What we did and what we found
Given the preliminary evidence that DS-CBT was helpful for DS, we were keen to ascertain whether potential treatment effects were generalisable, through a larger multicentre study where therapists varied in their knowledge and experience of DS.
Here we provide a synthesis of findings from the main trial and secondary analyses10–14 and put the trial in context, highlighting factors that might be relevant for future studies. There is a summary of the key findings from our studies in figure 2.
Figure 2Summary of key data from the CODES trial. This figure includes still images taken from an animated video created by Science Animated (sciani.com). The originals are available at www.codestrial.org and https://www.youtube.com/watch?v=pUFKbYH7BcQ. CBT, cognitive behavioural therapy; CGI, Clinical Global Impression of Change; CODES, COgnitive behavioural therapy versus standardised medical care for adults with Dissociative non-Epileptic Seizures; CORE-10, Clinical Outcomes in Routine Evaluation 10; EQ-5D-5L, European Quality of Life 5 Dimensions 5 Levels; GAD-7, 7-item Generalised Anxiety Disorder Assessment; NICE, National Institute for Health and Care Excellence; PHQ, Patient Health Questionnaire; QALY, quality adjusted life year; SF-12v2 MCS, 12-item Short Form Survey version 2 Mental Component Summary score; SF-12v2 PCS, 12-item Short Form Survey version 2 Physical Component Summary score; SMC, standardised medical care; WSAS, Work and Social Adjustment Scale.
The trial compared two treatment arms, DS-CBT plus SMC (n=186) versus SMC alone (n=182). Randomised patients were initially recruited from neurology/specialist epilepsy services in the National Health Service where treatment is free at the point of receipt. Importantly, rather than the equivalent of routine or standard care, SMC, as delivered in CODES, was ‘standardised and specialist’ care10 11 because we provided clinicians with guidelines regarding diagnosis and management (see online supplemental information). We also produced booklets specifically written for patients (see https://www.codestrial.org/INFORMATIONBOOKLETS). In addition to delivering the diagnosis in a recommended manner, neurologists and liaison/neuropsychiatrists saw patients for follow-up appointments to offer support and to prescribe pharmacotherapy when appropriate. Psychiatrists were asked not to include CBT techniques in their management.
DS-CBT, delivered by therapists (who had study-specific CBT training) was manualised (materials available from codestrial@kcl.ac.uk) but, importantly, allowed flexibility for the intervention to be formulation-based, that is, tailored to the individual. Therapist adherence to the treatment approach and therapeutic alliance were measured.
Main trial findings
Our prespecified primary outcome9 15 (monthly DS frequency at 12 months post randomisation) did not differ significantly between the two treatment arms, despite a 22% advantage in DS reduction in the DS-CBT+SMC group.10 11
There was a statistically significant advantage in the DS-CBT+SMC group in 9 of our 16 prespecified secondary outcomes, namely: more consecutive DS-free days in the past 6 months; less impact of DS on work, social and leisure activities as measured by the Work and Social Adjustment Scale (WSAS)16; better self-rated and clinician-rated improvement; greater satisfaction with treatment; finding DS less bothersome17; less distress rated on the Clinical Outcomes in Routine Evaluation-10 (CORE-10)18; lower symptom burden (modified Patient Health Questionnaire-15; PHQ-15)19 and better health rating on the European Quality of Life 5 Dimensions 5 Level (EQ-5D-5L) Visual Analogue Scale.20 See figure 2. No findings suggested better outcomes in the SMC-alone group. There was no evidence that any adverse or serious adverse events were related to the intervention.
The intervention was not found to be cost-effective considering the £20 000 to £30 000 cost per quality adjusted life years threshold recommended by the National Institute for Health and Care Excellence.10 The findings for cost-effectiveness were marginally better when estimated using the utility score (Short-Form 6-Dimension; SF-6D) derived from the 12-item Short Form Survey version 2 (SF-12v2)21 than from the more typically used EQ-5D-5L. It is possible however that our negative findings with respect to cost-effectiveness reflect the high levels of comorbidity10 22 and the very wide range of service use in our sample. In addition, we are not sure whether the EQ-5D-5L is a sufficiently sensitive measure for use in people with DS. Furthermore, as previously indicated,10 the time span may have been too limited to fully assess cost-effectiveness. Potentially, the effects of CBT may continue in the longer-term. We were limited to a 1-year follow-up for the trial and we did not undertake health economic modelling work to estimate the effects beyond this period.
Secondary analyses
Six-month outcomes
Our prespecified9 15 12-month follow-up evaluation did not include a formal analysis of outcomes at the end of treatment which broadly coincided with the 6-month post-randomisation measurement point. However, as the clarification of end-of-treatment benefits might be helpful for planning service provision, we conducted an additional secondary analysis which formally compared the groups at 6 months post randomisation. We found statistically significant differences for 13 outcomes, all in the same direction of benefit, that is, that DS-CBT+SMC was superior to SMC (figure 2). This included monthly DS frequency, that is, lower DS frequency in the DS-CBT+SMC group. Outcomes showing greater benefit in the DS-CBT+SMC group at 6 months that had not been significantly different at 12 months were self-rated seizure severity, >50% reduction in seizure frequency relative to baseline, the Mental Component Summary Score from the SF-12v2, anxiety23 and depression.24 All variables that had shown superiority in the intervention arm at 12 months showed a similar pattern at 6 months, but with larger effect sizes.
Visual and tabulated representations of the data10–12 suggested that rather than the benefit from DS-CBT+SMC diminishing between 6 and 12 months post randomisation, the more modest between-group results at 12 months largely resulted from improvements in the SMC-alone group across the follow-up period, accounting for the apparently smaller effect sizes at 12-month than at 6-month follow-ups. We speculate that this derived from support provided during SMC sessions which were spread across the 12-month follow-up period, and potential exposure to self-help websites and thereby seizure control strategies in the SMC group. Although both groups were exposed to SMC which included a discussion, but not the practice of distraction techniques, the DS-CBT intervention directly included DS-control strategies and may therefore have led to faster adoption of such techniques by participants than might have occurred in the SMC-alone group. In the absence of a completely inactive control group, we could not be sure, however, that this pattern of change in the SMC-alone group did not simply represent the natural history of DS and/or regression to the mean. Overall, inspection of our data does not support others’ suggestions that our treatment did not sustainably reduce DS frequency.25 It is possible though that extending therapy time beyond 12 sessions may have been additionally beneficial in continuing to assist participants practice and apply the DS-control techniques, as well as other therapeutic strategies, particularly given the complexity of the participants.10 22
This secondary analysis showed that it is important to undertake both short and longer-term follow-ups in treatment trials. Short-term outcome evaluations alone (eg,26) may give a distorted impression of the benefit or otherwise of the target intervention. It has been suggested27 that our cost-effectiveness analysis might have yielded better results if conducted at 6 months rather than 12 months post randomisation, since seizure frequency differed between groups at 6 months, but we did not investigate this possibility.
Treatment moderation
One issue that we did not address in our main outcome accounts10 11 was who might particularly benefit from DS-CBT, or what might predict treatment outcome. We explored this through a further secondary analysis to examine moderators of treatment effects, that is, which patient baseline characteristics might interact with CBT to influence outcomes. Putative moderators are baseline characteristics thought to be associated with more improvement in the DS-CBT+SMC arm.13 Where no moderators were found, baseline predictors of outcome across both interventions were investigated.13 Here we chose four outcomes, all at 12 months post randomisation: monthly DS frequency (our main primary outcome); the functional impact of DS as measured by the WSAS; and two quality of life measures, the Physical Component Summary score (PCS) and Mental Component Summary score (MCS), both from the SF-12v2. We prespecified putative moderators based on the literature.
Importantly, we found that DS-CBT was more effective than SMC-alone in reducing DS frequency for participants with at least one comorbid diagnosis on the Mini-International Neuropsychiatric Interview (M.I.N.I.)28 or having a score of at least 22/30 on the modified PHQ-15 (figure 2). For those with no comorbid M.I.N.I diagnoses or fewer symptoms on the modified PHQ-15, DS-CBT did not lead to a significantly greater reduction in DS frequency than SMC-alone. Our fear-avoidance-based DS-CBT approach might have specifically benefitted more complex patients because key intervention change techniques are likely to focus on aspects of comorbidity.13 Thus, for example, given the relatively high prevalence of agoraphobia (45%), major depressive disorder (31%), generalised anxiety disorder (29%), post-traumatic stress disorder (23%) and social anxiety disorder (20%) in our sample at baseline10 22 our focus on dealing with avoidance behaviour, anxiety, depression and to a more limited extent trauma may have been most useful for participants reporting these areas of difficulty.
Several baseline characteristics predicted outcome as measured by the WSAS, SF-12v2 PCS and SF-12v2 MCS across both intervention arms. Of note, predictors with larger effect sizes tended to be those related to socioeconomic characteristics such as being employed or in education, not being in receipt of state financial benefits and higher educational attainment. Some DS-related characteristics such as age at onset or duration of DS disorder, and anxiety and depression, were also predictive of outcome across both intervention arms. DS semiology did not predict outcome. Strength of belief in diagnosis was predictive of improvement across both arms on SF-12v2 MCS scores.
The suggestion that more complex patients might particularly benefit from CBT in terms of DS reduction when the comparison was SMC-alone raises the possibility that SMC, which was effectively enhanced treatment as usual, might be appropriate for less complex patients. However, services contemplating only offering DS-CBT to more complex patients would have to be able to deliver SMC to less complex patients, using the same pathway as evaluated in the CODES trial.
It has been speculated that more positive trial effects might not have been found due to high rates of maladaptive personality markers in participants (that is, scores on the Standardised Assessment of Personality Abbreviated Scale, Self-Report scale; SAPAS-SR29) and whether patients with this profile might have fared better with a more psychodynamically-oriented intervention.2 However, in our secondary analysis, our measure of maladaptive personality traits (SAPAS-SR) was found neither to moderate the treatment effect nor to predict treatment outcome. We acknowledge that a clinical assessment of personality disorder might have been more informative but this was outside the scope of the assessments we could undertake within the study. The heavy measurement burden for participants also meant that we did not include measures of dissociation or specific measures of trauma history that might also have helped clarify the patients for whom DS-CBT was particularly effective; this should be considered in future studies.
Mediation of treatment effects
Our final secondary analysis evaluated putative treatment mechanisms by examining putative mediators of the effect of DS-CBT+SMC on outcomes14 finding that DS-CBT effects on 12-month seizure frequency, functional impact and quality of life measures were mediated by changes in avoidance, anxiety and depression at 6 months (figure 2). DS-CBT did not bring about an improvement in another putative mediator, beliefs about emotions (as measured by the Beliefs about Emotions Scale; BES),30 although improvements in BES scores were associated with improvement in some outcomes. Estimates of mediated (indirect) effects tended to be small, with the largest mediated effects for avoidance and the WSAS.
The finding that avoidance behaviour mediated change in several outcomes supported our theoretical fear-avoidance model on which DS-CBT was partly based,3 8 9 and indicated that therapists were for, the most part, addressing avoidance behaviour during therapy. However, our analyses indicated that the mediated effects were relatively small; perhaps targeting beliefs about emotions and emotional avoidance more directly may be helpful in influencing outcomes. Of course, unmeasured mediators related to the therapeutic techniques used and to the therapeutic relationship may also have revealed more about the mechanisms of the intervention.
Choice of outcome measures
We chose monthly DS frequency as our trial primary outcome partly because the funder had identified seizure frequency as an important outcome (https://njl-admin.nihr.ac.uk/document/download/2023204) and partly because we had collected previous data on this measure,5 enabling data-based power analysis. It is also the symptom with which patients present, and for which they seek diagnostic evaluation. It is worth noting, however, that recording seizure diary data over a protracted period is challenging for patients. It can lead to variable amounts of missing data and measurement error, which requires careful consideration of how the outcome should be constructed from the weekly counts available.10 11 Moreover, DS frequency was a highly skewed outcome. Although we strictly complied with our analysis plan and were fully transparent about all statistical methods applied, including imputation steps, we therefore acknowledge that other solutions were available. Similar issues with recording seizure frequency in epilepsy have also been highlighted, suggesting it is not an optimal primary endpoint in a heterogeneous population. Other options might have been to adopt seizure freedom or the longest number of consecutive DS-free days as primary outcomes but we opted to include these as secondary outcomes, with data on these measures being reported in our main publications.10 11
Others31 have questioned whether another measure such as productivity might be more informative, and a recent meta-analysis has suggested that non-seizure outcomes are responsive to psychological treatment in patients with DS.32 Given that CODES secondary outcomes showed between-group differences in 9 of 16 secondary outcome measures10 11 it has also been questioned whether our primary outcome measure at 12 months post randomisation best represented the effectiveness of DS-CBT.33 Discussion of outcome measurement in FND more widely34 has advocated the use of core and supplementary measures. Several of these were incorporated in our selection, including the core symptom outcome measure recently recommended,34 that is, patient and clinician-rated global impression of improvement, based on the Clinical Global Impression of Change (CGI) measure.35 We found a between-groups difference in how bothersome17 participants found their seizures, even though there was no difference in seizure severity at 12 months post randomisation. This lends support to the future adoption of a self-rated CGI, or ‘seizure bothersomeness’/ ‘seizure severity’ rating, which is arguably a more specific form of self-rated seizure outcome, as a primary outcome.
Our qualitative work with trial participants36 supports the suggestion that changes in perceptions of the debilitating effects of DS may be much more important than the exact number of seizures participants have. A seizure in the middle of a busy workplace leading to an ambulance being called is very different from one occurring in bed just before sleep when the individual has found they can delay seizures using seizure control techniques. Therapists also acknowledged the value of patients being able to stall seizures to permit engagement with important activities.37
Finally, therapists beyond our study have expressed uncertainty over whether they can or should target DS symptoms directly rather than addressing quality of life more broadly.38 The CODES trial showed that specific seizure control techniques can be incorporated into a psychological therapy, analogous to panic control techniques. So even if a trial (or a clinical service) focuses on quality of life as a primary outcome, measuring seizure occurrence or impact in some manner appears relevant.
Choice of comparator group
The comparator condition in a trial can have a dramatic effect on the apparent effectiveness of the intervention. Participants in the SMC arm of the CODES trial did not receive ‘treatment as usual’ which in many parts of the UK and further afield would often mean ‘no treatment’ or attempted treatment from a mental health professional without experience of DS.
All participants experienced a neurologist/epilepsy specialist, trained in DS communication, explaining and providing well-presented written material about the diagnosis. Participants’ mean ‘agreement with the diagnosis’ after this process was 8 on Likert scale of 0–10 where 10 was complete agreement.10 So even before randomisation, most participants were perhaps more ready for treatment and understood the role of psychological therapy than a patient not having been given a clear diagnosis or written material. This may, therefore, have better prepared those patients subsequently receiving DS-CBT. However, the wider benefit of such an approach has been shown by a small study in Australia which highlighted how providing even these basic components to individuals with DS can be associated with much lower readmission rates and health economic costs.39
The standardised treatment in CODES consisted of seeing a liaison psychiatrist or neuropsychiatrist, experienced in DS, for up to a year. Although psychiatrists were asked not to include elements of the intervention, this is not a typical ‘treatment as usual’ scenario, and the comparator treatment is arguably one that could itself have been the focus of an intervention study. The fact that the DS-CBT plus SMC arm showed differences in so many clinically relevant secondary outcome measures in this context is arguably much more impressive than if the comparator arm had been ‘treatment as usual’. However, it is also necessary to acknowledge that our positive outcomes might in some, unspecified way, reflect synergy between DS-CBT and SMC and we cannot determine whether DS-CBT might have been less effective without adjunctive SMC. Future studies may wish to examine in greater detail the effectiveness of our model of SMC as compared with either a waiting list control or treatment as usual. The effect of spacing/frequency of SMC sessions could also shed light on its effectiveness.
Could the comparison group have been given a different psychological intervention? Even more psychodynamically-oriented treatments40 may contain some CBT-based elements targeting seizure control. It seems likely, based on other data, that the sample size needed to demonstrate a difference between two active psychotherapies would be so large as to not be feasible for this kind of design. However, a more inactive comparison such as progressive muscular relaxation could control for therapist time and attention and may be worth considering.
Context for implementation
In the CODES trial, we created a clinical pathway that included neurologists/epilepsy specialists, liaison or neuropsychiatrists and CBT therapists. This pathway was embedded in a UK health service with variable service provision for patients with DS and limited opportunities for neurologists to refer patients for psychotherapy.41 Funding regulations meant we could not directly pay clinicians for their time in the study with research funds, so we had to establish clinical pathways relying on the enthusiasm of interested services and clinicians. It is likely that SMC as provided in CODES was shaped by this enthusiasm. We did not specifically explore the moderating effect of site on outcomes but the effects of the randomisation stratifier 'site' were allowed for in all our analyses.10–14 Future studies might wish to consider systematically auditing the content of SMC for fidelity, rather than simply providing guidance to clinicians as to its content in the manner we employed here.
People attending their first psychiatric appointment were no different from those who did not, in terms of most baseline characteristics, including confidence in the diagnosis,42 suggesting that neurologists should be agnostic about who they refer for further treatment. Psychiatrists, who assessed patients 3 months after the neurologist, felt that patients they saw in the trial were generally more accepting of their diagnosis than other patients they saw, possibly as a result of the way neurologists had explained the diagnosis as well as the support materials made available to the neurologists and the patients.43 Therapists generally considered that those patients receiving DS-CBT in the trial seemed to present with a better understanding of their diagnosis than those seen in other contexts.10 37
We acknowledge that the CODES pathway infrastructure has not been maintained across all participating services although, following study completion, a survey of participating neurologists showed the majority were keen for the pathway to continue.41 Those therapists wanting to implement our DS-CBT approach should attempt to do so in association with input from neurologists and psychiatrists with expertise in this disorder. As noted previously the multidisciplinary CODES pathway10 may limit the extent to which patients feel ‘abandoned’ by medical staff following diagnosis and may prepare patients better for psychotherapy, as well as facilitate more cohesive interdisciplinary working.10 A consistent message from the multidisciplinary team is likely to be more powerful.
Our assessment of treatment fidelity10 involved the independent rating of pseudorandomly-selected recorded DS-CBT sessions, conducted by our therapists who had undergone study-specific training. The therapists were also allocated to supervision groups led by therapists experienced in delivering our treatment model. Our ratings suggested that DS-CBT+SMC can be implemented in a range of services. However, adhering to a DS-CBT treatment protocol, even if it is formulation-based and tailored to the individual, may be challenging for therapists who would otherwise adopt a more eclectic psychotherapeutic approach.37
A further consideration regarding implementation, especially in less specialised services is the provision of education about FND in general and DS in particular for staff. Some trial psychiatrists43 and therapists37 commented on the complexity of the patient group and the potential for damage to a patient’s progress by incorrect references to the patient possibly having epilepsy or needing anti-seizure medications.43 Both groups suggested that more experienced staff would be best placed to treat patients with DS. This means that services wishing to implement CODES-informed treatment pathways will not only need to consider the expertise of those delivering therapy but also of those providing supervision.
We also acknowledge the lack of diversity in our participant sample, with approximately 90% of our randomised patients self-identifying as white. While this may have derived from our eligibility criteria, whereby we excluded people who would not have been sufficiently fluent in English to complete measures and therapy without an interpreter, we cannot be sure about the extent to which our findings would apply to a broader demographic and the cultural competencies of therapists would also require consideration. Similarly, we cannot conclude that our findings would be generalisable to individuals with an intellectual disability or concurrent active epilepsy since such people were not included in the study. Nonetheless, as we have indicated elsewhere22 in addition to a typical proportion of women to men for a DS sample, our sample was broadly generalisable in terms of duration of symptoms, mean number and profile of comorbid psychiatric diagnoses (although people with the most severe psychopathology requiring crisis intervention were excluded from randomisation), at least moderately severe functional impairment and appreciable levels of avoidance behaviour and a range of symptoms on the modified PHQ-15 that were generally similar to those documented for other neurology outpatients with symptoms not explained by disease.
