LONGITUDINAL TRENDS IN MYOCARDIAL METABOLIC UPTAKE FOLLOWING STEREOTACTIC ARRHYTHMIA RADIOABLATION USING ¹⁸FDG-PET/CT

Abstract

Purpose/Objectives

Ventricular tachycardia (VT) is a life-threatening rhythm disturbance of the heart that has been shown to be amenable to treatment with stereotactic body radiation therapy, known as stereotactic arrhythmia radioablation (STAR). The metabolic impact of directly targeting arrhythmogenic regions of scarred left ventricle (LV) is unknown. In this study, the relationship between radiation dose and ¹⁸[F]-FDG-PET uptake in LV segments following STAR was interrogated.

Materials/Methods

Patients with VT in the ENCORE-VT trial received a 25 Gy single-fraction of STAR to an electrophysiology-guided target volume planned on 4D-CT. Patients underwent ¹⁸[F]-FDG-PET/CT +/− MRI at baseline, day 3 and day 90. Auto-segmentation of the LV into the AHA 17-segment model was performed on planning CTs & PET/CTs, allowing median radiation dose and standardized uptake value (SUV) to be extracted. Change in segment median SUV was correlated with dose and clinical outcomes (implanted cardiac defibrillator (ICD) shock, death).

Results

Nineteen patients received STAR. Median VT episodes were reduced from 119 in the 6 months before versus 3 in the six months after STAR (p<0.001). Targeted LV segments had lower baseline ¹⁸[F]-FDG SUV than non-targeted segments (p<0.001). Median SUV of targeted segments at baseline, day 3 & day 90 were 1.43, 1.67 & 2.52 (p=0.012). At day 3, a ≥20% SUV increase in targeted segments (n=7/19) was associated with improved survival (HR 0.10, p=0.005). At day 90, there was a trend toward weak association between increased median whole LV myocardium SUV and improved LV ejection fraction (R²=0.355, p=0.158). In logistic regression models, SUV values, together with cardiomyopathy type, VT storm occurrence, and amiodarone use, predicted ICD shocks (p=0.027) and death (p=0.016).

Conclusions

In this small hypothesis-generating study, late SUV increases (day 90) were observed in targeted segments, consistent with VT burden reduction. Further longitudinal studies with ¹⁸[F]-FDG-PET/CT after STAR will define how SUV dynamics can be optimally incorporated as a biomarker in practice.

INTRODUCTION

Up to 20% of deaths in Western societies are attributed to sudden cardiac death, which is predominantly caused by ventricular tachyarrhythmia (VT) (1). For patients with cardiomyopathy who have survived VT, the probability of fatal recurrence is high, and therefore treatment to reduce the risk of cardiac arrest are indicated (2). The therapeutic landscape for this condition, which is characterized by abnormal propagation of cardiomyocyte depolarization in a fibrotic/inflamed region of myocardium, typically involves antiarrhythmic drugs, an implanted defibrillator, and catheter-based ablation procedures (3). Single-fraction, high-dose stereotactic body radiation therapy (SBRT) has been shown to have utility in treatment-refractory VT cases in recent years, with the major advantages of being non-invasive and not requiring additional regular medication (4–7). In contrast with conventional treatment modalities, which attempt to compensate for heightened arrhythmogenicity, stereotactic arrhythmia radioablation (STAR) may directly improve the intrinsic electrophysiological properties of the lesion of arrhythmogenic substrate (8).

Understanding the long-term side effects of ionizing radiation in the heart has been an ongoing priority within radiation oncology, particularly since the publication of landmark studies in breast and lung cancer where heart radiation doses were associated with cardiac events and reduced survival (9, 10). Accordingly, the negative impact of radiation in the cardiovascular system has been prioritized in pre-clinical cardiac radiobiology studies (11). The therapeutic potential of radiation in the management of arrhythmia was first postulated in in vivo experiments in 2006 where it was shown that conduction velocity could be improved in infarcted left ventricular (LV) myocardium of rabbits using localized irradiation (12). These effects were associated with increased connexin 43 (Cx43) protein in cardiomyocytes, which plays a critical role in the gap junctions of intercalated discs for intercellular communication. In this study, increases in Cx43 were observed from 2 weeks post-irradiation, were dose-dependent (0–15 Gy) and were sustained at 1 year. A pig study published in 2010 confirmed that stereotactic body radiation therapy (SBRT) could be safely used to cause conduction blocks (13), and in 2013, the first human to receive SBRT for VT was successfully treated at Stanford University (14). A landmark pilot study at Washington University in St Louis published in the New England Journal of Medicine in 2017 served to consolidate the preliminary data for this novel VT treatment modality (15), now known as STAR. The ENCORE-VT study was a follow-up phase I/II trial that subsequently corroborated these pilot studies, demonstrating durable reductions in VT burden for a high proportion of patients and an excellent safety profile (7).

The biological mechanism of action for the therapeutic effects of STAR beyond enhanced expression of conduction-related proteins has not been elucidated. As some patients did not experience a durable clinical response in early-phase trials (4, 7), determination of additional factors by which STAR modulates cardiac conduction is urgently warranted. Functional imaging is routinely utilized by cardiologists for the assessment of cardiomyopathy, including 2-deoxy-2-[¹⁸F]fluoro-D-glucose (FDG) positron emission tomography–computed tomography (FDG-PET/CT) (16–18). Together with other biomarkers, the distribution of myocardial avidity on FDG-PET/CT protocols can be used to subtype underlying pathophysiology and assess therapy response. Alterations in cardiac tracer uptake following definitive chemoradiation have also been noted on FDG-PET/CT performed for tumor response assessment in intrathoracic cancers (19–27). Patterns of tracer uptake post-STAR have not been previously described, however, and therefore the association between diseased myocardial metabolic responses and clinical outcomes are unknown. In this planned translational sub-study of the ENCORE-VT trial, it was hypothesized that longitudinal FDG-PET/CT findings within the ventricular segments of the American Heart Association (AHA) 17 segment model would correlate with clinical outcomes, and in a dose-dependent manner.

METHODS

Study Population

The full methodology and primary results of the ENCORE-VT trial have been published previously (7). Briefly, ENCORE-VT was a prospective single-arm phase I/II trial conducted at a single academic medical center (NCT02919618, IRB 201901162). Eligible patients were adults with a history of either ≥3 episodes of sustained monomorphic VT, or cardiomyopathy related to monomorphic premature ventricular complexes (PVCs), with failure of ≥1 antiarrhythmic medication and ≥1 catheter ablation. Exclusion criteria included previous thoracic radiation therapy (RT) and heart failure dependent on inotropes or LV assist devices, a predicted survival of less than 12 months in the absence of VT and previous adverse VT morphologies. Patients provided informed consent for baseline FDG-PET/CT and cardiac MRI to enable target delineation during radiotherapy planning, and two further scans at day 3 and day 90 for a translational study, where possible. For this translational study using FDG-PET/CT, the main clinical endpoints were VT events, left ventricular ejection fraction and death.

Stereotactic Arrhythmia Radioablation

Patients underwent four-dimensional (4D) planning CT scans with 1.5 mm slice thickness to enable localization of the myocardial treatment target using the treatment planning system (TPS) (Eclipse, Varian Medical Systems, Palo Alto). Intravenous contrast enhancement was used in the absence of a clinical contraindication. Results of cardiac CT, magnetic resonance imaging (MRI), FDG-PET/CT imaging and electrocardiographic imaging from non-invasive programmed stimulation (NIPS) testing (28) were utilized together to define a target encompassing only those areas of ventricular scar containing or adjacent to the VT exit site. A clinical target volume (CTV) was delineated on the averaged 4D-CT and adapted to account for motion as assessed from the 4D-CT phases, generating an internal target volume (ITV). Notably, in all patients the portion of the heart being targeted was akinetic or hypokinetic due to pre-existing cardiac scar in that region. The planning target volume (PTV) was derived by adding an isotropic 3–5mm margin to account for setup and delivery uncertainty.

A total dose of 25 Gy in a single fraction was prescribed to be administered to the periphery of the PTV with a planning goal of achieving maximal dose coverage while not exceeding dose constraints of organs-at-risk (OARs), including the esophagus, stomach, lungs, and spinal cord (29). All plans were subject to standard local physics quality assurance. Treatment plans were devised using a volumetric modulated arc radiotherapy (VMAT) solution predominantly, with cone-beam CT-based image‐guidance for delivery on a linear accelerator (Edge, Varian Medical Systems, Palo Alto).

Follow-Up Assessments

Clinical assessments were scheduled on day 3, at weeks 2, 4, and 6, months 6 and 12, and annually thereafter. Adverse events were continuously assessed post-STAR, and implanted cardiac defibrillators (ICDs) were both remotely monitored and interrogated during study visits. ICD discharge in response to sensed arrhythmias were defined as having shock events, and the date of shock was recorded. Survival was defined as the time interval from treatment until death or last clinical contact. Further details are available on study assessments from previous publications (7, 15).

FDG-PET/CT

Functional imaging was obtained using a single dedicated PET/CT scanner (Seimens Biograph TruePoint/TrueView, Seimens Healthineers, USA) at baseline, day 3 and day 90. Prior to PET/CT scanning, patients were asked to fast for 12 hours. Approximately 370 MBq of FDG was intravenously injected in patients following a capillary blood glucose assessment. After 1 hr to allow for uptake, scanning was performed from the skull base to the proximal thigh in the supine position. First, an unenhanced spiral CT scan (120 kVp, 50 mAs effective) for anatomical information and attenuation correction was performed with slices at a thickness of 5 mm. Afterwards, PET scanning was performed at 1.5 min per bed position using 3D-OSEM with point spread function (PSF) at two iterations and 21 subsets with a 2 mm post-reconstruction Gaussian filter at a pixel size of 4.07 mm reconstruction algorithm with attenuation correction. Standardized uptake values (SUV) were calculated for each pixel based on the signal intensity, which accounts for FDG dose and body mass. SUVs were normalized for background signal using the median liver SUV, generated from a 2.5cm circular segmentation of liver parenchyma without major blood vessels and biliary tree. The median SUV (SUVmed) was then used for analysis. Total target SUVmed was calculated as the mean of the SUVmed values for all targeted segments, defined as those segments receiving a median dose ≥25 Gy. Total LV myocardium SUVmed was calculated as the median of SUVmed values for all LV segments.

Left Ventricle Segment Analysis

To evaluate change in SUVmed (ΔSUVmed) longitudinally and to correlate this with the radiation dose, all FDG-PET/CT were registered with the planning CT. Co-registration of these images for the heart directly is challenging due to the lack of contrast at the borders of the myocardial walls (the CT scan in the FDG-PET/CT was not acquired with contrast enhancement) and variability in respiratory status. Therefore, we chose to map each FDG-PET/CT and planning CT to the AHA 17 segment model, which allows a per-segment analysis of ΔSUVmed and radiation dose.

All planning CT and FDG-PET/CT and image datasets were imported into MIM software (MIM Software Inc, Cleveland OH). For all image datasets, the right ventricle and LV epicardial surfaces were contoured on CT according to a local protocol (Supp. Appendix 1) and the endocardial surface was also contoured for each excluding the papillary muscles. Prior to segmentation of the FDG-PET/CTs, the original FDG-PET/CT registration was visually reviewed prioritizing the accuracy of the registration of the LV contour structure and adjusted if necessary according to a local protocol (Supp. Appendix 2). In cardiac regions where the registration of the FDG-PET and CT images was not perfect after any adjustment, segmentations were altered to include avid regions out-with the CT contour. The LV was then divided into 17 volumetric segments according to the AHA guidelines (30) using a semi-automated method developed in-house, for each CT (three FDG-PET/CTs and one RT planning CT). The LV endocardial contour was used to remove blood pool from the SUVmed and radiation dose. Median radiation dose and SUVmed were extracted for each segment on all 4 image datasets for all patients, see Figure 1. Targeted segments were retrospectively classified as those receiving ≥25 Gy.

Figure 1. Spatial relationship between left ventricular wall FDG-PET SUV over time according to radiation dose distribution for a representative patient in the ENCORE-VT clinical trial.

Three-dimensional reconstruction of the 17 left ventricle segments from a representative patient (A–B), with bullseye plots of the radiation dose distribution (C) and the corresponding radiation planning scan (D), outer and endocardial ventricular surfaces annotated, plus planning target volume in cyan). Bullseye plots of the SUV uptake from FDG-PET/CT scans at baseline (E), day 3 (G) and day 90 (I) and a representative axial slice from the corresponding scan at the level of the planning target volume (F, H, J). This 64 year old male patient with non-ischemic cardiomyopathy (NYHA class II, LVEF 26%) was referred for STAR for ongoing VT despite a previous catheter ablation. Following treatment of two disparate substrates identified by electrophysiological mapping, the number of anti-tachycardia pacing events reduced from 127 to 29 in the 6 months preceding stereotactic arrythmia radioablation compared to the 6 months after. However, the patient died as a result of ventricular storm 19 months following treatment.

(FDG = 2-deoxy-2-[¹⁸F]fluoro-D-glucose; SUV = standardized uptake value; PET/CT = positron emission tomography–computed tomography; NYHA = New York Heart Association; LVEF = left ventricular ejection fraction; STAR = stereotactic arrythmia radioablation; VT = ventricular tachyarrhythmia)

Cardiac MRI

Patients underwent cardiac MRI when deemed safe at day 0, day 3 and day 90 on a 3.0 Tesla scanner (Vida, Siemens, Malvern, PA, USA) using a phased-array surface coil. T1-weighted gradient recalled echo LGE-MRI images of the left ventricle (5mm slice-thickness) were acquired in the short axis, horizontal-long axis and vertical-long axis in diastole to cover the heart. Images were acquired 10 min after intravenous injection of gadopentetate dimeglumine (0.2 mmol/kg). Other imaging parameters were; 3.6 ms time to repeat (TR)/1.1 ms time to echo (TE); 15° flip angle; maximum 380 mm field of view (FOV); 224 × 224 matrix. The inversion time was adjusted to null normal myocardium (200–300 ms). Gradient-recalled echo (GRE) cine MRI images of the left ventricle were acquired in the short axis to cover the heart and one image each in the vertical and horizontal long axis. The acquisition parameters for cine MRI were: 2.8 ms TR/1.52 ms TE; 60° flip angle; 360 mm FOV; 192 × 192 matrix; and 10 mm section thickness, and left ventricular ejection fraction (LVEF) was calculated. As per institutional protocol for patients with implanted electrical devices, T1 and T2 sequences were not performed. LV segments were scored for late gadolinium enhancement (LGE), a marker of fibrosis, according to a 4-point scale previously described (31) by an experienced cardiac radiologist (XX): 0 = no LGE, 1 = <25% thickness, 2 = ≥25% but <50% thickness, 3 = ≥50% but <75% LGE thickness, 4 = >75% thickness. Manual scoring was performed, given the high prevalence of device-related metal artifact.

Statistical Analysis

Descriptive statistics were used to summarize the data including box-whisker plots for segments and bar graphs for patients. The normality of ΔSUVmed values within 5 Gy bins was analyzed with Shapiro-Wilk tests, and these were not normally distributed at day 3 and day 90, except for 30–35Gy and 35–40Gy at day 3. Mann-Whitney tests were used to assess the significance between low (0–10 Gy versus 20–40 Gy). A 2-way ANOVA was used to assess the statistical significance of differences between timepoints. A planned exploratory analysis of the SUV change distribution included visually attempting to identify a threshold that defines a subgroup at higher or lower risk for shock and death events that is one-third the size of the total cohort. Kaplan-Meier analysis was undertaken for overall survival according to this arbitrary SUV change threshold and divided the cohort into one-third and two-thirds. The hazard ratio (HR) and it’s 95% confidence interval (95%CI) are presented along with a long-rank significance test result. Multiple logistic regression tests were performed for the endpoints of ICD shock and death according to absolute SUV (target) and relevant clinical covariables at both timepoints. The area under the curve (AUC) and p value compared with random, positive and negative predictive values and the Hosmer-Lemeshow statistics were calculated. Correlation between the mean LV myocardium ΔSUVmed and LVEF change were calculated with R² values and F tests. All statistics were performed using Prism version 10 (GraphPad Software, Boston, Massachusetts, USA).

RESULTS

Patient Characteristics

Nineteen patients received STAR. As shown in Table 1, seventeen were male and the median age was 66. Median follow up was 13 months, and no patients were lost to follow-up. The median PTV volume was 98.9 cc (range, 60.9–298.8). The mean radiation dose to the targeted segments was 33.1 Gy (range 25.0–41.9), and the mean dose to non-targeted segments was 7.9 Gy (range 0.0–24.6). Most patients (n=17, 89.5%) were treated with a volumetric modulated arc therapy technique as summarized in Table 1. Of the eighteen patients who survived 6 months, the primary efficacy endpoint of reduction in VT episodes or PVC burden was achieved in 17 of 18 (94%) patients. For the 16 evaluable patients with ICD-treated VT, the median number of VT episodes decreased from the 6-month pre-ablation period (119, range 4–292) to the 6-month post-ablation period (3, range 0–31, p<0.001), as previously described (7).

Table 1.

Patient, Cardiomyopathy & Radiation Characteristics

	n=19
Median Age (year) (range)	66 (49–81)
Sex (n) (%)  Male  Female	17 (89.5) 2 (10.5)
Ethnicity (n) (%)  Caucasian  Black/African American  Asian	17 (89.5) 1 (5.3) 1 (5.3)
Median Body Mass Index (kg/m2) (range)	33.0 (24.3–48.6)
Median Age-Adjusted Charlson Score (range)	4 (2–13)
Type of Cardiomyopathy (n) (%)  Ischemic  Non-ischemic   Idiopathic   Myocarditis (chronic)   Valvular	11 (57.9) 8 (42.1) 5 2 1
NYHA Class (n) (%)  I  II  III  IV	1 (5.3) 4 (21.1) 10 (52.6) 4 (21.1)
Median Left Ventricular Ejection Fraction (%) (range)	25 (15–58)
Median Number of Previous Catheter Ablations (range)	1 (0–4)
COPD / Emphysema (n) (%)	4 (21.1)
Type 2 Diabetes Mellitus (n) (%)	7 (36.8)
Hypertension (n) (%)	10 (52.6)
Chronic Kidney Disease, Stage ≥3 (n) (%)	9 (47.4)
Median Target Volume (cc’s) (range)  Gross target volume  Internal target volume  Planning target volume	25.4 (6.4–88.6) 31.0 (17.7–128.9) 98.9 (60.9–298.8)
Median Mean Lung Dose (cGy) (range)	158.06 cGy (44.53–339.49)
Median Maximum Stomach Dose* (cGy) (range)	795.57 cGy (18.23–3370.92)
Median Maximum Oesophagus Dose* (cGy) (range)	596.32 cGy (178.34–1533.14)
Immobilization System (n) (%)  Vacuum immobilization (Elekta BodyFIX)  Abdominal compression (CDR FreedomX)	3 (15.8) 16 (84.2)
Planning Technique  Volumetric modulated arc therapy  Fixed-field intensity modulated therapy	17(89.5) 2 (10.5)
Linear Accelerator  Varian TrueBeam  Varian Edge	3 (15.8) 16 (84.2)
Median Beam-On Time (min) (range)	15.3 (5.4–32.3)

^*= D0.03cc used for Dmax

(NYHA = New York Heart Association; COPD = chronic obstructive pulmonary disease)

Cardiac MRI

Cardiac MRIs were available for 9 patients; n=7 had baseline, day 3 and day 90; n=2 had baseline and day 3. LGE was non-evaluable due to artefact in 21/82 (25.6%) targeted and 102/343 (29.7%) non-targeted segments. As shown in the Supplementary Figures 1-2, while baseline fibrosis was higher in targeted segments than non-targeted segments (3.0 (2.3–4.0) vs 1.0 (0.4–1.3), p=0.008) as expected, there was minimal change after treatment (targeted p=3.05; non-targeted p=0.107).

Ventricular Segment SUVmed

Day 3 and day 90 FDG-PET/CT scans were performed at a median of 3 days (range 2–14) and 139 days (range 87–227), and the final timepoint was missing for two patients who expired prior to the 90-day timepoint. A median of 3 segments were targeted for STAR per patient (range 1–10). Targeted segments had lower baseline SUVmed than non-targeted segments, 1.84 versus 2.73, p<0.001. The SUVmed of targeted segments at baseline, day 3 and day 90 were 1.43, 1.67 and 2.52 (p=0.012), although these differences were driven by ischemic cases (p=0.026) rather than non-ischemic cases (p=0.161), see Figure 2. The magnitude of ΔSUVmed at day 3 and day 90 was similar for all segments (Supp. Figure 3), and there was no correlation between the ΔSUVmed at day 3, and day 90 (R²<0.001, p=0.876), across all segments.

Figure 2. Change in planning target volume SUV on FDG-PET/CT after stereotactic arrythmia radioablation.

FDG SUV (median, interquartile range) for all segments with median dose ≥25Gy are displayed for the total cohort (A) and for ischemic cardiomyopathy (B) and non-ischemic cardiomyopathy (C) subtypes. Significance of differences between timepoints tested by 2-way ANOVA.

(SUV = standardized uptake value; FDG = 2-deoxy-2-[¹⁸F]fluoro-D-glucose; PET/CT = positron emission tomography–computed tomography)

Radiation Dose and Change in SUVmed

At day 3 post-STAR, across all segments for all patients (n=323), there was no correlation between radiation dose (in 5 Gy dose bins) and median values for ΔSUVmed (R²=0.133, p=0.375). However, segments receiving lower RT doses demonstrated significantly greater SUVmed increases from baseline than those segments receiving higher doses (e.g. 18.4 for median SUVmed 0–10Gy (n=158) versus 1.38 for 20–40Gy (n=90), p=0.004), see Figure 3A–B. Similarly, on day 90 post-treatment, across all segments for all patients (n=323), there was no correlation between the dose (in 5 Gy dose bins) received and median values for ΔSUVmed compared to baseline (R²=0.106, p=0.431). At day 90, myocardial segments receiving higher RT doses exhibited significantly greater SUVmed increases than those receiving higher doses (e.g. 83.3 for 20–40Gy median SUVmed (n=76) versus 72.5 for 0–10Gy SUVmed (n=149), p=0.012) (see Figure 3C–D).

Figure 3. Relationship between FDG-PET SUV and radiation dose for all left ventricle segments.

FDG SUV (median, interquartile range) of all segments for all patients at day 3 (A) and day 90 (C), and a comparison of low (0–10 Gy) versus high (20–30 Gy) dose at day 3 (B) and day 90 (D). Significance of differences between dose groups tested by unpaired t test.

(FDG = 2-deoxy-2-[¹⁸F]fluoro-D-glucose; SUV = standardized uptake value)

Left Ventricle SUVmed and Ejection Fraction

The mean LVEF at baseline, day 3 and day 90 was 17.4% (range 8.0–44.6), 30.7% (range 11.4–56.2) and 28.9% (range 11.4–64.5) respectively and the differences versus baseline were statistically significant (Figure 4A). There was a non-significant negative association between total LV myocardial ΔSUVmed and change in LVEF both at day 3 (R²=0.2215, p=0.201), and a non-significant positive association at day 90 (R²=0.355, p=0.158), see Figure 4B–C).

Figure 4. Distribution of left ventricular ejection fraction and correlation with FDG SUV change in total left ventricle myocardium.

Ejection fraction as measured from a cardiac magnetic resonance study (A) plotted against FDG-PET SUV change at day 3 (B) and day 90 (C). Significance of differences between timepoints tested by 2-sided paired t tests, and significance of slope by F test.

(FDG = 2-deoxy-2-[¹⁸F]fluoro-D-glucose; SUV = standardized uptake value; CMR = magnetic resonance imaging; LVEF = left ventricular ejection fraction)

PTV SUVmed and Shock Events

Among the 17 patients with an ICD, there was no clear association between PTV ΔSUVmed at day 3 and the probability of an ICD shock being delivered (Supp. Figure 4A). Similarly, there was no clear association between PTV ΔSUVmed at day 90 and the probability of an ICD shock being delivered (n=16 as one patient with an ICD died prior to day 90 FDG-PET/CT) (Supp. Figure 4B). In a multiple logistic regression model for the outcome of ICD shock, no individual factors from PTV SUVmed (baseline or changes), cardiomyopathy type, previous VT storm, and amiodarone usage had a statistically significant relationship with shock events, however the model as a whole had a fair predictive capacity (AUC 0.828, p=0.027), see Table 2.

Table 2.

Multiple logistic regression models for shock events (left) and death (right) events after stereotactic arrhythmia radioablation

	Variable	Estimate	95%CI	|Z|	P value		Estimate	95%CI	|Z|	P value
		ICD Shocks (n=17*)					Death Events (n=19)
β0	Intercept	4.108	0.159–154.8	0.844	0.399		16.64	0.051–2685	0.920	0.360
β1	Baseline PTV SUV	0.402	0.017–1.968	0.915	0.360		0.7282	0.103–3.829	0.374	0.709
β2	Day 3 PTV SUV	3.17	0.539–58.15	1.078	0.281		0.2817	0.024–1.764	1.297	0.195
β3	Day 90 PTV SUV	0.428	0.060–1.901	1.044	0.297		1.417	0.207–16.88	0.355	0.723
β4	Non-Ischemic	0.616	0.021–13.06	0.318	0.751		0.3855	0.003–12.02	0.519	0.604
β5	Previous Storm	5.854	0.231–513.2	0.974	0.330		5.138	0.137–684.0	0.831	0.406
β6	Amiodarone	1.128	0.044–25.25	0.081	0.936		0.2306	0.0058–6.520	0.890	0.374
AUC			0.828				0.847
Std. Error			0.121				0.111
95% confidence interval			0.59–1.00				0.63–1.00
P value			0.027				0.016
Negative predictive power (%)			75.00				100.00
Positive predictive power (%)			75.00				82.82
Hosmer-Lemeshow Test Statistic			15.91	p=0.044			11.17	p=0.192

(95%CI = 95% confidence intervals; PTV = planning target volume; SUV = Standardized Uptake Value; AUC = area under the curve)

^*= 2 patients with PVCs did not have an ICD

PTV SUVmed and Death

The distributions of PTV ΔSUVmed for all patients at each timepoint are shown in Supp. Figure 5. PTV ΔSUVmed ≥20% at day 3 was associated with better survival than <20% (median not reached vs 14.8 months; HR 0.10 (95%CI 0.03–0.31), p=0.005), as shown in Supp. Figure 6A. Among the 17 patients with a day 90 FDG-PET/CT (n=2 expired prior to day 90), a PTV ΔSUVmed ≥140% was associated with worse survival than <140% (19.6 months vs median not reached, HR 4.80 (95%CI 0.96–24.1), p=0.008), as shown in Supp. Figure 6B. In a logistic regression model for death, no individual factors from baseline SUV or ΔSUVmed at day 3 or 90, cardiomyopathy type, previous VT storm, and amiodarone usage had a statistically significant relationship with death events, but the model as whole had fair a predictive capacity (AUC 0.847, p=0.016), see Table 2.

DISCUSSION

STAR has shown promise as a treatment option for VT refractory to standard therapies. Preliminary pre-clinical (32, 33) and clinical (34) data suggest that STAR of the arrhythmogenic myocardium can lead to durable reductions in the arrhythmia burden with a very low rate of grade 3 toxicity (34). Compared with catheter ablation, the current standard of care, STAR has the benefits of its non-invasive nature, reduced planning interval, and its suitability for reaching deeper and diffuse substrates (35). Animal experiments suggest that STAR may exert its positive effects by reprogramming the electrophysiology of viable cardiomyocytes adjacent to or within the arrhythmogenic region (8), and/or by homogenization of fibrotic scars (32). However, the complete mechanism(s) of action remain to be determined. This translational sub-study of the ENCORE-VT trial was designed to resolve the metabolic impact of STAR on the left ventricle myocardium with longitudinal FDG-PET/CT.

ENCORE-VT was the first phase I/II study of STAR, and it met both its efficacy and safety primary endpoints in that ICD shocks and anti-tachycardia pacing were appreciably reduced, and only 2 patients developed a grade 3 treatment-related toxicity (7, 36). Here we provide the first description of STAR effects on cardiac FDG-PET/CT imaging, with targeted segments displaying less tracer uptake than non-targeted segments at pre-STAR baseline, but with time-dependent increases in avidity seen after STAR (Figure 2). PTV segment SUV elevations were evident at the early timepoint of 3 days, and more pronounced at day 90. There was an observable dose-response, with those segments receiving a greater RT dose (≥20 Gy) exhibiting greater elevations in tracer uptake at day 90 (Figure 3C–D). In a subset of patients with cardiac MRI, there was no change in late gadolinium enhancement, suggesting that baseline myocardial fibrosis was unchanged by STAR. Although FDG-based PET/CT cannot resolve the cell population(s) driving the rise in SUV signal, these data are compelling evidence for focally-enhanced myocardial metabolism as a result of STAR.

FDG-PET/CT is an established tool for the characterization of myocardial and coronary disease (16–18), and scanning protocols and injected radioactive tracers are optimized based on the clinical question and patient factors (37). In contrast with unlabeled glucose, phosphorylated FDG cannot be metabolized and hence accumulates in tissues, allowing for its detection with a PET scanner. FDG is one of the most commonly used PET tracers as it serves as a marker of metabolic activity by enabling tissues with localized increased glucose concentrations to be identified (38). Importantly, avidity corresponds with tissue uptake of glucose rather than a direct measurement of glucose utilization. FDG-PET/CT is advantageous versus CT for many conditions as it can provide a multitude of functional data as well as anatomical information, in addition to being suitable for patients with poor renal function, a co-morbidity that is common in many patients with cardiomyopathy. Co-registration of longitudinal FDG-PET/CT images is often challenging due to differences in heart shape (e.g. due to decompensation), patient setup (e.g. fluctuant breathing) and patient geometry (e.g. weight loss) that can emerge over even short intervals (39). To circumnavigate these challenges, each individual imaging dataset in this study had independent LV myocardial segmentation. Registration between FDG-PET and CT images is also technically difficult, as the ground truth may not be ascertainable from the unenhanced CT and some myocardium may have no or low tracer uptake. A local protocol was therefore devised (Supp. Appendix 2) to ensure these uncertainties were managed systematically for this study.

Several investigators have reported myocardial changes on FDG-PET/CT following thoracic RT, as this imaging modality is commonly utilized for disease staging and treatment response monitoring for several thoracic cancers, including lung and esophageal cancers (40, 41). These reports have been small-to-modest-sized retrospective studies, with analyses typically applied to the whole heart, whole LV or a specific isodose line, which limits the comparability with the presented data using the 17-segment model. Nonetheless, increased SUV was observed in patients with increasing cardiac dose (19, 22, 27), including two studies where there was evidence of a dose-response relationship (21, 23). One study failed to identify a positive association between RT dose and SUV change, although that study analyzed dose to the entire LV despite evaluating SUV from LV walls (24). None of these studies provided radiation and tracer uptake data at the level of the AHA 17-segment model, and all but one study had only two timepoints available. However, overall, the included outcome data would broadly suggest that higher SUV increases are associated with favorable outcomes. Interestingly, studies that reported acute toxicity event outcomes were unable to establish a relationship between SUV levels and adverse cardiac outcomes (23, 24), and those analyzing survival showed a positive association between post-RT SUV and survival (20, 22, 25). It is controversial that irradiation in specific cases could be associated with a benefit to myocardial metabolism given the wealth of literature on the negative impact of cardiac dose in terms of acute cardiac events (42–44) and mortality (44–46). Although there are key differences in the myocardium of STAR patients compared with oncology patients, incorporation of FDG-PET/CT assessments alongside traditional imaging modalities and serum biomarkers in both STAR and cardiotoxicity studies in the future would be informative.

A limitation of FDG-PET/CT is that it is not possible to discern the precise location or cause of the increased metabolic activity. For example, whether the tracer is situated within one specific cell compartment or multiple cell types, and whether changes are due to active uptake processes, or secondary to other biological processes (e.g. capillary leak) is not known. Alternatively, raised local SUV could technically be a constitutive effect of high-dose radiation on the tissues, without any association with the cardiac response. However, it is probable that increases in SUV levels in the STAR-targeted segments are due to inflammation, increased contractility of the whole tissue secondary to increased Cx43 and/or voltage-gated sodium ion channels (8), and/or a primary improvement in cardiomyocyte viability. Some pre-clinical data supports the latter mechanism, such as a study in healthy dogs that found increased SUV in the myocardial region treated with a 20 Gy single-fraction (47, 48) and corresponding enhanced glucose uptake. In histopathological analysis of the canine hearts, glucose transporter 4 (GLUT4) protein, which is the main channel for glucose entry in cardiomyocytes, was upregulated in the myocardial scar compared with the control groups, and interestingly GLUT4 protein levels peaked at day 90, the final timepoint in the present study. Mitochondrial degeneration, swelling, and count reduction were also found in histological analysis of the cardiomyocytes of the irradiated groups. Extrapolating from radiation cardiotoxicity evidence, lung dose (49) or other factors (50) may also be important for interpreting such preclinical studies. In a rat model of high-dose whole-heart +/− pulmonary proton treatment, LV SUV was reduced at day 90, rather than increased (51). In vitro analyses have not been conducted in this area to date, but such studies may yield important insights into the post-radiation growth and repair kinetics of cardiomyocytes and other cell types.

Increased myocardial glucose usage in the ENCORE-VT trial, where the majority of patients derived a clinically meaningful benefit in terms of arrhythmia control, highlights myocardial metabolism as a further potential mechanism of action in STAR. In other disease settings, enhanced glycolysis normally occurs to provide additional adenosine triphosphate (ATP) to compensate for a fuel deficiency created by impaired mitochondrial oxidative phosphorylation, driven by an underlying cardiomyopathy pathology (52). That the low tracer uptake state of the targeted myocardial segments observed in this cohort was actionable may suggest that targeted segments are hibernating due to a chronic pathophysiological process (53, 54) and that STAR possibly augments the cellular viability, potentially by improving metabolic activity in targeted regions including via glycolysis and other pathways (55).

Focally intense tracer uptake could also indicate increased metabolic activity secondary to an influx of immune cells as part of a focal inflammatory response. However, while several components of the immune system appeared to be activated in prior studies (47, 48), the magnitude of these differences were small with 20 Gy, and not observed until 6 months post-radiation. The only pre-clinical studies to date that have performed STAR in an ischemic myocardial scar model did not report levels of inflammation histologically (8, 56). Cardiotoxicity models in mice and rats often have used single-fraction doses comparable to STAR, albeit largely in healthy animals, with various patterns in the immune cell compartment identified post-radiation. However, causative associations with toxicity and immune profiles are still not well-established. In these studies, during the first month, mast cell density generally decreases before subsequently increasing at 3–6 months, which persists to 18 months where it correlates with fibrosis severity (57, 58). Macrophage content, a recognized marker of cardiac remodelling, is generally increased after irradiation (59), however the opposite trend was observed in a comorbidity phenotype (60). The number of CD3+ T-cells in the myocardium is significantly increased at 20–40 weeks compared to unirradiated mice (61–63). Not all lymphocyte populations respond equally, for example, tissue CD2-positive T-cells were normal (64) or reduced (65) after cardiac irradiation. Neutrophils play an accessory role only, being found in the necrotic foci of myocardium exclusively (66, 67). It therefore seems unlikely that inflammation does not play a role in the cardiac response to RT, but it can most likely be expected at later time months (3–6 months) rather than over the course of days.

Several studies have sought to describe histological inflammation in wildtype animals and with a variety of irradiation setups. One group reported evidence of inflammation at 6 months in wildtype swine experiments treated with carbon ion STAR to various myocardial targets, with the most pronounced effects seen in the LV free wall-treated group (68, 69). Modest trends for dose-dependent CD45+ cell infiltration were also observed in a healthy dog model of breast radiation, using 1–5 Gy single-fractions delivered only to a small anterior volume of the heart (70). In a subsequent similar experiment, it was found that enhanced myocardial blood flow did not completely explain the increased metabolic activity, and this was particularly evident at day 90, consistent with the present study (71). In relation to perfusion, a greater benefit for STAR in SUV change was noted for patients with non-ischemic cardiopmyopathy, which may pertain to an underpinning vascular mechanism of action, but more likely is explained by greater tissue hypoxia and therefore less radiation effects in the ischemic subpopulation. Alternate PET/CT tracers which are retained in hypoxic tissue, such as ¹⁸F-fluoromisonidazole (¹⁸F-FMISO), could lead to greater insights into this potential etiology (17, 18).

Interpretation of all existing studies is restricted by the characteristics of the experimental design; however, the complex radiobiology underpinning cardiac effects is sensitive to species, gender, radiation dose fractionation and distribution, and follow-up time (11). Time-course experiments with robust immune endpoints and functional imaging in cardiomyopathy animal models are urgently required to better understand the role of inflammation in the efficacy and failure of clinical STAR. In summary, increased cardiomyocyte viability appears to be the most likely cause of the early rise in FDG-PET SUV observed post-STAR in the context of the existing literature, but whether changes in SUV are directly involved in the therapeutic effect of STAR, or are merely associated, is currently unknown.

In terms of the prognostic impact of ΔSUVmed at day 90, larger SUV increases were associated with a higher risk of death compared with smaller increments (Supplementary Figure 4). Given that the SUV increases were dose-dependent, and considering that the upregulation of important gap junction proteins was achieved with lower doses (15–20 Gy) in mouse models (8), these data could represent evidence for de-escalating the typical dose of 25 Gy. In contrast, the opposite trend was seen at day 3, as greater increments were associated with better survival. Interestingly, there was no apparent relationship between the day 3 and day 90 ΔSUVmed and the risk of an ICD shock during follow-up, although when combined in with relevant clinical variables, the models had good predictive capacity. Notably, neither day 3 nor day 90 myocardial SUV values were independently associated with shock or death events in multivariable models with established clinical factors (Table 2) and so larger studies are needed to better understand which other clinical factors modulate the myocardial metabolic response to STAR. Furthermore, baseline SUV was not associated with shock or death, meaning in this cohort, pre-treatment FDG-PET/CT alone has limited predictive capability.

In the subset of patients that had a paired cardiac MRI, raised SUV in both the PTV and the whole LV myocardium corresponded with improved LVEF at day 3 and day 90 (Figure 4), reinforcing the potential positive prognostic impact of RT-induced SUV increases. The observed acute improvement in LVEF was subject to a separate recent translational study which conveyed how low-dose irradiation was associated with attenuated heart failure in both ENCORE-VT patients and in murine models of heart failure (60). Aligning with that work, in this study, segments receiving a low radiation dose had greater early SUV rises than segments receiving the prescribed dose (Figure 3A–B). Furthermore, in a recent case report from Switzerland, a patient with sarcoidosis-related cardiomyopathy treated with 4 Gy in 2 fractions to the whole heart in conjunction with systemic immunomodulatory agents (and STAR for VT) maintained systolic function (72).

One possible unifying hypothesis resulting from this work is that the acute effects of STAR include enhanced glycolysis and other metabolic pathways via enhanced uptake, causing improved cardiomyocyte viability and contractility. In contrast, subsequent late tracer uptake could be driven by potentially detrimental inflammation, resulting in the negative prognostic signal from day 90 measurements. Overall, these data implicate FDG-PET SUV kinetics as a potential biomarker for outcome in STAR for VT, although further dedicated prospective clinical studies and translational work (73) are required.

Several key academic questions remain for the field, including the underlying radiobiology of early and late clinical effects, the relevance of substrate etiology, predictive biomarkers, and target volume and dose optimization.(74). The ongoing RADIATE-VT trial is a multicenter phase III randomized trial seeking to formally compare STAR with repeat catheter ablation (NCT05765175), and it has embedded several translational endpoints. (75)(76)(77)

The strengths of this translational study are its prospective nature, the application of the AHA 17-segment model, and its multimodality clinical evaluation, including cardiac MRI and FDG-PET/CT. The limitations of this translational study largely relate to the phase I/II nature of the main trial. For this single-arm study, a small heterogenous cohort was recruited, increasing the risk of spurious results in the absence of a validation cohort. The number of follow-up imaging investigations was restricted in part due to the poor fitness of the patients, but future studies should aim to extend the follow-up time captured. ΔSUVmed thresholds for identifying a high/low risk group for shock and death events were chosen post hoc and require validation in a larger cohort. Many of the day 90 scans were delayed for clinical reasons, which may impact the interpretation of the late FDG-PET/CT data. Ultimately, preclinical experiments will be helpful to resolve the nature of the FDG-PET/CT changes observed.

CONCLUSIONS

The FDG-PET-assessed metabolic activity of arrhythmogenic scar tissue after STAR is complex. Dose-dependent late SUV increases were observed in targeted segments, and there was a corresponding reduction in VT burden. Early SUV rises were observed in adjacent segments receiving a low radiation dose as well, and were associated with improved LV systolic function. Further studies of the metabolic impact of high-dose irradiation in STAR, and lower dose irradiation for cardiomyopathy are warranted.

Data Availability Statement

The data are not available for sharing at the time of publication, but requests will be considered.