogation captured part of the RT course, in which case only the fractions represented on the device data logs were included.

A review of clinical notes and interrogation reports immediately prior to RT, during treatment, and at all subsequent device follow-ups was conducted. All CIEDs with an ICD platform were designated “ICD” regardless of programming. Radiotherapy records were reviewed for treatment-specific variables including site, prescription dose and fractionation, RT type, and energy. Irradiated regions of the body were categorized as abdomen, brain, chest, head/neck, pelvis, and total body. Lower extremity sites were grouped with the pelvis, and upper extremity sites, with the chest. For cases in which several body sites were treated, the region in closest proximity to the CIED was assigned. Treatments producing substantial neutron fluence (15- or 18-MV photon beams) were categorized as neutron producing. Treatments using electrons, GammaKnife, or 6-MV photons only were categorized as non–neutron producing.⁶

Maximum photon or electron dose (excluding neutrons) received by the CIED was determined on the basis of accessibility of records. Where available, direct measurements using either TLD-100 (Thermo Fisher Scientific) or OneDose dosimeters (Sicel Technology Inc) were used. Measured TLD-100 doses were scaled down on the basis of their distance from the treatment field to correct for overresponse to neutrons during neutron-producing therapy.⁷ Measured TLD-100 doses were also scaled down to correct for overresponse to low-energy photons outside the treatment field.⁸ When direct measurement dosimetry was unavailable, treatment planning dose was used, scaled up to account for known underestimation based on distance outside the field edge.^9,10 In the absence of measured dosimetry or planning system estimates, published reference data for doses outside the treatment field were used.^11-13 For measurements and planning system calculations, uncertainty was dominated by application of necessary correction factors. As distance increased between the CIED and treatment field, larger corrections to the dose created larger uncertainty in the final dose estimate, ranging from approximately plus or minus 5% to 10% in near-field cases to a factor of 2 or more in far-field cases. In the 9 cases for which both planning system and measured dose data were available, the corrected dose to the CIED between the planning system and measurement agreed within a mean of 30%. This level of agreement and uncertainty is consistent with out-of-field dosimetric accuracy.^14,15

Cases with simultaneous treatment of multiple sites and isocenters were considered 1 course of treatment unless dosimetric measurements were available separately. When multiple regions were radiated to varying doses under 1 therapeutic course, the maximum prescription dose and associated fractionation were recorded. The number of lesions treated with GammaKnife was codified as the number of fractions to account for additional radiation delivered to distinct lesions and separate doses. A replanned treatment with an energy change that affected the presence of neutrons was considered a separate course.

Data regarding CIED malfunction were collected from clinical notes, CIED interrogation records, and correspondence with outside cardiologists and CIED manufacturers. All interrogations following RT were reviewed to evaluate long-term function. Any deviation from normal CIED behavior that merited a separate line item description in the interrogation report was documented. Data loss, parameter reset, and unrecoverable resets were categorized as single-event upsets, describing a change in memory state within the CIED circuitry resulting in device malfunction.¹⁶ Data-loss resets were defined as the loss of historical diagnostic data only. Parameter resets were characterized by a change in sensing, pacing, or therapy parameters. Resets that could not be reprogrammed were classified as unrecoverable. Other error categories included signal interference and the resulting CIED behavior (pacing inhibition, antitachycardia pacing, shock), lead threshold changes, and date display errors. All CIED data were reviewed by a single experienced investigator (M.A.R.) to confirm and categorize the event type. Event date and time were recorded where possible within a range of several days and correlated with the time of RT delivery.

Single-event upsets from RT represented the primary outcome observed for statistical analysis. Categorical and continuous variables were stratified by event occurrence and compared via Fisher exact and Wilcoxon rank-sum tests, respectively. Treatment-specific variables included prescribed radiation dose, incident radiation dose, body region receiving radiation, energy level of radiation received, and radiation modality, with separate courses of RT in the same patient counted as distinct analyzable units. Analyses of CIED-specific variables including type and age counted individual patients as distinct analyzable units. Univariate and multivariate analyses associating treatment and CIED-specific variables to single-event upsets were performed using logistic regression. A 2-sided significance α level of .05 was used in all statistical tests. Statistical analyses were conducted with SAS, version 9.3 (SAS Institute).

We identified 286 courses of therapy in 247 patients. In 13 cases, CIED interrogations were performed at an outside institution following RT completion, which cleared the device error log, and were thus excluded. In 9 cases, a pre-RT interrogation was not performed, but a mid-treatment device check captured event data from preceding fractions. In 56 cases, device data during the course of RT were unavailable because of either the absence of any follow-up institutional interrogation (40 cases), or the presence of an initial post-RT interrogation outside the institution that cleared the device error log covering the treatment period (16 cases). In 19 cases, however, mid-treatment interrogation allowed inclusion of preceding fractions. The remaining 37 cases were excluded, giving a total of 249 analyzable treatment courses in 215 patients (123 PMs [57%] and 92 ICDs [43%]). Among our cohort, 23 patients underwent 2 separate courses, 4 received 3 courses, and 1 patient received 4 courses of RT. Incident dose to the CIED was obtained by dosimeter measurement in 176 cases, treatment-planning software in 22, and reference data in 51.

Patient, CIED, and treatment characteristics are listed in Table 1. External-beam photon-only therapy was delivered in 203 treatment plans, 22 plans used a combination of photons and electrons, and 10 patients were treated with GammaKnife. Of the 14 electron-only treatments, therapeutic energies ranged from 6 to 16 MeV, with the majority using 6- or 9-MeV energies.

The cumulative time elapsed during all RT courses was 6217 days. The median (range) time from administration of the last RT fraction to institutional CIED interrogation was 1 (0-497) days. A malfunction in the CIED was detected in 18 courses (7%) as presented in Table 2. Single-event upsets occurred in 15 CIEDs, with loss of historical data in 5, recoverable parameter reset in 8, and unrecoverable reset requiring CIED replacement in 2. One CIED recorded 9 episodes of data loss, and 2 CIEDs each underwent 2 parameter resets. No devices exhibited runaway or no-output malfunctioning.

In 3 patients, transient signal disturbances recorded by the CIED were discovered at later interrogation with time stamps corresponding to RT delivery time in all cases. No inappropriate CIED therapy or change in parameters occurred as a result of these disturbances. In 1 patient, the pacing rate temporarily decreased from 75 to 55 beats per minute. Signal disturbances caused 1 ICD to detect ventricular fibrillation and initiate capacitor charging 9.2 seconds into the detected event. Charging completed in 9.5 seconds, but the disturbance ended 7.8 seconds into the charge, so therapy was aborted. No therapy was delivered because this ICD was incapable of antitachycardia pacing during charging. No clinical symptoms were noted in these instances.

All single-event upsets occurred in patients receiving neutron-producing RT. Table 1 stratifies treatment variables by single-event upset occurrence for this subgroup. Incident photon or electron dose received by the CIED was not correlated with the occurrence of single-event upsets (Table 3). In 46 cases, the incident dose to the CIED was at least 2 Gy, 11 of which exceeded 4 Gy (Figure). A consistent distribution of doses up to 5.4 Gy of incident radiation on the CIED was documented, with 2 outlying CIEDs receiving 12 and 30 Gy without malfunction. Median (range) incident doses to the CIED were as follows: non–neutron-producing RT, 0.42 (0-5.4) Gy; neutron-producing RT not causing a single-event upset, 0.47 (0-30.2) Gy; neutron-producing RT causing a single-event upset, 0.18 (0-4.1) Gy. Among the 34 repeated RT courses within our cohort, 3 resets occurred on the second course, with the preceding courses contributing respective incident doses to the device of 0.01, 0.22, and 0.48 Gy. Among cases with a defined error date, single-event upset occurred at the median fourth fraction of neutron-producing RT and a median incident dose of 0.28 Gy on the CIED. Overall incidences for single-event upsets are presented in Table 4.

Among the entire cohort, 3-dimensional conformal radiotherapy was associated with a higher incidence of single-event upset than intensity-modulated RT (P = .001), but this association disappeared when the cohort was limited to neutron-producing RT (P > .99). Among the neutron-producing RT group, body region remained a significant predictor of single-event upsets, with a higher incidence in abdomen and pelvis sites as compared with head/neck and chest sites. On multivariate analysis incorporating grouped RT site (abdomen/pelvis vs chest/head/neck/total body), incident dose (<2 vs ≥2 Gy), and CIED type (ICD vs PM) among the neutron-producing group, body region receiving RT remained significantly associated with single-event upsets (for abdomen/pelvis, hazard ratio [HR], 5.2 [95% CI, 1.2-22.6]; P = .03). Implanted cardioverter-defibrillator platform was associated with higher risk of single-event upset compared with PM on univariate analysis (P = .02) but was not significant on multivariate analysis (ICD vs PM, HR, 3.7 [95% CI, 0.99-13.9]; P = .052). Incident dose to the CIED was not predictive for single-event upset (≥2 Gy incident dose, HR, 1.4 [95% CI, 0.2-9.8]; P = .74).

Clinical symptoms developed in 6 of 10 patients experiencing CIED parameter reset: 3 experienced hypotension and/or bradycardia, 2 experienced abnormal chest ticking consistent with PM syndrome, and 1 developed symptoms of congestive heart failure. In the 9 patients with no clinical symptoms after single-event upset, 2 reported hearing an audible alarm from their CIED. The other 7 single-event upsets were identified only by follow-up interrogation.

We identified 93 patients with CIED interrogation records available at least 3 months following RT, with a median (range) time from RT completion to last interrogation of 11 (3-73) months. No events occurring after the RT course could be clearly attributed to RT, although 3 patients experienced an increase in the ventricular pacing threshold. In 1 pacing-dependent patient treated for lung cancer with 6-MV photons, ventricular pacing threshold increased from 1.0 V/0.7 ms before RT to 2.6 V/0.7 ms following RT and lead revision ensued for safety reasons. In this case, the maximum dose received by the heart at the lead implant site was 36 Gy and the CIED received 2.4 Gy. A second patient treated with 6-MV photons to the head/neck region with a maximum dose of 4.3 Gy at the CIED and minimal dose to the heart demonstrated a transient and intermittent increase in the ventricular pacing threshold from 1.4 V/0.6 ms to 2.0 V/0.6 ms, which then returned to 1.5 V/0.6 ms 1 week later. A third patient who received whole-body irradiation to 2 Gy in a single fraction with 18-MV photons experienced a transient ventricular pacing threshold increase from a pretherapy value of 1.0 V/0.5 ms to 1.25 V/0.6 ms on the day of therapy, which subseqently returned to 1.0 V/0.6 ms several months after RT. The ICD of 1 patient treated with 10-MeV electrons to the neck (0.09 Gy received by the CIED) showed a date error at interrogation 1 year after treatment.

The causality between RT and these events is unknown. No battery life issues, inappropriate battery longevity calculations, or premature battery depletions were found in patients with long-term follow-up.

Our cohort represents one of the largest studied clinical experiences regarding the effect of RT on CIED function. We report the absence of single-event upsets in 178 courses of non–neutron-producing RT (beam energies <10 MV, GammaKnife, and electrons only). We report 15 episodes of single-event upsets in neutron-producing RT (15/18-MV photons), corresponding to a 21% incidence per course or 1.2% incidence per neutron-producing fraction. In 2 cases, the malfunction required CIED replacement. Clinical symptoms developed in 6 of the 10 patients whose CIED underwent parameter reset. With a consistent dose distribution up to 5.4 Gy in our cohort, incident RT dose did not predict for malfunction. Single-event upsets occurred throughout the RT course with no dependence on cumulative dose and were significantly associated with abdominal/pelvic RT on multivariate analysis. The association between single-event upset an an ICD platform (versus PM) was significant on univariate (P = .02) but not multivariate (P = .052) analysis. We observed 3 cases of signal interference in the setting of both 6- and 18-MV photons without apparent clinical effect, 3 episodes of increased pacing threshold (2 of which were transient), and a date abnormality of uncertain etiology.

Guidelines from the American Association of Physicists in Medicine¹ for managing patients with PMs were published in 1994. In this report, increased risk was assigned to patients whose PMs received at least 2 Gy of direct incident RT. The principal malfunction recognized at that time was CIED failure from cumulative damage, based on ex vivo studies in which PMs exposed to direct RT developed output errors beginning at 10 Gy.^17-19 Subsequent clinical recommendations have adapted this 2-Gy threshold for PMs, with some recommending a 1-Gy limit to ICDs and CIED relocation if these limits cannot be observed.^2,20-24 Manufacturer-specific recommendations vary and are listed in eTable 1 in the Supplement.

The ability of high-energy radiation to cause single-event upsets was first described for CIEDs in 1998.²⁵ When high-energy photons (>10 MV) interact with material in the head of a linear accelerator, neutrons are emitted throughout the treatment room. Thermal neutrons are known in the setting of cosmic radiation to interact with ¹⁰boron found in the dielectric layers of integrated circuits, such as the complementary metal oxide semiconductor (CMOS) components present in contemporary CIEDs.²⁶ This boron-neutron interaction can produce charged α particles and result in disruption of electric currents. Of note, the baseline risk of single-event upset from background cosmic radiation in ICDs has been estimated at approximately 1 per 12 940 days.²⁵ The total number of days elapsed during all RT courses included in our cohort was approximately half this interval, giving an expected baseline risk of 0.48 background events during the study period.

The thermal neutron as the cause for CIED single-event upsets, rather than direct incident RT dose, has been implicated in case reports and small series describing single-event upsets in distantly located CIEDs for neutron-producing RT.^21,22,27-31 Higher failure rates have also been reported in ex vivo irradiations at 18 MV as compared with 10 MV.¹⁶ Such findings have prompted a recommendation against neutron-producing beam energies in some published recommendations.^2,27,32

Zaremba et al³³ recently reported a 3.1% rate of CIED malfunction among 453 RT courses delivered at 4 institutions in Denmark, 183 of which underwent a post-RT interrogation. They likewise report a strong association between malfunction and beam energy (odds ratio, 5.73 for beam energies ≥15 MV), with no association between incident dose and device malfunction. The incidence of malfunction among non–neutron-producing courses of RT is not reported in this study. Makkar et al³² published a series of 69 patients with CIEDs who underwent RT at the University of Michigan, with an incidence of single-event upsets per neutron-producing RT course of 6% for all CIEDs (0% for PMs and 17% for ICDs) and no malfunctions described in the setting of non–neutron-producing RT. Gossman et al³⁴ reported a 3% rate of malfunction in 112 patients with a CIED undergoing RT at 18 different institutions. Neutron-producing energies were used in 14% of cases, and the correlation between malfunction and RT energy is not reported. Elders et al³⁵ recently reported 17 courses of RT in patients with ICDs, 12 of which used photon energies greater than 10 MV. Among the neutron-producing RT group, 4 single-event upsets occurred (2 memory loss errors, 2 parameter resets), with an incidence of 33% per course and 2.1% per fraction.

We report a 0% incidence of single-event upsets in the non–neutron-producing cohort, with a 21% incidence in the neutron-producing group (34% for ICDs, 10% for PMs), and a 7% overall incidence among all courses. Our reported overall incidence of ICD single-event upsets in the setting of neutron-producing RT is generally consistent with that reported by Elders et al³⁵ and higher than reported by Makkar et al,³² although data to calculate the incidence per fraction are not available in the latter study. A comparison with incidences as reported in other studies is confounded by the inability to stratify reported events by neutron-producing RT and device type.

Mixed data exist with regard to the incidence and impact of signal interference on CIEDs. Earlier ex vivo studies report a high rate of sensing abnormalities in CIEDs exposed to direct 18-MV RT, at a threshold of 0.5 Gy and a dose rate of greater than 0.2 Gy/min.^36,37 Runaway malfunctions wherein the CIED persisted in inappropriate pacing or inappropriate shock delivery are described in many case reports in the setting of photon and neutron therapy, although most of these cases involved earlier-model CIEDs prior to the implementation of CMOS technology.^38-46 Subsequent studies have showed no episodes of sensing interference in contemporary CMOS-based ICDs irradiated with scattered 6-MV or direct 6/18-MV radiation.^30,47 In the aforementioned clinical studies, Elders et al³⁵ and Gossman et al³⁴ each report 1 episode of inappropriate tachycardia sensing.^34,35

Ex vivo studies support our finding that contemporary CIEDs show resilience to direct RT exposure, reporting malfunction thresholds of up to 150 and 30 Gy for directly incident 6- and 18-MV photons, respectively.^48,49 Indeed, in our cohort, treatments using non–neutron-producing RT in proximity to the CIED were not associated with any episode of single-event upset. In contrast, abdominal/pelvic RT was correlated with a higher risk of single-event upset, even when analysis was limited to the high-energy cohort. Zaremba et al³³ reported similar findings on univariate analysis, although the correlation was not significant when beam energy was taken into account.³³ The mechanism behind this finding is presently unclear and is the subject of ongoing investigation.

Limitations of the present study include its retrospective nature and lack of real-time electrocardiographic monitoring, which might have identified subtle effects such as sensing abnormalities or transient threshold changes. Single-event upsets, however, particularly those involving parameter reset, were clearly observable. Although we report a substantial number of patients with follow-up interrogations longer than 3 months, the inconsistency of this follow-up limits long-term effect analysis. Finally, multiple methods of determining RT dose to the CIED were necessary, including application of correction factors. Well-described methods of dose estimation were applied with a mean dose uncertainty of plus or minus 30%.

In a large patient cohort, we report all single-event upsets to have occurred in those receiving RT capable of substantial neutron production at an incidence of 21% per RT course. The contemporary CIEDs in our series tolerated a consistent range of incident RT doses up to 5.4 Gy with no increase in malfunction risk. Given the associated expense and potential morbidity, it may be safe to decrease the number of relocation procedures performed. Signal interference was uncommon and transient. In the absence of clinical benefit for higher-energy RT, the use of non–neutron-producing RT is recommended to avoid single-event upsets. If higher RT energies provide clinical benefit, however, the error rates and outcomes that we report will aid clinicians in weighing the risks of using neutron-producing RT.

Accepted for Publication: May 7, 2015.

Corresponding Author: Jonathan D. Grant, MD, Department of Radiation Oncology, University of Texas MD Anderson Cancer Center, 1515 Holcombe Blvd, Unit 0097, Houston, TX 77030 (jgrant@sero.net).

Published Online: June 25, 2015. doi:10.1001/jamaoncol.2015.1787.

Author Contributions: Drs Grant and Rozner had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Drs Gomez and Rozner served as co–senior authors, each with equal contribution to the manuscript.

Study concept and design: Grant, Jensen, Pollard, Kry, Dougherty, Gomez, Rozner.

Acquisition, analysis, or interpretation of data: Grant, Jensen, Tang, Kry, Krishnan, Gomez, Rozner.

Drafting of the manuscript: Grant, Tang, Dougherty, Rozner.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Grant, Tang.

Administrative, technical, or material support: Pollard, Dougherty, Rozner.

Study supervision: Grant, Krishnan, Gomez, Rozner.

Conflict of Interest Disclosures: Dr Kry has received partial support provided by Public Health Service grant CA180803 awarded by the National Cancer Institute, US Department of Health and Human Services. Dr Dougherty has received research grants from St Jude Medical and Biosense Webster and fellowship program support grants from Medtronic, St Jude Medical, and Boston Scientific. No other disclosures are reported.