The literature flow diagram (Fig. 2) summarizes the results of the search and study selection processes. Among 3,155 potentially relevant citations, we included 49 studies—1 systematic review (Frank et al.),8 34 studies included in that systematic review,,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49 and 14 new studies identified in our literature search.50,51,52,53,54,55,56,57,58,59,60,61,62,63 A total of 40,245 patients are included in the 14 new studies.
We prioritized evidence synthesis of: 1) studies conducted in a VHA setting or other outpatient settings; 2) studies that addressed serious harms such as overdose and suicide; and 3) studies that included enough detail regarding patient and tapering characteristics to evaluate the applicability of findings to VHA populations and settings. Nineteen studies (Table 1) (2 RCTs,18, 50 6 controlled observational studies,54, 56, 60,61,62,63 and 11 uncontrolled observational studies)27, 30, 32, 34, 36, 40, 55, 57,58,59, 65 met these criteria, which we refer to as “prioritized studies.” The remaining studies either had low applicability to VHA or outpatient settings or included patients or interventions that were not well-described. Also, unless they otherwise met our prioritization criteria, we do not discuss poor-quality studies included in Frank et al. in detail (see Table 1 for patient and taper characteristics of prioritized studies). Data tables for all included studies are available in online Appendix A and a list of all included and excluded studies are in our report’s supplemental materials.11 Table 2 provides an overview of results. Findings for specific patient outcomes are discussed below.
Table 1 Characteristics of 19 Prioritized Studies Table 2 Outcomes of 19 Prioritized Studies18, 27, 30, 32, 34, 36, 40, 46, 50, 54,55,56,57,58,59, 61, 63 Pain and Pain-Related Function
Studies reported mean changes in pain severity and intensity scores (as measured by tools such as the pain numerical rating scale, multidimensional pain inventory, and brief pain inventory). Most studies did not define clinically important changes on these scales or state whether opioid dose reduction resulted in severe pain requiring a change in management with a few exceptions.36, 63 Seven observational studies of opioid tapering in the context of intensive 3-4-week multimodal pain management programs requiring full-day participation Monday–Friday showed improvements in pain severity and intensity scores.27, 32, 34, 36, 40, 63, 65 Among studies reporting mean pain scores at baseline and endpoint, improvements were greatest (19-47%) in studies of patients on higher baseline MEDD (99-177 mg)27, 32, 36, 46 and more modest (8-10%) among studies of patients with lower baseline MEDD (47-61 mg).63, 66 Although these results are encouraging, studies of intensive multimodal pain management programs have limited applicability overall given that many patients lack access to or means to participate in such programs.
The best evidence of the impact of opioid tapering on pain severity with a moderate co-intervention is an RCT in which patients with chronic pain who wanted to taper opioids were randomized to usual care versus a tapering support intervention that included psychiatric consultation and 18 weekly meetings with a physician assistant to improve self-management skills. In both groups, patients had lower pain severity ratings at 22 weeks and patients in the intervention group also had improvements in pain interference and pain self-efficacy.18 Another RCT of a moderate-intensity intervention relied on voluntary opioid tapering and had a high drop-out rate, prompting the study authors to deem the intervention unsuccessful.50 The least intensive intervention was an observational study of 51 patients in a community pain clinic with high baseline MEDD (288 mg) who voluntarily participated in a slow individualized taper with the use of a self-help book and had mean improvements in pain scores of 10%. In another observational study of tapering, 50 VHA patients with high baseline MEDD (64% > 200 mg) who tapered opioids with usual care had less pain (40%) or unchanged pain (28%) at 6–12 months.30
Importantly, patients participated voluntarily in tapering interventions described above. Only 1 study that evaluated pain scores at baseline and endpoint provides information regarding patient outcomes after clinician-initiated or mandated tapers. In a retrospective study of 551 VHA patients with baseline MEDD 76 mg in which the majority (85%) underwent clinician-initiated tapers, pain scores improved by 3.8%.58 A limitation of this study is that it does not include patients who discontinued VA care or continued VA care but received pain management elsewhere.
Similar trends were noted for mean changes in pain-related function following opioid tapers. The most improvement was observed in a group of 1457 patients (baseline MEDD 117 mg) who participated in an intensive outpatient multimodal pain management program at the Cleveland Clinic.36 In this study, the mean score on the pain disability index (PDI) decreased from 42.95 at baseline to 18.29 at discharge (− 57.4%) and was 23.7 after 6–12 months of follow-up (− 44.8%).36 Similarly, in a VHA study of an intensive intervention in which 705 veterans (baseline MEDD 61 mg) voluntarily participated in a 3-week interdisciplinary pain program incorporating opioid cessation, scores on the VA Pain Outcomes Questionnaire-interference in Activities of Daily Living (POQ-ADL) decreased from 16 at baseline to 13 at 3-week discharge (− 18.8%).40 The smallest (and statistically nonsignificant) change came following the least intense intervention, in which an individualized taper was accompanied by a self-help book.55 While it is encouraging that mean pain scores and functional measures improved or did not significantly change for most patients who reduced or discontinued opioids, these types of outcome assessments may be of limited utility as they do not necessarily correlate with meaningful changes from the patient perspective.
We have low confidence in these findings (low quality per GRADE). The body of evidence has several limitations including a high proportion of uncontrolled observational studies (which introduce the potential for unmeasured confounders), unclear fidelity to interventions, and inadequate reporting of missing data and handling of missing data. Despite these limitations, within our subset of studies, findings regarding pain and pain-related function were consistent.
Serious Harms: Substance Use, Opioid Overdose, and Suicide
Evidence regarding the impact of opioid dose reduction and discontinuation on substance use, opioid overdose, and suicide is unclear and highlights the challenges of studying complex linkages among the problems of chronic pain, opioid dependence, opioid use disorder, suicidal ideation, and co-morbid mental health disorders. A 2019 retrospective study of Medicaid claims data in Vermont found that among a cohort of 694 Medicaid recipients who had a high prevalence of substance use disorders (60%) on ≥ 120 mg MEDD, almost half (49%) of 494 patients who discontinued opioids between 2013 and 2017 subsequently had an ED visit or hospitalization due to opioid poisoning or substance use disorder.59 In this study, opioids were most often discontinued without a gradual taper (median length of time to discontinuation was 1 day) and < 1% of patients were prescribed medication to treat substance use disorders. This study does not describe the circumstances regarding opioid discontinuation or exclude the potential for reverse causality (i.e., a diagnosis of substance use disorder was the reason prescription opioids were discontinued) but highlights a real-world pattern of abrupt opioid discontinuation as well as undertreatment of substance use disorders.
We identified 3 observational studies that examined overdose as an outcome of opioid dose reduction.54, 57, 61 Because opioid overdoses are rare events, larger population-level studies are more informative than small studies. The only large study of opioid overdose is a retrospective study of overdose rates following different phases of an opioid risk reduction initiative among patients in Washington’s Group Health practice (intervention group) compared to patients in Group Health’s contracted community clinics (control group).54 Both groups were subject to a change in Washington State’s opioid prescribing guidelines recommending against doses ≥ 120 mg MEDD, but providers in the intervention group also received “feedback and supervisory guidance by medical directors”. The within-group analysis demonstrated a significant decrease in overdose rates (relative annual change 0.83, 95% CI 0.70 to 0.99), but the between-groups analysis did not. Overall, the results of this study provide inconsistent support that reducing opioid doses leads to lower overdose rates.
The other 2 studies of overdose after opioid discontinuation are smaller. In a retrospective study of 572 patients in a primary care clinic on LTOT during 2010–2015, 17 (4.9%) patients who discontinued opioids died of an overdose and 4 (1.75%) patients who continued prescription opioids died of an overdose.61 LTOT discontinuation was associated with a hazard ratio for overdose death of 2.94 (1.01 to 8.61) after adjusting for age and race. In another retrospective study of 43 VHA patients who stopped opioids due to opioid agreement violations, no patients overdosed.57 A limitation of these 3 studies examining overdose is that they were not designed to evaluate causation (i.e., whether opioid overdoses occurred because patients had uncontrolled pain and sought illicit opioids when their prescription doses stopped or whether they had underlying opioid use disorder that was unmasked or worsened with a taper).
The best evidence on the association between opioid dose reduction and suicidal ideation and suicidal self-directed violence is a retrospective study of 509 VHA patients who underwent clinician-initiated tapers due mostly (75%) to aberrant behaviors. In this study, 47 (9.2%) patients had new-onset suicidal ideation and 12 patients (2.4%) had suicidal self-directed violence in the year following opioid discontinuation.56 Baseline PTSD (OR = 2.56, 95% CI 1.23 to 5.32) and psychotic disorders (OR = 3.19, 95% CI 1.14 to 8.89) were associated with suicidal ideation and suicidal self-direction violence, while other co-morbidities including substance use disorder and baseline MEDD were not. An important limitation of this study is that it excluded patients who had no VHA contact or who died in the year following opioid discontinuation, and therefore likely underestimates the actual proportion of patients who experienced suicidal ideation and suicidal self-directed violence.
Retention in Primary Care or Usual Source of Healthcare
The association between opioid dose reduction or discontinuation and retention in primary care or patients’ usual source of healthcare is unclear. The best evidence is a retrospective study of 1,624 patients on LTOT in an academic healthcare system in the Bronx, NY, which found that 78 of 207 patients who tapered opioids (4.8% of the total sample) terminated their care in the year following the taper (defined as no outpatient encounters in the healthcare system).62 In this study, opioid taper was significantly associated with termination of care (AOR 4.3, 95% CI 2.2 to 8.5) compared to continuing opioids.
Evidence on serious harms as well as the other outcomes that we examined in our full report including quality of life, resolution of opioid-related side effects, withdrawal symptoms, and patient satisfaction is unclear (very low quality per GRADE). Findings for these outcomes were supported by only a single or few observational studies with methodological weaknesses including lack of control groups and small sample sizes. Future studies of higher methodologic quality may have different findings for these outcomes.