Nonmedical Opioid Use After Short-term Therapeutic Exposure in Children: A Systematic Review

This systematic review was conducted according to methodologic standards defined by Cochrane,²⁷  and the protocol was registered with PROSPERO (identifier CRD42019122681). This study was exempt from ethics approval.

A medical research librarian developed the search strategy, which was peer reviewed by a second librarian. Eleven databases were searched from inception to May 12, 2019 (updated September 4, 2020), 5 of these were key gray literature sources (Search Strategy in the Supplemental Information, Supplemental Tables 5 and 6). Reference lists of relevant and included studies were checked, and primary authors were contacted as necessary. Scopus and Google Scholar were used to conduct a forward citation search based on included studies.

Screening forms were developed and piloted by the review team (M.A., M.P.D., and L.H.). Two reviewers (M.A. and a research assistant or M.P.D) independently assessed all titles and abstracts and potentially relevant full texts for inclusion. Discrepancies were resolved through discussion and with a third reviewer (M.P.D. or L.H.) when needed. Eligibility criteria included the following: (1) publication in English or French, (2) report of quantitative primary research, (3) study participants with short-term (≤14 days) therapeutic exposure to opioids before the age of 18 or during school years (kindergarten to 12th grade), and (4) report of OUD (encompassing the older terms “opioid addiction” and “abuse”)²⁸  or misuse (see Table 1 for definitions). Through discussion and consensus, the review team included one study with short-term exposure for a median of 17 days.²⁹  When the duration of exposure was unclear, studies were included provided there was no definitive language confirming long-term therapeutic exposure to opioids.

Data were extracted by 1 reviewer (M.A.) using a standardized form and verified by a second reviewer (research assistant or M.P.D.). Discrepancies were discussed and resolved by a third party when necessary (M.P.D. or L.H.). Data were extracted on study and population characteristics (eg, setting, age, ethnicity and race), exposure characteristics (eg, duration, dose), risk factors, and outcomes and results (eg, timing, follow-up period, effect estimate).

Methodologic quality of included studies was independently assessed by 2 reviewers (M.A. and L.H. or M.P.D.) using the National Institutes of Health (NIH) Quality Assessment Tool for Observational Cohort and Cross-Sectional Studies.³⁰  Disagreements were resolved through discussion.

Studies were grouped into comparative and noncomparative designs. Odds ratios, risk ratios or relative risks (RRs), and risk differences were extracted from comparative studies or calculated by using Review Manager 5.3 to provide consistent measures across studies. Data were not pooled because of methodologic, clinical, and statistical heterogeneity; as a result, data were summarized narratively.²⁷  Prevalence and incidence rates from the noncomparative studies are presented. Risk factors were extracted as reported by authors and summarized narratively; when available, measures of association were included.

We identified 5219 unique citations, of which 21 studies reported in 22 publications met our inclusion criteria (Fig 1). Reasons for exclusion are described in Fig 1.

Table 2 provides a summary of included studies; details for each study are provided in Tables 3 and 4. On the basis of studies in which sample size, sex distribution, and age were reported, 49 944 602 children and youth were analyzed (young adults were also included in the total population count when the youth population could not be separated)³¹ ; 48.4% were female, and the mean age was 15.7 years. Fifteen studies were conducted exclusively in pediatric or high school populations (including 18-year-olds),^{18,21–23,29,32–42}  whereas 6 studies included both children or youth and adults.^31,43–47 

A majority of the included studies were assessed as having lower methodologic quality; however, they were largely designed to address different types of research questions. Quality scores ranged from 6 of 14 to 13 of 14 (see Tables 3 and 4 and Supplemental Table 7). Common methodologic limitations were self-report of exposures and outcomes, no specification of the duration of exposure, and cross-sectional designs.

In 1 of the 5 comparative studies, the association between short-term therapeutic exposure and opioid abuse was explored⁴⁴ : Individuals (including adults and youth; mean age 21.8 [SD: 2.4] years) who were prescribed opioids were significantly more likely to have a diagnosis of abuse within 365 days than those who were not exposed (adjusted absolute risk difference 5.3% [95% confidence interval (CI) 5.0%–5.7%]; RR^* 15.1 [95% CI 12.4–18.3]). Within the opioid exposed cohort, those aged 22 to 25 were less likely to be diagnosed with opioid abuse at 1 year than those aged 16 to 18 years (adjusted odds ratio [aOR]^† 0.8 [95% CI 0.7–1.0]). Asian American race was associated with lower odds of abuse compared with White race (aOR^† 0.3 [95% CI 0.2–0.6]); female sex (aOR^† 11.5 [95% CI 9.4–14.8]) and previous nonopioid substance abuse (aOR^† 4.5 [95% CI 3.4–5.9]) were associated with higher odds of abuse.

The other 4 studies did not specify the duration of therapeutic exposure to opioids; however, each reported an association between therapeutic opioid exposure and future nonmedical use of opioids. McCabe et al²¹  found that adolescents who reported past-year medical exposure to opioids in year 1 were significantly more likely to have subsequent nonmedical opioid use in year 2 compared with those with no medical exposure (RR^* 2.36; 95% CI 1.28–4.37). McCabe et al¹⁸  showed that those who had any medical use of opioids by age 18 were significantly more likely to have past-year nonmedical opioid use at age 35 compared with those who had no medical or nonmedical use at age 18 (RR^* 1.83 [95% CI 1.15–2.89]; aOR^† 1.74 [95% CI 1.10–2.76]). Miech et al³²  reported a significant association between receipt of a prescription opioid by grade 12 and future opioid misuse by age 23 (RR^† 1.33; 95% CI 1.04–1.7); this association was strongest among individuals generally considered to be at lower risk for opioid misuse (ie, those with little or no previous drug use and who disapproved of marijuana use). Osborne et al³³  found that youth who indicated that they had lifetime medical use of prescription opioids had an increased risk of past-30-day nonmedical opioid use (ie, use of their own opioids in a manner other than directed or of opioids from another person) (RR^* 2.72; 95% CI 1.76–4.22) (Table 3).

A majority of the non-comparative studies reported prevalence of nonmedical opioid use after past therapeutic exposure of unknown duration; 2 of the 16 studies indicated that the study sample had short-term therapeutic exposure to opioids.^29,34,41  Bell et al³⁴  found that within a cohort of adolescent trauma patients, 11% of patients discharged with a prescription for opioids received an opioid antagonist injection and 8% were given an overdose diagnosis over the 5-year study period. Chung et al²⁹  identified 437 opioid-related adverse events (emergency department visit, hospitalization, or death) of 1 362 503 prescriptions in children aged 2 to 17 years (cumulative incidence 38.3 of 100 000 prescriptions); 71 of the events (0.005% of total prescriptions dispensed and 16% [71 of 437] of total adverse events) were attributed to abuse or self-harm (all in youth aged 12–17).

Across noncomparative studies, nonmedical use prevalence rates among those with medical exposure to opioids ranged from 0.005% to 36%, with a median of 27.8% (interquartile range [IQR]: 20.1%–30.7%; Table 4). However, this corresponds to a wide range of circumstances. In 3 studies, nonmedical use of opioids among those who had been prescribed opioids (20.1%–36%) was measured^22,23,35 ; in 1, the authors assessed the proportion of past-year nonmedical misusers whose only source of misuse was their prescription (14.4%).³⁶  Wei et al⁴⁵  examined 10-year trends in incident cases of OUD or overdose and reported annual rates of individuals who had received a prescription in the year before (22.6%–34.1%). There was a significant decrease in the proportion of youth receiving prescription opioids before being diagnosed with OUD or overdose between 2006 and 2016 (P value for trend = .001). The lowest value (0.005%) represents adverse events (emergency department visit, hospitalization, or death) as a result of abuse or self-harm from opioid prescriptions dispensed over a 5-year follow-up period.²⁹  Bell et al³⁴  also reported on opioid overdoses, with prevalence rates of 7.6% for an overdose diagnosis and 10.8% for an opioid antagonist injection over a 5-year period after an opioid was prescribed. These values can be contrasted with prevalence rates among samples including individuals who were and were not medically exposed to opioids (median 2.6%; IQR 1.3%–4.3%).

Age, sex, race and ethnicity, and previous substance abuse were commonly explored risk factors within the studies we included; however, results were inconsistent across and sometimes within studies. The impact of age was most constant, with 3 studies reporting that older adolescents were at higher risk of nonmedical opioid use than younger children and young adults.^29,34,44  Two studies determined that female participants were more likely to have nonmedical use than male participants,^33,36  although a difference was not found in 6 studies.^{23,29,31,32,34,42}  In 4 studies, authors examined previous substance abuse as a risk factor for nonmedical prescription opioid use, but the studies differed markedly in variables measured and analytic approaches used. Miech et al³²  determined that students who were at lower risk for opioid misuse (ie, had the least experience with illegal drug use and strongly disapproved of the use of marijuana) had the strongest association between prescription opioid use and nonmedical use after high school, whereas Schroeder et al⁴⁴  found that those with previous nonopioid substance abuse were at higher risk of opioid abuse. Bell et al³⁴  found no association with a positive drug or alcohol screen result on admission. All studies examining race and ethnicity found a significant effect, although findings differed: 3 studies found that White participants were more likely to misuse opioids than Asian American participants,⁴⁴  African American participants,²³  and racial and ethnic minority individuals,³²  whereas 2 others found that African American participants were more likely to misuse than White participants.^34,38  Currently, the evidence related to risk factors for nonmedical opioids use or OUD after short-term therapeutic exposure is unclear (Fig 2, Supplemental Table 8).

There is limited evidence to determine if short-term therapeutic exposure to opioids in childhood is definitively associated with future nonmedical opioid use or development of an OUD; however, this review suggests a link between lifetime therapeutic opioid use (unknown duration) and nonmedical opioid use. The existing evidence on risk factors for nonmedical opioid use or OUD after short-term therapeutic exposure is unclear; however, older adolescents with short-term therapeutic exposure during their lifetime may be at higher risk of nonmedical use than younger children.

There is a great deal of emerging research related to the harms associated with the therapeutic use of opioids; however, we were unable to clearly answer our original question. Current research has been focused on the use of opioids in adult populations, and the evidence suggests that a prescription for opioids for a long duration puts adults at risk for future nonmedical opioid use.⁴⁸  Data specific to pediatrics are scarce and, to date, have largely relied on cross-sectional, self-reported evaluations of usage patterns. Prospective studies are urgently needed in the area of acute pain to help inform decisions in emergency departments, in ambulatory clinics, and after surgeries.

The study by Miech et al³²  has been widely cited and has generated much controversy because the ones at most risk for opioid misuse after therapeutic exposure were those who would typically be least expected of being at risk: those with little to no history of previous drug use and with strong disapproval of illegal drug use. Miech et al³²  acknowledge that their “results do not support legitimate opioid prescription use, by itself, as a major contributor to chronic opioid misuse.”³²  Hence, health care providers should not use these conclusions as a basis for limiting prescribing when opioids are needed, but rather they should ensure some risk screening (ie, for preexisting OUD at minimum), judicious prescribing, and disposal mechanisms are in place. With the current stigma around opioid use, the undertreatment of pain and its associated short- and long-term consequences cannot and should not be ignored.⁴⁹ 

A broad framework for evidence-based prescribing post surgery identifies 3 key areas of responsible opioid prescribing: recognizing that prescription opioid use is associated with short- and long-term risks, optimizing pain management in the perioperative phase by using alternative nonopioid medication when possible, and educating families and patients about managing pain and using opioids safely before and after surgery.⁵⁰  Although evidence-based guidelines and validated risk screening tools for the use of opioids for children of all ages with acute pain are lacking, other resources may help guide decision-making on the use of opioids, such as Health Quality Ontario guidelines for prescribing opioids for acute pain for ages 15 and older⁵¹  and the World Health Organization’s guidelines for prescribing opioids for chronic pain.⁵²  The use of resources by caregivers and prescribers (eg, brochures developed by the Institute for Safe Medication Practices Canada)⁵³  may help navigate the challenging conversations with patients and families around opioid risks and use.

Our systematic review captures studies published up until September 4, 2020, and as a result, we would not take into consideration the evidence based on any newly published studies. For example, we did not include the 2021 study by Hadland et al⁵⁴  that reported incidence rates of overdose and OUD ranging from 0.1% to 0.3% among youth and adults who had acutely painful conditions and were therapeutically exposed to opioids. Nonetheless, these new findings would fall within the nonmedical use prevalence and incidence rates that we observed in our sample, which ranged from 0.005% to 36%.

This review provides a rigorous and comprehensive synthesis of the current literature; however, there were few studies that directly addressed our research question. In addition, there were a number of limitations with the existing evidence. First, study samples might not have been opioid naïve. Short-term exposure to opioids was evaluated in only 3 studies, of which only 1 had a nonexposed comparison group.⁴⁴  None of these studies were able to fully control for opioid naiveté. Studies that did not specify the duration of exposure were also included, unless there were clear indications of long-term exposure, such as long-term use of opioids for chronic conditions; therefore, our sample may have included some participants with longer-term exposure or excluded those who had chronic pain and were opioid naïve. Many studies were focused on past-year medical exposure rather than lifetime exposure and did not have measures in place to ensure that therapeutic exposure preceded opioid misuse at any point during one’s lifetime.

A second limitation is that exposure might have occurred during adulthood. Many of the surveys were conducted in high schools, where some of the participants were already 18 years old and therefore could have been exposed therapeutically in adulthood.^{18,21–23,32,33,35,36,40}  Similarly, in the most robust comparative study, the authors reported on associations for the entire sample, which included youth and adults and was not broken down by age group.⁴⁴  A third challenge in synthesizing this literature was inconsistency in outcome measurement and terminology. Some studies only captured past-year or past-30-day misuse, potentially underreporting the prevalence of opioid misuse. Study authors also reported a number of outcomes, such as addiction, abuse, overdose, self-harm, nonmedical use, medial misuse, diversion, opioid-specific substance use disorder, and OUD. Although a majority fall under the umbrella terms “misuse” and “nonmedical use,” they capture variable information. Fourth, a majority of the studies were considered of poor quality because of their cross-sectional design, reliance on self-reporting, inadequate measures for ensuring short-term therapeutic exposure preceded outcome, lack of controlling for confounders, and lack of a comparator group.

Finally, our included studies may be limited with respect to their generalizability. Some studies using clinical samples relied on commercial⁴⁴  and/or Medicaid^29,45  insurance claims. Data from other service providers (such as emergency medical services and the police) or opioid-related adverse events in which medical attention was not sought were not routinely reported. In addition, the risk of misuse may vary on the basis of location of exposure (eg, within a controlled hospital setting versus going home with prescribed opioids). Risk of nonmedical opioid use after short-term exposure in a controlled setting was not reported in any of the included studies.

To assess potential harms from short-term therapeutic opioid use, future research is needed in which studies reliably assess and report duration of opioid exposure, dosage, preparation type, and setting; compare those exposed with an unexposed group; control for opioid naiveté at baseline (ie, no previous long- or short-term therapeutic and/or nontherapeutic exposure); differentiate between reasons for nonmedical use (eg, for pain or pleasure); rigorously assess risk factors; and evaluate the potential misuse associated with exposure to different opioids because some may put patients at higher risk of misuse than others.⁵⁵  Existing administrative data may be more effectively used via record linkage to help construct more comprehensive data sets. The use of artificial intelligence, including machine learning, may also provide an opportunity to identify risk factors for misuse.⁵⁶ 

There is limited evidence to determine if short-term exposure is specifically associated with OUD or nonmedical use of opioids; however, a number of studies suggest an association between lifetime therapeutic opioid use and nonmedical opioid use. There is also limited evidence to identify risk factors for OUD or nonmedical use after short-term therapeutic exposure. Rigorous studies are needed to examine the association between short-term therapeutic exposure in children and youth and nonmedical opioid use or the development of an OUD. Until the risks are more clearly defined, it is recommended that before prescribing short-term opioids, health care providers carefully consider the risks, benefits, and alternatives and educate patients and caregivers about safe and appropriate use and when to seek reassessment.

Mr Ben Vandermeer provided assistance with interpreting statistical data and validating the statistical analysis. Mr Daniel DaRosa assisted as a second reviewer for screening and data extraction. Ms Jocelyn Shulhan-Kilroy assisted with data verification. Ms Linda Slater developed and ran the initial search, and Ms Tara Landry peer reviewed the search and ran the update. Ms Lana Atkinson and Ms Robin Featherstone also supported the development of the initial search. Ms Diana Keto-Lambert and Ms Sholeh Rahman conducted the gray literature searches.

aOR

adjusted odds ratio

CI

confidence interval

IQR

interquartile range

NIH

National Institutes of Health

OUD

opioid use disorder

RR

risk ratio or relative risk