Included Studies (See Table 1)

The 10 randomized controlled trials (See References) represented a total of 1125 participants of which four studies (Burgener 2008, Lichtenberg 2005, Samus 2014 and Stanley 2013)25,30,32,33 were pilot studies (N = 398). The intervention arm of included studies had a total of 527 participants, and the control arm had 598 participants. Sample sizes ranged from 20 to 303 across the ten studies: four studies had <50 participants at baseline, one had <100, three had between 100-200 participants and two had > 200 participants at baseline. Eight studies were carried out in the United States, one in the United Kingdom and one in Germany. Apart from two studies (Chapman 2007 and Lichtenberg 2005)27,30 developed in residential care settings, eight studies were conducted in community settings. With the exception of (Lichtenberg 2005)30, the other studies described detailed demography of participants. In these nine studies, participants had their mean ages ranged from 72.4 to 88, and 57% were female. Four depression outcome measures were utilized in the included studies: Cornell Scale for Depression - CSDD (n=6), Geriatric Depression Scale – GDS 15 (15 item version) (n=2), GDS 30 (30 item version) (n=2), and Montgomery Asberg Depression Rating Scale - MADRS (n=1). The (Lichtenberg 2005)30study used both CSDD and GDS 15 and only the data for CSDD was used for the purposes of this review. With regard to participant severity of depression (Kiosses, 2015; Lichtenberg, 2005)28,30mean depression scores of participants at baseline assessment were MADRS score 21.08 (SD 3.74) and CSDD score 13.1 (SD 7.0) indicating that participants in these two studies were depressed at baseline. In the other eight studies, mean depression scores at baseline varied with the majority being mild cases. The severity of dementia at baseline assessment for the 10 studies were mild (n=1), moderate (n=1), mild to moderate (n=3), mild to severe (n=1), and moderate to severe dementia (n=4).

Characteristics of Interventions and Control Conditions (See Table 1)

Out of 10 studies, four studies (Kiosses 2015, Lichtenberg 2005, Marriott 2000 and Stanley 2013)28,30,31,33 applied problem adaptation therapy, involving in pleasant activities, caregiver coping skills training, and cognitive-behavior therapy (CBT) for depression and anxiety respectively. In the other six studies BM was part of multimodal intervention (See Table 1). The duration of intervention ranged from eight weeks to 18 months in which groups received between five to 80 sessions of BM. Interventions were targeted on depression coupled with one or more symptoms of agitation, pain, functional ability and disturbing behavior in three studies (Chapman 2007, Kiosses 2015, Lichtenberg 2005)27,28,30 while the remaining seven studies focused on mood and other outcomes, e.g., anxiety, cognitive function and disability. Usual care control conditions were described as attention control (n=2), usual care (n=3), augmented usual care (n=2), standard or routine medical care for dementia (n = 2), and no intervention control (n=1).

Quality and Risks of Bias in the Included Studies (See Table 2)

All studies scored 6 and above in the Pedro Scale, and were judged as having a low risk of bias as well as using randomization designs. Except for (Lichtenberg 2005)30, all studies stated specific eligibility criteria for participants. Few studies (n=2) described allocation concealment. Regarding blinding, only one study (Callahan 2006)26 described for therapist and subject blinding; but most studies (n=8) blinded the assessors. Six studies had more than 85% follow-up of at least one key outcome. Concerning the attrition bias, seven studies described that they made intention to treat analyzes (ITT) while two other studies (Chapman 2007 and Lichtenberg 2005)27,30 had all of their participants involved in the post-intervention assessment and, therefore, did not need to perform ITT. All studies described intervention protocols, guidelines or manuals for treatment fidelity and adherence purposes, nevertheless other factors affecting outcomes, such as differences among therapists, were not controlled.

Efficacy of Behavior Management on Depression (See Table 3, Figure 2 and Figure 3)

After pooling of treatment effect data, the quantitative analysis (meta-analysis) yielded the following results. In subgroup analysis, the four studies that used BM as an independent intervention showed no significant treatment effect on depression (SMD -0.20; 95% CI -0.96 to 0.56; N = 154). Sensitivity analysis was undertaken due to considerable heterogeneity (p = 0.00; I² =80%) with the removal of an outlying study (Kiosses 2015)28, no heterogeneity resulted (p = 0.93; I² = 0) (SMD 0.16; 95% CI -0.28 to 0.60).

In the other six studies where behavior treatment was a component of a multimodal intervention, the analysis showed a trend favored the intervention, but it was not significant (SMD -0.12; 95 % CI -0.25 to 0.01; N = 971). There was no heterogeneity among these multimodal intervention studies (p = 0.79; I²= 0%). Only one study (Kiosses 2015)28used problem adaptation behavior therapy which showed a significant positive result (SMD -1.13; 95% CI -1.62 to -0.64).

The information from this meta-analysis suggested that BM lacked efficacy in reducing depression in dementia patients. However, ( Kiosses 2015)28 found an effect for BM for participants with mild to moderate dementia in a community setting. It is possible that BM has some efficacy for people with milder severity of dementia compared to those with advanced symptoms. The intervention used was problem adaptation therapy involving both patients and importantly caregivers. Multi-modal interventions modify the patient environment which may assist mood without requiring a high level of individual understanding and engagement with BM. The four studies in which BM was provided as a focused intervention yielded no significant treatment effect of the intervention on depression. (see Table 1) Interventions were exclusively focused on depression in only two studies (Kiosses 2015 and Lichtenberg 2005)28,30 where participants demonstrated higher mean depression scores at baseline assessment. The (Lichtenberg 2005)30 pilot study was undertaken in a residential care setting where participants had moderate to severe dementia. The intervention aimed to increase involvement in pleasant events and demonstrated no significant results. For the remaining two studies in this group of which one was a pilot, targeted symptoms in addition to depression and also was non-significant. However, there is a risk that pilot studies might not indicated a true effect for BM because of small sample sizes and might better represent the features of the specific participants studied.

There remains a question as to how depressed participants actually were in these studies. Baseline depression assessments in eight studies (92% of the total sample population) showed mean depression scores within mild to subclinical ranges, and in a few studies, e.g., Chapman 2007)27, baseline mean depression scores lay at the bottom of the respective rating scales. It might be the case the participants were not depressed enough to demonstrate significant change over the duration of BM. Depression research suggest higher symptoms score are more likely to significantly reduce as a result of intervention compared to low scores which result in a floor effect. It is also questionable if the depression rating scales used in these studies would detect atypical presentations of depression in the advanced dementia patient groups. In detecting depression in Alzheimer s disease, Muller-Thomsen, Arlt, Mann, Maß and Ganzer (2005)10 recommended the use of CSDD and MADRS, and stated the GDS was not an adequate for the detection of depression in this population because of sensitivity issues.

A considerable proportion of participants had severe dementia. The presence of declining language functions and increasing resistance to care in advanced dementia patients could limit adherence and cooperativeness with care, which might consequently reduce treatment effect. The prevalence of resistance to care might increase eightfold in patients with the most severe stage of dementia in comparison with patients with mild dementia (Volicer, Bass, & Luther, 2007)21. Due to the limited amount of published data that met the inclusion criteria a degree of heterogeneity between studies was tolerated. The limitation was related to the focus of the interventions with two studies targeted for depression only and one study of the multimodal group.

While meeting the criteria for BM, the interventions reviewed were based on different theoretical models, for example, behavior programming model in (Lichtenberg 2005)30, stress vulnerability family coping skills model in Mariott 2000, and unmet need model in Samus 2014;)32 it was delivered in different settings (e.g., family care and residential care). This heterogeneity was unavoidable due to the limited amount of literature available for this review. As a result it might have affected the comparability of outcomes. However, what appears reasonably robust in terms of findings is that the different approaches did not yield significant differences in effect which demonstrated them all as equally non-effective.

Another caveat could be the use of variable control conditions. Participants in the control groups were receiving some forms of usual service from their respective dementia care facilities. The intensities of treatment and services they received were varied due to a diversity of the control condition (from no intervention control to augmented usual care) which might have influenced the results of treatment.

The limitations of the literature found for this review reflects a disparity between the amount of research in the area of psychosocial interventions for depression in dementia patients, and the epidemiological importance of the problem. The Teri et al. (1997)20 study resulted in both the SIGN (Network SI, 2006)16and National Collaborating Centre for Mental Health - NICE-SCIE guidelines (NICE-SCIE D, 2006)11 suggested that BM may reduce depression in dementia patients living in the community with caregiver however there was a lack of empirical evidence for different stages of dementia. The findings of this review found no supporting data for the Teri et al. (1997)20 proposition.