Prolonged implantable electrocardiographic monitoring indicates a high rate of misdiagnosis of epilepsy—REVISE study

Aims Syncope, epilepsy, and psychogenic psuedo-syncope are the most common causes of transient loss of consciousness (T-LOC or blackout). All can present with similar features, including abnormal limb movements. It is reported that somewhere between 13 and 42% of patients with ‘epilepsy’ may be misdiagnosed. A UK Parliamentary working group found that at least 74 000 English patients are misdiagnosed with epilepsy, and taking antiepileptic drugs. The likely alternative diagnosis is ‘convulsive’ syncope, mimicking an epileptic seizure. We hypothesized that many patients misdiagnosed with epilepsy have convulsive reﬂex syncope, and that prolonged electrocardiographic (ECG) monitoring with an implantable ECG recorder (ILR) would show reﬂex cardioinhibition during T-LOC. This would respond to permanent pacing and allow antiepileptic drugs to be withdrawn. We also aimed to evaluate tilt testing and other tests done in these patients. study shows a high incidence of the cardioinhibition of reﬂex syncope in patients with convulsive T-LOC previously diagnosed as epilepsy and treated with antiepileptic drugs. We believe that reﬂex syncope with convulsive features mimics generalized epilepsy, leading to a misdiagnosis. This may be a widespread problem accounting for many wrong diagnoses of epilepsy. There was also poor correlation in ECG ﬁndings between tilt testing and ILR recording. - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - A 60-year-old man with Chagas disease, implanted with an implantable cardioverter deﬁbrillator (ICD) (Medtronic Maximo-VR7232) was hospitalized for electrical storm. Implantable cardioverter deﬁbrillator interrogation revealed 62 episodes of monomorphic ventricular tachycardia (VT). Analysing implantable cardioverter-deﬁbrillatordeﬁbrillator electrogram (ICD-EG), a difference in morphology was seen only in the HVA/HVB electrogram (a QS complex in VT1 and a Qr complex in VT2; see Figure 1 ). During hospitalization, he experienced several VT episodes and surface-electrocardiogram (ECG) showed two different QRS conﬁgurations, conﬁrming multiple morphology (MM)-VT. This difference in QRS morphology correlated to different ICD-EG, resembling those found before admission. At electrophysiological (EP) study, performed under ﬂuoroscopic navigation, VT1 was easily induced. With the use of conventional EP-mapping and entrainment techniques a reentry circuit in left ventricular (LV) posteroseptal zone (sites 4–6 from the LV mapping schema proposed by Cassidy and Josephson) was identiﬁed and successfully ablated. Repeated


Introduction
Syncope, epilepsy, and psychogenic psuedo-syncope are the three most common causes of transient loss of consciousness (T-LOC/ blackout). 1 -3 All three can present with similar clinical features, including abrupt T-LOC without warning, abnormal limb movements such as myoclonic jerks or tonic-clonic activity, 4 -8 and incontinence. Establishing a diagnosis is a clinical challenge, based on the history and eye-witness accounts. Abnormal movements are easily equated to a 'seizure', leading to a misdiagnosis of epilepsy, 1 but also occur frequently with cerebral anoxia during syncope. In order to confirm a clinical diagnosis objectively, an epsiode of T-LOC has to be captured during physiological recording, 7,8 as few clues exist between attacks. The surest method available to do this is videotelemetry, where patients are admitted for long periods under video surveillence during continuous electroencephalography (EEG) and electrocardiographic (ECG) monitoring. This is very costly, with very limited availability, and many patients thought to have epilepsy are treated on the basis on the clinical features alone. Clinical review and supplementary tilt testing in some cases disclose that between 13 and 42% 9 -12 of patients with epilepsy are misdiagnosed. The alternative diagnosis is syncope, by far the most likely cause of T-LOC. 1 Elsewhere, 8 it has been noted that patients with reflex syncope with abnormal movements have a higher rate of cardioinhibition during tilt table testing than those without abnormal movements. This serves to suggest that many patients with a diagnosis of epilepsy could have reflex syncope with abnormal limb movements, or 'convulsive syncope'. 3 We believed that prolonged electrocardiographic monitoring with an implantable ECG recorder (ILR) would show that many patients would have cardioinhibition during T-LOC. We anticipated that these patients would respond to permanent pacing and allow antiepileptic drugs (AEDs) to be withdrawn.

Aims
The aims of this study were to (i) determine the incidence of cardioinhibition during T-LOC/blackout using ILRs in patients with a previous diagnosis of epilepsy; (ii) correlate the results of tilt testing with the findings of the ILR; (iii) determine the value of other tests in this group of patients; and (iv) implant pacemakers in patients with cardioinhibition, stop their anticonvulsants and assess their progress.

Methods
We describe our experience of 103 patients previously diagnosed with epilepsy by neurologists and generalists, from two subgroups, one retrospective (n ¼ 62) and the other prospective (n ¼ 41). Both subgroups presented with T-LOC with convulsive movements and were evaluated in very similar ways. Patients with T-LOC but without any abnormal limb movements were excluded.

Retrospective group
The retrospective group came from 335 patients who underwent an ILR (Reveal w /Reveal Plus w ; Medtronic Inc.), for T-LOC at the Manchester Heart Centre, Manchester Royal Infirmary, UK, between 1996 and 2006. One hundred fifty-seven (47%) of these patients had been referred by the neurologists, out of which 62 of 157 (40%) of the referrals were for 'epilepsy'. The neurologists had made a 'possible' diagnosis of epilepsy in 45 of 62 (73%) and a 'confirmed' diagnosis in the remainder [17 of 62 (27%)]. Patients were referred if either they were unresponsive to AEDs or a clinical review indicated that the diagnosis might not be secure. All patients had a 12-lead ECG. Further cardiac investigations were undertaken as appropriate. In those patients who underwent a tilt test, a Finapres (Finapres Medical Systems BV, Paasheuvelweg 34a NL-1105 BJ Amsterdam ZO, The Netherlands) machine was used. Patients were tilted on a bed with footboard support for 45 min at a 608 angle, while monitoring their heart rate, blood pressure, and symptoms. No drug provocation was used. The tilt test was considered positive if hypotension and/or bradycardia were accompanied by reproduction of the patient's symptoms. Decisions about treatment were dependent on the treating cardiologist. Follow-up was in the cardiology and neurology departments at the Manchester Heart Centre, Manchester Royal Infirmary and the Greater Manchester Centre for Neurosciences, Hope Hospital, Salford, respectively.

Prospective group
The prospective group consisted of patients who were recruited into Reveal in the Investigation of Syncope and Epilepsy (REVISE), between 2007 and 2009 at the Manchester Heart Centre, Manchester Royal Infirmary, UK. The protocol was passed by the local ethics committee. All 41 patients had previously seen a neurologist, and they were reviewed by a neurologist with a special interest in epilepsy. Patients could be enrolled if he deemed that epilepsy had clearly been misdiagnosed, or that there was a significant doubt about a diagnosis of epilepsy, based on clinical features or a lack of response to AEDs.
Prospective patients were recruited from a total of 484 new patients with T-LOC seen during the period of prospective recruitment in an arrhythmia clinic, or a specialist nurse-lead Rapid Access Blackouts Triage Clinic. At enrolment, epilepsy was thought to be misdiagnosed in 28 patients, and there was a significant doubt about epilepsy in 13 patients. Most patients had received AEDs.
Other inclusion criteria for the prospective study were (i) three or more than three episodes of T-LOC in the last 12 months and a normal, equivocal or non-diagnostic 12-lead ECG, echocardiogram, 24 h ECG, standard unprovoked EEG and brain computed tomography (CT)/magnetic resonance (MR) scan. After recruitment, all patients underwent an ILR (Reveal Plus w /Reveal DX w ; Medtronic Inc., Minneapolis, USA) and tilt-table testing. Patients underwent follow-up every 3 months till at least 1 year after ILR implantation.
Tilt testing was undertaken using the Task Force w Monitor (APC Cardiovascular Ltd, Cheshire, UK), using the same protocol for tilt testing as in the retrospective group.
The protocol allowed for treatment to be given, if necessary, during the course of the study. Decisions about treatment were taken by the principle investigator. Patients who had symptomatic pauses of ≥3 s on the ILR or a decrease in heart rate below 40 beats/min for 30 s were offered a permanent pacemaker (PPM). If patients were on AEDs at the start of the study, they could continue, but AEDs could also be withdrawn or initiated during the study on the advice of the neurologist with a special interest in epilepsy.
At each follow-up visit, a record was kept of patient symptoms, including the number and date of episodes of T-LOC. All patients had their ILR downloaded, and ECG findings correlated with symptoms.

Statistical methods
Normalized data are given, with mean, standard deviations and ranges. Results are presented for all patients and separately for the prospective and retrospective groups. Student's t-test was used to compare percentages and means between the prospective and retrospective groups using the GraphPad Prism statistical package.

Results
Demographics (Table 1) The mean age of the whole cohort was 46.4 + 17.4 years (range: 18-80 years) with slightly more number of female patients (58 of 103; 56.3%). Patients in the prospective group were younger than the retrospective (40.2 + 16.2 vs. 50.5 + 17.0, P ¼ 0.0028). There was no difference in the age quartile or sex distribution.

Duration of symptoms
The mean duration of symptoms for the whole cohort was 126.8 + 131.6 months (range: 4-780 months) with no significant difference between the subgroups.

Origin of the initial diagnosis of epilepsy
In most cases (64 of 103, 62.1%), the diagnosis of epilepsy was made by a neurologist, otherwise it had been by a family doctor or general physician. In 17 patients the origin of the diagnosis was unknown [prospective vs. retrospective 15 of 41 (36.6%) vs.
Twelve different commonly used AEDs were prescribed, and most patients had been on more than one drug. More phenytoin had been prescribed in the retrospective group.

Transthoracic echocardiogram
Echocardiography was done in 76 patients (74%), including all prospective patients and half of the retrospective group. Few patients had structural heart disease and none had severe structural abnormalities.
Implantable electrocardiographic recorder data ( Table 2) Patients were followed up for a mean of 874 + 776 days after implantation of the ILR. Follow-up was longer in the retrospective group (1263 + 749 vs. 239 + 171, P ¼ 0.0093). The mean number of downloads from the ILR was known for the prospective subgroup were 2.24 + 1.88 (median: 2.00, range: 0-9).

and 2)
Sixty-seven percent (69 of 103) of patients had T-LOC symptom/ ECG correlation by ILR recording, at 35% of each subgroup. Sinus arrest was recorded in 13 of 103 (12.6%). The mean duration of asystole was 25.4 + 30.3 s (range: 4-89 s) for patients in the prospective group. Fifteen patients in the retrospective group had asystole . 6 s recorded. In 43 of 103 (41.8%) patients with T-LOC symptom/ECG correlation there was normal sinus rhythm. Muscle artefact suggestive of tonic-clonic seizures were recorded in a further four (3.9%) patients, while the underlying ECG appeared normal ( Figure 2). All of these patients were in the prospective group.

Further treatment
On basis of the results of the symptom/ECG correlation achieved by ILR, a pacemaker was offered to 22 of 103 (21.4%) patients and 21 (20.4%) consented (6 prospective, 15 retrospective). One prospective patient refused. Eleven (50%) of the patients offered pacemakers had confirmed epilepsy at baseline (prospective group: 6 of 18, 33.3%, retrospective group: 5 of 17, 29.4%). Seventeen patients (81%) who received a pacemaker were asymptomatic on follow-up (5 prospective, 12 retrospective). Four paced patients remained on AEDs. In 13 of 103 (12.6%) patients with a previous diagnosis of epilepsy, a pacemaker abolished symptoms. The mean duration of follow-up was 42 + 25.6 months (range: 15-91). All other patients in the prospective group continued to have episodes of T-LOC. This was despite treatment with AEDs in 7 of 41 (17.1%), midodrine in 6 and no active treatment was given in 15. Of the nine patients with a resting bradycardia on their initial ECG, only one patient had an asystolic pause on ILR and received a pacemaker.

Antiepileptic drugs on follow-up
In the prospective group 18 od 41 patients were on AEDs at follow-up which was similar to enrolment. Indeed, two of the prospective group started AEDs during the course of the study. Both of these patients had muscle artefact suggestive of tonic-clonic seizures recorded on the ILR at the time of symptoms. One of them has subsequently been asymptomatic for 24 months. In the retrospective group, AEDs were withdrawn in 60% of patients after pacemaker implantation.

Tilt testing
Tilt-table testing was undertaken in 81 of 103 (78.6%) patients, and was positive in a minority-14 of 103 (13.6%). Ten patients (10%) had a vasodepressor response and four patients had a cardioinhibitory response. The percentage of patients who had a positive test and the type of response was similar in the two subgroups.
Electrocardiography correlation between patients with a positive tilt test and implantable electrocardiographic recorder ( Table 3) In those with a positive tilt test, 12 of 14 (85.7%) patients had a positive symptom/ECG correlation by means of the ILR (prospective six of eight, retrospective four of four). Findings on ILR were sinus arrest: two of four (50%), sinus tachycardia: one of four (25%), and slow atrial fibrillation: one of four (25%). Among the four patients with a cardio-inhibitory positive tilt test, ECGsymptom correlation was achieved by ILR in two (50%) patients. One patient had sinus arrest and the sinus rhythm. The ECG features of a tilt test correlated poorly with findings of T-LOC symptom/ECG correlation by ILR.

Electroencephalogram
Seventy-three of the 103 patients had an EEG, which was abnormal in 18 (17.5%) patients. However, in all but one patient, nonspecific, non-diagnostic abnormalities were found. In the prospective group, 6   of 41 (7.3%) patients underwent EEG recording during video telemetry, but without any diagnostic findings.

Brain imaging
Brain imaging, by CT scan or MR imaging, was undertaken 82 of 103 (79.6%) patients. Ten scans were abnormal, but in the opinion of the neurologist none could explain the patients' symptoms.

Discussion
A likely diagnosis in T-LOC is made by clinical evaluation of symptoms, supplemented where possible by an eye-witness story. Laboratory investigations are usually done when the patient is well, and have to be interpreted in the light of the clinical presentation. Essential knowledge about any of heart rate and rhythm, blood pressure and brain activity during an attack have usually been unavailable until the advent of the ILR. Epilepsy is generally characterized by recurrent seizures with asynchronous neuronal discharge, preserved cerebral perfusion, and preserved heart rate and rhythm. Syncope is characterized by T-LOC where the cause is transient global inpairment of cerebral perfusion. Pathological changes in heart rate and rhythm may be the cause of failure to perfuse the brain, as may a pathological fall in blood pressure, or both. In psychogenic pseudosyncope, it is possible to provoke a blackout with tilt-table testing, but show that the ECG, blood pressure, and EEG remain apparently undisturbed. 10 It is now possible to record heart rate and rhythm during TLOC using an ILR. However, it is much more difficult to acquire information on the role of blood pressure and brain activity during spontaneous T-LOC. This is partly because of current limitations in implanted technology. Facilities for external monitoring of the ECG and EEG do exist in centres with videotelemetry, but typically in these settings blood pressure is not measured continuously. Also, videotelemetry facilities are very costly and thus have limited availability. 11 Clinical evaluation is therefore dependent on the history from the patient, who may recall little of an episode of T-LOC, supplemented by an eye-witness account if available. Laboratory physiological data are typically absent or largely incomplete, and it is not suprising that there may be a delay in diagnosis or a misdiagnosis. According to the All Party Parliamentary Group on Epilepsy, 14 there are 74 000 patients in the United Kingdom who are misdiagnosed and taking epilepsy drugs that they do not need. The direct cost of these misdiagnoses to the NHS £189m, 14 and place a strain on patients and on families. For the individual, there may be important implications for education, work, and childbearing. 15 Furthermore, if epilepsy is misdiagnosed, and the treatment is wrong, it raises the question, 'what is the diagnosis? ' We believed that convulsive syncope mimics an epileptic seizure in many cases. When the only clinical evidence available is a limited patient story supplemented by the eye-witness account of a relative or other lay person, 'having a fit' easily gets identified as epilepsy. We used ILRs to gather limited but reliable physiogical data at the time of a spontaneous event, capturing symptom/ ECG correlation. However, we had previously observed convulsive features accommpanying syncope with asystole recorded by ILR, and postulated that this would be a common phenomenon among the population with a misdiagnosis of epilepsy. This is the first study to systematically use the implantable ECG loop recorder (Reveal Plus w /Reveal DX w ) in the evaluation of patients suspected to be misdiagnosed with epilepsy or in those in whom the diagnosis of epilepsy was in doubt. We showed that one in eight (13 of 103, 12.6%) patients with syncope were misdiagnosed as epilepsy.
On the basis of treatment responses, the true diagnosis in these patients was reflex syncope. The convulsive movements noted during spontaneous episodes of T-LOC were manifestations of 'convulsive syncope'. These patients were asymptomatic after implantation of a PPM and withdrawal of AEDs. In some patients it was also possible to infer generalized epilepsy, 4 manifesting as tonic-clonic seizures, by the pattern of muscle artefacts found on the loop recorder during an episode of T-LOC in 4 of 103 (3.9%) patients.
The principle achievement of this study was to gather spontaneous physiological data, albeit limited to symptom/ECG correlation showing gradual slowing of sinus rate and increasing PR interval before atrioventricular block during marked sinus bradycardia, followed by resumption of sinus rhythm. This patient had a previous diagnosis of epilepsy, and was treated for various periods with sodium valproate, carbamazepine, clobazam, levetiracetam, and topiramate for 12 years prior to enrolment in the Reveal in the Investigation of Syncope and Epilepsy study. After permanent pacing she remains asymptomatic.
Prolonged implantable ECG monitoring in the REVISE study during a spontaneous convulsion, and then be able to diagnose cardioinhibitory reflex syncope. The diagnosis could then be changed, permanent pacing could be instituted, and AEDs stopped. The subsequent benefits of pacing were convincing. ECG/symptom correlation was achieved by ILR in 69 of 103 (67%) patients in this study. This rate of ECG/symptom correlation is higher than the 35% for patients presenting with unexplained syncope and 27% for patients with possible neurally mediated syncope, 16 but comparable to the 88% when used in highly selected patients. 16 In one other study, 17 the ILR been used to investigate patients with suspected misdiagnosis of epilepsy. However, in that study only10 of 74 (13.5%) patients had an ILR.
The second difference between this study and previous evidence is in the results of tilt testing. Zaidi et al. 9 used tilt testing to arrive at a diagnosis of syncope, being positive in 25.7% of patients. Tilt testing was positive in only 14 of 103 (13.6%) our patients. Moreover, there was poor correlation between the results of the tilt test and the ILR findings. None of the patients in the studies by Smith et al., 11 Josephson et al., 12 and Scheepers et al. 13 were tilted. Tilt testing relies on provocation to acquire evidence, rather than acquiring spontaneous data during T-LOC. There is a wide variation in the sensitivity and specificity of the tilt test, 1 probably accounting for the difference between this study and that of Zaidi et al. 9 Tilt testing is currently a class IIb indication for differentiating syncope from epilepsy in the latest European Society of Cardiology guidelines on syncope. 1 Also notable in this study is the apparent diagnosis of generalized epilepsy by the pattern of rhythmic skeletal myopotential artefacts noted in 4 of 103 (3.9%) patients during a spontaneous T-LOC ( Figure 2). It has also been shown to correlate with EEG findings 18 in a small study of 12 patients elsewhere. Two of the four (50%) patients with this pattern on the ILR were started on AEDs, and one of the two patients was asymptomatic after 18 months.
This study also provides some insight into the value of other tests in these patients. Seventy-three (70.9%) patients had an EEG. From the prospective group, 6 of 41(14.6%) patients   11 Scheepers et al. 13 also used the same tests in their study, although in unclear numbers. All patients in the study by Zaidi et al. 9 and 90% of patients in the study by Josephson et al. 12 had an EEG. However, like tilt table testing, the EEG has variable sensitivity and specificity for epilepsy, 4 and because of its limitations it is recognized that an EEG should not be used to make a diagnosis of epilepsy. It may be useful in under 35s to support a diagnosis and to help define an epilepsy syndrome where epilepsy is likely. 4 An EEG should be avoided in suspected syncope because of the possibility of false-positive results. 4 Patients in this study were considered to have a misdiagnosis based on a clinical review by a neurologist with special interest in epilepsy. Where an EEG was done there were no categorical features of epilepsy, so the results of the EEG were not used in the clinical evaluation. Misinterpretation of EEG findings is known to be a common reason for a misdiagnosis epilepsy. 7 MR scanning is an investigation of choice in patients with likely epilepsy. 4 The value of brain imaging in these patients is unclear from existing evidence. It was undertaken only in 28 of 186 (15.1%) patients in the study by Smith et al., 7 but the type of imaging done is unclear. MR scanning was undertaken in an uncertain number of patients in the study by Scheepers et al. 13 Josephson et al. 12 used CT in 51% of patients and MR brain in only 15% of patients in their study. No imaging was undertaken by Zaidi et al. 9 In this study 82 of 103 (79.6%) patients underwent brain imaging, by CT or MR scanning, but no diagnostic findings emerged. Seventy-five percent of patients also underwent external ambulatory ECG monitoring and an echocardiogram. These tests have not been used in any other studies, 9,11 -13 but the results here were unhelpful.

Limitations of the study
The whole cohort contains patients studied retrospectively and prospectively, and this may have a bearing on the results, although the two groups were very similar in baseline characteristics, clinical features, and investigations. Patients with a true diagnosis of epilepsy are reported to have asystolic pauses during an episode of spontaneous T-LOC. 17,18 It is possible that one patient in our study who did not respond to permanent pacing fell into this category. However, she continues to be symptomatic in spite of restarting AEDs and is now being considered for a vagal nerve stimulator as treatment for resistant epilepsy. 19 The possibility of a psychogenic cause of T-LOC has not been investigated in this study. It is possible that some patients with normal sinus rhythm on their ILR during an episode of T-LOC suffer from psychogenic pseudosyncope or non-epileptic attack disorder. This could still represent a misdiagnosis, and demonstrates the limitations of currently only being able to monitor the ECG in T-LOC patients. Elsewhere, tilt-testing during monitoring of ECG, blood pressure, and EEG has been used to provoke psychogenic pseudosyncope, 10 and more physiological data in long-term implanted monitors are needed.

Conclusions
This study used ILRs to show a high incidence of bradyarrhythmias and asystole in patients previously diagnosed with epilepsy where the diagnosis of epilepsy was in doubt. The most likely diagnosis in these patients was convulsive cardio inhibitory reflex syncope. Poor correlation between the results of tilt testing and ILR were found. In a small number of patients rhythmic skeletal myopotential artefacts on the ECG suggested a diagnosis of tonic-clonic epileptic seizures. An ILR represents the only readily available tool for the investigation of patients whose seizures may accompany T-LOC caused by convulsive cardio inhibitory reflex syncope or generalised epilepsy, and the diagnosis is in doubt. Given the high incidence of the misdiagnosis of epilepsy, and the profound effects that the diagnosis has on patients and their families, an ILR should be considered whenever a diagnosis of epilepsy is uncertain but being considered.