This is a retrospective registry of patients diagnosed with BrS and with previous EPS between 1998 and November 2020 in 1 Portuguese and 23 tertiary Spanish hospitals. The registry was conducted according to the Declaration of Helsinki and was approved by the institutional review board of each center. The participating centers included consecutive patients with BrS and EPS. BrS was diagnosed after an episode of aborted SCD, during syncope workup, in asymptomatic patients with a suggestive ECG pattern during routine examination, or as a consequence of familial screening after diagnosis of BrS in a family member.

Syncope was defined as loss of consciousness in accordance with the current literature consensus.12–15 For the purpose of this study, attempts were made to differentiate vasovagal from arrhythmic causes and only episodes of syncope of cardiovascular origin were included. Patients were included if they satisfied the screening criteria above and had a type 1 Brugada pattern on ECG at baseline on at least 1 occasion or after provocation with a class I antiarrhythmic drug (depending on availability in the participating hospitals, either ajmaline or flecainide/procainamide were used). A type 1 ECG pattern was defined as a prominent coved ST-segment elevation displaying J-wave amplitude or ST-segment elevation ≥ 0.2mV at its peak followed by a negative T-wave.14

The primary outcome included SCD or high-voltage defibrillator therapy (ie, shocks) for polymorphic ventricular tachycardia or ventricular fibrillation (VF), as determined through defibrillator interrogations during follow-up. We classified only defibrillator shocks for polymorphic ventricular tachycardia or fibrillation; VA terminated by antitachycardia pacing were not classified as meeting the primary outcome.

In the absence of symptoms or device therapy, patients were seen routinely every 6 to 12 months for clinical review and device interrogation, in accordance with the local practice protocol. ICD programming was left to the discretion of the referring electrophysiologist. However, following the 2006 recommendations, programming of a single VF zone above 210 to 220 bpm was advocated. Those patients who were not followed up in the original institution were contacted through telephone interview.

Programmed ventricular stimulation was consistently performed at the apex of the right ventricle (RV) and in some centers, also from the right ventricular outflow tract. Stimulation included 2 to 3 pacing cycle lengths (600, 500, and 430 ms) with up to 3 extra stimuli. The extra stimuli were anticipated in 10 ms decrements up to a shortest coupling interval of 200 ms. A positive EPS result was defined as the induction of a sustained VA (> 30seconds) or one that required shock. A negative EPS outcome was attributed when no VA or a short-lived, self-terminating VA was induced but did not require shock.

Statistical analyses were performed with SPSS software version 21 (SPSS Inc, United States) and R version 3.6 (R Foundation for Statistical Computing, Austria).16 Continuous variables are reported as mean±standard deviation and categorical variables as frequencies and percentages. A comparison of baseline characteristics between groups was performed using the chi-square test for categorical variables and the Student t test/Mann-Whitney U test to compare ranks of continuous variables. The c-index was used to assess the discriminatory capability of the model (concordance).

Cumulative incidence curves were plotted to assess the survival of patients according to their risk of competing events. Time 0 corresponds to the moment of the EPS. A competing risk is an event whose occurrence either precludes the occurrence of another event under examination or fundamentally alters the probability of the occurrence of this other event.

Fine and Gray competing risk regression models were used to predict the individual's overall risk (incidence) of an event. Risk was quantified as a Fine-Gray subdistributional hazard ratios (SHR), with 95% confidence intervals (95%CI). The reference category in each of the 4 scores corresponded to that with the lowest score (0 for the Delise score,7 <2 for the Sieira score,8 and ≤ 3 points for the Shanghai BrS score).9 For each cutpoint of the scores and EPS, the time-dependent positive predictive value (PPV) and negative predictive value (NPV) were estimated by inverse probability of censoring weighting, considering competing risks.

To date, the factor with the highest predictive value remains the baseline ECG pattern, often useful as the basis for decision-making.12,13 Nevertheless, it is important to highlight that one third of patients with events did not have a baseline type 1 ECG pattern, highlighting the dynamic nature of BrS. Besides the ECG, a history of arrhythmic events is considered a harbinger for recurrence.12–14,17 In our sample, the presence of previous syncope was also a robust predictive factor of future events. On the other hand, the role of a family history of SCD in predicting arrhythmic events remains controversial.7,12,17–20. In the study by Delise et al.,7 the presence of family history of SCD did not independently correlate with increased arrhythmic risk. Of note, the characterization of family history showed slight differences according to the score (“early SCD in first-degree relatives” in the Sieiras score; “before the age of 40” in the Delise score, and “family history in first- or second-degree relatives” (with different weights according to definitive diagnosis BrS [2 points], suspicious SCD [1 point] related to BrS or unexplained SCD at <45 years of age5 in the Shanghai BrS score). A recent study established a genetic predictive score based on the presence of the SCN5A mutation and polymorphism in other genes associated with arrhythmic risk.20 It remains to be seen whether the incorporation of both family history and genetic predictors could better identify those at high genetic risk of SCD. In our sample, neither the presence of family history nor a perceptible genetic mutation (performed in more than 300 patients) were reliable predictors for further events. The issue of the usefulness of EPS has been addressed by several consensus documents and multiple relevant meta-analyses. Unfortunately, conflicting evidence still exists regarding its prognostic value and, at present, it remains a IIb indication.21,22 However, it has gained value in the last few years with the publication of meta-analyses.5 Interestingly and in agreement with our own and previous data, induction with fewer extra stimuli was associated with higher risk. Nonetheless, and as already observed by Sroubek et al.,4 a negative EPS does not portent a low VA risk, particularly in patients with high-risk clinical features. Our data are even more disconcerting. We followed up 20 patients with a negative EPS (32 if the EPS is considered positive when VA were induced with 1-2 extra stimuli) and who developed cardiac events in the long term. Also relevant is the fact that its NPV and PPV are dependent on the follow-up period (table 1 of the supplementary data). Although this observation has never been analyzed, in our opinion, its predictive abilities seem to be time-dependent, and it could be argued that EPS should be re-evaluated at certain time intervals.

To overcome the limited value of the aforementioned individual factors, several authors advised the use of combined approaches. This is employed in some recommendations, for example, the Japanese Cardiology Society recommended ICD implantation for primary prevention in patients with BrS according to the number of risk factors (syncope, positive EPS, and family history of SCD).23 A potential advantage of the multiparametric approach lies in the enhanced specificity of combined risk factors compared with their individual components. One essential benefit of combined risk stratification is that patients without any risk factors have a sufficiently low incidence of arrhythmic events to justifiably circumvent specific therapy.3 In their multicenter studies (including 111 individuals, with EPS in 59), Letsas et al.10 evaluated the predictive ability of these risk scores. The Kaplan-Meier analysis showed that individuals displaying at least 4 risk factors (ECG, syncope, family history, EPS, fragmented QRS, and prolonged QRS in lead V2) were at increased risk of developing arrhythmic events. The ECG pattern, the presence of syncope and a positive EPS (albeit the Shanghai BrS score, which does not evaluate the role of the EPS) are common risk factors in the 3 models evaluated. Interestingly, the presence of a family history of SCD—although not statistically significant as an isolated variable at least in the Delise score—seems to help calibrate the final model and, although it was not the case in our cohort, it might be of help in borderline cases. More recently, Probst et al.11 evaluated the accuracy of the Shanghai BrS and Siera score. Predictive capacity was estimated by an area under the curve of 0.73 (0.67-0.79) and 0.71 (0.61-0.81). Of note, and in line with our results, risk scores did not allow stratification of the risk of arrhythmic events in intermediate-risk patients.

In our cohort, the 3 scores and a positive EPS with either 1 to 2 or 3 stimuli exhibited a low PPV and high mean NPV (taking into account competing risks). However, NPV exhibited a strong temporal dependence, being high (up to 100% at the beginning of follow-up, when no events were recorded) and with a steep decline at longer follow-up times (table 1 of the supplementary data). The FINGER trial reported that patients without symptoms and without a spontaneous type 1 ECG pattern were found to have a very low risk of SCD regardless of VF inducibility during EPS.11 In our sample, although the number of events in patients with few risk factors was low (table 2), it was not negligible. Two events occurred in patients with 0 risk factors in the Delise7 and Shanghai9 BrS scores, and events occurred in 12 patients with <2 points in the Sieira score.8 Likely, the inclusion of a significantly higher number of asymptomatic patients in the present study than in these series (154,7 74,10 269,8 and 2719 vs 627 in the present study) could have attenuated the predictive value of the scores. The addition of atrial fibrillation or sinus node dysfunction to the models (as proposed by the Shanghai9 and Sieria score,8 respectively) may have improved their predictive abilities, although the latter did not result in a significant change in our sample, most likely due to the low number of patients with these diagnoses. Indeed, it seems to be an uncommon diagnosis among patients with BrS, not only in our series. In Sieira series, only 2 out of 269 asymptomatic patients had this diagnosis.8

To sum up, we believe that these data emphasize the urgent need for better categorization tools in low-risk patients than the current conventional risk factors. Until such tools are available, particularly in the subset of asymptomatic BrS patients, and particularly in those with a spontaneous type 1 ECG pattern, we suggest the use of scores that incorporate the ECG and the EPS. However, as previously mentioned, due the lack of specificity of the EPS, it can lead to an overtreatment of patients at low risk. We believe that the management of these patients should be individualized and complemented with other advised markers such as the presence of fragmented QRS complexes, early repolarization, signal-averaged techniques, microscopic T-wave alternans,3 ventricular refractory period <200 ms,13 along with clinical factors such as atrial fibrillation and sinus node dysfunction.8,9 In contrast, a negative EPS should not be linked to low VA risk and these patients should also undergo periodic evaluation. Moreover, although we hypothesized that the adjustment of these risk scores by incorporating the number of extra stimuli required to induce VA could improve their predictive abilities, this was not the case (table 5), and induction of VA with 1 or 2 extra stimuli (compared with 3) did not improve the predictive performance measured by concordance index. Nevertheless, due to the positive value of the EPS with 1 or 2 extra stimuli at 1 right ventricular site compared with 3 extra stimuli in our model (together with age at diagnosis and type 1 ECG pattern) this protocol might be encouraged instead of the traditional one using 3 extra beats at 2 right ventricular sites.

Although the clinical profile of syncopal episodes was taken into account to exclude patients with noncardiogenic syncope, it is plausible that some vasovagal syncopal episodes could have been included, which could have swayed study outcomes. This study used a national registry, which may be subject to a referral bias and regional characteristics that vary among populations. Nevertheless, we believe that it is a good indicator of nationwide practice and outcomes. Moreover, the cohort included only BrS patients who underwent EPS. The referral for EPS was left to physician discretion rather than standardized indications and may have varied between centers. Therefore, the study results could not be uniformly extrapolated to the general BrS population. Arrhythmic events were classified according to SCD and ICD interrogation. Nonetheless, because ventricular tachyarrhythmias may terminate spontaneously, appropriate ICD shock is not synonymous with SCD. Moreover, not all SCDs can be attributable to BrS. This nuance could have led to an overestimation of the number of events.

An important limitation of the present study is that the risk scores could not be calculated in the entire study population due to missing information. This was particularly true of the Shanghai score (calculated in only 415 out of 831 individuals). As such, this study did not aim to compare the scores with each other, but instead to express their predictive abilities in our patients with BrS. We performed a sensitivity analysis removing all rows with missing values in the Shanghai score, without changes in either the interpretation of multivariable models or event prediction for the cohort. However, in asymptomatic patients, the loss of power due to removal of missing data caused the Delise score to be nonpredictive and a lack of convergence in the Shanghai score.

In a disease with a life-long risk of lethal arrhythmias, although comparatively longer than the vast majority of series, the present study might not have sufficient power to definitely exclude or confirm the predictive value of any of the aforementioned parameters. Moreover, the rate of cardiac events could be too small to allow firm conclusions to be drawn; further studies with a larger number of patients will be needed to confirm the conclusions of this study.

Finally, only the Shanghai BrS and Sieira scores seem to be adequately calibrated, whereas the Delise score showed poor calibration. The results of our study should be considered in light of this limitation.