Evaluating a primary care pharmacist-led intervention to reduce opioid use for persistent non-cancer pain: the PROMPPT cluster randomised controlled trial protocol

Background and rationale

Persistent pain, defined as long-term pain not caused by cancer, is a common problem affecting an estimated one-third to one-half of adults in the United Kingdom (UK)¹ and around one in five adults in the United States (US)². Opioids are commonly prescribed for persistent pain^3,4, despite a lack of evidence for their long-term efficacy and the risk of serious harm^5–7. A recent review estimated that almost one-third (31%) of patients with persistent pain are prescribed an opioid⁴ and reducing inappropriate opioid use is a key healthcare priority in many countries⁸. Most long-term opioid prescribing for persistent pain occurs in primary care and best practice guidelines recommend regular review of patients on opioids and gradual tapering of opioids (opioid deprescribing) if treatment goals are not met⁸. However, research suggests that opioid deprescribing is viewed as challenging and time-consuming by healthcare professionals⁹ and, with limited time available in UK general practitioner (GP) appointments, approaches to reviewing opioids in primary care are variable and often ‘light- touch’¹⁰. Furthermore, there is a lack of practical guidance on how to implement opioid deprescribing.

Systematic reviews have found limited evidence, often of low methodological quality, supporting interventions to reduce long-term opioid use and recommended further, larger randomised controlled trials (RCTs) of theoretically grounded behaviour change interventions to support opioid deprescribing in the context of persistent pain, which measure long-term follow-up and include a health economic evaluation^11–13. Sandhu et al. evaluated a multicomponent community-based group intervention, incorporating individual tapering support, to reduce opioid use (iWOTCH¹⁴). They found that, among UK patients with persistent non-cancer pain, a group-based educational intervention significantly reduced opioid use compared with usual care without affecting perceived pain interference with daily activities¹⁴. However, less than half of participants randomised to the group intervention attended all sessions. Given the diversity of patients who are prescribed opioids for persistent pain, not all of whom will be able or willing to engage with a group intervention, a range of evidence-based strategies to reduce opioid use are likely to be needed¹⁵.

To address the need for primary care-based interventions to support opioid deprescribing in the context of persistent pain, we developed the PROMPPT intervention (Proactive clinical Review of patients taking Opioid Medicines long-term for persistent Pain led by clinical Pharmacists in primary care Teams). PROMPPT capitalises on an expansion in the number of clinical pharmacists working in UK general practices (practice pharmacists), who play an increasing role in managing patients on long-term medicines by conducting consultations with patients in GP surgeries¹⁶. In line with Medical Research Council (MRC) guidance on the development and evaluation of complex interventions, the PROMPPT intervention was co-designed with stakeholders¹⁷ (patients and healthcare professionals), using a person-based approach¹⁸, combined with best practice guidance and theory on behaviour change. Extensive intervention development work included interviews with patients, pharmacists and GPs, an online qualitative study and in-practice testing of prototype PROMPPT consultations^19–21, followed by a non-randomised feasibility study²². At each stage, findings informed iterative refinement of the PROMPPT review and pharmacist training. We hypothesise that the PROMPPT intervention will increase the likelihood of patients with persistent pain reducing their opioid use, compared to usual care, and that this can be achieved without increasing pain and pain-related interference.

Primary objective

Secondary objectives

Inclusion criteria

Exclusion criteria

Recruitment and consent for outcome data collection

Participating practices search their electronic health record systems, using a Search and Report, designed by a health informatics specialist from Keele Clinical Trials Unit (CTU), to systematically identify adults prescribed opioid-containing pain medicines for 6 months or longer (with at least one prescription dispensed in the last two months), grouped according to the strength of opioid medicine (weak, intermediate, strong, see Table 2), based on a published categorisation for prescribed analgesics in primary care²⁸. Practices then screen patients identified according to study inclusion and exclusion criteria, excluding ineligible patients.

To avoid individual practices being overrepresented in the overall study population, the maximum number of patients invited to participate per practice is capped. Initially a maximum of 250 eligible patients per practice were invited, but the cap was increased to 350 from July 2023 to facilitate recruitment within the planned timeframe (Amendment 7, see Table 1). Where practices identify an excess of potentially eligible patients, a representative sample (randomly sampled by Keele CTU staff to reflect the distribution across the opioid strength groups (weak, intermediate, strong) in the practice overall) is invited by the GP practice.

Using general practices’ existing systems for bulk text messaging, potentially eligible patients with a mobile phone number on the practice record are sent a text message from their practice inviting them to participate in the Management of Opioids & Persistent Pain (MOPP) Study. The text message contains a secure online link to a web page, hosted by Keele University, with an embedded participant information leaflet, consent form and baseline questionnaire for online completion. A reminder text is sent after 2 to 3 days and includes information on how to contact Keele CTU for assistance completing the online questionnaire, or to request a postal version of the information leaflet, consent form and questionnaire. Potentially eligible patients who cannot be invited by text are sent a postal invitation pack via Docmail, a secure remote document compilation, print and mailing solution. To facilitate inclusion of patients, Keele CTU offer telephone support for questionnaire completion and reassurance that family members can assist; and for patients whose first language is not English, questionnaire completion can also be facilitated online by an interpreter on request.

Given the cluster RCT design, patient participants are not individually consented to randomisation; written informed consent for practices to participate in the trial is provided by the general practice lead acting as ‘guardian’ for patients in their care.

Written consent is obtained from patient participants for:

Written consent (optional) from patient participants is also requested for:

PROMPPT intervention

The PROMPPT intervention, comprising a proactive practice pharmacist-led review for patients who have been prescribed opioids for at least 6 months and an associated pharmacist training package, is described below, informed by the TIDieR guidance²⁹. Both the approach to delivering the PROMPPT review, and the content of the training package were shaped by earlier intervention development work^17,19–22. They incorporate best practice guidance^30,31, contemporary behaviour change theories and key communication skills to support shared decision-making^32–35. The intervention aims to respect patient values and autonomy, ensuring that decisions about when and how to change are patient centred.

PROMPPT review. General practices in the intervention arm invite MOPP study participants, by letter, to arrange a review with the practice pharmacist. The invitation letter is accompanied by an information leaflet called ‘Getting Ready for Your Pain Review’. This leaflet aims to inform participants about what to expect at the initial review consultation and help them identify their priorities to discuss with the pharmacist during the consultation. The initial PROMPPT consultation (approximately 30 min) is conducted face-to-face or remotely (by video or telephone) according to patient preference.

The consultation begins with a holistic assessment of the participant’s pain and the impact of persistent pain on their life, followed by a personalised discussion to explore the participant’s personal experience of the effects (wanted/unwanted, useful/bothersome) of opioids. Pharmacists are encouraged to explore participants’ reasons for considering changing their use of opioids, their readiness to change and any ambivalence, before agreeing an individualised management plan.

The agreed plan may include opioid tapering, but this is not mandatory, for example if a participant obtains continued useful benefit from moderate opioid doses without experiencing troublesome side-effects. Where changes to medicines are agreed, goal setting is used to facilitate translation of intentions into action. Important barriers to reducing opioids specific to the individual, such as fear of pain worsening and/or withdrawal symptoms following opioid reduction, are addressed. Practice pharmacists are encouraged to signpost to information resources, which are hosted on a dedicated website. These resources include strategies for living well with pain and reducing opioids. Pharmacists may also signpost or refer to appropriate community services (for example physiotherapy and mental health support services) and, if needed for more complex cases, discussion/collaboration with the GP and/or referral to specialist services is advised. Follow-up is encouraged and arranged according to pharmacists’ clinical judgment and participant preference. Follow-up appointments, conducted face-to-face or remotely by video or telephone, according to pharmacists’ assessment of clinical need and patient preference, are anticipated to be shorter in duration (up to 15 minutes) on average. Pharmacists are also encouraged to agree and document a clear plan for contact, if needed, between scheduled appointments. A written summary of the agreed plan (‘Your Pain Action Plan’), including a link to the information resources webpage, is provided in either paper or electronic form, according to patient preference.

Practice pharmacists delivering PROMPPT are independent prescribers and have completed trial-specific training to deliver the PROMPPT intervention, as described below. Intervention fidelity is assessed in the process evaluation (described in detail below) using pharmacist completed intervention delivery templates and audio-recorded consultations.

Pharmacist training. A comprehensive training package, developed by experts in pain management, primary care clinicians, behaviour change experts and medical/pharmacy educationalists, aims to equip practice pharmacists to deliver PROMPPT reviews. The training comprises an e-learning course and two half-day online workshops delivered via Microsoft Teams. The e-learning course, delivered via the PROMPPT study website, includes video presentations by experts in pain management and behaviour change, as well as videos of four simulated consultations aligned with readiness to change³⁵.

The training covers communication skills, how to conduct person-centred discussions about persistent pain and opioids, negotiating treatment plans, creating opioid tapering plans, optimising non-opioid pain management, supporting patients to live well with persistent pain, signposting to patient information resources and when to seek help (e.g. from a GP). Training also includes guidance to help practice pharmacists manage any emotional impact of conducting pain reviews.

The online workshops are facilitated by members of the research team with clinical and academic expertise in persistent pain management and behaviour change, supported by clinical champions, who are experienced clinical pharmacists and have completed the PROMPPT e-learning and received additional training in mentoring and supporting reflective practice. The online workshops incorporate role play, in simulated consultations, with feedback from the facilitators. In addition, practice pharmacists delivering PROMPPT reviews are mentored during the trial by one of the clinical champions.

Practice pharmacists are also provided with study-specific training including training on completion of study documentation, good clinical practice as applicable to research, and the maintenance of the study site file and study records. Reporting of serious adverse events and adverse events will also be covered.

Comparator - usual NHS primary care

MOPP study participants in control arm practices receive usual primary care arrangements for reviewing patients on long-term opioids for persistent pain, which are known to be variable¹⁰. Such reviews may be conducted by the GP or by another appropriately qualified healthcare professional and may take place as a face-to-face consultation or as a virtual review with the patient’s medication assessed from practice records without them being present.

Participating practices all have at least one practice pharmacist working in the practice at least some of the time. The roles adopted by practice pharmacists are variable, often including reviews for specific long-term conditions such as hypertension, diabetes and asthma, and Structured Medication Reviews (SMRs) for patients who may benefit from these, for example due to potentially problematic polypharmacy and / or use of potentially dependence forming medicines³⁶. Since the PROMPPT trial was conceived, reducing high-risk opioid prescribing has been included in NHS England’s Medicines Safety Improvement Programme and, as a result, practice pharmacists may increasingly be conducting SMRs with patients who are prescribed opioids, but there is no specific training available regarding SMRs in this population and it remains unlikely that a practice pharmacist in a control practice would offer a structured medication review in the same way as a PROMPPT review without having received the PROMPPT training. Usual care arrangements are investigated as part of the process evaluation (described in detail below) to explore whether and how this affects outcomes.

Post-trial care

All trial participants receive usual care from their treating primary care clinician(s).

Individual participant data

Baseline and follow-up questionnaires. The baseline and follow-up questionnaires are designed to collect information on descriptive characteristics of the participants and primary and secondary outcome measures (see Table 3). Patients are informed in their study invitation that they have been invited to take part in the Management of Opioids & Persistent Pain (MOPP) Study because they have long term pain and have been regularly prescribed opioid medicines.

To maximise data collection at all follow-up timepoints, non-responders (online) receive a reminder by email (with text alert) after 1 week and a postal questionnaire after 2 weeks. Non-responders (postal) receive a reminder postcard after 2 weeks and a new paper questionnaire after 4 weeks. All non-responders to the reminders are contacted by telephone by a blinded trial administrator for minimum data collection (MDC) after 6 weeks. A brief MDC questionnaire is sent by post to those who cannot be contacted after 3 telephone attempts (Amendment 4, Table 1). MDC collects primary and key secondary outcome data (opioid and non-opioid pain medicines use, Brief Pain Inventory (BPI) and EQ-5D-5L) along with date of birth and sex to ensure the data are provided by the intended participant. Participants opting to complete the questionnaire online are advised, in the participant information leaflet, that they can stop completing the questionnaire at any time point, without giving a reason, but that answers entered up to that point are retained unless they contact the study team to opt out of this (Amendment 4, Table 1). Where there are ambiguous or incomplete questionnaire entries that interfere with calculation of daily opioid use, a member of Keele CTU staff attempts to contact participants by telephone or in writing to obtain the missing data. Incomplete or ambiguous questionnaire entries regarding prescribed pain medicines can also be checked with the electronic health record, to obtain the missing data e.g. regarding drug strength, if necessary.

Medical Record Review (MRR). With participant consent, data is extracted from electronic practice health records to capture data on their prescriptions from baseline (defined as the period 90 days up to the date of their recruitment) up to the date of their 12-month follow-up to facilitate checking of ambiguous or incomplete questionnaire entries regarding pain medicines use if needed. MRR also captures information about PROMPPT-related consultations.

Practice-level anonymised and aggregated data of pain-related prescribing

Each participating practice provides anonymised electronic prescribing data for registered adult patients, aged ≥18 years, who are not receiving palliative care for cancer (defined as patients without both a code for cancer and for palliative care). We will compare prescribing of opioids and non-opioid medicines commonly prescribed for patients with persistent pain (specifically paracetamol, topical pain treatments, non-steroidal anti-inflammatory drugs (NSAIDs), nefopam, gabapentinoids (gabapentin, pregabalin), antidepressants, benzodiazepines and Z-drug hypnotics (zopiclone, zolpidem) at baseline (defined as the period 90 days up to the date the first participant was recruited at that practice) and 12-month follow-up (defined as the period 90 days up to the date 12 months after the first participant was recruited at that practice).

Co-primary outcome measures

When evaluating interventions aimed at reducing opioid use it is recommended to also measure the impact on pain⁴⁷, therefore we chose two co-primary outcomes, measured at 12-month follow-up:

1.   Reduction in opioid use (participant achieves at least a 25% reduction in daily morphine equivalent dose (MED) from their baseline (binary outcome, yes or no))

AND

2.   Non-inferiority of the Brief Pain Inventory (BPI)³⁷ total score

Members of the research team calculate a daily MED from self-reported pain medicines use data in participant questionnaires using a calculator created in Microsoft Excel (2016) by lead author JA. The calculator uses published opioid conversion factors^48–50 (see Table 4) and was refined following testing with members of the research team (SAH, TH, SW, RK) in a feasibility study²². As there is currently no accepted definition of a clinically meaningful reduction in opioid use^47,51, it was agreed with the study funder that, based on clinical experience, a 25% reduction in opioid use from their baseline reflects a clinically reasonable minimum threshold across a range of opioids and doses.

Defining the non-inferiority margin for the BPI total score

The BPI is a composite measure of pain intensity and pain-related functional interference, recommended for use in chronic pain trials and validated in primary care populations with persistent pain^52,53. The total score reflects the multidimensional impact of pain and has comparable responsiveness to the BPI pain intensity and interference subscales in this population^54,55. In defining the non-inferiority margin for the BPI total score we considered the following: (1) Feedback from patients that pain and pain-related interference does not noticeably increase as a result of reducing opioids and, (2) the Initiative on Methods, Measurement, and Pain Assessment in Clinical Trials (IMMPACT)^52,53 international consensus recommendations on interpreting the clinical importance of treatment outcomes in chronic pain trials. IMMPACT identified a 1-point change on the BPI pain intensity (0-10) scale as the smallest noticeable change in pain and a 1-point change on the BPI pain interference (0-10) scale as a clinically meaningful change⁵³. Therefore, our non-inferiority margin for the BPI total score needed to be less than 1.0. We chose 0.6 as a difference that is considered clinically non-inferior and aligns with the standard error of measurement of the BPI total score⁵⁴.

Our non-inferiority margin translates to an effect size of 0.3 (0.6/2) if a standard deviation for the BPI of 2⁵³ is assumed. We chose not to lower the non-inferiority margin further, for example to give a small effect size of 0.2 (based on effect size criteria from Cohen⁵⁶), as recent IMMPACT guidance on evaluating opioid reduction in clinical trials suggests that a somewhat larger non-inferiority margin can be considered for pain in ‘opioid sparing’ trials because of the potentially large benefits of reducing opioid use (e.g. reducing opioid-related adverse effects)⁴⁷.

Secondary outcome measures

To facilitate comparison and pooling of our results, our secondary outcomes are informed by IMMPACT recommendations for trials of interventions for persistent pain and opioid sparing^47,52,53. These include:

All outcome measures and data collection timepoints are presented in Table 3.

Health economics outcome data

Questions on additional persistent pain-related NHS and private practice health care appointments are included in the 3, 6 and 12-month follow-up questionnaires. The health care resource use questionnaire was developed for this research programme, tested in the feasibility study²² and refined considering the findings. Resource use information is obtained on primary care visits, (e.g. general practitioners, practice nurses, and physiotherapy), visits to other health care professionals (e.g. osteopath, chiropractor, psychologist / counsellor), pain medicines (prescribed and non-prescribed) including opioids, non-opioid analgesics and drugs to manage opioid-related side-effects e.g. laxatives and antihistamines, tests and investigations, non-drug treatments (e.g. acupuncture, injections), secondary care consultations, inpatient stays, and surgery during the previous 6 months. Given the nature of persistent pain, broader costs related to both out of pocket costs (e.g. over the counter medicines such as laxatives) and productivity losses (e.g. time-off work related to pain and reduced work performance (presenteeism) are also included. Information on time off work, occupation, typical work activities and the nature of their employment (full time or part time) is requested, and the Single-Item Presenteeism Question from the Work Productivity and Activity Impairment Questionnaire^42,43 is used to estimate productivity losses relating to presenteeism. The average wage for each respondent will be identified using UK Standard Occupational Classification coding and annual earnings data for each job type. The outcome of interest for the economic analysis is the quality-adjusted life years (QALYs) and these will be calculated using EQ-5D-5L⁴¹ responses obtained at baseline, 3, 6, and 12-months follow-up.

Baseline data and flow of patients

A CONSORT flow diagram will be produced to document the flow of clusters and participants through the trial and will include reasons for withdrawal if given. Descriptive statistics will be used to describe key patient and practice-level baseline variables overall, by treatment arm (to explore balance of characteristics across randomised treatment arms) and by whether participants returned a 12-month follow-up questionnaire (to explore the impact of loss to follow-up).

Co-primary outcome analysis

Primary analysis will use multilevel linear (BPI) and logistic (opioid reduction) regression to account for the clustering in the data and will compare the effectiveness of providing the PROMPPT intervention with usual primary care arrangements for review of patients prescribed opioids for persistent pain at 12 months, adjusting for baseline BPI score (for the BPI outcome) and baseline daily MED (for the opioid reduction outcome). The superiority and non-inferiority hypotheses will be assessed simultaneously; superiority will be concluded if the 95% two-sided confidence interval for the opioid reduction treatment effect does not contain one and non-inferiority concluded if the upper limit of the 95% two-sided confidence interval for the mean difference in BPI score is <0.6.

The primary analysis will be conducted using data from all clusters and patients randomised into the trial. A secondary analysis will be conducted to compare the effectiveness of accessing the PROMPPT intervention compared to usual primary care in participants who attended a PROMPPT review.

Secondary outcome analysis

Secondary (longitudinal) multilevel analyses, based on self-report questionnaire data, will explore whether change in the BPI total score (the co-primary outcome measure) and the secondary outcomes (BPI pain, BPI pain interference, daily MED, non-opioid pain medicine use, depression, anxiety, pain self-efficacy and treatment side effects) differ by treatment arm over time (i.e. using data from all time-points). The models will include a term to account for clustering by general practice and a time by treatment interaction. The beta coefficient and 95% confidence interval for the interaction term will be reported and graphical methods used to display the nature of any interactions found in the data. The continuous outcome, percentage change in daily MED (baseline to 12-month follow-up), will also be analysed using multilevel linear models, but as the data will be a single observation per participant, this model will not include a term for “time” in the multilevel model.

Descriptive statistics and 95% confidence intervals will be used to describe the aggregate data from the pseudonymised electronic health records at baseline and 12-months, to include a comparison of opioid and non-opioid analgesic use by treatment arm and to explore within-practice changes in between baseline (defined as the period 90 days up to the date the first participant was recruited at that practice) and 12-month follow-up (defined as the period 90 days up to the date 12 months after the first participant was recruited at that practice).

Subgroup analyses

Subgroup analyses will explore whether the conclusions of the primary analysis for opioid reduction depends on baseline levels of pain self-efficacy and baseline daily MED. This will be explored by including an interaction term in the primary model between treatment and the subgroup covariate of interest, and by running separate models for each covariate.

Adjusted analysis

The primary analysis will be adjusted for key a priori covariates specified in the analysis plan but are likely to include (1) stratification factors included in the randomisation process and (2) key patient-level variables likely to impact on outcome e.g. deprivation, age and sex.

Missing data

Missing data will be handled by using multi-level models as the analysis method or by multiple imputation of missing data as appropriate.

Health economics

An incremental cost-utility analysis will be undertaken from a base-case NHS/Personal Social Services (PSS) perspective to estimate the cost per quality adjusted life year (QALY) over 12 months follow-up, using patient level data on costs and trial outcomes. Alternative perspectives will include a health care and societal perspectives, taking into account private expenditures and out of pocket cost and productivity losses due to time off work.

The outcome of interest for the economic analysis is quality-adjusted life years (QALYs) and these will be calculated using EQ-5D-5L responses obtained in at baseline, 3, 6, and 12-months follow-up.

In line with updated NICE recommendations, individual EQ-5D-5L responses will be converted into index scores using a mapping algorithm developed by the Decision Support Unit. The 'EEPRU dataset' will be employed to convert EQ-5D-5L questionnaire responses onto the EQ-5D-3L utility scores⁶⁵.

Cost data is likely to have a skewed distribution and therefore a non-parametric comparison of means (e.g., bootstrapping) will be undertaken to estimate 95% confidence intervals around costs. Multilevel modelling statistical techniques, taking into consideration clustering and correlation in the cost and effect data will be considered in line with standard literature. QALYs will be calculated from EQ-5D-5L responses, using the “area under the curve” approach. Unit costs will be applied to all health care resource use items, and mean resource use (for each health care use category) and mean total costs will be calculated for all trial participants. Analysis of productivity losses will use the human capital approach. Multiple imputation will be used to impute all missing values for the EQ-5D-5L and total cost estimates for non-responders to ensure all participants are included in the analysis.

A cost-consequence analysis will initially be reported, describing all the important results relating to costs and consequences (across the full range of clinical and cost outcomes). Subsequently, incremental cost-utility analysis will be undertaken to estimate the incremental cost per QALY gained, adjusting for baseline covariates in line with the primary clinical evaluation. The analysis will be conducted on an intention to treat basis, and in accordance with current cRCT guidelines⁶⁶. This analysis will involve adopting multilevel modelling statistical techniques taking into consideration clustering in the cost and effect data⁶⁷. Separate generalised equation models, controlling for clustering will be used to estimate the mean incremental costs and QALYs for the intervention relative to the control.

The robustness of the results will be explored using sensitivity analysis. This will explore uncertainties in the methods employed to analyse the data (for example, a complete-case analysis as an alternative to using a multiple-imputed data set, and in estimation of costs), and any assumptions made in the analysis. Cost-effectiveness acceptability curves will be generated to reflect the probability the intervention will be cost effective at different cost per QALY willingness to pay thresholds. If the intervention is shown to be clinically and cost-effective within the trial, a simple budget impact analysis will be undertaken to calculate the potential overall cost for a typical primary care practice, in terms of delivering the intervention and potential cost savings from reduced opioid prescribing.

Decision modelling, using a Markov model will also be undertaken to estimate the potential longer-term cost-effectiveness of the intervention. The model-based analysis will be performed from an NHS perspective with appropriate discounting of costs and outcomes at 3.5%. The model will assess the impact of the intervention on quality of life, continued opioid use, escalation of opioid medication, and the potential for opioid-related serious adverse events. The model will be populated using the trial data, literature review of previously published relevant models, routine datasets and where appropriate consultation with clinical experts and Patient and Public Involvement (PPI). The model will be subject to deterministic sensitivity analysis by changing individual parameter values, broadening the perspective and changing model assumptions. Model parameters will be inputted as distributions and probabilistic sensitivity analysis will be undertaken to simultaneously incorporate all parameter uncertainty. Cost-effectiveness planes and cost-effectiveness acceptability curves will be presented to show the probability the intervention is cost-effective at different cost/QALY thresholds. Full details of the decision model methods, time horizon, structure and parameters will be detailed in the final health economics publication.

Identification and recruitment of participants

Patient participants. Participants from intervention arm clusters who provide written consent to being contacted about a future related research study and subsequently attend a PROMPPT review are invited to participate in the MOPP-2 study. Participants are invited to complete an acceptability questionnaire (described below) and/or take part in an interview (up to n=20) and may choose to consent to any, all or none of these options. Additionally, all participants who do not take up the PROMPPT review are invited to participate in an interview to explore their reasons for declining. Written consent is obtained from all participants in the process evaluation.

Clinicians (GPs and practice pharmacists). All practice pharmacists from intervention arm practices are asked to consent to audio-recording of a sample of PROMPPT consultations and to an interview. A sample of GPs from intervention arm practices (up to n=15) and practice pharmacists and GPs from control arm practices (up to n=12) are also invited to take part in an interview.

PROMPPT clinical champions. All PROMPPT Clinical Champions are interviewed to explore their experiences of mentoring practice pharmacists during the delivery of PROMPPT.

Data collection

Audio-recorded consultations. A sample of approximately 10% of PROMPPT consultations are digitally audio-recorded to allow assessment of key aspects of intervention delivery⁷⁰: treatment differentiation (Did the practice pharmacists only deliver PROMPPT and not other treatments?), treatment competency (Did practice pharmacists maintain the skills learned in training?), and treatment delivery (Were PROMPPT components delivered as intended?).

The research team work with participating practices to identify and approach suitable participants who have scheduled a PROMPPT consultation. Informed, written consent to audio-record the consultation is obtained from patient participants by a researcher prior to the consultation and confirmed with practice pharmacists prior to the start of the consultation. This process was used successfully in the feasibility study²². Audio-recording files are securely sent to Keele CTU for analysis.

Case-report forms. Practice pharmacists in the intervention clusters are required to complete an electronic case report form (eCRF) following each PROMPPT consultation. The eCRF collects information on what happened during the PROMPPT consultation (e.g. discussing the patient’s pain story and the impact of their pain, discussion of the patient perspective on making changes to opioid medicine, developing a personalised pain action plan) and will be used to assess intervention fidelity (i.e. Was PROMPPT delivered as intended?) and dose (How much was delivered?).

Semi-structured interviews. Consenting participants in the intervention arm are interviewed after attending at least one PROMPPT consultation. The interviews are based on semi-structured topic guides developed with our PPI group and include questions informed by our programme theories regarding behaviour change (Theoretical Domains Framework (TDF))⁷¹, implementation (Normalisation Process Theory (NTP))^72,73 and acceptability (Theoretical Framework of Acceptability (TFA))⁷⁴. Consenting participants who decline a PROMPPT review, and accept an invitation to speak to a researcher, are interviewed using a semi-structured topic guide that will explore reasons for declining. A similar process is followed for the clinicians, whereby they are interviewed once they have experience of PROMPPT intervention delivery. Topic guides are refined, as needed, during data collection to allow any emerging issues to be explored.

The sample sizes outlined above are expected to be sufficient to achieve information power⁷⁵, providing appropriate range and depth of insights, whilst being achievable within the scope of the project. Similar topics are covered with all interviewees to ensure comparisons between their responses, while at the same time enabling them to reflect on their specific expectations, understandings and experiences.

Informed consent is obtained prior to all interviews commencing. All interviews are conducted remotely via telephone by an experienced qualitative researcher, and audio-recorded with participants’ consent. The interviews are securely sent to a professional transcription company to be transcribed verbatim. Transcripts are checked for accuracy and anonymised by members of the research team ahead of analysis.

Acceptability questionnaire. All participants who attended the initial PROMPPT review are sent a short survey (known as the Acceptability Questionnaire) about their experiences of the consultation. The Acceptability Questionnaire comprises 12 questions from two existing measures: a modified version of the treatment acceptability and credibility measure⁷⁶ and the Theoretical Framework of Acceptability (TFA) questionnaire⁷⁴.

The modified acceptability and credibility measure, which has been used in previous similar studies^77,78, includes 4 items (scored on a 0-10 scale) and will assess (1) how logical the PROMPPT review seems to participants, (2) how confident participants are that it will be successful in helping them, (3) how confident participants would be in recommending it to a friend, and (4) how satisfied they were with the intervention overall.

The TFA questionnaire⁷⁴ comprises 8 items (scored on a 0-5 scale). The first is a global acceptability question and the remainder represent key constructs of acceptability relating to healthcare interventions (affective attitude, burden, intervention coherence, ethicality, perceived effectiveness, opportunity costs, and self-efficacy).

Practice profile form. A practice profile form is completed by all participating GP practices at multiple time points across the trial, including at the beginning and towards the end. It collects information about current practice activity relating to persistent pain management, practice pharmacist workload and responsibilities within the practice, and practice size and organisational characteristics. These data will facilitate assessment of practice context and effects of any variation in context.

Data analysis

Audio-recorded consultations. Researchers from the process evaluation team will examine the audio-recordings from the PROMPPT consultations and use a previously developed tick box score sheet to assess whether components of the consultation intended to be included, and focused on during training, were demonstrated by the practice pharmacists.

eCRFs. The intervention delivery template will be reviewed for each participant and the proportion of times that each component of the intervention was used will be reported. A judgement will also be made as to whether the patient received the intervention as intended, in line with the training, and whether the patient attended all scheduled follow-up appointments. The proportion of patients being treated in line with the training will then be calculated out of all participants who attended the first appointment with the practice pharmacist. For audio-recorded consultations, corresponding eCRFs will be triangulated with the audio-recordings to assess fidelity of the training and intervention.

Semi-structured interviews. The Framework approach⁷⁹ to analysis will be used starting deductively from pre-set aims and objectives, will enable a focused and efficient analysis within the timeframe of this study⁸⁰, and can be used within a multidisciplinary research team setting⁷⁹. Analysis will involve familiarisation, identification of an analytical framework (e.g. study aims, TDF⁷¹, TFA⁷⁴ and NPT^72,73,81), indexing, charting and mapping, and interpretation⁸². Analysis will include members of the research team from different professional backgrounds (e.g. GPs and pharmacists) to increase trustworthiness⁸³. Continuous team analysis of data will help to challenge interpretations, and coding will be refined and agreed through ongoing discussion⁸⁴. The data analysis will be facilitated using qualitative data-analysis software (NVivo: Qualitative Data Analysis Software Version 14⁸⁵.

Acceptability questionnaire. The mean and standard deviation, as well as the median and interquartile range, for the TFA questions and the acceptability/credibility questions will be calculated from items 1-8 and items 9-12 of the Acceptability Questionnaire respectively. A mean score of ≥ 5/10 for each of items 9-12 will be considered the threshold for acceptability/credibility, as in previous trials using this measure^77,78.

Practice profile form. A narrative review of practice profile form responses from participating practices will be undertaken. Responses from multiple time points across the trial will be examined and descriptive summaries produced to facilitate assessment of practice context and effects of variation in context.

Integrated analysis. When analysis of the quantitative and qualitative data is complete, a triangulation protocol^68,86 will be used. This technique enables integration of data in order to investigate completeness, convergence, and dissonance of key themes across datasets⁸⁷. Methods include following a thread and development of a convergence coding matrix. The matrix allows findings from different study components to be displayed side by side. Integration will aid interpretation of findings on implementation, acceptability, fidelity, mechanisms of impact and context.

Internal pilot study

An internal pilot study analysed recruitment, intervention uptake and fidelity of intervention delivery against pre-specified progression criteria between 20 May 2022 and 22 December 2022 from participants in the first batch of GP practices recruited and assessed retention at 3-month follow-up. Recruitment exceeded the pre-specified targets with 14 practices (target 12) and 388 participants (target 350, 50 per month on average) recruited. Fidelity of intervention (PROMPPT review) delivery was good with 89% of pain reviews delivered in line with the training (target 80%) based on the intervention delivery template. Intervention uptake was slightly lower at 62% than the pre-specified target of 70%. However, the first practice to invite participants sent all the invitations at once and had particularly low uptake (32%) which lowered the overall rate. Subsequently, practices were advised to send the review invitations in batches, according to pharmacist availability, which made it easier for practices to monitor uptake and send reminders. Rates of retention at 3-month follow-up were lower than expected, with 106 out of 174 (61%) of participants with sufficient time (8 weeks) to respond returning the 3-month follow-up questionnaire, whilst the sample size estimate assumed 70% retention at 12-month follow-up. These findings were reported to the DMC and TSC (March 2023). Both were satisfied that the trial should continue, and a range of measures were agreed and implemented to improve retention at follow-up and maximise primary outcome data collection at 12 months (See Amendment 4. Table 1). To mitigate any impact of lower follow-up rates on study power we also increased the number of general practices in the trial from 30 to 38 (Amendment 7. Table 1).