Article Text
PDF
Abstract
Objective To identify individual and initial prescription-related factors associated with an increased risk for opioid-related misuse, poisoning and dependence (MPD) in patients with non-cancer pain.
Methods Cohort study linking several databases covering 5 million inhabitants of the region of Valencia, Spain, including all adults initiating prescription opioids in the period 2012–2018. To ascertain the association between the characteristics of the initial prescription choice and the risk of opioid MPD, we used shared frailty Cox regression models. We additionally considered death as a competing risk in sensitivity analyses.
Results 958 019 patients initiated opioid prescription from 2012 to 2018, of which 0.13% experienced MPD. Most patients were prescribed tramadol as initial opioid (76.7%) followed by codeine (16.3%), long-acting opioids (6.7%), short-acting opioids (0.2%) and ultrafast opioids (0.1%). Initiation with ultrafast (HR 7.2; 95% CI 4.1 to 12.6), short-acting (HR 4.8; 95% CI 2.3 to 10.2) and long-acting opioids (HR 1.5; 95% CI 1.2 to 1.9) were associated with a higher risk of MPD when compared with tramadol. Initial prescriptions covering 4–7 days (HR 1.3; 95% CI 1.0 to 1.8), 8–14 days (HR 1.4; 95% CI 1.0 to 1.9), 15–30 days (HR 1.7; 95% CI 1.2 to 2.3) and more than one a month (HR 1.8; 95% CI 1.3 to 2.5) were associated with more MPD risk than initial prescriptions for 1–3 days. Treatments with >120 daily morphine milligram equivalents (MME) increased MPD risk (vs <50 MME, HR 1.6; 95% CI 1.1 to 2.2). Main individual factors associated with increased risk of MPD risk were male sex (HR 2.4; 95% CI 2.1 to 2.7), younger age (when compared with patients aged 18–44 years, patients aged 45–64 years, HR 0.4; 95% CI 0.4 to 0.5; patients aged 65–74 years, HR 0.4; 95% CI 0.3 to 0.5 and patients aged 75 years old and over, HR 0.7; 95% CI 0.6 to 0.8), lack of economic resources (2.1; 95% CI 1.8 to 2.5) and registered misuse of alcohol (2.9; 95% CI 2.4 to 3.5). Sensitivity analyses yielded overall comparable results.
Conclusions Our study identifies riskier patterns of opioid prescription initiation for non-cancer indications, as well as patient subgroups with higher risk of misuse, poisoning and dependence.
- adverse events, epidemiology and detection
- health services research
- medication safety
- pain
- pharmacoepidemiology
Data availability statement
Data may be obtained from a third party and are not publicly available. The datasets presented in this article are not readily available because legal restrictions on sharing the dataset apply as regulated by the Valencia regional government by means of legal resolution by the Valencia Health Agency (2009/13312), which forbids the dissemination of data to third parties (accessible at: http://www.san.gva.es/documents/152919/157920/resolucionsolicituddatos.pdf). On request, authors can allow access to the databases in order to verify the accuracy of the analysis or the reproducibility of the study. Requests to access the datasets should be directed to Management Office of the Data Commission in the Valencia Health Agency (email: solicitud_datos@gva.es; telephone numbers: +34 961-928207; +34 961-928198).
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
- adverse events, epidemiology and detection
- health services research
- medication safety
- pain
- pharmacoepidemiology
WHAT IS ALREADY KNOWN ON THIS TOPIC
Studies in the USA have shown that some patient factors, such as young age or low socioeconomic status, are associated with a higher risk of opioid-related misuse, poisoning and dependence (MPD) in patients prescribed opioids for non-cancer pain, but there is a lack of data on patient and prescription-related factors associated with MPD in Europe.
WHAT THIS STUDY ADDS
In a population-based cohort study including 958 019 patients prescribed opioids, prescription-related factors (initiating with fentanyl, long-acting oxycodone and short-acting buprenorphine, using benzodiazepines, gabapentinoids and antipsychotics concurrently, initiating therapy with a daily morphine milligram equivalents >120 or for >2 weeks) as well as patient characteristics (male sex, young age, use of tobacco and alcohol, lack of resourcesand a series of comorbidities) were associated with increased risk of opioid MPD.
HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY
By identifying patients at elevated risk for MPD as well as potentially hazardous prescription patterns, our findings may provide guidance for healthcare managers and clinicians towards a safer use of opioids in patients with non-cancer pain.
Introduction
Opioids provide small but significant benefits versus placebo in treating non-cancer pain,1–3 but evidence of their relative effectiveness in clinical practice remains inconclusive, whereas their use is unequivocally associated with a risk for serious harm.4–7 In the last 20 years however, use of opioids for non-cancer pain has increased in Europe and other regions,8–14 resulting in exponential growth of the number of patients exposed to potential harms, such as misuse, poisoning and dependence (MPD). In fact, rates of opioid-related overdose deaths have been rising globally in the last decade.7 15–18
Several studies in the USA have shown that patient-related factors, such as past or current substance misuse, mental health disorders, younger age or male sex, increase the risk of opioid MPD among patients using prescription opioids.19 20 However, evidence on prescription-related factors to MPD is scarce; a recent study conducted in Oregon with a cohort derived from a population of 3.6 million found that the choice of opioid and the use of concomitant benzodiazepines were associated with opioid MPD.21 In Europe, opioid use-related adverse events such as hospitalisations, overdose, dependence and deaths have increased in the last years, but to a lesser extent than in the USA,22–24 arguably thanks to the characteristics of the healthcare systems in place,25 and data on the factors associated with opioid-related MPD are lacking.
In Spain, the yearly volume of prescription opioids and number of patients with at least one prescription doubled, whereas the intensity of treatments in terms of morphine milligram equivalents (MME) tripled from 2010 to 2018.26 Also, opioids were initiated for a vast array of non-oncological indications, and a significant number of patients were exposed to potentially high-risk patterns of initiation, such as treatments lasting >14 days, treatments surpassing 50 daily MMEs, initiating therapy with long-acting opioids or hazardous overlapping with other therapies.27 Whether these patterns of prescription are associated with an increased risk of MPD remains unknown.
In this study, we aimed to assess the association of the choice of initial opioid prescription with opioid-related MPD outcomes, as well as to identify individual patient factors and characteristics of the initial prescription associated with an increased risk of MPD.
Methods
Design, setting and data sources
In this retrospective cohort study, we combined several population-wide databases from the Valencia Health System Integrated Database (VID), covering a population of 5 million inhabitants. The Valencia Health System is a universal, single-provider, free at the point of use, public health system, which is organised into 24 health areas (the administrative and territorial management units) that make up the healthcare provision network in the region. VID is a set of publicly owned, population-based healthcare, clinical and administrative electronic databases in the region of Valencia in Spain that can be linked by means of a single personal identification number and provides comprehensive information for the region’s population since 2008. VID includes sociodemographic and administrative data (sex, age, nationality) as well as healthcare information such as diagnoses, procedures, laboratory data, outpatient pharmaceutical prescriptions and dispensing (including brand and generic name, formulation, strength and dosing schedule/regime), hospitalisations, mortality, healthcare utilisation and public health data.28 Both specialists and primary care physicians can prescribe opioids in Spain, and dispensing is subject to a tight, formal control and registry. Treatments for one specific medication in VID may be established for as long as 12 months and may be extended for longer periods, but prescriptions are commonly refilled on a monthly basis or shorter (eg, one treatment for 6 months may include, for instance, 6 monthly prescriptions). Finally, clinical guidance in Spain is aligned with general, internationally shared recommendations on the use of prescription opioids for pain management, such as the adherence to the WHO analgesic ladder and the endorsement of the use of opioids only after employing non-opioid alternatives, the selection of short-acting morphine as third-step treatment of choice and the recommendation of using the lowest effective potency and for the shortest possible time.29–35
Participants
We included all patients aged 18 years old and over who received an initial opioid prescription from 1 January 2012 to 31 December 2018 for non-cancer pain. Patients prescribed opioids for cancer pain or with active cancer diagnostic codes in the electronic medical record of the region were excluded (see online supplemental table 1 for codes employed). Patients with a registered opioid-related MPD diagnosis before the index date were also excluded. Finally, people without general healthcare coverage (mainly certain Spanish government employees whose prescriptions are reimbursed by mutual societies for civil servants and are thus not included in the pharmacy databases of the Valencia Health System), and patients not registered in the municipal census (non-residents or temporary residents), were excluded because of limitations on follow-up (figure 1).
Supplemental material
Flow chart. *Patients could initiate more than once during the study period. Only first initiations were considered for analysis (n=1 091 769). Subsequent initiations (n=594 277) were not included in the analyses.
Initial opioid prescription
The index date (date of the initial prescription) was defined as the date when an opioid treatment was prescribed, without having an opioid prescription in the previous 6 months. This time spam sits in the middle of the range of possible definitions found in the literature to define naïve opioid user, and it appears reasonable for medications that may be taken chronic, but intermittently, leading to separate episodes of treatment. In this way, patients could have more than one initiation in the period (which would constitute different episodes of care), only the first was considered for analysis (figure 1). Opioids initially included were classified as: tramadol, codeine, long-acting (morphine extended and controlled-release, transdermal fentanyl, oxycodone extended-release, transdermal buprenorphine, hydromorphone extended-release, tapentadol extended-release), short-acting (standard-release formulations of morphine, oxycodone or buprenorphine) and ultrafast opioids (transmucosal fentanyl formulations, see online supplemental material S1). Low-dose codeine formulations containing <30 mg per dose unit were excluded as these are mainly indicated for the treatment of influenza and cold symptoms in our setting. Other opioids excluded from analysis were methadone (whose use is restricted to drug addiction programmes in Spain), pethidine and pentazocine due to marginal volume.
Variables
We included patients’ sociodemographic and lifestyle variables (age, sex, nationality, income range based on pharmaceutical co-payment status information, alcohol misuse and tobacco use), comorbidities at baseline (hypertension, ischaemic cardiomyopathy, heart failure, congestive obstructive pulmonary disease (COPD), dementia, depression, diabetes, renal and liver disease) and presence of surgical procedure in the 30 days before the index date (see online supplemental material S1 for International Classification of Diseases (ICD) and Anatomical Therapeutic Chemical codes used for covariates and drugs included in the study, and online supplemental material S2 for detail on the identification of surgical procedures).
Treatment initiation was characterised using the following variables: choice of opioid, duration (mean, median and range), daily MME (mean, median and range), indication of use and percentage of patients with concomitant use (or overlap) of benzodiazepines, gabapentinoids or antipsychotics. Overlap was ascertained when patients received at least one prescription for at least one of these drugs with days’ supply overlapping with the index date (date of initial opioid prescription). To estimate daily MME of the initial prescription, we used prescription information to identify milligrams prescribed and days’ supply (using the dosing regimen specified in the prescription), and an MME conversion table (see online supplemental material S3). Indication of use was classified as: musculoskeletal (MSK) pain (back pain, osteoarthrosis, arthritis, other MSK), gastrointestinal disorders (mainly dental pain), respiratory indications (including diagnoses of influenza and fever) and miscellaneous. Miscellaneous group includes several categories of pain diagnoses that individually account for a small percentage of patients (see online supplemental material S4 for detail on ICD codes included in each indication category and the more prevalent codes under each category). Finally, we included the health area as a contextual variable to account for the potential influence of the area of residence in MPD risk.
Outcomes
Main outcomes were hospitalisations, emergency department visits or ambulatory visits with a first diagnosis of opioid misuse, poisoning or dependence (see online supplemental material S1). Patients were followed from the index date until the first MPD diagnosis, death or end of study (31 December 2018). Only the first MPD diagnosis was considered for analyses. Events occurring on the index date (n=83) were not considered.
Analysis
First, we described baseline patient characteristics and the characteristics of the initial prescription, overall and per choice of initial opioid (tramadol, codeine, long-acting, short-acting and ultrafast opioids). P values were estimated using χ2 for categorical variables and analysis of variance for continuous variables. For continuous covariates with skewed distribution, the Kruskal-Wallis test was used. Second, we estimated the unadjusted incidence rates by 1000 person-years of first MPD, overall and for each category of opioid. We also plotted Kaplan-Meier survival curves over the observation window of 7 years stratified by initial opioid choice. Third, we used an intention-to-treat approach to ascertain the association between initial prescription choice and the risk of opioid MPD. For this purpose, we used shared frailty Cox models (with gamma-distributed heterogeneity) including all individual and prescription initiation variables available as fixed effects and the healthcare department as random effect. Initiation with tramadol was selected as reference group due to its relative volume. Individual baseline variables included were age (reference group: 18–44 years), sex (reference group: male), nationality (reference group: Spanish), income range (reference group <€18.000 per year), alcohol misuse and tobacco use, comorbidities and presence of surgical procedure in the 30 days before the index date (introduced as categorical variables). Therapy initiation variables included were daily MME (reference group: <50 MME), duration (reference group: 3 days or less), indication of use (reference group: back pain) and risk of overlap (introduced as categorical variable). Fifth, we repeated the analysis of the association between the initial choice of opioid and MPD per specific drug. Sixth, we performed sensitivity analysis. We excluded small initial treatment groups (short-acting and ultrafast opioids), and we ran a competing risk regression analysis (using the Fine and Gray method) considering mortality as the competing event. Two-sided p values of <0.05 were considered significant. Analyses were performed using STATA V.14 and R V.3.6.0. The investigators were granted access to the databases used to create the study population and performed several quality checks (consistency assessments, range and logic checks) before the linkage of the databases using a pseudonymised single identification number. The study was reported using the RECORD-PE checklist for observational studies using routinely collected health data.
Results
After applying all inclusion and exclusion criteria, a total of 958 019 patients initiated prescription opioids from 2012 to 2018, of which 1289 experienced MPD after the index date (0.13%; figure 1). Mean age was 56.5 years old, most patients were female (58.7%) and Spanish (83.9%), with a yearly income of €18 000 or less (76%). Most prevalent comorbidities were hypertension (39.9%), depression (15.7%) and diabetes (15.4%), and 11.8% had registered tobacco use (table 1). Median follow-up was 3.8 years (IQR 1.9, 5.6).
Patient characteristics at baseline, overall and per type of opioid, n (%)
Most patients were prescribed tramadol as initial opioid (76.7%) followed by codeine (16.3%) and long-acting opioids (6.7%). Short-acting (0.2%) and ultrafast opioids (0.1%) were less used. Median duration was 9 days, ranging from 5 days for codeine to 29 days for long-acting opioids (p<0.001); 30.1% of initiations lasted for >2 weeks. Median daily MME was 15.0 mg, ranging from 13.3 mg for tramadol to 41.4 mg for long-acting opioids (p<0.001), and 2.9% patients were initiated with a daily MME of 90 mg or higher. MSK pain was the most common cause of opioid initiation (65.1%), followed by a miscellaneous group of conditions (17.4%, see online supplemental material S3) and respiratory conditions (13.0%, see online supplemental material S4 for a detail of the most prevalent diagnostic codes used under each indication category). Overlapping with benzodiazepines, gabapentinoids and antipsychotics was observed in 24.7%, 4.6% and 2.4% of initiations, respectively (see table 2).
Characteristics of the initial prescription, overall and per type of opioid, n (%)
The overall unadjusted rate of opioid MPD was 0.36 (95% CI 0.34 to 0.38) per 1000 person-years, 0.18 (95% CI 0.15 to 0.22) for codeine, 0.37 (95% CI 0.34 to 0.39) for tramadol, 0.79 (95% CI 0.68 to 0.92) for long-acting opioids, 3.76 (95% CI 1.79 to 7.89) for short-acting opioids and 3.9 (95% CI 2.30 to 6.84) for ultrafast fentanyl (see figure 2 and online supplemental material S5 for Kaplan-Meier survival curves). In adjusted analysis, when compared with tramadol as initial prescription, initiation with ultrafast (HR 7.2; 95% CI 4.1 to 12.6), short-acting (HR 4.8; 95% CI 2.3 to 10.2) and long-acting opioids (HR 1.5; 95% CI 1.2 to 1.9) were associated with a higher risk of MPD, whereas starting with codeine was associated with a lower risk (HR 0.6; 95% CI 0.5 to 0.8; see figure 2 and online supplemental material S6). Effect estimates for groups with low event counts (ultrafast and short-acting groups, or for analyses per specific drug) should be interpreted with caution due to low event counts. In analyses per specific drug, ultrafast fentanyl (HR 7.2; 95% CI 4.1 to 12.6), long-acting morphine (HR 3.4; 95% CI 1.3 to 9.1), fentanyl (HR 2.7; 95% CI 2.0 to 3.6) and oxycodone (HR 1.5; 95% CI 1.1 to 2.0), as well as short-acting buprenorphine (HR 23.6; 95% CI 8.8 to 63.3) and oxycodone (HR 4.5; 95% CI 1.1 to 18.2) were associated with an increased risk versus tramadol (see figure 3 and online supplemental material S7).
Incidence rates for misuse, poisoning and dependence (MPD) per type of opioid and forest plot of the association between type of drug and MPD.
Incidence rates for misuse, poisoning and dependence (MPD) per specific drug and forest plot of the association between specific opioids and MPD.
When compared with initial prescriptions covering 1–3 days, initiations for 4–7 days (HR 1.3; 95% CI 1.0 to 1.8), 8–14 days (HR 1.4; 95% CI 1.0 to 1.9), 15–30 days (HR 1.7; 95% CI 1.2 to 2.3) and for >1 month (HR 1.8; 95% CI 1.3 to 2.5) were associated with more risk. Patients initiating treatment with a daily MME higher than 120 mg were at higher risk of MPD when compared with treatments with <50 mg MME (HR 1.6; 95% CI 1.1 to 2.2). When compared with back pain, patients using opioids for other MSK pain (HR 1.5; 95% CI 1.2 to 1.8) and miscellaneous (HR 1.4; 95% CI 1.2 to 1.7) were exposed to a higher risk. Using concomitant antipsychotics (HR 2.7; 95% CI 2.2 to 3.2) benzodiazepines (HR 2.1; 95% CI 1.8 to 2.4) and gabapentinoids (HR 1.7; 95% CI 1.5 to 2.1) when initiating opioid therapy was associated with an increased risk of opioid-related MPD (see online supplemental material S6). Regarding contextual factors, the health area was associated with the risk of MPD (frailty variance=0.10, p=0.04).
Patient factors associated with a higher risk of MPD were male sex (HR 2.4; 95% CI 2.1 to 2.7), younger age (when compared with patients aged 18–44 years, patients aged 45–64 years, HR 0.4; 95% CI 0.4 to 0.5; patients aged 65–74 years, HR 0.4; 95% CI 0.3 to 0.5 and patients aged 75 years old and over, HR 0.7; 95% CI 0.6 to 0.8), lack of economic resources (HR 2.1; 95% CI 1.8 to 2.5), registered misuse of alcohol (HR 2.9; 95% CI 2.4 to 3.5) and tobacco use (HR 1.3; 95% CI 1.1 to 1.5) as well as several comorbidities: liver disease (HR 1.5; 95% CI 1.3 to 1.8), depression (HR 1.4; 95% CI 1.2 to 1.6), COPD (HR 1.5; 95% CI 1.3 to 1.8), heart failure (HR 1.6; 95% CI 1.3 to 2.1), ischaemic cardiomyopathy (HR 1.2; 95% CI 1.0 to 1.5) and kidney disease (HR 1.4; 95% CI 1.1 to 1.7, see online supplemental material S6).
In sensitivity analyses, considering death as a competing risk resulted in a downshift of risk estimates for ultrafast (HR 4.5; 95% CI 2.2 to 9.5), short-acting (HR 3.3; 95% CI 1.4 to 7.8) and long-acting opioids (HR 1.4; 95% CI 1.1 to 1.8, see online supplemental materials S8 and S9). Excluding smaller groups (patients with an initial prescription of ultrafast or short-acting opioids) yielded comparable results to of our main analysis (see online supplemental materials S10 and S11).
Discussion
Summary
In this study, we identified several prescription-related characteristics and patient risk factors that were associated with a higher risk of MPD when initiating prescription opioids. These findings may help healthcare managers to address riskier patterns of prescription and may be helpful as well for clinicians to identify, provide counsel and monitor opioid-naïve patients who may be at higher risk of harm when prescribed opioids.
Prescription factors
With regard to the choice of initial opioid, when compared with tramadol patients initiating with ultrafast fentanyl, long-acting and short-acting formulations were exposed to a higher risk of MPD. In exploratory, adjusted analyses by specific drug, for short-acting opioids buprenorphine and oxycodone were associated with a significantly increased risk. The magnitude of the association for buprenorphine suggests that there may exist a pattern of off-label use of monotherapy short-acting buprenorphine as a treatment for opioid-dependence, when its use is only authorised in Spain for the treatment of pain. With regard to long-acting and ultrafast initiations, both patterns of use warrant investigation. Clinical guidelines in our setting recommends against the use of ultrafast fentanyl for non-cancer pain,29–32 and different interventions have been implemented in our country and our region to reduce its use; however, the latter have had mixed results,36 resulting in a small number of patients still being exposed to a greater risk of MPD. Starting opioid therapy with long-acting formulations may be, per se, a questionable pattern of use, as the WHO analgesic ladder recommends initiating opioid therapy with either codeine or tramadol (weak opioids). In our setting and based on our results, treatment initiation with long-acting fentanyl and long-acting oxycodone should be closely monitored and reviewed, and alternative therapeutic options may be considered. On the other hand, codeine was associated with a lower risk of MPD. This association may be mediated by the particular use of codeine in our country by healthier and younger patients, and in mild pain or non-pain conditions. The patterns of use of opioid medications for mild respiratory-related conditions, such as colds and influenza (96% of which are fixed-dose combinations of codeine or tramadol with paracetamol or ibuprofen, commercial presentations that mimic and appear to be used as substitutes to traditional paracetamol combinations with other drugs such as antihistamine or decongestive agents, that in turn are massively used in our setting to treat mild respiratory symptoms), could constitute potential inappropriate care and warrant further investigation.27 Also, we previously showed in this same population that, when compared with tramadol, patients initiating with short-acting, long-acting and ultrafast opioids were more likely to be older and have more comorbidities.27 In this sense, and despite we adjusted for observed confounders, unmeasured differences in individual characteristics may still be affecting our risk estimates.
Treatments prescribed for more than >2 weeks appeared to be riskier than shorter therapies. In fact, longer initial prescription is associated with an increased risk of long-term opioid use, which in turn leads to a higher risk of MPD.37 In this sense, guidance on opioid use in non-cancer pain recommends establishing treatment as short as possible and to avoid chronic opioid prescription.29–35 However, the observed patterns of initiation in our setting, with almost a third of initial treatments prescribed for more than >2 weeks, a notable proportion of initiations with long-acting opioids, and a large mean duration of initiations with tramadol or long-acting opioids, suggest otherwise.
Patients initiating with a daily MME >120 MME had also more risk of MPD than patients using <50 daily MME, which is again consistent with clinical recommendations preventing from exceeding that threshold.29–35 Even if some dosing categories were not significant, our results suggest that risk of MPD is dose dependent. Finally, appropriateness of prescription should be evaluated in the case of hazardous medication overlapping, notably for benzodiazepines: one in four patients started opioids while using benzodiazepines.
Individual factors
With regard to the characteristics of patients associated with MPD, even if some clinical conditions were associated with MPD risk, socio-economic factors such as male gender, young age, registered alcohol use or lack of economic resources appeared to play a more important role than clinical conditions. On the other hand, non-European population had a lower risk of MPD (when compared with Spaniards), as well as people earning €18 000–100 000 a year, when compared with those earning <€18 000; however, these estimates should be interpreted with caution as factors such as age may be mediating the association between these variables and outcomes. Even if we lack some potentially relevant individual risk factors such as illicit drug use or use encouraging social or family environments, these findings are consistent with the literature on individual factors associated with opioid MPD19–21 and may provide guidance for identification of naïve patients at a higher risk when initiating opioid therapy in daily practice in our setting.
Limitations
Our study is subject to some limitations. First, the VID databases gather several real-world clinical practice data and contain information as registered by health professionals during routine clinical practice, but data are not specifically prepared for research. In this sense, studies based on real-world clinical information like the VID are at risk of well-known biases such a differential recording, misclassification bias or missing data. Second, despite accounting for many relevant covariates in the adjustment, we may have missed some potentially relevant information and thus we cannot rule out residual confounding. For instance, MPD risk grows exponentially in chronic opioid users, therefore the risk of chronic use after initiation may be partially explaining the association between the initial prescription and MPD outcomes. The association between therapy initiation and the risk of chronic opioid use is warranted. Also, given the observational nature of the study, we could expect the presence of indication bias. In this sense, our results must be interpreted with caution. Third, by design we excluded some groups of patients, such as initiators with excluded opioids or those who initiated two opioids at the same time. Also, we failed to include 4% of initiations in the analyses due to incorrect prescription diagnostic coding, which hindered the classification of indication of use of opioid initiations. In this sense, we are missing a fraction of the population treated. Fourth, some of the categories we used to define indication of use may be hard to interpret, as they include a great number of diverse indications and of ICD codes employed to initiate opioids. Also, coding decision-making in real-world practice may be subject to variability, a potential bias frequent in retrospective real-world studies. Fifth, information on inpatient medication is not available in VID, which may result in some misestimation. Sixth, we used prescription information to define the initial prescription, in an intention-to-treat analysis, therefore we may be overestimating the exposure of primary non-adherent patients (those who do not fill the first prescription). Seventh, we included some groups with a relatively low number of patients and events such as the ultrafast and short-acting opioids groups, where risk estimates should be interpreted with special caution and considered as exploratory findings. We confirmed the main analyses with sensitivity analyses excluding these groups. Finally, the generalisation of our results to other settings outside Spain, or even to other Spanish regions, should be approached with caution.
Conclusions
In this population-based cohort study including 959 286 patients, we linked several databases to identify prescription-related factors (initiating with fentanyl, long-acting oxycodone and short-acting buprenorphine, concurrent use of benzodiazepines, gabapentinoids and antipsychotics, prescribing a daily MME >120 or treatment for >2 weeks) as well as patient characteristics (male sex, young age, use of tobacco and alcohol, lack of resources and comorbidities) that were associated with opioid MPD. By identifying individuals at greatest risk, as well as prescription patterns associated with a greater risk for opioid MPD, these findings may provide guidance to clinicians, healthcare managers and policymakers to improve the safety of the therapeutic management of non-cancer pain.
Data availability statement
Data may be obtained from a third party and are not publicly available. The datasets presented in this article are not readily available because legal restrictions on sharing the dataset apply as regulated by the Valencia regional government by means of legal resolution by the Valencia Health Agency (2009/13312), which forbids the dissemination of data to third parties (accessible at: http://www.san.gva.es/documents/152919/157920/resolucionsolicituddatos.pdf). On request, authors can allow access to the databases in order to verify the accuracy of the analysis or the reproducibility of the study. Requests to access the datasets should be directed to Management Office of the Data Commission in the Valencia Health Agency (email: solicitud_datos@gva.es; telephone numbers: +34 961-928207; +34 961-928198).
Ethics statements
Patient consent for publication
Ethics approval
This study was approved by the Ethics Committee for Drug Research of the 'Hospital Clínico-Universitario de Valencia' (Reference: F-CE-GeVA 14 v1.2; 2019, March 21), who waived the informed consent requirement given the retrospective design with pseudonymised data and minimal risk to participants.
Footnotes
Correction notice This article has been corrected since it was first published online. The figure 959 286 has been updated to 958 019 in the results and what this study adds sections.
Contributors GS-G is the guarantor of the study. IH, AG-S, SP and GS-G were responsible for the study concept, design and data acquisition. CR and IH carried out the data preparation and the statistical analysis and AG-S drafted the manuscript. All authors participated in the analysis and interpretation of data, critical revision of the manuscript for important intellectual content, approved the final version submitted for publication and agreed to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work were appropriately investigated and resolved.
Funding This study was funded by Instituto de Salud Carlos III (grant no.: PI21/01413, RD21/0016/0006).
Competing interests The authors are employed by FISABIO, a research body in Spain affiliated with the Health Department of the Valencia Government. FISABIO signed a Collaboration agreement (2018–2019) with Grünenthal Pharma, to conduct independent research on 'Patterns of use of opioids in the National Health System', unrelated to the present work.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.