Arrhythmogenic right ventricular cardiomyopathy: the never-ending quest for a risk calculator (2024)

,

This editorial refers to ‘Importance of genotype for risk stratification in arrhythmogenic right ventricular cardiomyopathy using the 2019 ARVC risk calculator’, by A. Protonotarios et al., https://doi.org/10.1093/eurheartj/ehac235 and ‘Arrhythmic risk prediction in arrhythmogenic right ventricular cardiomyopathy: external validation of the arrhythmogenic right ventricular cardiomyopathy risk calculator’, by P. Jordà et al., https://doi.org/10.1093/eurheartj/ehac289.

Arrhythmia risk stratification is probably one of the most important challenges in the management of patients with inherited cardiac diseases at risk of sudden cardiac death (SCD). Among them, arrhythmogenic right ventricular cardiomyopathy (ARVC) is characterized in some, but not all patients with a high burden of ventricular arrhythmias (VA), which usually presents as monomorphic ventricular tachycardia (VT).¹ The risk of lethal VA is especially present in males as well as younger and athletic patients.

Over the past years, several risk scores have been proposed to predict the 5-year VA risk in inherited cardiac diseases such as in hypertrophic cardiomyopathy,² Brugada syndrome,³ and laminopathy.⁴ In 2019, Cadrin-Tourigny et al. proposed a new prediction model to predict the 5-year VA risk in ARVC (ARVC Risk Calculator, www.arvcrisk.com) based on retrospective data from a derivation cohort of 528 ARVC patients in primary prevention. This score includes seven variables: sex, age, recent cardiac syncope (<6 months), non-sustained VT, 24 h PVC count, number of leads with T-wave inversion in precordial and inferior leads, and right ventricular ejection fraction.⁵ However, this score was not validated externally in the original publication. Several studies investigated the performance of the ARVC Risk Calculator in external independent cohorts showing good discrimination of patients at risk of VA with the model. However, some studies pointed out a possible VA risk overestimation in low- to intermediate-risk patients⁶ or, in contrast, underestimation in biventricular/left dominant forms,⁷ albeit being underpowered due to small cohorts.^6–8

In this issue of the European Heart Journal, two important validation studies of the ARVC risk calculator are provided. The same group who developed the ARVC risk calculator provides an external validation for it based on an independent international cohort of 429 ARVC patients (probands and relatives).⁹ It is important to note that the authors corrected the predictive model during the current validation study as they detected an inaccuracy in the formula of the original ARVC risk calculator published in 2019, which led to an overestimation of the predicted risk.¹⁰ They found that the corrected 2019 model performed well in the external validation cohort showing acceptable discrimination (although lower than in the derivation cohort, C-index 0.77) with a C-index of 0.7 and good calibration with an overall agreement between the predicted and observed events. In the same issue of EHJ, Protonotarios et al. propose an independent external validation of the ARVC risk calculator in a retrospective independent cohort of 554 ARVC patients in primary prevention and found a similarly good discrimination with a C-index of 0.75, with the best discrimination in gene positive patients with a C-index of 0.82.¹¹ However, they found that the ARVC risk calculator suffered from an overall overestimation of the predicted risk compared to the observed risk, which was only partially improved by the correction of the model.

Risk prediction scores present advantages in comparison to decision algorithms based on single risk criteria as they allow the integration of a combination of variables with different weight into a predictive model. However, there are several limits for the use of these models in clinical practice.

First, to be useful for clinicians in the decision-making process for ICD implantation, it is important to know which threshold may be the most accurate to distinguish between patients at risk of being harmed by an ICD implantation or those who might profit from it. We emphasize that, because of the high VA burden in ARVC, most patients present a high predicted risk of VA according to the model. As an example, 94% of patients have a predicted 5-year risk >5%, 78% have a predicted risk >10%, and 45% have a predicted risk >20%. This certainly does not mean that all patients would benefit from an ICD in primary prevention, especially considering the high rate of ICD-related complications in ARVC patients with an annual rate of complications of 4.2% and of inappropriate shocks of 3.9%.¹² Therefore, applying a threshold >5% (which is close to the 6% threshold used in hypertrophic cardiomyopathy) would lead to ICD implantations in more than 90% of ARVC patients in primary prevention, in whom 71% ICDs would be unnecessary. If the sensitivity of the model is high, the specificity for thresholds <15% remains under 50%. As stated by the authors, assessing the right threshold is not so easy as it is a balance between the need for high protection against SCD (sensitivity) and the risk of unnecessary ICD implantation (specificity) and ICD-related complications. Clearly, these risks are not directly comparable, but we should be aware that applying low thresholds would lead to ICD implantations in large numbers of patients who would probably not benefit from it. The acceptable risk also depends on the patient context. Even if the score appears to perform better than the current algorithm-based guidelines, a C-index of 0.7 is not perfect and leaves room for improvement especially regarding its accuracy and specificity. Additional risk markers such as magnetic resonance imaging, genetics, physical activity, or electrophysiological study may possibly improve the prediction accuracy.

A second important question is the endpoint of the model. As the clinical utility of the score is to identify individuals at risk of SCD and who would thus benefit from ICD implantation, the use of all sustained VA (including ICD therapy and slow VT >100 bpm) as an endpoint is questionable for this purpose, as it may overestimate the rate of clinically relevant events regarding the ICD utility. Contrary to Brugada syndrome or hypertrophic cardiomyopathy where sustained VA are usually fatal, VA in ARVC are mostly monomorphic VT that can be well tolerated and consequently cannot be considered as a surrogate for SCD. Therefore, in regards to ICD decision-making, which aims to prevent SCD, it may be clinically more relevant to use a predictive model focusing on malignant VA events. Interestingly, a second model focusing on life-threatening VA events (defined by SCD or sustained VT >250 bpm) was proposed by the same authors last year.¹³ Of the eight prespecified clinical predictors, only younger age, male sex, premature ventricular complexes count, and number of leads with T-wave inversion were predictive of life-threatening events. A model based on these four parameters showed quite a good discrimination with a C-index of 0.74; however, this second predictive model still needs to be validated in external independent cohorts.

Third, previous data have suggested that despite a good discrimination, there may be a possible overestimation of the risk predicted by the ARVC calculator, especially in lower-risk patients.⁶ In the validation cohort, Jordà et al. suggest that the VA risk may be overestimated in relatives and in non-ICD carriers.⁹ In this issue of EHJ, Protonotarios et al. also observed an overall overestimation of the predicted risk in a large independent cohort of 554 patients.¹¹ The overestimation was lowered after the formula correction but still remained significant, especially in gene elusive patients. Interestingly, the score performed differently according to the underlying genotype with rather good performances in pathogenic PKP2 variant-carriers but lower performance in other genes and gene elusive patients. Their data suggest that the risk markers may be different according to the underlying genotype. These data highlight the importance of genotyping for risk stratification and accurate use of the model. To help with communication of results at the lower end of the spectrum, the comparison of the results with other risk stratification tools such as the 2015 task force consensus statement¹⁴ or a tool for the exclusion of arrhythmic risk such as the one provided by Vischer AS et al.¹⁵ may be helpful to reassure both physicians and patients.

Fourth, in the derivation and validation cohorts for the risk calculator at hand, only patients fulfilling the 2010 ARVC criteria were included. Accordingly, the vast majority of patients included presented a right dominant, i.e. ‘classical’ form of ARVC. Therefore, it is important to remember that the model cannot be applied in patients with left dominant forms, rare genotypes and may be less performant in gene elusive patients.

Finally, we must keep in mind that this model still needs to be validated prospectively.

In conclusion, similarly to risk calculators proposed in other inherited cardiac disease, the ARVC risk calculator is not a magic tool and should be used cautiously by clinicians. The calculator should only be used in patients with classical right dominant form fulfilling the 2010 ARVC criteria and the clinicians should be aware of its limits, i.e. the possible risk overestimation, low specificity for thresholds <15%, and lower performance in non-PKP2 mutation carriers and gene negative patients. As no threshold has been validated to date, this score can only provide an estimation of the 5-year VA risk, which, importantly, is different from the SCD risk. ICD indication should remain a shared decision, not only based on the result of the score but also on the overall evaluation of the patient.

References

Author notes

The opinions expressed in this article are not necessarily those of the Editors of the European Heart Journal or of the European Society of Cardiology.

Conflict of interest. None declared.

© The Author(s) 2022. Published by Oxford University Press on behalf of European Society of Cardiology. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com

Issue Section:

Editorial

Email alerts

See also

Citing articles via

More from Oxford Academic