Principal outcomes

We evaluated all outcomes at 12 months post randomisation, in line with our published SAP.⁹² In terms of clinical outcomes, all estimated trial arm differences were in favour of CBT + SMC. For our primary outcome, both trial arms appeared to show decreased DS frequency over the course of the trial, but the between-group difference was not statistically significant; inferential statistics nonetheless suggested a 22% advantage for the CBT + SMC arm. For continuous secondary outcomes, this pattern of results favouring the CBT + SMC arm is illustrated in Figure 8. Although 9 out of 16 secondary outcomes were better in the CBT + SMC arm at a p-value < 0.05, five of these reached significance at the more conservative level of a p-value ≤ 0.001. This included one seizure-related outcome (longest period of DS freedom in months 7–12 of the study) as well as clinically important aspects of psychosocial function, global outcome and satisfaction with treatment. Compliance with CBT sessions by those allocated to receive them was good and similar numbers of SMC sessions were offered to the two trial arms. There was no evidence that the CBT + SMC arm experienced greater levels of harm than the SMC-alone arm during the trial, suggesting that our approaches can be offered safely within the NHS.

Health service use decreased in both trial arms over the course of the trial, but there was no overall comparative benefit from CBT + SMC; owing to the therapy cost, the total NHS costs were higher in the CBT + SMC arm than in the SMC-alone arm, and the cost of the therapy did not offset other changes in service use. The addition of CBT to SMC did lead to increased QALYs, but the difference between arms was small. The ICER for CBT + SMC compared with SMC alone was £120,658, and was £116,815 when the SF-6D was used. We estimated that the cost of a reduction in monthly seizure frequency was £611 per seizure. Whether or not this represents ‘value for money’ is unclear. We would need to have an estimate from participants themselves concerning the value of a reduction in one seizure per month to understand its worth and the possible benefit of such a reduction on psychological and psychosocial functioning. From a societal perspective, we noted the often high level of informal care reported by participants at baseline. Within the timescale of this study, it may have been unrealistic to assume that this would reduce significantly given the potential range of other comorbidities in this patient group and family support systems that may have developed.

In addition to the self-report health and informal care use measures, we obtained objective measures of health service use (A&E, outpatient and inpatient use) and found a generally similar pattern of results to that shown by the CSRI for A&E and outpatient use. There were slightly higher costs estimated for inpatient stays based on the CSRI than on HES data. To our knowledge, this is the first time that HES data have been analysed for patients with DSs. Although the numbers of patients contributing to the results from the CSRI and HES data differed slightly, the current data suggest that DS patients are able to provide broadly accurate accounts of their hospital health service use. It is possible, however, that the length of hospital stays was slightly over-reported. It is not possible in any case to be sure that HES data are themselves entirely accurate given that their completion depends on the diligence of the many different people who complete the records. We suggest, therefore, that future studies rely mostly on self-report measures of health service use unless very fine-grained analyses are required. Certainly, HES data do not permit the estimation of medication costs, and these were of relevance in this sample.

The current study has yielded data showing areas of differential clinical effectiveness in favour of CBT + SMC when considering secondary outcomes. We cannot determine whether or not the current high cost-effectiveness ratio of £120,658 per QALY is due to the relatively short follow-up period. Both costs and QALYs were measured for the whole follow-up period, but we do not know what longer-term impact of CBT there may be. If gains were prolonged, this would reduce the cost-effectiveness ratio. However, this assumes that no ‘top-ups’ are provided. For both trial arms, patients may have received SMC sessions as a function of what was recommended as part of the trial rather than what may have occurred in everyday clinical practice. Taking into consideration scores on the modified PHQ-15 and the number of SAEs (of which relatively few were related to DSs), it is clear that there was a wide range of reported comorbid medical disorders, many requiring medical service input, and it is perhaps unlikely that our CBT intervention would have reduced these. Patients were also characterised at baseline by high levels of psychiatric comorbidity on the M.I.N.I. (see Table 9); therefore, it is possible that at follow-up the failure to show reduced costs may be a function of this comorbidity having been identified during the trial, with participants being directed for further treatment. The fact that we evaluated our DS-specific intervention in a sample of people with DSs characterised by generally high levels of deprivation and low quality of life at baseline may also have had an effect on our outcomes.

Despite the lack of demonstrable cost-effectiveness, according to NICE guidelines it is important to note what patients reported about their experiences. There was greater satisfaction with treatment in the CBT + SMC arm than in the SMC arm (see Chapter 3). Relatively few AEs (30/118) and SAEs (10/58) were deemed to have been related to DSs. In addition to the value placed on seizure control techniques by those receiving them, there was a perception that different members of the multidisciplinary care team involved in the study could potentially offer a greater range of support than would have been available in the SMC-alone arm (see Chapter 5). The value of the multidisciplinary care pathway implemented for the CODES trial was acknowledged by the neurologists, psychiatrists and therapists, and supports the view that clear interdisciplinary communication is important for the management of patients with FND. A clear difficulty was that after their involvement in the trial it was not possible for all services to maintain this multidisciplinary care pathway (see Chapter 5), which has implications for service availability after this study. Nonetheless, the value placed on aspects of the diagnostic process, the guidance given to neurologists in terms of diagnosis delivery and the written materials made available for neurologists and psychiatrists to give to patients suggest that there are aspects of our SMC that could be adopted more routinely in clinical services. However, the use of such material may benefit from enhanced education of professionals about FND more generally and DSs specifically.¹⁵²

It is important to note that our analyses reflect group outcomes and hide variability within the different outcomes. Given that we followed our published SAP,⁹² we have not yet examined whether there were subgroups of patients showing different patterns of outcome either overall or at different times, or whether there were factors clearly moderating outcome. We have commented (see Chapter 3) on the fact that our DS-specific CBT was not a trauma-focused intervention and that in certain cases patients may have been deemed to need further therapy at the end of their time in the study. Although we did not formally monitor this clinical need, we are also aware that we did not systematically measure participants’ abuse history. We chose not to administer a trauma/life events checklist at baseline because we did not want to cause distress to participants or make them feel that disclosing abuse was essential early in the study. We acknowledge that we cannot examine whether or not abuse histories relate to outcome. We did, however, have a proxy measure of abuse history, the PTSD subscale of the M.I.N.I. (see Chapter 3), although this will not necessarily reflect the range and impact of different histories of abuse that may have been present in this population. We know from our qualitative work¹³⁷ that some patients were apprehensive about being allocated to the CBT + SMC intervention arm (see also Chapter 5). CBT therapists were not able to undertake an assessment of individuals’ suitability for a CBT intervention prior to them starting therapy. Being able to do this in routine practice may have led to patients being better prepared for therapy and all it entails, and may have allowed treatment to be deferred where this may have allowed better engagement, something not possible in a RCT. Nevertheless, patients had provided informed consent to be randomised possibly to a talking therapy, indicating a willingness to engage to some extent in change.

We are aware that some within-group variability may have come from the fact that for some patients a longer time between diagnosis delivery and start of therapy may have suited their personal and other family circumstances better in terms of attending treatment sessions, rather than having to work within the timelines of the funded study. Of note, however, is the observation (see Chapter 3) from our CACE analysis that the effectiveness of the intervention (i.e. the outcome seen for all randomised patients allocated treatment) was the same as its efficacy (i.e. the outcome seen in those receiving at least nine CBT sessions), which suggests that being compliant with attending CBT sessions was not straight-forwardly related to outcome. This indicates that a more fine-grained analysis of patients’ pattern of improvement could be informative.

Strengths and limitations

To the best of our knowledge, the CODES trial is by far the largest trial to have been undertaken worldwide in the treatment of DSs (n = 368). It is a study that has attracted international interest. It had high levels of PPI/SU input to study governance via our oversight committees throughout the study. The study recruited people from 27 neurology/epilepsy services and randomised patients in 17 liaison/neuropsychiatry services across the UK, thereby removing the bias associated with single centres that may attract patient referrals corresponding to specific clinicians’ interests, although we did involve services with interest and experience in working with patients with DSs. However, the broad spread of recruitment sites, with different organisational infrastructure, may have made it difficult to get accurate estimates of the overall number of patients initially considered for eligibility to the first phase of the study. We experienced excellent recruitment and retention by randomising above target and obtaining primary outcome data (monthly DS frequency at 12 months) for 85% of those randomised. We maintained a high level of masking for our data collectors and statisticians. We recruited patients with a wide range of comorbid psychopathology and other background characteristics that posed a stringent test of our psychotherapeutic intervention. Our baseline measures were well balanced across the trial arms. Our SMC materials were developed by experts in the field, but there was also SU input and advice from a hospital information officer to ensure the readability of clinical materials for participants.

The trial demonstrated that the CBT intervention could be applied satisfactorily by a range of therapists with diverse backgrounds who, following CODES-specific training, displayed a level of adherence to the therapy manual rated at 86% while the delivery of CBT was rated at 79%; therapists’ therapeutic alliance with participants was high. It is possible that higher levels of fidelity with the seizure-related components could have led to better effectiveness; however, our current trial design does not permit evaluation of the relative importance of the DS-specific components compared with a more generic or transdiagnostic CBT approach. Supervision of therapists was provided by experts in the application of our therapy model to patients with DSs, although their many years of experience with the treatment model may not be on offer elsewhere from future supervisors. Our therapists were from a range of professional backgrounds and levels of clinical experience; therefore, we were not testing our CBT using only therapists with high levels of prior experience in working with patients with DSs, which is important for generalisability. Given that we involved substantially more therapists than initially anticipated, this may mean that some therapists had insufficient opportunity to undertake the therapy with enough patients to develop confidence in applying the DS-specific CBT skills.

We commented earlier (Chapter 3) that our SMC might be better conceptualised as specialist and standardised medical care. Therefore, a possible weakness of the study is that, with respect to SMC, we may have inadvertently created an intervention in its own right, rather than allowing services to continue to treat patients as they would usually have done. Some services changed what they generally offered patients during the study (e.g. one site began to offer a psychoeducation group to patients with FND following diagnosis in neurology and prior to individual psychotherapy). By referring to the need to provide standardised rather than usual care, it was possible to explain the need for them to keep consistent what was offered to all of that site’s patients in the trial rather than starting mid-trial to offer potential CODES patients other interventions. Thus, it was possible to maintain consistency within the trial.

It is also the case that the frequent seizure diary data collection and contact with the research team, which would not occur normally in clinical practice over such a long period, may also have served as an intervention. It may have regularly focused patients’ attention on their disorder and its perceived severity. The quality of SMC might, however, have been such that a particularly high level of belief in diagnosis was enabled in the RCT patients, which may have rendered them less typical than other patients with DSs.

Given our inclusion and exclusion criteria, there will be some other limits to the generalisability of our findings. We recruited participants into the RCT according to a strict care pathway, including only people who had initially received their diagnosis from neurologists/epilepsy specialists and who had attended their psychiatric assessment around 3 months later. Thus, people with DSs who presented initially to other services or who were referred directly to secondary care CODES psychiatrists by their GP were not included. Including both neurology and psychiatry services in the study meant that, in some settings, we needed to create care pathways that did not exist prior to the study and that ceased to function after the study ended (see Chapter 5). The care pathways, therefore, offered benefit to patients who might not otherwise have had treatment opportunities but makes our findings less generalisable to settings in which both treatment components are not available.

The study was conducted in the context of the NHS where treatment provision is free at the point of use; therefore, this health context may not generalise to other healthcare systems. Similarly, healthcare costs may reflect service characteristics that are specific to the NHS and social care context; state financial benefits changed during the course of the study and we did not, therefore, estimate these. At times it was necessary to encourage services not to discharge patients for non-attendance at early psychiatry or CBT sessions where waiting list pressures might otherwise have led to discharge. Therefore, it is possible that our uptake rates might have been lower had we not been running a trial and encouraging services to engage reluctant or non-attenders, although we cannot estimate by how much. Similarly, we offered patients up to £25 towards travel to the initial psychiatry assessment and the CBT sessions and, although some patients able to receive hospital transport might not have claimed this money, this option might also have increased CBT attendance by others.

We excluded people with documented intellectual disabilities because we wished to implement our CBT package without further modification and make it more uniform; we similarly excluded people with insufficient fluency in English, as we could not guarantee the availability of interpreters and our outcome measures were all standardised in English. The majority of patients were white, with very few from other ethnic backgrounds, possibly as a result of this exclusion criterion. The delivery of our intervention, therefore, occurred within a relatively limited cultural context. The level of flexibility required by therapists to take account of cultural beliefs or other influences was, therefore, not tested.

Several baseline (pre-randomisation) measures suggested an appreciable level of psychopathology in our sample. We used the SAPAS-SR¹¹² as a screen for maladaptive personality traits rather than frank personality disorder, but note that the mean value obtained for our sample was very similar to that observed in people with generalised anxiety disorder.¹⁷³ This finding may reflect the confounding problem of anxiety symptoms in our sample rather than providing a meaningful estimate of maladaptive personality characteristics.

A particular limitation to the discussion of the findings presented here is that, although we adhered to our SAP⁹² and focused our outcome evaluation at the 12-month post-randomisation time point (see explanations in Chapter 2), we were not able to evaluate the statistical significance of any between-group differences on our many measures at the 6-month post-randomisation time point. We chose to evaluate outcomes at 12 months post randomisation to test our intervention in a stringent manner, to minimise the number of comparisons undertaken and to be in line with the commissioning brief, which stipulated at least 12 months of follow-up. Although we presented the two arms’ 6-month data (see Tables 14 and 15, and Figures 6 and 7), we cannot comment here on between-group differences at this time point using our current statistical modelling. However, it will be important for the clinical field to know the limits, in terms of both time and scope, of any intervention offered to patients with DSs, and we would recommend that any future trials formally evaluate outcomes at a point broadly consistent with the end of treatment as well as after longer-term follow-up. It might be appropriate, therefore, to avoid having to adjust statistically for a large number of related analyses, for future work to evaluate fewer outcomes but at both time points.

The funder had identified DS frequency as an important outcome and we adopted this as we could conduct a power calculation based on our previous findings.¹ However, others have questioned whether or not DS frequency is a useful outcome measure⁵² and whether or not economic productivity might be more informative, despite the fact that it may be challenging to improve this in a short period of time. It is also possible that functional status (e.g. as measured by the WSAS in our study) might offer a more informative measure.

We did not insist that all participants had received their diagnosis of DSs based on video EEG to be eligible for the trial; just over half of the sample was diagnosed in this manner, which is broadly consistent with management of patients with DSs in the UK.⁷¹ Video EEG is not always available or deemed cost-effective in clinical services. Others have considered it appropriate to initiate treatments with greater diagnostic uncertainty.¹⁷ We overcame some of this difficulty by accepting a consensus diagnosis or by facilitating review where the patient had been diagnosed by only one clinician without video EEG, but we acknowledge the possibility of misdiagnosis in some cases. The psychiatric assessment some weeks later allowed for further scrutiny of patients’ presentations and led to some exclusions. We did not, however, classify diagnostic levels of certainty according to recommended suggestions, so our data cannot be compared more widely on this basis.¹⁷

In this trial, patients were unblinded to the treatment arm allocation. We did not attempt to blind treating clinicians and we cannot be sure that doctors did not change their practice depending on the treatment their patients received in the study. In particular, this could have been relevant to the use of pharmacological interventions as part of SMC. LaFrance et al.⁶⁸ studied the effect of antidepressant medication on DS occurrence, and, although we indicated that psychopharmacological interventions could be implemented by the SMC doctors, we did not formally record when these were prescribed or what they were. Therefore, we cannot know whether or not there was a difference in specific types of medication use between arms.

Given the trial’s pragmatic design, we did not attempt to control for therapists’ attention and time between the two treatment arms. Although our CBT package did permit therapists to address trauma in patients’ histories, this was not a specific trauma-focused therapy and after the 12-month follow-up point additional therapy to address trauma may have been warranted for some patients.

Implications for health care

Outside the CODES study, clinical services for people with DSs remain variable (see Chapter 5, Rawlings et al.¹⁵³ and Hingray et al.¹⁵⁵). The CODES trial shows that within an NHS context and through the provision of materials, guidelines and training, it is possible to facilitate diagnosis delivery and subsequent medical plus psychotherapeutic care in a manner that leads to a number of benefits in patients’ psychological and psychosocial functioning. Our 12-session intervention was based on our previous work regarding numbers of sessions.^1,82 However, there may remain patients who require further treatment, for example for complex trauma, and services need to consider remaining open to patients who need ongoing clinical input. Equally, our CACE analysis (see Chapter 3) suggests that in clinical applications of the intervention, patient-by-patient assessment of the number of sessions needed to bring about benefit in terms of DS reduction may be necessary and it cannot automatically be assumed that better results are achieved with more sessions. Clinicians, as well as patient participants, valued the provision of written educational materials, suggesting that these could helpfully be used in routine clinical practice. It is possible that greater education about DSs (and FND in general) as part of HCPs’ training¹⁵² will lay the groundwork for better-quality service delivery more broadly. Wider possibilities for delivering DS-specific CBT (e.g. via Improving Access to Psychological Therapies services) could be explored and training provided, but such services could then be encouraged to foster firm links with referrers (neurologists, psychiatrists and GPs) to provide ‘joined up’ care rather than simply accepting referrals and then managing patients completely independently. We did not record whether or not all CBT sessions were offered during standard working hours and, although we consider it highly likely that they were, future service provision might consider whether or not people with DSs who are able to maintain active employment or study despite their seizures might be disadvantaged by therapy that is provided in normal working hours only.

Recommendations for future research

Further research is needed to clarify who benefits most from DS-specific CBT and what mediates change. Planned explorations of mediators and moderators of outcome in this study will address this issue. In addition, given that it is not possible to determine this from the current study of our complex intervention, future research could determine which components of the DS-specific CBT are important in improving functioning in patients and whether briefer interventions could yield similar or greater benefits and, if so, for whom. Our qualitative data (see Chapter 5, Wilkinson et al.,¹³⁵ Read et al.,¹³⁷ and Jordan et al.¹⁵²) may also offer helpful insights into aspects of future interventions that might enhance uptake and allow realistic expectations of what participants should expect from treatment. Other, third-wave, CBT approaches (e.g. acceptance and commitment therapy) for patients with DSs¹⁷⁴ could possibly be a useful comparison with our DS-specific CBT as a means of comparing and evaluating therapeutic modalities. Future research could also try to establish estimates of clinically meaningful change for people with DSs in a variety of outcomes, such as those measured here. This could lead to improvement of future treatment studies in this area.

Finally, the cost-effectiveness analysis was based entirely on trial data and future research might also use modelling to investigate longer-term cost-effectiveness.