Limited antidepressant efficacy in depression in dementia,in the context of limited evidence …

We comment on the contribution of Banerjee et al. (2011) to the limited literature concerning antidepressant efficacy in elderly populations with dementia. Together with a Cochrane review (Bains et al., 2002) and a 2011 systematic review of studies that enrolled 330 patients (Nelson and Devanand, 2011), it appears the evidence base for antidepressant efficacy in these groups is equivocal. The potential importance of the Banerjee study lies in the influence exerted by its size, with 326 patients enrolled, making this the largest single study to examine the question of antidepressant efficacy in this group of patients. However, as we and others have argued (Brodaty, 2011), the study has significant limitations impacting on its findings. We believe evidence of limited antidepressant efficacy, in the context of limited evidence itself, prompts consideration of how clinicians should approach treatment of depression in those suffering from dementia – as well as having potential health policy implications.

The Banerjee study

A total of 664 patients were screened (Banerjee et al., 2011), drawn from nine specialist older persons’ mental health services across England. Of these, 326 were enrolled and were randomly allocated to receive placebo (n = 111), sertraline (n = 107) or mirtazapine (n = 108). The primary outcome measure used was the Cornell Scale for Depression in Dementia (CSDD) (Alexopoulos et al., 1988). A cut-off score of 8 or more on the 38-point CSDD was used as an inclusion criterion. Respectively, 95, 78 and 85 participants remained in the study at 13 weeks. Of those assessed at week 39 (82, 68 and 76 in each group) over half (49, 41 and 41 in each group) had been either withdrawn from treatment or from the trial between weeks 13 and 39.

The average Mini-Mental State Examination (MMSE) across all groups was 18.1, though no MMSE data were available for 23% of the sample. Only 15.3% of the sample (n = 50) still lived at home, the remainder residing in residential care.

There were no significant differences in outcome at either 13 or 39 weeks between groups. All groups demonstrated average improvements on CSDD of between 3.9 and 5.6 points.

Limitations of the study

There are a number of limitations of the study. The authors report that their sample was drawn from nine old-age psychiatry services in England, yet state that ‘… because participants were not drawn from specialist research clinics or tertiary care, but from nine geographically diverse areas with a large number of clinicians representative of services in general, the external validity of our results will be strong’. The majority of patients treated for depression, however, are managed in primary care (Goldberg, 1995; Schurman et al., 1985), so the subgroup referred for management by older persons’ mental health teams are likely to have not responded to treatments initiated by their general practitioners. With a prevalence rate for depression of more than 20% in those with dementia (Banerjee et al., 2011), and with the total number of patients with dementia in England in 2010 estimated to be in excess of 634,000 (Alzheimer’s Society (UK), 2012), this translates to a depression ‘caseness’ within this group of 126,800, clearly not all of whom are being managed by older persons’ mental health teams.

The study limited recruitment to those suffering with Alzheimer’s dementia as defined by NINCDS-ARDRA criteria. This is valid in order to recruit a diagnostically ‘pure’ patient population. However, caution must be applied when interpreting the results of this study in relation to the treatment of depression that might occur in non-Alzheimer’s dementia, estimated at 38% of all cases (Alzheimer’s Society (UK), 2012).

Although target doses for sertraline and mirtazapine were 150 mg and 45 mg daily, respectively, the mean dose achieved for the sertraline group was only 70 mg (rising to 95 mg for those who completed the study), whilst for the mirtazapine group the mean dose achieved was only 24 mg (rising to 30 mg for those 35 patients who remained on mirtazapine by week 39 and thus completed the study). Both active treatment groups could thus be considered to have been under-treated.

The low doses achieved during the trial become of greater relevance when considering the duration of untreated depression in the treatment groups. Whilst patients were eligible to participate if they had been depressed for 4 or more weeks, 67.8% of those recruited (n = 221) had suffered depression for more than 6 months. This raises the question whether a group that was treatment-resistant had been selected, and whether these patients might better have been described as suffering a dysthymic disorder rather than a major depression. That dysthymic disorder might have been the diagnosis might explain the relatively low average CSDD score on trial entry (12.97 across all groups). The CSDD does not validly discriminate between major depressive disorder and dysthymic disorder (nor other depressive disorders), with a score of 8 or more merely suggesting ‘significant depressive symptoms’ (Alexopoulos et al., 1988). The CSDD, like the Geriatric Depression Scale (GDS) (Yesavage et al., 1982–1983), is a screening instrument, yet an arbitrary score of 8 or more on a screening instrument was considered adequate evidence of depression in this study. The study’s authors acknowledge that ‘very few people with CSDD scores less that 8 would have clinically significant depression’, yet 43.7% of their study subjects only had scores within the range of 8–11.

The implication of the potential inclusion of a large number of participants with dysthymic disorder in this study is that this group of patients can be expected to require treatment with higher doses of antidepressants and for much longer periods before response could be expected (BMJ Best Practice, 2011; Keller et al., 2000).

The screening use of the CSDD in demented populations requires caution. The Neuropsychiatric Inventory (NPI) (Cummings et al., 1994) was also administered during the course of the study. The items on the NPI are descriptive and non-aetiological, and include such items as dysphoria, anxiety, aggression, apathy, and irritability. Amongst the domains canvassed in the CSDD are items including anxiety, sadness (‘sad expression, sad voice, tearfulness’, perhaps synonymous with dysphoria in the CSDD), lack of reactivity to pleasant events, lack of energy (both of these readily confused with CSDD ‘apathy?’), irritability and agitation (‘restlessness, handwringing, hairpulling’, covered under ‘aberrant motor behaviour’ in the NPI). The CSDD may therefore ‘count’ symptoms as indicating depression which may instead represent non-specific Behavioural and Psychological Symptoms of Dementia (BPSD), and accordingly are less likely to respond to antidepressant medication. The lower the cut-off used on the CSDD for a diagnosis of depression, the more likely this misattribution error is to occur.

Since all three treatment groups in the study were observed to improve, it is possible that some BPSD were classified and treated as depression. BPSD are known to be transient, with a high rate of spontaneous remission. Treatment for BPSD involves a multidisciplinary behavioural approach (‘normal care’ within an older persons’ mental health service). Outcomes may have been confounded by the prescription of other medications used to treat BPSD whilst the trial was in progress. These, of course, include SSRI antidepressant drugs such as those used as treatment interventions in this study (Pollock et al., 2007).

Significance of improvement in all groups

The latest estimates of 1-year remission rates for depressive symptoms in Alzheimer’s disease are in the order of 35% (Devanand et al., 1997; Fritze et al., 2011; Garre-Olmo et al., 2003); 67.8% of study participants had depression for 6 months or more. That all groups in the Banerjee study showed marked improvement may be an artefact of the relatively low severity of depression of many participants and a high spontaneous remission rate for reasons noted above. An alternative is that the ‘normal care’ provided by specialist mental health services was in itself an effective intervention, as Banerjee himself noted in September 2011 at the International Psychogeriatric Association Congress in The Hague.

Implications for practice

The principal implication for clinical practice of the Banerjee study is that clinicians should expect to achieve improvement in the depressive symptoms of people with dementia. Inaction due to therapeutic nihilism cannot be justified. Choice of treatment should be based upon comprehensive assessment of each individual, his or her history, mental state and risk. Clinicians need to attend more to the differentiation between major depression, dysthymia and BPSD in patients with dementia.

There is building evidence for equivocal antidepressant efficacy in this population, and clinicians must act on the evidence available until it is improved.

The recommendations of Banerjee et al. (2011) that patients with Alzheimer’s disease and depression referred to old age psychiatry services should receive a form of ‘stepped care’ starting with psychosocial interventions of increasing complexity, then consideration of antidepressants if there is non-response after 3 months, is appropriate. However, suicide risk precluded study participation and there is no guidance provided regarding the majority of patients managed in primary care. Therefore, we suggest consideration of possible indicators for an early antidepressant trial, including a past history of antidepressant response, suicide risk, and persistent high-distress levels. Antidepressant initiation must also be accompanied by a clear plan to monitor efficacy and side effects. A similar approach within primary care is also reasonable. There is a significant caveat that should guide practice in the treatment of depression at any age: the outcome of no effective treatment is an unacceptable likelihood of persistent depression.

Implications for policy makers

Conclusions about the use of antidepressants in the setting of dementia should conform to that of Banerjee et al. (2011), leavened by consideration of the limitations of that evidence, and exhortation to improve the evidence base.

The Banerjee study highlights the importance of individuals with dementia and depression being able to access high-quality ‘normal’ care. All healthcare practitioners providing such care must be trained and equipped to routinely provide specific non-pharmacological interventions. They must also have ready access to skilled specialist teams where non-response or illness severity indicates the need for involvement. Access to such training, resources and teams is highly variable in Australia. This must change.

If antidepressants alone are not the answer for the treatment of depression for many people with dementia, it is essential to continue to fund research to discover what is effective, and provide the training and resources needed to provide the best care available now.