Impact of delisting high‐strength opioid formulations from a public drug benefit formulary on opioid utilization in Ontario, Canada

Abstract Purpose High‐strength opioid formulations were delisted (removed) from Ontario's public drug formulary in January 2017, except for palliative patients. We evaluated the impact of this policy on opioid utilization and dosing. Methods We conducted a longitudinal study among patients receiving publicly funded, high‐strength opioids from August 2016 to July 2017. The primary outcome measure was weekly median daily opioid dose (in milligrams of morphine or equivalent; MME) of (1) publicly funded and (2) all opioid prescriptions irrespective of funding source, evaluated using interrupted time series analyses and stratified by palliative care status. Results Following policy implementation, the weekly median daily dose of publicly funded opioids decreased immediately among non‐palliative patients by 10 MME (95% confidence limit [CL], −16.8 to −3.1) from a pre‐intervention dose of 424.5 MME (95% CL, 417.8‐431.2) and fell gradually among palliative patients by 3.9 MME per week (95% CL, −5.5 to −2.3) from a pre‐intervention dose of 450.1 MME (95% CL, 432.5‐467.7). In contrast, among all opioid prescriptions, gradual reductions in weekly median daily doses were observed only for non‐palliative patients, which decreased by 0.7 MME per week (95% CL, −1.3 to −0.2) from a pre‐intervention dose of 426.2 MME (95% CL, 420.9‐431.5). Conclusion The delisting of publicly‐funded, high‐strength opioids was accompanied by changes in funding source and small reductions in the weekly median daily doses dispensed. Although observed dose reductions of less than 1 MME weekly are likely not clinically relevant, safety implications of these changes require further monitoring.


| INTRODUCTION
High-strength opioid formulations are often used to treat severe pain when lower doses fail to produce adequate analgesia. Yet higher doses of opioids are associated with numerous harms, including motor vehicle collisions, 1 opioid misuse, 2,3 hyperalgesia, 4 depression, 5 testosterone suppression, 6 suicide, 7 and nonfatal or fatal overdose. [8][9][10][11][12] Given these concerns, the current guidelines for opioid use for chronic noncancer pain in Canada 13 and the United States 11 recommend against escalating beyond 90 mg of morphine or equivalent (MME) per day and suggest carefully tapering high doses when harms outweigh benefits. To regulate the availability of high-strength opioids in communities and help reduce high opioid doses, some jurisdictions in Canada and the United States have restricted access to highstrength opioid formulations 14 or prescriptions with high daily doses. 15 A 2017 petition to the US Food and Drug Administration (FDA) called for banning opioid formulations that could achieve daily doses of 90 MME or more when taken as directed. 16 In Ontario, Canada, 40% of long-acting opioid prescriptions dispensed in 2016, a period immediately preceding the release of the new Canadian guidelines, were for daily doses exceeding 90 MME, 17 and approximately 4% of newly treated patients were initiated on a dose exceeding this threshold. 18 On January 31, 2017, Ontario's Public Drug Program (OPDP) delisted all high-strength opioid formulations that were listed on their formulary at the time, which included 75-and 100-mcg/h fentanyl patches, 24-and 30-mg hydromorphone capsules, and 200-mg morphine tablets. By delisting these products, they were no longer eligible for reimbursement from the public drug program. This change was implemented as part of a strategy to reduce the risk of addiction and opioid-related adverse events resulting from the misuse and diversion of these opioids. 14 An exception was made for palliative care patients 19 who could access these formulations through a prior authorization process or from physicians registered with the palliative care facilitated access (PCFA) program. The new policy did not impact coverage for lower strength opioid formulations that could be combined to achieve equivalent daily doses. In addition, high-strength opioids could still be obtained outside the public drug plan (ie, out of pocket and private insurers).
We conducted an evaluation to assess whether this delisting policy led to any potentially harmful changes in opioid dispensing patterns.
Specifically, we sought to evaluate the impact of this policy on opioid prescription utilization and dosing among public drug beneficiaries in Ontario who were already taking these medications.

| Study population
We constructed a cohort comprising all individuals affected by the policy implementation, defined as those who received a publicly funded prescription for a high-strength opioid where 120% of the days' supply overlapped the policy implementation date (January 31, 2017). This 20% grace period for the days' supply was used to account

KEY POINTS
• Among non-palliative care patients, delisting publicly funded high-strength opioids led to an immediate reduction in the weekly median daily dose of publicly funded opioids dispensed and an accelerated declining dose of all opioids dispensed (irrespective of payer).
• Among palliative care patients, there were no significant changes in weekly median daily dose following the policy when considering all opioid prescriptions dispensed.
• One in three non-palliative care patients and one in five palliative care patients transitioned to accessing highstrength opioids through private insurance or by paying out of pocket.
• Overall, we found no evidence of the policy leading to increased likelihood of complete opioid discontinuation in any patient population.
for late refills and incomplete adherence given the PRN (ie, "use as needed") nature of some opioids. To limit the cohort to chronic highstrength opioid recipients, we excluded individuals who were not receiving a prescription high-strength opioid at the start of the study period (6 months prior to policy implementation). We excluded individuals who died before the end of the follow-up period to prevent observing changes in dose that could be due to having a person's dose included in the prepolicy period and not in the postpolicy period. This is necessary as we do not allow new patients in the follow-up period due to our focus on those affected by the policy at the time of implementation. Cohort entry was defined as the date of the last highstrength opioid prescription dispensed prior to the policy implementation date. We stratified the cohorts by palliative care status, defined using physician billing codes for palliative care services (Appendix S1) in the 6 months prior to cohort entry. To test the specificity of our findings, we used the same methods to construct a historical cohort 1 year earlier (August 1, 2015, and July 31, 2016), using January 31, 2016, as the "dummy" policy implementation date.

| Patient characteristics
We reported baseline patient characteristics for each cohort, stratified by palliative care status, including age, sex, residence in a rural community or long-term care home, and neighborhood income quintile.
We also report eligibility for the public drug program, categorized into seniors, long-term care residents, and individuals enrolled in disability and other social assistance programs (ie, high drug costs relative to income, employment assistance, home care, resident of home for special care, or enrolled in the Ontario Disability Support Program).
Furthermore, we measured health service utilization in the 6 months prior to cohort entry, including hospitalizations, ED visits, and physician visits.

| Outcomes
Outcomes were assessed by including all opioid prescriptions dispensed to individuals in the cohort over the study period. In the primary analysis, we assessed changes in weekly median daily opioid dose (in MME) during the study period. Specifically, we calculated the average daily opioid dose for each person every week, defined as the sum of the daily dose for the days covered by the prescription divided by the number of days covered by the prescription in the same week. We reported the median of this measure across patients dispensed opioids each week, resulting in one group-level summary estimate per week that we refer to as weekly median daily dose. This analysis was conducted for (1) publicly funded prescriptions and (2) all prescriptions dispensed, irrespective of funding source.
In a series of secondary analyses, we assessed changes in opioid utilization using prescriptions dispensed at any point in the 6 months following the policy. These binary outcomes included the following: continued use of publicly funded high-strength opioids, continued use of high-strength opioids paid through other means, discontinuation (ie, no prescription) of any publicly funded opioids, discontinuation of any opioids paid through any means, and de novo initiation of buprenorphine/naloxone or methadone.

| Statistical analysis
For our primary outcome, we used interrupted time series analyses and fit linear segmented regression models to the population weekly measure of median opioid daily dose, which are commonly used to assess the impact of policies/interventions on time series trends. [20][21][22][23][24] We included parameters to estimate the pre-intervention dose (intercept), pre-intervention trend (slope), post-intervention change in level (step), and post-intervention change in trend. In the presence of autocorrelation, we used a backward stepwise approach to include autoregressive parameters in the model. For our secondary binary outcomes, we used logistic regression models to test for differences between the intervention and historical cohorts.   In cases where the number of users is less than six, this number has been suppressed to ensure confidentiality. In cases where there is only one record being suppressed, another record has been suppressed to provide a range in order to avoid residual disclosure.
b Public drug program eligibility assigns individuals hierarchically as follows: receiving high drug costs relative to income, high-income seniors, resident of home for special care, receiving home care, resident of long-term care, receiving employment assistance, enrolled in the Ontario Disability Support Program, and low-income seniors. For this reason, the numbers may differ slightly from those identified as seniors and those living in a long-term care residence.    (n = 1212) transitioned to accessing high-strength opioids through non-publicly funded means (compared with 0.2% in the historical cohort, P < 0.01; Table 4). Despite this, the prevalence of discontinuation of any publicly funded opioid was low, but significantly higher in the intervention cohort (5.3%) compared with the historical cohort (0.5%; P < 0.01). We observed no difference between the intervention and historical cohorts in overall rates of opioid discontinuation among all opioids (0.6% vs 0.4%; P = 0.17) or incidence of methadone or buprenorphine/ naloxone initiation after the policy (1.6% vs 1.4%; P = 0.35).

| Patients receiving palliative care
Approximately half (49.5%; n = 54) of palliative care patients in the intervention cohort continued to access publicly funded high-strength opioids following the policy, which was significantly lower than in the preceding year (98.6%; P < 0.01; Table 4). In contrast, a much higher percentage (21.1%; n = 23) transitioned to accessing these opioids through other means relative to the preceding year (less than or equal to 3.5%; P < .01).

| INTERPRETATION
Following Ontario's delisting of high-strength opioid formulations from its public drug program, we found statistically significant reductions in publicly funded weekly median daily opioid doses, which were more substantial among patients not receiving palliative care.
However, when considering opioids reimbursed through any means, reductions in weekly median daily dose were very small (change of less than 1 MME reduction per week), statistically significant only among patients not receiving palliative care, and present in both the intervention and historical cohorts. Although we observed changing patterns of access for high-strength opioids through public and private payers, we found no evidence of complete opioid discontinuation following the policy. These findings are similar to an Oregon study evaluating the impact of a prior authorization policy for opioid prescriptions exceeding 120 MME per day, which reported a 20.3% decline in the probability of receiving a high-dose opioid prescription. 15 However, in contrast to our study, the study by Hartung et al did not examine the impact on dose received and was unable to measure the extent to which prescriptions were paid out of pocket.
A key finding of our study is that, although the majority of non-

| Strengths and limitations
Our ability to identify all opioids dispensed in Ontario allows us to evaluate the impact of a delisting policy at a population level. However, some limitations warrant discussion. First, analyses were limited to individuals with a valid Ontario health card (97% of NMS prescriptions), which could lead to a small degree of misclassification of continued opioid use. Second, our definition of palliative care, while broad, might not have captured all such patients in Ontario. This is evidenced by our finding that 2.3% of individuals identified as being non-palliative care patients continued to receive publicly funded, high-strength opioids following the policy. It is possible that these patients were in fact palliative care patients, or that they were treated by a PCFA physician who wrote a prescription for these medications inappropriately. In both of these cases, the null findings in our historical cohort analyses suggest that these limitations were not likely to have influenced our main findings.