Background Persistent postconcussional symptoms (PCS) can be a source of distress and disability following traumatic brain injury (TBI). Such symptoms have been viewed as difficult to treat but may be amenable to psychological approaches such as cognitive–behavioural therapy (CBT).
Objectives To evaluate the effectiveness of a 12-session individualised, formulation-based CBT programme.
Method Two-centre randomised waiting list controlled trial with 46 adults with persistent PCS after predominantly mild-to-moderate TBI (52% with post-traumatic amnesia (PTA)≤24 hours), but including some with severe TBIs (20% with PTA>7 days).
Results Improvements associated with CBT were found on the primary outcome measures relating to quality of life (using the Quality of Life Assessment Schedule and the Brain Injury Community Rehabilitation Outcome Scale). Treatment effects after covarying for treatment duration were also found for PCS and several secondary outcomes, including measures of anxiety and fatigue (but not depression or post-traumatic stress disorder (PTSD)). Improvements were more apparent for those completing CBT sessions over a shorter period of time, but were unrelated to medicolegal status, injury severity or length of time since injury.
Conclusions This study suggests that CBT can improve quality of life for adults with persistent PCS and potentially reduce symptoms for some, in the context of outpatient brain injury rehabilitation services.
Trial registration number ISRCTN49540320.
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
Traumatic brain injury (TBI) is associated with a range of cognitive, emotional and physical symptoms. In the context of mild TBI (MTBI), the label postconcussional symptoms (PCS) is commonly applied to reported problems, including difficulty with concentration and memory, irritability, depression and anxiety, and physical symptoms such as fatigue, headaches and dizziness.1 Such symptoms are common in the first few days or weeks after injury but typically improve and remit. However, in a minority, these symptoms persist.2 When lasting beyond 3 months post-MTBI, they are viewed as problematic and likely to be chronic3 or even worsen4 and have previously been viewed as resistant to treatment.5 Although the term PCS is typically associated with MTBI, the same constellation of symptoms can be reported after more severe injuries, although their aetiology and course may differ.6 With this caveat, ‘PCS’ will be used here to describe the set of self-reported persistent postconcussional or PCS-like symptoms across a spectrum of TBI severity.
There is growing evidence indicating that psychosocial factors influence the persistence of PCS after MTBI7–11 and overlapping symptoms or outcomes such as anxiety, depression and employment status in studies including individuals with severe TBIs.12 ,13 Lishman14 suggested that while direct effects of brain injury may contribute towards early PCS, persistence may increasingly involve broader psychosocial factors and mechanisms. Similar diathesis–stress models have been described15 ,16 where symptoms are maintained via ‘vicious cycles’ involving factors indirectly related to or independent of any direct effects of brain injury. If this is the case, then psychological approaches to management may be applicable and effective in helping address or manage symptoms across a continuous spectrum of TBI severity, including more-than-mild TBIs, acknowledging that the likelihood and degree of persisting difficulties related to brain injury increase as injury severity increases.
Although supported by expert opinion,17 evidence for the effectiveness of cognitive–behavioural therapy (CBT) to tackle persistent PCS is limited.18 Trials incorporating CBT for individuals with TBI have tended to focus on specific symptoms such as depression,19 insomnia20 or headache.21 Typically using single-case or case series designs, results do point to possible improvements in these domains using CBT. There is some support for prophylactic interventions in patients at risk for persistent PCS in some22 but not all studies.23 Only one randomised controlled trial (RCT) has specifically focused on persistent PCS24 randomising 20 individuals with mild–moderate TBI (seen on average between 5 and 6 years postinjury) to a waiting list group or an intensive neuropsychological rehabilitation treatment programme (5 hours per week for 11 weeks) combining CBT and cognitive remediation. Improvements were demonstrated on a global symptom measure, in symptoms associated with anxiety and depression, and some measures of cognitive functioning (although not subjective functioning). The relative contribution of the CBT and cognitive remediation components could not be assessed, and it is unclear whether improvement could be achieved using CBT alone.
The current study tested the impact of individual, formulation-driven CBT, without explicit cognitive remediation or cognitive rehabilitation. The study was conducted as a preliminary RCT (with waiting list control) delivered in the context of two outpatient brain injury services. Eligibility criteria were designed to help ensure that the results would be generalisable to routine clinical practice. It was predicted that individuals would report reduced PCS and greater quality of life and health status after receiving CBT when compared with those in the control group. Supplementary analyses examined the impact on specific domains of PCS and quality of life, such as symptoms associated with depression, post-traumatic stress disorder (PTSD), pain and fatigue.
Participants were randomised to the immediate intervention group or waiting list control group: the latter were offered treatment at the end of the waiting list period. There was no financial incentive or other compensation offered to participate in the study. Ethical approval was received by the committees overseeing the two trial sites. The trial was registered with the International Standard Randomised Controlled Trial Number Register (ISRCTN49540320).
Participants were recruited through consecutive outpatient referrals to two National Health Service secondary/tertiary care brain injury clinics at the Lishman Brain Injury Unit at South London and Maudsley NHS Foundation Trust, London, UK, and the Brain Injury Rehabilitation Unit at Edgware Community Hospital, London, UK, between March 2003 and June 2009.
Inclusion criteria were of age between 18 and 65 at the time of initial assessment; evidence for (at minimum) an MTBI25 at least 6 months before; and symptoms consistent with the International Classification of Diseases (ICD-10) criteria for Postconcussional Disorder (F07.2), as laid out in the Diagnostic Criteria for Research (DCR-10).1
Exclusion criteria were non-fluent English; Mini-Mental State Exam26 scores of <20 and/or Frontal Assessment Battery27 scores of <10; moderate–severe physical disability (Barthel Index28 score <15); previous receipt of four or more sessions of CBT after their TBI; other neurological disorder independent of the TBI (eg, non-post-traumatic epilepsy); drug/alcohol misuse meeting ICD-1029 criteria for a dependence syndrome (F1x.2); and clinically assessed risk of self-harm or severe psychiatric illness necessitating involvement of a Community Mental Health Team.
Potential participants were screened for eligibility at an initial neuropsychiatric or multidisciplinary assessment, and offered a neuropsychological assessment and feedback session if this had not been completed previously. In the absence of consistent contemporaneous injury severity information from Glasgow Coma Scale (GCS) or post-traumatic amnesia (PTA), the latter was retrospectively estimated at initial assessment(s) using the Rivermead protocol30 which provides reasonable accuracy in terms of overall categorical classification of TBI severity.31 Median PTA duration was 24 hours (intervention: 30 hours, range 0–1440 hours; control: 16 hours, range 0–1800 hours).
All eligible participants were invited to participate. After providing informed consent, but before randomisation, participants completed the first (T1) set of outcome measures with the study therapist. Their general practitioner was requested that, as far as possible, any psychoactive medications were kept constant, and to inform the project team of any changes.
Randomisation was carried out independently by a Clinical Trials Unit using four preplanned categorical variables for minimisation: site (Maudsley; Edgware); injury severity (PTA≤24 hours (mild); PTA >24 hours but ≤7 days (moderate); PTA>7 days (severe)); length of time since injury (6 to <12, 12–24, >24 months); and medicolegal status (previous or current involvement vs no previous/current involvement). An 80/20 minimised/randomised weighting was applied case by case. A minimum of 40 participants were recommended for randomisation with four minimisation factors: sample size was determined on that basis, with a minimum of 20 participants per group.
Flow through the study is shown in figure 1, with 26 randomised to the intervention arm and 20 to the control arm. Median interval between injury and randomisation was 25 months (intervention: 28 months, range 6–171; control: 23 months, range 8–175). Demographic details of the participants are shown in table 1. There were no significant differences between the two groups at T1 on any minimisation variables (all p's >0.4) or gender (χ2(1)=0.450; p=0.502), age (t(1)=−0.86; p=0.393), education (t(1)=−0.109; p=0.914) or current occupational level (χ2(2)=0.073; p=0.964) (table 1). PTA duration did differ between the groups (Mann-Whitney U=225.5; p=0.44).
Outcome measures are shown in table 2. Three co-primary outcome measures examined broad aspects of PCS and their impact. Six secondary outcome measures were used to examine more specific domains (such as those relating to depression or fatigue), as well as a supplementary quality of life scale.
For the Quality of Life Assessment Schedule (QOLAS), individuals are asked to identify their two greatest problems/issues in each of the five different domains: physical health, cognitive problems, psychological issues, social issues and daily activities. Each problem is rated on a 0–5 scale (from ‘no problem’ to ‘it could not be worse’) and the items summed. Validity, reliability and sensitivity to change for the QOLAS have been shown in other clinical populations.35 ,44 It was judged particularly suitable for assessing PCS due to the potential heterogeneity of symptoms, and to help identify those symptoms most troubling to an individual and prioritise them during treatment. Blind expert rating of individual responses in the physical, psychological and cognitive domains indicated that 84.2% were directly/closely related to DCR-10 description of postconcussional disorder,1 10.2% as indirectly/partially related and only 5.5% not apparently/obviously related.
With the exception of the European Quality of Life (EuroQol), higher scores on all of the outcome measures indicate a greater degree of symptoms, distress or impact. All measures were self-report and therefore non-blind. Assessments were completed prior to randomisation (T1) and at the end of treatment (intervention arm) or after 4 months (control arm) (T2).
Treatment methods: CBT format and content
The planned intervention comprised 12 weekly 1-hour sessions of individual CBT, but the protocol allowed for longer intersession intervals. Treatment was provided by the same therapist (SP), a clinical neuropsychologist with previous experience of CBT in the context of TBI, depression and chronic fatigue syndrome.
Given the heterogeneity of PCS, an individualised, formulation-driven approach within a semistructured protocol was used. Agenda-based sessions, collaborative target setting and homework tasks were central features of treatment. Details of the intervention are described in more detail elsewhere16 and in the online supplementary information. A session structure similar to that described by Miller and Mittenberg45 was used. The first three sessions were broadly focused on problem identification, psychoeducation based on a range of sources, socialising the patient to the CBT model and formulation. Sessions 4–12 focused on the individual target problems identified collaboratively with the therapist. In the final three sessions, time was increasingly focused on relapse prevention and how to maintain therapeutic gains. Copies of the protocol and booklets are available from the authors upon request.
Central to individual CBT case formulations were (1) that persistent PCS could be maintained or exacerbated by vicious cycles,15 (2) that these cycles could be understood by reciprocal relationships between thoughts, emotions and actions and (3) that alleviation of symptoms, associated distress and functional limitations was possible by changing thinking and behaviour. While it was recognised in the treatment protocol that persistent cognitive difficulties attributable to the injury might exist for some individuals (especially for more-than-mild injuries),46 no attempt was made to provide explicit cognitive remediation or rehabilitation. With increasing injury severity and more-than-mild TBIs, psychological mechanisms were framed as likely to still play an important role in maintaining ongoing symptoms via ‘vicious cycles’, presenting a therapeutic target for CBT to help reduce symptoms and their impact. Overt resistance to psychologically orientated treatment did not appear to be a common issue for individuals completing the initial screening assessments but who did not wish to take part in the trial. Although the reasons for declining participation were not systemically collected, of the 21 individuals who declined, 9 were prepared to be referred for CBT outside the trial and another 5 cited distance as an obstacle to attending regular treatment sessions.
Control waiting list group
Individuals randomised to the control arm received a letter following randomisation with the date of an outpatient appointment to complete the T2 measures. Patients in this arm did not receive any form of additional information or psychological intervention from the service for the period that they were on the waiting list.
Participants in both groups were offered clinical follow-up after their CBT sessions finished.
Analysis was planned on an intention-to-treat basis. In practice, almost all individuals in the intervention arm completed their course of CBT and T2 outcome measures: three individuals stopped after 8, 9 and 10 sessions, respectively, due to significant symptom improvement and a lack of outstanding treatment targets. Only one patient discontinued treatment after six sessions and was lost to follow-up due to difficulties attending the sessions. No attempt was made to impute missing data for this one case.
Data were double entered from the completed paper forms on to computer and analysed using SPSS V.15.0. For the main treatment effects, analysis of covariance (ANCOVA) was used,47 comparing outcome between groups at T2 with T1 as a covariate: effect sizes are shown using partial η2 and an estimate of Cohen's d (d^), calculated as the mean difference between the contrasts divided by square root of the mean squared error.
It was predicted that the intervention arm would report fewer symptoms and better quality of life after CBT compared with those in the control arm after their time on the waiting list. Given the exploratory nature of the study and the paucity of evidence from relevant previous controlled trials, primary outcome measures were analysed independently without correction for multiple comparisons. For secondary outcomes, corrections for multiple comparisons were applied using the Benjamini and Hochberg method.48
Mean duration of the T1–T2 interval for the intervention arm was 29 weeks (SD=10.3, median=26, range 14–53) and 17 weeks for the control arm (SD=2.9, median=17, range 11–24) (Mann-Whitney U=46.0; p<0.001), producing a potential confound to the analyses. The analyses therefore also included T1–T2 interval as a covariate.
Data for the primary and secondary outcome measures at T1 (baseline) and T2 are shown in table 3. The two groups were broadly similar at baseline in terms of symptoms and quality of life reported: there were no significant differences between the intervention and control arms at T1 on any of the primary or secondary outcome measures (p>0.05) after correcting for multiple comparisons on the latter.
Effects of treatment on primary and secondary outcome measures
A significant treatment effect was noted for quality of life as assessed by the QOLAS (table 4 and figure 2). For PCS (Rivermead Post Concussion Symptoms Questionnaire (RPQ)), the findings were only significant when using T1–T2 interval as a covariate (table 4).
No treatment effect was noted in terms of global functioning (Brain Injury Community Rehabilitation Outcome Scale (BICRO-39)). Previous research into broader community rehabilitation using the BICRO-39 has indicated individual scales may be more sensitive to change than the total score:49 these scale scores are also given in table 4. Psychological Well-Being and Socialising showed evidence of a treatment effect (the latter only when using T1–T2 interval as a covariate), with little overall change in the other subscales in either group.
For secondary outcome measures, treatment effects were only noted after covarying for T1–T2 interval, for anxiety (Hospital Anxiety and Depression Scale, HADS-A), fatigue (Checklist of Individual Strength, CIS20R), anger (State-Trait Anger Expression Inventory, STAXI-2) and a quality of life (EuroQol). These results remained significant after correcting for multiple comparisons. No treatment effects were found for depression (HADS-D), symptoms associated with PTSD (Impact of Event Scale-Revised (IES-R)) or pain (McGill Pain Questionnaire (MPQ)).
Examination of the data indicated that, where T1–T2 interval was a significant covariate, shorter intervals were associated with better outcomes. This is illustrated by the data for PCS symptom severity (RPQ) (figure 3) after dividing the intervention arm into those completing treatment more quickly (n=13) and slowly (n=12) based on a median T1–T2 interval split of 188/189 days. Those taking longer to complete CBT show little change, while those who completed CBT more quickly demonstrate larger improvements. This was in contrast with the possibility that a longer interval might allow additional time for individuals to show recovery, and be associated with better outcomes.
Uncorrected post hoc t-tests on unadjusted means were completed to explore within-group changes from T1 to T2 on those measures showing a significant difference between the intervention and control conditions on ANCOVA at T2 after covarying for treatment duration. For primary outcome measures, the intervention (t(24)=3.32, p=0.003) and control (t(19)=4.00, p=0.001) groups showed statistically significant improvements on the RPQ over this period, with similar results for the QOLAS (intervention: t(24)=4.55, p<0.001; control: t(19)=2.29, p=0.034). The intervention group showed a significant improvement on the Psychological Well-Being subscale from the BICRO (t(24)=2.99, p=0.006), whereas the control group did not (t(19)=0.12, p=0.197). Neither group showed evidence of significant improvement on socialising on the BICRO (intervention: t(24)=0.12, p=0.903; control: t(19)=1.35, p=0.197). For secondary outcome measures, the intervention group showed significant improvements on the CIS-20R (t(24)=2.91, p=0.008), STAXI-2 (t(24)=2.20, p=0.038) and EuroQol (t(23)=2.18, p=0.040), while the controls did not show any improvements on these variables (t(19)=0.71, p=0.785; t(19)=0.90, p=0.382; t(19)=1.58, p=0.132). Neither group showed evidence of significant improvement on HADS-A (intervention: t(24)=1.01, p=0.323; control: t(19)=1.18, p=0.253).
Factors moderating treatment effects
To explore the possible influence of other factors on improvements in the intervention group, data from the QOLAS were examined, as it showed the largest effects of treatment. Those factors used as minimisation variables at randomisation were considered separately as additional categorical factors: data stratified by these variables are shown in table 5. Injury severity (F(2)=0.33; p=0.72), length of time since injury (F(2)=1.09; p=0.36), medicolegal status (F(1)=1.20; p=0.29) and treatment site (F(1)=0.02; p=0.90) were not related to outcome. Current (as opposed to either previous or current) medicolegal involvement was also unrelated to outcome (F(1)=0.19; p=0.67).
This study, consistent with positive findings from a previous RCT,24 found that individual CBT improved the quality of life for patients with persistent PCS. CBT also appeared to be effective in alleviating PCS in those patients who completed treatment more quickly. Possible benefits on a range of secondary outcomes including anxiety, fatigue and anger were also noted, albeit only after treatment duration was considered as a covariate. Some outcome measures showed medium to large effect sizes from before to after treatment, although the mean scores of the CBT group indicated that a majority of patients were not symptom-free or problem-free at the end of treatment (as with the previous trial24); 24% of treatment completers fell below a proposed clinical cut-off of 12 or less on the RPQ.33 The current study also demonstrates that clinical improvement in PCS is possible using CBT without an explicit cognitive remediation component, and with a programme of only 12 hours of patient–therapist contact. Contrary to expectations, medicolegal status, injury severity and length of time since injury had no significant effects on improvements in the CBT group.
Of the outcome measures used, the QOLAS gave the most robust evidence for an effect of CBT, perhaps because it captured specific symptoms identified by the patient as directly relevant to their quality of life. Its identification of ‘key’ problem areas also helped identify personal treatment targets for therapy, increasing its sensitivity to change and its clinical utility. The potential advantages for individualised patient-reported outcome measures and their responsiveness to change have led to their use in brain injury rehabilitation via methods such as goal attainment scaling.49
Contrary to expectations that persistent symptoms are static or may even worsen with time, individuals in the control group did show evidence of improvements over their waiting list period in terms of PCS (RPQ) and their impact on quality of life (QOLAS). The neuropsychiatric and neuropsychological assessment, feedback and prospect of either immediate or delayed treatment may have contributed to this change, by providing a coherent possible explanation for symptoms and offering hope that change and improvement was possible.50
There was no evidence for significant effects of treatment on depression or PTSD, both of which have been linked with PCS.9 This may have been due to floor effects as <50% of individuals scored above clinical cut-offs on the HADS-D and IES-R at the initial assessment. However, the results imply that general improvement in PCS and quality of life was not simply mediated by improved mood.
The impact of the time taken to complete CBT was noteworthy. Not using treatment duration as a covariate removed positive group differences for PCS (as measured by the RPQ), although those for quality of life (QOLAS and BICRO-Psychological Well-Being) remained. Session frequency may be a factor, with evidence that increased intensity may be associated with better outcomes in psychological therapies in other areas,51 although this may not always be the case.52 Faster completion may also have indicated greater engagement to therapy. Although the large majority completed therapy, a longer time to complete the CBT sessions was often due to patients repeatedly cancelling and rearranging appointments, perhaps reflecting lower commitment. Whatever the source of this effect, it does indicate that there may be significant variability between individuals with persistent PCS in their treatment response to CBT.
Caution should be exercised in the interpretation of the null results from the moderator analyses examining the impact of variables possibly affecting treatment response, such as injury severity and medicolegal status. The relationship between injury severity and treatment response may be complex: a residential programme focusing on treating PTSD in military veterans found that individuals with moderate–severe TBIs showed greater improvements in their PTSD symptoms compared with those with mild TBIs.53 Although compensation seeking has been associated with increased symptom report even after early intervention,54 medicolegal status itself may serve as a poor proxy for phenomena such as poor effort or malingering in neuropsychological assessments.55 The latter is more likely to affect treatment outcome, but formal cognitive measures of effort were not available for all participants. More generally, the smaller subgroups used and the corresponding decrease in power may well have obscured modest influences on treatment outcome and precluded examining possible interactions such as medicolegal status playing a more significant role in influencing outcome in mild but not severe TBI.8 Nonetheless, the null results for variables that might be expected to affect treatment outcomes suggest that they may not be major determinants of improvement in CBT, at least for the individuals seen in the current trial.
The variation in time taken to complete the sessions of CBT (with the CBT group taking longer to complete their treatment than the interval spent on waiting list in the control group) was undesirable, although the alternative of having a fixed assessment interval regardless of the completeness of treatment imposes its own difficulties. Use of a cross-over waiting list design complicates the controlled assessment of the maintenance of therapy benefits and longer term outcome. Data on other variables that might have influenced outcomes (such as homework completion, or perceptions of the usefulness of treatment) were not collected.
Another limitation stems from the use of a single therapist combined with individualised treatment. Although fitting the needs of patients, rather than using a highly manualised protocol, generalisability of the findings may be limited by the therapist's expertise. While typical of clinical practice, completion of study questionnaires given by the treating therapist may also contribute a potential response bias. Future larger/multicentre studies may usefully include a measure of therapist competence and protocol adherence, and outcome measures given by a non-treating researcher.
The current trial adds to the sparse evidence that the impact of persistent PCS, especially on individuals' quality of life, can be ameliorated even for individuals sustaining more-than-mild TBIs. Individual differences in treatment response were noted, as indicated by the impact of treatment duration, and some treatment effects were only statistically significant after controlling for this variable. An explicit, concomitant cognitive rehabilitation component did not appear necessary for these improvements: further work is needed to examine the value of this approach with or without CBT, whether certain symptoms respond differently to different treatment types and how best to integrate different treatments practically and theoretically. As research into psychological factors affecting the development and maintenance of persistent PCS continues, understanding the role of variables such as coping,11 symptom and injury perceptions,10 ,11 and broader personality traits2 may also help in refining CBT interventions and identifying mediators and moderators of change.16
The authors would like to thank Jen Attwood, Ionie Lyon and Nancy Akrasi for their help with data entry, Dr Mike Dilley for his assistance in reviewing the QOLAS data and Jill Hazan for her assistance in the trial's integration into the clinical service at the Brain Injury Rehabilitation Unit, Edgware Community Hospital.
Contributors All authors were involved in study concept and design, and in drafting/revising the manuscript for content. SDSP and SF were also involved in data collection, and SDSP in data analysis.
Funding RGB acknowledges salary support from the National Institute for Health Research (NIHR) Mental Health Biomedical Research Centre and Dementia Unit at South London and Maudsley NHS Foundation Trust and King's College London. The views expressed are those of the author(s) and not necessarily those of the NHS, the NIHR or the Department of Health.
Competing interests SDSP and RGB acknowledge grant support from the Maudsley Charity for an extension of the current study into remote-access/internet-based cognitive–behavioural therapy for persistent postconcussional symptoms.
Patient consent Obtained.
Ethics approval Ethical Committee (Research), Institute of Psychiatry, King's College London & Local Research Ethics Committee, Barnet, Enfield and Haringey Mental Health NHS Trust.
Provenance and peer review Not commissioned; externally peer reviewed.