Editorials
Importance of Validating Guideline Recommendations
2020 年 84 巻 12 号 p. 2136-2137
詳細
2020 年 84 巻 12 号 p. 2136-2137
Guidelines aim to both help physicians in individual countries or around the world in decision making on a specific topic, and to reflect the current status of the field by including all available relevant literature. Most frequently they deal with risk stratification in and treatment of specific disease entities. Within the guidelines, recommendations surface after extensive discussion between experts who have evaluated all that literature. Such recommendations usually require a threshold for consensus of ≥80%. Outside the scope of a guideline is the ultimate test whether the propositions in the form of recommendations perform sufficiently well in daily practice and therefore separate, independent, attempts to do so are most welcome.
Article p 2158
In this issue of the Journal, Wakamiya et al1 present a multicenter retrospective study involving 234 Brugada syndrome (BrS) patients from 7 Japanese hospitals. They compared the guidelines published in 2012 with the newest guidelines published in 2018,2 in terms of appropriate prediction of the risk of a (potentially) lethal event. Unfortunately, both guidelines are only available in the Japanese language, which I have not mastered but it is sufficiently clear that in the 2012 guidelines 3 risk factors, syncope, family history of sudden cardiac death and inducibility of VF by programmed electrical stimulation (PES), were weighted equally. A BrS patient who met ≥2 of these risk factors had a Class IIa indication for an ICD and a patient who met only 1 risk factor was Class IIb. The 2018 guideline refined the criteria for syncope with an emphasis on arrhythmic syncope (differentiating it from non-arrhythmogenic syncope) and for PES with a maximum of 2 extra stimuli for inducibility of VF. Recommendations for an ICD were similar; that is, Class IIa for a patient meeting ≥2 of these risk factors and a Class IIb indication for a patient meeting only 1 risk factor.
Both refinements are based on relatively recent studies. Syncope is a frequent symptom in the general population and most often non-arrhythmic in nature. There is no a priori reason to believe that this is different in BrS patients. Indeed, in a single center study comprising 118 BrS patients with syncope, almost 60% was diagnosed, based on accurate history taking, as suspected non-arrhythmic.3 Compared with patients with aborted cardiac arrest (ACA), patients with suspected non-arrhythmic syncope were younger at first event, more likely to be female and more likely to have prodromes (which were never present prior to a cardiac arrest). Follow-up of these patients was uneventful, whereas follow-up in patients with suspected arrhythmic syncope or in ACA patients was not (cardiac arrest in respectively 0%, 12% and 43% during a median follow-up of 54 months (quartiles 31–91)). An earlier, smaller, study provided similar data and outcome.4 Careful history taking is thus of major importance to discriminate between non-arrhythmic and potentially arrhythmic syncope. In the current study the yield of history taking was, however, surprisingly low, as 62% remained unexplained compared with for example 15% and 30% in the aforementioned studies.3,4 The authors explain this as potential selection bias because only patients who underwent an electrophysiological study were included.1
The second refinement concerns the use of the PES data, which was collected for all patients with pacing protocols at 2 sites (right ventricle apex and right ventricular outflow tract), at 3 cycle lengths, and with 3 extra’s with coupling intervals >180–200 ms). In the new guidelines only inducibility with ≤2 extrasystoles was counted as a risk factor instead of ≤3 extrasystoles as in the previous protocol. Inducibility as a risk factor for future events has been a heavily debated issue over many years, in particularly in asymptomatic patients with a (spontaneous) type 1 ECG.5–7 Differences in the applied PES protocol are often mentioned as contributing to the different outcomes.5–8 Indeed, in 2012 it was demonstrated that the number of extrastimuli that induced ventricular arrhythmia potentially may serve as a prognostic indicator of patients with Brugada type 1 ECG.9 A later pooled analysis confirmed these initial findings, with a clear statement that triple extrastimuli are non-specific and therefore of limited clinical utility.10
The results of the present study are interesting and show that in a significant subset of patients another decision as to ICD treatment would have been reached, assuming that physicians would implant only with a Class I and IIa indication. Following the former guidelines 150 would have been implanted (in reality this number was 175, indicating that the assumption above is not completely valid) whereas with the new guidelines 99 patients would be implanted (Class 1 and IIa respectively 46 and 45). Limiting the number of ICDs in young individuals is important because the complication rate is not trivial: in BrS an annual complication rate of 3.4% and inappropriate shocks in 3.9% of patients has been observed.11 In the new guidelines the majority of patients shift to the unclassified group (i.e., the group without 1 of the 3 defined risk factors). In this group the risk for future events is not zero but very low (n=2 out of 107 patients in mean 6.9±5.2 years). One of these patients was inducible with 3 extra’s and the other had an unexplained syncope (and a drug-induced ECG).
As indicated, the accuracy of the predicted risk is particularly low in asymptomatic BrS patients. The present study included only 88 asymptomatic patients (with and without spontaneous type 1 ECG) and only 1 patient developed ventricular fibrillation (VF) during follow-up. These numbers are not much different to another risk prediction study with 3 events in 192 patients with ‘0’ points (i.e., all without a spontaneous type 1 ECG; follow-up 80.7±57.2 months).12 These 88 asymptomatic patients were classified in the former and new guidelines as, respectively, Class IIa (29 and 0), Class IIb (49, including the patient with VF, and 32), and unclassified (10 and 56, including the patient with VF). In the pooled analysis referred to, the differences in annual event rate during follow-up between inducibility (with up to double extrastimuli) and non-inducibility is small (spontaneous type 1 ECG: 1.70 (95% confidence interval (CI): 0.73–3.35) vs. 0.78 (95% CI: 0.36–1.47); drug-induced type 1: 0.45% (95% CI: 0.01–2.49) vs. 0.23% (95% CI: 0.05–0.68)) compared with the much higher risk in individuals with syncope and a spontaneous type 1 ECG (5.60%, 95% CI: 2.98–9.58).10 It is likely that some of these events will be preventable with appropriate life-style adjustments and, importantly, as some of these events occur late during follow-up it is also possible that a changing clinical profile during follow-up would have led to a timely decision for an ICD.
In summary, the current Japanese guidelines seems to perform well in discriminating high-risk from low-risk patients. However, in the low-risk group rare events do occur (as in other studies),12 but I predict a hard time accurately identifying these rare patients. The alternative of ICD treatment for everyone is certainly not the right solution. At this end of the spectrum larger cohorts are needed to make a step forward.
The Netherlands Cardiovascular Research Initiative (grant no. CVON2018-30 PREDICT2).
