Perhaps Past Performance Does Predict Future Results After All: A Key to Evaluating Treatment Interventions in Vasovagal Syncope

Syncope differs from most other recurring medical conditions by virtue of the fact that susceptible individuals are usually entirely asymptomatic (at least with respect to the symptom, syncope) between events.1 This is certainly the case for the vast majority of patients with recurrent vasovagal syncope (VVS), most of whom are otherwise generally healthy and manifest no overt evidence of their VVS susceptibility. Of course, many VVS-susceptible individuals do experience other symptoms (lightheadness, dizziness, and near faint) that may at times be confused with and seem to be “markers” of VVS. However, complaints such as these are most often nonspecific and should not be deemed as surrogate endpoints for true VVS in clinical trials.2 To evaluate effectiveness of possible preventative treatments for VVS, clinical studies have largely relied on enrolling patients with multiple prior VVS events, most often ≥3 syncopes in the past 2 years.3–10 However, there is as yet no consensus regarding the characterization of the optimal VVS clinical trial population; variations in admission criteria used in previous reports are readily apparent (Table 1). For example, the Vasovagal Pacing Study 1 and 2 (VPS1 and VPS2) permitted inclusion of patients with ≥6 episodes ever,11 while the PC-Trial included patients who reported ≥1 syncope and 3 presyncope events in the year prior to enrollment.10 In terms of designing clinical trials, sample size calculation critically depend on precise estimate of the expected number of endpoints to be encountered during follow-up (in this case the number of syncope events). Often, these estimates seem to be excessively optimistic, and the event rates are much lower than had been expected. The result may necessitate extending the follow-up observation period or declaring termination of the study due to futility. In the case of VVS, absent any evident interevent surrogates of the disease, most investigators have chosen past performance as an indicator of future event rates. However, the choices have been largely empirical, although heavily influenced by prior work from Sheldon et al.12 In that 1996 report, the number of syncopal spells preceding head-up tilt table test (HUT) and duration of symptoms were found to be predictors of syncope recurrence in the subsequent 2 years. Moreover, these 2 variables interact to modify the risk of syncope recurrence. Patients who have had few syncopal spells over a short period of time were at much lower risk of a recurrence than patients who have had a large number of syncopal spells over a short period of time. In this issue of the Journal, the same group of investigators13 has re-evaluated the utility of prior events for predicting subsequent VVS recurrence, and has provided a less complex means of predicting those patients who are likely to exhibit a VVS recurrence in the near future. In this analysis, based on the Prevention of Syncope Trial (POST) clinical trial study population,8 it seems that event rates in the year prior to enrollment was a stronger discriminator of subsequent 1-year syncope risk than was prior lifetime events. Compared with prior lifetime events, the number of events in the previous 1 year provided a more pronounced ‘dose-response relationship’ with respect to risk of syncope recurrence. Thus, the 12-month risk of syncope recurrence ranged from approximately 22% for individuals with 0–1 syncope in the prior year to almost 70% for subjects with >6 events in that same period. By contrast, lifetime event counts exhibited a less pronounced gradation of effect; for example, a subsequent 12-month risk of approximately 30–35% was associated with lifetime syncope counts of <6, 6–9, or 10–20. Only after >20 events were reported was the subsequent 12-month risk markedly increased (approximately 70%). The observations reported by Sumner et al.13 may prove to be of clinical value in terms of aiding physicians manage expectations of patients presenting with recurrent vasovagal faints. Furthermore, the authors suggest that their findings may help to enhance the efficiency of clinical trial designs in VVS; more specifically, by permitting more precise estimates of endpoint events, they argue that the size of proposed studies may be determined more accurately. There is little doubt that the results of this study will prove helpful for clinical study designers. However, several limitations will need to be taken into account. First, as these same authors have pointed out elsewhere,14 the trigger for patients to present for evaluation is often an abrupt rise in the number of symptom events (i.e., a burst of VVS activity). Perhaps, the previous 1-year event count and follow-up analyzed by Sumner et al.13 is simply reflecting an ongoing burst of activity. Given more time, this burst might subside and syncope frequency will return to its mean value for a given patient. If that is the case, the observation may prove to be useful in clinical trial design, however may not be any more than a biostatistical inevitability (regression to the mean). Second, as the authors have indicated, the results of this study are susceptible to recall bias. Although one can assume that data on syncope frequency in the preceding 1 year are fairly accurate, one has to question the accuracy of data on lifetime syncope frequency. The lack of clear separation in syncope-free survival rates stratified by quartiles of lifetime syncope frequency may simply reflect exposure misclassification. It may be that patients cannot accurately remember the number of lifetime syncope episodes until they have had more than 20 episodes. The misclassification of patients reporting 1–5, 6–9, and 10–20 lifetime syncope episodes may have resulted in the apparently observed subsequent lack of difference in syncope recurrence rates among these groups. Third, as the authors admit, many of the patients included in this report were being treated with beta-blockers. They argue that, absent evidence that beta-blockers offered any therapeutic benefit in the POST study,8 it is therefore reasonable to include these individuals. However, given the regression to the mean concern raised above, the possibility exists that a drug effect could have played a role on recurrence rates but that the POST trial was underpowered to detect such an effect given the tendency for postevaluation event rates to fall anyway. It might have been desirable to separate treated and untreated populations in order to demonstrate that they were comparable. Finally, and most importantly, in their analysis Sumner et al.13 relied on HUT as one of the diagnostic markers of VVS. However, they did not determine whether the subjects had experienced faints due primarily to cardioinhibitory, vasodepressor or mixed mechanisms. This limitation may become important as suggested by the International Study on Syncope of Uncertain Etiology 2 (ISSUE 2) study findings.7 In ISSUE 2, clinical observations during VVS (as documented by insertable loop recorders [ILRs]) provided clear assessment of the basis of the faint and illustrated the relatively reproducibility of spontaneous events. Consequently (although not a randomized clinical trial [RCT] and conceding that mean age was greater in ISSUE 2 [approximately 66 years vs 37 years in POST]), ISSUE 2 investigators were better able to select therapy for their patients based on ILR subset identification; in fact, cardiac pacing appeared in particular to be beneficial in patients with cardioinhibitory/asystolic VVS, despite RCT reports to the contrary.4,5 Although the veracity of the ISSUE 2 observations is being re-assessed in ISSUE 3,15 the possibility is raised that in the assessment of VVS therapies it is not syncope per se that matters as much as the precise pathophysiologic VVS subtype. Thus, prior syncope event rates absent specific pathophysiological subclassification may not prove as helpful as Sumner et al.13 surmise; it may be crucial that a therapy be tested in a population with a specific subtype appropriate for the therapy under consideration. In the case of pacemaker studies, a study population enriched with cadioinhibitory VVS may be more desirable, whereas in the case of vasoconstrictor agents, a primarily vasodepressor subset would be a more reasonable target. “Mixed” pathophysiology would presumably be the preferred target for clinical trials examining combination therapies. In summary, the optimal selection of a patient population for clinical trials in VVS may well be far more complex than attempting to identify a “recurrence-rich” population. As is widely accepted to be the case with most other cardiac arrhythmias (e.g., supraventricular and ventricular tachycardias), and arrhythmia-susceptible states (e.g., various channelopathies) in which targeting a specific arrhythmia type is essential in order to define effectiveness of any treatment strategy, the same may be true of multifaceted autonomic dysfunction states such as VVS. Nevertheless, despite the several limitations outlined here, Sumner et al.13 are to be applauded for an important contribution to our better understanding of the natural history of VVS. Their work has focused our attention on the need to better define optimal enrollment criteria for clinical trials of VVS, and their observations will almost certainly help future researchers develop more efficient and cost-effective trial design. In essence, they have taught us that recent past performance may indeed predict future results after all.