The impact of opioid prescribing guidelines on opioid dispensation: A systematic review and meta-analysis Authors: Jihoon Lim, Parker Tope, Maximilian Schaefer, Isaiah Williams, Imen Farhat, Andrea Benedetti, Dimitra Panagiotoglou Published: 2026 ABSTRACT ABSTRACTBackgroundOpioid prescribing guidelines for chronic noncancer pain (CNCP) aim to reduce opioid volume and dosage while addressing risks such as misuse, poisoning, and death. However, these guidelines may have spillover effects on patient populations beyond the intended scope.AimsThis study aims to summarize the impact of prescribing guidelines on opioid use prevalence, incidence, and dosage across multiple jurisdictions and patient populations.MethodsIn October 2025, we systematically searched MEDLINE, Scopus, Web of Science, and PsycInfo for studies on the effects of opioid prescribing guidelines in adult populations. Primary outcomes included opioid use prevalence, incidence, dosage, and duration. We calculated study-level effect estimates using interrupted time series analyses and conducted meta-analyses with random effects models to estimate the pooled effect of the guidelines on the aforesaid outcomes.ResultsOf 190 records, 71 studies met inclusion criteria for meta-analysis. After guideline implementation, morphine milligram equivalent (MME) dosage per person and prescription duration decreased annually by 10.4% (95% confidence interval [CI] −19.2% to −0.6%) and 5.6% (95% CI −8.1% to −3.0%), respectively, among patients with CNCP. In the general population, MME dosage per person decreased annually by 7.6% (95% CI −11.3% to −3.7%), whereas duration remained stable over time (−1.4%; 95% CI −6.5% to 4.0%). Following guideline implementation, declines in prevalence, dosage, and duration were also observed in patients living with cancer or treated for acute pain.ConclusionGuidelines reduced opioid prescribing for CNCP. Future research is needed to examine whether and to what extent similar changes occurred among other patient populations.KEYWORDS: Opioids, prescriptions, practice guidelines, interrupted time series analysis, meta-analysis Introduction The early 2000s witnessed a global surge in opioid consumption,1 accompanied by rising rates of opioid use disorder, overdose, and mortality,2–4 culminating in epidemics in some countries.5,6 These crises prompted strategies to curb opioid use, including interventions targeting opioid supply, such as drug delistings, prescription monitoring programs, and prescribing guidelines for chronic noncancer pain (CNCP). Opioid prescribing for CNCP, which often leads to long-term opioid use, was particularly targeted due to evidence questioning the safety and effectiveness of long-term opioid therapy.7 Although opioids may relieve pain, the associated risks, including opioid use disorder, overdose, and death, frequently outweigh their benefits.7,8 Beginning in the mid-2000s, health institutions across the world released opioid prescribing guidelines to standardize opioid use for CNCP.9–20 Though intended to promote safer prescribing, these guidelines may have contributed to restricted access to adequate pain management, particularly for populations outside their scope, such as those with acute, cancer, or palliative pain. Misinterpretation by physicians also led to harmful practices like abrupt tapering or discontinuation for patients on long-term opioids, pushing some toward unregulated markets.8 Thus, though the guidelines reduced overprescribing for CNCP, they may have inadvertently encouraged underprescribing in other patient groups. Existing studies evaluating the impact of opioid prescribing guidelines have reported reductions in opioid dispensations and high-dose opioid prescriptions.21–24 However, most studies focus on a single guideline or jurisdiction, limiting insight into how guideline effects may evolve over time or vary across geographic regions. Studies also differ in the outcomes assessed and in how effects are reported (e.g., regression coefficients versus percentage changes), hindering comparison across guidelines and settings. To date, no comprehensive synthesis has evaluated the impact of opioid prescribing guidelines across multiple prescribing metrics using a common analytical framework. Moreover, evidence on spillover effects, particularly among guideline-exempt populations including patients with acute, cancer, or palliative pain, remains limited and underexplored. Consequently, the overall population-level impact of opioid prescribing guidelines and the extent to which their effects differ across pain populations remains unclear. Therefore, this systematic review and meta-analysis aimed to assess the impact of opioid prescribing guidelines on opioid dispensation and dosage on patients living with CNCP as well as with guideline-exempt pain types. Methods We report the results of our systematic review and meta-analysis in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses.25 Data sources and search strategy Identification of studies This review built upon and expanded a previous review commissioned by Health Canada that mainly focused on three guidelines in Canada and the United States by the College of Physicians and Surgeons of British Columbia,11 the U.S. Centers for Disease Control and Prevention,12 and Health Canada.9 For internal validation purposes, we manually reviewed the reference lists of studies included in the previous review to identify records missed in our initial systematic search. We then expanded the project to include other international guidelines, prompting a repeat of the previous search. On May 21, 2024, we systematically searched MEDLINE (PubMed), Scopus, Web of Science, and PsycInfo (Ovid) for peer-reviewed articles and gray literature (e.g., government reports) on opioid prescribing trends to date. Our search strategy included keyword roots (e.g., “opioid analgesics,” “practice guidelines,” “practice patterns,” “standards,” “prescribing,” and “chronic pain”), their wildcard expansions, and associated MeSH (Medical Subject Headings) terms (Table S1). The search was updated on January 3, 2025, and October 14, 2025 (Figure 1; more details in Figure S1). Preferred Reporting Items for Systematic Reviews and Meta-Analyses flowchart of study selection process. Identification of guidelines We conducted an environmental scan for opioid prescribing guidelines for chronic noncancer pain by searching Google and Google Scholar, as well as reviewing the references of eligible studies. Germane opioid prescribing guidelines provided definitions of and were specific to prescribing practices for chronic noncancer pain (i.e., did not include additional recommendations on prescribing for other pain types) and explicitly provided recommendations related to dosage and/or duration of initial or continued opioid prescribing. When multiple versions or updates of a guideline existed, earlier versions were not excluded, because our objective was to assess the impact of the guideline in effect during the study period of each included analysis. For studies with observation periods spanning multiple guideline updates, each guideline implementation was treated as a distinct intervention, and postintervention effects were evaluated separately for nonoverlapping periods. Guidelines that addressed opioid prescribing for indications beyond CNCP without providing CNCP-specific recommendations were excluded. Details regarding guideline exclusion decisions are provided in Table S2. Eligibility criteria and definitions Population and pain type We categorized records by pain type: (1) general/nonspecific, (2) CNCP, (3) chronic, (4) acute, (5) cancer related, and (6) palliative. General/nonspecific studies reported opioid prescribing outcomes without specifying pain type. CNCP studies focused on pain lasting 3 or more months unrelated to recent cancer diagnoses, consistent with guideline definitions. Chronic pain studies captured dispensations to patients for >3 months but did not exclude patients living with cancer. Acute pain studies assessed prescribing for short-term pain (<1 month) due to sudden health events (e.g., trauma, surgery) or in emergency department settings. Cancer-related pain studies examined prescribing for patients living with cancer or oncologists’ prescribing practices, whereas palliative pain studies focused on terminal pain excluding end-stage cancer cases. Intervention The intervention of interest was the implementation of opioid prescribing guidelines or clinical practice standards at the regional, national, or jurisdictional level. We classified studies as guideline specific if they explicitly estimated the impact of a guideline or non-guideline specific if they reported prescribing trends during periods of guideline implementation. Outcomes We grouped study outcomes into five types: (1) prevalence, (2) incidence, (3) dosage, (4) duration, and (5) tapering/discontinuation, reflecting guideline goals to reduce the frequency, dosage, and duration of opioid use. Prevalence outcomes described the proportion of individuals dispensed opioids at a specific time, and incidence outcomes measured the proportion of opioid-naïve individuals dispensed an initial opioid prescription within a time interval. Dosage outcomes assessed opioid potency per prescription, population, or time unit (e.g., morphine milligram equivalent [MME] units).26 Duration outcomes described the length of opioid supply per prescription or over a given time interval (e.g., number of days of supply in a year). Tapering/discontinuation outcomes measured dosage reductions, with qualitative review reserved for tapering due to variability in definitions across studies. Study design We included studies that employed an observational or quasi-experimental design (e.g., retrospective or prospective cohort studies, ecological studies, interrupted time series analyses, controlled or uncontrolled before–after studies, and repeated cross-sectional time series analyses), Eligible studies were required to provide repeated measures of these outcomes over time (i.e., time series data) sufficient to assess temporal trends before and after an intervention or policy change. We included studies that reported data as aggregate measures (e.g., prescription counts or rates) or individual-level summaries, provided they allowed for the reconstruction or extraction of time series information suitable for interrupted time series analysis. Eligibility for qualitative systematic review Studies were included in the qualitative systematic review if they (1) were published before October 14, 2025, and either (2a) estimated the effects of opioid prescribing guidelines (“guideline-specific studies”) or (2b) reported prescribing trends during guideline implementation (“non-guideline-specific studies”). We excluded preprints, abstracts, commentaries, systematic reviews, modeling studies, qualitative studies without quantitative data, case series studies, animal studies, and those focused on local interventions, prescription drug monitoring programs, opioid scheduling changes, drug take-back programs, or legal mandates limiting opioid dispensing. We excluded abstracts and articles at the full-text review stage published in a language other than English or French due to limited resources. Eligibility for meta-analyses Studies were eligible for inclusion in the meta-analysis if they reported an opioid prescribing outcome (e.g., prevalence, incidence, dosage, duration, or tapering/discontinuation) measured at eight or more distinct time points in a time series, with at least two measurements before and two after the implementation of the intervention (excluding the implementation date itself). For articles with overlapping study populations, we applied a predefined selection rule prioritizing studies with longer time series, unless insufficient pre- or postintervention observations warranted selection of an alternative study with an adequate number of time points. Studies from the same region were retained if they used different, nonoverlapping databases. Studies drawing on the same underlying databases were retained if they examined distinct outcomes, pain types, underlying conditions, or nonoverlapping time windows. Screening Search records were imported into Covidence for duplicate removal and title and abstract screening. Two of six reviewers (M.S., I.W., J.L., I.F., P.T., and D.P.) independently screened titles and abstracts in Covidence,27 with conflicts resolved by consensus. Two of five reviewers (M.S., J.L., I.F., P.T., and D.P.) subsequently assessed full texts for inclusion based on predefined eligibility criteria. Records without full texts were excluded if authors could not be reached or were unable to provide full text. Data extraction Two of four reviewers (M.S., J.L., I.F., and P.T.) independently extracted predefined qualitative information (e.g., study population, period, data sources, and design) and validated results. Two reviewers (P.T. and J.L.) abstracted quantitative data from studies (i.e., outcomes reported and measures of effect) and one of three reviewers (I.W., M.S., and D.P.) validated the extracted data. For meta-analysis, time series data were extracted directly from studies or requested from authors. If unavailable, data were digitized from figures using PlotDigitizer,28 with extracted points independently reviewed to minimize error. Quality assessment We assessed the study quality using an adapted version of the “Quality Appraisal Checklist for Case Series Studies” from the Institute of Health Economics (see Table S3).29 Two of four reviewers (M.S., P.T., J.L., and D.P.) evaluated the quality of each study and resolved disagreements through consensus or adjudication by a third reviewer. Data analysis Systematic review For the systematic review, we visually assessed the extracted outcome measures to determine whether post-guideline trends were positive, negative, or unchanged relative to preintervention trends. Interrupted time series analyses For articles eligible for meta-analysis, we used interrupted time series (ITS) analysis to estimate immediate level and long-term trend changes in opioid prescribing outcomes after guideline implementation relative to the preintervention period. We applied segmented regression with outcome-specific distributions: negative binomial for counts (e.g., prescriptions dispensed), beta for proportions (e.g., percentage of patients dispensed opioids), and log-linear for dosage rates (e.g., MME per day). To account for autocorrelation, we calculated 95% confidence intervals using Newey-West standard errors,30–32 with a lag order equal to the number of time series measurements to the power of ¼, and otherwise specified lag orders based on assessment of residual autocorrelation.33 To standardize time, we centered the calendar time variable on the guideline implementation date, defined as the first day of the implementation year, quarter, or month. For studies that had intervals that exceeded 1 year, we conducted linear interpolation to estimate values at yearly time points. The intervention level was coded as 0 prior to and including the guideline implementation date and 1 post-guideline implementation. For each time series, we fit the following regression model: Our outcomes of interest (Y) included annual measures of prevalence, incidence, dosage, and duration. In each model, β1 estimated the percentage change in the outcome over time before the intervention (preintervention trend). β2 estimated the immediate percentage change in the outcome level following the intervention (level effect), representing an abrupt increase or decrease occurring directly after the guideline implementation. β3 estimated the change in the postintervention trend relative to the preintervention trend (trend effect). When exponentiated, β2 and β3 represent percentage changes in the outcome level and trend, respectively. In other words, the level effect reflects an immediate shift in the outcome at the time of the implementation, whereas the trend effect reflects a gradual change in the trajectory of the outcome following the intervention. Meta-analysis We conducted meta-analyses by pain type (e.g., general/nonspecific, chronic, CNCP, acute, and cancer) if at least three studies reported the same outcome. Using study-specific ITS estimates, we calculated pooled immediate percentage changes (level effects) and percentage changes per year (trend effects) attributable to guidelines. Heterogeneity was assessed using the I2 and τ2 statistic. We reported random effects estimates derived from the restricted maximum likelihood method to account for the potentially high level of between-study heterogeneity (I2 > 70). We conducted all statistical analyses using R statistical software v4.3.2.34 Results Guideline selection and characteristics We identified 27 pertinent opioid prescribing guidelines for CNCP implemented between March 2007 and January 2022 (Table 1).9–20,34–48 These guidelines provided recommendations for the duration and dosage of opioid prescriptions as well as criteria for initiating, monitoring, and continuing therapy. Across all guidelines, there was a consistent emphasis on prioritizing nonopioid treatments and considering opioids only when other options have failed, basing opioid initiation on objective clinical evidence rather than continuation of previous prescriptions, implementing rigorous patient monitoring throughout therapy, and seeking specialist consultation or providing clear justification for high-dose regimens. Twenty-five out of 27 included guidelines reported a hard threshold (e.g., 90 MME/day, 200 MME/day, etc.) for reassessment, tapering, or discontinuation of opioid therapy.9–20,34–37,39–41,43,45–48 Sixteen guidelines were disseminated at the national level,9,10,12–16,19,34,37,39,41–43,46,47 whereas 11 were administered by authority groups at the provincial or state levels.11,17,18,20,35,36,38,40,44,45,48 Overall, the included guidelines originated from the United States (n = 9), Europe (n = 10), Canada (n = 4), and Australia/Asia (n = 4). Table 1.Characteristics of included opioid prescribing guidelines for chronic noncancer pain.ContinentOpioid prescribing guidelineAuthority/author groupAbbreviationGeographic jurisdictionImplementation monthKey recommendationsNorth AmericaCanadian Guideline for Safe and Effective Use of Opioids for Chronic Non-Cancer PainNational Opioid Use Guideline GroupNOUGG (2010)CanadaJune 2010(1) Initiate opioid therapy only after thorough assessment and goal‑setting2) <200 MME dose per day(3) Regular reevaluation of treatment and discontinuation of opioid treatment if benefit is insufficient or harms predominate The 2017 Canadian Guideline for Opioids for Chronic Non-Cancer PainHealth CanadaHealth Canada (2017)CanadaMay 2017(1) Optimize nonopioid pharmacotherapies rather than immediately starting opioids(2) Maximum 90 MME/day (strong recommendation) but tapering to the lowest effective dose (or discontinuation) if already >90 MME/day(3) Advises against opioid initiation in patients with active substance use disorders or uncontrolled psychiatric comorbidity Douleur chronique et opioïdes: l′essentielCollège des médecins du QuébecCMQ (2009)Quebec, CanadaMay 2009(1) Clear consent and patient engagement(2) Prioritization of nonopioid treatment but establishment of clear plan if opioid treatment becomes necessary(3) Rigorous monitoring and assessment of risks and benefits of opioid treatment (200 MME/day threshold) College of Physicians and Surgeons of British Columbia Safe Prescribing of Opioids and Sedatives Practice StandardCollege of Physicians and Surgeons of British ColumbiaCPSBC (2016)British Columbia, CanadaJune 2016(1) Long-term opioid treatment based on a clear diagnosis, objective evidence, and comprehensive assessment(2) Doses >50 MME/day require caution; >90 MME/day requires strong justification along with close monitoring(3) Avoid high-risk and unsafe prescribing practices (e.g., coprescription of opioids and sedatives, prescription of large quantities of opioids) Clinical Guidelines for the Use of Chronic Opioid Therapy in Chronic Noncancer PainAmerican Pain Society, American Academy of Pain MedicineAPS (2009)United StatesFebruary 2009(1) Rigorous patient selection with risks vs. benefits assessment(2) Treat opioid therapy as a trial, with informed consent and monitoring plan(3) Ongoing monitoring, reassessment, and action on escalation or risk (200 MME/day threshold) CDC Guideline for Prescribing Opioids for Chronic PainCenters for Disease Control and PreventionCDC (2016)United StatesMarch 2016(1) Prioritization of nonopioid and nonpharmacologic treatments(2) Special caution at or above 50 MME/day, with strong justification required at or above 90 MME/day(3) Minimization of opioids in higher risk patients (e.g., active substance use disorders or psychiatric comorbidities) while implementing tapering when needed Washington State Interagency Guideline on Opioid Dosing for Chronic Non-Cancer PainWashington State Agency Medical Director’s GroupWA (2007)WashingtonMarch 2007(1) Warning threshold of 120 MME/day; pain management consultation for patients at or above this dose whose pain and function have not substantially improved(2) Regular assessment of safety and effectiveness by tracking sustained improvement in function and pain(3) Best practices for mitigating risk (e.g., opioid agreements and assessment of mental health and substance abuse history) Washington State Interagency Guideline on Opioid Dosing for Chronic Non-Cancer Pain (Update)Washington State Agency Medical Director’s GroupWA (2010)WashingtonJune 2010(1) 120 MME/day dosage threshold, not to be exceeded unless the patient shows sustained improvement in both function and pain or a consultation with a pain management specialist is obtained(2) Opioid therapy as last resort after a trial of appropriate nonopioid medications, physical therapy, and behavioral therapies has failed(3) Mandatory monitoring for compliance (via urine drug testing) and efficacy along with explicit opioid agreement with the patient Assessment and Management of Chronic PainInstitute for Clinical Systems ImprovementICSI (2011)MinnesotaNovember 2011(1) Focus on improving patients’ function rather than solely eliminating pain via the biopsychosocial model(2) Physical rehabilitation and psychosocial management with functional goals before considering specialty referrals or more complex treatments(3) Careful patient selection and close monitoring for all chronic opioid patients, with consultation with a pain specialist for doses exceeding the suggested maximum of 200 MME/day Opioid Prescribing Guidelines for Treatment of Chronic, Non-Terminal PainOhio Academy of Family PhysiciansOAFP (2013)OhioOctober 2013(1) Prioritization of nonpharmacologic and nonopioid therapies; if opioids are considered, obtain informed consent, define functional goals, and check the state prescription monitoring database(2) 80 MME/day threshold, above which providers should reassess the treatment plan, review functional outcomes, reconsent, and consider specialist referral or alternative treatments(3) Avoidance of high-risk combinations (e.g., opioids with sedatives) without careful review; greater caution in patients with substance use or mental health risks Pennsylvania Guidelines on the use of Opioids to Treat Chronic Non-Cancer PainPennsylvania Medical SocietyPMS (2014)Pennsylvania2014(1) Multimodal treatment, where opioids are rarely the sole treatment modality and are considered within the context of multimodal therapy (e.g., physical therapy and behavioral health)(2) Careful treatment consideration and consultation for specialty care when dose above the 100 MME/day threshold(3) Mandatory monitoring for efficacy, adverse events, and adherence and discontinuation if patients show no progress toward therapeutic goals or engage in repeated aberrant behaviors Massachusetts Medical Society Opioid Therapy and Physician Communication GuidelinesMassachusetts Medical SocietyMA (2015)MassachusettsAugust 2015(1) A thorough initial assessment, documented risk stratification (low, moderate, or high risk) of the patient, informed consent process, and opioid treatment agreement before initiating chronic opioid therapy(2) Consultation with a pain specialist before prescribing a daily dose that exceeds 100 MME/day or if the patient is categorized as high risk(3) Integration of multidisciplinary care by incorporating nonopioid therapies, physical therapy, behavioral health, and alternative strategies to improve function Prescribing Guidelines for Pennsylvania. Treating Chronic Non-Cancer PainCommonwealth of PennsylvaniaCOP (2018)PennsylvaniaJune 2018(1) Prioritize nonopioid/nonpharmacologic approaches and establish clear goals(2) Use of opioids only when clearly indicated, at the lowest effective dose, with rigorous monitoring (90 MME/day threshold)(3) Avoidance of high-risk prescribing combinations and mitigation of overdose riskEuropeNational Clinical Guideline on Opioid Treatment of Chronic Nonmalignant PainDanish Health AuthorityDHA (2018)DenmarkDecember 2018(1) Prioritization of nonopioid treatments and use opioids only for selected patients(2) Restriction and careful monitoring of opioid dosage (not to exceed 100 MME/day, while considering limitation to 50 MME/day)(3) Avoidance of opioids in high-risk settings, while mandating reassessment or tapering when needed Opioid Use for the Management of Chronic Non-Cancer Pain: French GuidelinesSociété Française d’Etude et Traitement de la DouleurSFETD (2016)FranceJune 2016(1) Multimodal assessment and nonopioid treatment options before opioids; administration of this treatment modality only if the patient has a multimodal treatment failure and the expected benefit outweighs the risks(2) Discontinuation of opioid treatment after 3 months in the absence of clinically significant improvement in pain and quality of life(3) 150 MME/day dosage threshold and expert review recommended for dosage that exceeds this threshold German Pain Society Guideline, first editionGerman Pain SocietyGPS (2009)GermanySeptember 2009(1) Consideration of long-term opioid therapy only after failure of nonopioid and nonpharmacologic treatments and in patients with well-defined pain conditions likely to respond (e.g., osteoarthritis, chronic low back pain, or neuropathic pain)(2) Treatment initiation with the lowest effective dose, typically not exceeding 120 MME/day and longer than 6 months without documented functional improvement and regular specialist review(3) Strict monitoring of pain, function, side effects, and misuse risk and discontinuation if goals are unmet, tolerance develops, or adverse effects outweigh benefits German Pain Society Guideline, second editionGerman Pain SocietyGPS (2015)GermanyJanuary 2015(1) Long-term opioid therapy is recommended for specific chronic pain conditions (e.g., osteoarthritis, lower back pain) but not for primary headaches and functional/psychiatric pain disorders(2) Mandatory efficacy review: Opioid therapy should be discontinued unless the patient shows significant pain reduction or functional improvement(3) Physicians are advised to adhere to 120 MME/day threshold; doses exceeding this threshold require special justification and expert consultation German Pain Society Guideline, third editionGerman Pain SocietyGPS (2020)GermanySeptember 2020(1) Restrictive indications (e.g., osteoarthritis, diabetic polyneuropathy, postherpetic neuralgia, and low back pain) and contraindications (e.g., primary headaches, functional somatic syndromes, and mental disorders)(2) Long-term therapy (defined as >26 weeks) is only justified if there is a positive therapy response; regular assessment of both clinical response and adverse effects to justify continuation of opioid treatment(3) Conservative dosing: Maintains the 120 MME/day threshold set in the previous version of this guideline Scottish Intercollegiate Guideline Network Management of Chronic PainScottish Intercollegiate Guidelines Network, Healthcare Improvement ScotlandSIGN 136 (2013)ScotlandDecember 2013(1) Multimodal and biopsychosocial assessment to identify the pain type and functional impact as well as to inform the selection of the most effective treatment options(2) Prioritization of nonpharmacological strategies, including exercise and exercise therapies for the management of chronic pain(3) Specialist advice if the patient needs >180 MME/day or has concerns about rapid dose escalation, with patients being reviewed at least annually Guía de Consenso para el Buen Uso de Analgésicos OpioidesSocidrogalcoholS-MS (2017)SpainJune 2017(1) Prioritization of nonopioid and nonpharmacologic treatments as first-line therapy and consideration of opioid analgesics only when the potential benefits (pain and function) clearly outweigh the risks(2) Prior to initiation, patients must receive comprehensive information about opioid therapy (including risks such as tolerance, hyperalgesia, addiction, overdose); structured treatment with defined functional goals, frequent reassessments, and a plan for dose reduction or discontinuation if outcomes are insufficient(3) Strict monitoring of dosage and risk factors: Close monitoring above 50 MME/day and clear justification in the patients’ medical records for doses >90 MME/day Guía de Atención a Los Pacientes con Dolor Crónico no Oncológico Utilizando Analgésicos OpioidesDepartament de Salut. Generalitat de CatalunyaDPS (2018)Catalonia, SpainDecember 2018(1) Long‑term opioid therapy for CNCP should only be considered after thorough assessment(2) Opioid prescriptions must be based on a clinical diagnosis and supported by objective evidence, not simply continuation of previous opioid prescriptions(3) Rigorous monitoring, including documentation of discussions with the patient regarding risks (including misuse and dependence) and readiness to adjust or discontinue therapy when benefit is inadequate Dolor Crónico no-Oncológico: ¿Opioides? INFACOsakidetza-Basque Health Service, Gobierno Vasco: Departamento de SaludO-GV (2022)Basque Country, SpainJanuary 2022(1) Opioid medications should not be routine for CNCP due to uncertainty in long-term benefit and substantial risks (e.g., dependence, hyperalgesia, and adverse effects)(2) Optimization of nonopioid pharmacologic and nonpharmacologic therapies, by assessing psychological and social factors and setting function-based goals rather than just pain relief(3) Opioid use to be considered only for highly selected patients, with clear documentation of risks and benefits, ongoing monitoring of function versus harm, and readiness to stop if benefit is not achieved (with 50 MME/day and 90 MME/day thresholds) Opioids for Persistent Pain: Good PracticeBritish Pain SocietyBPS (2010)United KingdomJanuary 2010(1) Opioids should not be used as first-line therapy when evidence-based nonopioid and nonpharmacologic treatments are available(2) If there is no meaningful improvement (e.g., after reaching 120–180 MME/day), referral to specialist care is strongly recommended(3) Clinicians should assess for factors such as prior or current substance misuse, mental health disorders, and social context and ensure long-term monitoring is in placeAsia/OceaniaFaculty of Pain Medicine, Australian and New Zealand College of Anaesthetists Recommendations Regarding the Use of Opioid Analgesics in Patients with Chronic Noncancer PainFaculty of Pain Medicine, Australian and New Zealand College of AnaesthetistsAustralia/New Zealand (2015)Australia, New ZealandDecember 2015(1) Multimodal treatment and thorough assessments required, while prioritizing nonopioid treatments as first-line therapy(2) Caution at 40 MME/day with prompt reassessment and specialist advice at doses above 100 MME/day(3) Ongoing review, monitoring, and readiness to taper or cease opioid treatment A Guide to the Use of Strong Opioids in Chronic Non-Cancer PainMalaysian Association for the Study of PainMASP (2015)MalaysiaAugust 2015(1) Optimize nonopioid and interventional treatments, while limiting opioid treatment to being part of a multimodal treatment plan(2) A 3- to 4-week trial period for opioid therapy and discontinuation if no significant improvement in pain and function is observed after this period(3) Detailed documentation and monitoring of pain relief, activities of daily living (function), adverse effects, and aberrant drug-related behaviors Pain Association of Singapore: Evidence-Based Guidelines on the Use of Opioids in Chronic Non-Cancer PainPain Association of SingaporePAS (2013)SingaporeMarch 2013(1) Opioids are not recommended as first-line treatment for CNCP, but they may be used as second- or third-line therapy, preferably within a multimodal approach (maximum of 200 MME/day)(2) Careful screening and assessment of patients prior to starting opioids, with an opioid treatment agreement that may include urine drug testing(3) Trial duration of up to 2 months is necessary to determine efficacy of opioid treatment (e.g., pain relief, improvement in function and quality of life), with regular reviews to assess efficacy, adverse effects, and aberrant behavior Guidelines for Prescribing Opioids for Chronic Non-Cancer Pain in KoreaKorean Pain SocietyKPS (2017)South KoreaJanuary 2017(1) Prioritization of nonopioid and nonpharmacologic therapies(2) Opioid use based on clear diagnosis and reassessment, by relying on objective evidence, not continuation of past prescriptions, along with regular monitoring of pain, function, and risk(3) Careful reassessment of dosage at ≥50 MME/day and justification of exceptional benefit for doses ≥90 MME/day; tapering or discontinuation recommended if functional improvement is insufficient Characteristics of included opioid prescribing guidelines for chronic noncancer pain. Study selection and characteristics Our systematic search yielded 15,243 records, of which 1092 were assessed for eligibility in full-text review and 129 studies were ultimately retained. Further, we added 61 studies from a previous review commissioned by Health Canada and its internal validation (Figure 1). In total, we included 190 studies reporting opioid prescribing time series related to prescribing guidelines for the qualitative systematic review (Table S4) and 71 studies for the meta-analysis (Table 2). Studies included for systematic review reported on opioid prescribing in the United States (n = 146), Europe (n = 16), Canada (n = 13), and Australia/Asia (n = 12), with 3 studies focusing on multiple countries. Table 2.Characteristics of studies included in the meta-analysis (n = 71).StudyLocationPeriod of observationGuidelineaPain typeData sourceOutcomesTime series length (pre, post)Unit of measurementAdalbert et al.165Pennsylvania2016–2020COP (2018)GeneralPennsylvania prescription drug monitoring programsNumber of prescriptions15 (7, 7)QuarterAdewumi et al.59Queensland, Australia1997–2018ANZCA (2015)GeneralMonitoring of Drugs of Dependence systemNumber of prescriptions, prevalence (%)22 (18, 3)YearAhn et al.137South Korea2010–2022KPS (2017)ChronicNational Health Insurance Sharing Service–National Health Information Database of South KoreaPrevalence (%)13 (7, 5)YearAli et al.168United States2005–2015APS-AAPM (2009)GeneralIBM MarketScan Commercial Claims and Encounters Database, Multi-State Medicaid DatabasePrevalence (N)11 (4, 6)YearAubry et al.75United States2010–2019CDC (2016)GeneralDrug Overdose Deaths (National); Total Overdose Deaths, Any Opioid Overdose Deaths and Prescription Opioid Overdose DeathsMME per person10 (6, 3)YearBandara et al.49United States2012–2019CDC (2016)CNCP, cancerIBM MarketScan Research DatabasesPrevalence (%)8 (4, 3)YearBedson et al.216United Kingdom2002–2012BPS (2010)GeneralCPRDNumber of prescriptions11 (8, 2)YearBhattacharya et al.84United States2012–2020CDC (2016)GeneralMedicareMME per day, duration, incidence (%)8 (3, 4)YearBohnert et al.21United States2012–2017CDC (2016)GeneralIQVIA transactional data warehouseMME per person, duration72 (50, 21)MonthBrett et al.72Australia, New Zealand, United States, Canada, United Kingdom, Denmark, South Korea2000–2020ANZCA (2015); APS-AAPM (2009); CDC (2016); Health Canada (2017); BPS (2010); DHA (2018); KPS (2017)General, CNCPAustralia: Administrative claims data, Pharmaceutical Benefits Scheme dataNew Zealand: Ministry of Health National Collections, Pharmaceutical CollectionUnited States: Merative Marketscan Commericial Claims and Encounters, MedicaidCanada: Institute for Clinical Evaluative Sciences databaseUnited Kingdom: CPRD GOLD databaseDenmark: Central Registration of Medication UseSouth Korea: Health Insurance Review and Assessment DatabasePrevalence (%)Australia: 7 (2, 4)New Zealand: 16 (11, 4)U.S. private: 10 (6, 3); 8 (3, 4)U.S. public: 19 (9, 9)Canada: 8 (4, 3)UK: 16 (5, 10)Denmark: 20 (17, 2)South Korea: 11 (7, 3)YearCanizares et al.153Ontario, Canada2004–2017NOUGG (2010)AcuteOntario administrative dataIncidence (%)14 (6, 7)YearCasagrande et al.60United States2008–2019CDC (2016)GeneralMedicarePrevalence (%)12 (8, 3)YearChai et al.76United States1997–2015APS-AAPM (2009)GeneralIQVIA National Prescription AuditNumber of prescriptions, MME per person19 (12, 6)YearChampagne-Langabeer et al.173United States2015–2017CDC (2016)GeneralTexas Prescription Monitoring ProgramNumber of prescriptions12 (5, 6)QuarterChen et al.50United States2003–2014APS-AAPM (2009)CNCPTruven MarketScan databasePrevalence (%)10 (5, 4)YearCohen et al.51United States2000–2019APS-AAPM (2009)CNCPWomen’s Interagency HIV StudyPrevalence (%)20 (9, 10)YearCrabtree et al.61British Columbia, Canada2015–2017CPSBC (2016)GeneralBritish Columbia administrative dataNumber of prescriptions, prevalence (%), incidence (%), MME per day30 (17, 12)MonthCurtis et al.77United Kingdom1998–2017BPS (2010)GeneralNHS DigitalMME per person20 (12, 7)YearDart et al.176United States2002–2013APS-AAPM (2009)GeneralRADARS System ProgramsNumber of prescriptions32 (12, 19)QuarterDavies et al.62Wales2005–2015BPS (2010)GeneralSAIL DatabankPrevalence (%)11 (5, 5)YearDay et al.161United States2010–2020CDC (2016)AcutePearlDiver Mariner91 national administrative claims databaseMME per day11 (6, 4)YearDiPrete et al.214United States2010–2018CDC (2016)GeneralMarketScanIncidence (N)9 (6, 2)YearFranklin et al.180United States1996–2010WA (2007)GeneralWashington State Department of Labor Industries Medical Information Payment SystemMME per day178 (134, 43)MonthGarcia et al.213Massachusetts2002–2017APS-AAPM (2009)GeneralMassHealth pharmacy claimsMME per day61 (28, 32)QuarterGarg et al.64Washington2004–2010WA (2007)GeneralWashington State Department of Labor Industries Medical Information Payment SystemPrevalence (%)27 (12, 14)QuarterGomes et al.78Canada2008–2016NOUGG (2010)GeneralIQVIA Canada CompuscriptMME per person95 (25, 69)MonthGoudman et al.128United States2012–2022CDC (2016)ChronicElectronic health records from TriNetXPrevalence (%), incidence (%)11 (4, 6)YearGupta et al.183United States2015–2017CDC (2016)GeneralUnityPoint Health Records SystemNumber of prescriptions36 (14, 21)MonthGuy et al.82United States2006–2015APS-AAPM (2009)GeneralQuintilesIMS Transactional Data WarehouseMME per day, duration10 (3, 6)YearHan et al.129United States2006–2015APS-AAPM (2009)ChronicMedicarePrevalence (%)10 (3, 6)YearHauser et al.57United States, Canada, Germany1980–2014APS-AAPM (2009); NOUGG (2010); GPS (2009)CNCPInternational Narcotics Control BoardMME per person35 (30, 4)YearHebert et al.22United Kingdom2005–2020SIGN (2013)GeneralPublic Health Scotland Prescribing Information SystemNumber of prescriptions184 (107, 76)MonthHechter et al.185California2013–2020CDC (2016)GeneralKaiser Permanente Southern California electronic health record databasesPrevalence (N), incidence (N)8 (3, 4)YearHedenmalm et al.79Germany2006–2016GPS (2009)GeneralIMS Disease AnalyzerMME per person, duration11 (3, 7)YearHolmer et al.117United States2012–2020CDC (2016)ChronicVA Corporate Data WarehousePrevalence (%)9 (4, 4)YearHu et al.81United States2011–2021CDC (2016)GeneralU.S. DEA Automated Reports and Consolidated Ordering SystemMME per person44 (20, 23)QuarterKang et al.23United States2011–2019CDC (2016)CNCPMerative MarketScan Commercial DatabaseMME per person, duration108 (62, 45)MonthKazanis et al.189United States2006–2013APS-AAPM (2009)GeneralIMS Health National Prescription Audit and Vector One: NationalNumber of prescriptions8 (3, 4)YearKimmel et al.121United States2011–2020CDC (2016)ChronicU.S. Renal Data System, MedicarePrevalence (%)10 (5, 4)YearLee et al.130United States2002–2015APS-AAPM (2009)ChronicCorrona RA RegistryPrevalence (%)14 (7, 6)YearLeja et al.138Michigan2015–2018CDC (2016)ChronicSaint Joseph Mercy hospital HERIncidence (%)41 (14, 26)MonthLiu et al.65United States2013–2021CDC (2016)GeneralMedicarePrevalence (%)9 (3, 5)YearLy et al.73Quebec, Canada1997–2018CMQ (2009)GeneralRAMQ, MED-ECHOPrevalence (%)11 (6, 4)YearMaierhofer et al.58United States2012–2018CDC (2016)CNCP, acutePrivate health insurance claims dataMME per day, duration72 (50, 21)MonthMauck et al.151United States2001–2020APS-AAPM (2009)AcuteIBM MarketScanIncidence (%)19 (8, 10)YearMojtabai66United States1999–2014APS-AAPM (2009)GeneralNHANESPrevalence (%)15 (10, 4)YearNahin et al.67United States1997–2014APS-AAPM (2009)GeneralMedical Expenditure Panel SurveyPrevalence (%)17 (12, 4)YearPanagiotoglou et al.24British Columbia, Canada2012–2020CPSBC (2016)CNCP, cancer, palliativeBC administrative dataMME per person84 (41, 42)MonthPensa et al.68United States2003–2013APS-AAPM (2009)GeneralClaims data on an industrial cohort from a specialty metals corporation with multiple manufacturing facilities geographically dispersed throughout the United StatesPrevalence (%)11 (6, 4)YearPritchard et al.52United States2011–2019CDC (2016)CNCP, acuteMedical Expenditure Panel SurveyPrevalence (%)9 (5, 3)YearRancu et al.147United States2011–2021CDC (2016)AcutePearlDiver Mariner 161 Insurance DatabasePrevalence (%), MME per person11 (5, 5)YearRancu et al.148United States2011–2021CDC (2016)AcutePearlDiver Mariner 161 Insurance DatabasePrevalence (%), MME per person11 (5, 5)YearRichter et al.136Minnesota2005–2015APS-AAPM (2009)ChronicRetrospective data from Rochester Epidemiology ProjectPrevalence (%)11 (4, 6)YearRogal et al.124United States2005–2014APS-AAPM (2009)ChronicVA Corporate Data WarehousePrevalence (%)10 (4, 5)YearRolova et al.74Denmark2010–2023DHA (2018)GeneralPublic Prescription Drug StatisticsPrevalence (%), MME per day14 (8, 5)YearSalvatore et al.200United States2015–2019CDC (2016)GeneralIQVIA Longitudinal Prescription databaseNumber of prescriptions, MME per day60 (14, 45)MonthSchieber et al.83United States2006–2017APS-AAPM (2009)GeneralIQVIA Xponent databaseNumber of prescriptions, duration12 (3, 8)YearSchieber et al.91United States2008–2018CDC (2016)GeneralIQVIA Total Patient TrackerPrevalence (%)11 (8, 2)YearScott et al.125North Staffordshire, UK2000–2015BPS (2010)ChronicConsultations in Primary Care ArchivePrevalence (%)16 (10, 5)YearSears et al.70Washington2008–2015WA (2010)GeneralWorkers’ Compensation dataPrevalence (%)48 (29, 18)MonthSears et al.69Washington2012–2017CDC (2016)GeneralWashington State Prescription Monitoring Program dataPrevalence (%, N), incidence (%, N)24 (17, 6)QuarterShen et al.53United States2006–2013APS-AAPM (2009)CNCPMedicare Current Beneficiary SurveyPrevalence (%)8 (3, 4)YearSmith et al.203Illinois2009–2018CDC (2016)GeneralNorthwestern Medicine Enterprise Data WarehousePrevalence (%)10 (7, 2)YearSmolina et al.80British Columbia, Canada2005–2013NOUGG (2010)GeneralBritish Columbia administrative dataMME per person9 (5, 3)YearSong et al.89South Korea2010–2019KPS (2017)CNCPNational Health Insurance Service (NHIS) databasePrevalence (%)10 (7, 2)YearStokes et al.54United States1999–2016APS-AAPM (2009)CNCPNHANESPrevalence (%)17 (10, 6)YearSurbhi et al.105United States2004–2018APS-AAPM (2009); CDC (2016)CNCPU.S. Renal Data SystemIncidence (%)15 (12, 2)YearTaqi et al.3449United Kingdom2000–2014BPS (2010)GeneralCPRDMME per person15 (10, 4)YearWhitney et al.127United States2011–2017CDC (2016)ChronicOptum Clinformatics Data Mart DatabasePrevalence (%)81 (59, 21)MonthZamora-Legoff et al.133Minnesota2005–2014APS-AAPM (2009)ChronicRochester Epidemiology ProjectPrevalence (%)10 (4, 5)YearZhu et al.71United States2012–2017CDC (2016)GeneralBlue Cross Blue Shield Axis claims databasePrevalence (%), incidence (%), MME per day66 (44, 21)MonthaSee Table 1, Abbreviation: ANZCA = Australian and New Zealand College of Anaesthetists; APS-AAPM = American Pain Society-American Academy of Pain Medicine; BPS = British Pain Society; CDC = Centers for Disease Control and Prevention; CNCP = Chronic non-cancer pain; COP = Commonwealth of Pennsylvania; GPS = German Pain Society; KPS = Korean Pain Society; MME = Morphine milligram equivalent; NOUGG = National Opioid Use Guideline Group; SIGN = Scottish Intercollegiate Guideline Network; S-MS = Socidrogalcohol-Ministerio de Sanidad ; UK = United Kingdom; USA = United States of America. Characteristics of studies included in the meta-analysis (n = 71). aSee Table 1, Abbreviation: ANZCA = Australian and New Zealand College of Anaesthetists; APS-AAPM = American Pain Society-American Academy of Pain Medicine; BPS = British Pain Society; CDC = Centers for Disease Control and Prevention; CNCP = Chronic non-cancer pain; COP = Commonwealth of Pennsylvania; GPS = German Pain Society; KPS = Korean Pain Society; MME = Morphine milligram equivalent; NOUGG = National Opioid Use Guideline Group; SIGN = Scottish Intercollegiate Guideline Network; S-MS = Socidrogalcohol-Ministerio de Sanidad ; UK = United Kingdom; USA = United States of America. Most reported time series overlapping with the CDC12 (n = 90) and the American Pain Society-American Association of Pain Medicine (APS-AAPM)10 guideline implementation dates (n = 59), with additional studies reporting time series overlapping with the Canadian National Opioid Use Guideline Group (NOUGG; n = 9)14 and Australian/New Zealand (n = 8).13 Among the articles included for the qualitative systematic review (Table S4), the majority of studies examined opioid dispensation in the general population (n = 97), followed by CNCP (n = 34), acute pain (n = 29), chronic pain (n = 26), and cancer pain (n = 9) populations. Quality assessment Of the 190 studies included for the systematic review, 26 were guideline-specific studies and 164 were non-guideline-specific studies. Upon quality assessment, we observed that 159 studies had a clearly stated objective, 144 had a clearly defined study population, 185 defined outcome measures a priori (i.e., prespecified), and 163 presented study conclusions that were consistent with the results. Through examination of disclosure and ethics statements, we identified 147 studies that explicitly stated both competing interests and financial support and 143 studies that explicitly stated ethics approval or exemption (Supplementary Materials Table S5). Meta-analyses Chronic noncancer pain For patients with CNCP (Figure 2), guidelines did not have an immediate level effect (6.5%, 95% confidence interval [CI] −1.0% to 14.6%) or trend effect (−2.6%, 95% CI −5.8% to 0.7%) on the prevalence of people prescribed opioids (N = 7 from 10 different CNCP populations).49–55 Similarly, there was no meaningful immediate level change in MME per person (−2.3%, 95% CI −10.9% to 7.3%), but there was a 10.4% decrease in the yearly trend (95% CI −19.2% to −0.6%; N = 4 from 6 different patient populations; Figure 3).23,24,56,57 Finally, there was no level change in prescription duration (2.1%, 95% CI −4.5% to 9.0%) immediately following implementation, but there was a 5.6% decrease in prescription duration year to year (95% CI −8.1% to −3.0%; N = 3) following guideline implementation (Figure 4).23,55,58 Forest plot of the percentage changes in level and trend for the prevalence of patients with CNCP prescribed opioids following guideline implementation. APS-AAPM = American Pain Society-American Association of Pain Medicine; AS = ankylosing spondylitis; CDC = Centers for Disease Control and Prevention; CNCP = Chronic non-cancer pain; LCL = Lower confidence limit; PsA = psoriatic arthritis; RA = rheumatoid arthritis; SLE = systemic lupus erythematosus; UCL = Upper confidence limit. Forest plot of the percentage changes in level and trend for the MME dose per person in the chronic noncancer pain population following guideline implementation. APS-AAPM = American Pain Society-American Association of Pain Medicine; BPS = British Pain Society; CDC = Centers for Disease Control and Prevention; GPS = German Pain Society; LCL = Lower confidence limit; MME = Morphine milligram equivalent dose; NOUGG = National Opioid Use Guideline Group; UCL = Upper confidence limit; US = United States. Forest plot of the percentage changes in level and trend for the average duration of opioid prescriptions in patients with CNCP following guideline implementation. APS-AAPM = American Pain Society-American Association of Pain Medicine; CDC = Centers for Disease Control and Prevention; LCL = Lower confidence limit; UCL = Upper confidence limit. General population For the general population (Figure 5), there was no immediate level change (0.5%, 95% CI −2.8% to 3.9%) but an 7.6% yearly trend decrease (95% CI −10.5% to −4.7%) in the prevalence of opioid use following guideline implementation (N = 16 from 22 different populations).59–74 There was limited level change (2.0%, 95% CI 0.0 to 4.1) but a 7.6% yearly trends (95% CI −11.3% to −3.7%) in MME per person (N = 8; Figure 6).21,75–81 Further, we did not find level or trend changes for the duration of opioid prescriptions following guideline implementation (N = 5; Figure 7).21,79,82–84 All other meta-analytic findings, including incident opioid use, MME per day, number of prevalent users, and number of opioid prescriptions, are presented in the Figures S2 to S7. Forest plot of the percentage changes in level and trend for the prevalence of patients in the general population prescribed opioids following guideline implementation. APS-AAPM = American Pain Society-American Association of Pain Medicine; ANZCA = Australian and New Zealand College of Anaesthetists; BPS = British Pain Society; CDC = Centers for Disease Control and Prevention; CMQ = Collège des médecins du Québec; CPSBC = College of Physicians and Surgeons of British Columbia; DHA = Danish Health Authority; KPS = Korean Pain Society; LCL = Lower confidence limit; NOUGG = National Opioid Use Guideline Group; UCL = Upper confidence limit; WA = Washington state. Forest plot of the percentage changes in level and trend for the MME dose per person in the general population following guideline implementation. APS-AAPM = American Pain Society-American Association of Pain Medicine; BPS = British Pain Society; CDC = Centers for Disease Control and Prevention; GPS = German Pain Society; LCL = Lower confidence limit; NOUGG = National Opioid Use Guideline Group; UCL = Upper confidence limit; WA = Washington state. Forest plot of the percentage changes in level and trend for the average duration of opioid prescriptions in the general population following guideline implementation. APS-AAPM = American Pain Society-American Association of Pain Medicine; CDC = Centers for Disease Control and Prevention; GPS = German Pain Society; LCL = Lower confidence limit; UCL = Upper confidence limit. Qualitative synthesis Findings from the qualitative synthesis are presented in Table S6. Chronic noncancer pain The effect of the guidelines on prevalence remains unclear, with nine studies reporting an increasing trend52,53,85–91 and nine studies reporting a decreasing trend.23,49,91–97 Four studies reported no changes or post-guideline prevalence that varied depending on the specific guideline examined (e.g., APS-AAPM versus CDC).50,54,96,98 However, the guidelines’ effects on the incidence were clearer, with seven studies showing a decrease,87,90,99–103 two studies showing no trend changes,104,105 and two studies showing an increase in prescribing incidence trend.106,107 As for dosage, four studies described a decreasing trend,23,24,49,108 two studies described an increasing trend,56,109 and two studies described no change58,110 following implementation. Five studies found a decline in the prescription duration trend post-guideline implementation.23,49,58,90,106 Three studies showed an increasing trend for tapering and discontinuation of opioid treatment, including abrupt or rapid discontinuation.111–113 One study showed limited changes in opioid discontinuation among patients undergoing long-term opioid treatment,114 and one study showed a decreasing post-guideline trend for aggressive tapering.24 Chronic pain For studies focusing on chronic pain patients, 13 showed decreasing trends,115–127 6 showed increasing trends,128–133 and 4 showed no changes in the prevalence following guideline implementation.134–137 Two studies observed that the proportion of patients dispensed their first opioid prescription decreased after guideline implementation,122,138 with one study observing an increase in incident prescriptions.128 Related, three studies demonstrated decreasing post-guideline trends in dosage prescribed.55,126,139 One study demonstrated decreasing post-guideline trends in the duration of opioid prescriptions,55 and another post-guideline trend.129 Acute pain In the acute pain population, 10 studies (with 11 outcome measures) showed a decreasing trend,119,140–148 2 showed an increasing trend,52,149 and 3 showed no change97,98,150 in prevalence following guideline implementation. Two studies found there was a declining trend in opioid use incidence,151,152 and another study found no trend changes following guideline implementation.153 Meanwhile, 14 studies reported decreases in dosage trends following guideline rollouts,58,141,147,148,150,152,154–161 with one study showing an increase in total MME per prescription.141 With respect to prescription duration, 3 studies showed decreasing trends,58,154,160 whereas 2 described an increase in duration over time.141,162 General population For studies reporting prevalence outcome measures,22,51,59–63,65–74,77,163–213 38 outcome measures showed negative trends, 26 showed positive trends, 8 showed no change, and one showed differing prevalence by country and database. The guideline effect on incident opioid use remains less clear. Six studies reported a decrease,71,84,209,214–216 3 studies showed no change,61,185,217 one study showed an increase,218 and one study showed incident prescribing trends varied by database.219 For dosage, 18 studies reported decreasing trends,21,63,71,74,75,81,82,84,165,169,200,209,211,220–224 whereas 7 studies reported an increasing trend in dosage.57,77,79,80,210,225,226 Four studies showed trend changes in dosage that varied by guideline, analgesic type, or database.78,219,227,228 Of the 8 studies reporting duration of prescriptions dispensed, 2 observed an increasing trend post-guideline,71,79 4 observed a decreasing trend post-guideline,21,84,169,229 and 2 observed no changes.82,230 One study showed an increasing post-guideline trend in abrupt discontinuation but a decreasing post-guideline trend in tapered discontinuation.231 Cancer pain Four studies reported a trend decrease in the proportion of individuals receiving opioid prescriptions,49,87,232,233 with two studies reporting no changes188,234 and one study indicating an increase in the mean number of opioid prescriptions per person.49 Two studies reported no changes in the proportion of incident opioid use post-guideline implementation.87,235 For dosage, three studies showed a decreasing trend,24,49,232 and one showed no changes.110 For duration, two studies reported a decrease in trend following guidelines.49,232 Palliative pain One study showed an increasing post-guideline trend for opioid prescription claims among nononcologists,188 and another study showed a decreasing post-guideline trend for MME per person.24 Discussion This systematic review and meta-analysis consolidated evidence on the impact of opioid prescribing guidelines across jurisdictions and populations. Our findings reveal year-to-year declines in opioid prescription prevalence, incidence, and dosage following guideline implementation. More specifically, among patients with CNCP, the primary target group, the guidelines led to decreasing year-to-year trends in MME per person and prescription duration. Similarly, in the general population (i.e., guideline exempt), there were declining annual trends in the prevalence of opioid use and MME per person as well as MME per day and number of opioid prescriptions. The results from the qualitative synthesis demonstrate that the majority of studies across acute pain, chronic pain, and cancer pain populations reported decreasing trends in the prevalence, dosage, and duration of opioid prescriptions following guideline implementation. Opioid prescribing guidelines contributed to shifts in opioid prescribing within clinical practice, aligning with public health goals to mitigate access to opioids. This shift likely reflects heightened awareness of opioid risks and increased use of alternative pain management strategies, such as nonopioid analgesics (e.g., gabapentinoids, paracetamol) and multimodal approaches incorporating psychosocial and personalized care.92,236–238 However, these changes in opioid prescribing (e.g., dose reduction, lower prevalence, and lower incidence) may also reflect the unintended consequences of the guidelines.239,240 In the general population, significant decreases in opioid prevalence and prescriptions were observed, but this may have been due to abrupt tapering and discontinuation of opioid treatments along with patients being denied opioids.239 Additionally, declining trends in opioid dosage for patients with acute pain and prevalence among patients with cancer further suggest spillover effects on guideline-exempt populations. From a policy standpoint, these findings indicate that prescribing guidelines can meaningfully influence clinical practice, but they also underscore the need for greater nuance in guideline design and implementation. Future guidelines should balance opioid stewardship with clinical flexibility by incorporating safeguards against spillover effects in guideline-exempt populations and emphasizing patient-centered outcomes beyond prescribing volume alone. The findings from the meta-analysis warrant careful interpretation, because most of the included studies were conducted in the United States. Consequently, the effect estimates predominantly reflect the level and trend changes associated with the implementation of the CDC guidelines and the APS-AAPM guidelines.10,12 Importantly, pre-guideline opioid prescribing trends in the United States differed from those observed in other countries. For instance, in North America, a decline in opioid prescriptions began in the early 2010s, preceding the introduction of many guidelines in the mid-2010s.82,140,241,242 In contrast, countries such as Australia, the United Kingdom, and Germany experienced increasing trends in opioid dispensing and prescribing during the same period.56,77,243–245 As a result, the impact of guidelines on clinical practice in the United States may have been less pronounced, because the decline in opioid prescribing was already underway due to increased awareness of potential risks associated with opioids.61,82 However, by fostering a climate of caution, the guidelines likely reinforced or accelerated this downward trend, pushing many physicians away from opioid-based treatments. At the same time, the guidelines appear to have prompted a significant shift in opioid initiations, resulting in fewer new patients being prescribed opioids and reductions in both dosage and prescription duration. This shift reflects a broader change in clinical practice paradigms, moving away from reliance on prescription opioids.238 Variations in the observed effects may also reflect differences in implementation fidelity and contextual factors across jurisdictions. For instance, the APS-AAPM guideline10 was published before the onset of the opioid crisis in the United States and endorsed chronic opioid therapy as a potentially effective pain management strategy.10 In contrast, the CDC guideline,12 developed in response to the escalating crisis, adopted a more restrictive and cautious approach, specifying dosage and duration limits and recommending nonopioid treatments whenever possible.12 A similar evolution occurred in Canada: the NOUGG predated the national opioid crisis and emphasized risk management and the safe and effective use of opioids,14 whereas the College of Physicians and Surgeons of British Columbia (CPSBC) practice standard and Health Canada guideline emerged as crisis-driven responses and closely mirrored the CDC’s restrictive recommendations.9,11 Notably, the CPSBC practice standard was legally enforceable, demanding a higher level of implementation fidelity than earlier advisory guidelines. In North American contexts, therefore, guidelines from the 2009–2010 period reflected clinical frameworks for appropriate long-term opioid use, whereas those introduced from 2016 onward represented public health–driven interventions aimed at minimizing opioid exposure. These temporal, regulatory, and contextual differences along with variation in the degree to which guidelines were implemented likely contributed to the heterogeneity in observed effects across studies and populations. Limitations Our study has several limitations. First, our database search focused on prescribing practices and patterns, the primary target of guidelines, capturing opioid prescription volumes rather than opioid use, potentially missing eligible studies. With guidelines’ intended audience being prescribers, we assumed a reduction in prescription volume led to a reduction in opioid use. To mitigate this concern, we consulted a librarian to validate our search strategy in January 2023 for the initial review commissioned by Health Canada, repeated the search in May 2024, and updated it in January 2025 and October 2025. Second, some studies lacked clear mention of prescribing outcomes in their titles or abstracts, complicating eligibility assessment. Internal validation helped identify and reinclude originally excluded studies. Third, most studies focused on opioid prescribing in the general population, limiting insight into how CNCP pain guidelines affected target and exempt populations, separately and specifically. Additionally, the lack of studies on certain outcomes or pain types prevented comprehensive meta-analysis across all patient groups. To address this, we qualitatively synthesized guideline effects to guide future evaluations. Fourth, plot digitization for time series data extraction may have introduced measurement bias. To minimize this bias, all plot digitized figures were independently reviewed and validated by an additional reviewer. Fifth, some papers using aggregate-level data were not clear in their included age range. Assuming a very low percentage of pediatric patients with opioid prescriptions in studies covering the entire population, we treated these studies as adult population studies. Sixth, we acknowledge the influence of concurrent policy changes and co-interventions across jurisdictions. During the study periods, some jurisdictions implemented multiple, overlapping initiatives alongside the prescribing guidelines, such as restrictions on high-dose formulations, prescription monitoring programs, and public awareness campaigns.246–249 Admittedly, these co-interventions make it challenging to isolate the specific causal effect attributable solely to guideline implementation. However, most interventions were not concurrent, with interventions like narcotics monitoring systems predating guideline rollouts by years. Though some spillover effects of other interventions are likely captured, we consider this study an ecologically valid measure of the real-world changes driven by a multipronged public health response. Seventh, several outcomes, such as opioid use and prescription, were indirectly measured using administrative and prescription claims data, which may not fully capture actual medication consumption or patient-level clinical outcomes. However, these measures are widely used in pharmacoepidemiological and health services research studies and provide valid and reproducible population-level indicators of prescribing behavior. Eighth, we did not systematically evaluate the robustness of our estimates using alternative ITS models (e.g., Prais-Winsten, restricted maximum likelihood, or autoregressive integrated moving average), because the combination of non-Gaussian outcomes and very short series in some included studies (i.e., only two to three pre- or postintervention time points) limited the feasibility of such sensitivity analyses. We acknowledge that some residual sensitivity to modeling assumptions may persist. However, our ITS analyses with the Newey-West heteroscedasticity- and autocorrelation-consistent standard errors are expected to provide reasonably robust estimates in this context. Ninth, because the majority of studies included in the meta-analyses were conducted in the United States and Canada, these findings may not be fully generalizable to health care systems with different regulatory or cultural contexts. However, the concentration of research in these jurisdictions enabled robust pooled analyses of prescribing outcomes that may not have been feasible only with studies from elsewhere. The concentration of evidence in these settings likely reflects the disproportionate impact of the opioid crisis and the intensity of guideline implementation and evaluation efforts in North America. Finally, we excluded abstracts and articles in languages other than English or French due to limited resources. Though we were able to screen many abstracts of articles published in neither English nor French because these abstracts were often available in English, we excluded five studies at the full-text review stage. Given the small share of these studies among those excluded (5 out of 958), we believe that the risk of language bias in our study is likely minimal. Conclusion Our study provides key evidence that opioid prescribing guidelines have reduced prescriptions not only for patients with CNCP but also for those with other pain conditions. These findings suggest that the guidelines have been broadly interpreted and applied beyond their intended scope. However, changes in opioid prescribing practices likely reflect the combined influence of multiple, overlapping policy, regulatory, and educational initiatives introduced over the past decade, as well as evolving clinical attitudes and public awareness of opioid-related harms. Thus, though our results underscore the important role of guidelines, they should be interpreted within the broader policy and health care context that collectively shaped prescribing behaviors. Analysis of a breadth of outcome measures offers the most current evidence to enhance patient safety and improve targeted aspects of opioid prescribing practices. At the same time, further evaluation of the impact of guidelines on exempt populations and the development of tailored recommendations for these groups may be necessary to strengthen and sustain evidence-informed best practices in pain management. Supplementary Material Supplementary Materials_February 2026.docx Acknowledgment We thank Genevieve Gore (McGill University Liaison Librarian, Schulich Library of Physical Sciences, Life Sciences, and Engineering) for assisting with the search strategy. Funding Statement This project was supported by a Canadian Institutes of Health Research Project Grant (No. 461668). DP holds a Tier 2 Canada Research Chair in the Economics of Harm Reduction. Disclosure statement No potential conflict of interest was reported by the author(s). Data availability statement The data used to run meta-analyses are available upon request. The R code used in the current study is available in the GitHub repository (https://github.com/jihoon-jay). Supplementary Material Supplemental data for this article can be accessed online at https://doi.org/10.1080/24740527.2026.2650299 References Associated Data Supplementary Materials Supplementary Materials_February 2026.docx Data Availability Statement The data used to run meta-analyses are available upon request. The R code used in the current study is available in the GitHub repository (https://github.com/jihoon-jay).