This study is the most comprehensive study conducted to date to observe opioid prescribing trends over 19 years in 20 different study sites. We noted marked geographical differences in opioid prescribing with signs of convergence toward 2019. In most countries, there were more incident opioid users than nonincident users, more opioid users were female than male, and opioid prescribing increased with age. The dissimilar trends and patterns suggest varied opioid prescribing approaches and potential differences in opioid access across and within geographical regions, sexes, and age groups.

4.1 Varied but potentially converging opioid prescribing

In our study, multinational opioid prescribing was converging from 2001 to 2019, i.e., decreased in countries with high use at the beginning of the study period and increased in sites with low use. This pattern suggests that the tension between concerns about under-treatment and the risk of opioid-related harms is slowly being resolved by a more middle-ground pattern. Our results align closely with the observation in a study using country-level consumption data, reaffirming that opioid prescribing decreased in historically high-use countries, while increasing in previously low-use settings since the early 2010s [7]. Despite this being an encouraging observation, there were concerns about people being abruptly cut off from long-term pain treatment in countries with a sharp decline [10]. Patients may be at risk of withdrawal or seek opioids from nonregulated sources that are not captured in administrative datasets, especially if supportive interventions, such as psychological and nonpharmacological services, were not available [11]. This challenge has been particularly salient in the USA, given pressures on physicians to reduce prescribing, resulting in a sharp reduction in opioid use relative to other countries [12].

Nonetheless, similar to earlier studies, opioid use remained unbalanced across and within geographical regions over the study period [6, 13, 14]. In Europe, opioid prescribing prevalence ranged from 4.3% in Germany to 19.7% in Iceland, a quadruple difference. East Asia had the lowest regional opioid prescribing prevalence. Sites with low prescribing should examine if barriers to opioid access including poor physical availability and practical accessibility, cultural biases, or restrictive regulations, should be addressed [2]. Conversely, the potentially excessive number of users in countries with high opioid prescribing suggests that a large proportion of patients could be at risk of opioid-related harm [5]. While opioid prescribing in each site is influenced by differences in health systems, opioid availability, pain prevalence, and regulatory policies, given that the human development indexes of our study sites were similar [15], this raises questions about how much variability in clinical opioid prescribing is reasonable and whether or not at least some degree of the identified heterogeneity should be reconciled.

4.2 Incident and nonincident opioid prescribing in different sites

To identify the people at higher risk of opioid-related morbidities and mortality (i.e. those with more frequent opioid exposure), we further classified prevalent users into incident and nonincident users. More people were incident opioid users than nonincident users, except in the USA, South Korea, and Sweden. Use of opioids to manage acute pain, cancer pain, and terminal pain is well-accepted clinically. However, their use for chronic noncancer pain remains controversial as potential harms outweigh the benefits.[16] In particular, nonincident opioid prescribing in South Korea was much higher than in other Asian sites. The sharp decline in nonincident use in North America suggests progress in restraining long-term use, a shift away from long-term use while countries where nonincident use is high and increasing relative to incident use like South Korea are concerning. Between the two US sites, a higher incident opioid prescribing was observed in individuals covered by private insurance than in Medicaid. While the observed differences between private and public reimbursed incident opioid prescribing could in part reflect potential differences in health, the direction of the finding was not as expected, given that Medicaid caters for the population with disability and greater health needs. People on Medicaid generally have poorer health and more chronic pain and would typically be thought to receive more opioids. The lower rates observed in our data suggest that other factors, not captured in administrative claims, may also play a role. There is evidence of racial, socioeconomic, and educational bias in opioid prescribing, with lower-income and less-educated individuals, many of whom are on Medicaid, receiving fewer prescriptions even when presenting with similar pain indications and severity as their privately insured counterparts [17]. Consequently, policies aimed at curbing opioid over-prescribing may have disproportionately impacted Medicaid populations, as reflected in our finding that the greatest average annual absolute reduction in opioid prescribing prevalence occurred among Medicaid recipients.

4.3 Sex and age differences in opioid prescribing

Sex differences in opioid prescribing were consistent across most data sources, where more females used opioids than males. This may be due to differences in pain conditions, pain experiences, and health-seeking behaviors [18,19,20]. Notably, females in the UK were twice as likely to be incident users of opioids than males, a much higher ratio than in other countries. Opioid prescribing also increased with age. The higher prevalence of chronic musculoskeletal and end-of-life conditions may explain the higher opioid prescribing in older adults when compared with other age groups. High pediatric opioid prescribing in the US MarketScan data is also notable, at over six times the pooled average. This aligns with recent US studies showing that opioids remain commonly prescribed to children, especially for dental and postoperative procedures [21]. Other single-country studies have similarly reported higher rates of pediatric opioid prescribing in the USA compared with countries such as Denmark and Norway [22, 23]. By age 18, nearly one in five children have received at least one opioid prescription, raising concerns about the risk of later opioid misuse and unintended prolonged use [23]. The differences in opioid prescribing by sex and age may reflect the different pain management needs in each subpopulation [24,25,26]. Sex-related differences in opioid metabolism, hormones, body composition, and menstrual cycles, may contribute to differences in analgesic effect and safety profile of opioids [24]. Similarly, age- and gene-related pharmacodynamic and pharmacokinetic characteristics affect pain sensitivity, clinical efficacies, and occurrence of adverse events with opioid prescribing [25, 26]. Older adults, for instance, are more susceptible to side effects such as respiratory depression, impaired motor coordination, dizziness, and falls [27].

4.4 Strengths and limitations

This study presents the most comprehensive analysis to date of opioid prescribing by sex, age, and geographical location over 19 years in 20 different study sites. This study has several limitations. Firstly, although the common protocol enabled us to standardize the opioid prescribing and population measures evaluated in this study, heterogeneity of data sources did exist. For instance, the Icelandic register also captured drugs dispensed in nursing homes. This may inflate opioid prescribing rates when compared with a purely outpatient population. Also, opioid administration during hospitalization were often not captured. Secondly, we could not collect information on the indication of use and several opioids may have therapeutic indications beyond pain management. Thirdly, clinical practice differs, and the opioids included per site were different. Finally, we only assessed the number of people using opioids, not the volume of use.

Our results should be interpreted considering the regulatory interventions or guidelines changes during the study period. For example, the European Medicines Agency recommended the withdrawal of dextropropoxyphene in 2009 but the effect on opioid prescribing was largest in France in 2011 [28]. In Denmark, there were considerable media attention and regulatory actions since 2017 to decrease tramadol use and subsequently other opioids [29]. In Finland, there was a nationwide intervention in 2017 for decreasing paracetamol–codeine prescribing in large packages for new patients [30]. Reimbursement status of a medication may also affect the data captured. In Finland, paracetamol–codeine products were out of reimbursement status in the years where the greatest decrease was seen (2001–2008). However, they were still used despite not being captured in the prescription register. In the USA, the Centers for Disease Control and Prevention released key guidelines on opioid prescribing in 2016 amidst the epidemic of opioid overdoses and substance use disorders [31].

4.5 Implications and future research directions

Patients who suffer from pain symptoms and conditions require adequate pain relief while avoiding opioid-related harms such as addiction and overdose. However, decisions on pain management remain complex, particularly for patients who have become established on long-term opioid treatment [32]. Cross-disciplinary collaboration is needed to streamline evidence-based recommendations for different types of pains at different severity levels. To ensure adequate but prudent opioid prescribing, pain assessment should be routinely implemented. Other preventive measures to mitigate unintended misuse and diversion of opioids include setting realistic expectations about pain relief goals, promoting analgesia stewardship, highlighting the addictive properties of opioids, short prescription durations, providing novel delivery devices or routes of administration where appropriate, and follow-up of care after prescription with careful evaluation on the need for treatment continuation [33]. The WHO Analgesic Ladder also recommends first considering nonopioid analgesics where appropriate when administering pain treatment [34].

Our study identified clear differences in opioid prescribing by patient demographics—age, sex, and geographical regions. There is currently a paucity of research on the comparative safety and effectiveness of opioids for pain relief in specific populations, especially females, older adults, and nonCaucasians [27, 35]. Given the biopsychosocial differences that affect both opioid prescribing and responses, extrapolating treatment evidence from general populations may not be appropriate. Future safety and effectiveness evidence on opioid use, stratified into sex, age, and strong and weak opioids, should be generated with comprehensive information on potential influencing factors and categorized by clinical indication and comorbidities to inform safe prescribing in different subpopulations. Comparative studies incorporating biopsychosocial approaches to pain are also needed.