A proof of principle study of atomoxetine for the prevention of vasovagal syncope: the Prevention of Syncope Trial VI

Abstract

Aims

There are few effective therapies for vasovagal syncope (VVS). Pharmacological norepinephrine transporter (NET) inhibition increases sympathetic tone and decreases tilt-induced syncope in healthy subjects. Atomoxetine is a potent and highly selective NET inhibitor. We tested the hypothesis that atomoxetine prevents tilt-induced syncope.

Methods and results

Vasovagal syncope patients were given two doses of study drug [randomized to atomoxetine 40 mg (n = 27) or matched placebo (n = 29)] 12 h apart, followed by a 60-min drug-free head-up tilt table test. Beat-to-beat heart rate (HR), blood pressure (BP), and cardiac haemodynamics were recorded using non-invasive techniques and stroke volume modelling. Patients were 35 ± 14 years (73% female) with medians of 12 lifetime and 3 prior year faints. Fewer subjects fainted with atomoxetine than with placebo [10/29 vs. 19/27; P = 0.003; risk ratio 0.49 (confidence interval 0.28–0.86)], but equal numbers of patients developed presyncope or syncope (23/29 vs. 21/27). Of patients who developed only presyncope, 87% (13/15) had received atomoxetine. Patients with syncope had lower nadir mean arterial pressure than subjects with only presyncope (39 ± 18 vs. 69 ± 18 mmHg, P < 0.0001), and this was due to lower trough HRs in subjects with syncope (67 ± 30 vs. 103 ± 32 b.p.m., P = 0.006) and insignificantly lower cardiac index (2.20 ± 1.36 vs. 2.84 ± 1.05 L/min/m², P = 0.075). There were no significant differences in stroke volume index (32 ± 6 vs. 35 ± 5 mL/m², P = 0.29) or systemic vascular resistance index (2156 ± 602 vs. 1790 ± 793 dynes*s/cm⁵*m², P = 0.72).

Conclusion

Norepinephrine transporter inhibition significantly decreased the risk of tilt-induced syncope in VVS subjects, mainly by blunting reflex bradycardia, thereby preventing final falls in cardiac index and BP.

Introduction

Recurrent vasovagal syncope (VVS) is a common medical condition that reduces quality of life and can be difficult to treat.¹ Although most patients do improve substantially without medical treatment many remain under-treated who might benefit from effective therapy. No medical therapies are proven effective by randomized controlled trials, although there is modest evidence for the effectiveness of fludrocortisone, midodrine, and beta blockers.¹

Norepinephrine transport (NET) inhibition shows promise as a novel treatment. Hypotension is very common in VVS,² and inhibition of the norepinephrine reuptake transporter increases synaptic norepinephrine.³ This might increase both cardiac preload and systemic vascular resistance, thereby ameliorating the hypotensive response. Alternatively, it might increase heart rate (HR),⁴ thereby increasing cardiac index. Both reboxetine and sibutramine, which inhibit NET, substantially reduce the proportion of positive tilt tests in healthy subjects⁵ and in a case series sibutramine reduced syncope frequency in highly symptomatic patients.⁶ However, neither sibutramine nor reboxetine are available in North America.

Atomoxetine, which is used to treat attention deficit disorder,⁷ is a highly selective NET inhibitor that might be useful clinically. Prior to embarking on a large clinical trial we conducted a proof-of-principle, randomized, placebo-controlled trial of the efficacy of atomoxetine to prevent VVS induced on tilt tests. While doing so we used continuous non-invasive beat-to-beat blood pressure (BP) and cardiac haemodynamic monitoring to test whether atomoxetine was efficacious by preventing venodilation, arteriolar dilation, or bradycardia.^2,8

Methods

Consent

The Sixth Prevention of Syncope Trial (POST 6) was approved by the all Institutional Review Boards and registered on Clinicaltrials.gov as NCT02500732. POST 6 complied with the Declaration of Helsinki. All subjects provided informed, written consent.

Definition of vasovagal syncope

Vasovagal syncope was defined¹ as a syncope syndrome that usually (i) occurs with upright posture held for more than 30 s or with exposure to emotional stress, pain, or medical settings; (ii) features diaphoresis, warmth, nausea, and pallor; (iii) is associated with hypotension and relative bradycardia, when known; and (iv) is followed by fatigue.

Study population

Patients were eligible if they had ≥1 syncopal spell in the year preceding enrolment, were ≥18 years, and had ≥−2 points on the Calgary Syncope Symptom Score.⁹ It has 90% accuracy in distinguishing VVS from other causes of syncope in subjects with structurally normal hearts.⁹ Patients were excluded if they had other causes of syncope; did not give informed consent; significant valvular, coronary, myocardial or conduction abnormality, or significant arrhythmia; hypertrophic cardiomyopathy; a permanent pacemaker; a seizure disorder; hypertension defined as >150/90 mmHg; pregnancy; glaucoma; medications with known effects on BP or NET inhibition; known hypersensitivity to atomoxetine and derivatives; known liver abnormalities; past or present pheochromocytoma; uncontrolled hyperthyroidism; or other factors which in the investigator’s opinion would prevent the subject from completing the protocol.

Enrolment/recruitment

Patients were recruited from the syncope clinics of the University of Calgary (n = 42); McMaster University, Hamilton (n = 6); Hopital Sacre Coeur, Montreal (n = 5); and University of Sherbrooke (n = 3).

Study design

POST 6 was a randomized, double-blind, parallel-arm study in which the subjects underwent a tilt table test following two doses of atomoxetine 40 mg (or matching placebo) ingested the evening before and morning of study. Heart rate and BP using a finger volume clamp method were recorded continuously and calibrated with intermittent brachial cuff BP measurements. Data were collected for offline analyses (see below).

Randomization

Randomization in permuted blocks of four was carried out using a computerized algorithm. Patients were randomized in a double-blind fashion to receive atomoxetine 40 mg PO ×2 or matching placebo with a 1:1 randomization ratio. Medication containers were centrally filled and labelled with the randomization code number.

Tilt table protocol

A period of at least 20 min elapsed before a 10-min baseline supine data collection period. The table was rapidly raised to 80° for up to 60 min.¹⁰ The study was terminated if the subject developed syncope (not simply presyncope), at the subject’s request, or at the completion of the protocol. Patients with presyncope were kept upright until they developed frank syncope, or to study termination at 60 min.

Outcomes

The primary outcome was the proportion of subjects who fainted associated with hypotension and bradycardia. Presyncope was defined as a state of lightheadedness that substantially reproduced the clinical presyncope of the patient. Given the unexplored effect of atomoxetine on tilt test haemodynamics we did not define a priori diagnostic values of haemodynamic variables. Secondary outcomes included the proportion of subjects who developed isolated presyncope, defined as presyncope that did not progress to syncope; the proportion of subjects who developed ≥1 of either syncope or isolated presyncope; nadir HR and BP; and changes in estimated stroke volume index (SVI), cardiac output index (CI), and systemic vascular resistance index (SVRI).^2,8,11 Unless otherwise noted haemodynamic variables were sampled in 10-s windows and are reported as the means.

Data analysis

Continuous HR and BP data were obtained during the tilt table study at each site and analysed in the Calgary core lab. In Calgary, continuous HR and BP data were acquired using a Nexfin System (BMEYE, The Netherlands) and advanced haemodynamic measurements (SVI, CI, SVRI) calculated with waveform-based Modelflow software. McMaster and Sherbrooke acquired continuous data using a Task Force Monitor (TFM; CNSystems Medizintechnik GmbH, Graz, Austria) Model 3040i, while Montreal used a TFM Model 3030i. Data from the TFM were downloaded and analysed in Calgary using LabChart (AD Instruments, Dunedin, New Zealand) by L.L. and G.B., blind to the randomized intervention. Analyses consisted of 1-min windows for averages of haemodynamic values during baseline and during tilt. Prior to and during presyncope and syncope, 10-s windows were averaged.

Statistical analysis

Coordination, data management, and analysis were performed at the University of Calgary. Parametric data are reported as mean ± standard deviation and analysed with a Student’s t-test. The Fisher–Freeman–Halton exact test was used to assess the statistical significance of the difference in outcome proportions. Time-dependent outcomes are displayed in a Kaplan–Meier analysis and the significance of differences estimated with the log-rank test and hazard ratios. Patients who exited the study before syncope or study end were censored in the syncope analysis. In the absence of a priori competing risks we did not perform competing risk analysis or linear mixed models. Data were entered into a Redcap database at the University of Calgary. SPSS version 21.0 (IBM Corporation, Armonk, NY, USA) was used for data analysis, and GraphPad Prism 6.0 (GraphPad Software, San Diego, CA, USA) to generate figures.

Sample size calculation

Sample size was calculated based upon the primary clinical endpoint. In a review of four studies¹² comprising 304 patients utilizing this tilt test protocol (at 60° head-up tilt) the likelihood of a positive test was 69%. We reported¹⁰ that the likelihood of a positive tilt test outcome was 1.1% per minute at 80°. After 60 min of 80° head-up tilt, we expect a positivity rate of about 66%. The effect of other NET inhibitors was reported using a slightly different tilt test protocol,⁵ and the positive outcome rate was 49% in the placebo arms. In the NET intervention arms, the positive outcome rate was 18%, of which 4% was due to diagnostic presyncope and 14% was due to orthostatic intolerance without bradycardia or hypotension,⁵ with a 92% relative risk reduction, which seemed higher than might be expected. A sample size of 56 syncope patients was chosen to provide 85% power to detect a 60% relative risk reduction from a placebo outcome rate of 65%, with alpha = 0.05.

Oral presentations

This work was presented at the European Society of Cardiology Congress in August 2018 and the Canadian Cardiovascular Congress in October 2018.

Results

Study population

Between August 2015 and January 2018, 56 consenting patients were randomized in four university hospitals in Canada. The mean age was 35 years (Table 1), and 73% were women. The subjects in the two arms were approximately evenly balanced: subjects in both arms had a median 3 syncopal spells in the preceding year and had very similar Calgary Syncope Symptom Scores. Study participants had a median 12 syncope spells over a median 15 years, with a median frequency of 1.4 syncopal episodes per year. Only one had a previous tilt test.

Symptomatic outcomes

All subjects either fainted or completed the study except for five whose tilt tests were stopped prematurely due to intolerable presyncope; they were included in the analysis. Atomoxetine significantly reduced the proportion of subjects who fainted on the tilt test and increased the proportion of subjects who developed isolated presyncope (Table 2). In the placebo arm, 19/27 (70%) subjects experienced frank syncope on tilt test compared to only 10/29 (35%) in the atomoxetine arm (P = 0.003) with a relative risk of 0.49 (confidence interval 0.28–0.86, P = 0.012). The absolute risk reduction was 36% and the number needed to treat to prevent a positive tilt test is 3. In contrast, the proportion of subjects who developed at least one of either isolated presyncope or syncope was similar in both arms [23/29 (79%) in the atomoxetine arm vs. 21/27 (78%) in the placebo arm; P = 1.0]. Strikingly, 13/15 patients with isolated presyncope had received atomoxetine and only two had received placebo. Only one subject had a possible drug-related unintended harm, which was mild transient insomnia.

Figure 1A displays the difference in survival free of syncope induced by tilt testing, with 65% vs. 30% fainting in the placebo and atomoxetine arms, respectively (P = 0.024, log-rank test). Figure 1B displays the similarities in survival free of either isolated presyncope or syncope induced by tilt testing, with 78% vs. 76% developing either isolated presyncope or syncope in the placebo and atomoxetine arms, respectively (P = 0.76).

In summary, atomoxetine does not prevent the development of the vasovagal reflex, but highly significantly prevents the progression of presyncope to syncope.

Haemodynamics in asymptomatic patients

Figure 2 displays the haemodynamic variables HR, mean arterial pressure (MAP), SVI, SVRI, and CI in subjects during the first 5 min of head-up tilt. Patients who received atomoxetine had no early significant differences between the groups in MAP, SVRI, or CI. In contrast, patients who received atomoxetine had slightly but significantly higher HRs (P = 0.008 to 0.05, t-tests) and significantly lower SVI than those who received placebo (P = 0.0007–0.01, t-tests). The directionally opposite changes in HR and SVI balanced each other to maintain a similar CI between the two groups.

Figure 3 displays the haemodynamic variables HR, MAP, SVI, SVRI, and CI for 60 min of head-up tilt in the 11 asymptomatic subjects, grouped by treatment allocation. Five patients received atomoxetine and six patients received placebo. During head-up tilt, there were no significant differences between the two treatment groups in any of the haemodynamic variables (P = 0.11–0.92).

Haemodynamics during isolated presyncope and syncope

Table 3 reports haemodynamic differences (and similarities) between syncope and isolated presyncope. Patients with syncope had lower nadir MAP than subjects with isolated presyncope (39 ± 18 vs. 69 ± 18 mm Hg, P < 0.0001), and this was due to lower nadir HRs in subjects with syncope than in subjects with isolated presyncope (67 ± 30 vs. 103 ± 32 b.p.m., P = 0.006), and a trend to a similar change in CI (2.84 ± 1.05 vs. 2.20 ± 1.36 L/min/m², P = 0.075). There were no significant differences in SVI (32 ± 6 vs. 35 ± 5 mL/m², P = 0.29) or SVRI (2156 ± 602 vs. 1790 ± 793 dynes*s/cm⁵*m², P = 0.72). Systemic vascular resistance index was not significantly different during isolated presyncope (P = 0.11) and syncope (P = 0.37) than after their respective 2-min head-up tilts (Table 3). During syncope HR and CI were both lower after their respective 2-min head-up tilts (Table 3, P < 0.01). The respective responses in subjects with isolated presyncope were more moderate: during isolated presyncope HR (P = 0.48) and CI (P = 0.08) were not significantly lower after their respective 2-min head-up tilts (Table 3).

This suggests that HR accounted for the increased CI and lesser drop in BP in subjects receiving atomoxetine. To test this we compared directly the haemodynamic changes in patients with isolated presyncope on atomoxetine and those with syncope on placebo. Figure 4 displays the marked changes in trough haemodynamic variables recorded during the 5 min preceding either syncope or the lowest MAP during presyncope, while receiving either atomoxetine or placebo respectively. In patients whose tests culminated in syncope there is an abrupt terminal collapse in MAP, HR, and CI but not in SVI or SVRI.

The MAP was significantly lower during syncope than isolated presyncope (34 ± 13 vs. 65 ± 12 mm Hg, P < 0.0001), while SVRI was not significantly different (2109 ± 675 vs. 2034 ± 525 dynes*s/cm⁵*m², P = 0.74). The lower MAP during syncope was associated with a lower CI (1.72 ± 0.92 vs. 2.73 ± 1.02 L/min*m², P = 0.009). In turn, the lower CI was associated with a lower HR (56 ± 32 vs. 100 ± 32 b.p.m., P = 0.0005) but not a lower SVI (30 ± 11 vs. 27 ± 6 mL/m², P = 0.47). Only preserved nadir HR accounted for the increased CI in subjects receiving atomoxetine.

Nadir heart rate

Atomoxetine effectively prevented profound bradycardia. A nadir HR <50 b.p.m. was noted in 2/23 symptomatic subjects receiving atomoxetine compared to 9/20 symptomatic subjects receiving placebo (P = 0.012). For 34 patients in whom RR intervals could be measured directly, a trough longest RR interval >1000 ms was noted in 1/14 symptomatic subjects receiving atomoxetine compared to 11/20 symptomatic subjects receiving placebo (P = 0.009). Figure 5 depicts the distribution of haemodynamic outcomes, in patients with presyncope or syncope, based on the VASIS classification.¹³ VASIS 3 outcomes (preserved or increased HR) were seen in 15/23 atomoxetine-treated patients and 5/20 placebo-treated patients (P = 0.023, χ² 2×4). In contrast, VASIS 2 outcomes (predominantly cardioinhibitory) were seen in 1/23 atomoxetine-treated patients and 7/20 placebo-treated patients (P = 0.023, χ² 2×4).

Discussion

Atomoxetine prevents the progression from presyncope to syncope, and this is associated with a reduction in bradycardia, in turn lessening the reduction in CI. It does not prevent the onset of presyncope, and appears to have no significant effect on systemic vascular resistance or stroke volume.

Treatment of vasovagal syncope

Generally, patients improve markedly with reassurance and education about salt, fluid, and counterpressure manouevres.¹ Some patients, however, might benefit from medical therapy.¹ Fludrocortisone was modestly effective only in a secondary analysis¹⁴ and cannot be used in the setting of even moderate hypertension. Beta blockers are ineffective in younger people, and the evidence for benefit is modest in patients over 40 years.¹⁵ Midodrine appeared to be effective in a meta-analysis of randomized trials,¹² but the studies had methodological flaws, and it cannot be used in even moderate hypertension. The role of dual-chamber permanent cardiac pacing in cardioinhibitory VVS remains debatable,¹ with an ISSUE 3 substudy¹⁶ suggesting a very limited role while the SPAIN study¹⁷ reported high efficacy.

Norepinephrine transport inhibition

Three small controlled studies showed that reboxetine and sibutramine reduced the likelihood of positive responses to tilt testing in healthy subjects without a history of syncope,⁵ and a small dose-ranging study of sibutramine in highly symptomatic, treatment-resistant syncope patients appeared to show benefit.⁶ Neither reboxetine nor sibutramine is available in North America. Whether any benefit of a NET inhibitor in VVS is a class action is unknown. Finally, there has been no randomized controlled study of NET inhibition to prevent syncope in patients with VVS.

Atomoxetine

Atomoxetine is a highly selective NET inhibitor used to treat attention deficit disorder.⁷ It markedly increases seated and standing systolic BP in patients with central autonomic failure with orthostatic hypotension.¹⁸ Given the effects of reboxetine and sibutramine, a randomized trial of atomoxetine with clinical outcomes is appealing. However, we felt it important to first perform an acute proof-of-principle study, and to understand its haemodynamic effects.

Haemodynamic effect of norepinephrine transporter blockade

Norepinephrine transporter inhibition with desipramine¹⁹ and reboxetine⁴ has both central and peripheral effects, both by increasing synaptic norephinephrine levels. Although peripheral NET inhibition increases HR and vasoconstriction, central NET inhibition reduces sympathetic outflow. The overall effect of oral NET inhibition is little if any effect on BP. However, and particularly in conditions such as orthostatic stress, NET inhibition increases HR.⁴ This is due to a direct effect on the sinus node, where synaptic norepinephrine is recycled through active transport by the presynaptic NET.³ Norepinephrine transporter recaptures about 90% of released norepinephrine in the heart, making it a critical mediator of NE inactivation and presynaptic catecholamine homeostasis.^3,19 There is a marked difference in clearance mechanisms in synaptic clefts in the sinus node and the vasculature, where only a small fraction of NET clearance is due to NE. This may be due to narrower synaptic clefts in the sinus node.¹⁹ Therefore, under stress NET inhibition predominantly causes sinus tachycardia rather than vascular effects.

The ability of maintaining HR as a means to prevent VVS may seem surprising, given the repeated failure of most clinical trials of permanent cardiac pacing to show benefit. However, Santini et al.²⁰ in an acute, placebo-controlled randomized study demonstrated highly significant benefit of atropine in preventing syncope induced by the drug-free Westminster tilt test protocol. Schroeder et al.⁵ reported that reboxetine induced sinus tachycardia while preventing VVS on tilt tests, and the SPAIN study reported¹⁷ a remarkable benefit of permanent cardiac pacemakers with a novel CLS sensor. Our study aligns well with others, in reporting early and sustained reduction in stroke volume, with a terminal collapse in HR. It might be that the CLS sensor accurately detects the onset of the terminal phase, while earlier sensors did not.

Clinical and mechanistic implications

This is the first (and positive) randomized study of NET inhibition in VVS. Given that all three NET inhibitors that have been assessed (reboxetine, atomoxetine, and sibutramine) decrease the likelihood of syncope on tilt testing,¹² it suggests that this is a class effect of NET inhibitors, and that this class of drugs merits assessment in a randomized clinical trial.

Atomoxetine did not appear to prevent the initial reduction in cardiac output and hypotension. Hypotension was mainly due to a decrease in stroke volume with preserved systemic vascular resistance.⁸ The fall in cardiac output suggests that atomoxetine does not cause venoconstriction to maintain preload. However, atomoxetine prevented profound bradycardia, defined as a trough HR <50 b.p.m., in nearly all patients. This suggests that atomoxetine prevents syncope by preventing the terminal bradycardia. This may be due its effect on CI, which is the product of HR and stroke volume. It maintains enough cardiac index and BP to prevent the progression of presyncope to syncope. This is not to imply that therapies that target vascular responses are not also effective.

Limitations

The subjects were young, and whether the findings can be extrapolated to older subjects is unknown.² We excluded potential subjects who were taking antidepressants with partial NET inhibitory activity, and whether atomoxetine interacts pharmacodynamically with these drugs remains to be determined. Third, the estimates of cardiac output and systemic vascular resistance are highly modelled and should be taken as only preliminary evidence. Their relative change over time may be more important than their absolute values. The unusual finding of a predominant effect on HR requires further exploration. Indeed, we have not ruled out an effect on brain NETs. We studied only VVS that could be induced by tilt testing, and it simply provides early evidence of a biological effect under orthostatic stress. Patients did not have a qualifying tilt test, and therefore, we do not know whether the electrocardiogram outcome of a positive tilt test predicts response on a tilt test to atomoxetine. Confirmation of these findings in a prospective, randomized, double-blind trial with clinical syncope recurrence as the outcome is required. Whether a suitable clinical balance can be struck between efficacy and side effects in clinical practice is unknown.

Conclusions

The NET inhibitor atomoxetine prevents the progression of presyncope to syncope on head-up tilt testing by attenuating sinus bradycardia, in effect, preventing syncope in patients with bradycardia during syncope. It is a suitable candidate for a formal randomized controlled clinical trial as an effective treatment for VVS.

Funding

This work was supported by the Grant #15-P01-001 from the Cardiac Arrhythmia Network of Canada (CANet), a Network of Centres of Excellence (NCE) of Industry Canada, Ottawa, Canada, and an unrestricted donation from John and Leslie Bissett. Atomoxetine was provided at cost by Apotex, Toronto, Canada, which also donated matching placebo.

Conflict of interest: none declared.

References

Author notes