Impact of chiropractic care on opioid use for noncancer spine pain: systematic review and meta-analysis

Supplemental Digital Content is Available in the Text.

Early access to chiropractic care may reduce reliance on opioids for patients with noncancer spine pain; however, rigorously designed randomized controlled trials are needed.

Abstract

Opioids are commonly prescribed for spine-related pain; however, emerging evidence suggests that access to chiropractic care may reduce reliance on opioids. We conducted a systematic review and meta-analysis to assess the impact of chiropractic care on new or continued prescription opioid use among adults with noncancer spine pain. We searched for eligible randomized controlled trials (RCTs) and observational studies in MEDLINE, Embase, AMED, CINAHL, Web of Science, and the Index to Chiropractic Literature up to March 20, 2025. Paired reviewers independently assessed risk-of-bias and extracted data. We performed random- and fixed-effects meta-analyses and used GRADE to assess the certainty of evidence. In total, 2 RCTs (838 participants) and 18 cohort studies (6,035,220 participants) were included in our analyses. We found very low certainty evidence that, compared with standard medical care alone, receipt of chiropractic care may reduce the odds of receiving prescription opioids by 64% (odds ratio [OR] = 0.36; 95% confidence interval [CI], 0.25–0.52; absolute risk reduction [ARR] 15%). However, we found a credible subgroup effect that earlier receipt of chiropractic services (within the first 30 days of presenting with spine-related pain) is associated with a greater decrease in the odds of receiving prescription opioids (OR = 0.33; 95% CI = 0.22–0.51; ARR = 15%) than later (≥30 days after presentation: OR = 0.73; 95% CI = 0.53–0.99; ARR = 8%; test of interaction, P < 0.001), but both with very low certainty evidence. Rigorously designed RCTs are needed to confirm these results.

1. Introduction

Opioid-related harms, including accidental or intentional overdose, addiction, and death, have risen over the past 25 years, particularly in Canada and the United States.^{3,8,29,43,56,65} In Canada, the most recent estimates indicate that there were, on average, 15 hospitalizations, 67 Emergency Department visits, 99 Emergency Medical Services responses (ie, paramedic or 9-1-1 calls), and 20 deaths, per day, due to opioid-related causes between January 1, 2024, and December 31, 2024.²⁹ In the United States, as of July 6, 2025, there were over 50,000 opioid-related deaths reported in the past 12 months.³ The Centers for Disease Control and Prevention estimated the annual cost (including direct and indirect expenses) of the opioid crisis at over $1 trillion USD in 2017, equivalent to 5% of the US gross domestic product.^43,53

Establishing the relative contribution of prescribed, diverted, and illicit opioids to the current opioid crisis in North America and elsewhere⁴³ is complex. However, a study of 2,910 opioid-related deaths in Ontario, Canada, found that, in 2016, a third of those who died had an active opioid prescription and more than 75% had been dispensed an opioid within 3 years of death.³⁵ Consequently, there is increasing interest in management strategies to reduce reliance on opioids for patients with acute or chronic spine pain.³⁰ Current clinical practice guidelines recommend optimizing nonopioid pharmacotherapy and nonpharmacologic treatments, such as exercise, education, cognitive behavioural therapy, acupuncture, soft-tissue massage, and spinal manipulation, rather than prescribing opioids as a first-line therapy for noncancer back or neck pain.^16,23,30 However, prescription opioid use among patients with noncancer spine pain remains high.^26,76,80 Several studies have reported that utilization of chiropractic services for spine-related pain may be effective in reducing opioid prescribing^{1,26,34,40,47,78,79} and long-term opioid use^{1,27,40,47,51,60}; however, the magnitude of these effects and overall quality of the evidence are uncertain.

A 2020 systematic review and meta-analysis of 6 uncontrolled studies¹⁹ found a large, inverse association between chiropractic management and receiving prescription opioids among patients with noncancer spine pain (pooled odds ratio [OR] = 0.36; 95% confidence interval [CI], 0.30–0.43). However, the literature search informing this systematic review was last conducted on April 18, 2018, and several studies investigating the effect of chiropractic care on new and existing prescription opioid use have since been published.^{1,26,27,34,40,47,51,61,78,79} Moreover, the pooled estimate did not account for potential confounding variables, assessments of risk of bias and heterogeneity were suboptimal,⁶⁴ and the certainty of evidence was not assessed.¹⁹

1.1. Objectives

We conducted a systematic review and meta-analysis to assess the impact of chiropractic care on initiation or continued use of prescription opioids among adults with noncancer spine pain. For the identified studies, we also conducted an exploratory analysis of secondary study outcomes that were co-occurring with hypothesized (or potential) effects of chiropractic care on prescription opioid use. We explored whether our results were influenced by the year the study was conducted, methodological quality, and earlier vs later chiropractic exposure.

2. Methods

We followed the PRISMA 2020 statement⁵⁹ (see Appendix A, supplemental digital content, http://links.lww.com/PR9/A364) and MOOSE guidelines⁶⁹ to report our findings, and registered our protocol²⁸ on PROSPERO (registration number: CRD42023432277).

2.1. Eligibility criteria

Our eligibility criteria^24,33 are listed in Table 1. We included randomized and nonrandomized (quasi-experimental) controlled trials and observational studies (including cohort and case-control studies) that reported an adjusted analysis exploring the association between receipt of chiropractic care for noncancer spine pain and prescription opioid use. We excluded case reports, case series, cross-sectional studies, pre/post designs (eg, time series, single cohort), protocols, letters, editorials, commentaries, books and book chapters, gray literature (eg, dissertations, conference abstracts, pre-print/non-peer-reviewed publications), qualitative studies, and secondary sources of evidence (eg, clinical practice guidelines or any type of review article).

Table 1.

Eligibility criteria.

Domain	Description
Participants/population	We included adults ($\ge$18 y of age) with noncancer back or neck pain (with or without radicular symptoms) of any duration*
Intervention/exposure	Our exposure of interest was receipt of chiropractic care, defined as care provided by a chiropractor, including, but not limited to, spinal manipulation, soft-tissue therapy, education, reassurance, and self-care advice (eg, icing, stretching, and strengthening exercises)
Comparator/control	The comparison was nonreceipt of chiropractic care (eg, usual medical care, physiotherapy)
Outcomes	Primary outcomes: Our primary outcomes were (1) prescription opioid receipt, and (2) continued prescription opioid use (measured as the number and/or dose of opioid prescriptions) Secondary outcomes: We extracted data on all other patient-important outcomes24 that were reported, including: (1) pain intensity, (2) physical functioning,† (3) emotional functioning,† (4) sleep quality, (5) patient satisfaction, and (6) adverse events‡

^*We excluded adults with spinal neoplasms or other contraindications to chiropractic treatment (ie, “red flag” diagnoses such as fractures, infections, inflammatory arthritis, or cauda equina syndrome).³³

^†Outcomes for physical and emotional functioning were only considered if they were reported over a minimum 4-wk follow-up period.

^‡We included all adverse events reported in eligible studies.

2.2. Information sources

We searched MEDLINE, Embase, AMED, CINAHL, Web of Science, and the Index to Chiropractic Literature without geographic or language restrictions from the inception of each database to March 20, 2025. Our search strategy was developed by an academic librarian (R.J.C.) and reviewed by a second librarian using the PRESS checklist⁵⁷ (see Appendix B, supplemental digital content, http://links.lww.com/PR9/A364). We hand-searched the bibliographies of eligible articles and contacted 2 content experts to identify any additional references.

2.3. Study selection

Pairs of reviewers (P.C.E., C.C., J.D.) independently screened titles and abstracts of identified citations and full texts of potentially eligible studies using online systematic review software (DistillerSR, Evidence Partners, Ottawa, Canada; https://www.distillersr.com/). Disagreements on eligibility were resolved by discussion or adjudication by a third reviewer (K.L.C. or B.C.C.). We assessed agreement for title/abstract and full-text screening using an adjusted kappa (k) statistic.⁴⁹

2.4. Data collection process

Pairs of reviewers (P.C.E., A.L.B., C.C., J.D.) independently extracted data from included studies using standardized, prepiloted data extraction forms. Extracted information included: (1) first author's name; (2) year of publication; (3) study design; (4) country where the study was conducted; (5) time period of data collection; (6) sample size; (7) participant demographics (ie, age, sex, primary pain complaint); (8) chiropractic care and control group information (ie, proportion of patients receiving chiropractic or usual medical care; type of usual medical care provided, such as primary or specialist care; number of days between the index visit date and initiation of chiropractic care); (9) details on opioid use (ie, proportion of sample prescribed opioids and, when available, total number and dose of opioid prescriptions); (10) all patient-important outcomes that were reported (eg, pain intensity, physical and emotional functioning, sleep quality, patient satisfaction, adverse events^36,54); (11) length of follow-up; and (12) source of funding. If a study reported outcomes at several time points, we used data from the longest follow-up in our analyses unless there was ≥20% missing data, in which case we used the next longest follow-up that had <20% missing data. We only included data from studies with the best adjusted model (ie, best model fit) or largest sample size in instances where 2 or more articles' study populations overlapped by ≥50%. Discrepancies between reviewers were resolved as previously described. We contacted study authors when necessary to request unpublished or missing data or to seek clarification regarding eligibility.

2.5. Risk of bias in individual studies

Pairs of reviewers (P.C.E., K.L.C., B.C.C.) independently assessed risk of bias of eligible randomized controlled trials (RCTs) using a risk-of-bias tool developed by the CLARITY group (https://www.distillersr.com/resources), according to the following domains: sequence generation; allocation concealment; blinding of patients, healthcare providers, data collectors, outcome assessors, and data analysts; infrequent missing data (>20% was considered high risk of bias); selective outcome reporting; and other sources of bias (eg, industry funding). To assess for selective outcome reporting, we identified all eligible studies that had been registered and then reviewed information provided on clinical trial registries (eg, clinicaltrials.gov) to compare studies' prespecified outcomes with their published results. When protocols were not available, we compared the methods and results in each trial publication. Response options for each item were dichotomized as “definitely or probably yes” (assigned as low risk of bias) and “definitely or probably no” (assigned as high risk of bias) (see Appendix C, supplemental digital content, http://links.lww.com/PR9/A364).

We also used criteria suggested by the CLARITY group to assess the risk of bias of observational studies, including: selection bias, assessment of exposure, temporality of exposure and outcome, control of confounding variables (with adjustment for age, sex, and severity or duration of noncancer spine pain, at a minimum, considered as an adequately adjusted model), assessment of prognostic factors, validity of outcome assessment(s), loss to follow-up (≥20% was considered high risk of bias), and assessment of co-interventions (including other pharmacologic and nonpharmacologic therapies) (see Appendix D, supplemental digital content, http://links.lww.com/PR9/A364). Disagreements between reviewers were resolved by consensus or adjudication by a third reviewer (P.C.E., K.L.C., or B.C.C). If a reviewer was an author on an included article, the study was reviewed by the other 2 members of the research team.

2.6. Data synthesis

We pooled all binary outcomes on opioid use (ie, prescribed opioid receipt, long-term opioid use) that were reported by more than 1 study using ORs and associated 95% CIs. When studies provided hazard ratios (HRs) or relative risks (RRs), we converted these to an OR using a baseline risk (ie, proportion of patients in the nonchiropractic care control group who experienced an event) before pooling.⁷⁵ We pooled the effect of chiropractic care on adverse events in RCTs using RRs and their associated 95% CIs. Continuous outcomes (ie, pain intensity and physical functioning) were pooled as weighted mean differences (WMDs) with associated 95% CIs after converting different instruments that reported on the same domain (eg, pain) into the most commonly reported scale among studies eligible for review.^46,71 We used change scores from baseline rather than end-of-study scores to account for interpatient variability. If the study authors did not report change scores, we calculated them using the baseline and end-of-study scores and the between-group end-of-study standard deviation (SD).⁴¹ If SDs were not reported directly, we estimated these from standard errors (SEs) or CIs.⁴¹ For all outcomes, we conducted separate analyses for RCTs and observational studies and prioritized adjusted over unadjusted associations if both sets of data were available in the same study.

We conducted all meta-analyses of 3 or more studies using random-effects models⁴¹ with the DerSimonian-Laird method,²¹ and fixed-effects models when pooling 2 studies.²² We performed a qualitative synthesis when data could not be pooled. To facilitate interpretation of outcomes amenable to meta-analysis, we calculated the absolute risk for each binary outcome using the median control group risk across studies. We used the mean control group risk when there were only 2 studies. For continuous measures, we modelled the risk difference (RD) of achieving at least the minimally important difference (MID) (ie, the smallest amount of improvement in a treatment outcome that patients recognize as important).^46,71 We used anchor-based MIDs in our analyses for pain intensity (1.5 cm on a 10-cm visual analogue scale [VAS]⁷⁴) and physical functioning (3 points on the 0–24 point Roland-Morris Disability Questionnaire [RMDQ] scale¹²).

Modelling assumptions for estimating the RD of achieving the MID assume that the SDs of outcome measurements are the same in both the treatment and control groups, and that change scores in both groups are normally distributed. To verify modelling assumptions, we compared SDs between treatment and control groups, and calculated the mean effect score ± 2 SDs in each treatment group for all trials that contributed to our analyses to identify any cases in which the distributions were substantially skewed.^41,71 All analyses were performed using Stata V.18 (StataCorp, College Station, TX), and comparisons were 2-tailed using a statistical significance threshold (α) of 5%.

2.7. Subgroup, meta-regression, and sensitivity analyses

Heterogeneity was examined using I² for all fixed-effects models and τ² for random-effects models,⁶⁴ as well as through visual inspection of forest plots.⁴⁴ We considered heterogeneity of a pooled estimate to be problematic if I² was ≥75% for pooled effects from RCTs,⁴¹ or if the range of estimates was beyond the pooled estimate ±2 τ for pooled effects from observational studies.⁶⁷ We explored sources of heterogeneity with 2 prespecified subgroup hypotheses, assuming larger associations with: (1) higher vs lower risk of bias, evaluated on a criterion-by-criterion basis, and (2) early vs later chiropractic exposure. We defined “early” chiropractic exposure as receipt of chiropractic services within the first 30 days after an index visit for acute or chronic noncancer spine pain.^{1,26,40,78,79}

We used meta-regression to explore the impact of time period on the association between chiropractic care and prescription opioid use,²⁸ assuming larger associations with studies conducted in earlier vs later calendar years—a proxy for increased pressure on physicians to reduce opioid prescribing.^16,23 In instances where study data were collected over a range of years, we used the median year of data collection. Tests for interaction were performed to establish whether subgroups differed significantly from one another, and we assessed the credibility of all significant subgroup effects (test for interaction P < 0.05) using modified ICEMAN criteria.⁶⁶

We performed sensitivity analyses to test the robustness of our results by excluding studies in which we derived measures of association from HRs or RRs, or effect estimates from converted change scores.

2.8. Certainty of evidence

We evaluated the certainty of evidence for all measures of association using the GRADE approach.^37,39,44 With this approach, RCTs begin at high certainty evidence, and observational studies reporting on treatment effects begin at low certainty evidence, and both can be rated down based on risk of bias, inconsistency, indirectness, imprecision, and publication bias. We rated down for imprecision if the 95% CI included the null effect, but did not rate down the same effect estimate twice for both inconsistency and imprecision when inconsistency was the cause of imprecision.⁸¹ The certainty of evidence for observational data can also be rated up 1 or 2 levels because of a strong association, a dose-response gradient, or when all plausible confounders or other biases increase our confidence in the estimated effect.³⁸ When there were ≥10 studies available for meta-analysis,⁴¹ publication bias was assessed for each outcome by visual assessment of funnel plots for asymmetry and calculation of Egger test.²⁵

3. Results

Of 952 unique citations, 2 RCTs^11,34 and 18 cohort studies^{1,4,6,9,10,26,27,31,40,42,47,50,52,62,73,77–79} were included in our analyses. We excluded 1 eligible cohort study⁷² as their study population overlapped with a larger cohort study⁷³ exploring the association between receipt of chiropractic care and prescription opioid use (Fig. 1). Agreement between reviewers at the title/abstract (k = 0.65) and full-text screening (k = 0.70) stages was substantial. A list of all excluded full-text articles is provided in supplemental digital content (see Appendix E, http://links.lww.com/PR9/A364). We contacted 3 authors, all of whom responded, 1 for clarification on eligibility⁶¹ and 2 who provided additional data for analysis.^4,34

Figure 1.

PRISMA flow diagram. ^a Twenty studies were included in our primary analysis; among these, 1 article reported 3 separate cohorts.^{77 b} One study⁷² included an overlapping population with a larger study⁷³ included in our primary analysis.

3.1. Study characteristics

Characteristics of the 2 RCTs and 18 cohort studies are provided in Table 2 and supplemental digital content (see Appendix F, http://links.lww.com/PR9/A364). The median of the mean age of participants across all 20 included studies was 47 years (interquartile range [IQR], 42–51), and 57% of participants were women. Most studies (1 RCT³⁴ and 16 cohort studies^{1,4,6,9,10,31,40,42,47,50,52,62,73,77–79}) were conducted in the United States, and the remainder (1 RCT¹¹ and 2 cohort studies^26,27) were conducted in Canada. Among the 18 cohort studies that were included, 1 used prospective data collection methods³¹ and the other 17 were retrospective reviews of administrative data.^{1,4,6,9,10,26,27,40,42,47,50,52,62,73,77–79} Follow-up in the 2 RCTs ranged from 3 to 4 months, and from 7 days to 12 months across cohort studies (Table 2 and see Appendix F, supplemental digital content, http://links.lww.com/PR9/A364).

Table 2.

Characteristics of the 20 included studies.

Study characteristic	Median (IQR)*†
	RCTs (n = 2)	Cohort studies (n = 18)
Year of data collection, median	2011 (2009–2012)	2014 (2012–2017)
Length of follow-up, days	105 (range, 90–120)	365 (range, 7–365)
No. of participants	419 (254–585)	35,368 (3,356–187,683)
Age, mean, year	34 (33–36)	48 (43–51)‡
Female, %	42 (33–51)	57 (56–61)
% of sample that received chiropractic care	51 (50–51)	24 (17–40)
% of sample that received comparison intervention§	50 (49–50)	52 (47–71)
Cumulative % of sample prescribed opioids during follow-up‖	42 (37–46)	40 (22–54)¶
Type of spine pain represented, no. of studies (no. of participants)
Low back pain	1 (88)	9 (4,220,612)
Mixed back and neck pain	—	5 (639,963)
Neck pain	—	2 (429,668)
Mixed low back and radicular pain	1 (750)	1 (478,981)
Radicular low back pain	—	1 (744,942)

IQR, interquartile range.

^*Unless otherwise indicated.

^†The mean was used when there were only 2 studies.

^‡In 13 studies.^{1,4,6,9,10,26,31,42,47,50,52,62,73}

^§Comparison interventions consisted of usual medical care/primary care^{6,9–11,26,27,34,40,42,47,50,52,73,78,79} or nonreceipt of chiropractic care.^1,4,31,62,77

^‖Measured at the longest point of follow-up.

^¶In 16 studies.^{1,6,9,10,26,27,31,40,42,47,52,62,73,77–79}.

3.2. Risk of bias in studies

Both included RCTs were at low risk of bias for allocation sequence generation and concealment, blinding of data collectors and outcome assessors, loss to follow-up, selective outcome reporting, and other sources of bias; however, 1 RCT¹¹ did not blind data analysts, and neither RCT blinded patients or healthcare providers (see Appendix G, supplemental digital content, http://links.lww.com/PR9/A364). All 18 cohort studies were rated at low risk of bias for cohort selection, assessment of prognostic factors, and assessment of outcome (see Appendix H, supplemental digital content, http://links.lww.com/PR9/A364). Most (94%) were at low risk of bias for assessment of exposure and adequacy of follow-up, 9 (50%) clearly reported that the outcome of interest was absent at baseline, and 3 (17%) were at low risk of bias for assessment of co-interventions. Only 1 cohort study (5%) reported an adjusted regression model adequately controlling for the minimum set of identified potential confounders (see Appendix H, supplemental digital content, http://links.lww.com/PR9/A364).

3.3. Primary outcomes for receipt of chiropractic care

3.3.1. Initiation of prescription opioids

We found very low certainty evidence from 2 RCTs involving 838 participants^11,34 that, compared with usual medical care alone, receipt of chiropractic care may result in a 34% lower odds of receiving prescription opioids for noncancer spine pain (OR = 0.66; 95% CI = 0.50–0.86). The absolute risk reduction (ARR) among chiropractic recipients was 10% less initiating prescription opioids (95% CI = 4%–17%) compared with nonrecipients (Fig. 2, Table 3, and see Appendix I, supplemental digital content, http://links.lww.com/PR9/A364). We also found very low certainty evidence from 18 cohort studies (involving 3,669,280 participants)^{1,4,6,9,10,26,40,42,47,50,52,62,73,77,78a-c,79} that, compared with usual medical care alone, receipt of chiropractic care may be associated with a 64% lower odds of initiating prescription opioids for noncancer spine pain (OR = 0.36; 95% CI = 0.25–0.52), with an ARR of 15% (95% CI = 11%–18%) (see Appendix J, supplemental digital content, http://links.lww.com/PR9/A364). We found a credible subgroup effect in 14 cohort studies^{6,10,26,40,42,47,52,73,78a-c,79} for timing of chiropractic exposure, with significantly lower odds of receiving prescription opioids among participants who received chiropractic services within 30 days (OR = 0.33; 95% CI = 0.22–0.51; ARR = 15%; 95% CI = 10%–18%) vs later (OR = 0.73; 95% CI = 0.53–0.99; ARR = 8%; 95% CI = 0.3%–15%) in their complaint (test of interaction, P < 0.001; moderate credibility for subgroup effect) (Fig. 3 and see Appendix K, supplemental digital content, http://links.lww.com/PR9/A364).

Figure 2.

Effect of receiving chiropractic care on initiation of prescription opioids for noncancer spine pain in 2 randomized controlled trials. CI, confidence interval; OR, odds ratio.

Table 3.

GRADE evidence profile for the impact of chiropractic care on initiation of prescription opioids and other patient-important outcomes for noncancer spine pain.

Study characteristics	Quality assessment					Summary of findings			Overall certainty of evidence
No. of studies (participants)	Risk of bias	Inconsistency	Indirectness	Imprecision	Publication bias	Treatment association (95% CI)	Absolute risk
							Baseline risk	Risk difference (95% CI)
Impact of chiropractic care on prescription opioid receipt* (yes vs no)
2 RCTs (838)	Serious risk of bias†	Serious inconsistency (I2 = 98%)‡	Serious indirectness§	No serious imprecision	NA (only 2 studies)	OR 0.66 (0.50–0.86)	65/100‖	10 fewer per 100 (−4 to −17)	Very low
Early DC care: 14 cohort studies¶ (3,487,063)	Serious risk of bias#	No serious inconsistency (τ2 = 0.65)**	No serious indirectness	No serious imprecision	Undetected; Egger test, P = 0.85	OR 0.33 (0.22–0.51)	24/100‖	15 fewer per 100 (−10 to −18)	Very low††
Later DC care: 5 cohort studies‡‡ (77,548)	Serious risk of bias#	No serious inconsistency (τ2 = 0.09)	No serious indirectness	No serious imprecision	NA (only 5 studies)	OR 0.73 (0.53–0.99)	51/100‖	8 fewer per 100 (−0.3 to −15)	Very low
Impact of chiropractic care on long-term opioid use§§ (yes vs no)
6 cohort studies (2,597,028)	Serious risk of bias#	No serious inconsistency (τ2 = 0.45)**	No serious indirectness	No serious imprecision	NA (only 6 studies)	OR 0.27 (0.15–0.47)	4/100‖	3 fewer per 100 (−2 to −3.4)	Very low††
Impact of chiropractic care on pain intensity (10-cm VAS for pain; lower is better; MID = 1.5 cm)
2 RCTs (838)	Serious risk of bias†	No serious inconsistency (I2 = 56%)	No serious indirectness	No serious imprecision	NA (only 2 studies)	WMD –0.64 (−1.01 to −0.28)	20/100‖‖	8 more per 100 (3–14)	Moderate
Impact of chiropractic care on physical functioning (0–24 point RMDQ scale; lower is better; MID = 3 points)
2 RCTs (838)	Serious risk of bias†	No serious inconsistency (I2 = 0%)	No serious indirectness	No serious imprecision	NA (only 2 studies)	WMD –2.03 (−3.15 to −0.91)	36/100‖‖	9 more per 100 (3–15)	Moderate
Impact of chiropractic care on nonserious adverse events (yes vs no)¶¶
2 RCTs (838)	Serious risk of bias†	No serious inconsistency (I2 = 0%)	Serious indirectness##	No serious imprecision	NA (only 2 studies)	RR 1.97 (1.17–3.34)	5/100‖	5 more per 100 (1–8)	Low
Impact of chiropractic care on serious opioid-related adverse drug events (yes vs no)
1 cohort study (744,942)	Serious risk of bias#	NA (only 1 study)	No serious indirectness	No serious imprecision	NA (only 1 study)	RR 0.29 (0.25–0.32)	0.3/100	0.2 fewer per 100 (−0.2 to −0.2)	Very low††

^*We found a significant subgroup effect in 14 cohort studies.^{6,10,26,40,42,47,52,73,78a-c,79} between early vs later receipt of chiropractic care (moderate credibility; interaction, P < 0.001).

^†Outcomes were subjective (ie, patient-reported) and may have been influenced by the lack of blinding of patients and healthcare providers.

^‡One study suggests a moderate reduction in the risk of initiating opioids with chiropractic care, and 1 study suggests that no patient receiving chiropractic care will be prescribed opioids.

^§We rated down for indirectness because the authors in 1 trial³⁴ did not stratify by opioid-naive vs opioid-using or opioid vs nonopioid analgesics (including over-the-counter) in their results.

^‖Based on the median control group risk across studies. The mean control group risk was used when there were only 2 studies.

^¶Association in 14 cohort studies between prescription opioid receipt and “early” chiropractic exposure, defined as receipt of chiropractic services within the first 30 d after an index visit for an acute or chronic noncancer spine pain diagnosis.

^#Most included cohort studies (94%) were at high risk of bias due to insufficient control of confounding. We did not rate down an additional level for risk of bias because our subgroup analyses found no credible subgroup effects between the pooled estimate and other risk-of-bias components.

^**We did not rate down for inconsistency because all studies suggest benefit with chiropractic care, and uncertainty appears to be the magnitude of association rather than whether there is an association.

^††The measure of association showed large effects, but we elected not to rate up our certainty in the evidence as the studies contributing to the pooled estimate were at high risk of bias.

^‡‡Association in 5 cohort studies between prescription opioid receipt and “later” chiropractic exposure, defined as receipt of chiropractic services >30 d after an index visit for an acute or chronic noncancer spine pain diagnosis.

^§§Defined as receiving at least 5 opioid prescriptions, or the equivalent of $\ge$144 days' supply of opioids, over a 12-month period.

^‖‖Based on the mean control group probability of achieving at or above the MID across studies.

^¶¶Adverse events attributed to chiropractic care in 1 trial³⁴ included muscle soreness or stiffness (n = 37), or transient paresthesia (n = 1). Adverse events reported for usual medical care were related to prescribed medications (n = 3), epidural injections (n = 4), or muscle or joint stiffness (n = 12) attributed to physical therapy or self-care recommendations.³⁴ The second trial¹¹ reported that there were no adverse events to chiropractic treatment.

^##We rated down for indirectness because the authors in 1 trial¹¹ did not report adverse events for the control group.

CI, confidence interval; DC, doctor of chiropractic; GRADE, grading of recommendations assessment, development, and evaluation; MID, minimally important difference; NA, not applicable; OR, odds ratio; RCT, randomized controlled trial; RMDQ, Roland-Morris Disability Questionnaire; RR, relative risk; VAS, visual analogue scale; WMD, weighted mean difference.

Figure 3.

Subgroup analysis of 14 cohort studies reporting early vs later chiropractic exposure and odds of initiating prescription opioids for noncancer spine pain (moderate credibility⁶⁶; test of interaction, P < 0.001). CI, confidence interval; OR, odds ratio. Whedon, 2020a is for 1,394 patients with early receipt of chiropractic services in Connecticut; Whedon, 2020b is for 77,611 and 71,373 patients with early and later receipt of chiropractic services in Massachusetts; Whedon, 2020c is for 33,338 and 4,122 patients with early and later receipt of chiropractic services in New Hampshire; Whedon, 2022* is for 937 patients with receipt of chiropractic services 31 to 90 days after the index visit; Whedon, 2022** is for 258 patients with receipt of chiropractic services 91 to 120 days after the index visit.

Very low certainty evidence from 6 cohort studies involving 2,597,028 participants^{1,6,31,40,47,52} suggested that, compared with usual medical care alone, receipt of chiropractic care may be associated with a 73% lower odds of initiating long-term opioid therapy for noncancer spine pain (OR = 0.27; 95% CI = 0.15–0.47). The ARR among chiropractic recipients vs nonrecipients was 3% less initiating long-term opioid use (95% CI = 2%–3.4%) (Fig. 4, Table 3, and see Appendix I, supplemental digital content, http://links.lww.com/PR9/A364).

Figure 4.

Association between receiving chiropractic care and initiation of long-term opioid use for noncancer spine pain in 6 cohort studies. CI, confidence interval; OR, odds ratio.

3.3.2. Continued prescription opioid use

We identified very low certainty evidence from 1 cohort study involving 210 participants²⁷ that investigated the association between initiating chiropractic care and continued prescription opioid use among adults receiving opioid therapy for noncancer spine pain (see Appendix F and I, supplemental digital content, http://links.lww.com/PR9/A364). Compared with receiving usual medical care alone, patients who initiated chiropractic services may receive between 34% and 73% fewer opioid prescriptions (ie, opioid fills: incidence rate ratio [IRR] = 0.66; 95% CI = 0.52–0.83; opioid refills: IRR = 0.27; 95% CI = 0.17–0.42), and may be 78% less likely to receive a higher (ie, >50 mg morphine equivalents daily) opioid dose (OR = 0.22; 95% CI = 0.08–0.62), over 12-month follow-up.²⁷ Patients who initiated chiropractic services may also be more than 3 times as likely to discontinue using opioids by 12 months compared with nonrecipients (29/49 vs 50/161, respectively; OR = 3.22; 95% CI = 1.66–6.23).

3.4. Secondary outcomes for receipt of chiropractic care

3.4.1. Pain intensity

Moderate certainty evidence from 2 RCTs (838 participants)^11,34 showed that, compared with usual medical care alone, receipt of chiropractic care for noncancer spine pain probably results in a small decrease in pain intensity (WMD = −0.64 cm on a 10-cm VAS; 95% CI = −1.01 to −0.28 cm), which equates to 8% (95% CI = 3%–14%) more patients experiencing pain relief at or above the MID (modelled RD for achieving at least the MID of 1.5 cm on a 10-cm VAS,⁷⁴ Fig. 5 and Table 3).

Figure 5.

Effect of receiving chiropractic care for noncancer spine pain on change in pain intensity in 2 randomized controlled trials. The red dashed vertical line represents the minimally important difference of 1.5 cm for the 10-cm visual analogue scale for pain. The solid red vertical line represents the overall pooled measure of effect. CI, confidence interval; SD, standard deviation; WMD, weighted mean difference.

3.4.2. Physical functioning

Moderate certainty evidence from 2 RCTs (838 participants)^11,34 showed that, compared with usual medical care alone, receipt of chiropractic care for noncancer spine pain probably results in improved physical functioning (WMD = −2.03 points on the 0–24 point RMDQ scale; 95% CI = −3.15 to −0.91 points), which equates to 9% (95% CI = 3%–15%) more patients experiencing an improvement in physical functioning at or above the MID (modelled RD for achieving at least the MID of 3 points on the 0–24 point RMDQ scale,¹² Fig. 6 and Table 3).

Figure 6.

Effect of receiving chiropractic care for noncancer spine pain on change in physical functioning in 2 randomized controlled trials. The red dashed vertical line represents the minimally important difference of 3 points for the 0 to 24 point Roland-Morris Disability Questionnaire scale. The solid red vertical line represents the overall pooled measure of effect. CI, confidence interval; SD, standard deviation; WMD, weighted mean difference.

3.4.3. Patient satisfaction

Moderate certainty evidence from 1 RCT (750 participants)³⁴ showed that, compared with usual medical care alone, receipt of chiropractic care for noncancer spine pain probably results in clinically important greater patient satisfaction⁶⁸ (mean difference favouring chiropractic care = 2.5 points on a 0–10 point satisfaction scale; 95% CI = 2.1–2.8 points).

3.4.4. Adverse events

Low certainty evidence from 2 RCTs (838 participants)^11,34 suggested that, compared with usual medical care alone, receipt of chiropractic care for noncancer spine pain may result in a small increase in the proportion of patients experiencing nonserious adverse events (eg, transient stiffness or muscle soreness) (RR = 1.97; 95% CI = 1.17–3.34; absolute risk increase [ARI] 5%; 95% CI = 1%–8%) (Fig. 7 and Table 3).

Figure 7.

Effect of receiving chiropractic care for noncancer spine pain on nonserious adverse events in 2 randomized controlled trials. CI, confidence interval; RR, relative risk.

Very low certainty evidence from 1 cohort study (involving 744,942 participants)⁷³ suggested a 71% reduced incidence of serious opioid-related adverse drug events (ie, opioid-related poisoning, overdose, or death) over 1-year follow-up among adults who initially received chiropractic services compared with matched controls who initially received standard medical care (no. of serious opioid-related adverse drug events: 335/372,471 vs 1,117/372,471, respectively; RR = 0.29; 95% CI = 0.25–0.32; ARR = 0.2%; 95% CI = 0.2%–0.2%) (Table 3).

3.5. Additional analyses

Both trials^11,34 that were used for modelling the RD of achieving the MID for pain intensity and physical functioning reported mean effect scores with an associated SD or SE, or MD and 95% CI, suggesting that trial authors concluded their data met normal distribution assumptions. When we calculated the mean effect scores ± 2 SDs in each treatment group for all trials that contributed to our analyses, we found no case in which the results exceeded the range of the study's pain intensity or physical functioning instruments, providing support that distributions were not substantially skewed (see Appendix L, supplemental digital content, http://links.lww.com/PR9/A364). SDs between the treatment and control groups for both outcome measures also proved to be very similar (Figs. 5 and 6, and Appendix L, supplemental digital content, http://links.lww.com/PR9/A364).

Meta-regression showed no significant relationship between the time period of data collection and the association of chiropractic care on initiating prescription opioids for noncancer spine pain (P = 0.99; Appendix M, supplemental digital content, http://links.lww.com/PR9/A364). Aside from the timing of receipt of chiropractic care, we found no other credible subgroup analyses (see Appendices N and O, supplemental digital content, http://links.lww.com/PR9/A364) or evidence of publication bias among outcomes reported by at least 10 studies (Table 3, see Appendices P and Q, supplemental digital content, http://links.lww.com/PR9/A364). Our findings were robust to sensitivity analyses; however, the effect of chiropractic care on pain intensity became nonsignificant when the larger of the 2 RCTs in which change scores were calculated from baseline and end-of-study scores was excluded (see Appendix R, supplemental digital content, http://links.lww.com/PR9/A364).

4. Discussion

4.1. Summary of main findings

We found very low certainty evidence that, compared with standard medical care alone, adults who receive chiropractic services for noncancer spine pain may be between 34% and 64% less likely to be prescribed opioids compared with nonrecipients and may be 73% less likely to initiate long-term opioid therapy. Furthermore, the impact of chiropractic care on reducing the odds of initiating prescription opioids for spine-related pain may be larger if access is provided within the first 30 days of presentation for care. We found moderate certainty evidence that, compared with standard medical care alone, receipt of chiropractic services for noncancer spine pain probably increases the proportion of patients experiencing important improvements in pain intensity (RD 8%) and physical functioning (RD 9%), as well as patient satisfaction. Low certainty evidence suggests that receipt of chiropractic care may double the risk of nonserious adverse events (eg, transient stiffness or muscle soreness) compared with not receiving chiropractic care. Very low certainty evidence suggests that access to chiropractic services may decrease the risk of serious opioid-related adverse drug events (ie, opioid-related poisoning, overdose, or death) by 71%.

4.2. Strengths and limitations

Strengths of our review include a comprehensive search for eligible studies in any language that identified 2 RCTs and 14 cohort studies not included in the most recent prior review.¹⁹ We excluded 2 observational studies that were included in this review¹⁹ because these studies did not meet our eligibility criteria. One was cross-sectional and the other did not provide an adjusted analysis between chiropractic receipt and opioid use (see Appendix E, supplemental digital content, http://links.lww.com/PR9/A364). We converted all pooled associations and mean effects in our current review to absolute risks to facilitate interpretation, used predefined subgroup analyses to explore sources of heterogeneity, and assessed the credibility of all potential subgroup effects. We conducted sensitivity analyses to confirm the robustness of our findings and used the GRADE approach to appraise the certainty of evidence.

Our review has limitations. First, the evidence for our primary outcomes was only of very low certainty, which limits the strength of inferences from our results. Second, although pain severity, symptom duration, chiropractic visit frequency, and type of opioid prescriber(s) (eg, general practitioner vs specialist/Emergency Department physician) may influence treatment effects,²⁸ there was either insufficient data or lack of variability among included studies to explore these issues. Third, most of our data were from observational studies, and patients accessing chiropractic services may be more resistant to using opioids than those not receiving chiropractic care.^26,27 Fourth, all studies eligible for our review enrolled patients presenting for care in North America, and the generalizability of our findings to other jurisdictions is unclear. Finally, we only included studies with outcome data for opioid use. As such, we did not capture all RCTs and observational studies investigating the impact of chiropractic care on other patient-important outcomes, including pain intensity, physical functioning, patient satisfaction, and adverse events.

4.3. Comparison with relevant literature

Our findings are consistent with a 2020 systematic review of 6 observational studies that found adults with noncancer spine pain who received chiropractic services were 64% less likely than nonchiropractic users to be prescribed opioids.¹⁹ This review reported large heterogeneity associated with their pooled effect (I² = 93%) that they were unable to explain; however, we found a credible subgroup effect that largely explained between-study variability. Specifically, we found that receipt of chiropractic care within 30 days of presenting with noncancer spine pain was associated with much larger effects vs receiving chiropractic care later. Furthermore, the prior review only explored the impact of chiropractic care on initiation of prescription opioids, whereas we captured all patient-important outcomes, including pain intensity, physical functioning, patient satisfaction, long-term opioid use, and adverse events. Our findings on the impact of chiropractic care on these outcomes were consistent with previous research.^{14,20,45,51,54,60,61,63}

Only 2^11,34 studies included in our review identified the specific treatments that patients received during chiropractic care (eg, spinal manipulation, soft-tissue therapy, exercise, education, reassurance, self-care advice), and the association with reduced opioid prescribing may be similar to care by other nonpharmacological healthcare providers (eg, physiotherapists).^1,15,19,62 If so, then patient preferences may be an important consideration. Patients are more willing to engage in treatments that match their preferences,⁴⁸ and show improved results when receiving preferred care.⁷ Furthermore, with respect to the use of nonpharmacologic care to help reduce opioid use among patients with noncancer spine pain, the results may differ whether augmented care is provided in isolation or as part of a coordinated effort between a prescriber, a patient interested in opioid tapering, and a nonpharmacological healthcare provider.⁷⁰ Several studies have shown that involuntary opioid tapering for chronic musculoskeletal pain is associated with net harms.^2,55,58

We anticipated that recent policy changes aimed at reducing opioid prescribing^5,13,16,23 would attenuate the relationship between chiropractic care and opioid use among studies in our review that were conducted in later vs earlier calendar years; however, we found no effect modification for time period with meta-regression. This is consistent with 2 previous cohort studies involving participant data from an Ontario community health centre between January 1, 2014, and December 31, 2020,^26,27 that found receipt of chiropractic care was associated with less prescription opioid use even after controlling for calendar year.^26,27 We also found that the results did not differ in our review between studies that did and did not control for co-interventions (eg, nonopioid analgesics, other pharmacotherapies, nonpharmacologic treatments). These findings suggest that the inverse relationship between receipt of chiropractic services and initiating^{1,4,6,9–11,26,31,34,40,42,47,50,52,62,73,77–79} or continuing^27,34 prescription opioids across studies in our review was independent of confounding by time or use of co-interventions.

There are several reasons why the utilization of chiropractic services might lead to reduced opioid use in adults with spine-related pain. First, chiropractic care, including spinal manipulation, has been found effective for some patients with back or neck pain.^{14,20,32,45,63} Patients who obtain pain relief from chiropractic treatment might therefore be less likely to require prescription opioids or, in collaboration with their prescribing physician, choose to taper opioid prescriptions. Physicians might also delay or prescribe fewer opioid medications or choose lower dosages if they can refer patients to chiropractors as a first-line treatment for pain management. This notion was supported in our review, where the impact of chiropractic care on initiating prescription opioids was found to be most pronounced among studies in which patients saw a chiropractor within the first 30 days of treatment. Furthermore, we found preliminary evidence to suggest that initiating chiropractic services among patients already receiving opioid therapy for noncancer spine pain may result in lower rates of opioid prescriptions (ie, fills and subsequent refills), and reduced odds of being prescribed a higher opioid dose.²⁷

4.4. Implications and future research

Our findings suggest that early receipt of chiropractic services for spine-related pain is associated with reduced odds of opioid prescribing and long-term opioid use. Among studies that met our inclusion criteria, chiropractic services were also associated with improvement in pain intensity, physical functioning, and patient satisfaction. Findings from our review further suggest that early receipt of chiropractic services for spine-related pain may increase the incidence of nonserious adverse events but reduce the risk of serious opioid-related harms. As such, our study represents a timely contribution to the important broader question of how to tackle the ongoing opioid crisis in North America and beyond⁴³ and the potential role that guideline-concordant nonpharmacologic treatments for pain, such as chiropractic care, can play in this process.^16,30,33 However, because the current evidence informing the impact of chiropractic care on new or continued prescription opioid use for noncancer spine pain is only of very low certainty, rigorously designed RCTs are needed to confirm these findings. Future studies should report all recommended patient-important outcomes for spine-related pain (eg, pain intensity, physical and emotional functioning, sleep quality, patient satisfaction, adverse events).^24,36,54 Studies should also incorporate objective outcome measures for prescription opioid use (eg, electronic medical records, pharmacy claims). Both trials in our review^11,34 captured patient-reported opioid use, which is susceptible to respondent and recall bias.^17,18

5. Conclusion

Our systematic review found very low certainty evidence that receipt of chiropractic care may be associated with lower odds of receiving prescription opioids or initiating long-term opioid use among adults with noncancer spine pain, particularly when chiropractic care is provided earlier vs later. Low certainty evidence shows that receipt of chiropractic care may increase the risk of nonserious adverse events, such as transient stiffness or muscle soreness; however, very low certainty evidence suggests the likelihood of serious opioid-related harms may be reduced. Rigorously designed RCTs are needed to confirm these results.

Disclosures

P.C.E. was supported by a postdoctoral fellowship from the Michael G. DeGroote Institute for Pain Research and Care (IPRC) at McMaster University. P.C.E. is also supported by grants from the Canadian Institutes of Health Research (CIHR), the Michael G. DeGroote IPRC, and the Canadian Chiropractic Research Foundation for postdoctoral research outside of the submitted work. J.W.B. is supported, in part, by a CIHR Canada Research Chair in the prevention and management of chronic pain. The remaining authors have no conflict of interest to declare.

Supplemental digital content

Supplemental digital content associated with this article can be found online at http://links.lww.com/PR9/A364.