Do stroke clinical practice guideline recommendations for the intervention of thickened liquids for aspiration support evidence based decision making? A systematic review and narrative synthesis

Abstract Rationale Aspiration is a common sequela post stroke as a result of oropharyngeal dysphagia. It is primarily managed using the poorly empirically supported intervention of thickened liquids. Where evidence is limited, clinicians may rely on clinical practice guidelines to support decision making. The purpose of this systematic review and narrative synthesis was to evaluate the evidentiary bases of recommendations made by stroke clinical practice guidelines regarding the thickened liquids intervention. Methods A systematic review was conducted on stroke clinical guidelines retrieved via searches conducted across a range of databases including Academic Search Complete, CINAHL, MEDLINE, and the Cochrane Library as well as through association websites. Guidelines were eligible for inclusion if they focused on adult stroke populations, made recommendations relating to the thickened liquid intervention and were published between January 2010 and December 2018. Four independent reviewers rated methodological quality using the AGREE‐II instrument. Intervention recommendations were extracted and analysed using the Criteria for Levels of Evidence Reported from the Canadian Stroke Best Practice Recommendations and a novel framework examining the appropriateness of the supporting evidence. Results Thirteen clinical guidelines were included in the review. Methodological quality was variable with seven rating as good‐excellent overall. Thirty recommendations regarding the intervention were extracted. Of these, 16 recommendations were classed as a recommendation to use the treatment and all guidelines made this recommendation. Much of the evidence used to scaffold recommendations did not directly support the intervention. Conclusions Despite the limited evidence base for the thickened liquid intervention, there was consensus among stroke guidelines in recommending it. This is despite limited empirical support. Furthermore, much of the evidence used to support recommendations was not appropriate, suggesting less than satisfactory evidence‐based practices in formulating recommendations. In this case, clinical guidelines may not be reliable decision‐support tools for facilitating clinical decision making.


| The intervention of thickened liquids
Stroke is a significant cause of oropharyngeal dysphagia which is known to increase the risk of malnutrition, dehydration, aspiration, pneumonia, and mortality. 1,2 It is estimated that approximately half of stroke patients experience aspiration and one-third of these develop aspiration pneumonia. 3 Furthermore, pneumonia is responsible for up to one-third of post-stroke deaths. 4 It is hypothesized that the early detection and treatment of aspiration is important for both individual patients and health care providers, as successful management of aspiration decreases medical complications and related health care expenditures including length of hospital stay. 5 It is widely accepted that swallowing thin liquids can pose a challenge for people with dysphagia (PWD) including those with dysphagia post-stroke, because they flow too quickly and can enter the airway below the vocal cords. 6 Consequently, aspiration is primarily treated by recommending thickened liquids (TL). This is a technique whereby a thickening agent is added to liquids to increase viscosity and slow bolus speed. 7 It is a common practice worldwide. Approximately 78% of speech and language therapists working with a range of clinical populations and 97% of those working with people with stroke employ the intervention. [8][9][10] Despite this, the evidence to support TL is not strong. [11][12][13][14][15] While some recent systematic reviews indicate emerging support, 12,15 the evidence base remains limited and contradictory with some reviewers even making a weak recommendation against its use. 16 Decision support tools such as clinical practice guidelines (CPGs) may therefore be especially useful for interventions such as TL.

| Clinical practice guidelines
CPGs are "systematically developed statements that assist clinicians to provide appropriate evidence-based care" (p. 57). 17 They typically focus on specific conditions such as stroke or dementia and provide evidence based recommendations regarding a range of interventions which are supported by reviews and evaluations of research findings. They aim to bridge the gap between research and practice and have complementary goals of optimizing patient care, decreasing variation in clinical practice, reducing costly, preventable mistakes, and adverse events. 18,19 Wellconstructed CPGs based on the most up-to-date evidence can improve the quality of clinical decision making. 18 Clinicians tend to view them as key sources of reliable guidance with as many as 78% of health care professionals employing them. 20 The use of CPGs by clinicians can be interpreted as positive clinical behaviour, illustrating a commitment to evidence based practice and, according to Hurdowar et al, as an indicator of quality of care. 17 Therefore, establishing the validity of CPG recommendations is important, perhaps especially in cases of interventions which are poorly empirically supported.
Quality CPGs should reflect evidence based practice and thus contain recommendations supported by current research evidence where present.
They should also ideally incorporate clinical expertise and patients' perspectives. This triad of evidence may be especially important in cases where the "existing literature is ambiguous or conflicting or where scientific data is lacking or an issue" (p. 6). 21 The use of practice evidence rather than research-based evidence to support CPG recommendations is not unusual. A recent review of CPGs for preoperative care for surgical antimicrobial prophylaxis, for example, found that most recommendations were based on clinical experience. 22 The increased emphasis on personcentred care has seen a growth in patient evidence being incorporated into decision support tools. A recent systematic review of decision aids by Clifford et al however, suggests this area remains considerably underdeveloped. 23 In the case of the TL intervention, there are known palatability issues with impacts on independence and quality of life and associated treatment adherence issues. 7,24 Such patient evidence should be reflected by guidelines when making recommendations.
The evolution of CPGs as sources of reliable clinical information has naturally resulted in research on guideline quality. 25,26 Recent systematic reviews of stroke CPGs suggest some issues in this area including regarding rigour of development. 16,27 Navarro-Puerto et al, for example, reported that the majority of ischaemic stroke CPGs reviewed in their study, received an overall "would not recommend" rating. 27 There has been considerably less focus on assessing the quality of recommendations made within CPGs-specifically the evidentiary bases of recommendations. Cosgrove et al's examination of recommendations in a major psychiatry CPG showed that fewer than half of the included studies supporting the recommendations met the criteria for high quality. Furthermore, they identified that one-fifth of the references were not congruent with the recommendations. 28 As stroke is the leading cause of long-term adult disability, the quality and content of stroke CPGs may be especially pertinent especially as organized stroke care has been shown to be beneficial and stroke evidence continues to grow rapidly. 17 30 There is a risk that such CPGs may result in the dissemination of inaccurate knowledge, be translated into poor clinical practice, result in waste of resources, be potentially harmful to patients, 19,30 and contrary to the overall aims of evidence-based practice generally.

| Study aims
The purpose of this systematic review and narrative synthesis is to evaluate the evidentiary bases of stroke CPG recommendations regarding the TL intervention.

| METHOD
There are no specific standardized reporting guidelines relating to systematic reviews of CPGs. To this end, relevant sections of the Preferred Reporting of Items for Systematic Reviews (PRISMA) guidelines 31 were referenced to standardize the conduct and reporting of the review. The protocol for this review has not been published elsewhere.

| Identification of CPGs
A search strategy was developed to identify all stroke CPGs that were published between January 2010 and December 2018. For the purposes of this study, stroke CPGs were defined as CPGs focused primarily on the condition of stroke in both acute and chronic phases and which made recommendations for the rehabilitation of the stroke patient. The search strategies were based on the processes recommended by Moher et al (Appendix S1). 31 Databases, stroke association websites, and guideline websites were the main sources searched (Appendix S2). Hand-searching of references of included CPGs was also conducted. Where further information was required and not available in the main document, the main author was contacted.
A sample search strategy for Scopus is presented with the main search terms and related terms (Appendix S3). Core search terms were stroke, guidelines, and TL. Inclusion criteria were as follows: • CPGs focused on the rehabilitation of stroke populations; • General stroke CPGs which are not intervention specific (eg, bolus modification) or discipline specific (ie, produced by individual allied health groups); • CPGs making recommendations relating to the intervention of TL for the management of dysphagia in adults (>18 years) post-stroke; • CPGs produced between January 2010 and December 2018; • Most recent versions of CPGs; and • Published in English.
CPGs should be reviewed or updated regularly, ideally within 3 to 5 years. [32][33][34] This was extended in our study to allow for the time required to produce and disseminate them by organizations. Titles and abstracts and guidelines were reviewed by the main author (A.M.) and those that did not meet the criteria were excluded. The remaining publications were reviewed independently by three reviewers (A.M., M.K., C.R.) and consensus regarding inclusion reached on discussion.

| CPG quality
The methodological quality of the included CPGs was evaluated using The Appraisal of Guidelines for Research and Evaluation Second Edition (AGREE-II) tool. 35 The AGREE-II has been established internationally as valid and reliable for quality assessment of CPGs. 36,37 To facilitate accurate reporting, the organization websites of included CPGs were visited to retrieve relevant documentation, for example, evidence tables, working group declaration of interests, and CPG development frameworks.
Where available, these were used in the evaluation process. The and met to agree final ratings for items where discrepancies in scoring were evident. Each reviewer provided a rationale for their score and consensus was reached between the four reviewers. Scores across all domains were converted to percentages for each CPG. The AGREE-II developers advise against calculating an "average" or "overall" score for CPGs representing a combination of the six AGREE-II domains as each domain is considered more individually informative to the user. Thus, a method previously applied by other authors to determine "high-quality" was also used, where CPGs achieved a standardized score of ≥50% on all six domains. 38 Table 2. the remaining sourced through stroke organization and guideline database websites. For most of those retrieved via bibliographic searches, full copies of guidelines and supporting documentation were obtained via the developers' websites. One CPG (Germany 2013) was primarily evaluated using the journal publication as its authors advised that the obtained CPG was the only English version available.

| Quality of CPGS
The CPGs generally scored well, with 7/13 scoring above the >50% quality threshold in all six domains. The others tended to score vari-  A AHRQ Evidence Report Summary 1999 63 Cook and Kahrilas 1999 64 Logemann et al 1998 65 Elmstahl et al 1999 66 Huckabee and Cannito 1999 67 Klor et al 1999 68 Rosenbek

| The evidentiary bases of recommendations
The consensus across all CPGs was that the intervention of TL should be used for people with aspiration subsequent to stroke, either in isolation or as part of a general dysphagia treatment programme. The evidence base pertaining to the intervention of TL is less than robust and therefore a mismatch between evidence and recommendations is apparent. Clinicians who access CPGs may do so due to a lack of time to read, appraise, and synthesize all the available evidence on the topic, leading to a reliance on CPGs to provide evidence-based recommendations. Therefore, the unequivocal recommendations by the included CPGs to administer TL for PWD post-stroke are problematic.  Furthermore, the more recently produced GPGs did not tend to include newer systematic reviews 11,12,14,15 including that of Beck et al who recommended against the use of TL. 16 While such deficits may in part result from the time required to develop and produce new versions of guidelines, an efficient process for retrieving and including most recent research is warranted to reflect the most up-to-date evidence. Martínez García et al noted that such a mechanism is imperative given their findings that CPG recommendations quickly become outdated, with one out of five recommendations being out of date after 3 years. 87 Protocols for removing outdated CPGs from circulation are therefore also warranted. In many cases, previous versions of the included CPGs were still available and a number of "older" CPGs were still in use.
The employment by guideline developers of other CPGs as sources of evidence and support for recommendations was also noted. This reliance can be understood to be less than ideal, especially in the context of the findings from this study. Furthermore, the range of sources to support recommendations was generally not extensive and did not always reflect the ideal triad of evidence from practice and patient perspectives. Where consensus was employed, this usually reflected consensus within the working group and not a broader representation of individuals directly involved in prescribing and implementing highlighted interventions. 23 There also tended to be poor consideration of other important factors such as contextual evidence which has been clearly shown to factor in treatment implementation in a number of health disciplines. 88 What can be positively inferred is the attempt by most CPGs working groups to systematically review the evidence. Furthermore, some CPGs do provide recommendation rationales. Scotland 2010 #119, for example, notes that despite the sparse available data, it seems "prudent" to include dysphagia-specific management as part of the standard protocol of stroke management in the acute care setting: "Diet modification has been shown to be effective in specific individuals using video-fluoroscopy and is standard management of dysphagia following stroke" (p. 12). 49  • Guidelines developers should consider the use of expert panels with intervention-specific expertise when evaluating evidence and making and reviewing recommendations. Where recommendations are based on such evidence, this should be explicitly stated to ensure readers understand the nature of the evidence supporting that recommendation and can act accordingly.
• Care should be taken to ensure that traditional clinical practices are not automatically recommended or assumed to be best practice in the absence of supporting evidence.
• To ensure that publicly available guidelines remain current, processes for updating CPGs need to be explicit and rigorous. For example, UK 2013 CPG authors designed ongoing external peer review elements to their guideline development process. A process for removal or CPGs is also indicated based on the findings of this review.

| For individuals and organizations employing CPGs
• Individuals, teams, and organizations employing CPGs should be aware that CPG recommendations may not be wholly evidencebased and should review the evidence where possible to ensure recommendations are evidence-based.
• Clinicians may benefit from training in CPG critical appraisal in order to be able to evaluate the supporting evidence used by CPGs. Such investments would maximize the applicability of CPGs as well as increase the number of clinicians who may, in the future, contribute to CPG content themselves.

| STRENGTHS AND LIMITATIONS
Systematic methods were used to identify, select, appraise, and synthesize the findings from the included CPGs. The PRISMA standardized reporting guidelines were referenced to illustrate the flow of studies in the review. The interdisciplinary review team reflected a mix of academics and clinicians. This is first review to examine the evidence underpinning recommendations in stroke CPGs.
Some limitations are evident. Comparatively, this review includes a small sample size. Possible explanations for this include the use of TL as a core term in the search string and the timeframe imposed to maximize retrieval of the most current versions of CPGs. Additionally, although it is advised that CPGs are updated every 3 to 5 years, this was not the case for a number of CPGs in the review. This means that some included CPGs could not reflect up to date evidence. Furthermore, the inclusion criteria did not capture CPGs published by professional speech and language therapy associations which may reflect deeper discipline-specific knowledge regarding the research evidence and the intervention itself. This may limit the external validity of the findings. Where possible the evidence referred to in "other guidelines" was also extracted. In some cases this was not possible due to guideline age, lack of access to original documentation or being superseded by a subsequent guideline. Some CPGs, such as Germany 2013, may not have been best reflected in this exercise due to lack of accessibility to guideline development documents.