Effectiveness of interventions to reduce opioid use

Summary of study characteristics

Thirty-three papers, reporting on 27 studies, assessed the effectiveness of tapering of opioids in people with chronic non-cancer pain. Five of those studies also reported on AEs.110,111,112,113,116 Eleven papers reported on comparative studies. Of these, seven were RCTs,69,71,75,111,112,113,116 one was a prospective matched cohort (PMC) study,81 one was a retrospective randomly matched cohort study70 and two were retrospective cohort studies.67,77 Sixteen papers were non-comparative studies; 14 were single cohorts (6 were prospective and 8 were retrospective) and 2 were studies that did not follow a patient cohort but reported on changes in outcomes following a change in practice.83,117

None of the included studies were done in the UK. Twenty-one studies were done in the USA,66,68,69,70,71,72,73,77,78,79,80,81,82,83,88,89,90,110,111,113,117 two each were done in Australia,112,116 Denmark75,76 and Germany.67,74 All comparative studies except one,67 were done in outpatient clinics and three studies71,77,81 were done in US Department of Veteran Affairs (VA) facilities. Two non-comparative studies74,78 were done in inpatient facilities and three73,80,117 were done in US VA facilities. The recruitment period of the studies varied from 1 month to 6 years and the earliest study started recruiting in 2001.74

The type of intervention and the comparator varied across the studies. Of the comparative studies, two studies67,75 examined a tapering-only intervention based on sequential dose reduction, four studies77,111,112,116 examined acupuncture, three studies69,71,113 examined cognitive, psychological and behavioural interventions, and two studies70,81 examined pain management programmes. The comparator groups received treatment as usual in six studies,70,75,77,81,111,113 support groups in two studies,69,71 sham acupuncture in two studies,112,116 and the remaining study67 assessed two different dose reduction schedules. Of the 16 non-comparative studies, 7 studies66,73,80,82,83,89,117 examined interventions aimed at changing a service (e.g. linked to the US Opioid Safety Initiative [OSI]), 3 studies78,90,110 examined a pharmacological intervention and 6 studies68,72,74,76,79,88 examined pain management programmes.

None of the authors of the comparative studies reported receiving commercial funding. Eleven non-comparative studies reported a funding source and one study110 received funding from a commercial company that manufactured the intervention under study (CBD Hemp extract).

Of the comparative studies, one69 reported a conflict of interest due to the author being ‘the Director of the Center on Mindfulness and Integrative Health Intervention Development’, which indicated a potential commercial interest in the intervention under study. Two non-comparative studies reported that members of the research teams either were in receipt of grants from private companies or were principal investigators (PIs) in commercially funded studies.83,88 For four studies,66,69,80,112 there was no information about whether there were conflicts or not. For the remaining 19 studies, it was reported that there were no conflicts of interest.

All studies reported outcome measures at baseline, the end of treatment/discharge and at least one later follow-up assessment. The length of follow-up varied between 3 and 8 months. A range of outcome measures were used in the studies but all but two studies68,69 reported on the change in opioid use. The timing of outcomes was not always reported clearly, although most studies reported values at baseline and the end of the intervention.

Details of the study characteristics are shown for comparative studies in Table 6 and non-comparative studies in Table 7.

Intervention characteristics

Comparative studies

Details of the intervention characteristics are shown in Table 8.

Of the 11 comparative studies, 2 examined sequential opioid dose reduction interventions (‘taper only’),67,75 4 examined acupuncture (only 1 of which combined acupuncture with active tapering),77,111,112,116 3 examined cognitive, psychological and behavioural interventions (only 1 of which combined the intervention with active tapering),69,71,113 1 examined a pain e-consult programme70 and 1 examined an integrated pain team.81

The comparator groups received treatment as usual in six studies,70,75,77,81,111,113 support groups in two studies,69,71 and sham acupuncture in two groups.112,116 The remaining study assessed two different dose reduction schedules, that is fixed opioid withdrawal compared to individualised withdrawal.67

Interventions could aim to either reduce or discontinue opioid use. The aim of the intervention in this respect is important when looking at the outcomes on opioid use. For nine of the comparative studies the aim was to reduce opioid use, whereas two aimed to discontinue opioid use.67,75 One study70 aimed to improve prescriber education and practice.

All but 1 study67 was conducted in an outpatient setting and 10 used face-to-face delivery.67,69,71,75,77,81,111,112,113,116 For three of these studies, telephone contact was also used.75,81,113 The intervention in one study was based on e-consults.70

Most interventions were delivered individually, though the two studies assessing mindfulness delivered the intervention in groups;69,71 the motivational interviewing (MI) and cognitive–behavioural therapy (CBT) element of the Sullivan 2017 study113 was also delivered in groups.

The duration and intensity of the interventions ranged from 6 weeks112 to 12 months.111

Non-comparative studies

Details of the intervention characteristics are shown in Table 9.

Of the 16 non-comparative studies, 7 were interventions aimed at changing the service, for example OSIs.66,73,80,82,83,89,117 Two of these did not follow the same cohort of patients through the study.83,117 Of the remaining nine studies aimed at the individual, three used a pharmacological intervention78,90,110 and six were pain rehabilitation programmes.68,72,74,76,79,88 The aim of 10 studies was to reduce the level of opioid use,66,68,76,79,80,82,83,89,110,117 whereas 5 studies aimed for participants to discontinue opioids.72,74,78,88,90 The remaining study aimed to reduce opioid prescribing rates.73

Where applicable, interventions were conducted in outpatient settings for all but two studies.74,78 The interventions that focused on individuals instead of changing a system were all conducted face to face, predominantly individually, although four studies also reported using group sessions.68,72,79,88

The duration of therapy was reported by seven of the nine studies aimed at the individual and ranged from 15 days88 to 0.8 years.90 Two studies did not report the duration or intensity of the intervention.76,79

Further analysis of intervention components

We examined the composition of the interventions examined across the included studies to understand more about what components of behaviour change were involved. Studies were first grouped by type of intervention as follows:

• pain management programme, eight studies68,70,72,74,76,79,81,88
• tapering curriculum/guidelines/initiative, six studies66,73,80,83,89,117
• pharmacological-assisted opioid tapering, three studies78,90,110
• tapering only, two studies67,75
• acupuncture, four studies77,111,112,116
• cognitive, psychological and behavioural, three studies69,71,113
• clinical encounter, one study.82

Two studies were excluded from the further analysis as insufficient information was provided about the components of the intervention.82,89 In 19 studies,67,68,69,71,72,74,75,76,77,78,79,81,88,90,110,111,112,113,116 the intervention components targeted patient behaviour only, in 5 studies intervention components targeted healthcare professional’s behaviour only,66,70,73,80,83 and in 1 study,117 intervention components targeted both patient and healthcare professional behaviour. Using the Behaviour Change Wheel book as a guide,118 each study was examined to identify which of nine possible intervention functions (education, training, environmental restructuring, modelling, persuasion, coercion, incentivisation, restriction and enablement) were used in the intervention.

Patient behaviour

For 20 studies that targeted patient behaviours, interventions had between 1 and 3 intervention functions (Table 10). The most frequent functions were Enablement (18 studies) and Restriction (12 studies). Restriction refers to the use of rules to reduce the opportunity to engage in the target behaviour and interventions that involved opioid dose reduction or cessation were mapped to this function. Enablement functions aim to increasing means or reduce barriers to increase capability and were mapped to interventions that aimed to bring about a change in psychological capability. Intervention functions identified less frequently were Training (four studies) and Education (three studies). In 15 studies, interventions were mapped to more than 1 function and 5 combinations of functions were found:

• Training + Restriction + Enablement, five studies68,72,78,81,88
• Restriction + Enablement, five studies67,74,76,90,111
• Education + Restriction + Enablement, two studies75,79
• Training + Enablement, two studies69,71
• Education + Training + Restriction + Enablement, one study.113

Healthcare professional behaviour

For six studies that targeted healthcare professional behaviours, interventions had between two and three intervention functions (see Table 10). The most frequent functions were Environmental restructuring (four studies) and Enablement (four studies). The functions Training (three studies), Education (three studies) and Persuasion (two studies) were also identified. Environmental restructuring refers to functions which aim to change the physical or social context. In the context of the interventions that targeted healthcare professional behaviour, an example of this function involved notes or recommendations being placed on patient’s electronic health record. All the interventions across the six studies were mapped to more than one intervention function as follows:

• Training + Environmental restructuring + Enablement, one study70
• Education + Persuasion + Training + Enablement, one study117
• Education + Persuasion + Environmental restructuring, one study83
• Education + Training, one study66
• Environmental restructuring + Enablement, two studies.73,80

Population characteristics

Comparative studies

Details of the population characteristics are shown in Table 11.

The 11 comparative studies included a total of 1606 participants, with individual studies ranging from 3575,112,113 to 665 participants.70 The participants were evenly divided between the intervention and comparator groups for seven studies, whereas three studies70,111,116 enrolled more participants to the intervention group than the comparator group and one study enrolled more participants to the comparator group than the intervention group.67

Four studies reported notably more females than males in the study, three studies reported notably more males than females and four studies reported equal proportions of males and females. The mean ages of participants ranged from 47.469 to 65 years.77

The type of chronic pain that participants were experiencing was reported by six studies and was varied. The mean duration of pain was reported by four studies and varied from 10.8116 to 19.8 years.112

The types of opioids participants were taking was only reported by three studies and differed between studies. For the five studies reporting the duration of opioid treatment, the values varied between 6.675 and 14.1 years.81 The average daily opioid dose was reported by seven studies and varied between 76.970 and 367 mg/day.75 Two further studies reported weekly doses of 461 and 295 mg/week112 and 463 and 620 mg/week.116

Non-comparative studies

Details of the population characteristics are shown in Table 12.

The two non-comparative studies that did not follow a cohort of patients are not included in the description of the population characteristics, though the sample sizes and details of opioid use are reported in Table 12. The remaining 14 studies included 5710 participants and individual studies ranged from 3479 to 249282 participants.

The sex of participants in the non-comparative studies was less evenly spread than in the comparative studies with seven studies reporting notably more females than males, five studies reporting notably more males than females and only two studies reporting equal proportions of males and females. Two studies failed to report the proportion of males and females. The ages of participants ranged from 44.588 to 64 years.73 Three studies did not report the average age of participants.72,82,89

The type of chronic pain that participants were experiencing was reported by six studies and was varied. The mean duration of pain was reported by four studies and varied from 8.668 to 12.7 years.78

The types of opioids participants were taking was only reported by two studies68,74 and differed between the studies. For the five studies reporting the duration of opioid treatment, the values varied between 3.988 and 10 years.76 The average daily opioid dose was reported by 11 studies and varied between 41.0479 and 367 mg/day.74

Quality assessment

Comparative studies

Randomised controlled trials

For the RCTs, the Cochrane Risk of Bias 2 tool36 was used and overall quality was assessed rather than for each outcome. The results are shown in Table 13 and Figure 3. Only one RCT was rated as being at low risk of bias, with low risk across all domains;69 all other RCTs had an overall rating of ‘some concerns’, with at least one domain assessed as having some concerns. No studies were assessed as being at high risk of bias for any domain.

FIGURE 3 Percentage of RCTs in each risk category overall and for each domain (Review 1). (Figure not reproduced here; see the original article.)

Non-random comparative studies

For non-random comparative studies, CASP tools were used to assess study quality. For the PMC study, the CASP case-control checklist was used.119 For the two retrospective cohort studies, the CASP cohort study checklist was used.38 The results for these studies are shown in Table 14.

Non-comparative studies

As all of the non-comparative studies were single cohort studies, the CASP cohort study checklist was used.38 The results for these studies are shown in Table 15.

Outcomes

Of the outcomes of interest to this review (see Table 2), seven were reported by three or more studies and are synthesised in this section. Full data extraction tables for each outcome are included in Appendix 5 with brief summary tables shown here for pain outcomes (Table 16) and change in opioid dose (Table 17).

Pain

Pain was reported by 15 studies either in terms of severity and/or interference.

Four non-comparative studies that assessed pain management programmes reported pain severity as an outcome.68,74,79,88 Three studies used an 0- to 10-point scale and one study used the Multidimensional Pain Inventory (MPI), a 52-item, 12-scale inventory. All four studies reported a decrease in pain scores from baseline to discharge/end of the intervention. A decrease was still apparent at follow-up in two studies that reported outcomes at 6 months and 12–24 months, respectively.74,88 Neither of the comparative studies assessing pain management programmes reported pain severity or pain interference as an outcome.70,81

Three non-comparative studies that assessed pharmacological-assisted opioid tapering78,90,110 all reported at least one pain outcome. Two studies reported decreases in pain severity from baseline to discharge/end of intervention, both on a 0- to 10-point scales. Two studies also reported a decrease in pain interference. None of the studies reported outcomes over a longer follow-up period.

Two studies of tapering-only interventions reported pain severity, both at discharge/end of the intervention and over a longer follow-up period. One study75 compared a taper intervention with stabilisation finding no significant difference between groups at follow-up. The other study67 of different starting doses found a reduction in pain scores from baseline in both groups at follow-up.

Two comparative studies assessing cognitive, psychological and behavioural interventions reported both pain severity and interference; measured in both studies with the Brief Pain Inventory (BPI). The study of mindfulness training compared to a support group, found statistically significant effects in favour of the intervention for both pain severity and interference.69 The study of tapering and MI/CBT found no difference in pain severity between the intervention and taper-only control at follow-up.113 For pain interference, a statistically significant effect in favour of the intervention was found at follow-up.

Four comparative studies that assessed acupuncture reported pain severity as an outcome. None of the studies found a statistically significant difference between the intervention and control groups.

As statistically significant evidence by itself is insufficient to indicate whether the magnitude of the effect on pain is clinically important,120 we explored minimally important differences across the studies that reported pain outcomes. This exploratory analysis was based on a reduction of at least 2 points on 0- to 10-point scale. The change from baseline reached a difference of 2.0 points or more in three non-comparative studies: two studies68,79 of pain management programmes and one study90 of pharmacological-assisted tapering.

Change in opioid dose (including cessation)

All but two studies68,69 reported at least one measure of change in opioid use. Twenty-two studies reported on the outcome of reduction in opioids with differences in how the outcome was reported. The different outcome types are reported in separate tables in Appendix 5.

Change in mean morphine equivalent daily dose or equivalent

Five studies that examined pain management programmes reported on the reduction of opioid dose. Two comparative studies found statistically significant effects in favour of the intervention at follow-up.70,81 Seal et al.81 compared an integrated pain team with usual care and Gersch et al.70 a pain e-consult programme with usual care. Three non-comparative studies all reported reductions in opioid dose between baseline and the end of treatment/discharge.72,79 Two further studies74,88 reported details for participants still taking opioids at the end of treatment, with all patients having reduced their doses.

Three studies66,80,117 that examined an opioid-tapering curriculum, guideline or curriculum reported changes in opioid dose. Austin et al.66 reported the change in dose between a baseline cohort and patients who remained on opioid therapy, finding no significant change. Two studies80,117 about the implementation of recommendations related to the US OSI reported reductions in opioid dose. One further study83 of OSI implementation reported ‘modest reductions’ in the annual rate of opioid prescribing.

One study that examined pharmacological-assisted opioid-tapering interventions reported that 51.5% of participants using the CBD hemp extract were able to reduce opioid medications at week 8.110

Both studies that examined tapering-only interventions reported on the reduction of opioid dose. Kurita et al.75 reported a reduction in dose in the taper-only intervention group, but there was no significant difference between the taper only and stabilisation groups at follow-up. The authors state that the taper phase of the study was underpowered due to dropout. Bienek et al. reported reductions in dose in both the fixed starting dose group and individualised starting dose group.67

Two comparative studies71,113 that examined cognitive, psychological and behavioural interventions reported changes in opioid use. Hudak et al.71 found a statistically significant reduction in opioid dose that favoured mindfulness training compared to a support group. Sullivan et al.113 reported reductions in opioid dose in both the taper and MI/CBT intervention group and taper-only control group but found no statistically significant difference between the groups at the end of treatment or at follow-up.

All four comparative studies that examined the use of acupuncture reported on the reduction of opioid dose.77,111,112,116 Three studies, including two that examined electroacupuncture and one that examined taper and acupuncture, reported reductions in the both the intervention and control groups. However, none of the studies found a statistically significant difference between the intervention and comparator groups at the end of treatment or follow-up.

One study that examined clinical encounters reported that 29% of clinical encounters were followed by a reduction in opioid prescribing.82

Cessation of opioid use

All but one study75 whose aim was for cessation reported on the number of participants succeeding in ceasing opioid use. An additional study that aimed to reduce opioid use also reported cessation rates.66

Four non-comparative studies that examined pain management programmes reported cessation rates. The studies reported a high proportion of participants ceasing use at the end of the intervention – ranging from 69.5%72 to 100%.68 Three studies reported cessation rates at follow-up and a proportion of patients had relapsed to opioid use following cessation; the proportion of patients that remained opioid free in these three studies were 48.0%,74 77.3%88 and 29.9%,72 respectively.

Two studies that examined tapering curriculums, guidelines or initiatives reported low proportions of participants ceasing use – 9.5%73 and 26.6%,66 respectively.

Two studies that examined pharmacological-assisted opioid tapering reported that 100% of participants78 and 27.9% of participants,90 respectively, had ceased opioid use at the end of treatment.

One comparative study of a cognitive, psychological and behavioural intervention that compared taper plus MI/CBT with taper only found that only two participants in each group ceased opioid use. Two comparative studies that examined tapering interventions reported cessation rates. Kurita et al.75 reported that no participants in either the taper-only intervention group or stabilisation comparison group ceased opioid use. Bienek et al.67 compared a fixed starting dose taper regimen with an individualised starting dose regimen, finding that 64.8% and 50.2% of participants, respectively, ceased opioid use.

Opioid withdrawal-related symptoms or dependence

Opioid withdrawal-related symptoms were poorly reported across the studies with only three studies67,68,75 reporting symptoms assessed using a validated scale.

One non-comparative study68 reported that the peak score on the Clinical Opioid Withdrawal Scale (COWS) was higher for patients on higher opioid doses but not significantly different based on the morphine equivalent dose (MED). The mean peak COWS score occurred at 80% reduction from the initial opioid dose for patients on a MED dose of 100 mg/day or less, 64% for a MED dose of 100–200 mg/day, and 72% for a MED dose of > 200 mg/day.

Two comparative studies examined tapering-only interventions. One study75 did not report the results of either the Objective Opiate Withdrawal Scale (OOWS) or Subjective Opiate Withdrawal Scale (SOWS) that they used. The final study did report that those on a fixed starting dose had a significantly lower 10-day mean daily SOWS scores than those on an individualised starting dose (comparator group).67

Acceptability of the intervention and patient satisfaction

Acceptability of the intervention and patient satisfaction were reported in four studies. One study examined pharmacological-assisted tapering, and participants were asked to score overall satisfaction on a scale of 0–10, the average rating was 8.3.78 Two studies examined cognitive, psychological and behavioural interventions. Garland et al.69 measured treatment credibility based on three items about attitudes towards treatment. There was no statistically significant difference between the intervention and control groups on this measure (21.4 vs. 19.3). In the study by Sullivan et al.,113 intervention group participants were asked to rate the helpfulness of the taper plus MI/CBT intervention on a 1- to 5-point scale. The intervention was rated as very or extremely helpful by 81.3% of participants at the end of the intervention and by 73.3% at 34 weeks follow-up. Finally, Sharp et al.82 reported that the number of patients with a mean satisfaction score > 9 was 88.9%.

Emotional functioning

Eight studies reported the outcome depression,68,69,74,75,76,88,111,113 and of those, four also reported anxiety outcomes.75,76,111,113

Four of the studies reporting depression as an outcome were evaluations of pain management programmes. Two non-comparative studies74,76 reported no change in depression measures, at 12- to 24-month follow-up or at the final visit, respectively. Two further non-comparative studies, both of 3-week rehabilitation programmes reported statistically significant reductions in depression on the Centre for Epidemiological Studies Depression Scale (CES-D) at the end of the programme.68,88 Townsend et al.88 followed patients up at 6 months post treatment, finding that the significant reduction in depression remained.

Four comparative studies69,75,111,113 reported depression outcomes measured with the Hospital Anxiety and Depression Scale (HADS),75,111 Calgary Symptoms of Stress Inventory (C-SOSI)69 and the Patient Health Questionnaire-9 items (PHQ-9).113 The intervention types explored in these studies varied and included two studies of cognitive, psychological and behavioural interventions,69,113 acupuncture111 and sequential opioid dose reduction.75 None of the studies found a statistically significant difference between the intervention and comparator groups at the end of treatment or follow-up.

Four studies reported anxiety outcomes. Three were comparative studies (of acupuncture,111 sequential opioid dose reduction75 and a cognitive, psychological and behavioural intervention113) and one was a non-comparative study of a pain management programme.76 None of the comparative studies found a statistically significant difference between the intervention and comparator groups at the end of treatment or follow-up.75,111,113 The non-comparative study did not find a statistically significant change between baseline and the end of treatment.76

Sleep quality

Sleep quality was reported by four studies.75,78,110,113 Two non-comparative studies examined pharmacological interventions78,110 and both reported statistically significant improvements in sleep quality from baseline to the end of the intervention. Two comparative studies75,113 that examined sequential opioid dose reduction and acupuncture interventions, respectively, found no statistically significant difference between the intervention and comparator groups at either the end of the intervention or at longer-term follow-up.

Quality of life

Quality-of-life outcomes measured with the 36-item Short Form Survey (SF-36) were reported by five studies.68,74,75,76,112 Two comparative studies, of sequential opioid dose reduction75 and electroacupuncture,112 respectively, reported no statistically significant difference between the intervention and comparator groups at either the end of the intervention or at longer-term follow-up.

The three remaining studies were all non-comparative studies of pain management programmes.68,74,76 One study reported a statistically significant improvement from baseline in the mental component summary of the SF-36 but not in the physical component at 12–24 months follow-up.74 Another study68 reported a statistically significant improvement from baseline on the general health perceptions domain of the SF-36. One study reported no significant changes from baseline.76

Two further studies reported that they assessed SF-36 scores but did not report results in their published papers.75,112

Summary

The literature searches identified 27 relevant studies, including 11 comparative studies and 16 non-comparative studies. There were some concerns about the risk of bias across the included studies but only significant concerns about the quality of one study.67 With differences in population, such as the number of recruited patients and patient characteristics, study design (including intervention), collected outcome data and duration of follow-up, it was unsurprising that for the outcomes of pain severity, emotional functioning, sleep quality and quality of life, some studies reported improvements while other studies reported no differences. Although only reported by two studies, there was consistency in terms of the acceptability of interventions which was high. Confidence in this finding is enhanced by the safety review results (see Chapter 7) as none of the seven studies that reported AEs reported any serious AEs, and no patients were reported to have withdrawn from a study due to an AE.

All except two of the included studies68,69 reported at least one measure of change in opioid use. While cessation of opioid use is an outcome that is clinically important, defining the level of reduction in opioid use that is clinically relevant is problematic. Findings from a systematic literature review121 of minimum clinically important differences (MCID) for chronic pain relief showed that reported MCID levels vary considerably, although baseline pain was strongly associated with absolute, but not relative, measures. Further, but to a much lesser degree, MCID was also influenced by the operational definition of relevant pain relief and possibly by clinical condition.

The relevance to NHS practice of some of the interventions described in the included studies is unclear. For example, four of the comparative studies assessed the effect of acupuncture on reduction of opioid use; however, acupuncture is only sometimes available on the NHS, access is limited and most patients wanting acupuncture pay for private treatment. Further, five of the non-comparative studies examined how the introduction of OSIs affected patient use of opioids; such initiatives will only be relevant if the assessed measures are not already implemented in the NHS.