The review methods followed the general principles outlined in the Centre for Reviews and Dissemination guidance28 for conducting reviews in health care (Reviews 1–4), Thomas and Harden’s29 ‘Methods for the thematic synthesis of qualitative research in systematic reviews’ (Review 3) and the Joanna Briggs Institute mixed-methods guidance30 (Review 3). The reviews are reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) checklist.31
Search strategy
The search strategies were designed by an experienced information specialist (MM) in collaboration with the wider review team and stakeholders to ensure that the search terms (particularly the opioid terms) were relevant to our requirements (e.g. relevant to NHS clinical practice). The following information sources were searched: databases, websites, academic repositories, references of topic relevant systematic reviews, citation searches, reference lists of included studies, trials registries (Review 1 only) and suggestions from experts and clinicians working in the field. Table 1 outlines the specific sources searched for each of the reviews. We also used an adapted version of the CLUSTER approach32 using citation searching, tracking lead authors, targeted Google Scholar searches and PubMed-related articles, to identify sibling or related studies.
| Review 1 | Review 2 | Review 3 | |
|---|---|---|---|
| Search sources | Effectiveness | Adverse effects | Barriers and facilitators |
| MEDLINE | ✔ | ✔ | ✔ |
| EMBASE | ✔ | ✔ | ✔ |
| CENTRAL | ✔ | ✔ | ✔ |
| CINAHL | ✔ | ✔ | ✔ |
| PsycInfo® (American Psychological Association, Washington, DC, USA) | ✔ | ✔ | ✔ |
| Science Citation Index | ✔ | ||
| ICTRP Unavailable at this time redirected to primary sources: | |||
| EU-CTRwww.clinicaltrialsregister.eu/ | ✔ | ✔ | |
| ISRCTNwww.isrctn.com/ | ✔ | ✔ | |
| ANZCTRwww.anzctr.org.au/ | ✔ | ✔ | |
| ClinicalTrials.gov | ✔ | ✔ | |
| NIHR be part of research3https://bepartofresearch.nihr.ac.uk/ | ✔ | ✔ | |
| NICE Evidence Search | ✔ | ✔ | ✔ |
| HMIC | ✔ | ||
| Repository searches via Google ScholarCORE.ac.uk3 | ✔ | ||
| British Pain Society Members area | ✔ | ✔ | ✔ |
Three searches were conducted in parallel: Search 1: identified relevant studies for the effectiveness review (Review 1); Search 2: identified relevant studies for the adverse effects review (Review 2); and Search 3: identified papers that were either barrier and facilitator studies or were qualitative studies (Review 3). Studies used to inform the inequalities review (Review 4) were identified from the studies included in Reviews 1–3.
Initial exploratory searches were developed in an iterative manner in MEDLINE from keywords identified by the review team and published topic-relevant systematic reviews. Keywords included opioid terms AND tapering terms AND pain terms. This structure was then adapted to identify relevant studies for each review according to the relevant eligibility criteria (see Appendix 1 for search strategies). All databases were searched from inception to 11 January 2021. All searches were updated on 20 September 2022. Search results were downloaded into EndNote [Clarivate Analytics (formerly Thomson Reuters), Philadelphia, PA, USA] and records were deduplicated and exported to a specialist systematic review management system, Rayyan33 for screening.
Eligibility criteria
The population, intervention, comparators, outcomes, study designs process was used to frame the inclusion/exclusion criteria for the effectiveness and adverse effects reviews (Reviews 1 and 2) (Table 2). The SPICE framework34 was used to frame the inclusion criteria for the barriers and facilitator review (Review 3) and the inequalities review (Review 4) (Table 3). For the inequalities review (Review 4), the PROGnosis RESearch Strategy partnership-Plus (PROGRESS-Plus)35 framework was used to consider socially stratifying factors that may widen inequalities (Table 4 for defining PROGRESS-Plus characteristics).
| Inclusion | Exclusion | |
|---|---|---|
| Population | Adults (18 years and older) with chronic non-cancer pain and prescription opioid use of at least 3 months | People with a diagnosis of opioid use disorder (opioid addiction) |
| People using illicit opioids | ||
| Intervention | Any intervention (pharmacological or non-pharmacological) focused on managed reduction of opioid use over time | Interventions not focused on managed reduction of opioid use |
| Aim of the study must include a treatment goal of dose reduction or cessation of opioid medication | (Rapid detoxification – such as over 1 up to × days) | |
| Comparator | Any comparator or none | Not applicable |
| Outcomes | Effectiveness | Other reported outcomes |
| Pain intensity/severity, measured using a validated scale (e.g. NRS, VAS) | ||
| Physical functioning, measured using a validated scale (e.g. ODI, BPI) | ||
| Prescribed opioid use | ||
| Psychological functioning/emotional functioning, measured using a validated scale (e.g. BDI, HAD, SF-36) | ||
| Opioid withdrawal symptoms, measured using a validated scale (e.g. ShOWS) | ||
| Quality of life, measured using a validated scale (e.g. EQ-5D, SF-36) | ||
| Dropout rates and reasons | ||
| Mortality | ||
| Patient global impression of change, measured using a validated scale (e.g. PGIC) | ||
| Sleep quality, measured using a validated scale (e.g. Pittsburgh Sleep Quality Index) | ||
| Social or economic activities (e.g. work) | ||
| Rescue treatment use | ||
| Non-prescribed medication use | ||
| Adverse effects | ||
| AEs | ||
| Mortality | ||
| Design | Effectiveness | Effectiveness |
| RCTs | Uncontrolled observational studies and case series with fewer than 25 participants | |
| Controlled observational studies | Case reports | |
| Uncontrolled observational studies (with at least 25 participants) | Qualitative studies | |
| Reviews (systematic, non-systematic) | ||
| Adverse effects | Adverse effects | |
| RCTs | Qualitative studies | |
| Controlled observational studies | Reviews (systematic, non-systematic) | |
| Uncontrolled observational studies (with no limit on participant numbers) | ||
| Case reports | ||
| Limits | Published in 2000 and onwards | Pre-2000 |
| English-language publications | Non-English language | |
| Publication types: editorials, theses, conference abstracts, letters, commentaries |
| Inclusion | Exclusion | |
|---|---|---|
| Setting | Any community or health or social care setting using prescription opioids for management of chronic non-cancer pain | |
| Perspective | Adults (18 years or older) with chronic non-cancer pain and prescription opioid use of at least 3 monthsor | People with a diagnosis of opioid use disorder (opioid addiction) |
| Healthcare providers/carers of patients with chronic non-cancer pain and prescription opioid use of at least 3 months | People using illicit opioids | |
| Intervention | Opioid tapering – any intervention (pharmacological or non-pharmacological) focused on managed reduction of opioid use over time | Interventions not focused on managed reduction of opioid use |
| Comparison | Not applicable | Not applicable |
| Evaluation | Studies that reported (explicit or non-explicit) barriers or facilitators to safe and effective reduced use of prescribed opioids for chronic non-cancer pain from patient, professional and service perspectives | Conference abstracts |
| Qualitative, quantitative, mixed-method or case series study designs as long as they reported barriers/facilitators in the results (rather than reported anecdotally within the discussion) | ||
| Relevant systematic reviews and study protocols were retained only for checking for completed primary studies | ||
| English language only studies |
| PROGRESS-Plus factor | Definition |
|---|---|
| Place of residence | Rural, urban and inner-city places of residence |
| Race, ethnicity, culture and language | Racial, ethnic and cultural background and language (e.g. patient does not speak the same language as the healthcare provider) |
| Occupation | Unemployed/employed/retired; full-time or part-time employment; graded hierarchies measuring occupational status; skilled or unskilled work |
| Gender and sex | Male or female |
| Religion | Religious affiliation, beliefs or values held |
| Education | Level of educational attainment or qualifications; institutions attended (e.g. school/further education/higher education); years in full-time education |
| Socioeconomic status | Poverty or income level (continuous level or proportion falling into income brackets), asset-based measures (e.g. housing tenure) or receipt of state welfare or health payment assistance |
| Social capital | Social support (e.g. marital status, social support from friends, family or provider) or social network |
| Plus factors | Age |
| Comorbidities | |
| Status as long-term opioid user (inequalities in access to opioid tapering only) |
Study selection
A single screening process was adopted across all reviews. Two reviewers independently screened all titles and abstracts to identify studies for inclusion against all review eligibility criteria (see Tables 2 and 3). The full text of papers was retrieved if studies fulfilled the eligibility criteria for at least one review, or if it was unclear from the abstract whether or not the paper fulfilled the inclusion criteria. The full-text papers were independently reviewed by two reviewers. The full text had to fulfil the eligibility criteria for at least one of the reviews (see Tables 2 and 3) to be included. In the event of disagreement, a third reviewer or clinical expert was consulted.
Quality assessment
For each review, quality assessment was undertaken by one reviewer and cross-checked by a second reviewer. Any disagreements about risk of bias were resolved by discussion, or by consulting with a third reviewer.
For the effectiveness and AEs review (Reviews 1 and 2), randomised controlled trials (RCTs) were assessed using the Cochrane tool for assessing risk of bias in randomised trials,36,37 whereas cohort studies and case-control studies were assessed using the relevant Critical Appraisal Skills Programme (CASP) checklists.38
For the barriers and facilitators review (Review 3), the methodological limitations of qualitative studies and qualitative data from mixed-methods studies were assessed using an adapted version of the CASP Qualitative tool39 based on the following CASP domains: context, sampling strategy, data collection, data analysis, support of individual study findings in the underlying data, reflexivity, ethical considerations and other concerns. For the quantitative studies, the relevant CASP tool was applied.38,40 CASP tools were modified to use ‘can’t tell’ when insufficient information was available to make a judgement (i.e. a reporting issue) and an additional category was added (‘somewhat’) and used when relevant information including both strengths and limitations was reported (i.e. a methodological issue).41 We made a judgement on the overall assessment of the limitations of the study42 as follows:
- where assessments for most items in the tool were ‘yes’, this indicated no or few limitations
- where assessments for most items in the tool were ‘somewhat’ or ‘cannot tell’, this indicated minor limitations
- where assessments for one or more items in the tool were ‘no’, this indicated major limitations.
Data extraction
Data from the included studies were entered into a pre-piloted data extraction form tailored to the effectiveness and AEs reviews (Reviews 1 and 2) and the barriers and facilitators review (Review 3). Data were extracted by one reviewer and independently checked for accuracy by a second reviewer. Data from multiple publications of the same study were extracted and reported as a single study. The data extraction elements for the effectiveness and AE reviews (Reviews 1 and 2) and barrier and facilitators review (Review 3) are presented in Appendices 1 and 3. Relevant data extracted and included in these reviews were examined for evidence of inequalities in access to, acceptability of and benefit from interventions to reduce use of opioids (Review 4).
Data synthesis
Effectiveness and adverse effects reviews (Reviews 1 and 2)
Meta-analysis was not possible for this review because of significant differences in study designs, methods, timing of assessment and definition of outcomes. Consequently, a narrative synthesis was performed. Outcomes are presented in evidence tables with an accompanying description of the evidence.
Barriers and facilitators review (Review 3)
Qualitative data were analysed following Thomas and Harden’s three-step approach to thematic synthesis of qualitative research in systematic reviews:29 (1) coding of text line by line; (2) development of descriptive themes; and (3) development of analytical themes. First, text was coded on a line-by-line basis. An iterative approach to coding was undertaken and codes were reviewed and revised as new data emerged. A single reviewer coded the data, and a second reviewer checked the assigned codes. Any discrepancies were resolved by discussion. Second, the codes were reviewed and arranged into related descriptive themes. Finally, analytical themes were derived from the related descriptive themes in collaboration with the wider review team. A convergent integrated approach to synthesis was adopted that was consistent with the Joanna Briggs Institute mixed-methods guidance of ‘qualitising’ quantitative data, that is ‘qualitizing refers to quantitative data being converted into themes, categories, typologies, or narratives’.30
Confidence in review findings
Currently, there is no guidance on how to assess confidence/certainty in mixed-methods reviews. By transforming quantitative data into qualitative data,30 we aimed to be inclusive when assessing confidence in the review findings. Confidence in the Evidence from Reviews of Qualitative Research (CERQual) uses four components to assess confidence in review findings: methodological limitations, adequacy of data, coherence and relevance. For each review finding, each component was assessed as either having ‘no, or very minor’, ‘minor’, ‘moderate’ or ‘serious’ concerns. An overall assessment of confidence in the review findings was then assessed as ‘high’, ‘moderate’, ‘low’ or ‘very low’.43,44,45,46,47 Confidence in the review findings were assessed using the information available in the thematic analysis and results from the CASP quality assessment.
Mapping barriers and facilitators to theoretical frameworks
Identifying barriers and facilitators is enhanced by using theoretical frameworks to guide the process.48,49,50 Framing barriers and facilitators within theoretical domains can help explain why an intervention works49,51 and can aid in the identification of tailored implementation strategies for successful opioid tapering. Opioid-tapering initiatives are more likely to be effective if they target the determinants (barriers and facilitators) of tapering. To facilitate implementation of the review findings, we mapped identified barriers and facilitators to opioid tapering in people with chronic non-cancer pain to two theoretical frameworks; the Theoretical Domains Framework (TDF)52,53,54 and a modified version of the Levesque et al. conceptual framework of access to health care.55,56
Theoretical domains framework
The TDF version 2 [TDF(v2)]52,53,54 was used to identify what affective, cognitive, social and environmental barriers and facilitators to tapering exist. The TDF(v2) is a validated framework based on psychological theory that has been used to identify determinants of behaviour change. The TDF(v2) consists of 84 constructs organised into 14 theoretical domains (knowledge; skills; social/professional role and identity; beliefs about capabilities; optimism; beliefs about consequences; reinforcement; intentions; goals; memory, attention and decision processes; environmental context and resources; social influences; emotions; and behavioural regulation) focusing on individual- and group-level behaviours derived from 33 behaviour change theories. The TDF(v2) offers advantages for assessing barriers and facilitators to interventions that operate at individual, group/team or organisational level. This framework has been used to guide recent evidence synthesis on barriers and facilitators to opioid monitoring and deprescribing,26 a recent realist review57 on opioid tapering and in primary research on the use of opioid prescribing.58
We followed the guidance on analysing data using the TDF(v2) of behaviour change to investigate implementation problems.59
Developing a coding guideline
To increase the reliability of coding, we developed a coding framework. One reviewer developed the coding framework; a second reviewer checked the coding. Any discrepancies in the interpretation of the codes were resolved by discussion.
Deductive analysis
Data were extracted and coded according to which TDF domain they were judged to represent. Coding was undertaken in an iterative manner, with modifications made upon familiarisation with the data.
Inductive analysis
If data did not fit any of the TDF domains, a new theme was created.
Identifying important domains
Barriers and facilitators (coded to constructs) were rank-ordered according to the numbers of studies that reported them to identify the most common determinants to opioid tapering. To facilitate recommendations for practice, barriers and facilitators within each TDF domain were organised using the socio-ecological model (SEM).60 The SEM allowed us to highlight barriers and facilitators to opioid tapering within multiple levels of the healthcare system (i.e. at the individual, interpersonal, organisational and environmental levels).
Levesque’s conceptual framework of access to health care
Review findings were also mapped to a modified version of Levesque’s conceptual model of access to health care56 for two reasons. First, to help gain a deeper understanding of whether barriers and facilitators arise from dimensions of accessibility of services, from the abilities of healthcare workforce (provider-related barriers and facilitators) and/or from the abilities of patients to interact with the dimensions of accessibility to generate access (patient-related barriers and facilitators). Second, the model allows for the exploration of barriers and facilitators at various time points in the patient’s care pathway, thereby enabling a greater understanding of when barriers and facilitators may arise during the opioid-tapering process. A modified version56 of Levesque et al.’s. conceptual framework55 was used as it recognised the ‘“human fit” between the needs and abilities of the population with the capacity and abilities of the healthcare workforce’,56 (p. 7) highlighting that ‘organisational structures and processes are mediated by individuals with different capabilities and with varying capacity to engage’56 (p. 4). Thus, the modified version allowed for consideration of barriers relating to the abilities of providers (e.g. confidence in their ability to taper, fears of difficult patient interactions) as well as the abilities of patients.
Inequalities review (Review 4)
Due to the clinical heterogeneity observed in opioid-tapering outcomes in studies reporting quantitative data, we adopted a two-stage narrative approach to synthesis. In stage 1, all studies included in Reviews 1–3 were reviewed for PROGRESS-Plus characteristics (see Table 4) reported in baseline characteristics and PROGESS-Plus characteristics controlled for, or adjusted for, in statistical analysis. In stage 2, only studies that explored one or more of the following were narratively synthesised in detail:
- stratifying effect analyses by different categories of a PROGRESS-Plus characteristic
- associations between PROGRESS-Plus characteristics and changes in opioid use or tapering trajectory following a tapering intervention
- the impact of opioid-tapering initiatives on PROGRESS-Plus-related outcomes (e.g. ability to work)
- any patient PROGRESS-Plus factor identified as a barrier or facilitator to opioid-tapering interventions or any PROGRESS-Plus factor associated with participation (e.g. number of dropouts/completers, reasons for dropout).
Data within each PROGRESS-Plus factor were then organised using the SEM60 to present the findings. The SEM allowed for highlighting of inequalities in the effectiveness of, and access to, opioid tapering within multiple levels of the healthcare system. Findings were also mapped to a modified version of Levesque’s conceptual model of access to health care55,56 for two reasons. Firstly, to help gain a deeper understanding of whether inequalities in access arise from dimensions of accessibility of services or abilities of healthcare workforce and/or from the abilities of patients to interact with the dimensions of accessibility to generate access. Secondly, the model allows for the exploration of inequalities at various time points on the tapering pathway, thereby enabling a greater understanding of when inequalities may arise during the opioid-tapering process.
Stakeholder engagement
To explore the completeness of and to contextualise the findings of the evidence reviews, a summary of findings was presented to stakeholders and discussed at an online workshop using an approach based on NICE fieldwork methods.61
Deviations from the protocol
We specified in the protocol that we would use an emerging method, qualitative comparative analysis (QCA) to complement the synthesis of quantitative and qualitative evidence. The QCA would have been used to identify components of interventions that were linked to the safe reduction of opioids. However, QCA was not feasible as we did not find sufficient evidence. The heterogeneity and reporting of the evidence also hampered our use of other novel methods specified in the protocol (harvest plots and a menu of intervention effects) that would have provided a visual extension to the results of the synthesis.
When we developed the project, two UK intervention trials, I-WOTCH and the EMPOWER study, were in progress. We intended to have access to the I-WOTCH study data, but this was not forthcoming, and the EMPOWER study did report within the time frame for inclusion of data in our project.