Trajectories of opioid consumption as predictors of patient‐reported outcomes among individuals attending multidisciplinary pain treatment clinics

This study aimed to identify opioid consumption trajectories among persons living with chronic pain (CP) and put them in relation to patient‐reported outcomes 6 months after initiating multidisciplinary pain treatment.


| INTRODUCTION
Chronic pain (CP) is defined as pain persisting or recurring for more than 3 months. 1It is a major public health concern due to its potential consequences on the physical, psychological, and social functioning of those affected. 2Furthermore, the economic impact of CP is high. 3In Canada, the prevalence of CP increased by 5%, from 16% in 2000 to 21.0% in 2014, 4 which is similar to rates of CP in the United States at 20.4% in 2016. 5propriate CP management involves a multidisciplinary approach and multimodal interventions (i.e., combining pharmacological and non-pharmacological treatment modalities). 2,6,7However, these approaches are often inaccessible due to the limited number of such clinics in Canada and their long waiting lists. 81][12] Evidence indicates no to modest effects of prescription opioids on pain and function among individuals with CP. 13 Available clinical trials studies 14,15 are also typically of short follow-up duration or do not assess the harms of long-term opioid therapy, 11,16 such as opioid consumption for other reasons than pain, opioid diversion, opioid use disorder, and overdose deaths.It is striking that over the past two decades, the United States and Canada have been ranked the top worldwide consumers of prescription opioids. 2 Concurrently, the number of opioid overdose deaths have increased significantly over time, leading to the so-called opioid crisis. 17To tackle this crisis, the United States and Canada have introduced guidelines and policies 11,18 to better control and limit opioid prescriptions.[21] Nonetheless, a minority of patients with CP treated with opioids can benefit from long-term opioid therapy. 16,22However, studies to date have failed to identify predictive factors associated with positive opioid treatment response.Notwithstanding, opioids are often prescribed in specialized pain clinics as part of multimodal and multidisciplinary treatments.Yet, it remains difficult to predict who will benefit the most from these multidisciplinary programs in terms of pain, function, and quality of life. 236][27][28] To the best of our knowledge, no prior study has modeled trajectories of opioid consumption among people living with CP and their associations with patient-reported outcomes after initiating treatment in a multidisciplinary pain clinic.
The present study aimed to: (1) model opioid consumption trajectories among persons living with CP in the first 6 months after initiating multidisciplinary pain treatment using a novel individual-centered approach, (2) examine the sociodemographic and clinical profiles of participants who fell into each of the identified trajectory groups, and (3) assess the associations between opioid consumption trajectory memberships and patient-reported outcomes (average and worst pain intensity, pain interference, severity of depressive symptoms, and health-related quality of life) 6 months after initiating multidisciplinary pain treatment.

| Study design and data sources
This retrospective cohort study was based on data from the Quebec Pain Registry (QPR), which contains patient-reported outcomes (PROs) and data from nurse-administered questionnaires. 29 under social assistance, and workers and their families without access to a private drug insurance program.This represents $45% of the Quebec population 30 and such database was shown to be valid. 31R data were linked to the RAMQ database using the unique
We retained only people covered by the Quebec prescription drug insurance plan in the first 6 months following the initial visit to the pain clinic (n = 4136).Pain management at the multidisciplinary pain treatment clinics was personalized according to the patients' needs and characteristics, and could include pharmacological, procedural, physical, and psychosocial interventions. 29We excluded 1989 participants who did not meet the inclusion criteria (Figure 1).Finally, we excluded one participant with clearly outlying morphine milligram equivalents during follow-up (weekly average > 4000 morphine milli-

| Opioid consumption
Data on opioid consumption were extracted from the RAMQ prescription claims database in the 6 months after initiating multidisciplinary pain treatment.The opioid consumption calculations 18,31,35,36 can be found in Appendix A in Data S1.

| Self-reported outcomes of interest
Average and worst pain intensity, pain interference, severity of depressive symptoms and health-related quality of life (physical and mental) 6 months after initiating multidisciplinary pain treatment were selected as our dependent variables.These outcomes were chosen because they are relevant to the bio-psycho-social nature of CP 37,38 and can help to identify whether some trajectories were more favorable than others.Pain intensity: Average and worst pain intensity over the past 7 days were assessed using a standardized Numeric Rating Scale ranging from 0 (no pain) to 10 (worst possible pain).Such rating scales are reliable, valid, and responsive among pain patients. 37in interference: It was measured using the average of the interference items of the Brief Pain Inventory-10 (BPI-10).37,39,40 Total score ranges from 0 to 10. Severity of depressive symptoms: It was assessed using the Beck Depression Inventory-I (BDI-I), which has 21 items scoring from 0 to 3. The total score ranges from 0 to 63, with higher scores indicating higher levels of depressive symptoms.[41][42][43] The BDI-I showed adequate psychometric properties. 43 Healt-related quality of life: It was assessed using the SF-12v2 (4-week recall).Two 0-100 composite norm-based scores representing physical healthrelated quality of life and mental health-related quality of life were derived.Higher scores indicate better quality of life.44 This questionnaire is available in numerous languages [45][46][47] and has adequate psychometric properties in patients with pain.48

| Covariables
Covariables included sociodemographic data, pain characteristics at the time of the initial visit to the pain clinic, 49,50-52 lifestyle habits, and baseline measures of our outcomes of interest.Longitudinal health administrative data allowed to compute and consider opioid consumption profile before the initial visit to the pain clinic.We also derived variables such as polypharmacy, 52,53 excessive polypharmacy, 53,54 use of opioid antagonists (e.g., naloxone, naltrexone) or benzodiazepines, and the combined Charlson and Elixhuser comorbidity indices 55,56 during the 12 months before the initial visit to the clinic.Details of how the variables were operationalized can be found in Table 1.

| Statistical analysis
The statistical analysis included descriptive statistics, opioid consumption trajectories, trajectory groups profile comparisons, associations between opioid trajectory group membership and 6-month patientreported outcomes, subgroup analyses, and sensitivity analyses.
Appendix B in Data S1 has the details of the statistical analysis.

| RESULTS
Participants' characteristics are shown in

Neuropathic pain
Neuropathic pain was evaluated using the self-reported portion of douleur neuropathique 4 questions (DN4) 48 1.Having neuropathic pain if there was a DN4 ≥ 4; 2. Non-neuropathic pain was defined as a DN4 score <4

Pain catastrophizing
The Pain catastrophizing Scale (PCS), a well-validated measurement tool which assesses the extent to which individuals tend to ruminate, magnify, and feel helpless in the presence of pain, [49][50][51] was used.It has 13 items scored from 0 (not at all) to 4 (all the time).
A total score based on the sum of the 13 items was calculated.It ranges between 0 to 52, with higher scores indicating higher levels of pain catastrophizing

Comorbidities
The combined Charlson comorbidity index and Elixhauser comorbidity index 56,76 was calculated using a SAS Macro developed by the Institut national de santé publique du Québec 56 based on Quan's algorithm. 55The index documents the presence of 32 health conditions during the year before the initial visit to the pain clinic.
Index score (Continues) characteristics, clinical variables, and drug-related variables can be found in Table 2.

| Opioid consumption trajectories
The    The multivariate multiple linear regression models evaluating the associations between trajectory groups and patient-reported outcomes at 6 months are presented in Tables 5 and 6.There was only an association between trajectory groups and depressive symptoms.

| Associations between opioid trajectory group membership and 6-month patient-reported outcomes
The "increasing" opioid consumption group had a significantly greater severity of depressive symptoms than the "no or very low and stable" opioid consumption group (Beta = 2.21; 95% CI = 0.14, 4.28) after adjusting for several covariates (Table 6).

| Subgroup analyses
Appendix D in Data S1 has the details of the trajectory analysis by neuropathic pain and Appendix E in Data S1 has the trajectory analysis among oxycodone users.There were three, four, and two trajectories in the neuropathic group, non-neuropathic group, and oxycodone users, respectively.
One trajectory in both neuropathic groups and both trajectories among oxycodone users were different from the trajectories of the complete data.

| Sensitivity analyses
Sensitivity analyses confirm the robustness of our results as shown in Appendix F in Data S1, the trajectory analysis in the full data, Appendix G in Data S1, the weighted kappa of trajectories distributions between complete data vs. full data, and Appendix H in Data S1, trajectory analysis among those covered by the public prescription drug insurance between 1 year before and 6-months after the initial visit.Furthermore, the distribution of the incomplete sample (1149) across trajectories was similar to the complete data: "no or very low and stable" opioid consumption (n = 1097, 95.5%), "increasing" opioid consumption (n = 30, 2.6%), "decreasing" opioid consumption (n = 22, 1.9%).Those included (n = 2146) were slightly older (59.7 SD 14.7 vs. 55.5 SD 16.2) and slightly more likely to be women (60.2% vs. 55.0%)compared with those not included (n = 1149) but those differences do not appear clinically important.In the first two sensitivity analyses, the multivariate linear regression models did not show any association between the trajectory groups and the six evaluated outcomes at 6-month.Additionally, the binary logistic regression model did not show an association between trajectories and global impression of change.(Continues) This study identified three opioid consumption trajectories in the first 6 months after initiating multidisciplinary pain treatment: "no or very low and stable" opioid consumption, "increasing" opioid consumption, and "decreasing" opioid consumption.among the "increasing" trajectory compared to the "no or very low and stable" trajectory).

| Trajectory groups
The great majority of patients (n = 96.3%)did not consume or consumed very low doses of opioids (<90 MME) 11 in the 6 months following initiation of multidisciplinary pain treatment.Only 1.9% of the participants increased the opioid regimen over time while a small percentage (1.8%) was tapering opioids.These findings concur somewhat with Groenewald et al.'s study, 57 in general population (n = 31 916) from the United States, 4.4% (95% CI, 3.8%-5.1%)reported opioid used without other treatment for CP management. 57Furthermore, we found 10% had >90 MME of opioid exposure in the 12 months prior to the visit, yet so few patients (<4%) were in any group other than the "no or low and stable use" group after treatment visit.It could be explained by the fact that chronic pain management and pharmacotherapy is suboptimal in primary care. 58When starting a follow-up in a multidisciplinary facility with pain specialists, some patients were probably tapered off opioids because of suspected opioid induced hyperalgesia or non-optimal opioid pharmacotherapy.
As far as we know, no other research has employed data collected with comparable methods to ours in terms of a linkage of a pain registry and administrative data and opioid trajectory analysis in a population with CP.As a result, direct comparisons of studies related to opioid consumption trajectories among persons living with CP remain difficult.However, some comparisons are possible with studies conducted in other populations which used other types of statistical individual-centered approaches.For example, Elmer et al. 59 found three trajectories of prescription opioid use among adult aged 18-45 from Allegheny County, Pennsylvania, between 2010 and 2011.Two of these trajectories were similar to those identified in our study, as we both found a low, filled few, if any, opioid prescriptions (83%), and a declined opioid consumption group (12%).The third trajectory had a sustained high level of opioid prescriptions filled over the duration of observation.

| Trajectory groups baseline profile
The "no or very low and stable" opioid consumption group had the better physical and psychological functioning at the initial visit, and a lower percentage of participants classified as currently smoking, being on polypharmacy, having an excess of polypharmacy, and few opioid users and benzodiazepine users in the months prior to the initial visit to the pain clinic.The "increasing" opioid consumption group showed the greatest scores of pain interference, the greatest percentage of participants currently smoking and excessive polypharmacy, and a high proportion of benzodiazepine users in the past 12 months, although the 95% CI did not show significant differences in comparison with the "decreasing" opioid consumption group.Concurrent use of opioids and benzodiazepines should be avoided or at least minimized as the use of both drugs at the same time put patients at a greater likelihood of T A B L E 4 Patient-reported outcomes at 6-month follow-up by trajectory groups."increasing" opioid consumption 18 as well as potentially fatal overdoses. 60The "decreasing" opioid consumption group had the greatest percentage of opioid users in the past 12 months, although it was not significantly different from the "increasing" opioid consumption group.
These findings could be attributed to the fact that multidisciplinary pain rehabilitation programs have a long history of including opioid tapering as part of their program package. 61,62Thus, it is possible that in our study, those patients with a history of opioid consumption in the past year may be encouraged to reduce their opioid use as part of the multidisciplinary pain clinic management.A systematic review found that F I G U R E 3 Boxplot of the observed outcomes by opioid trajectory membership at the initial visit and at 6-month follow-up.
opioid tapering in multidisciplinary pain rehabilitation programs reduces pain or maintains the same level of pain in 95% of the included studies. 63This finding may be explained by opioid-induced hyperalgesia, in which a patient receiving opioids for pain treatment could become more sensitive to certain painful stimuli. 64Moreover, a high percentage of patients discontinued long-term opioid use because of a lack of evidence on long-term effectiveness. 183 | Associations between opioid trajectory group membership and 6-month patient-reported outcomes We found that opioid consumption trajectories were not significantly associated with PROs related to pain in the first 6 months after initiating multidisciplinary pain treatment.However, those classified in the "increasing" opioid consumption group had more severe depressive symptoms than those classified as "no or very low and stable" opioid consumption group after adjusting for several covariables.
Our findings are consistent with previous literature showing an association between mental health symptoms and prescription opioid use among patients with CP. [65][66][67] Additionally, prolonged pain warranting opioids may increase the risk of developing depression.However, the presence of depressive symptoms may also lead to impaired coping and, thus increased reliance on opioids to manage pain. 68erefore, future studies should consider using repeatedly measured data (exposures, covariates, and outcomes) to correct for this potential bias. 69

| Strengths and limitations
Our study had several strengths.To our knowledge, this is the first study to identify the trajectories of opioid consumption among patients with CP, and then relate these trajectories to PROs.The data we used came from the PAIR Project and offered a unique opportunity to harness the strengths of longitudinal administrative claim data, PROs, and precise pain diagnoses.The validity of administrative databases used alone for identifying persons living with CP is highly questionable 70 and self-report data are essential to capture the biopsychosocial nature of CP.Another strength of our study was the use of an individual-centered statistical approach based on principal factor analysis and cluster analysis that allowed the identification of opioid consumption trajectories.Furthermore, we performed additional sensitivity analyses to confirm the robustness of our results and comparability between those included and excluded in this study.
Although we followed the CDC guidelines for MME calculations, 18 there are other ways of calculating the MME 71 that were not explored within this manuscript and could show different results.Quebec prescription claims only cover a portion of the Quebec population (lower socioeconomic status) which reduced our sample size and affected the external validity of high socioeconomic status individuals. 72However, the databases still captured a substantial proportion of high socioeconomic status individuals.
Although we use the term consumption, our data speak directly to opioid prescription claims and not to the directly observed consumption by patients.In addition, we relied on diagnosis codes to calculate the comorbidity index, whose validity will therefore be dependent on the validity of the individual codes. 73This study includes participants currently seeking treatment in tertiary care settings who may have suffered severe CP for a long time and are not necessarily representative of opioid effectiveness soon after CP onset.Moreover, data from 2008 to 2014 may not reflect current opioid prescribing practices.Furthermore, variables such as patients' beliefs, anxiety levels, and fear avoidance behaviors were not recorded and could be potential confounders of the associations between opioid consumption trajectories and PROs.Similarly, granular data on all of the specific treatments each individual received was not available within the registry.

| Conclusions
In this retrospective cohort of patients with CP, we identified three opioid consumption trajectories among persons living with CP in the first 6 months after initiating multidisciplinary pain treatment: (1) "no or very low and stable" opioid consumption, (2) "increasing" opioid consumption, and (3) "decreasing" opioid consumption.However, those trajectories could be different based on type of pain and type of opioid.Participants included in the "no or very low and stable" opioid consumption trajectory had a better clinical profile.Finally, this study revealed that opioid consumption trajectories do not seem to be important determinants of most PROs 6 months after initiating multidisciplinary pain treatment and underscores the importance of continuing to question the long-term effects of opioids for the treatment of chronic pain.
The QPR was linked to the province of Quebec (Canada) administrative health insurance data.The QPR (https://quebecpainregistry.com/) recruited participants scheduled for an initial visit across five multidisciplinary pain treatment clinics in the province of Quebec, Canada, between 2008 and 2014 (n = 9480 eligible to enroll).Patient-reported and nurse-administered questionnaires were completed before the initial visit to the pain clinic (index date), and at 6-month follow-up.Diagnoses made by physicians of the pain clinics were also recorded in the QPR.The administrative data was obtained from the prescription claims of the Quebec public health insurance plan databases (Régie de l'assurance maladie du Québec [RAMQ]).The RAMQ prescription claims database includes information from reimbursed prescription drugs dispensed to participants aged ≥65 years, those

( 1 )
Drinking habitsFrequency of alcohol consumption in the past 12 months (1) Less than two;(2) two or more times per week Currently smoking Smoking status (1) Yes; (2) No Drug use consumption in the past 12 months Derived variable calculated based on at least one affirmative response to the use in the past 12 months out of the following drugs: marijuana, cannabis, hashish, cocaine, crack, heroin, ecstasy, or other drugs.(1) Yes; (2) No Opioid risk tool Opioid risk tool (ORT) is a 10-item scale designed to assess risk factors for opioid abuse, including family and personal history of substance abuse, age, history of preadolescent sexual abuse, and psychological distress.High score represent severe risk of opioid abuse. 74,750 to 24 Variables from the administrative data (Régie de l'assurance maladie du Québec [RAMQ]) Opioid use past 12 months MME/day calculated in the year before the initial visit to the pain clinic was averaged and used to classified opioid users > 90 MMEs ≤ 90 MMEs Polypharmacy Use of ≥5 drugs at the time of the initial visit to the pain clinic.Yes; (2) No Excessive polypharmacy Use of ≥10 drugs at the time of the initial visit to the pain clinic.(1) Yes; (2) No Use of opioid antagonists Use of opioid antagonists in the past 12 months (1) Yes; (2) No Use of benzodiazepines Use of benzodiazepines in the past 12 months (1) Yes; (2) No descriptive statistics of the 24 statistical indicators (Step 1) and the three indicators selected by the factor analysis (Step 2) can be found in Appendix C in Data S1.Individual patterns were classified into one of the following three opioid consumption trajectory groups: "no or very low and stable" opioid consumption (n = 2067, 96.3%, weekly mean MMEs = 31.5 ± 99.2), "increasing" opioid consumption (n = 40, 1.9%, weekly mean MMEs = 473.4± 283.7), and "decreasing" opioid consumption (n = 39, 1.8%, weekly mean MMEs = 411.2± 302.7; Figure 2).

Figure 3
Figure 3 presents the outcomes at 6-month follow-up along with their baseline values.At the index date and 6-month follow-up, the "no or very low and stable" opioid consumption group had low pain interference scores.The "increasing" opioid consumption group showed a high median of pain interference scores (Figure 3C.Md = 72.0;95% CI = 64.0,77.0) at 6-month follow-up, but it was not significantly different from the "decreasing" opioid consumption group (Md = 60.0;95% CI = 54.0,70.0).The "decreasing" opioid consumption group obtained a low median of physical health-related quality of Sociodemographic and clinical profiles varied according to opioid trajectory membership.Surprisingly, the results revealed that opioid consumption trajectories were not associated with most PROs at 6 months.Nevertheless, there was an association between opioid consumption trajectories and severity of depressive symptoms at the 6-month follow-up (worse symptoms being recorded T A B L E 2 (Continued) Abbreviations: BPI-10, Beck depression inventory; DN4, douleur neuropathique; MME, morphine milligram equivalents; PCS, pain catastrophizing scale; SF-12, 12-item short form survey.

Table 2
Covariables included in the study.

Table 3
shows participants' characteristics at the initial visit across the different opioid consumption trajectory groups.At the initial visit, in comparison with the "increasing" and "decreasing" opioid consumption groups, the "no or very low and stable" opioid consumption group showed a low score of pain interference (61.0), a low percentage of current smokers (29.8%), low percentage of participants being on polypharmacy (54.0%), low percentage of patients showing polypharmacy excess (19.4%), few opioid users (8.0%) and benzodiazepine users (41.0%) in the 12 months preceding the index date.

Table 4
Participants' characteristics at the initial visit to the pain clinic.
Participant characteristics at the initial visit to the pain clinic by trajectory groups.
Abbreviations: BPI-10, beck depression inventory; DN4, douleur neuropathique; IQR, interquartile range; MME, morphine milligram equivalents; PCS, pain catastrophizing scale; SD, standard deviation; SF-12, 12-item short form survey. *Opioid risk tool was available in a subsample (nÀ969).F I G U R E 2 Observed average weekly morphine milligram equivalents (MMEs) during the first 26 weeks of multidisciplinary pain treatment stratified by opioid consumption trajectory membership.T A B L E 3 Adjusted for age, sex, education level, civil status, family income, neuropathic pain, disability, benzodiazepine user in the past 12 months, polypharmacy status, excessive polypharmacy, drugs consumption in the past 12 months, currently smoking, heavy drinking in the past 12 months, lumbar pain, lower limb pain, generalized syndromes, cervical pain, pain duration at the initial visit, pain frequency at the initial visit, average pain intensity at the initial visit, worst pain intensity at the initial visit, pain interference at the initial visit, pain catastrophizing at the initial visit, depression at the initial visit, Combined Charlson and Elixhauser Index, mental health-related quality of life at the initial visit, physical health-related quality of life at the initial visit, average pain intensity at 6 month, worst pain intensity at 6 months, pain interference at 6 months, and physical health-related quality of life at 6 months.Multivariate multiple linear regression model exploring associations between trajectory groups membership and average pain intensity, worst pain intensity, pain interference, and physical health-related quality of life at 6 months (n = 2133).Adjusted for age, sex, education level, civil status, family income, neuropathic pain, disability, benzodiazepine user in the past 12 months, polypharmacy status, excessive polypharmacy, opioid user in the past 12 months, drugs consumption in the past 12 months, currently smoking, heavy drinking in the past 12 months, lumbar pain, lower limb pain, generalized syndromes, cervical pain, pain duration at the initial visit, pain frequency at the initial visit, average pain intensity at the initial visit, worst pain intensity at the initial visit, pain interference at the initial visit, pain catastrophizing at the initial visit, depression at the initial visit, Combined Charlson and Elixhauser Index, mental health-related quality of life at the initial visit, physical health-related quality of life at the initial visit, depression at 6 months, and mental health-related quality of life at 6 months.
a T A B L E 5 a