Manualised cognitive behavioural therapy in treating depression in advanced cancer: the CanTalk RCT

Definition of advanced cancer

For this study, we defined ‘advanced cancer’ as cancer that is considered by oncology experts not to be amenable to cure. People vary in terms of prognosis and the impact of the disease on their physical functioning. Advanced cancer may include metastatic disease for which standard curative therapies have failed, disease that is already widespread at the time the patient presents and that is unlikely to be cured, and cancers with a poor prognosis, such as lung or pancreatic cancers.

Depression and advanced cancer

In this study, we defined ‘depression’ using the Mini-International Neuropsychiatric Interview (MINI). It is distinct from adjustment disorders that might be expected in those with a terminal illness, for whom appropriate sadness is a common response.

Depression is one of the most common mental disorders in people with cancer. 1 Among those with advanced cancer, the prevalence of depression measured using structured clinical interviews ranges from 5% to 45%2–4 and may vary with type of cancer. 5,6 A meta-analysis3 of depression in advanced cancer found the pooled prevalence of clinical depression to be 16.5%. There is a considerable burden on public finances presented by depression; the overall economic cost of depression generally in England was estimated to be £9B in 2000. 7

Depression in people with cancer is associated with several negative health outcomes. It undermines quality of life (QoL) for both patients and their carers,8–10 can reduce adherence to medications and treatment,10 and may prolong episodes of hospitalisation and increase health-care costs. 11 Psychological distress,12 and a diagnosis of depression in particular, predicts elevated mortality among cancer patients,13 as do higher levels of depressive symptoms. Among those with advanced cancer, untreated depression is an independent predictor of early death. 14

Therapy for depression in advanced cancer

There is a scarcity of evidence to guide the management of depressive symptoms in advanced cancer. 15–17 Among cancer patients with mixed prognoses (not limited to advanced cancer), evidence suggests that a nurse-led intervention including education about depression, problem-solving and behavioural activation is effective in treating depression in people with cancer (not advanced). 18,19 However, this intervention was also found to be effective among patients with poor-prognosis (lung) cancer. 20 Guidelines developed for the European Palliative Care Research Collaborative suggest that patients whose cancer is unlikely to be cured and who present with depression should be referred to specialist palliative care. 21 Treatment will include psychosocial support and the use of antidepressants and/or psychosocial therapy. 22

Three Cochrane reviews23–25 have looked at which psychosocial therapies are effective in advanced cancer. One of these, a meta-analysis23 of psychosocial therapies for depression in advanced cancer, cited six studies: four26–29 used supportive psychotherapy, one30 used group CBT and one31 used problem solving. The results of this meta-analysis23 were heterogeneous, and the authors concluded that further, well-designed trials were needed to evaluate if CBT is effective at treating depression in patients with advanced cancer. The other two Cochrane reviews24,25 looked at psychological interventions in women with metastatic breast cancer. One24 identified five well-designed studies. 26,30–33 The first two of these studies30,32 suggested that CBT resulted in a short-term improvement in the Profile Of Mood States (POMS) score; however, the effects were lost at the 6-month follow-up, possibly because group therapy did not sufficiently address the needs of specific individuals. An updated review of psychological interventions in women with metastatic breast cancer and depression25 identified 10 well-designed studies. 26,30–38 Three of these studies26,31,33 showed evidence of a small improvement in the POMS score among patients receiving group CBT, although this finding did not reach statistical significance.

Cognitive–behavioural therapy

Cognitive–behavioural therapy is an empirically effective treatment for major depression. The rationale behind CBT is that depression is associated with negative ways of thinking and unhelpful ways of behaving, and teaching the individual to challenge and modify negative thoughts and unhelpful behaviours helps to improve mood. CBT for treating depression compares favourably with antidepressant treatments and has been shown to be associated with significant therapeutic gains over time. Trials of CBT have demonstrated some evidence of efficacy in treating depression in cancer patients. A recent Cochrane review39 of trials of psychosocial interventions for women with non-metastatic breast cancer indicated that patients receiving CBT were less depressed than patients in control groups. CBT is an approach that may be pertinent to treating a population who may experience significant symptom burden from advanced cancer and palliative treatments, such as nausea and pain. A review40 of treatments for depression in patients with advanced cancer has suggested that CBT approaches are the best evaluated and show the most encouraging results; the studies are summarised here.

Screening and recruitment of advanced cancer patients

The European Association for Palliative Care calls for the screening and treatment of depression in patients with advanced cancer. 41 A number of different methods have been used and recommended to screen for depression in cancer patients. 42 The simplest method asks two simple questions43 that have been shown to exclude depression in non-depressed individuals in cancer and palliative care with a 97% negative predictive value. 44,45 The first two questions of the Patient Health Questionnaire-9 (PHQ-9), known as the Patient Health Questionnaire-2 (PHQ-2), are routinely used to screen for depression in primary care. Such brief screening is acceptable if followed by a clinical interview to confirm the clinical diagnosis of depression. 46 The MINI47 provides a brief and reliable method for diagnosing depression according to Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV) and International Classification of Diseases, Tenth Edition (ICD-10) criteria in cancer patients. 47–55 It has been validated against the Structured Clinical Interview for DSM,56 Diagnostic and Statistical Manual of Mental Disorders, Third Edition, Revised57 and the Composite International Diagnostic Interview58 and has been widely used in cancer patients.

Recruitment into, and retention of participants in, palliative care studies can be difficult for several reasons. 59,60 First, associated mental and physical exhaustion present major hurdles to recruitment and retention. 61 Second, there may be high rates of attrition due to early death. 61 Third, well-meaning health professionals may be protective towards patients by discouraging them from making the extra effort required to participate in a study. For example, in previous trials of CBT for depression in advanced cancer, 9–19% of patients approached agreed to participate in the research. 30,62,63 Follow-up rates as low as 44% at 10 weeks have been reported in severely ill palliative care patients,62 although higher follow-up rates are possible and some studies have reported 75% follow-up at 3 months. 30,63 Previous research conducted by our research team within the London cancer networks have achieved follow-up rates of at least 65%. 64,65 A number of strategies and recommendations have since been made to increase recruitment of people with cancer in clinical trials. 59,60

In trials conducted in our research group to evaluate CBT in older people with depression,66 CBT in cancer patients67 and advance care planning discussions in cancer,68 follow-up rates of > 85% were recorded. In each of these studies we offered both treatment at home and telephone interview follow-ups, significantly minimising attrition. Telephone CBT has been shown to be both feasible and clinically effective and cost-effective. 69,70 Individualised, rather than group, CBT is likely to facilitate recruitment and minimise attrition. Individual CBT has been found to be preferred by patients with head and neck cancer. 15,71 Benefits of individualised CBT have also been reported in a study of CBT for depression in women with metastatic breast cancer. 63

Rationale for providing therapy for depression in advanced cancer through the NHS

The research outlined suggests that CBT is an effective treatment for depression and may be a promising treatment for depression among advanced cancer patients. However, advanced cancer patients are not routinely screened and treated for depression within the NHS, despite National Institute for Health and Care Excellence (NICE)’s recommendations that this be done. 72 There is currently no manual-based therapy aimed at treating depression in this patient group and there is not currently a sufficient evidence base to determine that CBT is clinically effective and cost-effective in advanced cancer patients.

Our therapy, consistent with the evidence given, consisted of individual as opposed to group CBT. The UK agenda for treating depression is to widen access to psychological treatment through Improving Access to Psychological Therapies (IAPT) centres that operate in primary care and provide a stepped care approach provided by trained mental health practitioners. In order to provide a pragmatic trial of the effectiveness of CBT in treating patients with advanced cancer, we utilised this existing IAPT infrastructure to provide CBT to advanced cancer patients who screen positive for clinical depression.

Evaluation of cognitive–behavioural therapy provided, fidelity to the intervention and its principles

Moncher and Prinz73 proposed guidelines to enhance treatment fidelity, which have been further developed by Lichstein et al. 74 and Bellg et al. 75 They recommend assessing whether or not a psychological treatment is delivered by the therapist and is understood and carried out by the client; these stages have been respectively described as delivery, receipt and enactment. The primary purpose of the CanTalk trial was to evaluate the addition of CBT to treatment as usual (TAU) compared with TAU alone. We adopted a pragmatic approach within the constraints of the resources available by focusing on evaluation of the treatment delivery, and assessing whether or not the patient seemed to have understood the principles of CBT.

The first question, of treatment delivery, was assessed using quantitative methods. The second question, about whether or not the treatment was received, was explored using qualitative semistructured interviews with 10 participants who had received CBT. The background to this is reported in Background to qualitative work and the methodology in Chapter 3, with a summary of the findings in Chapter 5. A fuller report on qualitative experience of therapy will be prepared for publication elsewhere. The third question, relating to enactment and determining whether or not the CBT model prompted behaviour change in the patient, was beyond the aims of the study and, therefore, was not evaluated.

Concerning treatment delivery, two areas are worthy of consideration. The first important consideration is whether or not the CBT is being delivered competently, including whether or not the therapist is sufficiently flexible and able to use a range of techniques to engage the patient. Competent delivery of CBT, evaluated using the Cognitive Therapy Scale (CTS), has been shown to be related to outcome,76 although the relationship between therapist adherence and competence in determining symptom change is less clear cut than originally thought. 77 The second important consideration is whether or not the therapist adheres to the treatment manual described by Moorey, Mannix and Serfaty. The CanTalk treatment manual was intended to be made available for download on the NIHR webpage for this project. However, the manual was based on work published in Moorey78 and it was not possible to obtain permission from the publisher to make this document available. Please contact the corresponding author for more information about the manual. Indeed, we would like to point out that the revised Consolidated Standards of Reporting Trials (CONSORT) guidelines for reporting non-pharmacological trials79 have suggested that a description of different components of the intervention is to be provided when evaluating non-pharmacological interventions.

With respect to the first question concerning treatment delivery, we decided that two issues were worthy of consideration. First, an assessment of the therapist’s competence and, second, a measure to confirm that adherence to the CanTalk manual had taken place. In effectiveness studies such as the current one, it can be more difficult to ensure competence and adherence owing to the challenges inherent in delivering the treatment in a routine clinical service. Therapists may be time pressured, there may be differences between services in how CBT is delivered and, crucially in a study of patients with advanced cancer, therapists may not have experience in treating people with physical conditions.

Training of therapist for CanTalk to deliver cognitive–behavioural therapy specific for people with advanced cancer

To maximise competence and adherence to the protocol, mental health therapists with existing CBT skills were selected and trained to apply these skills to people with advanced cancer. This seemed appropriate given that the Five Year Forward View for Mental Health80 aims to expand the IAPT services for people with long-term conditions (LTCs). However, we chose to stipulate that, within IAPT services, therapists would need to be accredited by the professional organisation of the British Association for Behavioural and Cognitive Psychotherapies (BABCP). Accreditation is not in itself sufficient assurance that high-quality treatment is delivered; however, it does improve the likelihood. When adherence is concerned, it needs to be acknowledged that accreditation cannot ensure that therapists deliver the therapy in accordance with a specific protocol for the treatment of depression associated with cancer. However, selecting high-level therapists minimises any distraction caused by needing to learn CBT techniques and allows them to apply their skills to cancer patients. This was done by training all therapists and some of the supervisors on how to assess and deliver the Moorey, Mannix and Serfaty model to advanced cancer patients; more details are provided in Chapter 3, Training Improving Access to Psychological Therapies therapists.

Clinicians’ experience of the CanTalk trial

During the course of the study, we became aware that cancer clinicians involved in identifying patients who did not come from a mental health background had a wide range of views about psychological research in people with advanced cancer. We informally detected a range of views about the relevance of psychological research in advanced cancer patients. Some clinicians were strongly supportive and others suggested that CanTalk was outside their remit or that they felt uncomfortable with the psychological nature of the trial. Further examination of the literature suggested that there was a dearth of research in the area. As cancer clinicians play an essential role in facilitating recruitment, we decided to conduct qualitative work to determine the views of clinicians about psychological research in advanced cancer, their experience of referring into the CanTalk trial, any obstacles they may have experienced and how to improve recruitment into similar trials.

Therapists’ views of treating people with advanced cancer

Both the client and therapist play a role in the outcome of CBT,81 with the therapeutic alliance having an impact on the outcome of therapy. 82 When randomised controlled trials (RCTs) have failed to demonstrate an impact on cancer populations,83,84 interpretations of the findings are purely theoretical. Empirical literature suggests that a number of therapist-specific factors play a major role in the outcome of CBT. 85,86 Despite the importance of the therapists’ role in this type of therapy, there is a dearth of qualitative research conducted from a therapist perspective that addresses their personal experiences of delivering therapy. Indeed, most of published research into experiences of CBT has used quantitative methods, and concerns about such methods have been raised,87 with the focus being purely on client perspectives of CBT with no attempts to explain things qualitatively and directly from a therapist perspective.

In order to determine whether or not the costs outweigh the benefits of this particular treatment in this population, we took a holistic approach so that, as well as quantitatively addressing the effectiveness of treatment, we undertook a qualitative assessment (not required in the brief) of treatment from the therapists’ perspectives on (1) their overall views and experiences of delivering CBT, (2) how services and therapy may be improved, (3) specific training requirements and (4) issues in delivering therapy sessions. This was to help inform the optimum use of resources.

Experience of cognitive–behavioural therapy in people with advanced cancer

There is evidence to suggest that CBT is an effective treatment for people with depression and advanced cancer63 but little is known about how patients in this group perceive CBT or about their thoughts and experiences of it.

For cancer patients attending group CBT, their experiences have been positive; patients enjoy the interpersonal and social environment of the group88 and learn skills to challenge and solve problems. 89 Feedback from patients receiving individual CBT has also been encouraging. Omylinska-Thurston and Cooper90 interviewed eight patients with primary cancers who had received a course of psychological therapy within a NHS service for cancer patients and found that participants found talking about their feelings to someone outside their family and problem solving helpful. In a study in Australia,91 cancer patients with metastatic disease commented that CBT allowed them to share their thoughts and feelings with an understanding, caring therapist. Finally, Anderson et al. 92 found that hospice patients reported CBT to be acceptable and effective.

Although qualitative work has focused on cancer patients’ experience of individual CBT, there is little information about the experience of advanced cancer patients. The remit of IAPT services in the UK is to be expanded to cover patients with chronic health conditions,93 including cancer. Despite the London Cancer Alliance94 having some reservations about the ability of IAPT therapists with brief training to treat older people with complex needs, there are few data evaluating how patients should best be managed.

The qualitative work was designed to elucidate what aspects of CBT participants found helpful, their thoughts about their therapist and the impact of CBT on their QoL, and patients’ views about the best way to support them emotionally.

Trial registration

This trial is registered as ISRCTN07622709. The full protocol is available online and published in Trials. 95

Trial design

This is a parallel-group RCT, stratified by antidepressant prescribed at baseline (yes/no), comparing TAU with TAU plus up to 12 sessions of manualised CBT. Allocation ratio for the two trial arms was 1 : 1.

Patient and public involvement

The trial was designed in response to a HTA programme call. Janet St.John-Austen, a user of cancer services, was an active contributor to the design of the project, the preparation of the materials (including the layout and wording for clarity and sensitivity) and ethical considerations. She attended regular steering group meetings and was invaluable in commenting on methods to boost recruitment. She also contributed to the interpretation of the results and write-up.

Ethics approval

A favourable opinion for the conduct of the study was granted by the London–Camberwell St Giles National Research Ethics Service committee, Central London (REC3 reference number 11/LO/0376). This study formed part of the National Cancer Research Network (NCRN) clinical trials portfolio (registration number 10255, ISRCTN07622709). The trial was conducted in compliance with the Declaration of Helsinki. 96 Informed written consent was obtained from all participants.

Changes to protocol

Several amendments were made to the study’s procedures during the course of the study, including changes to streamline recruitment methods and the addition of three qualitative substudies. These amendments are outlined in Table 1.

Inclusion criteria

1. People with a diagnosis of cancer not amenable to curative treatment as assessed by their clinician and defined as those receiving palliative radiotherapy or chemotherapy, and those with metastatic disease or subsequent incurable recurrence. The diagnosis was verified by oncologists or general practitioners (GPs).

2. A DSM-IV diagnosis of depressive disorder using the MINI. 47

3. Sufficient understanding of English judged by clinic staff to enable them to engage in CBT.

4. Eligible for treatment in an IAPT centre. Either the patient or their GP had to be located in an appropriate IAPT catchment area.

Exclusion criteria

1. Clinician-estimated survival of < 4 months, verified by the patients’ oncologists or GPs.

2. People at high risk of suicide, established through module C of the MINI.

3. Currently receiving, or having received in the last 2 months, a psychological intervention recommended by NICE aimed at treating depression (e.g. interpersonal psychotherapy, CBT).

4. Suspected alcohol dependence using the Alcohol Use Disorders Identification Test. 97

We did not recruit in areas where the local palliative care service includes routine access to CBT, to avoid contamination of TAU arm by non-study CBT.

Timeline and setting

Recruitment of participants commenced on 1 September 2012. Recruitment finished on 14 December 2015 (the end of the study’s agreed recruitment period). The first participant was recruited into the study on 27 November 2012. The first 6-week follow-up took place on 15 January 2013 and the final 24-week follow-up took place on 19 May 2016.

Participants were identified in four ways: (1) from oncology centres, (2) through GP practices, (3) through the Marie Curie Hospice, Hampstead and (4) through self-referral using leaflets left in GP surgeries and oncology clinics.

Oncology centres

Either support staff or UCL researchers screened suitable patients for depression using the PHQ-2,101,102 the first two questions of the PHQ-9,103 a valid screening measure for depression routinely used in general practice.

Patients who scored ≥ 3 points were provided with a pre-screening information pack and asked if they would be willing to be assessed for the study. If they scored ≥ 3 points but did not wish to participate, their permission was sought for their GP or oncology team to be informed that they may be depressed. If they agreed in principle to take part, a researcher undertook a further assessment, using the MINI to establish a DSM-IV diagnosis of depressive disorder. If a DSM-IV diagnosis of depression was confirmed, the patient was given an information pack. The patient was then given at least 48 hours to reflect on whether or not they wished to participate in the study before giving written consent for participation. If a patient consented to take part, the researcher then conducted baseline assessments and passed the participant’s details to an independent trial administrator, who arranged randomisation through the PRImary care and MENTal health (PRIMENT) Clinical Trials Unit (CTU). The study administrator informed the participant by telephone of their group allocation. For those randomised to the treatment arm, the administrator liaised with IAPT to set up the therapeutic sessions. Both administrator and PRIMENT were situated separately from the trial research team to maximise masking of trial arm allocation from the researchers.

General practitioner practices

University College London researchers attended the practice and trained a practice research nurse in the standard operating procedure on how to identify potential suitable participants. Practice administrators identified people on the cancer register and consulted the GP on whether or not the patient had advanced cancer as defined in the protocol. By mutual agreement according to availability at the practice, either the practice nurse or PCRN staff contacted the patient by telephone or face to face in practices to explore whether or not they were willing to answer the two PHQ-9 screening questions for depression. The procedure was then the same as for the oncology centres but, in this case, the practice nurses/PCRN nurses collected follow-up data at the relevant time points.

Marie Curie Hospice, Hampstead

University College London researchers consulted clinic records and checked the eligibility criteria for those identified as suitable by hospice staff. They then approached potential participants for screening as outlined above. Those suitable were then told about the trial and consented once they had had 48 hours to consider participation.

Self-referral

The leaflet contained the PHQ-2 for patients to conduct a quick assessment of their mood themselves, suggesting to people with a score of ≥ 3 points that they may have depression and that they should either (1) approach the clinical team within the site or (2) contact the study team directly using the reply slip attached to the leaflet. The process of recruitment was the same as that previously outlined.

Screening data

In instances in which the UCL researchers undertook the screening, they were able to collect data about the number of patients screened – the proportion of whom satisfied entry criteria – and, if patients declined to take part, the reasons (if given). It is important to highlight that the ethical principle that patients are free to decline to take part or withdraw from a study without giving a reason was included in the project’s ethics approval submission. When Comprehensive Local Research Networks (CLRNs) were undertaking identification, selecting and screening of patients, comprehensive data for numbers of people screened and reasons for declining to take part in research were not always available as CLRNs indicated that they did not have the resources to collect these data.

Masking

Once a participant had been randomised, the trial administrator unblinded that participant by clicking on an ‘unblind’ link on the Sealed Envelope system that generated an e-mail to themselves with details of the group allocation.

It is not possible for patients or therapists to remain blind to the treatment group. The trial manager was unblinded only if needed, for example if there was a problem referring a patient to therapy. The trial team worked at UCL and was based in a different location to the therapy teams that conducted the trial intervention.

Assessment of blindness

The UCL researchers who were blinded were asked to guess group allocation (TAU alone, TAU plus CBT or do not know) at 3 months (post intervention) and 6 months (follow-up). Although the PCRN assessors were blinded, they requested that any additional data collection was kept to a minimum and, therefore, they did not make an assessment of blindness.

Unmasking for those conducting the analysis did not occur until databases were closed.

Treatment as usual

All participants received TAU from all clinicians involved in their care. This consisted of routine support, such as appointments with GPs, clinical nurse specialists, oncologists and palliative care clinicians. Participants’ physical health and medication were reviewed and treatment was modified according to symptoms, such as pain. Psychotropic medication was allowed to be prescribed as necessary, by either the GP or the oncologist. In line with NICE’s guidance,107 specific psychological support should have been available for those who presented with psychological needs at any time, and study participants were not exempt from receiving external psychological support. We discouraged specific psychological interventions aimed at treating symptoms of depression (e.g. CBT or interpersonal psychotherapy), but, ultimately, we could not interfere with usual care for ethical reasons. We recorded the numbers of participants receiving any psychological therapy during the trial, although we predicted that the numbers were likely to be small. 66 We did not stipulate post randomisation that antidepressant medication could not be used or that the dose should be fixed. Withholding a recognised treatment for depression would be unethical and would not reflect TAU.

Cognitive–behavioural therapy (in addition to treatment as usual)

The CBT was delivered through IAPT93 and well-being centres. IAPT/well-being centres train, supervise and supply therapists to treat people in primary care with mental health problems. For the purpose of the study, only step 3 and 4 (high-intensity) therapists who had experience of CBT were used. They were given 1 day’s training by the CanTalk team (SM, MS and KM) so that their existing CBT skills could be adapted to use a specially developed treatment manual for people with advanced cancer. The manual detailed modifications in the structure of therapy and its content; in particular, it took into account physical health problems, existential issues and communication with loved ones.

Identifying and contacting Improving Access to Psychological Therapies centres

The following methods were used to identify IAPT/well-being collaborating centres where the intervention could take place. IAPT/well-being leads were approached for selected boroughs across London for the pilot stage and further areas within, and outside, London for the definitive trial. Areas were selected for several reasons: first, where study team personnel had existing links with oncology teams, hospices and GP centres for recruitment to be feasible; second, where study personnel had existing links with IAPT/well-being services and IAPT/well-being leads expressed an interest in research; third, IAPT/well-being services were approached if they had a mature service running for ≥ 2 years and, therefore, would be more likely to be able to participate in the delivery of specialist CBT; and fourth, we were also approached by a number of services who identified the CanTalk trial through trust research co-ordinators.

Our approaches were initially made by telephone to IAPT/well-being leads, identified from websites and by personal contact. These were followed by an e-mail summarising the project with the advantages to IAPT/well-being services highlighted as follows:

1. free training for IAPT/well-being high-level CBT practitioners

2. improvement in IAPT/well-being therapists’ CBT skills

3. developing the delivery of IAPT to patients with long-term physical health conditions, which is consistent with national aims

4. effective publicity for IAPT/well-being so that services are properly funded

5. demonstration that, if effective, CBT, delivered through IAPT, would represent a good model of care for LTCs.

Training Improving Access to Psychological Therapies therapists

The IAPT/well-being services were asked to supply at least two high-intensity IAPT/well-being therapists for manualised training. IAPT/well-being supervisors were also encouraged to attend.

Training was delivered by Stirling Moorey, Marc Serfaty and Kathryn Mannix. Therapists were supplied with a therapists’ manual, presented with Microsoft PowerPoint^® 2010 (Microsoft Corporation, Redmond, WA, USA) slides giving an overview of sessions in the manual as well as videos and role plays and were asked to participate in practising skills using a variety of scenarios. How they could access the resources online was also explained. Finally, they were taken through processes required for, and associated with, the research.

Training took place at the following sites: UCL for London, the south of England, the Midlands and the west of England (SM and MS); Chester for the North West (MS and KM); and Newcastle upon Tyne for the North East (KM).

Training took place in 27 IAPT/well-being services. Of these, 25 participated in the study. Chester and Newcastle upon Tyne did not participate as we could not set up recruitment centres in these areas. Details of the 25 services engaged in the study, including the number of therapists (124 in total) represented from each service and the IAPT/well-being leads, are presented below.

The number of people trained in applying CBT skills to patients with cancer came from the following services.

• London:

  • North London – 13 people from five IAPT services from Barnet, Camden and Islington, Enfield and Haringey.

  • East London – 19 people from nine IAPT services from City and Hackney, St Bartholomew’s and The London (Hospital), Tower Hamlets, Redbridge, Homerton, Newham, Waltham Forrest, Havering, Barking and Dagenham.

  • South London – 30 people from seven IAPT services from Lambeth, Lewisham, Bromley, Bexley, Croydon, Southwark and Greenwich.

• Outside London:

  • the south of England – four people from two IAPT services (Sussex and Brighton and Hove).

  • the west of England – two people from one IAPT service (Avon and Wiltshire).

  • Midlands – four people from Coventry and Warwick IAPT services.

  • North East – 11 people from one IAPT service (South Tyneside).

  • North West – five people from one IAPT service (Stockport).

Evaluation of training

Training was evaluated using a feedback questionnaire with a Likert scale asking about specific aspects of training; boxes enabled additional free-text comments. These findings are presented in Chapter 5.

Location of therapy

Patients were offered the opportunity for face-to-face therapy in their local IAPT/well-being centre. In some cases, this was at a local GP practice, depending on the set-up of the service. We did consider delivering therapy in the patients’ own homes but were constrained by limits on safe lone working and, thus, therapy could not be delivered in this way through the IAPT/well-being service. When patients were too frail or reluctant to continue to attend therapy in an IAPT/well-being centre, we allowed for the delivery of telephone CBT, providing the patient had seen the therapist at least three times and it was deemed by the therapist to be consistent with safe working practices. We asked therapists to record whether therapy was delivered face to face or by telephone.

Measure to assess competence of delivery of cognitive–behavioural therapy

A scale for measuring therapist competence in cognitive therapy, based on the original CTS,109 is the 12-item CTS-R. 110

This revised version improves on the original CTS by eliminating the overlap between items, improves on the scaling system and defines items more clearly. In this trial, we assessed competence by having an independent rater listen to audio recordings of therapy sessions using the CTS-R.

Measure of adherence to cognitive–behavioural therapy in cancer manual

In this context, we defined therapist adherence as the extent to which the therapist adhered to the essential ingredients described within the treatment manual developed for use in the trial by Marc Serfaty, Stirling Moorey and Kathryn Mannix.

Detailed information about the content of the intervention was collected using a ‘Therapy Components Checklist’ (TCC) (Table 2). The role of the TCC was threefold. First, it provided the therapist with a guide to which elements were delivered (or not) from the manual. Second, it provided information about which elements had been delivered by therapists during the course of the trial. Third, it enabled us to evaluate whether a therapist’s self-report of what they said they did was consistent with what they actually did, judged by an independent rater.

The general procedures and main interventions for successful treatment are as follows:

1. general procedures (eight elements)

2. behavioural techniques (four elements)

3. cognitive techniques (13 elements)

4. cognitive–behavioural techniques (nine elements)

5. specific cancer topics (12 elements).

We also collected information from the therapist about what they thought were the three most important aspects of the therapy and why, whether or not they felt that there was anything missing from therapy and whether or not they had any general comments.

Analysis

Ensuring safety

This study was not a drug trial and our main concern was centred around risk of self-harm or suicide. UCL researchers and the research nurses screening in oncology centres were given training by Marc Serfaty covering the serious adverse events (SAEs) protocol (see Appendix 1), detailing what action should be taken if patients assessed were considered to be high risk, even though they would be excluded from entry into the study. These appropriate governance procedures and good clinical practice applied to all patients seen, to ensure safety at all times. For those who were detected as being at high risk at follow-up, similar procedures were actioned.

Examples of good practice and the difficulties associated with risk were discussed, including ensuring that there was an opportunity for individual supervision on request with a senior member of the team who was always available. This culture of transparency ensures that researchers are able to always raise concerns about their participants. Time was also taken within supervision to highlight the importance of behaving ethically and safely in all aspects of clinical work.

We considered the possibility that patients being seen by IAPT/well-being therapists may be detected as being at increased risk. However, practitioners are bound by their own governance procedures in their assessment of risk and are very familiar with managing suicidal patients. Because of the number of IAPT/well-being centres collaborating in this study and minor variations in the procedures on how to manage risk, we stipulated that IAPT/well-being therapists adopt their own governance procedures, as this is what would happen if the intervention were to be rolled out across the country, but that they also complete and return a SAE form. The chief investigator, Marc Serfaty, would then contact the IAPT/well-being team within 2 working days to discuss the case and consider whether or not the participant should be withdrawn from the study.

Embedded qualitative study

In this report, we have included information about a qualitative substudy exploring clinicians’ views on referring into the CanTalk trial, which we conducted independently. However, as qualitative research was not commissioned by the HTA programme, only a brief summary of the methodology and results has been provided (see relevant sections of this report):

• an evaluation of clinicians’ experience of referring into the CanTalk trial evaluated through qualitative interviews (described in Clinicians’ views of the CanTalk trial)

• an evaluation of the patient experience of therapy was evaluated through qualitative interviews (described in Patient interviews to determine experience of cognitive–behavioural therapy)

• an evaluation of the therapists’ experience of delivering CBT to advanced cancer patients using semistructured interviews (described in Therapists’ views of delivering cognitive–behavioural therapy to people with advanced cancer).

General interview procedures for qualitative data capture

Semistructured, one-to-one interviews using topic guides (presented in Appendices 2–4) ensured that all participants within these three groups were asked the same questions to minimise researcher effects. The topic guide initially consisted of a number of open-ended, non-leading questions. We aimed to cover the six types of questions described by Patton114 for qualitative interviews. These include experience/behaviour questions, opinion/belief questions, feeling questions, knowledge questions, sensory questions and background/demographic questions.

The interview began with introductory questions in order to help establish a good relationship with interviewees to encourage rapport. The topic guide followed a logical sequence with topics being grouped together with corresponding categories. The questions were formulated to avoid influencing the participant. We attached probes to each question to aid rapport building. The topic guide ended with a closing question to encourage participants to discuss topics or issues that were not mentioned previously. Individual interviews were audio recorded and transcribed verbatim. Flexibility allowed the interviewer or interviewee to divert from questions to pursue other areas where necessary. 115

The interviews took place where there were minimum distractions for the participant and for the researcher to aid the dialogue. All interviews were audio recorded for transcription at a later date. We aimed to recruit up to 20 interviewees for each of the three areas of interest from the CanTalk trial to allow a sufficient number to generate themes from the data until no new themes emerged from the interviews.

Clinicians’ views of the CanTalk trial

Any clinician who had been involved in referring patients to the CanTalk trial was considered eligible for inclusion. The study aimed to obtain a good cross-section of participants including consultants, registrars, nurses and radiotherapists. The study also aimed to include the views of 15 clinicians from each of the referring sites.

Clinicians were asked to describe their role in oncology clinics, their previous involvement in research, their views on non-drug trials, the CanTalk trial, any patient feedback and their views about future psychological studies. A detailed topic guide is provided in Appendix 2.

Therapists’ views of delivering cognitive–behavioural therapy to people with advanced cancer

We used a purposive sampling frame and contacted all therapists who delivered CBT as part of the CanTalk trial. We aimed to recruit up to 20 IAPT therapists for one-to-one, face-to-face, semistructured interviews in a quiet setting agreed by both the researcher and therapist.

The therapists were asked to describe their role and how it applied to the CanTalk trial, what they knew about the patient, their views on working with patients with advanced cancer and, with CanTalk patients in particular, any components of CBT that were or were not useful, and any other important views. A detailed topic guide is given in Appendix 3.

Patient interviews to determine experience of cognitive–behavioural therapy

All participants recruited from London sites who had reached the 24-week follow-up who had not died or withdrawn from the study and who had received at least one session of CBT were approached by a member of their clinical team via post and asked for permission to contact them. Those agreeable to being contacted or who had already provided consent to contact were sent an invitation letter offering them the chance to provide feedback about their experiences of CBT.

Interviews took place in the participants’ homes or within the department where the research was taking place. The interview schedule was as follows: participants were asked to describe what therapy they had, what knowledge they had of CBT, what they found helpful or not, their views of the therapist, how CBT had an impact on their life and their views about CBT for low mood in cancer patients. The precise topic guide is provided in Appendix 4.

Qualitative analysis

All interviews were audio recorded and transcribed. Thematic analysis was used to analyse data using NVivo version 11 software (QSR International, Warrington, UK) by two researchers who were ‘immersed’ in the data before proceeding with data analysis, as this can strengthen data analysis. 116

Researchers used a matrix-based analytical method of framework analysis to analyse the data. 117 The method of ‘complete line-by-line coding’ was used to broadly identify any code or theme that emerged without restricting analyses to detecting particular themes. 116 Codes were then grouped into broader themes. For the purposes of triangulation, researchers compared their findings with each other once they had identified broader themes. Any discrepancies were discussed until agreement was reached.

Screening measures

Demographic information

A baseline assessment form was used to ensure that entry criteria were satisfied and that the patient had consented. It also included demographic information on sex, date of birth, marital status, ethnicity, employment status, highest level of education, previous history of depression, cancer diagnosis and whether the tumour was defined by clinicians as primary or secondary, along with date of diagnosis of the primary or secondary tumour type. We also noted other treatments and prescribed medication, dose and frequency.

Quantitative

We collected a number of measures at the different time points summarised (see Table 3).

Primary outcome

Secondary outcomes

Measures to reduce bias at baseline

Timing of measures

Sample size and power calculations

Published data for trials of CBT suggest that initial reductions in BDI-II with time may not be linear. A separation in depression scores favouring therapy has been observed within 6 weeks of starting treatment133,134 and continues after the treatment phase has finished. 135,136

For ethical reasons, participants were entitled to withdraw from the study without giving a reason; however, we recorded the reason for withdrawal from the trial if known. We recorded the timing of any attrition. Differential dropout may occur early on because people are not satisfied with their trial arm allocation. Dropout at later phases is more likely to be due to factors such as advancing disease or death, which are less likely to be influenced by group allocation. Details of the assumptions made concerning attrition and the effect of treatment are described below and summarised (see Table 4).

Analysis plan

Outcomes

Quality-adjusted life-years were calculated from EQ-5D scores at baseline, 12 and 24 weeks’ follow-up. The EQ-5D is a non-disease-specific measure for describing and valuing health-related QoL. 143 The measure includes a rating of own health in five domains (1, mobility; 2, self-care; 3, usual activities; 4; pain/discomfort; and 5, anxiety/depression) and a rating of own health by means of a VAS [a ‘thermometer’ (score of 0–100)]. The EQ-5D is widely used in economic evaluations for common mental health disorders. The health states from the EQ-5D were given a utility score using responses from a representative sample of adults in the UK. 144 From these, QALYs were calculated using the area under the curve approach as defined by the utility values at baseline and at each follow-up.

Costs

Permissions

Table 5 summarises the process taken for research to take place.

Project set-up started in December 2011. Because of limited resources, some services were not able to complete data for the number of people assessed for eligibility, namely oncology centres 7, 9 and 16 (see Table 5).

Furthermore, the time taken from the submission of documentation to the approval for research to start is shown in Table 5. The time taken for submission and documentation varied between as little as 1 month for oncology centre 16 and 9 months for oncology centres 1, 2, 3 and 6.

Oncology centres 4 and 14 encouraged early involvement to prepare the documentation for review and discussion prior to submission to R&D, for approval for research to proceed. As this process would normally be considered part of the submission and approval process, we have included it in Table 5 in order to make a comparison with other centres. As the research nurse undertaking recruitment from oncology centre 11 left her role, oncology centre 11 had to withdraw from the study.

Recruitment rate

Figure 1 shows cumulative original, amended and actual recruitment by month and year from the start of the project in August 2012. This figure is supplemented by the data in Table 5. No participants were recruited until December 2012 because of the delay in permissions; this is represented by the relatively flat recruitment line from August 2012 to February 2013, when recruitment was initially slow.

FIGURE 1.

Cumulative original, amended and actual recruitment time.

As permissions came through, recruitment accelerated to around 7.8 participants per month (93 participants over 12 months for the period from June 2013 to the end of May 2014).

Our recruitment target and methods had to be revised to allow for delays in permissions that were beyond our control. We overcame these hurdles by increasing the number of clinics within a locality from which recruitment was taking place and enrolling other areas of the UK into the study. This required us to establish whether or not IAPT services had the capacity to take on additional cases and whether or not IAPT services from different regions could actually participate, especially as changes in the configuration of IAPT services nationally were taking place. This made them cautious about being able to commit to the trial.

It became evident that the same patients were being seen in both the oncology clinics and hospices. Therefore, we chose to expand the project to centres with large numbers of oncology patients, such as oncology centre 14. It is also notable that, in the final 6 months, the recruitment rate fell to 3.3 participants per month (20 participants were recruited in the final 6 months). This is because the reservoir of patients available became depleted.

A CONSORT flow diagram is provided in Figure 2. Of those assessed for eligibility, only 2.7% (230/8398) were suitable for the trial. However, at least one outcome measure was available on 80.4% (185/230) of participants.

FIGURE 2.

Participant flow and recruitment; a CONSORT flow diagram.

Reasons for withdrawal from the study

Twenty-one (9.1%) of the 230 recruited participants died and 51 (22.2%) participants withdrew. Although the ethics committee stipulated that patients did not need to give a reason for withdrawal from the study, we aimed to collect this information when possible (these reasons are summarised in Table 10). The number of participants who withdrew from the study during the first 6 weeks was twice as high in the CBT group as in the TAU group.

Reason for missed follow-up

The CONSORT flow diagram (see Figure 2) provides participant flow. It is also of note that follow-up was missed at different time points. The specific reasons why participants were not followed up at 6, 12, 18 and 24 weeks are given in Table 11.

Demographics of cancer

Two-thirds of patients had tumours of one of the five main groups (breast, colorectal, lung, prostate and haematological), with the majority, around one-third, having a primary diagnosis of breast cancer. Of the 41 ‘other’, 14 were neurological, four were oesophageal, three were bowel, two cervical, two liver, two bone, two anal, two sarcoma, one nose, one throat, one ovarian, one testicular, one parotid gland, one mesothelioma, one myelofibrosis and three patients had cancer of unknown primary diagnosis (Table 13). The mean time since diagnosis of cancer was skewed by 44 participants with a haematological cancer, among whom the mean time since diagnosis was 2970 days. The median length of time since the primary diagnosis was just over 2 years (770 days).

Diagnosis of depression, previous psychiatric history and treatment

All participants included in the study satisfied a MINI diagnosis of depressive disorder. Three-fifths of participants had a previous history of depression, with the mean number of previous episodes being just over two. The duration of the current episode of depression was also skewed, with the median duration being around 12 weeks. One-tenth of participants had previously received CBT, with less than one-third receiving treatment for depression (Table 14).

Antidepressant usage

Fifty-five (23.9%) of the 230 participants were taking an antidepressant, 26 out of 115 in the TAU group and 29 out of 115 in the CBT group. There was no significant between-group difference as to whether or not an antidepressant was prescribed. Although the names of the antidepressants were recorded, the dose was available in only 35 cases. The mean dose-equivalent of fluoxetine prescribed was 23.4 mg (SD 8.2 mg) for the TAU group (n = 17) and 31.8 mg (SD 8.2 mg) for the CBT group (n = 18). There was no significant difference in the mean dose by group allocation.

Other psychological therapies

The record of any psychological intervention reported by patients or recorded in their case notes during the period of the trial is summarised in Table 15.

Expectations at baseline132

Prior to randomisation, participants were asked to predict the degree to which they thought that their mood would or would not improve on a 7-point Likert scale ranging from –3 to 3.

Treatment preference

Patients’ preferences for treatment were collected on a 4-point Likert scale (0–3), as in Serfaty et al. 66

Main outcome

Data for the main outcome measure, the BDI-II, are shown in Table 16 and on the histograms of Figure 3. Mean BDI-II score by time and group is also presented (Figure 4).

FIGURE 3.

Histograms of BDI-II score at baseline and at 6, 12, 18 and 24 weeks. (a) TAU, baseline; (b) TAU, 6 weeks; (c) TAU, 12 weeks; (d) TAU, 18 weeks; (e) TAU, 24 weeks; (f) CBT, baseline; (g) CBT, 6 weeks; (h) CBT, 12 weeks; (i) CBT, 18 weeks; and (j) CBT, 24 weeks.

FIGURE 4.

Mean BDI-II score, by time and randomisation group.

At baseline, the data were normally distributed, as exemplified by the similar mean and median scores (see Table 16). A BDI-II score of 20–28 points suggests the severity of depression to be moderate and a score of 29–63 points suggests severe depression. A total of 194 out of the 230 participants had at least moderate depression, 17 participants had mild mood disturbance (a BDI-II score of 14–19 points) and 19 participants had minimal depression (a BDI-II score of 0–13 points). BDI-II scores decreased for both groups by around 5 points during the course of the study. There appears to be a slight skewing of the data for CBT at 18 and 24 weeks. This is shown in Table 16 and Figure 3. The histograms show that there appears to be a shift to the left by week 12 for the CBT group with respect to the BDI-II score with TAU plus CBT (see Figure 3). The BDI-II scores for CBT and TAU, from baseline to 24 weeks, are also shown (see Figure 4).

Statistical modelling was conducted in accordance with the methods described in the original analysis plan (Table 17). There were no baseline differences for BDI-II, antidepressant use or clustering by therapist or IAPT service. Nor, using this model, were there any differences in previous history of depression, baseline EQ-5D, baseline duration of depression, or length between primary diagnosis and baseline visits. It is noteworthy that the ICC for clustering by therapist was low. Indeed, the analyses of the primary outcome variable (BDI-II) over all time points were identical, whether clustering by therapist or IAPT service, or whether no clustering at all was included in the model. This is because, averaged over time points, there was no evidence that the components of variance associated with therapist and IAPT service were other than zero.

Although we did not find a significant benefit of CBT plus TAU versus TAU alone for the treatment of depression, participants who were widowed, divorced or separated had significantly poorer outcomes on the BDI-II without the addition of CBT to TAU (Table 18 and Figure 5). There was no clustering by therapist or IAPT service.

FIGURE 5.

Mean BDI-II score by marital status and randomisation group with time.

As shown in Figure 5, those in the TAU group who were widowed, divorced or separated remained unwell, whereas those who received CBT behaved similarly to the other participants, with some improvement in BDI-II scores.

Secondary outcomes

Secondary outcomes for the PHQ-9, ECOG-PS and satisfaction with care are shown (see Tables 20–22).

Patient Health Questionnaire-9 scores and the changes observed were similar to the BDI-II. Data for the PHQ-9 are shown in Table 20.

Selection criteria stipulated that participants needed to have a PHQ-2 score of ≥ 3 points to enter into the trial. It is of note that one participant in the TAU group had a score of only 1 point on the PHQ-9 at baseline. This is because the baseline measures were repeated immediately prior to randomisation and scores may change in the time from the week at screening for entry into the trial to when the baseline measures were collected. Recovery in IAPT services is taken as < 10 points on the PHQ-9. The number of participants with scores of < 10 points in the TAU group at baseline, 12 and 24 weeks was 20, 33 and  35, respectively, and in the TAU plus CBT group at baseline, 12 and 24 weeks the number was 26, 37 and 33, respectively.

The PHQ-9 suggests that a score of 10–14 points indicates moderate depression and a score of 15–19 indicates moderately severe depression. As shown in Table 20, the scores are not skewed and consistent with our target population, namely people with a diagnosis of depression of moderate severity. At the end of treatment, both groups had median depressive symptoms at the mild end of the severity range (range 5–9 points).

Eastern Cooperative Oncology Group Performance Status scores suggested that only one-fifth of participants were fully active, with two-fifths being of restricted mobility (Table 21). Participants were equally balanced between the groups with respect to disability and this appeared to remain constant with time. Although the ECOG-PS questionnaire contains the category ‘dead’, these data were excluded in the analysis to maintain consistency with the primary outcome measure, as it is not possible to collect mood ratings from deceased participants. The CanTalk data suggest that participants’ physical functioning was towards the active end of the scale, with the greatest proportion being fully active or restricted, a similar portion being classed as ambulatory, fewer participants being limited and no participants being classed as disabled.

The mean satisfaction with care was 80% (40 out of a total score of 50 points) at baseline (prior to randomisation) and did not change with time or by treatment group (Table 22). This finding is also represented graphically (Figure 6).

FIGURE 6.

Satisfaction with care total score, by randomisation group and time for each follow-up.

Training for the intervention

Training was delivered to 129 therapists and information regarding their experience of the training session was collected from 22% of them (29/129) by asking them to rate the questions on a 4-point Likert scale (Table 23).

Take-up and engagement with cognitive–behavioural therapy

A total of 543 (39.3%) sessions out of a potential total of 1380 sessions (12 × 115) were taken up, of which 32 were by telephone (5.9%): seven participants took up one telephone session, three participants took up two sessions, one participant took up three sessions, one participant took up four sessions and one participant took up 12 sessions. The mean time from being referred to being seen by IAPT was 29.4 days (SD 26.7 days).

Among the total of 115 participants randomised to CBT, the mean number of sessions received was 4.7 (SD 4.9). It is notable that, of the 115 participants allocated to CBT, 41 (35.6%) did not take up any sessions and 36 (31.3%) had at least eight sessions. The reason for sessions not being received was predominantly participants withdrawing after randomisation (Table 24). The median time from being referred to being seen by IAPT was 21 days [quartile range (QR) 13–37 days].

Quality of cognitive–behavioural therapy and adherence to manual

Health-care workers’ views about the CanTalk trial

An additional qualitative substudy was conducted, looking at the views of 14 health-care workers. This study looked at health-care workers’ views of psychological [non-Clinical Trial of Investigative Medicinal Product (CTIMP)] research in general and of the CanTalk trial more specifically. An initial analysis of the data generated the following emerging themes.

Therapists’ views of treating patients with advanced cancer

Sixteen therapists were interviewed qualitatively using semistructured questionnaires and their repsonses transcribed and coded onto a Microsoft Excel^® 2010 (Microsoft Corporation, Redmond, WA, USA) spreadsheet. Although such interviews were not required for the purpose of this report, a brief summary is presented here.

Patients’ views of cognitive–behavioural therapy

Costing the intervention

In total, 1380 sessions were available for participants in the CBT arm; however, fewer than half (n = 543) were taken up. A total of 65% of CBT arm participants attended at least one session and 31% attended at least eight sessions. The mean number of sessions attended in the CBT arm per participant was 4.8. It was assumed that when a session was offered but a participant did not attend it still constituted the IAPT’s therapist opportunity cost. The mean cost of CBT intervention was £948.75 (SD £427.00) per participant.

Quality-adjusted life-years

Line plots for EQ-5D VAS and histograms for EQ-5D index scores are presented in Figures 7 and 8, respectively.

FIGURE 7.

Median EQ-5D VAS scores at baseline, 12 and 24 weeks, by time and randomisation group.

FIGURE 8.

Mean EQ-5D values at baseline, 12 and 24 weeks, by randomisation group.

There were no differences in EQ-5D median scores at baseline, nor was there any advantage of CBT over TAU at 12 weeks. At 24 weeks, there was a trend towards improvement with CBT; however, this was not significant. A similar pattern was shown for the VAS on the EQ-5D. There was no statistically significant improvement in QALYs at 24 weeks. Both these figures and the data on which they are based show that any mean changes over time or mean differences between trial arms in either of these scores were minimal.

In the CBT group, the mean EQ-5D-5L values were 0.64 (SD 0.24) at 12-week follow-up and 0.68 (SD 0.22) at 24-week follow-up. The figures for TAU were similar: 0.64 (SD 0.24) and 0.63 (SD 0.27), respectively.

Summary statistics are presented in Table 29 and the results of the statistical analysis in Table 30.

Table 29 shows the EQ-5D scores at baseline and at each follow-up period, and total QALYs over the 24-week follow-up. The data are presented as both median scores and lower and upper QRs (Q1 and Q3), as well as mean, SD and the number of participants available.

The maximum QALY accrual was 0.82. The mean QALY accrual was 0.504 for the CBT group and 0.386 for the TAU group. The difference adjusting for baseline was 0.118 in favour of CBT; this difference in QALYS was not statistically significant (p = 0.12).

Cost–utility analysis

The cost–utility results are presented in Table 30, showing the incremental cost-effectiveness for each CBT session compared with TAU for the period of the trial.

Analysis suggests that CBT has higher costs and produces more QALYs than TAU. To reduce the degree of uncertainty around the estimates, bootstrapping is applied.

Cost-effectiveness

Figure 9 is a scatterplot of incremental costs and QALYs and shows the cost-effectiveness plane of 1000 bootstrap-replicated ICERs for CBT compared with TAU, based on health and social care costs and QALYs over 24 weeks, adjusted for baseline costs and utility.

FIGURE 9.

Cost-effectiveness plane: CBT vs. TAU, ratio of incremental cost to QALY gained.

The CEP indicates that there is a 15.5% chance that CBT is cheaper and produces more QALYs, and a 74% chance that CBT is more expensive and produces more QALYs.

The CEAC indicating the probability that CBT is cost-effective compared with TAU, based on health and social care costs and QALYs over 24 weeks, is shown in Figure 10.

FIGURE 10.

Cost-effectiveness acceptability curve.

Willingness to pay (WTP) for an improvement in QoL per participant per year is around £20,000–30,000. 151 As illustrated by the CEAC (see Figure 10), at a WTP for an improvement in QALY of up to £20,000, the probability that CBT is cost-effective compared with TAU is not > 51%. At a WTP for an improvement in QALY of up to £30,000, the probability is at most 52.5% and it never rises above 56% at a WTP of up to £100,000.

Furthermore, translating ICER results into monetary terms from the adjusted analysis, we can estimate that an additional £47,985 would need to be invested to generate a unit increase in the participant’s QoL.

Cost-effectiveness of improving negative symptoms

In comparison with the TAU group, participants in the CBT group showed an improvement in QoL at 24 weeks; however, the difference was not deemed statistically significant at the 0.05 level.

Complier average intention-to-treat analysis has estimated that, on average, every CBT session would be expected to decrease the total BDI-II score by 0.3 points compared with TAU. Although this effect was not found to be statistically significant (p = 0.213), dividing the cost by the difference-in-change scores shows that CBT had higher costs and better outcomes; however, the difference between the groups was marginal. Taking account of the ICER, which is the cost incurred by CBT to produce an extra unit improvement on this scale, Figure 9 indicates there was a 14.5% probability that CBT was dominant (i.e. less costly and more effective) compared with TAU. The north-east quadrant, defining a scenario of CBT having higher costs and better outcomes, has a 62% proportion of ICERs, whereas the south-west quadrant, indicating lower costs and worse outcome, is represented by the lower proportion of bootstrapped ICERs.

Results in the light of previous findings

Our main finding is that CBT (plus TAU) did not show any benefit over TAU in terms of either our primary outcome or our other measures. Although CAITT analysis suggested that the estimated mean change in BDI-II score per therapy session was –0.3 points for CBT compared with TAU, the wide CI (95% CI –0.760 to 0.170 points) was consistent with the main result, showing no significant benefit of CBT (p = 0.21). A subanalysis has suggested that for a particular subgroup of participants, those who were widowed, divorced or separated, there may be a benefit from CBT (mean change –7.21, 95% CI –11.15 to –3.28; p < 0.001).

Previous work reviewing the use of psychological interventions for depression in advanced cancer has raised questions concerning research design, generalisability of the findings and, more specifically, uncertainty about whether or not CBT, either group or individual, is effective in an advanced cancer population. Although three published Cochrane reviews23–25 have suggested that psychosocial therapies are effective for advanced cancer, the last two24,25 covered women with metastatic breast cancer only. The largest study30 that used CBT for low mood in metastatic breast cancer had only 62 participants and suggested no effect of CBT at 3 and 6 months. A more recent study by Savard et al. 63 suggested a benefit of CBT for treating depression in a similar population, but numbers were limited to a small sample size of 45 patients. The small sample size calls into question how much authority these results should hold.

In addition to the finding by Savard et al. 63 that cognitive therapy (as opposed to CBT) was effective for depression in metastatic breast cancer, the SMaRT (Smart Management Research Trials) Oncology-3 trial20 found that integrated collaborative care delivered in a secondary care setting for people with lung cancer was substantially more efficacious than usual care. However, both of these studies20,63 delivered interventions to specific populations of oncology outpatients.

It is generally agreed that CBT is an effective treatment for depression, so what might explain our findings in people with advanced cancer? There are two broad possibilities. One is that limitations in the CanTalk trial obscured a treatment effect. The other is that CBT, as delivered in accordance with the CanTalk manual, is ineffective in this patient group. We will consider these possibilities in turn.

Reason for a treatment effect in widowed, divorced and separated individuals

The CanTalk trial demonstrated a significant effect of CBT for people who were widowed, divorced or separated. Although we found a highly significant effect, the trial was not powered to detect a significant difference by marital group and it is possible that this was a type I error.

The effect of demographic variables as moderators of response to CBT in adults has been previously reported. 152–154 Fournier et al. 153 found that being married, unemployed and having more antecedent life events were associated with a better response to CBT than to antidepressants.

A study by Button et al. 152 was consistent with our findings, suggesting that being widowed, divorced or separated was associated with better responses to CBT. Although no explanation was offered by Button et al. 152 for this finding, we considered two main possibilities. The first is that this is a random finding. However, our finding is a highly significant effect (p < 0.01) and is consistent with previous reports. Second, people who are emotionally isolated or have suffered a bereavement or separation may benefit from the non-specific component of CBT, by having a warm and friendly person to talk to, with positive effects on their mood. We have not explored the outcome when broken down into separate categories for the subgroup (widowed, divorced or separated) as this was not a part of our pre-analysis plan. It might have also been useful to know whether or not the participants’ status was a recent event. Although there appears to be some indication that CBT is helpful, it would be premature to specifically recommend CBT to people who are widowed, divorced or separated until further research validates these findings.

Establishing whether improvement is associated with the specific effects of CBT or, rather, a non-specific treatment effect such as having a warm, friendly person to talk to remains to be evaluated. There are very few studies that examine the impact of non-specific interventions with CBT in depression. Although older people with depressive disorder have reported feeling that just talking helped, this did not improve their depression scores. 66,67 However, preliminary findings from more recent feasibility work, comparing acceptance and commitment therapy (a new-wave therapy derived from CBT that uses cognitive diffusion and mindfulness methods rather than attempting problem-solving) with a talking control for people with advanced cancer155 indicates that being given the space to simply talk and be heard is well received. It is possible that additional psychosocial support, possibly in the form of CBT, should be considered in this demographic group.

Subsidiary findings

Power

We recruited a total of 230 participants. Our original power analysis suggested that 240 participants (accounting for attrition at follow-up) would provide 90% power to detect a 6-point difference at 12 weeks and 80% power to detect a 6-point difference at 24 weeks (see Table 4). Retention in the trial exceeded our original estimates so that, with 230 participants, we had sufficient numbers to detect a 3-point change on the BDI-II at 90% power, suggesting that our trial was overpowered. Savard et al. 63 reported a clinically significant improvement with CBT in metastatic breast cancer patients but these findings should be treated with caution as the sample size was small, with 21 in the CBT group and 16 in the ‘wait list’ group. One of the strengths of our study is that it is sufficiently powered to produce robust findings and to answer the question of the benefit of CBT delivered by an IAPT service for people with advanced cancer.

Attrition is common and unavoidable in studies of those with advanced, progressive disease. It is always a serious consideration when considering the validity of the findings, as differential dropout in the two study arms may reduce the internal validity of the findings. We conducted a sensitivity analysis, assuming that all people in the CBT group who dropped out recovered and we still could not demonstrate any benefit of CBT.

Representativeness of the sample population

Reliability of the diagnosis of depression

Typically, the CanTalk trial targeted people with a depressive episode (ICD-10 F32), a recurrent depressive disorder (ICD-10 F33) or a mixed anxiety and depressive disorder (ICD-10 F41.2), as identified by the MINI. However, it is not always easy to distinguish depressive disorder from an adjustment disorder with a prolonged depressive reaction (ICD-10 F43.21).

Rayner et al. 22 found that 69% (27/39) of patients in palliative care who were diagnosed with major depressive disorder at baseline had remitted 4 weeks later. Our target population, after adjusting to a recent major life event around their cancer diagnosis, could experience remission of their symptoms. This is consistent with the observation that people allocated to TAU alone also improved. This is also a common finding in RCTs because of regression to the mean.

The mean time from the initial diagnosis was almost 4 years in our study population. Therefore, it is unlikely that a diagnosis of cancer per se was the trigger for distress; however, it is possible that people were adjusting to the realisation that cure was no longer likely. Unfortunately, the entry definition for the study was for ‘people with cancer not amenable to cure’ and this group is more heterogeneous in term of adjustment disorders. For example, patients with lung cancer that is not amenable to cure, which is often known from the time of diagnosis, may have had more time to adjust than people who entered into the study with a recent 1 month reoccurrence of breast cancer. In hindsight, it may have been useful to ask participants whether or not they had had recent upsetting information about their disease prognosis or recurrence.

Reliability of the main outcome measure

The BDI-II was used as the main outcome measure. The BDI has been used in trials of psychotherapy for patients with advanced cancer. 63,120,121 However, in the light of our findings, we need to consider whether or not the BDI/BDI-II is a valid scale for measuring change in depressive symptoms in advanced cancer. The difficulty with assessing severity of depression in cancer is that there are a number of items on depression scales that complicate estimations of severity. For example, people may lose weight because of the cancer itself or nausea associated with treatment, sleep may be increased or decreased because of the use of opiates or decreased because of associated physical disease, such as pain.

Prior to setting up the study, we considered the different scales to be used for measuring depression in cancer and also reflected that people with advanced cancer, our target population, are likely to be older than average.

Unfortunately, there is no ideal scale for measuring depression in cancer158 and a variety of scales have been used. 159,160 These include the BDI-II,118 Brief Symptom Inventory,161 Centre for Epidemiologic Studies–Depression Scale,162 Geriatric Depression Scale,163 Hospital Anxiety and Depression Scale,164 PHQ-9,102 Profile of Mood States – Short Form,165 Zung Self-Rating Depression Scale166 and the Modified Edinburgh Postnatal depression Scale. 167

As we were evaluating the effects of CBT, we decided on the BDI-II for two main reasons. First, the BDI-II, or its previous version (the BDI), has been used in previous studies to measure depressive symptoms in advanced cancer. Second, the BDI-II contains a balance of cognitive and somato-affective elements. The rationale behind CBT is that targeting cognitions associated with depression results in an improvement in mood. Although other scales may ask some cognitive-related questions, such as suicidal thoughts and hopelessness, they tend to be less weighted towards the measurement of cognitive elements. The BDI-II includes thoughts about being punished, worthless, being a failure, self-dislike and self-criticism, all of which, along with addressing existential issues, were a target of the CanTalk intervention. Nevertheless, like all measures of depression, the BDI-II also includes somatic symptoms, for example cachexia, which, in this population, are related to the underlying physical disease, rather than to depression per se. Even with CBT, these elements may be less amenable to change, making it harder to detect a significant treatment effect using the BDI-II.

Finally, a criticism associated with the BDI-II is that it takes quite a long time to complete, and this may increase the number of missing items, which makes it unreliable. This was not the case in this study, as we found that the number of missing items was low.

Screening using the PHQ-2 suggested significant depressive symptoms, and this was confirmed with the MINI, giving a DSM-IV diagnosis of depression in all participants who entered the trial. Nevertheless, in 36 out of 230 (15.6%) cases, the BDI-II score was < 17 points, suggesting mild mood disturbance (score 11–16 points) or normal ups and downs (0–10 points). Even though our analysis adjusted for baseline depression score, the inclusion of participants with low depression scores generally reduces the ability to influence significant change. This observation was also consistent with qualitative feedback from some therapists, who indicated that, by the time some patients were seen, they were no longer ‘depressed’.

Randomisation

In this randomised trial, various demographic and other factors that are known to predict depression outcome were well balanced in both arms.

However, more people were employed in the CBT group. Unemployment is known to be associated with poorer prognosis of depression. This imbalance would be expected to favour CBT, but we did not find this to be the case. Although our population included a lower proportion of single, never married people (17%) than in the total UK adult population (33%), this may be explained by the fact that the mean age of our population was higher than the UK average, and the proportion of single and never married people is generally lower among the older population. This age distribution may also explain why people with a degree or higher degree in our population was lower than for the UK population as a whole.

Both groups were also balanced on our recorded potential sources of bias, such as antidepressant use, use of other psychological therapies, predicted improvement and treatment preference. Therefore, these factors are unlikely to account for our lack of apparent treatment effect.

Dropout

There was considerable morbidity among our population, which accounts for the fact that almost one-third of people were not able to provide complete data at any any particular time point. However, this finding may be misleading because, as shown by Figure 2 for example, although 24-week follow-up data were available for 130 out of 230 (56%) participants, people dip in and out of follow-up. Analysis was possible because at least one post-baseline data point was available for 185 out of 230 (80.4%) participants. Furthermore, missing data for people who died can, in this study, be regarded as missing completely at random,168 with respect to whether or not they have received CBT (potential cases of suicide excluded; none occurred).

The patterns of early and late dropout may be relevant, especially as they may not be completely random. Dropout early on (within 6 weeks) may occur because people are unhappy about their group allocation, whereas dropout late in the treatment may occur because people are unable to comply with the rigours of having to attend therapy sessions. Over 75% of participants indicated a preference for CBT; however, withdrawal in the early stages was greatest for the CBT group. It is unlikely that dropout from the CBT group is explained by people receiving their preferred treatment. One explanation for the greater dropout in those allocated to CBT is that their expectations were not met because they had to wait to be seen by IAPT services. It is notable that people who engaged well with therapy were seen within a median of 16 days, whereas those who engaged poorly had to wait a median of 26 days. We are aware of the need for caution when interpreting these data, especially as it could be argued that a delay for people to be seen by IAPT services occurs because people are physically unwell. However, using ECOG-PS, which is a measure of disability, we found no significant differences in physical function at baseline between those who were seen quickly and those who were not. Furthermore, overall satisfaction with care was equally high in both groups. Our belief that delays in being seen by IAPT services explained some of these findings was supported by qualitative interviews, in which participants voiced dissatisfaction at ‘having to wait’ to be seen. If waiting to get CBT elicits dissatisfaction, then reduction in waiting time must be considered. This observation is further supported by our finding that attrition later in the study (weeks 12–24) was balanced across the two arms.

Satisfaction with care was high and this did not differ between the TAU and TAU plus CBT group. Satisfaction with care as measured in this study related to a general satisfaction. In hindsight, it may have been more helpful to collect quantitative data specifically related to participants’ satisfaction with CBT.

Engagement with therapy is an important issue when considering whether or not a particular intervention can be delivered in a NHS setting. A total of 22% (25/115) of participants withdrew before starting therapy. Data from IAPT for 2014/15169 suggest that just under 70% of participants referred to IAPT enter treatment. The CanTalk trial found that, although 65% (75/115) of participants attended at least one therapy session, only 31.3% (36/115) completed at least eight therapy sessions. A major reason for dropout was that participants had significant physical ill health. Our quantitative and qualitative data suggested that a proportion of participants withdrew either because of health reasons or because of the rigidity of IAPT practice, the policy of which tends to be to discharge people if they fail to attend on two occasions. Although ethics stipulated that patients did not need to give a reason for not taking up treatment, it was perceived by the study researchers that patients felt that attending therapy sessions for some was burdensome. 170 Although it is costly, attrition may be minimised by seeing people in their own homes and by allowing great flexibility for those with chronic physical health problems. Unfortunately, we were not able to offer treatment in a home setting because of restrictions on lone working within IAPT. Although we did offer the opportunity for telephone CBT, few sessions (5.9%) were taken up. Moreover, these telephone sessions were accounted for by one participant who received all 12 sessions over the telephone as they were unable to attend face-to-face therapy. Although our protocol recommended that at least three sessions were offered face to face, we decided to be flexible. Although the alternative option of offering computerised CBT should be considered, the benefit of these packages is questionable170 and we cannot make recommendations on these interventions based on the results of our study.

A total of 36 out of 115 (31.3%) participants referred for therapy were recorded as having ‘completed’ it. However, the proportion of those who engaged was slightly higher (32.7%), as only 110 out of the 115 would have been able to take up therapy, with five participants dying by 3 months’ follow-up. This figure compares with recent IAPT data171 suggesting that 41% of people complete therapy. Of the 110 participants who could have received all 12 therapy sessions, 41 participants had withdrawn or data were missing. Although for a proportion of these patients the decision to terminate therapy was agreed between the patient and the therapist, it would have been useful to have had data on why these termination decisions were made.

Seventeen out of 115 (14.8%) patients in our population were discharged from IAPT, in keeping with the IAPT policy of discharging people who fail to attend. Qualitative interviews suggested that therapists tried to be more flexible with our patient population. If IAPT is to treat patients with LTCs, such as advanced cancer, consideration must be given to physical problems and attendances. Furthermore, six people were not referred for therapy; despite us sending an e-mail to IAPT and confirming receipt in a follow-up telephone call, we were informed by IAPT that a referral had never been received. Patients were subsequently re-referred for ethical reasons, but the delay in this process meant that they had reached the final follow-up point in the trial. Such administrative errors may be minimised by closer tracking of participants. However, changes in staffing, IT systems and administrative errors do occur and remain potential sources of error. These errors could be minimised in future trials by asking patients allocated to CBT to be proactive in contacting the trial team and IAPT if they have not heard from a therapist within 3 weeks of allocation.

The quality of the intervention

We are aware that research-active sites tend to deliver higher-quality care. This could be reflected in improved outcomes from IAPT. However, as 25 IAPT services from across England were involved, we would argue that these were representative of a range of IAPT services. Engaging IAPT centres that have an interest in participating in the study is a prerequisite for the study and, if anything, would be likely to bias the outcome in a positive way.

Although a meta-analysis on therapist adherence and competence suggests that these factors play little role in determining symptom change,77 competence is routinely measured in IAPT, and adherence to treatment protocol is considered essential in evaluating trials of CBT. Therefore, we have to consider the possibility that CBT failed to effect an improvement in depressive symptoms because the quality of therapy was poor and that therapists did not adhere to the treatment manual. However, the CTS-R suggests that the quality of delivery of CBT was at the upper end of the proficient range.

Number of therapy sessions

National Institute for Health and Care Excellence’s guidelines142 for the treatment of depression suggest that a range in number of sessions may be required, with 16–20 individual CBT sessions over 3–4 months. When planning the study, the consensus view was that fewer sessions would be needed as it was felt that most patients with cancer would have reasonable pre-morbid adjustment and that one of the main hurdles in their lives is adjusting to a life-threatening illness. In CanTalk, a mean of 4.7 sessions (SD 4.9 sessions) were taken up, which is lower than a mean of 6.3 treatment appointments attended overall for IAPT in 2014/15. 171 However, the number of people with physical health problems treated by IAPT, although increasing, remains low.

Complier average intention-to-treat analysis found a change in BDI-II score with CBT of 0.3 points per therapy session and the wide CI suggests no benefit of CBT. We chose to adhere to a consensus opinion among clinicians, notably that a change of 6 points is considered clinically significant on the BDI-II. It is possible that a 6-point difference would be achieved by simply delivering more sessions. Although speculative, as a linear relationship between the number of CBT sessions and a reduction in the BDI-II score cannot be assumed, one could envisage a 6-point change with around 20 CBT sessions. More cautiously, scaling up the ‘per-session’ estimate by the average number of sessions experienced within the treatment group for those who took up at least one session (this comes to 7.14 sessions) produces an estimate of –2.11 points (95% CI –5.43 to 1.21 points); this is an estimate of the average benefit in the treatment group for ‘compliers’ (defined here as those who had at least one session of CBT), which can be compared with the principal ITT analysis estimate (–0.84, 95% CI –2.76 to 1.08).

Our original assumptions were that our sample population had better pre-morbid resilience than a population with repeated depressive episodes. In hindsight, it may be that the severe physical consequences of their disease necessitate considerably more sessions for clinically significant change. In reality, it is going to be very difficult and there may not be enough time to engage people for 20 sessions. Furthermore, even if this were possible, resources are not currently available in the NHS to deliver sufficient therapy for treatment to be successful. It follows that CBT is not a pragmatic treatment for people with depression and advanced cancer.

It is notable that, when the CanTalk project was set up, we stipulated that, when possible, IAPT should fast track people into therapy, ideally within 2 weeks. Our study findings suggested that there was an association between the time for the IAPT appointment to be received and the number of therapy sessions completed; the faster an appointment with IAPT, the more sessions were taken up. People with more physical problems may be less able to attend appointments and may also be less likely to engage in therapy. However, attendance was not found to be related to physical health status judged by the ECOG-PS. Furthermore, IAPT services send out appointments irrespective of an individual’s health status. It follows that, in order to encourage people to take up therapy sessions, rapid referral into therapy is desirable.

Although we requested that participants be fast tracked into therapy within 2 weeks of referral by the research team, in reality this was not possible because of service constraints. Nevertheless, the mean time for participants to be seen was 29 days, which is better than routine IAPT referral, for which waiting times for a course of treatment (for those finishing a course of treatment in March 2015) are quoted as being < 6 weeks to enter treatment for 37,097 (78.0%) people and < 18 weeks for 45,589 (95.9%) people. In contrast, the presence of a CBT practitioner within a palliative care service ensures rapid access to treatment. We deliberately excluded areas that offer this service from recruitment to avoid contamination from the TAU group.

We are confident that the CanTalk trial is one of the largest and methodologically most robust trials of therapy in psycho-oncology. So, if the results are sound, the outstanding question is why CBT delivered using the CanTalk manual was ineffective in the IAPT setting.

Depression in advanced cancer is more treatment resistant

A previous study62 of CBT in palliative care found an effect for anxiety but not depression. The nature of depression in advanced disease may be different (e.g. depression in advanced disease may have a biological underpinning that makes it less amenable to treatment). Despite the lack of effect on depressive symptoms, therapists reported useful work with other problems and goals that patients brought to therapy. CBT, as a collaborative effort in problem solving, works with what the patient brings and it may be that problems other than depression were shifted by the therapy. Unfortunately, CanTalk did not use measures of personal problems or anxiety, so it is not possible to determine if CBT had an impact on areas other than depression. For many people with advanced cancer, anxiety may be a more disabling symptom than depression, but we do not know if our treatment improved anxiety. CBT in palliative care is less disorder focused and much more problem focused. For instance, symptoms of fatigue or pain may be the main ones addressed in therapy and, although these are related to depressive symptoms, they may change without effecting changes in depression. This raises questions of whether or not a disorder-specific approach and disorder-specific measures are the best outcome measures in advanced diseases.

The intervention was not sufficiently well designed

There was no difference in mean satisfaction with care between patients allocated to TAU and TAU plus CBT. This is probably because questions referred broadly to their total care and CBT made up only a small element. However, the mean score of around 40 on a scale ranging from 0 to 50 suggests that considerable improvement could not be made with the care delivered. Engagement with therapy may provide an indication of whether or not the intervention was acceptable. Although 65% of people took up the offer of therapy, it is notable that around one-third completed 8–12 sessions. Data from IAPT for 2014/15169 suggest that the rates of uptake were similar to IAPT. However, our population had many more physical problems and, therefore, continued engagement as health deteriorates is likely to be challenging. Although telephone CBT was offered to participants, the majority elected to be seen face to face. Qualitative interviews with therapists and patients suggested that the treatment could be delivered and that it was well received. Generally the therapists and the patients liked the materials used and we were pleased that they adhered to the protocol by showing flexibility in which elements were tackled and when.

The CanTalk manual was constructed on the basis of considerable clinical experience in working with cancer patients and is geared towards addressing the cancer-related issues faced by patients with advanced disease. The entry criteria for the study did not differentiate between depression resulting from the disease and pre-existing depression or even depression subsequent to the disease but not related to cancer. Some therapists reported that the problems with which patients presented were not cancer related. Thus, the manual may have restricted them in their delivery of therapy.

An alternative explanation is that the therapists did not focus enough on cancer-related issues. It is interesting that the TCC reveals a discrepancy between the observer and therapist ratings in the areas of impact of illness, impact of disease and mood on behaviour, beliefs and expectations about illness, concerns about current and future ability to cope, and impact of disease and death on loved ones. Therapists were more likely than raters to have believed that they covered these issues in a session.

Although the problem-solving focus of CBT can help patients with early-stage cancer to develop effective coping strategies and evaluate unrealistic hopeless and helpless beliefs, in advanced disease there is less scope for problem solving. It may be that ‘third wave’ therapies, such as acceptance and commitment therapy, which are based more on acceptance, could be more relevant. However, this is not the clinical experience expressed by clinicians working in palliative care, in which problem-solving approaches to fatigue, pain and avoidance of activity for example seem to work very effectively.

We need to consider whether or not a cognitive–behavioural approach is not the treatment of choice in this target population. More recent work has been conducted by Walker et al. 20 in the SMaRT Oncology-3 trial, in which 142 people with lung cancer (poor prognosis) were randomised to integrated collaborative care, which uses elements of CBT or usual care. This found that the active arm was substantially more efficacious than usual care alone, although data were available for 113 (80%) and 97 people at (68%) at 12 and 24 weeks, respectively. It may be that the use of CBT in combination with this integrated collaborative care approach is more suitable for advanced cancer patients. However, this remains to be evaluated for cancer patients with different tumour types.

Engagement with the intervention

A low uptake of therapy may explain failure to find a treatment effect. However, the CanTalk trial was a trial of the clinical effectiveness, not efficacy, of CBT. The uptake of therapy for people referred to IAPT is around 70%. In the CanTalk trial, of 99 people referred to therapy, 66 (67%) took up at least one therapy session, which is similar to IAPT. Although we offered people the opportunity to tell us their reasons for withdrawal, it is not ethical to pressurise people into providing a response. In qualitative interviews with participants and therapists, there was a suggestion that CBT was well received. Physical problems meant that people were not always able to attend therapy in a local IAPT centre or GP premises. In addition, there may be stigma associated with attending a centre that is known to address mental health problems. Despite this, only a small proportion of participants took up the option of telephone CBT. A strategy to boost engagement could be to offer patients the opportunity to receive therapy in their homes or at the oncology centre, but it is unclear whether or not this would be taken up. Delivering integrated care through palliative care teams may facilitate engagement with psychological interventions and is recommended by the Improving Outcomes in Supportive and Palliative Care for Adults with Cancer guidance. 72

Psychological interventions are less effective than previously thought

Cuijpers et al. 172 suggest that the effectiveness of CBT for depression generally may have been overestimated, possibly owing to publication bias, small sample size and a lack of suitable control groups. Although the therapeutic outcomes from various psychological interventions are broadly similar173 and the effectiveness of an intervention may be enhanced by improving retention with treatment, the findings from CanTalk trial suggest a lack of treatment effect in patients with advanced cancer.

Service use

Costs

The increase in costs throughout both follow-ups for both groups may be partially explained by more frequent inpatient admissions and general uptake of secondary care.

When planning services, one needs to take into account the full costs of an intervention, including the IAPT therapist’s time allocated to a participant who may not, for example, attend a therapy session, as was the case in this trial.

The intervention and supervision

• It was possible for CBT therapists to be trained to deliver CBT to an advanced cancer population.

• IAPT therapists may have been reluctant to explore some of the physical health, deterioration and death issues with patients.

• A variety of CBT methods were used, with predominantly cognitive and cognitive–behavioural methods, as well as discussion of specific cancer topics, being undertaken.

• The number of sessions taken up was consistent with other studies.

• Independent ratings of therapy, judged by the CTS-R as proficient, suggested that high-quality therapy was delivered.

• Qualitative interviews suggested that therapy was well received.

• Although this was a pragmatic study, qualitative interviews suggested that therapists would have preferred more specialist supervision than currently exists in IAPT services.

• Therapists’ self-report appeared to be an accurate reflection of what was delivered.

• No AEs were recorded.

Structure of services

• The lowest proportion of people recruited was from GP databases (2.6%), with the numbers of people available for screening being small; the greatest proportion was from oncology clinics (10.2%).

• IAPT discharged 11% of participants because they did not satisfy requirements of IAPT for therapy.

• There was considerable replication of documentation required for the research to be approved by local trusts, which held up the trial in some centres. The time taken from submission to approval for the research to proceed varied from 2 to 9 months.

Clinical

Research implications

Summary of service implications

• CBT delivered through IAPT is neither clinically effective nor cost-effective for people with depression and advanced cancer.

• People who are widowed, divorced or separated and who have advanced cancer may benefit from screening and referral to IAPT for CBT.

• Clinical supervision may benefit from involving people with additional expertise for specialist client groups, possibly through a centralised supervision system.

• Therapist’s self-report sheets may be helpful in guiding the components of therapy delivered.

Research recommendations

• Recruitment should be considered over a wide recruitment base from oncology and hospices, but this should be no longer than 6–12 months to avoid reservoir effects.

• Further studies should consider researching the use of an integrative care model, such as that described in the SMaRT oncology trials,18–20 for all people with advanced cancer.

• The mechanism of change in the treatment of people who are widowed, divorced or separated and who have advanced cancer and depression needs to be evaluated to see whether specific or non-specific effects bring about a change with CBT.

• We support the NHS Health Research Authority aims of making it easier for research studies to be set up and the elimination of duplicate application. We would recommend a target for approvals to be processed within 2 months.

1. Suicide ideation or intention

Suicide ideation is characterised by thoughts of suicide without clear intention or plan to attempt suicide. Suicide intention is characterised by intentions or plans to attempt suicide. If a participant reports possible suicide ideation or intention they will be assessed for suicide risk using the MINI Section C: Suicidality at screening or the BDI-II during follow-ups.

1. Low to moderate risk:

   1. At home – they will be encouraged to contact the GP if their mood continues to deteriorate. They will be asked for consent to call their GP on their behalf and a family member. If they refuse contact with a GP or family, but are still at high risk to themselves or others, proceed to call GP.

   2. In a health-care setting – ask the patient to speak with their GP/oncologist or see the psychological support services available.

2. High risk (i.e. the participant reports immediate suicide intent):

   1. At home – they will be asked to contact their GP immediately, either themselves or by someone on their behalf. Check if a relative can be contacted. Remain with them until a family member or the GP arrives. Follow up with the research site to inform them of event (see lone worker policy). When the participant refuses help and risk is deemed immediately life-threatening then the police should be notified.

   2. In a health-care setting – they will be encouraged to contact their GP immediately. Remain with them until a family member or the GP arrives. If in the patient’s home, inform research site of event (see lone worker policy). When the participant refuses help and risk is deemed immediately life-threatening then the police may also be notified.

2. Self-harm

1. If participants exhibit any signs of self-harm, they will be encouraged to contact their GP that day. In the event that they refuse to seek help and their safety is a concern, their GP will be notified.

2. When the participant refuses help and risk is deemed immediately life-threatening then the police should also be notified.