Prevalence of opioid use in adults with spinal cord injury: A systematic review and meta-analysis

Abstract

Objective

To determine the prevalence, reported harms and factors associated with opioid use among adults with spinal cord injury (SCI) living in the community.

Study design

Systematic review and meta-analysis.

Methods

Comprehensive literature searches were conducted in PubMed (MEDLINE), EMBASE, CINAHL, Web of Science and Scopus for articles published between 2000 and 2023. Risk of bias was assessed using a prevalence-specific tool. Random-effects meta-analyses were conducted to pool prevalence data for any context of opioids. Sensitivity and subgroup analyses were also performed. The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines were followed, and the study protocol was registered via Prospero (CRD42022350768).

Results

Of the 4969 potential studies, 38 were included in the review. Fifty-three percent of studies had a low risk of bias, with a high risk of bias in 5% of studies. The pooled prevalence for the 38 studies included in the meta-analysis (total cohort size of 50,473) across any opioid context was 39% (95% confidence interval [CI], 32–47). High heterogeneity was evident, with a prediction interval twice as wide as the 95% CI (prediction interval, 7–84%). Mean or median opioid dose was unreported in 95% of studies. Opioid dose and factors related to opioids were also rarely explored in the SCI populations.

Conclusions

Results should be interpreted with caution based on the high heterogeneity and imprecise pooled prevalence of opioids. Contextual details including pain, cohort-specific injury characteristics and opioid dosage were inconsistently reported, indicating a clear need for additional studies in a population at greater risk of experiencing opioid-related adverse effects.

Introduction

Chronic and severe presentations of pain following a spinal cord injury (SCI) are common (1). Pain often has a large impact on an individual’s capacity to participate in daily activities, contributing to poorer quality of life and higher rates of depression (2, 3). Similar to other forms of chronic non-cancer pain, SCI-related pain has broader societal and economic burden associated with reduced participation in work and pain-related hospital admissions (4, 5). Consequently, pain in people with SCI is regularly considered a clinical management priority, and opioids are one medication class often prescribed. The large escalations in the global prescribing and dispensing of opioids and related harms including respiratory depression, potential for addiction, drug-related misuse and hospitalizations; and death have been well-documented among general populations (6, 7). However, far less is known in specific populations, including people with SCI. This is concerning, as many factors following a SCI can exacerbate opioid-related risks.

Secondary complications following a SCI are typical, often contributing to reduced respiratory functioning and bowel problems (8), the first of which can be exacerbated by the mechanistic action of opioids; and the latter being a common side-effect of opioids. Furthermore, many people with SCI are taking multiple other medications (e.g. anticonvulsants, antidepressants) concurrently which increases the potential of drug–drug interactions (9). Further risks in people with SCI relate to altered physiological functions affecting how analgesics (including opioids) are absorbed, distributed within the body, metabolized and excreted (10). SCI-specific pain classifications (i.e. the International Spinal Cord Injury Pain Classification (11)) have enabled standardization of various types of SCI-specific pain (e.g. neuropathic pain) which are used to guide clinical management options (11). Current clinical practice guidelines recommend that opioids are considered as a fourth-line option in the management of neuropathic pain, with evidence supporting the prescription of gabapentin and pregabalin ahead of opioids (12). Despite current evidence-based management guidelines, opioids are frequently prescribed to those with SCI, and compared to the general population, they are more likely to receive prescriptions for high morphine-equivalent doses of long-acting opioids (13).

The majority of opioids are prescribed in the community by general practitioners (14), who are not necessarily familiar with the complexity of SCI, including injury-specific pain and the nuances of its management. There has been some previous research examining opioid prescribing practices in people with SCI, including a recent scoping review on the topic (15). However, this review considered a limited scope of articles published between 2014 and 2021 and while prevalence of opioid use was reported (identified among 7 of the 16 included articles), it was not the main outcome. Consequently, there remains a clear need for a deeper understanding of historical and current prescribing trends within this population. The aim of this systematic review and meta-analysis is to synthesize evidence relating to the prevalence of opioid use among adults with SCI.

Review questions

The current systematic review sought to address the following primary research question: what is the prevalence of prescription opioid use among adults with SCI? In addition, secondary research questions were: (a) what are the associated harms, adverse effects or events related to prescription opioid use for adults with SCI? and (b) what personal characteristics are associated with prescription opioid use?

Methods

Study protocol

This systematic review followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (16). The review was prospectively registered in the PROSPERO database (CRD42022350768).

To assist with reader clarity, the term ‘opioid use’ will be used throughout this review irrespective of the method in which opioid data was measured. For example, dispensing data reflects medications that are dispensed through pharmacies but may not necessarily be consumed. Similarly, charted medications may reflect either medications prescribed by practitioners at the medical facility or self-reported medications being used by patients. This is distinct from either prescribing data or self-reported use of opioids. However, each of these methods of data capturing reflects a general approach to measuring and understanding opioid use among a group of participants. As such, opioid use will be used throughout, unless otherwise specified.

Main outcomes

The main outcome was prevalence of opioid use among SCI populations. Both point prevalence (opioids used at a specific time, e.g. current opioid use) and period prevalence (opioids used over a specified time, e.g. 12 months) were included (17). For studies where data was collected directly from participants that captured both period and point prevalence, only point prevalence was used to reduce errors from participant recall.

To avoid inadvertently overestimating prevalence, studies that captured ‘pain medications’ without specifying opioids were excluded on the basis that participants may report non-opioid medications being taken for pain. When prescription opioid prevalence was reported, secondary outcomes were also extracted (if reported). Secondary outcomes were: (1) opioid dose, (2) adverse effects relating to prescription opioids; (3) opioid effectiveness and (4) characteristics associated with prescription opioids. Considered characteristics included but were not limited to personal, health and injury factors.

Searches strategy

A comprehensive search from 1 January 2000 to 7 September 2023 was conducted in five databases: PubMed (MEDLINE), EMBASE, CINAHL, Web of Science and Scopus. The search strategy was designed in conjunction with an experienced librarian, and searches were undertaken by the principal investigator. Language was limited to English, and key terms relating to opioid prescribing, dispensing or use among people with SCI were included. The full search strategy is detailed in Supplement 1.

Eligibility criteria

Empirical quantitative observational studies without intervention (e.g. prospective or retrospective cohort studies, cross-sectional studies) for adults (≥18 years) with SCI and living in community-based settings were considered. With respect to aetiology of the SCI, eligible studies included those which examined people with traumatic SCI only, people with non-traumatic SCI only or studies in which people with both traumatic and non-traumatic aetiology (henceforth referred to as mixed aetiology) were participants. Studies which did not specify whether the aetiology of the SCI was either traumatic or non-traumatic (i.e. non-specified aetiology) were also included. Studies based on prescription opioids inclusive of any context (prescribing, dispensing, self-reported use, charted medications) or duration (point or period prevalence) were eligible. Studies were required to be published in English and since the year 2000.

Studies examining illegal use of prescription opioid analgesics or illegal opioids (e.g. heroin) were excluded. Additional exclusion criteria were: (1) reviews, meta-analyses, case-series and case-reports, (2) prevalence of prescription pain medicines when not explicitly identified as opioids (e.g. analgesics, ‘pain medication’), (3) inpatient acute or rehabilitation settings, and (4) studies reporting on participants with dual diagnoses of SCI and cancer. Studies involving mixed populations (e.g. other types of injury or disease in addition to SCI) were included if data from participants with SCI was able to be extracted separately.

Study selection process

Search results were imported to Endnote (18) for duplication detection prior to being uploaded to Rayyan (19), where blinded title and abstract, and full-text screening were conducted by two independent reviewers (SB and AR). The same two screeners (SB and AR) assessed full-text articles for eligibility criteria, and recorded exclusion reasons. Conflicts during the selection process were resolved by a third investigator (VM).

Data extraction

Data were extracted using an Excel (20) spreadsheet designed specifically for the study and piloted on a small sample of selected studies in the first instance. Extracted data included: study design, sample size, cohort demographics (e.g. sex, age), injury characteristics, opioid prevalence (incidence and denominator), opioid context (i.e. prescribing, dispensing, self-reported use), prevalence type (i.e. point, period), opioid-related harms, opioid efficacy and factors relating to opioids. Relevant injury and pain characteristics were also extracted, including time since injury, aetiology of SCI (i.e. traumatic only, non-traumatic only, or mixed cohorts), completeness of injury (i.e. complete, incomplete), level of injury (i.e. cervical, non-cervical) and the presence of pain (including pain type if reported).

Risk of bias (quality) assessment

Study-level risk of bias was independently assessed by two reviewers (SB and AR), with any discrepancies resolved by a third reviewer (VM). The risk of bias was determined based on a prevalence-specific tool developed by Hoy and colleagues (21). The risk of bias tool includes ten questions, with five questions relating to each of (1) external and (2) internal validity. The ten risk of bias questions and the modifications made to contextualize the tool to the current study’s population are detailed in Supplement 2.

Data analysis

All analyses were conducted in R (22) using the RStudio environment (23). Descriptive summaries of study characteristics were completed for: country, data collection method (prescribing or dispensing data, self-reported use or charted medications), prevalence type (point or period), aetiology of SCI (traumatic, non-traumatic or mixed), participant sex and pain presence reported across included studies.

Meta-analyses

The prevalence of opioid use was pooled from individual studies. We assumed that the true prevalence would differ across populations and accounted for this in the modeling. To do this, prevalence of opioid use was determined as the incidence of opioids (either used, dispensed, charted or prescribed) divided by the number of individuals with SCI in the study. Where studies reported opioid incidence for a subset of participants (e.g. only participants who reported pain), the subset was used as the denominator. This was consistent with choosing the lower risk level as determined by the risk of bias assessment utilized for this review (21). Studies that included propensity-score matched cohorts (e.g. SCI with neuropathic pain and SCI without neuropathic pain) and reported the prevalence of opioid dispensing separately for each matched cohort were included in the meta-analysis as separate cohorts. As such, the number of populations included in the meta-analysis exceeds the number of studies included in this review.

To determine the overall prevalence of opioid use among adult populations with SCI, a random-effects meta-analysis was conducted using the metaprop function from the meta package (24). A generalized linear mixed model with Logit transformation approach was used (25). This approach is considered an appropriate method compared to the Freeman-Tukey double arcsine transformation, in the use of single proportion meta-analyses (26). The Clopper–Pearson method was implemented to adjust the confidence intervals (CIs) of the overall estimate. Prevalence and 95% CIs were presented using forest plots.

Between-study heterogeneity was assessed using prediction intervals. In interpreting the results, we based our conclusions on the prediction interval as it accounts for between-study heterogeneity (unlike the 95% CI) and has a more practical interpretation than other measures of heterogeneity (27). The I² and the variance of (the distribution of) the true effect sizes (τ²) are also reported for transparency. Where data could not be pooled, a narrative synthesize was completed.

Sensitivity analyses

Following the examination of an overall effect size for individual studies, a sensitivity analysis was conducted to determine whether studies assessed as high risk regarding the main outcome (opioid use) had an effect on the pooled prevalence. The same meta-analysis approach described above was carried out for studies after exclusion of those with high risk of bias. The effect of individual studies on the overall pooled prevalence was assessed based on the leave-one-out approach, conducted using the metainf function (28).

Subgroup analyses

It was assumed that there would be heterogeneity across the included papers and differences in prevalence were explored through subgroup analyses. Subgroup analyses explored the effect of the data collection method (self-reported, charted medication, prescribing or dispensing data) and prevalence type (point or period). These two characteristics were chosen as they were deemed likely to have an impact on the prevalence of opioid use, pending whether the data was self-reported or derived from pharmaceutical datasets; and whether opioid use was considered over a period or at a single point in time. A subgroup analysis was also conducted for studies that reported a prevalence of opioid use based on duration of opioid action (short- and long-acting). A final subgroup analysis was conducted based on the time since injury of included studies (1 year; 2–10 years; 10–15 years; over 15 years), given pain and subsequently opioid use is likely to fluctuate over time following injury. All sensitivity and subgroup analyses were conducted using the same random-effects meta-analysis approach described above.

Results

Search results

There were 4969 citations identified from the search process, of which 40 articles met eligibility criteria. Two articles were found to report results from the same data source and with the same sample of people with SCI (29, 30). Neither of these studies offered additional information relating to the secondary aims of this review and consequently were excluded to avoid duplication in the meta-analysis. As such, 38 articles were included in this review. The PRISMA study selection flow diagram is presented in Figure 1.

Figure 1.

Preferred reporting items for systematic review and meta-analyses (PRISMA) flow diagram.

Study characteristics

A description of cohorts, the opioid data source, prevalence type, injury index period, and observation period for the 38 studies included in the review are provided in Table 1 (4, 13, 31–66). The majority of studies were based in the USA (20/38, 53%) and Canada (7/38, 18%) (42–46, 58, 59), with two studies from each Sweden (2/38, 5%) (56, 57) and Denmark (2/38, 5%) (41, 55). The remaining studies were each conducted in a single country (e.g. Australia (32), Iran (35)).

Table 1.

Cohort description, opioid data source, injury index period, prevalence type, observation period and sample size (n = 38).

Country	Author	Year	SCI cause	Cohort description	Opioid data source	Injury index period	Prev. type	Prev. period	Observation / recruitment period	Sample size
Iran	Behnaz	2017	Mixeda	Males, aged ≥18, with chronic SCI and erectile dysfunction.	Hospital medical records	N/A	NR	–	Jan 2013–Mar 2015	319
Ireland	Burke	2019	Mixed	Aged ≥18 and a member of Spinal Injuries Ireland (SII).	Survey data	N/A	Period	6-m	Not reported	643
USA	Carbone	2013	Trauma	Male veterans with traumatic SCI for ≥2yrs. Identified by Veteran Affairs Spinal cord dysfunction registry.	VA pharmacy benefits management group prescription database	N/A	Period	5-yr	FY2002 to FY2007	7447
USA	Cardenas	2006	Trauma	Traumatic SCI, aged ≥18, with chronic pain (pain within the previous 3-months before responding to the survey).	Survey data	N/A	Pointd	-	Not reported	117
USA	Carlozzi	2021	Mixeda	Aged ≥18 with a medically documented SCI, fluent in English and able to consent.	Survey data	N/A	Point	-	Not reported	173
Europe	Cragg	2016	Mixed	Adults able to consent and participate in the first study assessment within 6 weeks of the SCI.	Survey data	N/A	Point	-	2017	225
USA	DiPiro	2021	Mixeda	Individuals with a newly diagnosed SCI in 2013–14, living in South Carolina and alive during the 2–3 year follow-up period.	South Carolina Reporting & Identification Prescription Tracking System (SCRIPTS)	2013–2014	Period	2-yr	2015–2016e	503
USA	Durga	2011	Mixeda	Convenience sample of male Veterans with chronic SCI, presenting for annual evaluation from July 2006 to Apr 2007.	Milwaukee Veterans Affairs Medical Centre	N/A	Periodc	NR	July 2006 to April 2007	60
DEN	Finnerup	2001	Mixed	Outpatients of the Viborg rehabilitation center for SCI. Minimal cohort details provided.	Survey data	N/A	Point	-	Not reported	330
UK	Gore	2013	Mixeda	Aged ≥18 with SCI on or after 1 July 2004 who initiated pregabalin at least 9 months after their SCI diagnosis, with 9-month pre- and 9-month post-index data.	The Health Improvement Network (anonymized general practice records)	Diagnosed on/after 1 July 2004	Period	9-m	Not reported	72
CAN	Guan	2021a	Non-trauma	Non-traumatic SCI, aged >18, eligible for provincial drug coverage program.	Ontario Drug Benefit Database	1 Apr 2004 to 31 Mar 2015	Period	12-m	2004–2015	3468
CAN	Guan	2021b	Trauma	Individuals eligible for the provincial drug coverage, who had a traumatic SCI during the index period.	Ontario Drug Benefit Database	1 Apr 2004 to 31 Mar 2015	Period	12-m	2004–2015	934
CAN	Guilcher	2018	Trauma	Aged ≥66 at the time of hospitalization for traumatic SCI.	Ontario Drug Benefit Database	1 Apr 2004 to 31 Ma 2014	Period	12-m	2004–2014	418
CAN	Guilcher	2021	Trauma	Traumatic SCI acquired during the index period.	Narcotics Monitoring System Database	1 Apr 2004 to 31 Mar 2015	Period	12-m	1 Apr 2016 to 31 Mar 2017	1842
CAN	Gupta	2020	Mixed	Aged ≥18, community dwelling, living in Canada and prescribed ≥1 medication at the time of the survey.	Survey data	N/A	Period	12-m	May to Oct 2018	160
USA	Hand	2018	Mixeda	Aged: 18–64 years. Propensity-score matched opioid users with SCI and opioid users without SCI. Required to have ≥18 months of continuous insurance coverage from the date of the first opioid claim.	MarketScan Commercial Claims and Encounters Database	1 Jan to 31 Dec 2012	Period	12-m	2012–2013	1454
USA	Hatch	2018	Mixed	Veterans, aged ≥18 with SCI/dysfunction, using Medicare or Veteran Affairs pharmacy benefits.	VA Managerial Cost Accounting National Data Extract Pharmacy Databases	Injuries prior to 2011	Period	12-m	2011	13,442
NL	Heutink	2011	Mixed	Aged ≥18, chronic pain, living in the community & attended the Rehabilitation Centre ‘De Hoogstraat’ from 1990 to 2005.	Survey data	N/A	Pointd	-	1990–2005	215
USA	Hwang	2015	Mixeda	Adults with pediatric onset SCI who received rehabilitation services at one of the SCI specialty programs in Chicago, Philadelphia or Northern California.	Survey data	N/A	Periodc	NR	Jan 2011	159
AUS	Karran	2022	Mixed	Aged ≥18, with SCI, who experience persistent pain.	Survey data	N/A	Period	6-m	Not reported	43
USA	Kokorelis	2019	Trauma	Males aged ≥21, who were evaluated over a 10-year period at a SCI-specialized centre for trauma-induced SCI	Medical records from a specialized SCI center.	N/A	Point	-	June 2005 – June 2015	279
USA	Kratz	2018	Mixeda	Aged >18, at least >1-year post-SCI, with chronic pain (defined as >4/10 average pain in the past month).	Survey data	N/A	Point	-	June 2014 to Jan 2016	120
USA	Mann	2013	Mixeda	Aged >18, with chronic SCI (≥1 year) and diagnosed with NeP managed at their physician’s practice for ≥6 months.	Medical records from 14 community-based physician practices	Injuries prior to Sept 2010	Period	6-m	Sept 2011 –June 2012	103
USA	Margolis	2014a	Mixed	Aged ≥18 and commercially insured. All individuals with SCI, propensity-score matched with and without NeP.	MarketScan Commercial Database	1 Jan 2006 to 30 June 2011	Period	12-m	1 Jan 2005–30 June 2012	7048b
USA	Margolis	2014b	Mixed	Aged ≥18 at injury, with SCI and propensity-score matched with and without NeP. Required to have ≥6 months of continuous Medicaid eligibility prior to injury.	MarketScan Multi-state Medicaid Database	1 Jan 2006 to 30 June 2011	Period	12-m	1 Jan 2005–30 June 2012	1092b
USA	McCasland	2006	Trauma	Aged ≥20; convenience sample of traumatic SCD registry	Survey data, validated with medical records	N/A	NR	-	Jan 2000–Dec 2002	63
DEN	Nielsen	2017	Mixeda	Members of a Danish SCI organization.	Survey data	N/A	Point	-	24 Sept-1 Dec 2015	130
SWE	Norrbrink Budh	2003	Mixed	People with SCI attending Spinalis SCI unit for a yearly assessment were asked to fill in pain questionnaires.	Data from annual medical assessment.	N/A	NR	-	1999	130
SWE	Norrbrink Budh	2005	Mixed	People with SCI attending Spinalis SCI unit who completed the previous survey from Norrbrink et al., 2003	Survey data	N/A	Point	-	2002	90
CAN	Patel	2017	Mixeda	Adults with a documented SCI, attending a primary care clinic in Ontario.	Medical records from a primary care clinic	N/A	Point	-	Aug 2012 – Mar 2013	19
CAN	Rouleau	2011	Mixed	All SCI patients through Interval Rehabilitation Centre in Quebec, Canada.	Medical records from a Rehabilitation Centre	N/A	NR	-	Not reported	151
USA	Tsai et al.	2021	Trauma	Aged ≥18, with non-NeP, ≥ 1-year post-injury and enrolled in one of six participating SCI centers.	Survey data	N/A	Period	12-m	Mar 2017–July 2019	190
USA	Turner	2001	Mixeda	Aged ≥18, with SCI.	Survey data	N/A	Periodc	NR	Feb 1997–July 1998	384
USA	Ullrich	2014	Mixed	Veterans with SCI diagnosed with and without depression. Reported as 2 distinct samples.	Veterans Health Administration National Registry for Depression	FY1997 to FY2007	Period	12-m	2007	2796
USA	Warms	2002	Mixeda	Aged ≥18, with SCI and pain. Reported as two samples: Sample 1 is reported in Turner et al., 2001. Only Sample 2 included in this review.	Survey data (only Sample 2, as Sample 1 is accounted for already).	N/A	Pointd	-	Aug 1998 – June 2000	163
China	Wen	2013	Trauma	Patients who sustained a SCI in the 2008 Sichuan earthquake.	Survey data	May 2008	NR	-	Oct 2012	26
USA	Widerstrom-Noga	2003	Mixeda	Aged ≥18 years, chronic SCI (≥18 mths), who reported chronic pain/nonpainful sensations in the preceding study.	Survey data	N/A	Period	18-m	Not reported	120
USA	Wilkinson	2023	Trauma	Aged ≥18 years, traumatic SCI, with ≥12 months of pre-injury and ≥15-months of post-injury data.	MarketScan Research Database	2009–2019	Period	12-m	2009–2019	7187

Abbreviations: FY = financial year; N/A = not applicable; NeP = neuropathic pain; NR = not reported; SCI = spinal cord injury; VA = Veterans Affairs.

^aSCI onset not specified (traumatic or non-traumatic), assumed to be a mixed cohort.

^bReported in article as two distinct sub-populations.

^cPeriod of prevalence not specified, identified as ‘routinely’ or ‘regularly’ taken medications.

^dPeriod (past medications) and point prevalence both reported but only point prevalence extracted and used for this review.

^eObservation period inferred from methods section details.

Self-reported use of opioids was the most common data collection method, used in 47% of studies (18/38), followed by dispensed pharmaceutical datasets (11/38, 29%) and medications extracted from medical charts (7/38, 18%). Prescription datasets were only used in two studies (2/39, 5%). Fifty-eight percent (22/38) of studies utilized period prevalence. The most common period used to capture opioid use was 12-months (12/22, 55%), however, periods ranged from 6-months to 5-years (Table 1). Three studies did not provide enough details to identify the period for which prevalence of opioid use was captured (40, 50, 61). Point prevalence was used to capture opioid use among 11 studies, while 5 studies did not provide sufficient detail to classify whether prevalence type was period or point (35, 54, 56, 59, 64).

Many studies did not specify the aetiology of the SCI of participants (14/38, 37%), and for these studies the aetiology was assumed to include a mix of both traumatic and non-traumatic SCI (i.e. mixed aetiology). The remaining studies specified aetiology for participants, including 13 (34%) with mixed aetiology, 10 (26%) that were exclusively of traumatic aetiology and one study that was exclusively non-traumatic SCI. Four studies (11%) included males only populations (34, 35, 40, 51). The presence of pain (inclusive of all types) was reported for nearly two-thirds of studies (22/38, 58%, Table 2) and a criterion for the individual participant eligibility in 29% (11/38) of studies (Table 1).

Table 2.

Included studies for meta-analysis (41 cohorts reported from 38 articles).

First author	Year	Country	Design	Cohort type	SCI cohortb (N)	Male (N)	Mean age (yrs)	Injury characteristics				Opioid type	Opioid incidence (N)	Denominator (N)	Prevalence	Pain
								Cervical (N)	Complete (N)	Mean TSI (yrs)	Trauma (N)					All types	NeP	NI	MS
Use
Finnerup	2001	DEN	CSS	Mixed	330	230	43c	113	159d	9c	258	SR	36d	255	14	255	–	–	–
Turner	2001	USA	CSS	Mixeda	384	283d	43	196d	143d	12	308a	SR	186d	384	48.5	304	–	–	–
Warms	2002	USA	CSS	Mixeda	163	114d	42	85	–	4c	142a	SR	48	163	28.8	163	–	–	–
Widerström-Noga	2003	USA	CSS	Mixeda	120	94	41	62	-	10	-	SR	27	120	22.5	120	-	-	-
Norrbrink Budh	2005	SWE	CSS	Mixed	90	44	53	-	-	14	70	SR	31	90	34.4	90	43	20	-
Cardenas	2006	USA	CSS	Trauma	117	85	49	56d	-	17	117	SR	25	117	21.4	117	-	-	-
McCasland	2006	USA	CR/CSS	Trauma	63	61	58	32e	-	-h	63	SR	4	63	6.3	44	-	-	-
Heutink	2011	NL	CSS	Mixed	215	133	51	80e	92	12	138	SR	13g	207	6.3	215	149	65	131
Wen	2013	China	PC	Trauma	26	12	51	3	6	-	26	SR	2	22	9.1	22	19	6	6
Hwang	2015	USA	CSS	Mixeda	159	100	35	81	-	21	-	SR	23	159	14.5	-	-	-	-
Cragg	2016	Europe	PC	Mixed	225	180	48	120e	82	1	216d	SR	63	225	28.0	161	63	96	-
Nielsen	2017	DEN	CSS	Mixeda	130	94	56	-	-	31	-	SR	9	41	22.5	40	-	-	-
Kratz	2018	USA	CSS	Mixeda	120	88	47	46e	65	15	106	SR	49	120	40.8	120	-	-	-
Burke	2019	IRE	CSS	Mixed	643	447	52	218	172	17	456	SR	123	458	26.9	458	236	206	-
Gupta	2020	CAN	CSS	Mixed	160	71d	47	53d	-	18	93d	SR	85	160	53.1	-	-	-	-
Carlozzi	2021	USA	CSS	Mixeda	173	110	50	76e	-	-	-	SR	52	173	30.1	-	-	-	-
Tsai	2021	USA	CSS	Trauma	190	144	51	87	82	20	190	SR	84	190	44.2	190	0	-	-
Karran	2022	AUS	CSS	Mixed	43	31	49	20	17	16	31	SR	25	43	58.1	40	-	-	-
Charted
Norrbrink Budh	2003	SWE	OBS/CS	Mixed	130	65	-	-	-	-	-	CR	36	130	27.7	130	62	30	-
Mann	2013	USA	OBS/CS	Mixeda	103	72	49	-	-	-h	-	CR	35d	103	34.0f	103	103	-	-
Durga	2011	USA	OBS/CS	Mixeda	60	60	54	31	24	22	-	CR	18	60	30.0	-	-	-	-
Rouleau	2011	CAN	OBS/CS	Mixed	151	97	-	67e	28	-	82	CR	87	151	57.6	-	-	-	-
Behnaz	2017	Iran	OBS/CS	Mixeda	319	319	47	43	169	-	-	CR	37	319	11.6	-	-	-	-
Patel	2017	CAN	CRV	Mixeda	19	14d	47	-	-	-	-	CR	6	19	31.6	7	-	-	-
Kokorelis	2019	USA	RC	Trauma	279	279	-	165e	142	-h	279	CR	91	279	32.6	-	-	-	-
Prescribing
Carbone	2013	USA	RC	Trauma	7447	7447	-	2138e	2724	-h	7447	PD	5106	7447	68.6	-	-	-	-
Gore	2013	UK	RC	Mixeda	72	34	48	-	-	-	-	PD	44	72	61.1	72	72	-	-
Dispensing
Margolis	2014a	USA	RC	Mixed	3524	1916	48	569d,e	-	1	1343d	DD	2764d	3524	78.4f	3524	3524	-	1887d
Margolis	2014a	USA	RC	Mixed	3524	1950	48	614d,e	-	1	1118d	DD	1762d	3524	50.0f	1826d	0	-	1826d
Margolis	2014b	USA	RC	Mixed	546	240	40	111e,d	-	1	166d	DD	474d	546	86.8f	546	546	-	272d
Margolis	2014b	USA	RC	Mixed	546	247	42	130e,d	-	1	122d	DD	338d	546	61.9f	273d	0	-	273d
Ullrich	2014	USA	RC	Mixed	2615	2482d	-h	1123e,d	-	20	-	DD	1164d	2615	44.5	-	-	-	-
Ullrich	2014	USA	RC	Mixed	181	179d	-h	80e,d	-	21	-	DD	45d	181	25.1	-	-	-	-
Guilcher	2018	CAN	RC	Trauma	418	265	75	321	-	1	418	DD	332d	418	79	-	-	-	-
Hand	2018	USA	RC	Mixeda	1454	720	46	364	-	-	-	DD	1190d	1454	81.8f	-	-	-	-
Hatch	2018	USA	RC	Mixed	13442	12948	-h	5526e	-	-h	4176i	DD	5890	12161j	48.4	-	-	-	-
DiPiro	2021	USA	RC	Mixeda	503	357	-h	303	-	3	-	DD	269	503	53.5	-	-	-	-
Guan	2021a	CAN	RC	NTSCI	3468	1746	70c	-	-	1	0	DD	2062	3468	59.5	-	-	-	-
Guan	2021b	CAN	RC	Trauma	934	664	63c	626	-	1	934	DD	510	934	54.6	-	-	-	-
Guilcher	2021	CAN	RC	Trauma	1842	1372	51c	1143	-	6c	1842	DD	644	1842	35.0	-	-	-	-
Wilkinson	2023	USA	RC	Trauma	7187	3849	54	4031	-	1	7187	DD	3888	7187	54.1	1597	320	-	424
Total	-	-	-	-	52115	39643	-	18713	3905	-	27328	-	27677	50473	-	8318	5137	423	4819

Note. Three studies (Margolis et al., 2014a, Margolis et al., 2014b & Ullrich et al., 2014) described distinct sub-samples, and as such, this table details 41 cohorts from 38 studies.

Abbreviations: CR = chart review; CSS = cross-sectional survey; DD = dispensing database; MS = musculoskeletal pain; NeP = neuropathic pain; NI = nociceptive pain; NL = Netherlands; NR = not reported; NTSCI = non-traumatic spinal cord injury; OBS = observational study; PC = prospective cohort study; PD = prescription database; RC = retrospective cohort study; RCT = randomized controlled trial; SR = self-reported opioid use; TSI = time since injury (years).

^aSCI onset not specified (traumatic or non-traumatic), assumed to be a mixed cohort.

^bReflective of total SCI cohort, which may not reflect full cohort in paper (e.g. excluding non-SCI control cohorts).

^cReported in study as median.

^dNumber not reported, n is calculated based on the reported proportion (i.e. the conservative calculated n was used when multiple numbers would equate to the same percentage).

^eCervical injuries not reported in study, n is spinal cord injuries classified as tetraplegia.

^fStudy reports on different durations/strengths of opioids (e.g. short-acting opioids), the highest prevalence is reported.

^gBased on current opioid use, as study captures self-reported use ‘ever’ and ‘currently’.

^hReported as categorical.

ⁱThirty-seven percent (n = 4964/13442) of the cohort were not able to be identified as traumatic or non-traumatic.

^jn = 1281 individuals excluded from prevalence as data is from two types of pharmacy dispensaries, and it is not certain the level of overlap.

Risk of bias

The risk of bias assessments and scores are presented in Supplementary Table 3. Risk of bias was determined to be low (20/38, 53%) or medium (16/38, 42%) across most studies. Two studies were determined to have a high risk of bias (40, 54).

Meta-analyses

Details of the study samples included in the meta-analysis are provided in Table 2. Three studies included propensity-score SCI matched populations, which were reported as two distinct subgroups of participants (4, 53, 62). Consequently, there are 41 samples from the 38 studies included in the meta-analyses. The mean age of participants ranged from 35 to 75 years. The 41 cohorts included a total of 52,115 people with SCI (Table 2). The largest sample consistent of 13,442 individuals from a national Veterans database in USA (48), while the smallest sample was 19 people who had information extracted from medical records following attendance at an Ontario primary care clinic (58).

Overall pooled prevalence

The overall pooled prevalence of opioid use from the 38 studies was 39% (95% CI: 32–47%, Table 3). Considerable between-study heterogeneity was evident with the prediction interval being two times wider than the 95% CI (prediction interval: 7–84%; Figure 2). This suggests a high degree of variability or uncertainty in the estimated prevalence of opioid use across the different studies.

Table 3.

Meta-analysis results from studies reporting opioids in the SCI population.

Prevalence of	Number of cohorts*	Sample size	Random effects
			Prevalence (95% CI)	I2	τ2	95% prediction interval
All non-overlapping cohorts	41 (from 38 studies)	50,473	39% (32–47)	99%	1.06	7–84%
Excluding studies with high risk of bias	39 (from 36 studies)	50,350	41% (33–49)	99%	1.00	8–84%
Opioid context
Charted medications	6	931	31% (20–45)	95%	0.47	5–78%
Dispensed medications	14 (from 11 studies)	38,903	59% (49–69)	99%	0.65	19–90%
Prescribed medications	2	7519	68% (67–70)	45%	0	67–70%
Self-reported use	19	3120	26% (20–34)	93%	0.59	6–65%
Prevalence type
Period prevalence	25 (from 22 studies)	48,099	51% (42–60)	99%	0.74	15–87%
Point prevalence	11	1689	25% (18–32)	87%	0.33	8–56%
Not reported	5	685	18% (8–38)	96%	1.17	1–91%
Time since injury
1 year	10 (from 8 studies)	21,826	65% (53–76)	99%	0.64	21–93%
2–10 years	5	2883	29% (19–43)	97%	0.41	4–80%
10–15 years	4	801	28% (13–52)	96%	1.02	0–98%
Over 15 years	11 (from 10 studies)	4343	30% (22–39)	96%	0.48	8–69%
Not reported or listed categorically	11	20,620	35% (24–49)	99%	0.85	6–83%
Opioids by length of mechanistic action
Short-acting opioids	7 (from 5 studies)	10,157	55% (35–74)	100%	1.23	6–96%
Long-acting opioids	7 (from 5 studies)	10,157	17% (11–25)	99%	0.38	4-54%

Note. The number of cohorts does not directly correlate to the number of studies as three studies report on discrete propensity-score matched cohorts (4, 53, 62). If the number of studies is not specified, the number of cohorts is equivalent to the number of studies.

Figure 2.

Meta-analysis of prevalence of opioids among people with SCI, stratified according to opioid context (charted medications, dispensed medications, prescribed medications, self-reported use) and ordered by year of publication. CI, confidence interval.

Sensitivity analysis

The sensitivity analysis identified no change in the pooled prevalence of opioid use when the two studies with high risk of bias were removed (pooled prevalence: 41%; 95% CI: 33–49, prediction interval: 8–84%). There was also no difference in the overall pooled prevalence based on the leave-one-out approach (Supplementary Figure 4).

Subgroup analysis

Subgroup analyses were conducted for four variables, with findings presented in Table 3.

Pooled prevalence by opioid data collection method. There was considerable variation in the prevalence estimates when comparing the data collection methods for capturing opioid use (Figure 2, Table 3). The pooled prevalence of opioid use for prescribing data was highest at 68% (95% CI: 67–70; I²= 45%, τ²= 0), and lowest for self-reported opioid use at 26% (95% CI: 20–34, I²= 93%, τ²= 0.59, prediction interval: 6–65%).

Pooled prevalence by prevalence type. The pooled period prevalence of opioid use was 51% (95% CI: 46–60; prediction interval: 15–87%). The pooled point prevalence of opioid use was lower at 25% (95% CI: 18–32; prediction interval: 8–56) (Table 3).

Pooled prevalence by time since injury. The pooled prevalence of opioid use was highest for the first year following SCI at 65% (95% CI: 53–76; prediction interval: 21–93%). Prevalence for the periods of 2–10 years; 10–15 years and over 15 years since injury remained consistent and was less than half the prevalence in the first-year post-injury (Table 3). The pooled prevalence for the 11 studies where time since injury was not specified was 35% (95% CI: 24–19; prediction interval: 6–83%).

Pooled prevalence of opioids by duration of action. Five studies reported opioids categorized into duration of action, specifically short- and long-acting (4, 42, 46, 53, 53). The pooled prevalence was 55% (95% CI, 35–74, prediction interval: 6–96%) for short-acting opioids and 17% (95% CI, 11–25, prediction interval: 4–54%) for long-acting opioids (Figure 3; Table 3). High heterogeneity was also evident, with prediction intervals 2–4 times wider than the 95% confidence intervals. A sixth study also reported on duration of action, however, this included a more complex classification of opioids into 8 distinct categories by dose and duration of action. These categories were not able to be collapsed into short- and long-acting categories and were excluded from this subgroup analysis. The 8 classification categories from this paper were: short-term low dose long-acting opioids (81.8%), long-term low dose long-acting opioids (10.2%), short-term high dose long-acting opioids (1.4%), long-term high dose long-acting opioids (6.6%), short-term low dose short-acting opioids (59.4%), long-term low dose short-acting opioids (40.6%), and short- and long-term high dose short-acting opioids (both 0%) (13).

Figure 3.

Forest plot of meta-analysis for studies that report (A) short-acting and (B) long-acting opioids. CI, confidence interval.

Figure 2 shows studies plotted by year of publication, as a proxy for the chronological order of the data, which shows no clear pattern. Year of data collection was not considered as this was unreported in 21% (8/38) of studies (see Table 1). Based on the high heterogeneity (prediction intervals for subgroups being up to 4 times wider than the 95% CIs) and the smaller number of included studies, a meta-regression was not completed.

Opioid dosage

Mean opioid dose across studies was not able to be determined, as it was not reported in 95% (37/39) of studies. DiPiro et al. report a mean daily morphine milligram equivalents (MME) of 41 ± 70 (range = 0.17–750) for 503 individuals with mixed onset of SCI (39). The second study including dose reports a median daily dose of 32.6 mg (interquartile range [IQR], 19.3, 87) MME for 1842 individuals with traumatic SCI (46). A further three studies included categorized opioid dose within their methods. This included varying classifications of high opioid dose: ≥ 120 mg/day (13), > 225 mg/day (34), while Guilcher et al. defined high-dose use as individuals with ≥90 mg/day for ≥90 days (46).

Adverse events relating to opioid use

Four studies (11%) identified adverse outcomes relating to opioid use (full details in Supplementary Table 5). Carbone et al. found a positive relationship between opioid use and incidence of lower limb fractures in a cohort of male veterans with traumatic SCI (hazard ratio: 1.82, 95% CI, 1.59–2.09) (34). Two studies identified associations between opioid use and hormone levels in male only chronic (>2 years) SCI populations: biochemical androgen deficiency among a cohort of 319 (35) and lower serum testosterone levels among 60 male Veterans (40). Neither study was examined beyond the univariate level. A 2021 study explored associations between opioid analgesics and measures of cognitive performance in cohort of 173 individuals with mixed aetiology SCI and identified poorer outcomes for those on opioids in five of the nine cognitive measures (e.g. verbal fluency, working memory) (33).

Perceived effectiveness of opioids

Among the 18 studies that captured self-reported opioid use; only five also collected measures of effectiveness or helpfulness of opioids (37, 49, 60, 63, 65). Each study used a separate measure to capture effectiveness. Despite this, opioids were often perceived to be the most effective method for pain relief among the pharmaceutical and non-pharmaceutical approaches investigated. Cardenas and Jensen used a 10-point scale to measure pain relief, ranging from 0 (no relief) to 10 (complete relief), where opioids offered the highest level of relief among the 10 various medication groupings (37). Another study using a similar 5-point Likert scale from 1 (not at all helpful) to 5 (extremely helpful) and also found opioids to be the highest rated medication to help with pain (63). Heutink and colleagues used a 3-item response to capture effectiveness of current treatments (not at all helpful; somewhat; to a large extent), where opioids had the highest rating among the six medication groups for being helpful ‘to a large extent’ (49). Tsai and colleagues measured helpfulness of treatments using a simple ‘no’ or ‘yes’ response, where opioids were the highest rated treatment among a list of twelve pharmaceutical and non-pharmaceutical approaches (e.g. heat therapy, massage) (60). The fifth study used a 5-point response to measure the perceived effectiveness of treatments (response options: worse; no effect; slightly better; considerably better; pain free) (65). Only four medications were rated by participants as contributing to being pain free, where opioids had the highest pain-free rating (65). For full details of measures and outcomes for effectiveness, see Supplementary Table 6.

Factors associated with opioid use

Five studies included models of personal characteristics that were associated with opioid use (see Supplementary Table 7) (39, 43, 44, 46, 52). Studies were varied in relation to the population considered (e.g. traumatic only (43, 46), non-traumatic only (44), or a mixed cohort (39, 52)); the outcome variable (e.g. dichotomized opioid use or chronic opioid use); and the number and type of variables considered (with modeling exploring association between 4 and 12 variables (39, 43); Supplementary Table 7). Age, sex and time since injury were important across multiple studies (39, 43, 44, 46). However, these studies were inconsistent in the direction of an effect for sex and included inconsistent formats of variables (e.g. age included as continuous or categorical). They also exclusively represented samples of people with SCI in Canada (n = 3) or the USA (n = 2). Other variables, for example, chronic obstructive pulmonary disease or osteoarthritis (43), and measures of functional capacity (44), were only considered in a single study, and their level of importance was even less certain. Given the heterogeneity across models used and the variables considered, a meta-analysis for factors related to opioid use was not conducted.

Discussion

This is the first systematic review and meta-analysis to be conducted on the prevalence of opioid use in people with SCI living in the community. The meta-analysis identified an overall prevalence of opioid use of 39% among adult SCI populations, with extremely high heterogeneity, evidenced by a prediction interval two times wider than the 95% confidence interval (Figure 2; Table 3). In the case of wide prediction intervals, as in the current study, caution is needed when interpreting findings of meta-analyses (67). Consequently, subgroup analyses reflecting the data collection method offered a better understanding of the prevalence of opioid use. Secondary outcomes were seldom reported, with factors associated with opioid use and adverse events reported in less than 15% of studies (Supplementary tables 5 and 6). This review highlights a clear need for additional studies and stronger, more consistent reporting of main outcomes and injury characteristics to enable better understanding of opioid use in this high-risk population.

A major theme evident from this review was the considerable variability and lack of consistency in the reporting of methodological and injury characteristics. Prevalence type (point or period prevalence) was unreported in 13% of studies, while 3 of the 11 studies reporting period prevalence did not specify the period for which they measured opioid use (Table 1). The recruitment or observation period was unreported in approximately 20% of studies (Table 1), often making it difficult to determine whether populations across the included studies were distinct. Rates of reporting of SCI-specific injury characteristics were highly varied, including high rates of missing details for basic SCI variables. For example, injury aetiology not specified in 37% of studies (Table 1). There has long since been a recommendation for standardization regarding the reporting of characteristics for samples of people with SCI (68). Yet even the more recent articles included in this review were not consistent with these standards. These reporting deficits limit the generalizability of findings and highlight the need for standardized and clearer reporting.

Subgroup analysis for opioid context followed a clear pattern of higher proportions prescribed or dispensed opioids (59–68%), compared to those who self-reported opioid use (26%, Figure 2). This was mostly consistent with the findings regarding prevalence type. Period prevalence (51%, Table 3) which was typically obtained through prescribing or dispensing data was higher than point prevalence (25%, Table 3), which was captured through self-reporting. Dispensing and prescribing data can routinely be captured across geographical catchment areas, and in the current study, it accounted for nearly three quarters of the total pooled participants. As such, this data collection method is often ideal for ascertaining population-level trends. Such data monitoring sources can directly influence clinical practice and policy (69), for example, when escalations in medication prescribing are detected. Yet, the difference in prevalence between opioids dispensed and those reported as being taken (a difference of 33%) has implications about the potential overestimation of opioid use among people with SCI. While self-reporting may account for underreporting of opioid use in surveys and dispensing data does not directly relate to opioids that are consumed, the wide difference brings into question the reliability of the data monitoring information compared to what may actually be happening. This again reiterates caution when interpreting findings from large datasets that may overestimate opioid use.

Secondary outcomes and opioid context were also rarely reported. Opioid dose was unreported in 95% of studies. This was despite pharmaceutical data or medical records accounting for approximately half of the data sources of included studies. When opioid dose was reported, this was typically considered over the duration of the study period (any opioid use over 12 months) and prevalence over smaller increments (e.g. 30-day intervals) were not considered. The remaining studies reported dose using varying cut points for what they considered ‘high’ dose opioids (Supplementary table 5), again making comparisons difficult. The effectiveness of opioids was also seldom considered, and for the five papers that did consider opioid efficacy for pain relief, each study utilized a different measure. Adverse effects were also inconsistently reported. For example, a common side effect of opioids is constipation (70), and people with SCI are a population known to experience high rates of bowel problems (8). Yet bowel functioning was not among the reported adverse effects, again suggesting opioids as an under-researched area among this population. For the secondary outcome of factors associated with opioid use, studies also differed in what variables they considered in their statistical modeling. As such, it was not possible to provide any clear indication of which characteristics were important when considered collectively. Studies that can capture full opioid details (e.g. pharmaceutical data that includes detail of medication dose and dispensing dates) should focus on presenting more in-depth opioid information to help contextualize opioid use among people with SCI. There is also a need for further and more robust investigation into associations between characteristics and opioid use as a means of identifying subgroups within SCI populations who may be at greater risk of opioid-related harms.

The heterogeneity and imprecision of prevalence in the current study make it difficult to make clinical inferences about opioid use among SCI populations. However, the pooled prevalence for self-reported opioid use (28%, Figure 2) was similar to a recent meta-analysis on opioid use among chronic non-cancer pain (27%, 95% CI, 23–31) (71). In the current review, two-thirds of included studies had cohorts with mixed aetiology of SCI, yet opioid use was rarely contextualized to injury aetiology. Of the remaining studies that considered aetiology-specific populations, only one study focused on non-traumatic SCI, compared to a quarter of studies that focused on traumatic SCI. It is well recognized that characteristics vary within SCI populations, particularly in relation to injury aetiology. Traumatic SCI is historically more common among younger and male populations, while those with non-traumatic onset are often older and frequently considered medically complex irrespective of their SCI (72–74). Such differences are likely to have an impact on the prevalence of opioids dispensed to or used by these subgroups. Yet, the lack of reporting by aetiology type and a reduced focus on non-traumatic cohorts makes such comparisons unfeasible based on currently available studies.

This systematic review presents a rigorous exploration of opioid use in people with SCI living in the community; and in doing so acknowledges several practical considerations for moving forward. Firstly, caution should be exercised when interpreting studies reporting opioid use among SCI populations given the lack of generalizability, and in consideration of the significant heterogeneity demonstrated by the prediction intervals. Secondly, future studies focused on adults with SCI should ensure a standard practice of reporting injury-based characteristics for contextualizing opioid uses (and other outcomes). Recommended variables to include for reporting generally among studies, as well as in SCI or opioid-focused study are listed in Table 4. Thirdly, in consideration of the well-documented characteristic differences within SCI populations (e.g. in relation to SCI aetiology), greater emphasis needs to be given to subpopulations when studies include SCI of mixed aetiology. Fourthly, future studies need to facilitate a more in-depth picture of opioid use, including dose and prevalence over shorter increments given the fluctuation of pain, and in consideration of opioids being intended for shorter-term use.

Table 4.

Recommended variables for reporting.

Focus	Information for reporting
General methods information	Recruitment and observation period
	Year of data collection
	Prevalence type: point or period prevalence (including specification of the period)
	Age and sex
Information for studies that include people with SCI	Aetiology: traumatic, non-traumatic or mixed aetiology (including specific numbers when reported as mixed populations)
	Neurological level of injury: tetraplegia vs. paraplegia, and specification of SCI level if possible (e.g. C1–4, C5–8)
	Degree of impairment as complete vs. incomplete SCI, or AIS Grading A­–E, if known
	Time since injury
	Type of SCI-specific pain classification (ISCIP Classification)
	Presence of other major injuries, illnesses or disease at the time of SCI that may also result in chronic pain
Information relating to opioids	Prevalence of opioids
	Opioid data collection method
	Context of pain
	Duration of opioid use (e.g. number of days of prescription coverage)
	Dose of opioids, including changes over time, where possible
	Type of opioid (e.g. long vs. short-acting opioids)
	For studies focusing on the immediate post-injury period, opioid use during initial inpatient rehabilitation
	Side effects relevant to SCI populations: including bowel and bladder problems, cognitive and behavioral side effects and respiratory problems.

Abbreviations: AIS = American Spinal Injury Association (ASIA) Impairment Scale; ISCIP = International Spinal Cord Injury Pain; SCI = spinal cord injury.

Study limitations. This systematic review is not without limitations. We acknowledge that despite a comprehensive search strategy, relevant articles published in non-English languages may have been missed. Another primary limitation is the smaller number of included studies and high heterogeneity across study populations, which prevented further explorations into underlying reasons for variations in prevalence. This high heterogeneity could be a contraindicator for inclusion of a meta-analysis. However, as with other studies that experienced high levels of heterogeneity (1), the authors felt that completion of a meta-analysis was appropriate both in consideration of the analysis approach which focused on opioids by context; and in the broader concept of highlighting the importance of further research needed in this space.

The scope of this review was limited to studies focusing on prevalence of prescription opioids; and the misuse of prescription opioids, and other illegal narcotics were not considered. Consequently, this review does not capture the full breadth of opioid issues or harms among SCI populations. A noticeable limitation is the heterogeneity within the SCI populations among the included studies, largely influenced by the diverse participant eligibility criteria, limiting the generalizability of study findings. Data for approximately 70% of studies originated from North America, creating a strong representation for veteran populations and individuals with varying health insurance eligibilities or coverage (Table 1). This largely limits the generalizability of findings for populations outside of North America.

Conclusion

This study identified a pooled prevalence of opioid use among adults with SCI living in the community, demonstrating high heterogeneity across the included study populations. Despite many studies reporting prevalence of opioid use, the wide variation in populations included means that much remains to be elucidated about who uses opioids after SCI, for low long and at what dose. Additionally, opioid context and secondary outcomes were rarely reported, indicating a clear need for additional studies in a population at greater risk of experiencing opioid-related adverse effects.

Disclaimer statements

Funding This review was not funded.

Conflict of interest The authors declare they have no competing interests.

Supplemental data

Supplemental data for this article can be accessed online at https://doi.org/10.1080/10790268.2024.2319384.