Individual and group psychotherapy with people diagnosed with dementia: a systematic review of the literature

Psychotherapy provides a means of helping participants to resolve emotional threats and play an active role in their lives. Consequently, psychotherapy is increasingly used within dementia care. This paper reviews the existing evidence base for individual and group psychotherapy with people affected by dementia.


Introduction
The emphasis within psychotherapy on helping people to resolve emotional threats, to take greater control over their lives and to adjust to illness means that psychotherapy has potentially has much to offer within dementia care. Psychotherapeutic approaches, for instance, may be one way to address the powerful emotional responses to dementia (Connell et al., 2004;Aminzadeh et al., 2007) and the desire of most people to know about their illness (Ouimet et al., 2004;Elson, 2006). There are, however, many challenges in using psychotherapy for this client group: not only is there the impact of the neurological impairment, but the emotional weight of a diagnosis and the residual social difficulties in talking about dementia can all make it difficult for clinicians to find ways to engage meaningfully with people affected by dementia. Almost twenty years ago, Cheston (1998) provided a narrative review of the psychotherapy and dementia care. Although the review identified examples of the main domains of psychotherapy, the empirical literature was limited, Studies involving people with Alzheimer's disease, vascular dementia, Lewy-body dementia or a mixed condition were all included. We excluded studies which focussed exclusively on people with mild cognitive impairment or people with rarer forms of dementia (i.e. frontal-temporal dementia, Human Immunodeficiency Virus, Creutzfeldt-Jakob Disease, Huntington's Disease, Parkinson's Disease and Down's Syndrome) as our clinical experience is that there are often subtle, but important differences between these populations, for instance in the nature of the psychological challenge that they face.

Language
This review was restricted to publications written in English.

Intervention/exposure
We reviewed group or individual psychotherapeutic interventions for people with dementia that meet the definition provided by the British Association of Counselling and Psychotherapy (BACP). Thus, in order for psychotherapeutic interventions to be included, the intervention must: focus on "talking about life events, feelings, emotions, relationships, ways of thinking and patterns of behaviour"; occur regularly at specific times and within a specific context and aim to help individuals to understand themselves and their illness, to promote effective change of thinking or behaviour or otherwise to enhance the person's wellbeing. Consequently, we excluded Art and Music therapy (as these did not focus primarily on talking) as well as Cognitive Stimulation Therapy, Cognitive Rehabilitation, Life Review and Reminiscence Therapy (as these interventions do not meet the BACP criteria of explicitly aiming to change thinking or behaviour). Family or couples therapies were also excluded as we wished to focus on change at the individual level (see Benbow and Sharman (2014) for a recent review of this literature). Similarly, the literature on support groups for people with dementia has also been reviewed recently by Toms et al. (2015) and by Leung et al. (2015).

Outcomes and comparators
In order to increase the range of studies that we included, we did not specify either outcomes or comparators.

Study types
We included randomised and non-randomised controlled trials, as well as studies using repeated measured designs (i.e. non-controlled studies) as these are the most robust methodologies for the research question. Those papers reporting case studies, cross-sectional questionnaire studies or qualitative studies were excluded and will be reported on elsewhere.
"Alzheimer's Disease" ") AND ("psychotherapy" OR "counselling" OR "cognitive therapy" OR "validation therapy" OR "support groups" OR "peer support") NOT ("cognitive stimulation" OR "rehabilitation"). We gathered additional papers by searching the grey literature (including SIGLE and Zetoc), by crosschecking against the reference lists of studies that we had already identified and from studies already known to RC. Study selection followed the PRISMA guideline for reporting flow of information in systematic reviews of literature (Moher et al., 2009). AI screened articles first by reading titles, before checking abstracts for eligibility (and, where this was still not clear, then by reading the full text). RC read 10% of these abstracts as a validity check, with disagreements resolved through discussion. See Figure 1 for more details of this process Time period We limited the review to those studies that appeared after Cheston's (1998) review, i.e. which were published between 1 January 1997 and 31 March 2015.

Data extraction
Following the TIDieR guideline for reporting therapeutic interventions (Hoffmann et al., 2014), a data collection form was developed to extract data. This contained a series of broad domains (e.g. therapy type, aims, mode of delivery, number and duration of sessions) and was pilot tested on a random selection of 10 studies prior to conducting the full review. AI initially entered data onto the form, and all entries were then checked by RC, with disagreements resolved through discussion.

Risk of bias
The risk of bias tool for randomised and nonrandomised controlled trials was adapted from the PEDro-P Scale for randomised and non-randomised controlled trials (Murray et al., 2013), with two additional items added: "Was the allocation sequence adequately generated?" was taken from the Cochrane Collaboration's tool (Higgins et al., 2011); and "The therapy across the intervention was standardised (e.g. through training, supervision and use of manuals)" was specifically added for this review. For repeated measures studies, we followed the procedure adopted by Toms et al. (2015) and rated studies in terms of the relevant 28 CONSORT items (Moher et al., 2010).

Results
The database search yielded 1596 citations, with an additional 93 reports identified through other means. After removing duplicates, 1397 papers were screened. The flow of records through the review is set out in Figure 1.

Synthesis of results
In all, 26 papers were identified. We followed the procedure adopted by Toms et al. (2015) and have categorised papers according to their study design using the system described by Arbesman and Lieberman (2011): of the 26 papers, 19 articles concerning 16 studies were identified as Level I (RCTs); 2 were Level II (controlled non-randomised studies) and 5 were categorised as Level III (repeated measure designs). Where preliminary or follow-up results were reported on separately, then papers have been brought together and described as a single study. Interventions were categorised in terms of broad therapeutic domains with the main characteristics of the included studies being shown in Table 1.

Risk of bias assessment
Risk of bias assessments were conducted independently by the two authors with disagreements resolved through discussion (see Tables 2 and 3). Potential risks of bias include inadequate blinding of therapists and assessors, and partial reporting of results. The agreement level for Level I and II studies was 80.34% (weighted Kappa = 0.681), and for level III studies it was 78.57% (weighted Kappa = 0.602).
Overall, the majority of level I and II studies had an unclear or high risk of bias in the areas of participant, therapist and assessor blinding. Amongst the 19 Level I studies, 10 papers either did not provide outcome data for 85% or more of participants who were randomised into the study or did not provide enough information to allow reporting on this. Amongst Level III studies, a recurring failure was the absence of appropriate baseline and follow-up data. Four of the five studies only took measures at one point before the group began, whilst two studies (Gaugler et al. (2011) and Putman and Wang (2007)) did not collect follow-up data, making it difficult to determine whether changes in measures during therapy were related to the intervention, or to general trends. Table 4 reports study outcomes.

Cognitive-behaviour therapy (CBT)
A total of six studies assessed a CBT-based therapy for people with dementia. The only Level I CBT study to be adequately powered was the CORDIAL study (Kurz et al., 2012), which evaluated a multi-modal intervention for people with mild levels of cognitive impairment caused by Alzheimer's disease that combined behavioural strategies (e.g. activity planning and day structuring) with Cognitive Rehabilitation, a support group and instructions to carers in the use of validation therapy (VT). Although the primary outcome (i.e. daily functioning) was unchanged, quality of life and depression levels improved for a sub-set of female participants.
Three level I pilot studies incorporated modified forms of CBT. Spector et al. (2015) found strong but non-significant improvement in anxiety and a significant fall in depression levels for individuals with a mild to moderate cognitive impairment and clinically significant levels of anxiety. Their intervention involved working with participant and their carer together, and was delivered by four Clinical Psychologists who were also CBT therapists in 10, weekly sessions. Spector et al. suggested that CBT therapy was cost-neutral with a short-term reduction in health and social care costs being balanced against the cost of the intervention itself. Stanley et al. (2013) reported the effects of the Peaceful Mind intervention originally described by Paukert et al. (2009,  None stated, although CBT group followed guidelines by Teri and Gallagher-Thompson (1991), and the support group that set out by Yale (1995).
Person-centred counselling Phung et al. (2013), Waldorf,  Inclusion criteria: a diagnosis of Alzheimer's disease; mild dementia indicated by CDR of 1; a MMSE score of 15 or above; living in their own home with a carer in regular contact and the ability to communicate verbally.
Intervention: 6 × 50 min individual sessions with a carer involved provided by trained psychotherapist Control: standard care (general advice regarding the diagnosis and treatment of dementia plus outparticipant review) with an option to receive the therapy after end of study Psychotherapy was manualised with treatment fidelity ensured by regular supervision using audiotapes. One session from each individual therapy was rated for adherence to the model. Deponte and Missan (2007). N = 30 (VT = 10, SR = 10, C = 10); Nursing Homes (Italy).

Validation therapy
Inclusion criteria: Nursing Home resident for at least 6 months; diagnosis of dementia; aged over 70 and lack of concomitant psychiatric pathologies Validation Therapy: group therapy 2 times per week for three months; 45 to 60min duration Sensorial Reminiscence: same schedule as for VT by a different therapist, but further details not given Control: further details not given None stated. Toseland et al. (1997) N = 88 (VT = 31/ SC = 29/ UC = 28); 4 nursing homes (USA) Inclusion criteria: a clear diagnosis of dementia; willing to attend groups; at least a moderate level of dementia and displayed "problem behaviours" (e.g. physical aggression, verbally abusive behaviours, disruptive Validation Therapy: Four group sessions of 30 min each week for a total of 52 weeks. Separate therapists for each home received 4 days training, and regular supervision. Social contact: equal length of intervention to VT, Therapists in both VT and SC conditions received weekly phone and monthly in-person supervision. One session each month in both arms was randomly reviewed for threats to treatment integrity. (Continues) Participants were excluded if: they did not have a documented diagnosis of dementia; they could not communicate adequately (indicating too severe a level of impairment); were aged under 60 or anxiety was not a problem for them.
Up to 12 weekly individual sessions for the first 3 months (30 to 60 min), were provided in the participant's home, followed by a brief telephone call. In the second 3 months of treatment, telephone booster sessions occurred weekly for 4 weeks and biweekly for 8 weeks. The intervention provided jointly between a clinician and a "collateral" (a friend or family All sessions were audiotaped, and supervision was provided by experienced clinical psychologists and a social worker. (Continues) Inclusion criteria: a diagnosis of dementia, resident at nursing home, MMSE score of 10-25, GDS score of severe, symptoms interfering with daily functioning, requiring frequent staff intervention following the consensus of treatment team; score of at least 1 and a severity of 5 on the CMAI Group counselling-two group sessions per week for 2 years; 120 min per session. "Naomi Feil's principles of validation were used whenever possible when communicating with participants" (p. 168) None stated. Inclusion criteria: dyads of people with early stage dementia and their partners; an MMSE score above 18. An earlier paper on the same intervention by Zarit (2004) stated that inclusion was based on the person's awareness of their memory loss. Exclusion criteria: not specified by Gaugler. However, Zarit states these as people with primarily psychiatric symptoms (consistent with FD and LBD); people who were unable to recognise changes in themselves because of dementia or if they could not acknowledge memory problems.

Generic group psychotherapy
Between 10 and 13 weekly sessions, each lasting between 90 and 120 min moderated by two facilitators. Sessions involved: joint interaction with both the person with dementia and their carers; separate group sessions and a "wrapping-up" session in which the carer/ person with dementia dyads reunited. Sessions were organized around specific topics related to early-stage dementia and also included expert speakers.
Session topics and order of sessions varied across three sites because of local factors. In order to ensure consistency there were regular telephone conference calls to discuss issues of clinical concern.
Cheston, Jones and Gilliard  ) 2010) with people with mild and moderate levels of dementia. Their modified form of CBT incorporated religious elements and a simplified package of training in skills such as breathing, calming thoughts and sleep hygiene. The authors reported significant improvement in participants' anxiety and quality of life compared to the control group. In the third pilot study, Burgener et al. (2008) combined bi-weekly CBT with Taiji (or Tai Chi) exercises and a support group over 40 weeks for people in the early and mid stages of dementia, suggesting limited improvement in participants' cognitive functioning and self-esteem compared to the control group.

Person-centred counselling
Three level I person-centred studies were identified. The Danish Alzheimer's Disease Intervention Study or DAISY was the most methodologically sophisticated study that was reviewed. Within this study, counselling based on constructivist principles was the central part of a multi-faceted and semi-tailored support programme. This package was offered both to people who had been diagnosed with dementia in the previous year and had mild to moderate levels of cognitive impairment, and to their carers. The primary aim of the intervention was to reduce levels of depression and to improve health-related quality of life in participants affected by dementia at 12 months. To control for the possibility of finding spurious effects from multiple testing, the authors adopted an extremely conservative level of p < 0.0005 for statistical probability. Although, participants' depression levels improved, this did not reach this increased level of significance (Waldorf et al., 2012). A cost utility evaluation of the DAISY intervention found that whilst none of the observed costs of the intervention and control arms were significantly different, there was a tendency for psychosocial care to lead to informal care cost increases (Søgaard et al., 2014). Two other, person-centred studies were identified: both of which involved people affected by severe cognitive impairments who were residents in long-term care facilities Tappen and Williams (2009) described Therapeutic Conversations which "provides the Multi-dimensional approach combining pharmacological treatment, group therapies (music, movement and psychodynamic), sociotherapy as well as individual and family therapy. Participants attend the therapeutic community two to three times per week for a 6-h day. During each attended week, each participant participates in four mixed-gender groups of a maximum 10 participants. "The care team includes two residents in psychiatry, one senior resident, one movement therapist, one music therapist, one psychologist, one social worker and four nurses" (p. 93) None stated.
Cheston, Jones and Gilliard (2003) Weber et al. opportunity to share feelings and concerns with a skilled listener who can understand their attempts to communicate" (p. 270); whilst Hirazakura et al. (2008), reported the use of group emotional therapy which "appealed to the feelings of the patients and elicited emotions" and in which therapists sought to "share in the same emotions" (p 304) as participants. Hirazakura et al. reported increases in cognition after intervention whilst Tappen and Williams found improvements in affect and depression when compared to the control arms. However, both studies had a series of methodological limitations including small numbers of participants and relatively poor standard of reporting.

Psychodynamic interpersonal
Carreira et al. (2008) was the only paper included in the review that compared the impact of psychotherapy on people with and without a cognitive impairment. This study presented a sub-group analysis of a larger RCT (Reynolds et al., 2006) comparing maintenance paroxetine and interpersonal psychotherapy (IPT) in participants aged 70 years of age or older who had depression. Carreira et al. looked at 52 people in the pill placebo arm who had received either monthly maintenance IPT sessions or clinical management (CM). Their analysis suggested that participants with cognitive impairment who received IPT fared significantly better than those who received just CM (relapsing on average after 58 weeks compared to 17 weeks). No differential benefit of IPT over CM was observed for individuals without impairment. The authors suggested that IPT may have helped to resolve interpersonal conflict with caregivers in the cognitively impaired group. In a small trial of 40 people with mild levels of cognitive impairment who were randomised to either receive six, 50-min individual sessions of psychodynamic interpersonal therapy (PIT) or usual care, Burns et al. (2005) did not find any significant differences on their main outcome measures.

Validation therapy
Two level I studies tested VT which incorporates a range of recognised psychotherapy and counselling techniques including empathic listening (Feil, 2003;Neal and Briggs, 2003). Both studies were set in long-term care facilities in which the level of cognitive functioning of participants was relatively low and both compared VT with both a usual care control arm and an active intervention: sensorial reminiscence for Deponte and Missan (2007) and a social contact group for Toseland et al. (1997). Although the adoption of a third treatment arm has potential methodological advantages, both reports are unclear about a number of design issues, including blinding. The results from both studies were inconclusive: Deponte and Missan found decreased behavioural distress in both the VT and the reminiscence arms, whilst the reminiscence arm had also improved cognitive functioning. Toseland et al. found lowered levels of verbal and physical aggression in the VT group at both 3 months Putman and Wang (2007) Gaugler et al.

Other information Funding 25
Overall score 23 23 25 23 27 Quality of life and self-esteem were improved in the intervention group compared to control group, but this feel short of statistical difference after adjusting for baseline differences. The (Continues)    Significant improvement in the psychotherapy group for CSDD (9.5, range 0-14 to 7.00, range 1-11; p = 0.013) and RAID (9. 00, range 0-17 to 7.25, range 2-14; p = 0.05). Changes not significant when baseline differences are accounted for.

Validation therapy
Tondi et al. Person with dementia Behavioural Distress (NPI and BANSS) In the VT group, NPI scores decreased in 23 of 27 participants, stayed steady in 4 and did not increase in any participants. NPI scores in the control group increased in 10 of 23 participants, stayed steady in 9 and decreased in 4. Carer's distress score in VT group reduced from 8.6 to 3.5, no change in the control group.
Level III studies Cognitive-behavioural therapy Carer Distress at partner's anxiety (NPI-A distress question). At 3 months, 86% of participants had a 20% or more reduction of anxiety from baseline, according to the NPI-A. At Anxiety (RAID and HADS) CSDD scores at baseline (7.58,SD 2.19) and at the start of the intervention (8.32, SD 2.19) significantly improved at the end of therapy (6.42, SD 2.04) and were maintained at follow-up (6.37, SD = 3.09).
(Continues) and 12 months, but nursing staff reported greater improvements in levels of aggression in the two control group arms. A level II study (Tondi et al., 2007) compared Nursing home residents with dementia individuals receiving VT and group therapy, and found that that VT participants showed lower levels of behavioural distress and carer distress. A level III study (the Closing Group of Putman and Wang, 2007) also reported results for a VT group which ran twice a week for two years within long-term care facilities, but which did not show any significant changes in outcome measures.

Generic group psychotherapy
Although psychotherapies such as CBT can be delivered within a group as well as an individual format, some therapies rely specifically on the dynamics created by a group in order to function. Two level I, one level II and two level III studies using group therapy were identified, all of which were aimed at people with mild to moderate levels of cognitive impairment. The Early Stage level I Memory Loss Support (ESML) group involved nine, weekly group sessions in which family members attended the first part of the group. Logsdon et al. (2010) randomised 96 participants (with mild or moderate levels of impairment) to ESML and 46 to usual care with preliminary findings being reported by Logsdon et al. (2006). This study was based on an established body of previous research (e.g. Snyder et al., 1995;Yale, 1995;Snyder et al., 2007) and as such was powered to find significant differences. After controlling for baseline differences and changes in cognition, the authors reported significant improvements in quality of life and depression. However, the study only provides pre and postintervention scores, with no follow-up. Marshall et al. (2014) report a level I pilot study of the "Living well with Dementia" (LivDem) intervention, which had similar inclusion criteria, session length and session frequencies to the ESML. However, whilst ESML sessions were provided by three or four trained and experienced facilitators, at least two of whom were master's level professionals, the LivDem therapists were memory clinic staff who had attended a two day training course, but otherwise had little experience of therapy. After adjusting for baseline differences between the two groups, they found a nonsignificant trend for improvements in self-esteem and quality of life in the intervention arm, with an effect size similar to that of Logsdon et al. A level II study (Cheston and Jones, 2009) compared attendance at a therapy group and psycho-educational group for a small number of participants with mild or moderate levels of dementia. Changes in depression were not significant after adjusting for baseline differences. In a level III repeated measures study, Cheston et al. (2003) found significant improvements in levels of depression and anxiety during the intervention compared to a six week baseline period, which were maintained at follow-up. Gaugler et al. (2011) also described a level III repeated measures study of a 10 to 13-week intervention which aimed to develop the coping skills of people living in the community during the early stages of dementia. Carers reported a number of significant improvements in coping.

Multi-component interventions
Three level I and one level III studies reported interventions that described eclectic combinations of different forms of therapeutic work. In a level I study, Bakker et al. (2011) tested the impact of a multidisciplinary 13-week combined group and individual intervention described as Integrative Interactive Rehabilitation (IRR). This involved elements of cognitive and behavioural therapies, counselling and family therapy. The IRR arm of the study comprised 81 participants with mild or moderate levels of dementia, and who have at least three neuropsychiatric symptoms. Compared to the control arm, the intervention arm showed significant reductions in both the number and the severity of psychiatric symptoms, as well as providing significant benefits for carers. Although overall, the study was at a low risk of bias, the validity of results may have been compromised by a failure to blind outcome measures evaluation.
Hilgeman et al. (2014) described a level I study testing an eclectic intervention named Preserving Identity and Planning for Advanced Care or PIPAC. The foursession individual intervention employed a combination of self-adjusting, future planning and selfmaintaining, reminiscence-based work. After controlling for baseline differences, results revealed clinically meaningful differences between intervention and control arms at post-treatment for depressive symptoms, quality of life, health-related quality of life indicators, coping styles and decisional conflict. Jha et al. (2013) reported a level I peri-diagnostic intervention in which participants with suspected dementia were referred to a specialist mental health team and received pre-diagnostic well-being assessment and counselling followed by a diagnostic consultation with written feedback and six monthly home visits for post-diagnostic support. Although well-being was improved compared to usual care, there was no significant change in other outcomes variables. A level III repeated measures study by Weber et al. (2009) found significant improvements in anxiety and apathy after combining pharmacological treatment, group therapies (music, movement and psychodynamic), individual and family therapies with people with dementia referred to a Day Hospital.

Methodological rigour and risk of bias
The standard of reporting of papers was of mixed quality. For example, whilst the DAISY study (Waldorf et al, 2012;Phung et al., 2013) in particular stood out for its methodological rigour, other studies including Hirazakura et al. (2008) and Tondi et al. (2007) had weak designs and were poorly reported. Moreover, the literature is marked by profound variability: differences in aims and outcome measures; in populations and domains of working and in how interventions are delivered and by whom. Because of this heterogeneity it was not possible to carry out a meta-analysis.
The majority of the Level I studies that we have reported on were either pilot studies, preliminary reports or gave no indication that their sample sizes had been based on a power calculation. Only four studies (CORDIAL, DAISY AND ESML and the Integrative Psychotherapeutic Nursing Home Programme reported by Bakker et al., 2011) provided evidence of being adequately powered to find statistical change.

Conclusions
In this review we have attempted to summarise the main findings of individual and group psychotherapy interventions with people with dementia. However, our conclusions are tempered by a number of limitations: we only reviewed papers which reported in English and we thus excluded a range of reports of psychotherapy (e.g. Fabris, 2006;Scheurich et al., 2008;Scheurich and Fellgiebel, 2009). In addition, we focussed on psychotherapy with people with Alzheimer's disease, vascular, Lewy body or mixed dementia. We also excluded both support groups and family therapy, which have both been reviewed recently elsewhere. Finally, whilst we tried to maintain a broad definition of psychotherapy, it is possible that we excluded some interventions which did not meet the BACP definition but which still have psychotherapeutic characteristics. Table 5 summarises the key findings from the studies which we did review: When compared to those studies reported in a review of the same area, 18 years ago (Cheston, 1998), not only did we identify many more studies examining the impact of psychotherapy, but the quality of the design of the studies is much higher. We will now examine the strength of the evidence available.

Quality of evidence
Where participants are in the early stages of dementia, the strongest evidence that we found was from Logsdon et al.'s study demonstrating that a 9-week group intervention delivered by experienced therapists significantly reduced levels of depression and improved quality of life. There was also preliminary evidence supporting the potential of cognitive behavioural or behavioural interventions, although here studies were limited by their relatively small sample sizes, and, for two studies by the inclusion of additional, non-psychotherapeutic interventions. The evidence to support a person-centred counselling approach is also inconclusive. The DAISY study showed that an early psychosocial counselling and support intervention reduced levels of depression, but despite a substantial effect size, their findings did not meet the conservative level for significance which they had set. Although two other level I studies identified significant effects, once again these trials were relatively small.
For people with mild to moderate levels of impairment living in Nursing Homes, then Bakker et al.'s study provides some evidence that individually tailored and eclectic packages of psychotherapy interventions can help to reduce challenging behaviour. Whilst the use of a number of different psychotherapies means that it is not possible to be clear about the impact of individual interventions, in many ways, multi-component psychotherapies are more representative of the eclectic forms of psychotherapies used by many professionals, especially Clinical Psychologists. However, the intensive, multi-disciplinary intervention that was required to achieve this impact is one that many clinical services will struggle to replicate.
There is less evidence to support the use of psychotherapy for people who have more severe impairments. Carreira et al. showed that after monthly maintenance psychotherapy sessions, lower cognitive performance was associated with longer time to recurrence. Two level I studies using person-centred principles within long-term care facilities also found promising results, albeit within the context of underpowered studies. There were also mixed results for the impact of VT with the quality of reporting of these studies also limiting the conclusions that can be drawn.
Even where studies focused on people with relatively mild levels of cognitive impairment, the language, memory and other deficits inherent in the diagnosis of dementia, inevitably meant that adaptations to the usual psychotherapeutic process were made. However, reporting of these adaptations was limited, and there was no consensus about the critical areas for adaptation. Typically, skills based therapies often emphasised behavioural rather than cognitive interventions and took therapy at a slower pace with more repetition of core skills. However, other studies (e.g. Marshall et al) encouraged adaptation to these deficits through discussion and the sharing of experiences.
More generally, the psychotherapeutic literature is marked by heterogeneity. For instance, the experience, training and supervision of therapists varied widely, and relatively few studies provided any evidence that the delivery of the intervention was standardised, for instance through the use of treatment manuals. Whilst using experienced and qualified therapists may increase the likelihood of finding positive benefits from psychotherapy, it is also likely to increase the costs associated with its implementation, and to reduce the accessibility of services for people with dementia. In this respect it is encouraging that one of the only studies to provide a cost-analysis (Spector et al.) also employed highly qualified Clinical Psychologists and concluded that the intervention was cost-neutral.
Despite our exclusion of family and marital therapy from this review, nevertheless many studies actively involved carers, with one study (Stanley et al.) recruiting a friend or family member to act as a cotherapist to provide skills training. However, even when families were not involved to this extent, then there are still important clinical and practical reasons for therapists to work alongside them. What is less clear is how therapy might take into account the quality of marital relationships.
A number of design flaws also limit our ability to interpret these results. First, a statistically significant result is not equivalent to a therapeutically significant impact. Many studies did not report effect sizes, whilst for those that did, the effect sizes were often modest. One exception, was the DAISY trial (Waldorff et al., Phung et al.) which found a effect size of À1.58 for the reduction in levels of depression which the researchers treated as non-significant. Second, the validity of many studies was threatened by the absence of a psychological placebo as a control condition in which non-specific elements of therapy were included. Where the control condition is treatment as usual, then it is not possible to know whether any changes in outcome were related to the intervention rather than to non-specific elements of the therapeutic process. For instance within group therapies, simply convening people who share the same condition into a group may have an impact, whilst in individual therapy clients may value the process of meeting someone who is interested in their life for one hour per week, regardless of the impact of any specific form of therapy.
At the same time, the notion of a psychological placebo is, in itself, challenging: psychotherapy simply cannot be prescribed for someone in the same way as medication can be (Bannister and Fransella, 1980). Thus, the application of the RCT model to psychotherapy tends to favour those interventions which incorporate a discrete set of techniques that can be reliably taught and operationalised, such as cognitive-behaviour therapy, over more complex, longer-lasting psychodynamic interventions (e.g. Sinason, 1992;Davenhill, 2007). Additionally, RCTs do not, on their own address the challenges of defining appropriate outcomes: for instance whether the goal of psychotherapy be to improve insight, to reduce anxiety and depression or to reduce challenging behaviour and carer stress? Importantly, there is a need to develop a more nuanced understanding of those elements of therapy that help people affected by dementia to change or, equally relevantly, not to change. This would help to identify those generic factors that are common across different forms of therapy-such as how people manage shame and stigma, or how fears of a loss of internal control are contained (Cheston, 2015). Thus, arguably, the most significant impact of psychotherapy in work with people affected by dementia resides not in doing therapy, but in helping everyone involved in dementia care to be more therapeutic.

Key points
• Psychotherapy is increasingly used to help people affected by dementia to adjust to their illness. However, the evidence base for this is limited and uncertain. • This review screened 1397 papers evaluating the impact of group or individual psychotherapy with people affected by dementia published in English between 1997 and 2015, with 26 papers being included in this review. • Four trials were adequately powered to find statistical change. Of these, one study provided evidence that post diagnostic group therapy improved quality of life and reduced depression whilst a second suggested that an intensive, multi-faceted intervention that included psychotherapeutic elements lessened distress for Nursing Home residents. • Currently, the evidence base for psychotherapy with people affected by dementia is limited. If the promise of this clinical intervention is to be realised, then it is important to identify the change processes that lead to successful outcomes.