UEG framework for the development of high-quality clinical guidelines

Abstract

The 48 national member societies and 17 specialist member societies which operate under the umbrella of United European Gastroenterology (UEG) increasingly develop clinical practice guidelines for both national and international implementation. The methodologies and strategies used in these guidelines vary considerably. The UEG Quality of Care Taskforce aimed to provide a framework for quality guidelines in order to assist member societies in the process of developing guidelines, and to provide a tool for readers of guidelines to critically appraise their quality. We outline the steps necessary to begin the guideline development process, how to build working groups, how to search for evidence, how to grade the quality of the evidence, how to reach consensus on statements and how to write the guideline document. We believe that using this framework will increase the potential to produce a high-quality guideline which is transparent, independent, reproducible and implementable.

Introduction

Clinical guidelines are fundamental to assist clinicians and other healthcare professionals optimize the care of people and patients, based on the latest and best available scientific knowledge and, where evidence is scarce, consensus opinion of the experts from the respective field. Sometimes, guidelines are expressed in other formats such as consensus statements, expert recommendations or position papers. Hence, guidelines are not uniform and the process by which guidelines are developed can be very heterogeneous. Factors that influence the quality of the guidelines are the composition of the guideline panel, the methodology of the evidence review, the evidence and recommendation rating, the decision-making and consensus process and, finally, the implementation of the guideline. There are guidelines with a very stringent process, including different guideline committees, systematic literature reviews, PICO (population, intervention, comparator, outcome) questions, Delphi technique and/or consensus conferences, which are all important for objective and unbiased evidence-based recommendations. However, feasibility and also costs are potential obstacles to universal application. Other guidelines are written by only a few experts with a very lean systematic process. A lean guideline process might allow for a faster response and timely publication of practice recommendations when there are new developments on a topic. However, the potential bias of a small expert steering group needs to be considered. A very sensitive topic to deal with is the conflict of interest (COI) of the expert panel. The experts for a particular topic might be involved in clinical trials or research activities that are supported by industry. This COI has the potential to influence the expert’s position with respect to the subject matter being considered or, at the very least, to raise concerns regarding bias. However, excluding these experts from the guideline process may significantly reduce the panel’s expertise and experience. Thus, all aspects of the composition of the guideline panel and the methodology have advantages and disadvantages.

National and international health organizations and medical societies have increasingly issued their own guidelines and, as a result, the quality of the guidelines can vary considerably. In addition, there are often various guidelines on a topic, some of which contain differing, sometimes even contradictory, recommendations. Different recommendations may be due to national circumstances but also due to a bias in the guideline process. Thus, it is important to develop standards for guideline development which can help organizations optimize their guideline process and ensure that recommendations being developed are transparent, unbiased and evidence-based. In 2003, the Guidelines International Network was founded with the aim to develop and implement international consensus for minimum standards for high-quality guidelines (Appendix 1). These standards can help users to better classify the quality of guidelines.

Several national and international organizations have proposed standards for guideline development. These can assist guideline developers to use appropriate methodologies, and similarly assist guideline readers to assess the quality of a guideline. One of the most widely applied tools for this purpose is the Appraisal of Guidelines for REsearch & Evaluation (AGREE II) Instrument (Appendix 1). In addition, several medical societies have proposed standards or guidance for their guideline process (Appendix 2). Accordingly, United European Gastroenterology (UEG) has developed a “guideline for guidelines” document that will discuss the key steps in the guideline process and thereby offer a framework for future UEG-supported guideline projects (Figure 1).

Figure 1.

Schematic overview of guideline development process.

This document was developed by the UEG Quality of Care Taskforce (QoC TF), which includes representatives from specialist member societies and the National Societies Committee. A six-member working group was formed, consisting of five QoC TF members and a methodologist (SB). We identified published guideline appraisal tools and standards of practice from different medical subspecialties. These documents were retrieved and critically reviewed by working group members. An appropriate methodology for the UEG framework was developed through consensus discussion at face-to-face and virtual meetings. Following approval of the draft by the working group, the final draft was approved by all QoC TF members.

How to start a guideline process

Identifying the unmet need

Before developing new guidelines, the need for creating a guideline should be carefully assessed. New guidelines should not duplicate an existing document, but rather focus on relevant questions which have not been addressed sufficiently.¹ Similarly, in non-English speaking countries, it may be preferable to translate and adapt an existing guideline rather than develop a non-English guideline de novo. Accordingly, the unmet need for a guideline should arise after a thorough investigation of existing guidelines and the current evidence available. Firstly, the need for new guidelines may arise due to the emergence of new diagnostic or therapeutic modalities. For example, the advent of new drugs in the area of inflammatory bowel disease or viral hepatitis may necessitate new or updated guidelines. Secondly, the need for a guideline may arise due to gaps in current knowledge or practice. For example, guidelines may be especially useful in the management of rare diseases since physicians have less personal experience to guide their decision-making. Thirdly, in the case of variations in epidemiology, a new guideline may seek to adapt pan-European recommendations to the local setting. For example, a national guideline for the management of Helicobacter pylori infection may be necessary in order to consider local antibiotic resistance patterns and efficacy data. In some situations, a new guideline may seek to address health inequalities and variations in the provision and quality of care.²,³ Where the need for a new or updated guideline is identified which is cross-disciplinary, co-operation between national or international groups and societies to develop the guideline is encouraged. The umbrella of UEG provides an excellent space to develop such collaborations.

Defining the scope

After the unmet need for a guideline is established, the scope of the guideline should be defined. This process involves deciding both the breadth and the depth of the document to be developed. It should consider what should and should not be covered in the guideline, with respect to the target audience, patient population, healthcare setting and treatment or clinical outcomes of interest. The scope of the guideline should be based on the gaps identified above, rather than aiming to produce an all-inclusive document.⁴ For example, a guideline on the management of cirrhosis should target one or more specific aspects of care, such as portal hypertension, encephalopathy or transplantation. Writing a list of priority topics will aid in this process. In this manner, the document will be balanced to provide a specific answer to selected questions within the guideline statements and provide a manageable explanation and focused solution. Further details on how to define the scope and purpose of a guideline can be found in Domain 1 of the AGREE II instrument (Appendix 1).

Building working groups and formulating PICO questions

Once the scope of the guidelines is defined, a steering group should be formed to continue the process.⁵ Members of the steering group should be chosen based on their pre-eminence in the field, their ability to contribute to the work process and their personal experience as effective team members (Figure 2). In addition to choosing recognized experts in the field, the steering group may wish to consult experts in methodology and research synthesis, health economists, representatives of stakeholders, representatives of potential users and, in some cases, experts in health systems or policymakers. These experts should be consulted at this early stage; however, they do not all need to be included in the steering group. For practical reasons, a steering group should not include more than a handful of participants who are able to effectively deliver a joint decision on key issues and literally steer the work process when problems arise. The steering group should also liaise with a secretarial team that can provide support throughout the work process.

Figure 2.

Structure of the guideline working groups.

Once a steering group has been formed, the group should identify and isolate relevant topics within the desired scope. The topics that need to be addressed should be listed and prioritized based on joint decision-making of the steering group. The chosen topics are then elaborated upon, and a list should be made of key issues to be addressed, including knowledge gaps and controversies. It is advisable to select key issues that are adequately focused, of manageable size and are amenable to practical change. The topics should be restricted as much as possible at this stage, as the topics often expand and multiply later in the development process.

Box 1.

How to involve patients in the guideline process

• Ask patients about unmet need (patient survey or focus groups)

• Involve patient representative organizations as stakeholders

• Have a patient representative in each working group

• Review patient feedback on draft PICO questions

• Ask patients to vote on importance of outcomes (GRADE)

• Create a “patient guideline” – a summary in simple language

After the steering group has been formed, the scope has been defined and priority topics selected, it is useful to formulate potential recommendations in the framework of PICO questions.⁶,⁷ PICO refers to four elements that should be in a question governing a systematic search of the evidence. The PICO formulation defines the population investigated, the intervention in question, the comparator for assessing an alternative option and the outcome(s) used to assess the intervention. PICO questions are influenced by the desired scope of the guidelines and the relevant topic. The PICO framework is simply a method that ensures a comparative answer and, therefore, could result in a narrow or broad answer, depending on the desired scope of the question.⁸,⁹ For example, a guideline for the management of acute non-variceal upper gastrointestinal bleeding might address the specific question: in haemodynamically stable patients with evidence of an adherent clot which cannot be dislodged with targeted irrigation (population, patient or problem), how effective is endoscopic therapy (intervention) compared to a proton pump inhibitor alone (comparator) in reducing the rate of rebleeding (outcome)?

Once PICO questions have been drafted, a broader developing group should be formed, involving a greater number of participants (ideally not exceeding 20). It is useful to draw a list of key external organizations and stakeholders who should be invited to participate in the development group based on their good standing and their achievements in the field. For example, a developing group for a guideline on the management of premalignant gastric lesions may include clinical gastroenterologists, endoscopists, pathologists and medical oncologists. Details on stakeholder involvement can be found in Domain 2 of the AGREE II instrument (Appendix 1). It is important for the developing group to be multidisciplinary so that bias from a like-minded group with an identical agenda is avoided. It is at this stage that patient groups may also become involved (Box 1). Although the role of patient contribution to medical guidelines has been debated, the involvement of patients in the development of guidelines may translate to a more implementable guideline. This is increasingly evident in the field of inflammatory bowel disease (IBD), where patients are also increasingly involved in defining PROMs (patient-reported outcome measures).¹⁰

Finally, an external review group should be formed, made up of experts and stakeholders who were not involved in the work process. The review group’s aim is to provide a peer-based high-quality critique of the drafted guidelines. Ideally, the review group should have a wide scientific, geographical and philosophical reach, being able to illuminate the gaps that remain and solidify the guideline’s statements. From a practical perspective, the review process should be balanced and should not impede the work process. It is worthwhile to consider that while an extensive validation of the guidelines is desirable, a review group that is too large has the potential to encumber the process.

Managing COIs

COIs should be addressed early in the process of establishing the working group. Since numerous COIs can arise during the work process and potentially introduce substantial bias in the final document, we recommend that all individuals provide a declaration of interest (DOI) prior to being allowed to participate in the work process.¹¹ These forms should include financial, academic and public disclaimers. COIs of a scientific nature are more difficult to identify and address. For example, a working group member who is involved in research which is directly related to a guideline statement may have a viewpoint which could introduce bias. The steering group should review participants’ declarations and decide if any conflicts exist that might preclude a specific participant from commencing the work process. During the work process, COIs may change.¹² Therefore, we recommend reassessing COIs at selected intervals of time during the work process (e.g. every 3–6 months, or at every major meeting). A practical approach could be re-signing a draft of the previously signed declaration. Any changes in a DOI should be re-approved by the steering group. In certain cases, it may be prudent to consult an unbiased, certified external consultant. All declared interests must be reported in the final guideline document.¹³

It is a matter of judgement as to what actually constitutes a potentially significant COI with regards to industry funding. Being employed by, consulting for, acting as a paid speaker for or owning shares in a company with a product related to the guideline should be considered a potential COI. On the other hand, it is our opinion that having received research funds from a company, either directly or indirectly, might not automatically constitute a COI.

Budget and timeline

Budget and sources of funding should be transparent to members of the working groups. Sources of funding should also appear in the final guideline document. Direct funding from industry should be avoided, whether or not they have an explicit commercial interest in the outcome of the guideline. Industry funding fundamentally undermines the quality and objectivity of any guideline and introduces concerns regarding commercial bias, which will be a barrier to publication and dissemination. Funding from institutional or government bodies is acceptable provided that they are not involved with the guideline development or review process. For example, the QoC TF under the auspices of UEG offers activity grants for the development of guidelines.^14–17

A timeline and a timetable should be addressed at the beginning and end of every meeting. Establishing a timeline early in the process and maintaining strict adherence to the timetable is of utmost importance. The timetable should be available to all participants. Realistically, a good quality guideline will take at least 12–36 months to complete (Box 2).¹⁸

Box 2.

How to start the guideline process

• Identify unmet need

• Define the scope

• Invite stakeholders

• Build working groups

• Develop PICOs

• Address conflicts of interest

• Set budget and timeline

How to search for evidence

Framing the question

Although preliminary questions in the form of PICOs may have already been drafted when deciding the guideline’s scope, these questions must be clearly defined, agreed upon and finalized prior to commencing the literature search (Figure 3). Several frameworks, besides PICO, have been described; however, PICO remains the most common and easy to apply.¹⁸ For example, the ECLIPSE framework addresses questions related to health policy and management, rather than clinical questions.^19–21 ECLIPSE is structured according to the mnemonic: expectation (why is the information needed), client group, location, impact (what is the change in service, what constitutes success and how is it measured), professionals and service. For example, when searching for data on continuity of care following discharge of patients with an exacerbation of ulcerative colitis, a question might be framed using the ECLIPSE model. The expectation is that improvement in the transition process among patients with ulcerative colitis (client group) being treated in the community (location) will improve health outcomes (impact). Relevant studies may involve gastroenterologists, IBD defined above nurses and social services (professionals) in community clinics or using eHealth platforms (service). If the search question pertains to a diagnostic test or device (as opposed to an intervention or health policy) then it may be useful to formulate the question using the PIRO model (patient/disease, index test, reference standard and outcome). For example, when searching for data on the effectiveness of colon capsule endoscopy for colorectal cancer screening, the patient is an average risk adult, the index test is colon capsule endoscopy, the reference standard is high definition colonoscopy and the outcome could be adenoma or advanced adenoma detection.

Figure 3.

Searching for evidence.

Search protocol and strategy

The quality of a literature search directly impacts the quality of the guideline. Evidence must be unbiased and comprehensive.²² Before a search is performed, a search protocol must be agreed upon. The aim of this protocol is to outline what strategy will be used in order to answer each question in the framework; which inclusion and exclusion criteria will be applied; which databases will be utilized; and how the results will be assessed and documented.

The most commonly used search protocol (favoured by National Institutes of Health (NIH), National Institute for Health and Clinical Excellence, World Health Organization (WHO) and UEG member societies; Appendices 1 and 2) is a step-by-step iterative approach that allows the working group to first consider high-quality evidence from systematic reviews and meta-analyses, then randomized controlled trials (RCTs) and then all other levels of evidence.⁴,²³,²⁴

It may even be necessary to perform a new systematic review and evidence synthesis if there are no high-quality reviews recently published. In order to decide whether to update an existing systematic review or commission a new document, we recommend the stepwise approach developed by the WHO. This involves developing key questions, identifying a relevant systematic review, assessing its quality and determining whether it is sufficiently up to date.²³

At this stage, a search for primary sources, which may not appear in the high-quality review, should be performed. It is important to utilize methodological filters rather than simply searching keywords or MeSH (medical subheading) terms as free text. Both EMBASE and TRIP allow searching according to the PICO format. Other databases such as PubMed and Medline have advanced search options and are also acceptable. It is preferable to use more than one database, and it may be useful to seek the assistance of a medical information specialist.

Search filters

At this stage it is not advisable to limit the search by excluding studies with a low quality of evidence. This could lead to the exclusion of topics where a paucity of evidence exists. Similarly, limiting the dates in the search may lead to the exclusion of pivotal studies. In some situations, however, such as when a guideline is being updated or when addressing a new development, limiting publications by date may be useful. In many cases, for practical reasons, the search language is restricted to English, because a reliable translation is not readily available. This could introduce publication bias. If a foreign language guideline is being produced and non-English local data are available, then a language restriction should not be applied. Including data published in abstract form only should be avoided, as these data are often preliminary and not subject to rigorous peer review.

Documenting and reporting

All stages of the search should be documented to ensure transparency and reproducibility, and to enable future guideline updates. We recommend that each manuscript is reviewed by at least two working group members who both agree to inclusion of the publication. The final search results should be summarized in an output document and include the key references together with a short explanatory text and level of evidence (see How to grade the quality of evidence below). Creating an evidence table (for each PICO question) is a helpful way to provide a structured overview and documentation of the available evidence.²⁵ We recommend using one of the published checklists (such as PRISMA or MECIR; Appendix 1) to document the search performed (Box 3).²⁶,²⁷ Further details on search methods, assessment of the evidence and formulation of recommendations can be found in Domain 3 of the AGREE II instrument (Appendix 1).

Box 3.

How to search for evidence

• Frame the question

• Approve search protocol and strategy

• Define search filters

• Choose bibliographic databases

• Document and report using checklist (e.g. PRISMA)

How to grade the quality of evidence

There are several systems available to grade the quality of evidence; however, we recommend adopting one of the three most commonly encountered: the Oxford system, Scottish Intercollegiate Guidelines Network (SIGN) and the Grading of Recommendations Assessment, Development and Evaluation (GRADE)²⁵,^28–30 (Appendix 1). Both the Oxford and SIGN classification systems focus on the quality of the individual studies, while the GRADE approach views the available evidence from the outcome perspective (Table 1).

Table 1.

Systems to grade the quality of evidence.

	OCEBM (Oxford)	GRADE system
Advantages	• Simple, easy to apply• Allows guideline statement to be made based on expert opinion when evidence is lacking• Applicable to public health questions	• Ability to raise or lower level of confidence in the results based on different domains such as bias, design and confounding• Permits greater transparency as to how judgements on quality of evidence are made
Disadvantages	• Judgments about quality of evidence often based on study design alone• No definitive practice recommendations given	• Multiple steps, more complex to apply• “Conditional” recommendations may be difficult to interpret• Potential lack of reproducibility• Limited ability to address public health questions since non-RCTs are penalized

The Oxford and SIGN grading systems

The Oxford Centre for Evidence-Based Medicine (OCEBM) appraises data according to five levels of evidence (Levels 1–5). This system may be applied to different types of studies, with seven different outcome measures, including disease prevalence, accuracy of diagnostic tests, disease prognosis, therapeutic effects, harms of screening and usefulness of screening. The level of evidence assigned depends upon which outcome is measured in the particular study. For example, a large population-based cohort study measuring the prevalence of Hepatitis C is considered Level 1 evidence, whereas a cohort study assessing treatment outcomes for Hepatitis C may be considered Level 4.³¹ SIGN provides an alternative, more general, but in many ways similar grading system.³¹ Previous versions of both the Oxford (2009 version) and SIGN (before 2013) methods each included grades of recommendation (A, B, C, D) that could be derived directly from the levels of evidence. In the most recently updated versions of Oxford (2011) and SIGN (2019), these grades of recommendations have been abandoned. Oxford (2011) explicitly refrains from making definitive recommendations and the current SIGN system (2019) assigns a “strong” or “conditional” recommendation based on the evidence while also considering benefits, harm and patient preferences.

The GRADE system

GRADE is unique in that it allows the level of evidence to be downgraded, such as in the case of limitations in study design (risk of bias), inconsistencies between studies, indirectness, imprecision or publication bias. For example, a single-blind RCT comparing different bowel preparations in which the endoscopist, but not the subject, is blinded to group allocation, may receive a lower grade than a double-blind RCT. Similarly, an RCT assessing clinical remission in Crohn’s disease could be penalized if a high rate of loss-to-follow-up is observed. GRADE also specifies circumstances in which evidence from a study may be upgraded, such as a large magnitude of an effect. For this reason, a well-designed case-control study may produce evidence on par with an average or low-quality RCT. GRADE requires that a reviewer judge each factor that determines the quality of evidence for each outcome. Because of its relative complexity, the GRADE system may take some time to master and obtaining methodological support can be useful when applying GRADE in the guideline development process. A computer program (GRADEpro) has been developed to aid in the construction of evidence tables and summary results using the GRADE methodology.²⁵,³² A simplified schematic summary of the GRADE system is provided in Figure 4.

Figure 4.

GRADE approach to assessing level of evidence.

How to evaluate the evidence for guidelines using consensus methods

After an evidence table has been constructed for a particular PICO question, a draft recommendation can be written. Once all draft recommendations have been completed, they should be discussed by the panel of participants as a whole. The participants should be presented with the evidence tables (indicating the number, type, quality and grade of evidence of studies for each review question; see Documenting and reporting and How to grade the quality of evidence). In the process of guideline development, consensus decision-making requires the consent of all guideline panel members with a unanimous decision resulting in a recommendation. Where a unanimous decision has not been reached, a process of informal consensus should be attempted. When disagreement persists, it is useful to use a formal consensus methodology.³³

The most common formal consensus method in use is the Delphi method (Figure 5). The Delphi method is a stepwise process in which an anonymous questionnaire is circulated among the developing group. The results are reported back to the group, after which the questionnaire is again circulated. This process is repeated until there is a convergence of opinion (to a predetermined level of agreement) or until no further changes in the replies are elicited. An advantage of this method is that it allows a large number of participants.³³ The RAND/UCLA appropriateness method (RAM) is a modified two-round Delphi process consisting of two independent groups: a core panel and an expert panel.³⁴ The core panel oversees the consensus process and provides synthesized literature evidence to ensure that the expert panel has all pertinent information available to guide evidence-based decision-making. Regardless of which method is used, it is important that the process is defined in advance and should contain at least two rounds of voting to establish whether agreement can be reached.³⁴ A second voting round allows for revision of statements and recommendations with a view to achieving consensus. Ideally, the final round of voting should occur at a face-to-face meeting of all participants to allow debate and discussion where a consensus on draft statements has been difficult to achieve.

Figure 5.

Methodology for reaching consensus.

The level of agreement required for a statement or recommendation (generally expressed as a mean rating or percentage of agreement) should be decided in advance, and this threshold should not be modified during the process. For example, a threshold of 80–90% of participants must agree to allow a given statement or recommendation to be included in the guideline. The level of agreement with the statement in the final voting should ideally be included in the final guideline manuscript. Where the evidence for or against an intervention is closely balanced, the guideline participants may decide not to formulate a recommendation and instead make a statement that no recommendation is possible and suggest that the question be the subject for further research. The participants may recommend more than one management option where they are of equal efficacy or where individual circumstance will determine efficacy. The GRADE evidence-to-decision framework should be used to provide a systematic and transparent summary of the discussions outlining the factors determining the direction and strength of the recommendation, and to help to achieve consensus and identify causes of dissenting opinion (Box 4).

Box 4.

How to evaluate evidence and reach a consensus

• Grade the evidence using an established system (e.g. Oxford, SIGN, GRADE)

• Construct an evidence tableWrite draft recommendations

• Attempt to reach unanimity via discussion

• Use consensus methodology if necessary (e.g. Delphi, RAM)

• Document discussions with evidence-to-decision framework (e.g. GRADE)

How to write the guidelines

Writing draft recommendations

Recommendations should be drafted that relate directly to the PICO questions formulated and the outcome of the literature search that was performed. The language used should ideally follow a uniform format and refer to the population, the intervention and the context. Recommendations should also each contain an indication of the strength of the recommendation and the quality of evidence that underpin them, followed by a summary of evidence. Domain 4 of the AGREE II instrument provides an outline for how to clearly present summary recommendations (Appendix 1).

Constructing a guideline statement

Guideline statements are generally written in English, but can also be drafted in the primary language for the country or geographical area to which they pertain. Whatever the language used, the statement should be drafted in the simplest language possible, while ensuring clarity of meaning. Statements should generally adopt a broad health-system perspective, based on the needs of the patient population on a national and/or European level. This perspective is key when determining benefit, harm and costs of any intervention. Recommendations should clearly state the assumed values and preferences of the population for which it is intended. Recommendations should be concise, unambiguous and actionable, with consistent language across the guideline. It is better to keep sentences short and avoid long statements with multiple elements as they are more likely to give rise to lack of agreement, in which case it may be difficult to define which element is generating the difference in opinion. Guidelines should use language such as “we recommend” and “should” for strong recommendations, and “we suggest”, “may” and “can” for weak recommendations. For example, it is inappropriate to state “we recommend head of bed elevation for the management of nocturnal reflux-related symptoms”, since a low level of evidence exists to support this practice. Recommendations should ideally be categorized into “strong” and “weak” as per the GRADE system, but a similar approach is beneficial even if GRADE is not used. The strength of the recommendation is determined by the balance between the relative benefit and harm of the intervention, taking into account the overall quality of the evidence and the magnitude of undesirable effects relative to the potential benefit. Other considerations include resource implications, feasibility, acceptability and equity. Evidence-to-decision tools can facilitate the process of determining the strength and direction of a recommendation, ensuring all criteria are considered and allowing judgements to be recorded.

The participant who drafts the guideline should provide justification for the recommendation and include relevant considerations about implementation strategies, practice implications, monitoring and future research priorities. This can then be included in the supporting text. It should be emphasized in the final guideline manuscript that the recommendations/statements should not read in isolation but considered in the context of the supporting text and evidence. Guidelines should support shared decision-making and allow for flexibility to reflect the need for individualized treatments – that is, it should be recognized that not all strong recommendations based on high-quality evidence will be suitable for every patient.

Final manuscript and publication

There should be a clear outline of who will write the final guideline document. This should contain all of the statements or recommendations that achieved the threshold for agreement in the final voting round. The evidence supporting each statement should also be included in the final manuscript (either in the body of the manuscript or as supplementary material that is readily accessible for anyone who has access to the final published manuscript). The supporting text is important as it allows a context to be offered for each recommendation and allows potential caveats to any recommendation to be outlined.

The structure of the guideline document should ideally also be decided in advance and may comprise an executive summary, a main body and one or more appendices. The executive summary is a concise summary of the recommendations, their strength and evidence quality. The background to the recommendations should include the overall objective of the guidelines, the target population and target users and the main clinical questions to be addressed. The main text of the guideline should include a table of contents, background, objectives, development process of the guideline, evidence summary, recommendations, resource and service implications and recommendations for future research. All participants, roles and affiliations should be listed, with their COIs and how these were managed. The systematic review(s), outcome ratings, summaries of findings, GRADE evidence profiles, evidence-to-decision frameworks and tables, and any other relevant documentation can be supplied in appendices and/or may be published electronically as supplementary material, as long as these remain readily accessible.

Where appropriate, guideline panels should actively consider including a specific section in the guideline manuscript addressing how the recommendations might be implemented in countries or populations with resource limitations, which will serve as a barrier to comprehensive guideline implementation. Ideally, implementation strategies should be proposed, and specific tools and/or strategies should be suggested to assess implementation. Strategies to aid guideline implementation are further explored in the Guideline Implementation Toolbox, published on behalf of the UEG QoC TF.

Prior to submission for publication, the AGREE-II appraisal instrument should be applied to ensure the guideline meets international quality standards (Appendix 1). COI declarations should also be shown clearly. A timeframe for updating the guideline should be agreed and outlined in the final published manuscript. It is useful to consider adding a disclaimer in any published format of a guideline reminding readers that a guideline does not replace the need for clinical decision-making for each individual patient and cannot always take account of variations in patient population, facilities and resource availability, thus avoiding the inappropriate use of the guideline in medico-legal proceedings.

Validation and review

A robust validation process is an integral part of guideline development and may involve consultation with relevant stakeholders, public consultation and external peer review. Completed draft guidelines should be published on an appropriate website with a designated feedback form to facilitate structured feedback from relevant stakeholders. External peer review may also take place during the consultation process and peer review by a panel of expert peer reviewers not involved in the guideline development or voting may serve as a further useful validation. Following submission of comments on the draft guideline, the participants or a panel of coordinators should respond to all feedback and meet to discuss if any changes are required to the guideline. Guidelines should be refined and prioritized prior to final publication, and comments and responses made during consultation should ideally be made available with the final publication of the guideline.

Guideline implementation

Prior to publication, strategies to increase the implementation of a guideline should be considered. Guidelines should be published on an open-access platform or made available for download free of charge to maximize the dissemination of the guideline and ensure it is accessible. For example, many English and non-English guidelines, including those produced by UEG-member societies, are freely accessible through the Standards and Guidelines Repository on the UEG website and via the newly launched GI Guidelines App.¹⁴ Short versions, patient versions, mobile app versions and computerized decision support systems can be made available together with the full guideline document. Translation of the guideline may increase its dissemination among healthcare professionals and policymakers who are not fluent in English. Following publication, it may be useful to identify opinion leaders and involve them in educational outreach in the form of workshops or symposia. Patient representative groups can be notified and policymakers can be approached. Social media may be an effective means for disseminating download links. Strategies should be considered to overcome hurdles in low resource settings. Local adaptation of guidelines, in collaboration with local stakeholders, may help overcome the economic, cultural and infrastructural challenges to implementation in low-income countries.

Guideline updates

In the medium- to long-term period following publication, the steering group should convene at predetermined intervals in order to assess the need for a guideline update (e.g. on the side-lines of scientific meetings). The steering group may wish to form a designated working group for this purpose. Literature searches, as described above, should be performed periodically in order to identify new evidence. The importance and relevance of this evidence should be assessed at guideline update meetings. If new evidence substantially impacts or alters the recommendations of the published guideline such that the guideline is no longer valid, then an update is required.

Conclusion

Clinical practice guidelines have the potential to improve the quality of patient care and population health in the regions where they are applied. However, in order to produce a high-quality guideline, it is crucial to adhere to a transparent and rigorous methodology. UEG supports the development of high-quality guidelines, which are independent, reproducible and implementable. Future UEG-funded guidelines will adhere to this framework.

Appendix 1. Useful links

AGREE II Instrument and Checklist: https://www.agreetrust.org/agree-ii/

Guidelines International network (GIN): https://g-i-n.net/library/tools

Institute of Medicine (IOM) publication – clinical practice guidelines we can trust: https://www.ncbi.nlm.nih.gov/books/NBK209539/

Methodological Expectations of Cochrane Intervention Reviews (MECIR standards): https://community.cochrane.org/book_pdf/558

National Institute for Health and Clinical Excellence: https://www.nice.org.uk/process/pmg20/chapter/introduction-and-overview

Oxford Centre for Evidence-Based Medicine 2011 levels of evidence: https://www.cebm.net/wp-content/uploads/2014/06/CEBM-Levels-of-Evidence-2.1.pdf

Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA): http://www.prisma-statement.org/

Scottish Intercollegiate Guidelines Network SIGN 50 guideline developer’s handbook: https://www.sign.ac.uk/media/1644/sign50_2015.pdf

World Health Organization WHO handbook for guideline development: https://apps.who.int/iris/bitstream/handle/10665/145714/9789241548960_eng.pdf?sequence=1&isAllowed=y

Appendix 2. Examples of guidelines and standards of practice by UEG member societies

EASL: https://www.journal-of-hepatology.eu/article/S0168-8278(18)32493-0/fulltext

ESCP: https://www.escp.eu.com/guidelines/processes-and-methods

ESGAR: https://link.springer.com/article/10.1007%2Fs00330-019-6002-9

ESGE: https://www.thieme-connect.com/products/ejournals/html/10.1055/a-1067-4657

ESPEN: https://www.clinicalnutritionjournal.com/article/S0261-5614(15)00185-5/fulltext

ESPGHAN: http://www.espghan.org/societal-papers/

Declaration of conflicting interests

The authors have no conflicts of interest to declare.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

ORCID iDs

Doron Boltin https://orcid.org/0000-0003-3357-613X

Stefanos Bonovas https://orcid.org/0000-0001-6102-6579