Methods

Sections of this chapter have been reproduced from the published OPTION-DM trial protocol.2 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: https://creativecommons.org/licenses/by/4.0/. The text below includes minor additions and formatting changes to the original text.3

Study design

This was a randomised crossover trial of treatment pathways to evaluate the superiority of at least one pathway (i.e. A-P, D-P and P-A) in reducing the 7-day average 24-hour pain in patients with DPNP. Eligible patients were randomised to one of six treatment sequences, with equal allocation to sequences (1 : 1 : 1 : 1 : 1 : 1). Each sequence examined all three treatment pathways in random order. Each treatment pathway consisted of two phases (i.e. 6 weeks’ monotherapy followed by 10 weeks’ monotherapy or 10 weeks’ combination therapy based on response to treatment).

Figure 1 shows a schematic representation of the study schedule. Following the screening, consent and initial washout visits (i.e. weeks –2 to 0), the visits from week 0 to week 16 were repeated until all three treatment pathways were completed. Face-to-face assessments were completed at the week numbers indicated in Figure 1. Weekly telephone calls were carried out between study visits.

FIGURE 1 Study schedule. (Figure not reproduced here; see the original article.)

Trial approvals and registration

The trial was approved by the Yorkshire and the Humber – Sheffield Research Ethics Committee (reference 16/YH/0459) on 9 December 2016. The Medicines and Healthcare products Regulatory Agency issued the clinical trial authorisation on 23 June 2017 (reference 21304/0262/001-0001, EudraCT reference 2016-003146-89).

The trial was registered as ISRCTN17545443.

Participant eligibility criteria

Inclusion criteria

Participants were required to meet all of the following inclusion criteria:

• Age ≥ 18 years.
• Neuropathic pain affecting feet and/or hands for at least 3 months or taking pain medication for neuropathic pain for at least 3 months.
• Bilateral distal symmetrical neuropathic pain confirmed by the Douleur Neuropathique 4 (DN4)38 questionnaire at screening visit. The participant was eligible if four or more questions were answered as ‘yes’.
• Bilateral distal symmetrical polyneuropathy confirmed by a modified Toronto Clinical Neuropathy Score (mTCNS)39 of > 5 points at screening visit.
• Stable glycaemic control [i.e. glycated haemoglobin (HbA_1c) < 108 mmol/mol].
• A mean total pain intensity of at least 4 points on an 11-point NRS (with 0 being ‘no pain’ and 10 ‘worst pain imaginable’) during 1 week off pain medications (i.e. the baseline period). Patients could be invited to attend the randomisation visit sooner if it was clear that their mean pain score for the week was ≥ 4, that is, as soon as the total sum of the pain scores was ≥ 28 points (e.g. randomisation could take place after 3 days if a patient scored 10 on each of the first 3 days of monitoring). This was to minimise the length of time patients remained off neuropathic pain treatments.
• Patient is willing and able to comply with all the study requirements and be available for the duration of the study.
• Patient is willing to discontinue current neuropathic pain-relieving medications.
• Informed consent form for study participation signed by participant.

Exclusion criteria

Patients were not eligible for the study if they met any of the following exclusion criteria:

• Non-diabetic symmetrical polyneuropathies.
• History of alcohol/substance abuse that would, in the opinion of the investigator, impair the patient’s ability to take part in the study.
• History of severe psychiatric illnesses that would, in the opinion of the investigator, impair the patient’s ability to take part in the study.
• History of epilepsy.
• Contraindications to study medications.
• Pregnancy/breastfeeding or planning pregnancy during the course of the study.
• Use of prohibited concomitant treatment that could not be discontinued with the exception of prior concomitant and safe use of selective serotonin reuptake inhibitors (SSRIs) with study medication (duloxetine and/or amitriptyline). Note that concomitant use of citalopram was not permitted.
• Use of a high-dose morphine equivalent (> 100 mg/day).
• Liver disease [i.e. aspartate aminotransferase (AST)/alanine aminotransferase (ALT) two or more times the upper limit of normal].
• Significant renal impairment [i.e. an estimated glomerular filtration rate (eGFR) of < 30 ml/minute/1.73 m²].
• Heart failure (i.e. New York Heart Association ≥ class III).
• Clinically significant cardiac arrhythmias on 12-lead electrocardiogram, current history of arrhythmia, second- or third-degree heart block or left bundle branch block (patients with right bundle branch block or first-degree heart block may be included following discussion with cardiology team).
• Patients with a recent myocardial infarction (< 6 months prior to randomisation).
• Symptomatic postural hypotension that, in the opinion of the investigator, is clinically significant and would be a contraindication to the study medication.
• Prostatic hypertrophy or urinary retention to an extent that would, in the opinion of the investigators, be a contraindication to the study medication.
• Patients with other painful medical conditions where the intensity of the pain is significantly more severe than their DPNP (patients were not excluded if the pain was transient in nature).
• Any suicide risk, as judged by the investigator or as defined by a score of ≥ 2 points on the Suicide Risk Questionnaire.
• Significant language barriers that are likely to affect the participant’s understanding of the medication schedule or ability to complete outcome questionnaires.
• Concurrent participation in another clinical trial of an investigational medicinal product (IMP).
• Major amputations of the lower limbs.
• Foot ulcers, only if, in the opinion of the local principal investigator (PI), they were likely to have a confounding/detrimental effect on study primary outcome or participation (e.g. localised foot pain from the ulcer site).

Withdrawals

Participants could choose to withdraw from the trial treatment or follow-up at any time. The local research team could also choose to discontinue the study treatment for clinical reasons or if the participant’s condition changed following randomisation so that they met one or more of the exclusion criteria. Outcome data were collected up to the end of the current treatment pathway, if possible. Data already collected up to the point of withdrawal were kept.

Settings and locations where the data were collected

Initially, the trial planned to recruit participants from eight secondary care hospital sites across the UK. Additional sites were added during the trial, and a total of 21 sites were opened to recruitment during the course of the trial.

Potential participants were identified directly through the database and via clinics at participating centres, as well as via participant identification centres, podiatry clinics and general practice mail-outs. Recruitment strategies differed between sites to reflect the local organisational structure at each site.

Participant treatment and outcome data collection was carried out by the recruiting hospital site. All research activity at the site was carried out by hospital employees trained in OPTION-DM trial processes.

Screening, assessment of eligibility and consent

Prior to any study procedures being completed, participants were required to give written informed consent for the study. The participant was given sufficient time to ask questions, consider the study and discuss it with family/friends prior to providing consent, which was taken by medically qualified site investigators.

Eligibility for the study was assessed by the local investigator. Participants were required to stop all existing treatment for neuropathic pain, except paracetamol, if applicable. Treatments were tapered, usually over a period of 3 days followed by a 4-day washout period. Participants then entered the baseline period and the pain scores collected during this period were used to determine eligibility.

Randomisation

Randomisation was completed using the Sheffield Clinical Trials Research Unit (CTRU) online randomisation system (SCRAM). Participants were assigned to one of six sequences (allocation 1 : 1 : 1 : 1 : 1 : 1) based on a predetermined randomisation schedule and stratified by site using permuted blocks of sizes 6 or 12. The trial statistician created the randomisation schedule. Each sequence consisted of three treatment pathways given in random order. The three treatment pathways were:

1. A-P (i.e. first-line amitriptyline, second-line pregabalin)
2. D-P (i.e. first-line duloxetine, second-line pregabalin)
3. P-A (i.e. first-line pregabalin, second-line amitriptyline).

Members of the research teams at participating sites were granted access to the SCRAM system with individual usernames and passwords. These members of staff were responsible for performing the randomisation process once eligibility had been confirmed by the local investigator. After the randomisation was completed, the pharmacy department was informed that a new participant had been randomised. A member of the pharmacy team then accessed the SCRAM system to obtain the unblinded treatment allocation for the participant to allow dispensing. Pharmacy staff were assigned a different level of access to SCRAM to ensure that they were the only members of the local site team with access to the treatment allocation information.

Interventions

Investigational medicinal product details

Study treatment was supplied in capsules with dose levels as follows:

• amitriptyline – 25-mg capsules
• amitriptyline – 50-mg capsules
• duloxetine – 30-mg capsules
• pregabalin – 75-mg capsules
• pregabalin – 150-mg capsules
• matching placebo capsules.

Capsules were supplied in bottles containing nine, 23 or 51 capsules.

Participants were instructed to take medication orally before breakfast and at bedtime. Participants on dose levels 1 or 2 took one tablet in the morning and one tablet in the evening. Participants on dose level 3 took two tablets in the morning and two tablets in the evening. The placebo ensured that the same dosing schedule could be followed for each study drug. For example, a participant on dose level 1 of amitriptyline would take one placebo capsule in the morning and one 25-mg amitriptyline capsule in the evening. A participant on dose level 1 of standard dose pregabalin would take one 75-mg pregabalin capsule in the morning and one 75-mg pregabalin capsule in the evening. This ensured that the medication schedule appeared the same to participants, regardless of which medication they were currently taking.

The total daily dose of each drug was dependent on the dose level prescribed to the participant (Table 1). Participants were provided with clear instructions on the dosing schedule, and this was reinforced with written instructions and a medication diary.

Dose titration

Participants were titrated to a maximum tolerated dose level on starting each new treatment. The schedule for dose escalation was the same in each pathway (Figure 2).

FIGURE 2 Dosing and titration schedule for treatment pathways. Each pathway had two treatment phases, each with a 2-week initial titration period towards maximum tolerated dose. Participants continued on maximum tolerated maintenance dose of the drug from the first treatment phase for the duration of the second treatment phase. For patients with an eGFR of ≤ 60 ml/minute, the maximum pregabalin dose was 300 mg/day. (Figure not reproduced here; see the original article.)

When a new treatment was started, all participants started at dose level 1 and the dose was escalated slowly, one dose level at a time, towards a maximum tolerated dose or maximum permitted dose, whichever was reached first (see Table 1). Dose titration decisions were based on treatment response (i.e. 24-hour pain NRS score), side effect profile and participant preference. Dose titrations were usually made during the first 2 weeks of a new treatment; however, investigators were permitted to make dose changes at any time if deemed necessary. At weekly intervals, the site research nurse evaluated the participant’s response to treatment and AEs and this information was used to guide dose titration.

Treatment response

Adequate pain relief was defined as a 24-hour pain NRS score of ≤ 3 for the purposes of dose titration decisions. Participants who experienced adequate pain relief at dose levels 1 or 2 did not have their dose escalated further.

Adverse events

Participants were asked to report all side effects. The local study team graded the side effects as mild, moderate or severe. The participant was asked to rate whether the side effects were tolerable or intolerable. If a side effect was rated as ‘tolerable’ and was non-severe, dose escalation was continued as indicated by the pain score assessment. However, sites were advised that if a patient was experiencing tolerable non-severe side effects, then it was also acceptable to maintain the current dose for a further week to allow side effects to improve before increasing the dose further. This decision was made by the local site team.

If side effects were severe or were rated as ‘intolerable’, then investigators considered reducing the dose by one dose level or discontinuing the medication based on the overall condition of the participant.

Participant preference

Participant preference was taken into account, where possible, when making dose titration decisions. However, the dose was not increased based on participant preference alone (i.e. the dose was increased only if the participant expressed a preference for an increase in dose and if this was indicated based on treatment response and side effect profile).

Treatment phases

Each treatment pathway was split into two treatment phases, as shown in Figure 3.

FIGURE 3 Two treatment phases per pathway. (Figure not reproduced here; see the original article.)

First treatment phase

During the first treatment phase, participants received monotherapy with the first-line treatment in the pathway. This lasted for a total of 6 weeks.

Responder/non-responder assessment

At the end of the first treatment phase, a decision was made either to continue on monotherapy or to start combination therapy with the addition of the second-line treatment in the pathway. This decision was based on the 7-day average pain NRS score during the week preceding the week 6 study visit. Participants were divided into ‘responders’ (with a pain score of ≤ 3 points) and ‘non-responders’ (with a pain score of > 3 points). Responders continued on monotherapy and non-responders commenced combination therapy.

Second treatment phase

The second treatment phase lasted for a total of 10 weeks. Responders continued on first-line treatment as a monotherapy for the remainder of the pathway. Non-responders commenced combination therapy, with the addition of the second-line treatment in the pathway, for 10 weeks.

Taper doses

At the end of a treatment pathway, participants were provided with a taper dose of their current medication. Participants were instructed to take the taper dose for 3 days and then to stop study medication completely for 4 days before commencing the next pathway. The taper dose was one dose level below the maximum tolerated dose, as shown in Table 2.

Where the patient experienced significant withdrawal side effects, the medication could be tapered down more gradually in accordance with the judgement of the investigator. However, the patient was still required to stop the medication completely for at least 4 days before starting the next pathway.

The first and second treatment phases were repeated until the participant had completed all three treatment pathways.

Permitted changes to the treatment schedule

To make the trial as pragmatic as possible, the following changes were permitted to the treatment schedule as needed:

• If there was significant intolerance to first-line treatment, participants were permitted to switch to the second-line treatment in the pathway as a monotherapy. This change could be made immediately at any time and without the need to wash out first-line treatment.
• Non-responders at week 6 who were on dose levels 1 or 2 of first-line treatment were permitted to increase the dose of first-line treatment rather than start combination treatment at the discretion of the local investigator.
• Second-line treatment could be added as a combination therapy up to week 13 if needed (e.g. if the participant was a responder at week 6, but then, subsequently, became a non-responder).
• For participants whose pain scores had not reduced at all at week 6, compared with baseline, or if their pain scores had increased first-line treatment was stopped and second-line treatment was started as a monotherapy for the remainder of the pathway.
• Participants who did not tolerate one treatment pathway could start the next treatment pathway early. In this case, the taper dose was dispensed early and the next pathway started following the appropriate washout period.

Compliance with intervention

Treatment compliance was assessed by the local study team at each study visit via pill counts and this was recorded on a treatment compliance log within the study database. Issues with compliance were discussed with participants and, if needed, participants were re-educated on the study requirements.

Blinding and masking

The study was double-blinded, whereby participants and the local research team were blinded to treatment allocation, with the exception of the site pharmacist who was unblinded. Blinding was maintained with over-encapsulation and matching placebos. As the study drugs have different dosing schedules (e.g. amitriptyline is given once per day, whereas pregabalin is given twice per day), the placebos were used to ensure that the dosing schedule was identical across the three pathways, with dosing twice per day on all treatments. The IMP bottles were supplied with a tear-off label, which the centre’s pharmacist removed prior to dispensing to the participants. Participants and sites were aware of whether monotherapy or combination therapy had been prescribed and of the dose level. Unblinding was considered only in the event of a medical emergency where knowledge of the participant’s treatment allocation would change the clinical management. All participants were unblinded at the end of the study, when the final statistical report was completed.

Data collection and management

Data management was provided by the University of Sheffield CTRU who adhere to their own standard operating procedures relating to all aspects of data management, including data protection and archiving.

Participant confidentiality was respected at all times and the principles of the UK Data Protection Act40 were followed. All participants were assigned a unique study identification number at screening that linked all of the clinical information held about them on the study database. Data were collected on standardised questionnaires and study-specific case report forms (CRFs) and were entered onto the CTRU’s in-house data management system (Prospect). Access to Prospect was controlled by usernames and encrypted passwords, and a comprehensive privilege management feature was used to ensure that users had access to only the minimum number of data required to complete their tasks. This was used to restrict access to personal identifiable data. After data had been entered onto the database, electronic validation rules were applied on a regular basis and discrepancies were tracked and resolved. All entries and corrections were logged, with the person, date and time captured within the electronic audit trail.

Regular site monitoring visits occurred throughout the study and additional visits were undertaken where required. At these visits, the monitor reviewed activity to verify that the data were authentic, accurate and complete. Accurate and reliable data collection was assured by verification and cross-check of the CRFs against investigator’s records (i.e. source document verification). The study monitor contacted and visited sites regularly to inspect CRFs throughout the study to verify adherence to the protocol and the completeness, consistency and accuracy of the data being entered on the CRFs. Monitoring visits also included a pharmacy visit to review processes, documentation and accountability of study drugs. CTRU staff reviewed entered data for possible errors and missing data points. A central review of consent forms was also completed, and sites were requested to post consent forms to CTRU on an ongoing basis. CTRU reviewed pharmacy dispensing logs for some patients centrally.

Study records will be stored for 25 years after the completion of the study before being destroyed.

Outcome measures

The study evaluated the superiority of at least one pathway in reducing the 7-day average 24-hour pain in patients with diabetic neuropathy.

Primary end point

Difference between 7-day average 24-hour pain (evaluated at patient level) among pathways on an 11-point NRS (0 = no pain and 10 = worst pain imaginable), measured during the final follow-up week of the treatment cycle (i.e. week 16). NRS 24-hour average pain is now considered the gold standard for the assessment of neuropathic pain and has been employed in almost all well-designed neuropathic pain studies over the past 10 years.19,28,37

Secondary end points

Efficacy

• Difference in 7-day average 24-hour pain (evaluated at patient level) on an 11-point NRS at week 6 among monotherapies.
• Difference in Short Form questionnaire-36 items (SF-36) physical mean scores (evaluated at patient level) at week 16 among pathways.41
• Difference in SF-36 physical mean scores (evaluated at patient level) at week 6 among pathways.41
• Difference in SF-36 mental mean scores (evaluated at patient level) at week 16 among pathways.41
• Difference in SF-36 mental mean scores (evaluated at patient level) at week 6 among pathways.41
• Difference in Hospital Anxiety and Depression Scale (HADS) mean anxiety scores (evaluated at patient level) at week 6 among pathways.42
• Difference in HADS mean anxiety scores (evaluated at patient level) at week 16 among pathways.42
• Difference in HADS mean depression scores (evaluated at patient level) at week 6 among pathways.42
• Difference in HADS mean depression scores (evaluated at patient level) at week 16 among pathways.42
• Difference in proportion of patients having treatment success (30%) at week 16 among pathways. Treatment success was defined as a reduction in 30% value at follow-up compared with baseline.
• Difference in proportion of patients having treatment success (50%) at week 16 among pathways. Treatment success was defined as a reduction in 50% value at follow-up compared with baseline.
• Difference in Brief Pain Inventory – Modified Short Form (BPI-MSF) measure of pain interference with function total score (evaluated at patient level) at week 6 among pathways.43
• Difference in BPI-MSF measure of pain interference with function total score (evaluated at patient level) at week 16 among pathways.43
• Difference in Insomnia Severity Index (ISI) total score (evaluated at patient level) at week 6 among pathways.44
• Difference in ISI total score (evaluated at patient level) at week 16 among pathways.44
• Difference in Patient Global Impression of Change (PGIC) (evaluated at patient level) at week 16 among pathways.45
• Difference in proportion of care pathway preferred by participants at week 50.

Cost-effectiveness

• The EuroQol-5 Dimensions (EQ-5D) is a routinely used generic HRQoL instrument. It is the instrument preferred by NICE for assessing HRQoL, and the newer five-level instrument [i.e. EuroQol-5 Dimensions, five-level version (EQ-5D-5L)] is more sensitive than the original three-level version.46
• The Client Service Receipt Inventory (CSRI) is an instrument routinely used to capture health resource use and personal expenses. A modified version of the CSRI, where unnecessary questions were removed to reduce participant burden, was used.47

Safety

• The frequency and proportion of patients reporting at least one AE for each of the pathways. In addition, the relationship to intervention (i.e. definite, probable, possible, unlikely, unrelated or not assessable) was reported (frequency and proportion).
• The frequency and proportion of AEs for each of the pathways.
• A list of AEs for each of the pathways.
• The frequency and proportion of patients reporting at least one SAE for each of the pathways. In addition, the following characteristics were summarised (frequency and proportion): intensity (i.e. mild, moderate or severe), relationship (i.e. definite, probable, possible, unlikely, unrelated or not assessable), whether or not a suspected unexpected serious adverse reaction (SUSAR), whether or not resulted in death.
• Frequencies of SAEs for each of the pathways.
• A list of SAEs for each of the pathways.

Subgroup

• Neuropathic Pain Symptom Inventory (NPSI) questionnaire for subgroup analysis relating pain phenotype to treatment response.48 There is emerging evidence that treatment response may be determined by a patient’s pain phenotype.49–51 In particular, the following outcomes were evaluated:
  • Difference in ‘burning (superficial) spontaneous pain’ NPSI mean subscores (evaluated at patient level) at week 6 among pathways.
  • Difference in ‘burning (superficial) spontaneous pain’ NPSI mean subscores (evaluated at patient level) at week 16 among pathways.
  • Difference in ‘pressing (deep) spontaneous pain’ NPSI mean subscores (evaluated at patient level) at week 6 among pathways.
  • Difference in ‘pressing (deep) spontaneous pain’ NPSI mean subscores (evaluated at patient level) at week 16 among pathways.
  • Difference in ‘paroxysmal pain’ NPSI mean subscores (evaluated at patient level) at week 6 among pathways.
  • Difference in ‘paroxysmal pain’ NPSI mean subscores (evaluated at patient level) at week 16 among pathways.
  • Difference in ‘evoked pain’ NPSI mean subscores (evaluated at patient level) at week 6 among pathways.
  • Difference in ‘evoked pain’ NPSI mean subscores (evaluated at patient level) at week 16 among pathways.
  • Difference in ‘paresthesia/dysesthesia’ NPSI mean subscores (evaluated at patient level) at week 6 among pathways.
  • Difference in ‘paresthesia/dysesthesia’ NPSI mean subscores (evaluated at patient level) at week 16 among pathways.
  • Difference in NPSI mean total scores (evaluated at patient level) at week 6 among pathways.
  • Difference in NPSI mean total scores (evaluated at patient level) at week 16 among pathways.

Patient’s perceived tolerability

• Difference in tolerability among pathways, evaluated at the patient level on an 11-point NRS at week 6.
• Difference in tolerability among monotherapies, evaluated at the patient level on an 11-point NRS at week 6

Changes to subgroup analyses and exploratory analyses

The study had intended to characterise pain phenotype by the individual domains of the NPSI; however, a new classification system for pain based on the NPSI was used instead.52 In addition, the following subgroups were investigated: age, pain score at baseline, anxiety and depression scores at baseline, previous medication and the COVID-19 lockdown restrictions. Additional analyses were performed to compare outcomes among patients on combination therapy against those who remained on monotherapy.

Economic evaluation

A within-trial economic evaluation was completed as part of the study to understand the relative cost-effectiveness of the three treatment pathways.

Data collection tools

A pro forma of each of the forms was available to be used as source documents if needed. These forms were provided to each site electronically, along with a paper copy of each form for reference.

Procedures for assessing efficacy

Numeric Rating Scale 24-hour average pain was assessed via pain diaries that were given to participants at each study visit. Participants were instructed to complete the diaries each morning during the study. Completed diaries were then collected at the subsequent visit. During the weekly telephone calls, the research nurses reminded participants to record their pain scores every day. Pain scores were also collected via daily text messages where participants had given additional consent for this.

Procedures for assessing safety

The following safety assessments were performed to assess safety:

• Blood tests were performed at week 16 of each pathway.
• Vital signs were assessed at week 16 of each pathway.
• AEs were assessed during each study visit or telephone call.
• Concomitant medications were reviewed during each study visit or telephone call.

Additional procedures for assessing neuropathic pain

The NPSI and BPI-MSF were completed to assess neuropathic pain.

Procedures for assessing quality of life, psychological well-being and health economics

The following questionnaires were completed to assess QoL:

• ISI
• HADS
• SF-36
• EQ-5D-5L
• modified CSRI
• Pain Catastrophizing Scale
• Suicidal Risk Questionnaire
• PGIC
• Tolerability Scale.

Questionnaires could be completed during the visit. Alternatively, questionnaires could be posted to participants in advance of the visit and participants could bring the completed questionnaires when they attended their study visit. In the event that the participant forgot to bring the questionnaires or had not completed them, they were provided with another copy to complete during the visit.

Study-related case report forms

The following additional CRFs were completed by the investigators or delegates (i.e. sub-investigators delegated to consent or collect data) during the study:

• pre-screening log
• screening consent
• demographics
• medical history
• mTCNS
• DN4
• previous medications
• weekly contact form
• pregnancy test
• confirmation of eligibility
• randomisation
• treatment decisions
• treatment compliance log
• unblinding
• unscheduled dose changes
• pregnancy information
• protocol non-compliance
• intervention withdrawal
• study completion discontinuation.

See Appendix 1 for the CRF completion schedule and Appendix 2 for the questionnaire completion schedule.

Ancillary substudies

Participants were given the opportunity to consent to blood sample collection for future research. This aspect of the study was optional, and participants could take part in the main study without consenting to the blood sample collection. Samples were obtained at the same time as other study blood samples (i.e. week –2 or week 16 of each pathway) and shipped directly to the central labs via Royal Mail (Royal Mail Group plc, London, UK).

Patient and public involvement

Patient and public involvement (PPI) representatives have been involved throughout the study, including involvement in the initial study design, as well as implementation and oversight. The Diabetes PPI Panel at Sheffield Teaching Hospitals NHS Foundation Trust reviewed the study at the proposal stage. The PPI representatives were supportive of the proposal, including the study design, and they contributed to the choice of end points for the study. The panel was later involved in the development of the patient information sheet, consent form and study medication diary, helping to ensure that the study documents were accessible for potential participants. The Trial Steering Committee (TSC) included a PPI representative who provided ongoing input into the oversight of the study. PPI representation on the TSC ensured that the patient perspective was considered throughout the trial, including in decision-making regarding protocol amendments and trial recruitment strategies.

Trial management and oversight

Trial Management Group

The Trial Management Group (TMG) consisted of the chief investigator, collaborators, site investigators, site research nurses and staff from Sheffield CTRU. The TMG was responsible for the day-to-day implementation of the trial.

Data Monitoring and Ethics Committee

A Data Monitoring and Ethics Committee (DMEC) consisted of an independent statistician and two independent clinicians with research expertise. The DMEC reviewed reports provided by Sheffield CTRU to assess the progress of the study, the safety data and the critical end-point data. The DMEC provided feedback to the TSC following each meeting. The chief investigator (or delegate) and members of staff from Sheffield CTRU attended the open sessions of the DMEC meetings as observers. An unblinded statistician from Sheffield CTRU produced the report and attended the closed sessions of the DMEC.

Trial Steering Committee

The TSC consisted of independent clinicians, an independent statistician and a PPI representative. The role of the TSC was to provide supervision of the protocol and statistical analysis plan, to provide advice on and monitor progress of the study, to review information from other sources and to consider the recommendations from the DMEC. The chief investigator (or delegate), a sponsor representative and members of staff from Sheffield CTRU attended the TSC meetings as observers. Sheffield CTRU produced reports for review during the TSC meetings.

Changes to the protocol

All protocol amendments are listed in Appendix 3. A summary of the key changes is provided below.

Eligibility criteria

In substantial amendment 1, the inclusion criteria were updated to clarify that patients must have neuropathic pain affecting both feet and to update the Toronto Clinical Neuropathy Score to the ‘modified’ version.39 The exclusion criteria were updated to allow investigator judgement to be used when assessing the criteria for alcohol/substance abuse, history of psychiatric illness and prostate hypertrophy or urinary retention. The permitted dose of morphine equivalent was reduced from 120 to 100 mg/day. The exclusion criterion relating to history of ischaemic heart disease was updated to exclude any patient who had suffered a recent myocardial infarction (< 6 months prior to randomisation). New exclusion criteria were added for major amputations of the lower limbs and active diabetic foot ulcers.

Substantial amendment 7 updated the requirements for neuropathic pain to allow the pain to be present in the feet and/or hands. The exclusion criteria were also updated to clarify that only patients with non-diabetic symmetrical polyneuropathies were excluded. Previously, any non-diabetic neuropathy was an exclusion criterion, for example patients with diabetic neuropathy and carpal tunnel syndrome would previously have been excluded from the trial, but this was not the intention. This point was clarified in substantial amendment 7 to ensure that patients were not excluded unnecessarily. The exclusion criterion relating to liver function tests was updated to clarify that only the AST/ALT results were relevant for the eligibility assessment. The exclusion criterion relating to the electrocardiography results was updated to clarify that patients with a current history of arrhythmia were not eligible for the trial.

In substantial amendment 8, the exclusion criteria were updated to allow patients taking concomitant SSRIs to join the study provided that they had prior concomitant and safe use of SSRIs with the study medication (duloxetine and/or amitriptyline). The exclusion criterion relating to active foot ulcers was also updated to allow for investigator discretion. Patients were only excluded if the investigator felt that the ulcer would have a confounding or detrimental effect on the primary outcome or on patient participation.

To minimise the amount of time participants were on no pain medications, substantial amendment 12 allowed participants to be randomised early if the pain scores were high, provided that the mean pain score for the week was > 4 points. The exclusion criterion for heart failure was updated to exclude patients with heart failure class III or above (rather than class II or above). The exclusion criterion for postural hypotension was also updated to allow investigator discretion in the decision. Previously, all patients with a postural drop of > 20 mmHg were excluded.

In substantial amendment 13, the exclusion criteria were updated to clarify that patients taking concomitant citalopram were not eligible for the study. In addition, an update was included to exclude patients with second- or third-degree heart block or left bundle branch block (patients with right bundle branch block or first-degree heart block were permitted to be included following discussion with the cardiology team).

Data collection for the primary end point

At an early meeting, the TSC noted that there were potential issues with recording pain scores using paper diaries, including retrospective completion. The TSC recommended that the trial team consider an alternative method for collecting the primary end-point data. A text message data collection system was included in substantial amendment 5 and implemented in January 2018. This allowed daily text messages to be sent to participants who had provided optional consent for this aspect of the study. The text messages reminded participants to take their study medication and asked them to reply with their pain score, allowing the scores to be captured in real time.

Study treatment

Substantial amendment 1 allowed participants to start the next treatment pathway early if they wanted to withdraw from their current treatment pathway. The requirements for performing a dose review (i.e. to reduce the dose or discontinue a drug) were updated in substantial amendment 2, and it was clarified that only side effects that were severe or intolerable would require a dose review (rather than side effects which were moderate).

In substantial amendment 5, a reduced pregabalin dosing schedule was introduced for participants with an eGFR of 30–59 ml/minute. This ensured that the protocol was in line with British National Formulary (BNF) guidelines for pregabalin dosing.53 Substantial amendment 5 also allowed participants to start second-line treatment up to week 13 if needed.

Substantial amendment 7 clarified that participant preference could be considered when making dose titration decisions.

In substantial amendment 8, an update was made to allow study medication to be tapered more gradually between pathways in the event of significant withdrawal side effects, at the discretion of the local investigator.

Statistical methods

Sample size

A mean change between groups of 0.5 points was chosen based on the effect size previously reported in a crossover study29 for comparison of two active interventions for neuropathic pain. It was estimated that this would equate to an 8% difference between groups in the proportion of people improving by at least 1 point,54 which is considered a minimally clinically significant change in an individual.55 By using a within-patient standard deviation (SD) of 1.65,3,9 an alpha of 0.0167 to allow for three comparisons and 90% power, it was calculated that 294 evaluable participants were required.56

The original plan was to screen 536 patients, in total, for participation in the study. Assuming a 25% dropout rate, the study intended to recruit and randomise 392 participants to ensure that 294 participants completed the study.

However, recruitment for this demanding trial, with multiple study visits and four washout periods, became challenging and difficult to justify, given that most previous similar trials29,31,57 had used a 1 NRS point difference. With approval from the TSC, our PPI Panel and the funder, a decision was made to continue recruitment to a fixed time (July 2019), at which point the trial had recruited 140 participants. Using our original assumptions (i.e. a within-patient SD of 1.65 and alpha of 0.0167), the trial would achieve 90% power to detect a difference of 1 NRS point, assuming at least 74 patients per arm provided outcome data. With a 25% dropout, as originally assumed, the trial would have 95% power to detect a 1-point change and was sufficient to estimate differences in average pain to within a standard error of 0.25 NRS points.

Statistical analysis

General principles

Analyses were limited to randomised and eligible participants who started their pathway. Any participant who withdrew from study in one pathway was excluded from any succeeding pathways. Analyses were undertaken using generalised mixed-effect modelling, with treatment group (i.e. A-P, D-P or P-A) and pathway order (i.e. first, second or third) as fixed-effect covariates and participant as a random intercept. Statistical comparisons used 98.3% confidence intervals (CIs) and a 0.0167 statistical significance level was used for pairwise comparisons. Analyses were undertaken using intention-to-treat principles, which evaluated the policy of pathways rather than adherence to therapies. All analyses were undertaken using Stata^® version 16 (StataCorp LP, College Station, TX, USA).

Treatment uptake and response

Treatment uptake was summarised separately for first- and second-line therapies as (1) the proportion of participants stopping treatment prior to week 16 and (2) the dose level being taken at the end of the pathway at week 16 among those on therapy. The uptake of second-line therapy was further categorised as (1) being in combination with first-line therapy or (2) as a switch from first-line monotherapy.

Treatment response at week 6 was defined in relation to first-line monotherapy. A patient was classified as a treatment responder if they remained on first-line monotherapy at the week 6 visit with an average pain score of ≤ 3 points over the previous 7 days. Conversely, non-responders were those who (1) discontinued first-line monotherapy for AE, toxicity or ineffectiveness on or prior to week 6, (2) started second-line treatment (as combination therapy or as treatment switch) on or prior to week 6 or (3) had a NRS pain score > 3 points at week 6.

Treatment response at week 16 was in relation to both therapies. Patients who remained on at least one study medication and reported a 7-day average NRS score of ≤ 3 points were defined as having responded to the pathway, whereas those who discontinued for AE, toxicity or poor effectiveness and/or had a NRS score > 3 points were classed as non-responders.

Numeric Rating Scale pain

Self-reported NRS scores were collected daily for the duration of the pathway, by short message service (SMS) and/or via patient diaries. On days when a participant had provided both SMS and diary data, the NRS was taken from the SMS. A weekly average was calculated only if NRS scores were available on at least 4 out of the 7 days. The time window for the 6-week outcome was –2 weeks to + 1 week (i.e. any consecutive 7-day period ending between 28 and 49 days post commencement), provided that this did not extend to the washout phase or the next pathway. The week 16 data time window was –3 weeks to + 1 week (i.e. any consecutive 7-day period ending between 91 and 119 days post commencement), again, provided that this did not extend to the washout phase or next pathway.

At weeks 6 and 16, NRS scores were analysed using linear mixed-effect modelling. The binary outcomes (i.e. 30% pain reduction from baseline, 50% pain reduction from baseline and NRS score of ≤ 3 points) were analysed by logistic mixed effects regression. Baseline NRS score was the pain score taken during the washout period (i.e. week –1) immediately prior to randomisation and changes from baseline were calculated in reference to this score.

Missing data

The primary analysis was performed using a mixed model on complete-case data, which may be inadequate in situations where differential levels of withdrawal (or differential reasons for withdrawal) occur between groups. To address this, the following three sensitivity analyses were undertaken, using statistical imputation to assess the impact of missing data on pain scores:

1. Last observation carried forward, using the last available weekly NRS data. Although this approach has been widely criticised as oversimplistic,58 it offers a conservative estimate of the treatment effect in the case of conditions that improve with time.59
2. Multiple imputation,60 which is unbiased under the missing at random assumption if missing data can be predicted by the imputations model’s characteristics.
3. Controlled multiple imputation,61 in which informative withdrawal had more pessimistic values imputed.

Approach 2 used chained predicted mean matching imputation, with 10 nearest neighbours and 100 imputations. Missing data were imputed based on age, sex, baseline total NPSI score, treatment arm, pathway and any previous weekly NRS data available. Trace plots were used to assess convergence and, on the basis of these, 1000 burn-in imputations were used.

Approach 3 assumed that the participants who withdrew from treatment because of toxicity and/or inadequate treatment response would have a worse response than the value imputed by multiple imputation. Specifically, the imputed NRS values y_i* created by step 2 were replaced with a pessimistic imputation (y_i* + δ), with δ ranging between 0.5 and 2.5. Imputed values were bounded at 10 where applicable.

Subgroups

The NRS responses were analysed in relation to the age, baseline pain, HADS anxiety and depression scores and pain phenotypes, as derived from NPSI scores.52 Additional post hoc analyses were undertaken to assess whether or not outcomes were temporally associated with the COVID-19 lockdown, which began 3 months before the last patient last visit. Subgroup analyses were undertaken by adding an interaction term to the model and reported as marginal means.

Preferred treatment

After completing all three pathways, participants were asked to choose their preferred treatment. This was reported as a single 3 × 1 contingency table. The hypothesis test of equal proportions was assessed by chi-squared test.

Other efficacy assessments

All other efficacy assessments were undertaken at 6 and 16 weeks after the start of the treatment pathway, which corresponded to the end of monotherapy phase and the end of the treatment pathway.

Harms

Adverse events were recorded at each follow-up visit and categorised prior to unblinding. Any AEs occurring prior to the first treatment pathway were excluded. AEs were presented as the number of patients experiencing each event type and the number of events of each type. Where data allowed, the proportions were compared between arms using a mixed-effect logistic regression approach, as per binary pain outcomes. The following summaries were presented:

• all AEs
• all AEs of moderate or severe intensity and related (probably or definitely) to either treatment
• all SAEs.

Additional post hoc analysis looked at the number of days affected by each event type and the treatment phase during which the event occurred.

Health economic methods

A within-trial cost–utility analysis was conducted alongside the clinical study. The cost–utility analysis estimated the mean differences in costs, QALYs and the incremental cost-effectiveness ratio (ICER) over 16 weeks for each treatment pathway. The cost–utility analysis was conducted in line with the NICE Guide to the Methods of Technology Appraisal62 and is in line with the Consolidated Health Economic Evaluation Reporting Standards (CHEERS) checklist.63 The analysis is presented from an NHS and Personal Social Services perspective.

Quality of life and quality-adjusted life-years

Quality-adjusted life-years were measured over the 16-week period using the EQ-5D-5L.64 The EQ-5D-5L is a preference-based QoL measure that can be used in economic evaluations. The EQ-5D-5L consists of five dimensions (mobility, self-care, usual activities, pain or discomfort and anxiety or depression). For each dimension, responders indicate which of five levels their health is at today, with levels ranging from no problems to unable to do/extreme problems. Responses are then scored on a 0–1 scale on which 0 represents death, 1 represents perfect health and negative values indicate states worse than death.

In the OPTION-DM study, and in line with NICE guidelines,65 preference weights were obtained from van Hout et al.’s66 mapping study in the main analysis, with Devlin et al.’s46 preference weights applied in sensitivity analysis.

Participants completed the EQ-5D-5L at baseline, prior to the study commencing, and at 6 and 16 weeks in each treatment pathway. Given that this was a crossover study and participants received all three treatment pathways in turn, for the treatment pathway received second or third in order it was assumed that participants’ EQ-5D scores returned to the baseline value during the washout period between treatment pathways. Area under the curve using the trapezium rule was then used to estimate QALYs, which are presented in years throughout this report.

Resource use

NHS resource use was measured for each participant between baseline and the final follow-up (i.e. before crossover/end of follow-up). Resource use included all medication costs, visits to health services and any social care and community support. Medical costs were taken from the study medication records. In addition, other NHS resources used were self-reported by participants using the widely used and validated CSRI questionnaire.47 Unnecessary questions in the CSRI were removed to reduce the burden for participants; however, questions relating to personal costs incurred and time off work (where relevant) were retained for sensitivity analysis.

Details of unit costs for hospital visits, general practitioner (GP) visits, and social and other health-care services are listed in Table 3 (further details are available from the authors by request). Unit costs for laboratory tests were obtained from the national cost collection for the NHS67 and unit costs for medications are taken from the BNF.53 For unit costs to be applied to the same year, medication costs were deflated back to 2018/19 prices using the inflation rates provided in the Personal Social Services Research Unit (section 15.2).76

Treatment costs

Treatment costs consisted of the costs of medication (i.e. amitriptyline, duloxetine and pregabalin), the cost of the clinic visit (face to face or via telephone) and the cost of laboratory tests. Table 4 lists the treatment costs and the source of unit costs. Treatment medication was costed as it would be delivered within the NHS (rather than during a research study) so that results are presented from an NHS perspective. For example, the smallest pack of amitriptyline 25-mg tablets contains 28 tablets, and so the medicine cost at the week 1 visit was £0.94. The frequency of study medication visits were assumed to mirror what would take place in an NHS setting and these were either face to face or via telephone. Laboratory tests were assumed to take place once at the beginning of each treatment pathway and were costed accordingly.

As this was a crossover study, the time period in which the participants received the different treatment pathways could affect resource use and QoL. Therefore, linear regression analysis was performed to establish whether or not there was a difference in costs depending on the time period the participant received the treatment pathway. A separate regression model was fitted to costs and QALYs per treatment pathway (i.e. A-P, D-P and P-A). Confidence intervals around the coefficients for differences in costs between those receiving a treatment pathway first and either second or third in order were calculated using 5000 bootstrap simulations. A p-value of < 0.05 was used to establish statistical significance of an ordering affect. If statistical significance was established, then a sensitivity analysis would be carried out to allow for the time period in which participants received the treatment pathway.

Costs are reported using the 2018/19 time frame.

Cost-effectiveness analysis

Results are presented as a within-trial analysis, using a pairwise ICER as cost per QALY gained. It is possible to present this analysis because, within the crossover trial, participants received all three treatments. A non-pairwise (conventional) ICER as cost per QALY gained is also presented. Three cost-effectiveness comparisons are carried out: (1) A-P compared with D-P, (2) A-P compared with P-A and (3) D-P compared with P-A. Results are presented on the cost-effectiveness plane and on cost-effectiveness acceptability curves. No discounting was applied, as the follow-up period was < 1 year.

A total of 5000 simulations were used to obtain 95% CIs using bootstrapping.

All analysis was undertaken using Stata.

Sensitivity analysis

To allow for uncertainty, the following sensitivity analyses were undertaken:

• Devlin et al.’s46 algorithm for EQ-5D-5L utility values was used as an alternative to van Hout et al.’s66 algorithm for utilities.
• Analysis was undertaken with a wider societal perspective for costs. Personal costs and time off work are included, as reported by participants using the CSRI questionnaire. Participants provided details of any out-of-pockets costs related to employing extra help, transport to health-care appointments, modifications to their home and equipment purchased as a result of their condition. Details were also provided on any time away from usual activities for themselves or for friends or relatives as a result of their condition. Participant and friend/relative time was costed at the average UK wage as detailed in the Office for National Statistics annual survey for hours and earnings (2019).77
• EQ-5D responses were missing for 27–32% of participants. Therefore, multiple imputation58 was carried out to impute the missing values assuming responses were missing at random. As with the statistical analysis, predicted mean matching with 10 nearest neighbours and 100 imputations were carried out. Missing data were imputed base on age, sex and treatment pathway.