These data demonstrate a link between impaired in vivo myocardial energetics, namely ATP depletion, and subsequent life-threatening ventricular arrhythmic events in people. In MEA, the ventricular arrhythmic risk associated with low cardiac ATP was approximately 3-fold higher than that in individuals with normal ATP and independent of established risk factors, including low LVEF. These findings in a high-risk patient population are consistent with prior preclinical data demonstrating a mechanistic role for impaired cardiac metabolism in experimental arrhythmogenesis. The current findings are an important first step in recognizing the metabolic-energetic aspects of life-threatening arrhythmogenesis in people.
Prior preclinical work demonstrated that metabolic inhibition and ATP depletion can cause both triggered and reentrant arrhythmias. Maintenance of intracellular ionic concentrations is critical for preventing arrhythmias, and the Na+/K+ ATPase and the sarco/endoplasmic reticulum Ca++-ATPase (SERCA) are 2 of the most important. Both are driven by ΔG ~ATP (14). Disruption of these energy-dependent gradients increases the risk of ventricular fibrillation, triggers extrasystoles, and alters refractoriness (12, 14, 17–20). In addition, local mitochondrial and high-energy phosphate abnormalities have been linked to spatiotemporal heterogeneity in the cardiac action potential that also predispose to reentrant arrhythmias (21). Finally, local mitochondrial dysfunction increases production of ROS that result in “metabolic sinks” that cause regions of conduction block and create yet another substrate for reentrant arrhythmias (22, 23). Thus, ATP depletion and metabolic abnormalities are mechanistic substrates for both triggered and reentrant experimental arrhythmias and may also contribute to arrhythmogenesis during times of acute metabolic stress.
Because the patients with low myocardial ATP levels did not exhibit differences in electrocardiographic or inducible ventricular arrhythmias at baseline as compared with patients with normal ATP (Supplemental Table 2) but later were more likely to develop life-threatening arrhythmias (Figures 2 and 3; Tables 2 and 3), we hypothesize that low cardiac ATP is a “substrate” for arrhythmogenesis (6) and that subsequent stressors or increased metabolic demand, due to physical or emotional stress or to induced-ischemia, may further jeopardize energetic balance and “trigger” arrhythmias in patients with already low ATP levels.
To gain some initial insight into the potential electrophysiological impact of energetic abnormalities like those observed in these patients, we explored a well-established computational model that incorporates mitochondrial bioenergetics and electrophysiology in the ventricular myocyte (24). When the energetic parameters measured in these patients were inserted in the computational model with all other input parameters held constant between the normal and low ATP groups, we observed that the action potential duration was modestly prolonged and that SERCA flux was reduced in low ATP conditions compared with normal ATP conditions (Supplemental Figure 3). A theoretical increase in action potential duration or heterogeneity could contribute to reentrant ventricular arrhythmias, and reduced SERCA could alter calcium handling and refractoriness, in turn increasing the propensity for triggered arrhythmias. It is worth noting that SERCA has one of the highest energy requirements (ΔG ~ATP ) of the ATPase reactions (25, 26) and that we observed a significantly reduced ΔG ~ATP in the patients with low ATP versus normal ATP (P = 0.0002; low vs. normal ATP, respectively; Supplemental Table 2). It is important to acknowledge that this computational analysis uses many input parameters that were not determined in these patients. Nevertheless, the ionic and current input parameters in the simulations are those used in the literature, treat the low and normal ATP groups the same, and now incorporate cardiac energetic parameters measured in these patients. As such, the simulations indicate that the degree of ATP depletion observed in these patients alone would prolong the action potential and impair calcium handling, 2 established ionic mechanisms contributing to ventricular reentry and triggered arrhythmias.
Future trials are needed to determine whether metabolic modulators or interventions that preserve or augment cardiac ATP reduce the risk of life-threatening arrhythmias in high-risk patients. Nevertheless, it is worth noting today that some conventional antiarrhythmic drugs (class I and III antiarrhythmics) neither alter energy metabolism nor reduce SCD risk, while medications that reduce energetic demand (beta blockers, aldosterone antagonists), also reduce SCD risk (27). Of note, there is emerging evidence that metabolic modulators like the new SGLT2 inhibitors reduce arrhythmias in animal models (28) and SCD risk in patients with reduced ejection fraction (29). Taken together, our observations of an independent relationship between myocardial ATP depletion and life-threatening arrhythmias in people build on prior basic scientific discovery and frame clinical ventricular arrhythmogenesis, in part, as a metabolic disease. They offer a noninvasive means to both identify increased metabolic arrhythmic risk and to determine the energetic consequences of metabolic interventions. Most importantly, these findings provide a foundation for future investigations of the antiarrhythmic potential of metabolic modulators.
Although a significant association of low myocardial ATP concentrations with SCD risk does not prove causality, detection of cardiac ATP depletion by 31P MRS offers a noninvasive metabolic means to identify increased SCD risk that may complement more established anatomic and functional SCD risk factors like LVEF, ventricular remodeling, and fibrosis/scarring measured by MRI on the same MRI scanners. Among biomarkers for SCD risk, reduced LVEF is one of the strongest and is a central clinical criterion used today for primary prevention ICD implantation. We show here that in patients with reduced LVEF who qualified for primary prevention ICD implantation, a low myocardial ATP level predicted individuals who were most likely to have an appropriate ICD firing or cardiac death and that independent association remained significant after adjusting for LVEF (Table 2 and Table 3). Further, in this entire study population of patients with low LVEF, ATP concentration did not correlate with ejection fraction (Supplemental Figure 1), mean LVEF was similar in the low ATP and normal ATP cohorts (Supplemental Table 2), and LVEF changed over the first 3 years after ICD implant in a similar fashion in the low and normal ATP groups (Supplemental Figure 2). Further, when both ATP and LVEF were dichotomized, ATP and LVEF were complementary, and low ATP still distinguished lower from higher SCD risks (Figure 3). From a risk prediction perspective, conventional demographic and clinical indices were not strong surrogates for distinguishing low from normal cardiac ATP levels (Supplemental Table 2), with the caveat that diabetes (P = 0.03), ischemic etiology (P = 0.061), and higher NT-proBNP (P = 0.043) tended to be associated with lower ATP, as might be expected.
More work is needed to replicate these initial findings in larger, more diverse populations and to better define the ultimate impact on clinical practice and risk prediction. Although it may be premature today to defer ICD placement in a patient with normal myocardial ATP levels based on this single report, it is encouraging to observe that only approximately 20% of the patients with normal cardiac ATP studied here had an appropriate ICD firing or cardiac death over 10–15 years (Figure 2), resulting in an annualized combined event rate of 1.3%–2.0% per year. Although ICD complications are more common in the first year after implantation, complications can and do occur years later. For the purpose of comparison, it is worth noting that the average rates of appropriate ICD firings in patients with normal cardiac ATP concentrations observed here (1.3%–2.0%) are much lower than the approximately 4% average annual rates of ICD complications (30) and the 2.5%–11% annual rates of inappropriate ICD firings (31). In addition, the annual arrhythmic rates are even lower (0.8%–1.0% per year) for individuals with normal ATP and higher LVEF (Figure 3). In the future, ATP levels may be one factor to consider when balancing the benefits and competing risks over 10+ years of ICD placement, especially considering the typical battery lifetime of ICDs. Arguably, the most relevant clinical time horizon for assessing SCD risk as part of the decision matrix to implant an initial ICD tends to be about 5–7 years, the lifetime of most ICDs (32). It is thus reassuring that the Kaplan-Meier curves for ATP prediction began to diverge within this window at about 2–2.5 years (Figure 2, A and B). It would have been difficult to detect differences sooner in this cohort, given the low number of early events in the first year.
Impaired myocardial CK metabolism, but not ATP depletion, was shown previously to predict heart failure events, including heart failure hospitalizations, left ventricular assist device implantation, and transplantation (33). Those observations are consistent with murine heart failure studies showing that restoration of reduced myofibrillar CK metabolism improves contractile indices, consistent with the energy reserve role of CK (34). Clinical heart failure outcomes differ from arrhythmic outcomes, and it is therefore perhaps not surprising that the energetic abnormalities associated with heart failure and arrhythmic risk also differ. Nevertheless, we believe this is the first report to relate a cardiac energetic abnormality, namely the depletion of ATP, important for maintaining ionic balance, to subsequent life-threatening clinical ventricular arrhythmias. Although mortality in patients with heart failure with a reduced ejection fraction was linked to a low myocardial PCr/ATP more than 2 decades ago (8), that study did not investigate arrhythmias or ICD firings. In the current study, when we used the same PCr/ATP criterion of less than 1.6 reported in the early work (8), we found that low PCr/ATP did not predict an increased arrhythmic risk. In fact, we observed a trend for low PCr/ATP to predict better, not worse, arrhythmic outcomes (Figure 2, G and H). This is because low ATP is important and appears in the denominator of the PCr/ATP ratio. The prior study of heart failure (33) and our current arrhythmic findings demonstrated that the cardiac energetic predictors of heart failure outcomes (CK flux) differed from those of arrhythmic risk (ATP levels). They also highlight that contemporary 31P MRS studies in patients that rely solely on PCr/ATP or assume constant ATP concentrations would underestimate the extent of metabolic abnormalities and potentially miss ATP depletion herein linked to life-threatening arrhythmic risk.
It should be acknowledged that our study was observational with a long enrollment phase and that ICD programming parameters were not prescriptive but determined by the implanting clinicians, as done in other contemporary ICD trials (35). Nevertheless, appropriate ICD firings were adjudicated by electrophysiologists as previously reported (15), and the appropriate ICD firing rates in our PROSe-ICD population were similar to those of contemporary trials where ICD programming was prescribed to minimize bias by nonsustained arrhythmias (3, 36). Although inappropriate ICD firing rates have declined over time in the PROSe-ICD and other ICD cohorts, likely influenced by published studies promoting changes in device programing (31), appropriate ICD firing rates have not declined (Supplemental Table 4 in ref. 36). Taken together, these observations suggest that the combination of clinically indicated ICD programming and blinded expert adjudication of appropriate ICD firings for ventricular tachycardia or ventricular fibrillation here reflect clinically relevant, life-threatening arrhythmias, the primary outcome measure of the current study that was predicted by cardiac ATP depletion.
The size of the study population was modest, which may have limited the number of potential confounders that could be adjusted for in MEA. Nevertheless, the population was large enough to detect significant associations of low myocardial ATP with subsequent life-threatening arrhythmias in multiple analyses, and many potential confounders were explored (Supplemental Table 2). Although SCD risk evolves over many years, it was not possible to repeat measures of myocardial energetics and ATP over time in these patients because once implanted, the ICDs create MRS artifacts, and an ICD is generally considered a contraindication to repeat 31P MRS studies, especially at 3 T. In addition, our follow-up data indicate that low cardiac ATP at baseline did not predict subsequent changes in NYHA class, LVEF, or NT-proBNP over approximately 2–3 years (Supplemental Figure 2). The limited spatial resolution and depth of interrogation of these older MRS studies may not fully characterize the spatial heterogeneity in energetics present throughout the heart (especially with prior infarction; ref. 37), but they did detect a significant association with human arrhythmogenesis. It is possible that newer, more sensitive metabolic MRS approaches offering improved spatial resolution may improve the detection of arrhythmia-relevant metabolic abnormalities in the future (38, 39). Finally, the cutoff values for normal myocardial energetic parameters were determined from previously published healthy controls because no healthy individuals qualified for a primary prevention ICD or were enrolled in PROSe-ICD. Nevertheless, identical acquisition and analysis protocols and MR scanners were used for healthy individuals and PROSe-ICD patients, and the studies were conducted contemporaneously.
In summary, these data demonstrate a significant link between cardiac ATP depletion detected noninvasively and life-threatening ventricular arrhythmogenesis in patients with reduced LVEF. These findings support testing of metabolic strategies that limit ATP loss as a treatment for or to prevent life-threatening cardiac arrhythmias. They highlight the potential of noninvasive 31P MRS as a complementary risk stratification tool in identifying individuals who would most benefit from ICDs for the primary prevention of SCD.