Guidelines for dementia or Parkinson’s disease with depression or anxiety: a systematic review

Background Depression and anxiety remain under-diagnosed and under-treated in those with neurologic diseases such as dementia or Parkinson’s Disease (PD). Our objectives were to first, to provide a synthesis of high quality guidelines available for the identification and management of depression or anxiety in those with dementia or PD. Second, to identify areas for improvement for future guidelines. Methods We searched MEDLINE, PsycINFO, and EMBASE (2009 to July 24, 2015), grey literature (83 sources; July 24-Sept 6, 2015), and bibliographies of included studies. Included studies were evaluated for quality by four independent reviewers the AGREE II tool. Guideline characteristics, statements and recommendations relevant to depression or anxiety for dementia and PD were then extracted. (PROSPERO CRD: 42016014584) Results 8121 citations were reviewed with 31 full text articles included for assessment with the AGREE II tool. 17 were of sufficient quality for inclusion. Mean overall quality scores were between 4.25 to 6.5. Domain scores were lowest in the areas of stakeholder involvement, applicability, and editorial independence. Recommendations for the screening and diagnosis of depression were found for PD and dementia. There was little evidence to guide diagnosis or management of anxiety. Non-pharmacologic therapies were recommended for dementia patients. Most advocated pharmacologic treatment for depression, for both PD and dementia, but did not specify an agent due to lack of evidence. Conclusions The available recent high quality guidelines outline several recommendations for the management of comorbid depression or anxiety in PD or dementia. However there remain significant gaps in the evidence. Electronic supplementary material The online version of this article (doi:10.1186/s12883-016-0754-5) contains supplementary material, which is available to authorized users.

Despite awareness of these comorbidities, depression and anxiety remain under-diagnosed and under-treated in those with neurologic diseases [1,3,[13][14][15][16][17]. Only 20% of PD patients diagnosed with depression receive therapy [18]. This represents a significant knowledge-topractice gap. One way to address this is through the use of Clinical Practice Guidelines (CPGs) [19]. CPGs synthesize available evidence based on a systematic review of the literature, clinical expertise and patient preferences [19]. CPGs are targeted at practitioners who apply the recommendations to clinical decision-making and reduce disparities in care [19][20][21][22].
Thus, in the setting of PD and dementia, CPGs should enable the appropriate management of depression and anxiety [23][24][25][26]. Despite available CPGs, these disorders remain under-managed, suggesting these CPGs are underused or lack sufficient recommendations [26][27][28]. Multiple available guidelines of varied quality leads to uncertainty as to which CPGs should be used in practice. Our primary aim is to synthesize the high-quality evidence-based CPGs available for diagnosis, and management of depression or anxiety in those with dementia or PD. We chose to summarize and evaluate guidelines as the majority of physicians will use CPGs as a tool to review evidence and inform practice. Secondarily we aim to, identify areas gaps within the existing guidelines to inform future guideline development. This provides a broad over view of evidence in the area and identifies areas for further study and development.

Methods
The study protocol follows the recommendations provided by the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA)-Protocols Statement [29] and guidelines and the protocol was registered with PROSPERO [30] (CRD: 42016014584).

Search strategy
The literature search was developed in conjunction with an experienced librarian (DL) and was verified independently by a second librarian (HLR), using the Peer Review of Electronic Search Strategies (PRESS) methodology [31]. Any recommendations were incorporated into the final search.
Databases included MEDLINE, EMBASE, and Psy-cINFO. Clusters of terms (controlled vocabulary and key words) were used to search each database; these include dementia, Parkinson's disease, depression, anxiety and CPGs (Additional file 1: Box S1). The search was completed by cluster, first searching the terms in each cluster (combined with the Boolean operator 'OR') and keyword searches of abstracts and titles. The clusters were then combined with ' AND'. We searched for several pathological variants of dementia including Alzheimer's disease, vascular, frontotemporal, Lewy Body disease, Huntington's Disease, CADASIL, primary progressive aphasia, and Creutzfeldt Jakob (Additional file 1: Box S1). We included relevant derivatives of terms or broad key words related to depressive or anxiety disorders (Additional file 1: Box S1).
This was augmented by a search of the grey literature (Additional file 2: Table S1). This search was limited from 2009 to search date, such that we would only capture CPGs developed within the past 5 years; given the evidence that CPGs may become out of date after only 3 years [32]. All languages were included in this search.

Selection & eligibility
All citations were reviewed for eligibility by two independent authors; citations meeting initial eligibility criteria were included in full text review. If there was disagreement at the abstract stage, the full article was pulled for review. Bibliographies for all included articles were searched. If multiple CPGs were identified from a single agency on the same topic the most recent was used.
At the first stage of abstract review, any article that represented a guideline for PD or dementia was included. Eligibility at the full text stage required that the CPGs included at least one recommendation related to depression and/or anxiety in patients with PD and/or dementia. The kappa statistic was used to quantify interrater reliability.
For non-English articles that met eligibility at the full text stage, the language was determined using online translation software. Citations were translated using the online (Google translate) function to determine if an article was a guideline. When included, the documents were searched using translated relevant terms; for example, if a guideline pertained to PD in the abstract, the text was searched for depression or anxiety (and all translated synonyms). If those criteria were met, the full guideline was translated and reviewed.

Assessment of quality
The Appraisal of Guidelines Research & Evaluation (AGREE II) tool was used to assess guideline quality [33]. This tool was designed to evaluate guideline quality and to aid in guideline development and reporting [33]. The tool includes 6 domains covering scope and purpose, stakeholder involvement, rigour of development, clarity of presentation, applicability and editorial independence [33]. Within each domain there are between 2 to 8 questions, to a total of 23 [33]. Each item is rated from 1 (not included or very poorly reported) to 7 (exceptional reporting of all criteria outlined in the AGREE II Manual) [33].
Each domain was scored independently by four reviewers, along with the assignment of an overall score. An initial assessment of 5 citations was done and compared across all 4 reviewers [33]. The 4 reviewers met to discuss discrepancies and address questions about rating, before the remainder of the guidelines were reviewed and scored. This also served to ensure that all raters were aligned in their understanding of the AGREE II items. Any further discrepancies were resolved by discussion.
Domain scores pooled across the 4 assessors were calculated, as outlined in the AGREE II user manual [33]. The higher score indicates a higher quality across rated items. It has been demonstrated that the quality across the AGREE II domains predicts guideline implementation [33]. The mean overall quality scores with standard deviations (SD) were calculated, as well as for each domain item. CPGs with a mean overall quality score 5 or greater were assigned at least moderate quality and included in further analysis. CPGs with a score below 3 were excluded due to low quality. A score less than 5 but greater than 3 were re-evaluated and inclusion status was decided by consensus.

Data extraction & synthesis of evidence
Guideline characteristics were extracted by one author (ZG) and independently verified by a second author (BM). Items extracted included the primary conditions covered, region/organizations, number of committee members, numbers of references, and sources of funding.
Two independent reviewers then extracted relevant recommendations (ZG, BM). Specifically, guidelines were searched for any mention of relevant recommendations and supporting text or statements. Three authors reviewed the extracted recommendations (ZG, BM and JHL). Recommendations were compiled across the guidelines into relevant categories and subcategories, and reported using descriptive statistics including the quality, number of guidelines supporting the statement and subpopulations included. As the evidence in the guidelines is represented by practice recommendations, it was not amenable to meta-analysis. The main output of this systematic review was an appraisal of the quality of all guidelines pertaining to comorbid depression or anxiety in PD or dementia, and a synthesis of the recommendations across the different guidelines. Data were analyzed using STATA 13.1 (Stata Corp. College Station, TX).

Study selection
The database search generated 4441 citations after duplicates were removed, with a further 3681 citations identified from the grey literature (Fig. 1). When screened for eligibility, 360 citations met criteria for full text review (κ = 0.88, 95.7% agreement). At this stage most articles were excluded because they were not relevant (n = 218), were not guidelines, or were unrelated guidelines. Other common reasons for exclusion at the full text stage were being out of the date range (n = 33) or a duplicate (n = 35). Excluded citations also included 26 mental health guidelines that did not address PD or dementia. Similarly there were 5 PD and 9 dementia guidelines that did not address depression or anxiety. The dementia guidelines primarily pertained to Alzheimer's disease, vascular dementia, general dementia care and one referred to Lewy Body Disease. Of these articles, 4 were identified to be summary documents of included guidelines and were used as supplemental material to these included guidelines. Twenty-six CPGs met all eligibility criteria and were evaluated using the AGREE II tool, of which 17 met the quality cut off for inclusion.

Study quality
These 26 CPGs were assessed for quality using all 23 items across the 6 domains of the AGREE II tool. Nine guidelines were excluded for low quality. Six were excluded with an overall mean rating ranging from 2.25 to 3.75. Three had ratings of 4-4.5, where decision to exclude was by consensus. A low rating was typically due to unclear methods; thus scoring low on rigour of development, applicability and editorial independence. Authors of guidelines were contacted for more information in the case that an item was unclear and responses were incorporated in the quality assessment.
The 17 included guidelines had mean overall scores from 4 to 6.5 (Table 2). When examining the individual domain scores, the highest rated domain was Domain 4: Clarity of Presentation (mean score 77.0; SD 11.4). This was followed by Domain 1: Scope and Purpose (mean score 72.1; SD 12.1). Domain 5: Applicability was the lowest rated domain (mean score 41.5; SD 22.6). Stakeholder involvement (Domain 2) also had a low score (mean score 54.5; SD 23.3).
The mean rating across each question in the domain scores were also examined to explore differences between domains (Additional file 4: Table S2). Question one pertaining to the overall objectives was the highest rated item at 5.88 (SD 0.61), followed by link between evidence and recommendations at 5.78 (SD 0.51). The lowest rated item was providing a procedure for updating the guideline is provided, with a mean rating of 3.16 (SD 1.73). The views and preferences of the target population have been sought was also rated poorly with a mean score of 3.25 (SD 1.92). All items in Domain 5 had low mean scores, ranging between 3.27 (SD 1.46) for resource implications and 3.72 (SD 1.53) for advice on putting recommendations into practice.

Guideline recommendations
The details of extracted recommendations are summarized in the Table 3 for PD and Table 4 for dementia. 21 categories of recommendations were extracted in total.

Parkinson's disease recommendations
Only two guidelines discussed anxiety in those with PD [34,35]. These stated there was little evidence for either the diagnosis or treatment of anxiety in PD, and that there was insufficient evidence for the treatment of anxiety with levodopa [34,35].
There were clear recommendations surrounding the diagnosis of depression in PD [34,37,38]. Clinicians should have a low threshold for the diagnosis of depression in PD given the difficulties making a diagnosis [34]. Use of a validated tool for detecting depression (or neuropsychiatric symptoms) was advocated by two guidelines, with varying levels of recommendations [37,38]. Tools that were recommended include the HDRS, the MADRS or the UPDRS-Part 1 Non-Motor, among others [37,38]. The diagnosis should be made based on a clinical interview and not based on the tool alone and should seek collateral information from carers [37].
Antidepressant therapy is recommended, however there is little evidence to support one agent over another (n = 2) [37,39]. Additionally, the choice of an agent must be individualized (n = 1) and the practitioner should consider side effects and drug interactions prior to initiation [34]. There have been prior studies on the tricyclic antidepressants (TCAs), specifically amitriptyline, and although they were beneficial for mood, this was offset by the side effects (n = 3) [34,37,39]. One guideline noted that selective serotonin reuptake inhibitors (SSRIs) showed some benefit in uncontrolled studies [39,40], but noted that the SSRIs could worsen PD symptoms of restless legs (RLS), periodic limb movement (PLM) and  REM sleep behaviour disorder (RBD) (n = 2) [39,40]. It is recommended to avoid amoxapine and lithium in those with PD, due to the risk of worsening motor symptoms (n = 1) [39]. There is some weak evidence supporting the use of dopamine agonists and monoamine oxidase inhibitors for the management of depression in PD (n = 3) [34,39,40]. Pramipexole was suggested to have an antidepressant effect not solely due a motor effect [40]. Selegiline has some antidepressant effects but further studies are needed [39]. If the mood symptoms are only present during off periods, it was suggested that patients might benefit from drugs addressing the motor symptoms [34]. However there was no evidence levodopa alone affected mood [40].
Other therapies for depression are not well explored in PD. The European Federation of Neurological Sciences (EFNS) concluded there was insufficient data to recommend psychotherapy, electroconvulsive therapy (ECT) or transcranial magnetic stimulation (TCMS) [40]. Other guideline assert that ECT has been used in PD, but that there are no specific trials in PD and is associated with risk (n = 2) [34,39].

Dementia recommendations
It is recommended that patients with dementia be assessed for anxiety (n = 2), however there is no clear consensus on what tools to use [41,42]. One guideline recommended the use of the Hospital Anxiety Depression Scale [42]. The evidence for the treatment of anxiety in dementia is lacking (n = 1) [42]. It is recommended that patients with dementia be evaluated and re-evaluated over time for depression (n = 5) [41][42][43][44][45]. As part of this assessment, patients should be evaluated for other secondary causes of depression. It is suggested that these patients be assessed for suicidality by one guideline [39], however another reported there was inconclusive evidence regarding this [46].

Discussion
This study provides a synthesis and quality assessment of available guidelines for the management of depression or anxiety in PD or dementia. We identified clear gaps in guideline quality and the evidence, which inform future research and knowledge translation. Guidelines AIAQS (2010) [42], NICE (2011) [41] Psychological Interventions can be considered for Anxiety in Dementia Guidelines NICE (2011) [41] There is little evidence about the treatment of Anxiety in those with Dementia. Cholinesterase Inhibitors can be considered for treating Dementiarelated behaviours, including anxiety. Comment These include: cognitive behavioural therapy, reminiscence therapy, multi-sensory stimulation, animal-assisted therapy, exercise, stimulation-oriented treatment (recreational or pleasurable activities), or improvements to a living situation. Consider the involvement of carers.
Although evidence is mixed, a trial of Anti-depressants could be considered for Depression in Dementia.

Guideline quality
Guidelines that were excluded due to low quality were typically those that lacked explicit development methods, thus ratings across all the domains were low. When examining the AGREE II ratings overall, the lowest rating was in assessing the guideline description of barriers and facilitators, implementation, resource implications, or monitoring/auditing criteria (Domain 5). In fact, few guidelines had discrete sections addressing knowledge translation. The concern about guideline applicability was explored in a 2015 systematic review [51], which found that applicability scored lower than any other domain [51,52]. If guidelines rarely address their implementation in practice, then there will be continued practice variation. There is clear evidence supporting the use of implementation tools to improve guideline uptake [51]. Thus making guidelines without a clear knowledge translation plan does a disservice to stakeholders [51]. The engagement of patients and caregivers was notably absent in CPG development. This process is important, as it is aimed at improving implementability, by ensuring the recommendations are comprehensive, adaptable and applicable to the target group and have an open process [53]. Given the constant changing nature of evidence, having up-to-date guidelines certainly makes a difference to the validity [32]. However, the lowest rated item was for the guideline update procedures.

Guideline content
There is an overall lack of recommendations related to the diagnosis or treatment of anxiety in either PD or dementia. This stems from the fact there is little evidence on how to approach the assessment. One guideline suggested the use Hospital Anxiety and Depression Scale for dementia, but they did not provide diagnostic accuracy information or suggestions for implementation [42]. There is also a concern that the medications traditionally used for anxiety can have major adverse effects [35], and there are few studies to guide treatment. Anxiety was less frequently mentioned than depression in the included CPGs, and in some cases was only mentioned in combination with other neuropsychiatric symptoms. The overall lack of evidence for anxiety care in PD and dementia is a major gap in the current research.
Guidance for depression was present in a higher proportion of guidelines. Despite this, there is variability in the reporting of levels of evidence and recommendations (Additional file 3: Figure S1). In some cases the recommendations for depression in PD only had 1 or 2 guidelines supporting them, indicating variance in guideline reporting. In other cases recommendations were vague, which can lead to difficulty with end user interpretation and implementation [36].
It is clear that screening for depression with a validated tool in PD is recommended, although evidence varies [37,38]. It is recommended, as a good practice point, that any diagnosis of depression is not made solely on a brief assessment tool, as these tools are more focused on case finding [37]. Although this is an important concept in detection, it was only recommended by one guideline [54]. A 2015 systematic review identified several validated tools for the detection of depression in PD, with the GDS-15 having the highest pooled sensitivity (0.81; 95% CI 0.64, 0.91) and area under the curve (0.94) [55].
Recommendations surrounding non-pharmacologic therapy were few, stating there was insufficient evidence for the use of psychotherapy, ECT or TMS [34,39,40]. Two recent trials demonstrated the effectiveness of cognitive behavioural therapy in PD [56,57]. This highlights the need for further large high quality studies on a range of non-pharmacologic therapies and the need for constant update of guidelines. Pharmacological therapy is recommended for managing depression in PD, but there is little evidence on choosing agents [39,54]. This has resulted in a variety of treatment recommendations, with little evidence to direct clinical practice.
Depression in dementia was more frequently addressed. However, these recommendations also had varied guideline and evidentiary support. Guidelines supported the evaluation of depression in dementia, but evidence ranged from high quality to good practice points [41][42][43][44][45]. Commonly recommended tools were the CSDD and GDS, with preference towards the CSDD due to better accuracy [39,42,45,47,48]. This was confirmed by a 2015 systematic review of depression tools for dementia, finding that the CSDD had a area under the curve of 0.89 [58].
Interestingly, the issue of evaluating for suicide risk was raised in two guidelines with divergent recommendations [39,46]. One stating there was inconclusive evidence [46] and another stating substantial evidence [39]. It is unclear why there is such a difference in reported evidence; perhaps development groups have different evidence available or differing interpretations of the evidence.
There are stronger recommendations for nonpharmacologic treatment in dementia than in PD, outlining several options [41,42,45,47,48]. The evidence for pharmacologic therapy is described as mixed with Grade 2A (Moderate Recommendation, High Level Evidence) to Class IV (Un-blinded Study, Expert Opinion) [39,41,44,45,49,50]. Again SSRIs and TCAs are the focus, with TCAs being less likely to be recommended due to side effects [39,42,45,47,50]. For those with dementia, there were more options recommended for therapy including stimulants, cholinesterase inhibitors and ECT [39,42].

Limitations
There is a well-recognized issue with heterogeneity in the terms used to refer to guidelines [52]. For our database search we used indexed terms from each of the three databases as well as key words using known nomenclature for guidelines and the comorbidities. It is also possible that the addition of the depression or anxiety criteria to the search may have been restrictive, however without these terms the search was impractical. To address this, we developed the search strategy with experts in the area of guideline systematic review and an experienced librarian, and we had an external reviewer independently assess the search strategy. To reduce the risk of missing literature not indexed in databases we contacted experts, searched references of included studies and performed an extensive search of the grey literature search.

Conclusions
Given the burden of comorbid mental illness in dementia and PD, it is key that we understand clearly the current knowledge base so we can improve care for these populations. This study provides a synthesis and quality assessment of the relevant guidelines. By synthesizing the recommendations, we identified areas of knowledge that are potentially ready to be translated into practice but also clear evidence gaps. This data was further evaluated in a subsequent study by stakeholders in focus groups to understand the other barriers and facilitators to the use of guidelines. This was to inform and help develop a comprehensive knowledge/end-user focused plan for addressing these gaps.

Funding
This project had funding from an Alberta Innovates Health Solutions-Knowledge to Action Grant.

Availability of data and material
The datasets during and/or analysed during the current study available from the corresponding author on reasonable request. All data from this study are presented in detail. For the quality assessment the individual ratings are not published as the AGREE group recommends publication f the mean scaled domain scores. The extracted recommendations are summarized in the tables and text, full details are available in the source guidelines.
Authors' contributions ZG and BM performed all citation/full text screening, quality assessments, data extraction and analysis and drafted the manuscript. ZG completed all statistical analysis. SG was involved in the grey literature search and quality assessment. JHL supervised all parts of the systematic review and analysis, was involved in the quality assessment and determination of inclusion. ZG, BM, SG, HH, SS, TP, NJ and JHL provided input and reviewed the proposal, protocol, analysis and manuscript. ZG registered the protocol with PROSPERO [58]. All authors had access to the data, reviewed and approved the final manuscript. ZG and JHL had full access to the data in the study and take responsibility for the integrity of the data and accuracy of the data.