This retrospective study received Ethics approval from The Joint Chinese University of Hong Kong – New Territories East Cluster Clinical Research Ethics Committee and is based on datasets that have already been made available in an online repository (Tse et al., 2019a, b, c). The study inclusion criteria include: (1) spontaneous, fever or drug-induced presentation of type 1 or 2 BrS; (2) undergone ICD implantation between 1997 and 2018. Type 1 BrP is defined as a coved-shape ST segment with elevation of >2 mm followed by a negative T-wave, and type 2 pattern is defined as convex ST segment with >0.5 mm elevation followed by variable T-wave, resulting in a saddleback-shaped morphology (Antzelevitch et al., 2005). The diagnosis of BrS is made based on the 2017 ACC/AHA/HRS Guideline (Shen et al., 2017), after reviewing documented patient history, and confirmed by analysis of all documented ECG by S.L. and G.T.

The cohort was divided into three groups based on patient history prior to ICD implantation: (1) asymptomatic; (2) syncope; (3) ventricular tachycardia/ventricular fibrillation (VT/VF). Asymptomatic was defined as the absence of syncope and VT/VF manifestation, whilst the syncope and VT/VF group was categorized based on the patient presentation of the respective symptoms. Patients presented with both syncope and VT/VF were included in both groups.

Clinical data was extracted from manual review of case records. The following baseline clinical data was collected: (1) sex; (2) age of initial BrP presentation; (3) family history of BrS and SCD/VF; (4) type of BrP presented, presence of evolution and fever; (4) presence of other arrhythmias [supraventricular tachycardia (SVT) and ectopic beats (SVE), atrial fibrillation (AF), and tachycardia (AT), atrial flutter]; (5) inducibility in electrophysiological study (EPS); (6) abnormalities in 24-h Holter; (7) presentation of syncope and VT/VF; (8) follow-up period; (9) quinidine use (dosage and duration); (10) baseline average ECG parameters recorded on the first available ECG. Follow-up period is defined as the period from initial presentation of BrP to patient’s death, time of contact loss, or until 22th April 2019. Patients are considered to present type 1 BrP if type 1 pattern is presented initially, or at any point in follow-up. Evolution of ECG pattern is defined by the change of BrP type, or resolution of pattern, during continuous follow-up. The baseline average ECG parameters include: (1) ventricular rate; (2) P wave duration; (3) PR interval; (4) QRS interval; (5) QT interval; (6) QTc interval; (7) P wave axis; (8) QRS axis; (9) T wave axis; (10) lead V1 S wave amplitude; (11) lead V5 R wave amplitude.

Furthermore, the following details on ICD implantation was collected: (1) date and type of initial and subsequent ICD implantation; (2) number of appropriate and inappropriate shocks; (3) reason of inappropriate shocks; (4) experience of lead malfunction, dislocation, and infection; (5) number of VT/VF episodes in total, prior to and after ICD implantation. Appropriate shocks are defined as shocks delivered to correct sustained VT/VF, whilst inappropriate shocks are shocks delivered for other reasons. Device setting to determine sustained VT/VF was based on guidelines at the time of device implantation, and subject to further adjustment given the patient’s condition. Episodes of VT/VF include both symptomatic and asymptomatic events detected by the ICD. Lead malfunction includes lead failure and fracture that required surgical extraction or replacement.

All statistical analysis was performed using Stata MP13. For clinical characteristics, categorical variables were compared between the three groups using Fisher’s exact test, whilst continuous variables were compared by Kruskal–Wallis one-way ANOVA. For baseline characteristics and ICD outcomes with significant inter-group differences, the continuous variables were compared using Dunn’s pairwise comparison test with Sidak correction, and the categorical variables undergone pairwise comparison by Fisher’s exact test again for further examination. Appropriate shock-free survival was estimated by the Kaplan–Meier survival curve and compared through the log-rank test. To examine the individual contribution of clinical characteristics to the execution of appropriate shocks during follow-up, univariate Cox proportional hazard ratio regression with Efron’s method for ties was performed on the entire cohort, and the three subgroups.

Variables with P-value < 0.05 in univariate analysis were then selected into the multivariate Cox regression model for the prediction of appropriate shock. The analyzed clinical features include: (1) sex; (2) age of initial BrP presentation; (3) presence of pattern evolution; (4) presentation of type 1 BrP; (5) fever-induced pattern; (6) presentation of syncope; (7) presence of other arrhythmia; (8) inducible EPS. Similarly, the Cox regression was used to examine the effect of the following clinical features on the occurrence of inappropriate shock: (1) age at which BrP ECG manifested; (2) evolving ECG pattern; (3) syncope; (4) other cardiac arrhythmias. Statistical significance is defined as P-value < 0.05.

The present cohort included 136 Han Chinese BrS patients receiving ICDs (mean age of initial BrP presentation: 50.0 ± 15.4 years, male: 96.3%, spontaneous type 1 pattern: 64.7%). Their baseline characteristics of the cohort categorized into asymptomatic (n = 26, 19.1%), syncope (n = 99, 72.8%), VT/VF (n = 36, 26.5%) groups are summarized in Table 1A. Of the 136 patients, 25 patients suffered from both syncope and VT/VF. ICD implantation took place for both primary prevention amongst patients who were asymptomatic or only presented with syncope (n = 100, 73.5%) or secondary SCD prevention in patients presented with VT/VF (n = 36, 19.1%). ICD implantation in asymptomatic patients were performed based on VT/VF inducibility in EPS (n = 16), and a combination of positive drug challenge test and family history of SCD (n = 1) (Priori et al., 2013). The remaining devices were implanted at the patient’s request after discussion with the attending physician. Amongst patients with type 1 BrP, 78 patients (88.6%) presented with type 1 pattern initially, and for 10 patients (11.4%) the pattern resolved or evolved into type 2 during follow-up. Other types of arrhythmia were presented in 35 patients (25.7%), notably SVT (n = 10), SVE (n = 3), AF (n = 10), and AT (n = 4) and atrial flutter (n = 3). 4 (2.94%) and 14 (10.3%) patients have a family history of BrS and SCD/VF, respectively.

The baseline ECG parameters are summarized in Table 2A. P-wave duration and R-wave amplitude in lead V5 differed significantly between the three subgroups (Table 2B). The syncope group had significantly longer P-wave duration than the asymptomatic group (p = 0.021). Moreover, R-wave amplitude in lead V5 differs significantly in the following descending order: asymptomatic, syncope, VT/VF (P-value: asymptomatic vs. syncope = 0.063, asymptomatic vs. VT/VF = 0.015, syncope vs. VT/VF = 0.034).

In the present cohort, 29 patients were prescribed quinidine over the course of follow-up (asymptomatic = 2; syncope = 23; VT/VF = 14; mean single dose = 404 ± 163 mg; mean daily dose = 670 ± 297 mg; mean duration = 1573 ± 2178 days). The reasons for quinidine prescription were (1) recurrent VT/VF (n = 14); (2) recurrent PVC (n = 4); (3) VT/VF prophylaxis before ICD implantation or replacement (n = 6); (4) VT/VF prophylaxis for patients who initially refused ICD implantation (n = 2). Indications for quinidine use in three patients were not documented. Two patients were prescribed quinidine for recurrent VF, but the drug was withheld for observation of ICD shock frequency instead. Out of the 16 patients who were prescribed and given quinidine due to recurrent arrhythmia, 11 patients were treatment-responsive, 4 were non-responsive, and one patient was intolerant with symptoms of vertigo and numbness.

The mean follow-up period is 97.5 ± 61.8 months (median = 95 months, interquartile range = 77.3 months), with 7 patients (mortality rate = 5.15%) passed away. The underlying causes of death, none of which was cardiogenic, include malignancy (n = 2), chronic obstructive pulmonary disease (n = 2), sepsis (n = 1), pneumonia (n = 1), and non-ST-elevation myocardial infarction (n = 1). Amongst the 68 patients (50.0%) who undertook EPS, 59 patients (20.6%) was found to be VT/VF inducible. Within the group of 28 patients (20.6%) who undergone 24-h Holter monitoring, rhythm abnormalities were found amongst 15 patients (53.6%). Insignificant intergroup differences were found in both the inducibility of EPS and the occurrence of abnormalities during 24-h Holter. Genetic testing was performed in 20.6% of the cohort (n = 28) and of these four patients tested positive for SCN5A mutation. All four patients belonged to the syncope group, and one patient is also a part of the VT/VF group.

For the baseline characteristics, significant inter-group differences were found in the family history of SCD, performance of EPS and follow-up period. Findings under further analysis are presented in Table 1B. Family history of SCD/VF was noted to be more commonly observed amongst the asymptomatic group than the VT/VF group (p = 0.017). Furthermore, the prevalence of EPS delivery differed significantly when the three subgroups undergone pairwise comparison, in descending order of prevalence: asymptomatic, syncope and VT/VF group (P-value: asymptomatic vs. syncope = 0.046; asymptomatic vs. VT/VF ≤ 0.0001; syncope vs. VT/VF = 0.017). Additionally, the follow-up period for the asymptomatic group is significantly longer than the VT/VF group (p = 0.010).

In total, 252 ICD has been implanted (n = 252, average per patient = 1.85 ± 1.00), of which the vast majority is automatic (n = 238, 94.4%), and the remaining devices were subcutaneous (S-ICD) (n = 14, 5.56%). 51.5% of the cohort undergone ICD replacement at least once (n = 60, mean years before replacement = 5.06 ± 1.81). The outcomes of ICD therapy are presented in Table 3A. The overall complication rate is 22.1% (n = 30), with device and infection-related complications present in 9.56% of the cohort (n = 13), including 5 patients with lead malfunction (3.68%), 6 patients with lead dislocation (4.41%), and 3 patients with post-operative wound infection (2.21%). No significant inter-group differences were found for the complications. 6 patients required lead repositioning due to lead dislodgement or noise sensing, and 4 patients ultimately undergone ICD removal after initial implantation, and/or subsequent ICD replacement due to infection (n = 4), lead erosion (n = 1) and infective endocarditis (n = 1). S-ICD was implanted in 1 case of lead malfunction, 1 case of lead dislocation, and 2 cases of infection.

Appropriate shocks were executed in 25.0% of the cohort during the follow-up period (n = 34, male = 100%, age = 49.3 ± 15.2). Amongst these patients, 19 patients (55.9%) presented with type 1 pattern, and 11 patients (32.4%) experienced pattern evolution. It is worth noting that three patients (8.82%) were previously asymptomatic, and 17 patients (50.0%) did not experience VT/VF prior to ICD implantation. Amongst the 17 patients, two patients have relevant family history (BrS: n = 1; VF/SCD: n = 2), 9 out of the 1010 patients with EPS performed showed inducible VT/VFS, and 7 of the 10 with drug challenge performed were tested positive. Ten patients presented with type 1 BrP, and three of these patients experienced pattern evolution to a type 2 pattern. Patients with appropriate shocks experienced an average of 4.71 ± 12.8 VT/VF episodes in total.

Significant differences were found when comparing the delivery of appropriate shocks between initially asymptomatic and symptomatic patients (Figure 1A, p = 0.007). Further comparison of survival estimates between the three patient groups is presented on Figure 1B), where the VT/VF group is presented to be most likely to receive an appropriate shock, followed by the syncope group (P-value: asymptomatic vs. syncope = 0.062, asymptomatic vs. VT/VF = 0.003, syncope vs. VT/VF = 0.034). Table 3B shows that under pairwise comparison, the VT/VF group is found to be significantly more likely to experience appropriate shocks in comparison to both the asymptomatic and syncope group. Whilst it is expected for the number of VT/VF episodes prior to ICD implantation to be significantly higher in the VT/VF group, it is worth noting that the difference in number of VT/VF episodes between the asymptomatic and VT/VF group was not significant post implantation (p = 0.063). Significant difference was only found when comparing between the syncope and VT/VF group (p = 0.014). VT/VF group remains to have significantly higher number of total VT/VF episodes in comparison to the remaining two subgroups (P-value: asymptomatic vs. syncope = 0.365; asymptomatic vs. VT/VF ≤ 0.0001; syncope vs. VT/VF ≤ 0.0001).

The results for the univariate Cox regression analysis of clinical predictors for the occurrence of appropriate shock in both the entire study cohort and the subgroups are presented in Table 4A. Concomitant presence of other arrhythmia is an independent predictor for appropriate shock delivery overall (p = 0.040, HR = 2.09, 95% CI = [1.03, 4.22]). None of the clinical characteristics showed statistical significance in their prediction for appropriate shocks for the asymptomatic group. The presentation of other arrhythmia is also predictive for the syncope group (p = 0.038, HR = 2.328, 95% CI = [1.05, 5.14]), while the male sex (p = 0.014, HR = 32.4, 95% CI = [2.02, 519]) and the presentation of type 1 BrP (p = 0.040, HR = 4.90, 95% CI = [1.07, 22.3]) are predictive for the VT/VF group. Given their statistical significance in the prediction of appropriate shock delivery, the three clinical characteristics: (1) male sex; (2) presentation of type 1 pattern; (3) concurrent presentation of other arrhythmia were included in the multivariate Cox regression analysis predictor model. However, as illustrated on Table 4B, type 1 BrP and the presence of other arrhythmia are significant predictors in the multivariate model (P-value: male = 0.674; type 1 pattern = 0.035; other arrhythmia = 0.041). The results of sensitivity analysis by excluding females are shown in Supplementary Tables 1A,B.

17.6% (n = 24) of the cohort experienced at least one inappropriate shock. The reasons underlying inappropriate shocks include SVT (n = 10), AF (n = 8), noise-sensing (n = 2), T-wave oversensing (n = 2), and transcutaneous electrical nerve stimulation (n = 1). Intergroup differences for the delivery of inappropriate shocks were statistically insignificant (p-value = 0.676). Given most cases of inappropriate shocks reported in this cohort is related to concomitant arrhythmias, it was found to significantly increase the risk of inappropriate shock (p ≤ 0.0001, HR = 5.09, 95% CI = [2.17, 11.9]). The univariate Cox proportional hazard ratio analysis of the clinical characteristic for the delivery of inappropriate shocks is summarized in Table 5. Amongst patients with inappropriate shock delivered, 41.7% (n = 10) received appropriate shocks as well. The results of sensitivity analysis by excluding females are shown in Supplementary Table 2.

Several limitations should be noted for the present study. Firstly, due to the multicenter nature of the present study, inconsistency may be present in data collection despite our best effort. Secondly, since this was a retrospective study, ICD technology and their settings have evolved over the course of follow-up. ICD programming may have differed between centers, and the definition of sustained VT/VF requiring ICD shock was determined by individual physicians who entered the case records without standardization. Thirdly, one patient was lost to follow-up. Fourthly, the use of appropriate shock as a surrogate for SCD may result in an overestimation of the death risk. Finally, we were not able to access data stored on the ICDs, such as program settings and ECG waveforms, but relied on case records written by the physicians. Future studies can use stored ICD data to determine the frequency and occurrence of pattern conversion between types 1 and 2, as well as correlation between appropriate shocks and conversion into type 1 close to the event.