Breathing control training for functional seizures: A multi-site, open-label pilot study

There are no


Introduction
Functional seizures (FS), also known as psychogenic non-epileptic seizures (PNES), are paroxysmal events that clinically resemble epileptic seizures but are not accompanied by changes on electroencephalogram, and which have a presumed psychological cause.FS have an annual incidence of around 3/100,000, and prevalence of approximately 24/100,000 [9].They make up approximately 20% of those who present to services for epilepsy, and commonly present emergency departments [6].The prognosis is generally poor [6], with significantly increased mortality, though deaths do not usually appear directly attributable to the seizures themselves [5].There is no universally accepted therapy for FS.Psychotherapy is most commonly used, however its benefits are moderate, with most people still experiencing seizures at follow-up [1].
Symptoms suggesting hyperventilation, such as dizziness, breathlessness, and paresthesiae, are very common in FS [2].Many patients are witnessed to hyperventilate before and during their FS attacks, and hyperventilation can be successfully used to induce attacks for diagnostic purposes [2].It seems possible, therefore, that disordered breathing plays a role in the pathogenesis of FS attacks, at least in some patients [3].If so, then improving breathing control may be a potential therapeutic avenue in FS.Breathing Control Training (BCT) involves teaching reducing overall ventilation and establishing a nasal/diaphragmatic breathing pattern [7], and is widely used to treat hyperventilation in patients with asthma and other respiratory disorders [7].We hypothesised that breathing control training would reduce seizure frequency in FS.

Methods
We carried out an observational, uncontrolled pilot study of the potential therapeutic effect, acceptability and safety of BCT in FS.Patients were recruited from Christchurch Hospital in New Zealand and from the Austin and Alfred hospitals in Australia.The two cohorts were recruited prospectively, and independently, at different times, the first participant in March 2014 and the last in July 2020.Inclusion criteria were of age 16 or over, a diagnosis of clinically established FS [4] based on clinical history and video EEG recording of typical events, and with reported seizure frequency of at least four per month.Exclusions were made of those with epilepsy, or clinically significant asthma/other respiratory disorder, or those unable to understand English or to give informed consent.The intervention (i.e.BCT) was delivered by respiratory physiotherapists specifically trained in the intervention [7], over one hour, with a half-hour 'booster' session a month later.The intervention focused on establishing a nose/stomach breathing pattern, and reducing overall ventilation.The slow breathing techniques were individualised, with participants encouraged to practice breathing at a rate lower than their usual resting respiratory rate.Participants were asked to practice breathing re-training 2 x 10mins daily, and use slow breathing as a rescue technique if they began to hyperventilate.
At baseline, and at 3-4 months, we collected the number of seizures over the previous month recorded by patients in daily seizure diaries, supported by retrospective report where diaries were not available, as well as the Nijmegen scale, measures of anxiety (The Beck Anxiety Inventory, BAI) and self-efficacy (The General Self-Efficacy Score, GSES).Outcome variables were collected at three months from baseline in Melbourne and at 4 months in Christchurch: since there was variability of 3 weeks in appointment date within the two groups it was considered acceptable to consolidate 3 and 4 month outcomes into one variable.The Nijmegen scale is a 16 item, self-report questionnaire which explores respiratory and non-respiratory symptoms associated with hyperventilation, and acts as a screening tool to detect hyperventilation.Each item is answered by the participant on a five-point Likert scale ranging from 'never' (equating to zero) to 'very often' (equating to four points).Total scores range from 0 to 64, with a score of over 19 indicating the presence of respiratory distress and dysfunction [8].We summarised data for all recruited participants, with statistical analysis of outcomes performed on those with available data at both timepoints.Non-parametric testing (i.e., Wilcoxon signed rank test) was used due to the small sample size.Analysis was conducted using SPSS version 29 (www.ibm.com/spss) with figures produced in R version 4.3.1.A goal of 20 participants was set as a pragmatic sample size.The study was approved by the Austin Health and the NZ Health and Disability Ethics Committees.

Results
We recruited a total of 18 patients, eight from New Zealand and 10 from Australia.Though no formal sample size was pre-determined, the study finished earlier than expected due to challenges with continuing during the COVID-19 pandemic.Eleven patients were female, seven were male.Fifteen identified as of European ancestry, two Maori and one Asian.Their median age was 36 (range 18-56), with median age at FS onset of 24 (range 15-53).Duration of FS was median 4 years, interquartile range (IQR) 1-8.5.Median seizure frequency per month was 45, IQR 10-60.We had baseline Nijmegen scores on 14 patients, median score 29, IQR 20.5 -35.5.We had baseline GSES and BAI on 8 patients, median scores 43 (range 21-50) and 26.5 (range 7-53), respectively.There were no differences in baseline demographic or clinical variables between those recruited from Australia and New Zealand.
At 3-4 months follow-up, seizure outcome data were available for 10 patients and Nijmegen scores for 8 patients.There were no differences in baseline variables for those who did and did not complete follow-up.Median seizure frequency at follow-up was 0.75 seizures per month, IQR 0.12-7.Median Nijmegen score was 19.5, IQR 17.75-22.Differences between baseline and outcome data were tested using the Wilcoxon signed rank test and suggested a trend toward improvement (p = 0.092, seizure frequency, and p = 0.090, Nijmegen Score).Seizure frequency was significantly correlated with Nijmegen score at follow up (Spearman's rank r = 0.75, p = 0.034).In terms of individual outcomes among those who completed follow up, seven patients had clearly improved seizure frequencies (i.e., lower frequency at follow up [median 0.5] compared to baseline [median 60]), of which three achieved seizure freedom and three had one or fewer seizures per month, while three had worse seizure frequencies (median frequency 14 at baseline and 74 at follow-up) (see Fig. 1).In six patients, the Nijmegen score improved (i.e., lower score at follow-up [median 18.5] compared to baseline [median 29.5]), in one it was unchanged, and in one it was worse (22 at follow-up and 11 at baseline).
Of the eight patients who did not participate in follow-up, there is clinical evidence for three, suggesting in each case that they had improved -two informed the physiotherapist that their seizures had stopped but declined further help, and one discontinued their (previously frequent) presentations to the emergency department -but they have not been included in our analysis.
No patient reported difficulty tolerating BCT or adverse effects.All patients who attended follow up attended both the intervention and 'booster' BCT sessions.One subject lost to follow-up was subsequently found to have died, but the cause of death was considered unrelated to the intervention.

Discussion
In carrying out this pilot study, we sought to gain a preliminary indication of the potential therapeutic effect, acceptability and safety of BCT in the treatment of people with FS.Recruitment was slow and follow-up challenging: three patients did not return either outcome variable at baseline, and 7 patients did not complete them at follow up.Of the patients who completed the study, 7 out of 10 had improved seizure frequencies at 3-4 months and 6 out of 8 had improved Nijmegen score.
No patient reported difficulty tolerating BCT or adverse effects.This tends to support our belief that BCT would be as well tolerated in the FS population as it is in the respiratory population.In terms of safety, a participant death is notable in a small study, even if not thought to be related to the intervention, but appears to fit with the increase in mortality observed among patients with FS [5].
Recruitment and compliance were challenging.While this is perhaps explicable in a study run by busy clinicians, without dedicated research assistants or specific compliance measures, it has two implications.Firstly, it means that in a definitive randomised controlled trial we might have to plan to recruit more patients in order for it to be sufficiently powered.Secondly, patients who do not attend study visits self-select, involving a significant likelihood of an effect of selection bias (though we noted evidence from 3 subjects that this was not the case, above).
There are a number of limitations to the study.We assessed at a single follow-up point, so that longer term effect is unknown.The measurement of seizures relied largely on retrospective reports, which reduces reliability, and did not assess the extent to which participants actually practised their techniques.The study was small, with a significant risk of selection bias as noted above.Well-motivated, engaged patients may preferentially have been recruited and completed the study, with the study investigators perhaps preferentially recruiting those they sensed would do well, because of their clinically observed panic or anxiety.Our sample was relatively young, which may be associated with a tendency to anxiety and to hyperventilate during seizures [3].Being uncontrolled and unblinded, the study could not in any event demonstrate efficacy of BCT in FS.Nonetheless, median seizure frequency and median Nijmegen score improved between baseline and follow-up visits.Most of the subjects who completed follow-up achieved remission from seizures, which compares well with other interventions.We are, however, not able to comment on whether, if real, the improvement was due to a therapeutic effect of BCT, a placebo effect, or to regression to the mean.That would require a randomised controlled trial, for which we believe these data offer support.A definitive RCT would need to address these limitations, with a control arm which provided the non-specific effects of care and attention, longer follow-up, measures of practice, and ideally a mechanistic biomarker, so that any benefit can be attributed to the ventilation changes, rather than, for example, the change in breathing attention which the practice may also induce.
BCT is relatively inexpensive, simple and easily-accepted intervention, in particular one which does not require participants to accept a psychological role in their illness, which can be a barrier for some [1].Even if only half of people with FS benefitperhaps those whose seizures can be induced by hyperventilation, or who are observed to hyperventilate before their seizures [3]-that would still be invaluable for the treatment of a condition for which there are currently few practical options.

Fig. 1 .
Fig. 1.Change in seizure frequency (left panel) and hyperventilation (right panel) from baseline to follow-up.Numbers on the panel are individual scores; points without a connecting line are those who did not complete follow-up.