Skip to main content
NCBI home page
Clinical Pharmacology and Therapeutics logoLink to Clinical Pharmacology and Therapeutics
. 2026 Apr 7;120(1):122–138. doi: 10.1002/cpt.70267

Opioid Toxicity Following Concomitant Use of Macrolide Antibiotics with Fentanyl, Hydromorphone, or Oxycodone: A Population‐Based Study

Tony Antoniou 1,2,✉, Joanna Yang 2, David N Juurlink 2,3,4, Joanne M‐W Ho 5,6,7, Muhammad Mamdani 2,8,9,10,11, Fangyun Wu 2, Mina Tadrous 2,8,12, Tara Gomes 2,8,13,14
PMCID: PMC13264458  PMID: 41944476

Abstract

Fentanyl, hydromorphone, and oxycodone are metabolized by cytochrome P450 3A4 (CYP3A4). Co‐administration with CYP3A4‐inhibiting macrolides (clarithromycin or erythromycin) may increase opioid concentrations and overdose risk, but the clinical impact of these interactions is unknown. We conducted three population‐based nested case‐control studies of Ontario residents aged 15 years or older who were prescribed transdermal fentanyl, oral hydromorphone, or oral oxycodone between 1997 and 2023. We defined cases as individuals who died of or were hospitalized with opioid toxicity and matched each case to up to four controls on age, sex, calendar year, and a disease risk score. We assigned controls random index dates based on the distribution among cases and estimated adjusted odds ratios (aOR) and 95% confidence intervals (CIs) for the association between opioid toxicity and macrolide use in the preceding 14 days. We matched 735 fentanyl‐treated (45.3% of eligible cases), 2,506 hydromorphone‐treated (64.5%), and 678 oxycodone‐treated (50.1%) individuals with opioid toxicity to at least one control. Across all studies, 70 cases and 120 controls were exposed to a macrolide. CYP3A4‐inhibiting macrolide exposure was associated with an increased risk of opioid toxicity among patients treated with fentanyl (aOR 3.80, 95% CI: 1.36–11.56), hydromorphone (aOR 3.38, 95% CI: 1.28–8.97), and oxycodone (aOR 3.11, 95% CI: 1.05–9.51). We observed no associations with azithromycin, a non‐CYP3A4 inhibiting macrolide. There was no significant effect modification by study period, and findings were robust in bias analyses. These findings likely extend to individuals exposed to clandestinely produced fentanyl and to other CYP3A4 inhibitors.

Study Highlights.

  • WHAT IS THE CURRENT KNOWLEDGE ON THE TOPIC?

Pharmacokinetic studies demonstrate that CYP3A4 inhibitors can increase opioid concentrations. However, the clinical consequences of these interactions in real‐world practice remain unknown.

  • WHAT QUESTION DID THIS STUDY ADDRESS?

Do CYP3A4‐inhibiting macrolides (erythromycin, clarithromycin) increase the risk of opioid overdose and death among patients prescribed opioids metabolized by CYP3A4 (fentanyl, hydromorphone, oxycodone)?

  • WHAT DOES THIS STUDY ADD TO OUR KNOWLEDGE?

In this population‐based study, co‐prescription of CYP3A4‐inhibiting macrolides was associated with a 3.1‐ to 3.8‐fold increased risk of opioid toxicity. Azithromycin, which does not inhibit CYP3A4, showed no such risk. While the use of CYP3A4‐inhibiting macrolides has declined, clarithromycin is still used for specific indications, such that combined use places patients at risk. Moreover, our findings likely extend to the use of other commonly prescribed CYP3A4 inhibitors.

  • HOW MIGHT THIS CHANGE CLINICAL PHARMACOLOGY OR TRANSLATIONAL SCIENCE?

Clinicians should preferentially select azithromycin or other non‐CYP3A4‐inhibiting antibiotics for patients receiving these opioids. When clarithromycin is required, careful opioid dose adjustment and monitoring is warranted. These findings have broader implications for other potent CYP3A4 inhibitors prescribed to opioid‐treated patients, as well as for the population of people who use illicitly derived fentanyl.

Opioid toxicity and overdose deaths remain public health crises in North America. In Canada, there have been over 50,000 opioid‐related deaths between 2016 and 2024, 1 with over 250,000 years of life lost in 2021 alone. 2 In the United States, more than 79,000 opioid‐related deaths were reported in 2023. 3 Although clandestinely produced nonpharmaceutical fentanyl and its analogues account for most opioid‐related deaths, 1 , 4 prescription opioids continue to contribute to harm. In Ontario, Canada, one in 10 opioid overdose deaths during the COVID‐19 pandemic involved only pharmaceutical opioids. 5 More recently, prescription opioids were identified in nearly one of every five opioid toxicity deaths in Canada in 2024, underscoring their ongoing role in the opioid crisis. 1 A similar pattern has been observed in the United States, where prescription opioids were involved in 18% of opioid overdose deaths in 2022. 3 Risk factors for prescription opioid‐related overdose and death include opioid dose, increasing age, male sex, substance use disorder, cardiorespiratory comorbidities, chronic kidney disease, and mental health conditions. 6 , 7

Interactions between prescribed opioids and other medications are an underrecognized and potentially preventable cause of accidental opioid overdose and death. Patients prescribed fentanyl, hydromorphone, or oxycodone are at especially high risk of accidental overdose resulting from drug interactions because these opioids are metabolized by cytochrome P450 3A4 (CYP3A4), the most abundant of the cytochrome P450 enzymes. 7 , 8 , 9 , 10 , 11 , 12 , 13 , 14 Volunteer studies show that strong CYP3A4 inhibitors can increase concentrations of oxycodone and its active metabolite oxymorphone by up to 3.6‐fold and 7.3‐fold, respectively. 15 , 16 , 17 , 18 , 19 , 20 , 21 Similarly, fentanyl exposure increased by 30% to 170% following the administration of CYP3A4 inhibitors. 22 , 23 , 24 Although no pharmacokinetic studies have examined CYP3A4 inhibition in individuals receiving hydromorphone, in vitro studies indicate that inactivation to norhydromorphone is reduced by ~50% with combined use. 13

Yet, despite these data, no studies have examined the real‐world consequences of co‐prescribing CYP3A4 inhibitors to patients receiving oxycodone, fentanyl, or hydromorphone. This is important for several reasons. First, patients receiving opioids for chronic pain are often prescribed interacting nonopioid medications. Previous research suggests that more than 5% of opioid‐treated patients are co‐prescribed a drug with the potential to trigger a life‐threatening interaction. 25 However, these studies document potential interactions, rather than actual harm to patients. Second, most opioid overdose deaths involve illicitly manufactured fentanyl and its analogs, accounting for nearly 70% of all opioid‐related deaths in the United States in 2023. 4 Because these substances are also metabolized by CYP3A4, individuals exposed to nonpharmaceutical fentanyl may similarly be at risk of potentially fatal interactions with CYP3A4 inhibitors. 26 , 27 Finally, oxycodone, hydromorphone, and fentanyl have historically accounted for most prescription opioid‐related harm in Ontario, collectively contributing to more than 80% of opioid‐related deaths and more than 70% of overdose hospitalizations among individuals prescribed opioids in 2015 and 2016, a period preceding the widespread introduction of fentanyl to the unregulated drug supply. 28

Given the ongoing contribution of prescription opioids to overdoses and deaths, the potential for dangerous drug interactions, and the absence of population‐based studies examining the real‐world consequences of co‐administering opioids with CYP3A4 inhibitors, we examined the risks of hospitalization and death from opioid toxicity among patients prescribed fentanyl, oxycodone, or hydromorphone with a macrolide antibiotic in clinical practice. We hypothesized that patients co‐prescribed these opioids with the CYP3A4‐inhibiting macrolides erythromycin and clarithromycin would have a higher risk of opioid overdose compared to those not prescribed macrolides. In contrast, we expected no such risk with azithromycin, which shares overlapping clinical indications with clarithromycin and erythromycin but does not inhibit CYP3A4.

METHODS

Setting

We conducted three population‐based, nested case‐control studies of Ontario residents aged 15 years and older who were eligible for public drug coverage and who were dispensed one of fentanyl, hydromorphone, or oxycodone between January 1, 1997, and June 30, 2023. We used a nested case–control design for several reasons. First, opioid toxicity events resulting in death or hospitalization are relatively rare, making a case–control study more efficient than a cohort study for examining these outcomes. Second, this design is well‐suited to studying acute outcomes following short‐term exposures such as antibiotics. Finally, the design allows time‐matched control selection through risk‐set sampling, allowing for control of temporal trends in clinical practice. We selected this study period to ensure that we would identify a sufficient number of cases for robust analysis. In Ontario, publicly funded drug coverage is available to individuals aged 65 years and older, residents of long‐term care homes, recipients of social assistance, disability benefits or home care services, and individuals with high drug costs relative to their net household income. 29 All Ontario residents receive publicly funded physician and hospital care.

Data sources

We used Ontario's administrative health databases, which were linked using unique encoded identifiers and analyzed at ICES in Toronto, Ontario (https://www.ices.on.ca). ICES is an independent, nonprofit research institute whose legal status under section 45 of Ontario's Personal Health Information Protection Act (PHIPA) privacy law allows it to collect and analyze health care and demographic data, without consent or review by a Research Ethics Board, for health system evaluation and improvement. We identified prescription records using the Ontario Drug Benefit Database, which contains comprehensive records of all publicly funded medications dispensed to Ontario residents. We obtained diagnostic information from inpatient hospital admissions, emergency department visits, and mental health‐related hospitalizations using the Canadian Institute for Health Information's Discharge Abstract Database (DAD), National Ambulatory Care Reporting System (NACRS) database, and Ontario Mental Health Reporting System database, respectively. We used the Ontario Health Insurance Plan (OHIP) database to identify claims for physician services. We identified cancer diagnoses using the Ontario Cancer Registry, the Cancer Activity Level Reporting database, and the OHIP, DAD, and NACRS databases. We obtained demographic information from the Registered Persons Database, a registry of all Ontario residents eligible for health insurance. We identified accidental opioid‐related deaths using the Drug and Drug/Alcohol‐Related Death (DDARD) database, which contains information from all completed investigations by the Office of the Chief Coroner of Ontario in which an opioid was determined to have directly contributed to death, and the manner of death as determined by the investigating coroner (accidental, suicide, homicide, natural, undetermined). 30 These databases were linked in an anonymous fashion using encrypted health card numbers and are routinely used to study the consequences of drug interactions.

Study population

We identified individuals who experienced a hospitalization (emergency department visit or admission) for opioid toxicity or opioid‐related death between January 1, 1997, and June 30, 2023. We excluded deaths classified as suicides, homicides, natural, and undetermined by the investigating coroner, retaining only those deemed accidental. We identified hospitalizations and emergency department visits for opioid toxicity using International Classification of Diseases, 9th and 10th (ICD‐9) and 10th (ICD‐10) revision diagnosis codes 965.0 and E850.2, and T40.0, T40.1, T40.2, T40.3, T40.4, or T40.6, respectively. These codes are used by the Public Health Agency of Canada and the Canadian Institute for Health Information for monitoring opioid overdose hospitalizations, 31 , 32 and a validation study found sensitivity, specificity, and positive predictive values of 97.2%, 84.6%, and 87.4%. 33 We defined the index date as the date of emergency department visit, hospitalization, or death, whichever occurred first.

From within this cohort, we conducted three separate studies of individuals treated with transdermal fentanyl, oral hydromorphone, or oral oxycodone (see Figures S1–S3 for case and control selection flow diagrams). To attribute events to the specific opioid of interest, we required individuals to have at least one prescription for that opioid within 100 days preceding the index date, with the prescription overlapping the index date. For individuals with multiple hospital encounters during the study period, we considered only the first because inclusion of multiple events would violate the independence assumption of standard case–control analyses. We excluded individuals with a cancer diagnosis or those receiving palliative care in the 6 months preceding the index date, as these conditions may lead to escalating opioid requirements and a higher baseline risk of overdose or death. We also excluded individuals who filled prescriptions for more than one unique macrolide in the 30 days preceding the index date to mitigate confounding by illness severity and varying antibiotic exposures. For each case, we selected up to four controls from the same opioid‐specific cohort who were alive and event‐free on their assigned index date, applying the same eligibility criteria as for cases. Index dates for controls were randomly assigned based on the distribution of index dates among included cases. Like case patients, controls were required to have at least one prescription for the study opioid overlapping their index date. All cases and controls were also required to have at least 6 months of continuous eligibility for public drug benefits prior to the index date.

To increase the comparability of cases and controls, we used a disease risk score as a confounder summary measure to estimate each individual's predicted probability of opioid overdose. 34 We selected this approach because of the large number of potential confounders relative to the number of events and to balance baseline opioid overdose risk between cases and controls. We derived the disease risk score using a nonparsimonious multivariable logistic regression model that included the composite study outcome as the dependent variable and an extensive list of demographic and clinical variables (Table S1a,b for components of disease risk score and ICD‐9 and ICD‐10 codes) associated with opioid toxicity risk as independent variables. In the oxycodone study, we also included CYP2D6‐inhibiting antidepressants in the disease risk score because oxycodone is also a CYP2D6 substrate, and co‐administration with these medications has been associated with increased overdose risk. 35 We matched each case with up to four controls on their disease risk score (within 0.2 standard deviations), age at index date (within 1 year), sex, and the calendar year of opioid dispensing. We used matching with replacement, allowing controls to be matched to multiple cases. When fewer than four control subjects were available for a case, we analyzed only those controls and maintained the matching process. We excluded cases that could not be matched to at least one control. Because a substantial proportion of eligible cases were not matched at the default caliper of 0.2 standard deviations on the logit of the disease risk score, we examined the impact of relaxing the matching caliper width to 0.25 and 0.3 standard deviations. The number of matched or exposed cases did not increase appreciably with these wider calipers (Table S2 ). We therefore retained the 0.2 standard deviation caliper and compared baseline characteristics between matched and unmatched cases.

Exposure to macrolides

For each patient, we identified prescriptions for CYP3A4‐inhibiting macrolides erythromycin or clarithromycin dispensed in the 14 days preceding the index date. We selected this exposure window to allow sufficient time for complete enzyme inhibition, opioid accumulation, and development of an outcome (hospital encounter or death) following macrolide treatment. To test the specificity of our findings, we also examined the association between opioid toxicity and azithromycin, a non‐CYP3A4‐inhibiting macrolide which shares similar clinical indications with erythromycin and clarithromycin but is not expected to inhibit opioid metabolism or contribute to opioid toxicity. 36

Statistical analysis

We used standardized differences to compare baseline characteristics of cases and controls, with values less than 0.1 indicating good balance for a given covariate. 37

We used conditional logistic regression to estimate odds ratios (ORs) and 95% confidence intervals (CIs) for the association between the composite outcome and receipt of a CYP3A4‐inhibiting macrolide or azithromycin within the preceding 14 days. Patients who did not receive a study antibiotic served as the reference group. Because of sparse data, we used exact conditional logistic regression in the hydromorphone and oxycodone studies. We incorporated matching weights into each model to account for matching with replacement. Weighted analyses adjust for controls' repeated contribution across multiple matched sets, resulting in an effective control sample size equal to the number of cases. We adjusted all models for baseline variables with a standardized difference exceeding 0.1. We repeated our analyses by adjusting all models for calendar year as a continuous covariate to assess potential residual temporal confounding beyond that controlled for by matching on calendar year. To examine whether the association between CYP3A4‐inhibiting macrolides and opioid toxicity varied over the study period, we also tested for effect modification by time period using an interaction term between macrolide exposure and time.

Sensitivity analyses

We conducted E‐value and probabilistic bias analyses to test the robustness of our findings to unmeasured confounding, selection bias and misclassification bias (see Appendix S1 ). 38 , 39 The E‐value quantifies the minimum strength of association that an unmeasured confounder would need to have with both the exposure (i.e. CYP3A4‐inhibiting macrolide) and the outcome (i.e. opioid toxicity), conditional on measured covariates, to fully explain the observed exposure–outcome relationship. High E‐values for both the point estimate and confidence interval suggest that the observed association is unlikely to be explained by unmeasured confounding.

We also performed a deterministic bias analysis to estimate the effect of excluding high‐risk cases, 40 focusing on substance use disorder and psychiatric comorbidity, with recent psychiatry visits serving as a proxy for the latter. We used literature‐based relative risks for the association of these variables with prescription opioid overdose to calculate bias‐adjusted odds ratios. 7 , 40 This analysis was intended to illustrate the potential magnitude and direction of bias rather than to produce corrected causal estimates.

All analyses were completed using SAS Enterprise Guide 8.3 (SAS Institute, Inc, Cary, North Carolina) and R Studio.

Protocol registration

This study was not prospectively registered in a public registry.

RESULTS

During the 26‐year study period, we identified 77,385 individuals who experienced a hospital visit for or died from opioid toxicity. After exclusions, 1,621 of these individuals had a prescription for transdermal fentanyl, 3,883 had a prescription for oral hydromorphone, and 1,353 had a prescription for oxycodone overlapping their index date. Of these, 735 (45.3%) fentanyl cases, 2,506 (64.5%) hydromorphone cases, and 678 (50.1%) oxycodone cases were successfully matched to at least one control. Overall, across the three opioid cohorts, 70 cases and 120 controls were exposed to a macrolide. Compared to matched cases, unmatched cases generally had greater health service utilization and a higher prevalence of comorbidities, including substance use disorder (Tables 4 , 5 , 6 ). Across all opioid cohorts combined, the proportion of matched cases exposed to CYP3A4‐inhibiting macrolides declined over the study period, from >50% (>5 of 10 matched cases) between 1997 and 2007, 54% (23 of 43 matched cases) between 2008 and 2016, and <29% (<5 exposed among 17 matched cases) between 2017 and 2023. (Table S3 ).

Table 4.

Characteristics of unmatched and matched cases—fentanyl

Variable Unmatched cases (n = 886) Matched cases (n = 735) Standardized difference
Age (median, IQR) 59 (46–73) 69 (52–80) 0.39
Female 484 (54.6) 492 (66.9) 0.25
Number of prescription drugs in previous year, (median, IQR) 18 (13–24) 17 (12–23) 0.06
Charlson co‐morbidity index
No hospitalization 250 (28.2) 255 (34.7) 0.14
0 278 (31.4) 198 (26.9) 0.10
1 113 (12.8) 101 (13.7) 0.03
2+ 245 (27.7) 181 (24.6) 0.07
Comorbidities in previous 5 years
Chronic kidney disease 159 (17.9) 115 (15.6) 0.06
Chronic liver disease and cirrhosis 81 (9.1) 35 (4.8) 0.17
Chronic lung disease 399 (45.0) 275 (37.4) 0.16
Rheumatoid arthritis 103 (11.6) 98 (13.3) 0.05
Seizure disorder 132 (14.9) 73 (9.9) 0.15
Stroke 176 (19.9) 160 (21.8) 0.05
Sleep apnea 43 (4.9) 34 (4.6) 0.01
Hypertension 495 (55.9) 493 (67.1) 0.23
Heart failure 205 (23.1) 178 (24.2) 0.03
Mood disorders 123 (13.9) 64 (8.7) 0.16
Substance use disorder 241 (27.2) 80 (10.9) 0.42
Anxiety disorder 124 (14.0) 63 (8.6) 0.17
Schizophrenia/Other psychotic disorders 28 (3.2) 12 (1.6) 0.10
Deliberate self‐harm 326 (36.8) 44 (6.0) 0.81
Other mental health conditions 55 (6.2) 15 (2.0) 0.21
Psychiatrist visit in past 180 days 168 (19.0) 93 (12.7) 0.17
Number of physician visits in past year (median, IQR) 17 (11–27) 15 (9–23) 0.19
Number of emergency department visits in prior year (median, IQR) 3 (1–7) 2 (0–3) 0.47
0 158 (17.8) 213 (29.0) 0.27
1 123 (13.9) 143 (19.5) 0.15
2–5 323 (36.5) 286 (38.9) 0.05
6–10 187 (21.1) 66 (9.0) 0.34
>10 95 (10.7) 27 (3.7) 0.28
Days in hospital in prior year (median, IQR) 3 (1–15) 1 (0–11) 0.42
0 170 (19.2) 345 (46.9) 0.62
1–10 441 (49.8) 206 (28.0) 0.46
11–20 105 (11.9) 88 (12.0) 0.00
>20 170 (19.2) 96 (13.1) 0.17
Medication use in preceding 120 days
Benzodiazepines 538 (60.7) 385 (52.4) 0.17
Antipsychotics 271 (30.6) 175 (23.8) 0.15
Antidepressants 640 (72.2) 495 (67.3) 0.11
Other CNS depressants 1–5 (0.11–0.56) 1–5 (0.14–0.68) 0.02
Gabapentinoid 313 (35.3) 182 (24.8) 0.23
CYP3A4 inhibitors 94 (10.6) 63 (8.6) 0.07
CYP3A4 inducers 73 (8.2) 51 (6.9) 0.05
Opioids other than fentanyl 720 (81.3) 546 (74.3) 0.17
Residence in long‐term care facility 59 (6.7) 121 (16.5) 0.31
Income quintile
1 (lowest) 319 (36.0) 229 (31.2) 0.10
2 219 (24.7) 168 (22.9) 0.04
3 125 (14.1) 133 (18.1) 0.11
4 112 (12.6) 116 (15.8) 0.09
5 105 (11.9) 84 (11.4) 0.01
Rural residence 186 (21.0) 140 (19.0) 0.05
Open in a new tab

Table 5.

Characteristics of unmatched and matched cases—hydromorphone

Variable Unmatched cases (n = 1,377) Matched cases (n = 2,506) Standardized difference
Age (median, IQR) 54 (42–68) 63 (52–76) 0.47
Female 735 (53.4) 1,423 (56.8) 0.07
Number of prescription drugs in previous year, (median, IQR) 17 (12–24) 16 (11–22) 0.17
Charlson co‐morbidity index
No hospitalization 342 (24.8) 843 (33.6) 0.19
0 455 (33.0) 657 (26.2) 0.15
1 203 (14.7) 314 (12.5) 0.06
2+ 377 (27.4) 692 (27.6) 0.01
Comorbidities in previous 5 years
Chronic kidney disease 259 (18.8) 494 (19.7) 0.02
Chronic liver disease and cirrhosis 239 (17.4) 235 (9.4) 0.24
Chronic lung disease 705 (51.2) 1,054 (42.1) 0.18
Rheumatoid arthritis 125 (9.1) 241 (9.6) 0.02
Seizure disorder 232 (16.8) 199 (7.9) 0.27
Stroke 264 (19.2) 410 (16.4) 0.07
Sleep apnea 67 (4.9) 135 (5.4) 0.02
Hypertension 661 (48.0) 1,577 (62.9) 0.30
Heart failure 290 (21.1) 540 (21.5) 0.01
Mood disorders 257 (18.7) 194 (7.7) 0.33
Substance use disorder 460 (33.4) 379 (15.1) 0.44
Anxiety disorder 288 (20.9) 241 (9.6) 0.32
Schizophrenia/Other psychotic disorders 66 (4.8) 79 (3.2) 0.08
Deliberate self‐harm 703 (51.1) 224 (8.9) 1.03
Other mental health conditions 96 (7.0) 90 (3.6) 0.15
Psychiatrist visit in past 180 days 346 (25.1) 337 (13.4) 0.30
Number of physician visits in past year (median, IQR) 17 (10–29) 14 (8–22) 0.29
Number of emergency department visits in prior year (median, IQR) 3 (1–7) 2 (1–4) 0.45
0 189 (13.7) 600 (23.9) 0.26
1 199 (14.5) 507 (20.2) 0.15
2–5 547 (39.7) 1,010 (40.3) 0.01
6–10 274 (19.9) 263 (10.5) 0.26
>10 168 (12.2) 126 (5.0) 0.26
Days in hospital in prior year (median, IQR) 5 (1–17) 1 (0–11) 0.40
0 238 (17.3) 1,009 (40.3) 0.52
1–10 685 (49.7) 841 (33.6) 0.33
11–20 157 (11.4) 270 (10.8) 0.02
>20 297 (21.6) 386 (15.4) 0.16
Medication use in preceding 120 days
Benzodiazepines 853 (61.9) 1,032 (41.2) 0.42
Antipsychotics 448 (32.5) 619 (24.7) 0.17
Antidepressants 956 (69.4) 1,634 (65.2) 0.09
Other CNS depressants 13 (0.9) 13 (0.5) 0.05
Gabapentinoid 529 (38.4) 1,200 (47.9) 0.19
CYP3A4 inhibitors 119 (8.6) 187 (7.5) 0.04
CYP3A4 inducers 117 (8.5) 151 (6.0) 0.10
Opioids other than hydromorphone 851 (61.8) 1,093 (43.6) 0.37
Residence in long‐term care facility 47 (3.4) 247 (9.9) 0.26
Income quintile
1 (lowest) 575 (41.8) 965 (38.5) 0.07
2 309 (22.4) 553 (22.1) 0.01
3 215 (15.6) 394 (15.7) 0.00
4 164 (11.9) 339 (13.5) 0.05
5 105 (7.6) 253 (10.1) 0.09
Rural residence 229 (16.6) 376 (15.0) 0.04
Open in a new tab

Table 6.

Characteristics of unmatched and matched cases—oxycodone

Variable Unmatched cases (n = 675) Matched cases (n = 678) Standardized difference
Age (median, IQR) 54 (43–67) 53 (45–65) 0.01
Female 357 (52.9) 357 (52.7) 0.00
Number of prescription drugs in previous year (median, IQR) 16 (12–22) 15 (10–20) 0.22
Charlson co‐morbidity index
No hospitalization 232 (34.4) 292 (43.1) 0.18
0 213 (31.6) 205 (30.2) 0.03
1 86 (12.7) 83 (12.2) 0.02
2+ 144 (21.3) 98 (14.5) 0.18
Comorbidities in previous 5 years
Chronic kidney disease 91 (13.5) 69 (10.2) 0.10
Chronic liver disease and cirrhosis 45 (6.7) 51 (7.5) 0.03
Chronic lung disease 302 (44.7) 264 (38.9) 0.12
Rheumatoid arthritis 85 (12.6) 69 (10.2) 0.08
Seizure disorder 107 (15.9) 53 (7.8) 0.25
Stroke 92 (13.6) 77 (11.4) 0.07
Sleep apnea 33 (4.9) 36 (5.3) 0.02
Hypertension 324 (48.0) 324 (47.8) 0.00
Heart failure 103 (15.3) 87 (12.8) 0.07
Mood disorders 100 (14.8) 106 (15.6) 0.02
Substance use disorder 193 (28.6) 139 (20.5) 0.19
Anxiety disorder 102 (15.1) 87 (12.8) 0.07
Schizophrenia/Other psychotic disorders 41 (6.1) 25 (3.7) 0.11
Deliberate self‐harm 348 (51.6) 116 (17.1) 0.78
Other mental health conditions 32 (4.7) 25 (3.7) 0.05
Psychiatrist visit in past 180 days 159 (23.6) 128 (18.9) 0.11
Number of physician visits in past year (median, IQR) 17 (10–28) 16 (10–26) 0.09
Number of emergency department visits in prior year (median, IQR) 2 (1–5) 1 (0–4) 0.34
0 126 (18.7) 200 (29.5) 0.26
1 126 (18.7) 150 (22.1) 0.09
2–5 270 (40.0) 234 (34.5) 0.11
6–10 91 (13.5) 64 (9.4) 0.13
>10 62 (9.2) 30 (4.4) 0.19
Days in hospital in prior year (median, IQR) 1 (1–9) 0 (0–6) 0.46
0 156 (23.1) 355 (52.4) 0.63
1–10 376 (55.7) 208 (30.7) 0.52
11–20 66 (9.8) 50 (7.4) 0.09
>20 77 (11.4) 65 (9.6) 0.06
Medication use in preceding 120 days
Benzodiazepines 478 (70.8) 426 (62.8) 0.17
Antipsychotics 204 (30.2) 172 (25.4) 0.11
Antidepressants 487 (72.1) 456 (67.3) 0.11
Other CNS depressants 1–5 (0.15–0.74) 1–5 (0.15–0.74) 0.03
Gabapentinoid 184 (27.3) 97 (14.3) 0.32
CYP3A4 inhibitors 51 (7.6) 56 (8.3) 0.03
CYP3A4 inducers 38 (5.6) 39 (5.8) 0.01
Opioids other than oxycodone 522 (77.3) 492 (72.6) 0.11
Residence in long‐term care facility 8 (1.2) 14 (2.1) 0.07
Income quintile
1 (lowest) 271 (40.1) 258 (38.1) 0.04
2 176 (26.1) 152 (22.4) 0.09
3 95 (14.1) 110 (16.2) 0.06
4 80 (11.9) 89 (13.1) 0.04
5 52 (7.7) 66 (9.7) 0.07
Rural residence 94 (13.9) 78 (11.5) 0.07
Open in a new tab

Overall, baseline characteristics of fentanyl‐treated cases and controls were well balanced, although cases were more likely to have a seizure disorder (n = 73 [9.9%] vs. n = 48 [6.6%]; SD = 0.12) (Table 1 ). Following adjustment, patients with recent CYP3A4‐inhibiting macrolide exposure were more than three times as likely to experience an opioid toxicity event compared to those without macrolide exposure (adjusted odds ratio [aOR] 3.80, 95% CI: 1.36–11.56) (Figure 1 ). Results were similar following adjustment for calendar year (aOR 3.75, 95% CI: 1.46–9.62), and the effect of CYP3A4‐inhibiting macrolide exposure did not vary significantly across time periods (p‐value for interaction = 0.49). As expected, we found no such association with azithromycin (aOR 1.10, 95% CI: 0.27–4.04) (Figure 1 ).

Table 1.

Baseline characteristics of fentanyl cases and controls

Variable Unweighted, No. (%)a Weighted, No. (%)a
Cases (n = 735) Controls (n = 1,632) Standardized difference Cases (n = 735) Controls (n = 735) Standardized difference
Age (median, IQR) 69 (52–80) 71 (53–81) 0.15 69 (52–80) 69 (52–80) 0.00
15–29 1–5 (0.14–0.68) 1–5 (0.06–0.31) 0.06 1–5 (0.14–0.68) 1–5 (0.14–0.68) 0.02
30–49 126–130 (17.1–17.7) 246–250 (15.1–15.3) 0.06 126–130 (17.1–17.7) 128–132 (17.4–18.0) 0.01
50–65 200 (27.2) 418 (25.6) 0.04 200 (27.2) 191 (26.0) 0.03
66–79 205 (27.9) 471 (28.9) 0.02 205 (27.9) 216 (29.4) 0.03
80+ 199 (27.1) 492 (30.2) 0.07 199 (27.1) 195 (26.5) 0.01
Female 492 (66.9) 1,176 (72.1) 0.11 492 (66.9) 492 (66.9) 0.00
Number of prescription drugs in previous year, (median, IQR) 17 (12–23) 17 (12–22) 0.12 17 (12–23) 17 (12–22) 0.08
Charlson co‐morbidity index
No hospitalization 255 (34.7) 705 (43.2) 0.18 255 (34.7) 283 (38.5) 0.08
0 198 (26.9) 376 (23.0) 0.09 198 (26.9) 177 (24.1) 0.07
1 101 (13.7) 174 (10.7) 0.09 101 (13.7) 83 (11.3) 0.07
2+ 181 (24.6) 377 (23.1) 0.04 181 (24.6) 192 (26.1) 0.03
Comorbidities in previous 5 years
Chronic kidney disease 115 (15.7) 230 (14.1) 0.04 115 (15.7) 110 (15.0) 0.02
Chronic liver disease and cirrhosis 35 (4.8) 73 (4.5) 0.01 35 (4.8) 38 (5.2) 0.02
Chronic lung disease 275 (37.4) 582 (35.7) 0.04 275 (37.4) 273 (37.2) 0.01
Rheumatoid arthritis 98 (13.3) 212 (13.0) 0.01 98 (13.3) 96 (13.0) 0.01
Seizure disorder 73 (9.9) 96 (5.9) 0.15 73 (9.9) 48 (6.6) 0.12
Stroke 160 (21.8) 306 (18.8) 0.08 160 (21.8) 138 (18.8) 0.08
Sleep apnea 34 (4.6) 51 (3.1) 0.08 34 (4.6) 26 (3.5) 0.06
Hypertension 493 (67.1) 1,089 (66.7) 0.01 493 (67.1) 483 (65.8) 0.03
Heart failure 178 (24.2) 389 (23.8) 0.01 178 (24.2) 179 (24.4) 0.00
Mood disorders 64 (8.7) 91 (5.6) 0.12 64 (8.7) 52 (7.0) 0.06
Substance use disorder 80 (10.9) 112 (6.9) 0.14 80 (10.9) 69 (9.4) 0.05
Anxiety disorder 63 (8.6) 102 (6.3) 0.09 63 (8.6) 52 (7.1) 0.05
Schizophrenia/Other psychotic disorders 12 (1.6) 24 (1.5) 0.01 12 (1.6) 12 (1.6) 0.00
Deliberate self‐harm 44 (6.0) 45 (2.8) 0.16 44 (6.0) 34 (4.6) 0.06
Other mental health condition 15 (2.0) 17 (1.0) 0.08 15 (2.0) 9 (1.2) 0.06
Psychiatrist visit in past 180 days 93 (12.7) 155 (9.5) 0.10 93 (12.7) 81 (11.1) 0.05
Number of physician visits in past year (median, IQR) 15 (9–23) 15 (9–24) 0.02 15 (9–23) 15 (10–25) 0.06
Number of emergency department visits in prior year (median, IQR) 2 (0–3) 1 (0–3) 0.17 2 (0–3) 2 (0–3) 0.02
0 213 (29.0) 579 (35.5) 0.14 213 (29.0) 228 (31.0) 0.04
1 143 (19.5) 330 (20.2) 0.02 143 (19.5) 136 (18.5) 0.03
2–5 286 (38.9) 582 (35.7) 0.07 286 (38.9) 278 (37.9) 0.02
6–10 66 (9.0) 102 (6.3) 0.10 66 (9.0) 67 (9.1) 0.00
>10 27 (3.7) 39 (2.4) 0.08 27 (3.7) 27 (3.7) 0.00
Days in hospital in prior year (median, IQR) 1 (0–11) 0 (0–9) 0.05 1 (0–11) 2 (0–11) 0.04
0 345 (46.9) 883 (54.1) 0.14 345 (46.9) 342 (46.6) 0.01
1–10 206 (28.0) 381 (23.4) 0.11 206 (28.0) 198 (27.0) 0.02
11–20 88 (12.0) 169 (10.4) 0.05 88 (12.0) 85 (11.6) 0.01
>20 96 (13.1) 199 (12.2) 0.03 96 (13.1) 109 (14.9) 0.05
Medication use in preceding 120 days
Benzodiazepines 385 (52.4) 835 (51.2) 0.02 385 (52.4) 380 (51.7) 0.01
Antipsychotics 175 (23.8) 344 (21.1) 0.07 175 (23.8) 165 (22.5) 0.03
Antidepressants 495 (67.3) 1,058 (64.8) 0.05 495 (67.3) 474 (64.5) 0.06
Other CNS depressants 1–5 (0.14–0.68) 1–5 (0.06–0.31) 0.01 1–5 (0.14–0.68) 1–5 (0.14–0.68) 0.03
Gabapentinoid 182 (24.8) 341 (20.9) 0.09 182 (24.8) 168 (22.9) 0.05
CYP3A4 inhibitors 63 (8.6) 158 (9.7) 0.04 63 (8.6) 69 (9.4) 0.03
CYP3A4 inducers 51 (6.9) 79 (4.8) 0.09 51 (6.9) 38 (5.1) 0.08
Opioids other than fentanyl 546 (74.3) 1,163 (71.3) 0.07 546 (74.3) 544 (74.0) 0.01
Residence in long‐term care facility 121 (16.5) 264 (16.2) 0.01 121 (16.5) 103 (14.1) 0.07
Income quintile
1 (lowest) 229 (31.2) 496 (30.4) 0.02 229 (31.2) 224 (30.4) 0.02
2 168 (22.9) 371 (22.7) 0.00 168 (22.9) 178 (24.3) 0.03
3 133 (18.1) 291 (17.8) 0.01 133 (18.1) 128 (17.5) 0.02
4 116 (15.8) 253 (15.5) 0.01 116 (15.8) 108 (14.7) 0.03
5 84 (11.4) 213 (13.1) 0.05 84 (11.4) 93 (12.7) 0.04
Rural residence 140 (19.0) 306 (18.8) 0.01 140 (19.0) 137 (18.6) 0.01
Open in a new tab
a

Unweighted values represent the crude distribution in the matched sample, while weighted values account for matching with replacement, where controls could be matched to multiple cases. Weighted and unweighted case counts are identical because each case contributes to only one matched set. Controls were weighted by the inverse of the number of controls matched to each case, such that the total control weight within each matched set equaled 1. The weighted effective sample size for controls equals the number of cases after adjusting for controls' repeated use across matched sets.

Figure 1.

Figure 1

Open in a new tab

Odds ratios and 95% confidence intervals for the association between macrolides and opioid toxicity among individuals receiving fentanyl, hydromorphone, or oxycodone. n = number of cases and N = total number of cases and controls within each exposure group.

In the hydromorphone cohort, cases were more likely than controls to have received a gabapentinoid in the 120 days prior to the index date (n = 1,200 [47.9%] vs. n = 1,066 [42.5%]; SD = 0.11) (Table 2 ). Following adjustment, recent CYP3A4‐inhibiting macrolide exposure was associated with an increased risk of opioid toxicity compared to no macrolide use (aOR 3.38; 95% CI: 1.28–8.97) (Figure 1 ). Results were similar following adjustment for calendar year (aOR 3.36, 95% CI: 1.27–8.90), and the effect of CYP3A4‐inhibiting macrolide exposure did not vary significantly across time periods (p‐value for interaction = 0.84). In contrast, no association was observed with azithromycin (aOR 1.07; 95% CI: 0.68–1.66) (Figure 1 ).

Table 2.

Characteristics of hydromorphone cases and controls

Variable Unweighted, No. (%)a Weighted, No. (%)a
Cases (n = 2,506) Controls (n = 7,488) Standardized difference Cases (n = 2,506) Controls (n = 2,506) Standardized difference
Age (median, IQR) 63 (52–76) 65 (54–77) 0.10 63 (52–76) 63 (52–76) 0.00
15–29 21 (0.8) 26 (0.4) 0.06 21 (0.8) 20 (0.8) 0.00
30–49 480 (19.2) 1,120 (15.0) 0.11 480 (19.2) 479 (19.1) 0.00
50–65 870 (34.7) 2,662 (35.6) 0.02 870 (34.7) 863 (34.4) 0.01
66–79 654 (26.1) 2,160 (28.9) 0.06 654 (26.1) 663 (26.5) 0.01
80+ 481 (19.2) 1,519 (20.3) 0.03 481 (19.2) 481 (19.2) 0.00
Female 1,423 (56.8) 4,414 (59.0) 0.04 1,423 (56.8) 1,423 (56.8) 0.00
Number of prescription drugs in previous year, (median, IQR) 16 (11–22) 16 (11–21) 0.04 16 (11–22) 16 (11–22) 0.01
Charlson Co‐morbidity Index
No hospitalization 843 (33.6) 2,658 (35.5) 0.04 843 (33.6) 803 (32.0) 0.03
0 657 (26.2) 1855 (24.8) 0.03 657 (26.2) 670 (26.7) 0.01
1 314 (12.5) 945 (12.6) 0.00 314 (12.5) 322 (12.9) 0.01
2+ 692 (27.6) 2030 (27.1) 0.01 692 (27.6) 712 (28.4) 0.02
Comorbidities in previous 5 years
Chronic kidney disease 494 (19.7) 1,395 (18.6) 0.03 494 (19.7) 465 (18.6) 0.03
Chronic liver disease and cirrhosis 235 (9.4) 630 (8.4) 0.03 235 (9.4) 254 (10.2) 0.03
Chronic lung disease 1,054 (42.1) 3,069 (41.0) 0.02 1,054 (42.1) 1,065 (42.5) 0.01
Rheumatoid arthritis 241 (9.6) 761 (10.2) 0.02 241 (9.6) 263 (10.5) 0.03
Seizure disorder 199 (7.9) 610 (8.2) 0.01 199 (7.9) 243 (9.7) 0.06
Stroke 410 (16.4) 1,205 (16.1) 0.01 410 (16.4) 415 (16.6) 0.01
Sleep apnea 135 (5.4) 426 (5.7) 0.01 135 (5.4) 146 (5.8) 0.02
Hypertension 1,577 (62.9) 4,671 (62.4) 0.01 1,577 (62.9) 1,488 (59.4) 0.07
Heart failure 540 (21.6) 1,612 (21.5) 0.00 540 (21.6) 534 (21.3) 0.01
Mood disorders 194 (7.7) 431 (5.8) 0.08 194 (7.7) 191 (7.6) 0.00
Substance use disorder 379 (15.1) 836 (11.2) 0.12 379 (15.1) 373 (14.9) 0.01
Anxiety disorder 241 (9.6) 631 (8.4) 0.04 241 (9.6) 253 (10.1) 0.02
Schizophrenia/Other psychotic disorders 79 (3.2) 141 (1.9) 0.08 79 (3.2) 68 (2.7) 0.03
Deliberate self‐harm 224 (8.9) 281 (3.8) 0.21 224 (8.9) 177 (7.1) 0.07
Other mental health conditions 90 (3.6) 178 (2.4) 0.07 90 (3.6) 82 (3.3) 0.02
Psychiatrist visit in past 180 days 337 (13.5) 831 (11.1) 0.07 337 (13.5) 337 (13.5) 0.00
Number of physician visits in past year (median, IQR) 14 (8–22) 14 (8–22) 0.03 14 (8–22) 14 (8–23) 0.03
Number of emergency department visits in prior year (median, IQR) 2 (1–4) 2 (0–3) 0.11 2 (1–4) 2 (1–4) 0.03
0 600 (23.9) 2005 (26.8) 0.07 600 (23.9) 603 (24.1) 0.00
1 507 (20.2) 1,626 (21.7) 0.04 507 (20.2) 514 (20.5) 0.01
2–5 1,010 (40.3) 2,935 (39.2) 0.02 1,010 (40.3) 1,008 (40.2) 0.00
6–10 263 (10.5) 678 (9.1) 0.05 263 (10.5) 276 (11.0) 0.02
>10 126 (5.0) 244 (3.3) 0.09 126 (5.0) 105 (4.2) 0.04
Days in hospital in prior year (median, IQR) 1 (0–11) 3 (0–10) 0.13 1 (0–11) 3 (0–11) 0.05
0 1,009 (40.3) 3,262 (43.6) 0.07 1,009 (40.3) 986 (39.3) 0.02
1–10 841 (33.6) 2,427 (32.4) 0.02 841 (33.6) 874 (34.9) 0.03
11–20 270 (10.8) 741 (9.9) 0.03 270 (10.8) 264 (10.5) 0.01
>20 386 (15.4) 1,059 (14.1) 0.04 386 (15.4) 382 (15.2) 0.00
Medication use in preceding 120 days
Benzodiazepines 1,032 (41.2) 2,987 (39.9) 0.03 1,032 (41.2) 1,076 (42.9) 0.04
Antipsychotics 619 (24.7) 1,586 (21.2) 0.08 619 (24.7) 585 (23.3) 0.03
Antidepressants 1,634 (65.2) 4,734 (63.2) 0.04 1,634 (65.2) 1,601 (63.9) 0.03
Other CNS depressants 13 (0.5) 20 (0.3) 0.04 13 (0.5) 9 (0.3) 0.03
Gabapentinoid 1,200 (47.9) 3,237 (43.2) 0.09 1,200 (47.9) 1,066 (42.5) 0.11
CYP3A4 inhibitors 187 (7.5) 601 (8.0) 0.02 187 (7.5) 199 (7.9) 0.02
CYP3A4 inducers 151 (6.0) 425 (5.7) 0.02 151 (6.0) 152 (6.1) 0.00
Opioids other than hydromorphone 1,093 (43.6) 3,244 (43.3) 0.01 1,093 (43.6) 1,168 (46.6) 0.06
Residence in long‐term care facility 247 (9.9) 755 (10.1) 0.01 247 (9.9) 234 (9.3) 0.02
Income quintile
1 (lowest) 965 (38.5) 2,773 (37.0) 0.03 965 (38.5) 946 (37.7) 0.02
2 553 (22.1) 1,616 (21.6) 0.01 553 (22.1) 531 (21.2) 0.02
3 394 (15.7) 1,268 (16.9) 0.03 394 (15.7) 425 (17.0) 0.03
4 339 (13.5) 1,041 (13.9) 0.01 339 (13.5) 346 (13.8) 0.01
5 253 (10.1) 782 (10.4) 0.01 253 (10.1) 255 (10.2) 0.00
Rural residence 376 (15.0) 1,195 (16.0) 0.03 376 (15.0) 402 (16.1) 0.03
Open in a new tab
a

Unweighted values represent the crude distribution in the matched sample, while weighted values account for matching with replacement, where controls could be matched to multiple cases. Weighted and unweighted case counts are identical because each case contributes to only one matched set. Controls were weighted by the inverse of the number of controls matched to each case, such that the total control weight within each matched set equaled 1. The weighted effective sample size for controls equals the number of cases after adjusting for controls' repeated use across matched sets.

Among oxycodone‐treated patients, cases were more likely to have been diagnosed with a mood disorder (n = 106 [15.6%] vs. n = 75 [11.1%]; SD = 0.13) and anxiety disorder (n = 87 [12.8%] vs. n = 63 [9.3%]; SD = 0.11) in the preceding 5 years (Table 3 ). Conversely, controls were more likely to have received a gabapentinoid (n = 128 [18.8%] vs. n = 97 [14.3%]; SD = 0.12) (Table 3 ). Following multivariable adjustment, recent CYP3A4‐inhibiting macrolide exposure was associated with an increased risk of opioid toxicity (aOR 3.11, 95% CI: 1.05–9.51). Results were similar following adjustment for calendar year (aOR 3.11, 95% CI: 1.20–8.03), and the effect of CYP3A4‐inhibiting macrolide exposure did not vary significantly across time periods (p‐value for interaction = 0.51). In contrast, no association was observed with azithromycin (aOR 0.77, 95% CI: 0.14–2.95) (Figure 1 ).

Table 3.

Characteristics of oxycodone cases and controls

Variable Unweighted, No. (%)a Weighted, No. (%)a
Cases (n = 678) Controls (n = 1767) Standardized difference Cases (n = 678) Controls (n = 678) Standardized difference
Age (median, IQR) 53 (45–65) 53 (45–66) 0.02 53 (45–65) 53 (44–65) 0.01
15–29 15 (2.2) 35 (2.0) 0.02 15 (2.2) 16 (2.3) 0.01
30–49 248 (36.6) 635 (35.9) 0.01 248 (36.6) 247 (36.5) 0.00
50–65 253 (37.3) 649 (36.7) 0.01 253 (37.3) 250 (36.9) 0.01
66–79 126 (18.6) 353 (20.0) 0.04 126 (18.6) 129 (19.1) 0.01
80+ 36 (5.3) 95 (5.4) 0.00 36 (5.3) 36 (5.3) 0.00
Female 357 (52.7) 951 (53.8) 0.02 357 (52.7) 357 (52.7) 0.00
Number of prescription drugs in previous year, (median, IQR) 15 (10–20) 14 (9–19) 0.14 15 (10–20) 14 (10–19) 0.04
Charlson co‐morbidity Index
No hospitalization 292 (43.1) 927 (52.5) 0.19 292 (43.1) 307 (45.4) 0.05
0 205 (30.2) 440 (24.9) 0.12 205 (30.2) 198 (29.3) 0.02
1 83 (12.2) 174 (9.9) 0.08 83 (12.2) 72 (10.7) 0.05
2+ 98 (14.5) 226 (12.8) 0.05 98 (14.5) 100 (14.7) 0.01
Comorbidities in previous 5 years
Chronic kidney disease 69 (10.2) 122 (6.9) 0.12 69 (10.2) 55 (8.0) 0.07
Chronic liver disease and cirrhosis 51 (7.5) 76 (4.3) 0.14 51 (7.5) 36 (5.3) 0.09
Chronic lung disease 264 (38.9) 597 (33.8) 0.11 264 (38.9) 253 (37.3) 0.03
Rheumatoid arthritis 69 (10.2) 203 (11.5) 0.04 69 (10.2) 76 (11.3) 0.04
Seizure disorder 53 (7.8) 137 (7.8) 0.00 53 (7.8) 62 (9.1) 0.05
Stroke 77 (11.4) 174 (9.9) 0.05 77 (11.4) 69 (10.1) 0.04
Sleep apnea 36 (5.3) 91 (5.2) 0.01 36 (5.3) 37 (5.5) 0.00
Hypertension 324 (47.8) 844 (47.8) 0.00 324 (47.8) 332 (48.9) 0.02
Heart failure 87 (12.8) 179 (10.1) 0.09 87 (12.8) 77 (11.4) 0.05
Mood disorders 106 (15.6) 152 (8.6) 0.22 106 (15.6) 75 (11.1) 0.13
Substance use disorder 139 (20.5) 246 (13.9) 0.18 139 (20.5) 127 (18.8) 0.04
Anxiety disorder 87 (12.8) 133 (7.5) 0.18 87 (12.8) 63 (9.3) 0.11
Schizophrenia/Other psychotic disorders 25 (3.7) 39 (2.2) 0.09 25 (3.7) 18 (2.7) 0.06
Deliberate self‐harm 116 (17.1) 163 (9.2) 0.24 116 (17.1) 105 (15.5) 0.04
Other mental health conditions 25 (3.7) 45 (2.6) 0.07 25 (3.7) 23 (3.3) 0.02
Psychiatrist visit in past 180 days 128 (18.9) 248 (14.0) 0.13 128 (18.9) 115 (17.0) 0.05
Number of physician visits in past year (median, IQR) 16 (10–26) 15 (9–23) 0.10 16 (10–26) 16 (10–24) 0.07
Number of emergency department visits in prior year (median, IQR) 1 (0–4) 1 (0–3) 0.14 1 (0–4) 1 (0–3) 0.05
0 200 (29.5) 673 (38.1) 0.18 200 (29.5) 223 (32.8) 0.07
1 150 (22.1) 380 (21.5) 0.02 150 (22.1) 149 (21.9) 0.01
2–5 234 (34.5) 542 (30.7) 0.08 234 (34.5) 217 (32.0) 0.05
6–10 64 (9.4) 123 (7.0) 0.09 64 (9.4) 64 (9.5) 0.00
>10 30 (4.4) 49 (2.8) 0.09 30 (4.4) 26 (3.8) 0.03
Days in hospital in prior year (median, IQR) 0 (0–6) 0 (0–4) 0.16 0 (0–6) 0 (0–6) 0.01
0 355 (52.4) 1,087 (61.5) 0.19 355 (52.4) 362 (53.4) 0.02
1–10 208 (30.7) 464 (26.3) 0.10 208 (30.7) 217 (32.0) 0.03
11–20 50 (7.4) 92 (5.2) 0.09 50 (7.4) 37 (5.4) 0.08
>20 65 (9.6) 124 (7.0) 0.09 65 (9.6) 63 (9.2) 0.01
Medication use in preceding 120 days
Benzodiazepines 426 (62.8) 1,007 (57.0) 0.12 426 (62.8) 408 (60.1) 0.06
Antipsychotics 172 (25.4) 327 (18.5) 0.17 172 (25.4) 149 (21.9) 0.08
Antidepressants 456 (67.3) 1,131 (64.0) 0.07 456 (67.3) 453 (66.8) 0.01
Other CNS depressants 1–5 (0.15–0.74) 1–5 (0.06–0.28) 0.03 1–5 (0.15–0.74) 1–5 (0.15–0.74) 0.04
Gabapentinoid 97 (14.3) 275 (15.6) 0.04 97 (14.3) 128 (18.8) 0.12
CYP3A4 inhibitors 56 (8.3) 126 (7.1) 0.04 56 (8.3) 51 (7.5) 0.03
CYP3A4 inducers 39 (5.8) 86 (4.9) 0.04 39 (5.8) 38 (5.7) 0.00
Opioids other than oxycodone 492 (72.6) 1,241 (70.2) 0.05 492 (72.6) 489 (72.2) 0.01
Bupropion 41 (6.1) 92 (5.2) 0.04 41 (6.1) 34 (5.1) 0.04
Duloxetine 48 (7.1) 120 (6.8) 0.01 48 (7.1) 54 (8.0) 0.04
Fluoxetine 23 (3.4) 58 (3.3) 0.01 23 (3.4) 26 (3.8) 0.02
Paroxetine 46 (6.8) 96 (5.4) 0.06 46 (6.8) 43 (6.4) 0.02
Residence in long‐term care facility 14 (2.1) 32 (1.8) 0.02 14 (2.1) 10 (1.5) 0.05
Income quintile
1 (lowest) 258 (38.1) 672 (38.0) 0.00 258 (38.1) 259 (38.1) 0.00
2 152 (22.4) 421 (23.8) 0.03 152 (22.4) 164 (24.1) 0.04
3 110 (16.2) 270 (15.3) 0.03 110 (16.2) 106 (15.6) 0.02
4 89 (13.1) 233 (13.2) 0.00 89 (13.1) 87 (12.8) 0.01
5 66 (9.7) 162 (9.2) 0.02 66 (9.7) 59 (8.7) 0.03
Rural residence 78 (11.5) 245 (13.9) 0.07 78 (11.5) 91 (13.4) 0.06
Open in a new tab
a

Unweighted values represent the crude distribution in the matched sample, while weighted values account for matching with replacement, where controls could be matched to multiple cases. Weighted and unweighted case counts are identical because each case contributes to only one matched set. Controls were weighted by the inverse of the number of controls matched to each case, such that the total control weight within each matched set equaled 1. The weighted effective sample size for controls equals the number of cases after adjusting for controls' repeated use across matched sets.

Sensitivity analyses

The E‐values for the odds ratio estimates were 7.06 for fentanyl, 6.22 for hydromorphone, and 5.67 for oxycodone. The E‐values for the lower bounds of the 95% confidence interval were 2.06 for fentanyl, 1.88 for hydromorphone, and 1.28 for oxycodone.

In the probabilistic bias analysis, the median bias‐adjusted odds ratios for the association between recent macrolide exposure and opioid toxicity were 3.86 (95% CI: 1.41–10.27) for fentanyl, 4.45 (95% CI: 1.23–9.51) for hydromorphone, and 3.18 (95% CI: 1.14–8.72) for oxycodone (Figures S4–S6 ). These findings were consistent across alternative bias term distributions and under a more conservative bias range of ±40% variation (Table S4 ).

In the deterministic bias analysis, the prevalence of substance use disorder among unmatched and matched cases was 27.2% and 10.9% for fentanyl‐treated individuals, 33.4% and 15.1% for hydromorphone‐treated individuals, and 28.6% and 20.5% for oxycodone‐treated individuals. For recent psychiatric care, the corresponding values were 19.0% and 12.7%, 25.1% and 13.4%, and 23.6% and 18.9%, respectively (Tables 4 , 5 , 6 ). Using a literature‐derived summary odds ratio of 6.21 (95% CI: 4.60–8.38) for the association between substance use disorder and prescription opioid toxicity, 7 the bias‐adjusted odds ratios were 5.88 for fentanyl, 5.18 for hydromorphone, and 3.74 for oxycodone. Based on a summary odds ratio of 4.07 (95% CI: 3.50–4.72) for psychiatric comorbidity, 7 the respective adjusted estimates were 4.33, 4.24, and 3.38.

DISCUSSION

In our population‐based study, we found that use of CYP3A4‐inhibiting macrolides was associated with an increased risk of opioid toxicity and death in patients treated with fentanyl, hydromorphone, or oxycodone. As expected, we found no such risk with azithromycin, an antibiotic with similar clinical indications but devoid of CYP3A4‐inhibiting activity. These findings were unlikely influenced by unmeasured confounding, stable in probabilistic bias analyses accounting for potential selection and misclassification bias, and further supported by deterministic analyses that quantified the impact of differential inclusion of high‐risk individuals. Overall, our findings support the notion of a potentially life‐threatening drug interaction between CYP3A4‐inhibiting macrolides and fentanyl, hydromorphone, and oxycodone.

Our findings build on prior research demonstrating that CYP3A4 inhibition can increase serum concentrations of fentanyl and oxycodone and alter hydromorphone metabolism. Although these data provide a biologically plausible and mechanistic basis for a drug interaction, they are derived under controlled experimental conditions in healthy volunteers who may not represent opioid‐treated patients in practice. Moreover, findings from pharmacokinetic studies do not always translate into clinically meaningful interactions. 41 , 42 Similarly, case reports describing opioid toxicity following co‐prescription of fentanyl or oxycodone with CYP3A4‐inhibiting macrolides are limited by their anecdotal nature, lack of comparators, and inability to control for confounding related to comorbidity and polypharmacy. 43 , 44 Our work extends this evidence by quantifying the real‐world consequences of these interactions in a population‐based study. The inclusion of azithromycin as a neutral exposure strengthens causal inference by demonstrating the absence of associations where none are expected.

Our findings have several important clinical and public health implications. First, while unregulated fentanyl is the primary driver of opioid‐related mortality, 1 , 4 prescription opioids continue to contribute to overdose deaths and hospitalizations. Furthermore, despite declining over time, prescribing of opioids remains common. In the United States, over 12 million individuals aged 18–64 were dispensed at least one opioid prescription in 2020–2021, representing 6.4% of the population. 45 Similarly, over 1.3 million Ontario residents, or ~8% of the provincial population, were dispensed an opioid for pain in 2024. 46 Second, although use of CYP3A4‐inhibiting macrolides, particularly erythromycin, declined markedly over the study period, clarithromycin remains guideline‐recommended first‐line therapy for Helicobacter pylori eradication, a common indication affecting millions of patients annually in North America, and for nontuberculous mycobacterial infections.

Thus, the risk of serious drug interactions between CYP3A4‐dependent opioids and clarithromycin remains particularly relevant in these specific clinical scenarios. In addition, our temporal sensitivity analyses demonstrated that effect estimates remained stable following adjustment for calendar year, with nonsignificant tests for effect modification by study period. These analyses suggest that the risk of the drug interaction is similar regardless of calendar period. Third, the small number of macrolide exposed cases, particularly in later years of the study period, suggests the absolute burden of preventable prescription opioid deaths and hospitalizations attributable to interactions with CYP3A4‐inhibiting macrolides is likely very small in contemporary practice, particularly in North America, where use of erythromycin and clarithromycin has declined over time. However, market analyses demonstrate that global use of these drugs is expected to increase. Specifically, the clarithromycin global market is forecast to grow from ~2.1 billion United States dollars (USD) in 2024 to 4.5 billion USD in 2032, representing a compound annual growth rate (CAGR) of 9.8%. 47 The Asia‐Pacific region is expected to experience the fastest growth, with a projected CAGR of 7% over the forecast period. 47 Similarly, the erythromycin market is expected to grow at a CAGR of 3.8%, from 1.8 billion USD in 2024 to 2.5 billion USD in 2031. 48 As with clarithromycin, growth is expected to be particularly strong in Asia due to the rising prevalence of bacterial infections and the increasing demand for antibiotics in countries such as China and India. Furthermore, moderate growth is expected in the Middle East, Latin America and Africa due to improving access to healthcare facilities and a growing focus on infectious disease management. 48 Thus, while the number of macrolide exposed individuals in our study was small, our findings may have greater relevance in regions where clarithromycin and erythromycin use is expected to increase. Fourth, the mechanism underlying this interaction is not specific to macrolides. We focused our study on macrolide antibiotics because they are short‐course therapies with overlapping indications, allowing us to define a narrow, clinically plausible exposure window around the index date. Moreover, azithromycin provides a negative control, as it has similar indications to CYP3A4‐inhibiting macrolides but is not expected to interact with opioids. However, our findings likely extend to other CYP3A4 inhibitors. Thus, our study has broader implications for CYP3A4 mediated interactions with opioids, with a similar risk for opioid toxicity and a potentially higher burden existing with other commonly prescribed CYP3A4 inhibitors. Finally, although our study focused on prescribed fentanyl in patients with chronic noncancer pain, similar interactions are expected among individuals receiving the drug for cancer‐related pain and those using illicitly manufactured fentanyl who are co‐prescribed CYP3A4‐inhibiting macrolides and other CYP3A4 inhibitors in routine care. Given the high and variable fentanyl doses in the unregulated drug supply and that nonpharmaceutical fentanyl accounts for most opioid toxicity in North America, our findings may underestimate the population‐level impact of this drug interaction. 1 , 4 , 49 Moreover, fatal opioid overdose resulting from drug interactions is unlikely to be recognized as the cause of death, particularly among individuals with substance use disorder, who are at higher risk of these events. Overall, our study highlights a potentially unrecognized and entirely preventable drug interaction as a component cause of fatal and nonfatal opioid toxicity among patients treated with fentanyl, hydromorphone, or oxycodone. This risk can be mitigated through the use of drugs that do not inhibit CYP3A4.

Our study has some limitations. First, we used administrative data and had no access to information regarding illicit drug use, alcohol use, nonprescribed medications or treatment adherence. However, these limitations apply equally to CYP3A4‐inhibiting macrolides and azithromycin. Moreover, our findings were robust to unmeasured confounding in E‐value analyses. Second, our analysis was restricted to fatal events and outcomes resulting in hospital encounters, and we therefore did not identify cases of opioid toxicity managed in ambulatory or prehospital settings. Our study may therefore underestimate the consequences of this drug interaction. Third, our study population comprised individuals eligible for public drug coverage in Ontario, which may not be representative of all individuals using opioids, particularly younger adults without public drug coverage. Furthermore, individuals with public drug coverage may have greater medical complexity and polypharmacy relative to the general population. However, the biological mechanism of CYP3A4 inhibition would be expected to operate similarly across age groups and drug coverage mechanisms. Fourth, matching rates in our studies were low, resulting in underrepresentation of individuals with mental health comorbidities and substance disorder, groups at elevated risk of opioid toxicity, in our main analyses. Our findings may therefore underestimate the burden of macrolide–opioid interactions. This differential case inclusion likely biases our estimates toward the null, as unmatched cases had higher baseline risk of opioid toxicity. Our deterministic bias analyses resulted in higher estimates, supporting this interpretation. Moreover, probabilistic bias analyses showed consistent findings, suggesting that our results are robust to plausible bias scenarios. Fifth, the number of macrolide exposed cases was small, particularly for oxycodone, thereby limiting statistical precision. Accordingly, our findings should be interpreted as hypothesis‐generating. However, the consistency of effect estimates across opioids, the null findings with azithromycin and the biological plausibility of the interaction support a cautious approach in which clinicians consider non‐CYP3A4‐inhibiting drugs in patients receiving fentanyl, oxycodone or hydromorphone, pending confirmation with larger studies. Sixth, our 26‐year study period could have potentially introduced heterogeneity in prescribing practices, as reflected by the decreased use of CYP3A4‐inhibiting macrolides over time. However, our findings were stable in temporal sensitivity analyses, suggesting that the biological mechanism for the interaction and effect size is consistent over time. Finally, we could not conduct separate analyses for erythromycin and clarithromycin due to the small number of cases exposed to the former drug. However, both are potent CYP3A4 inhibitors with similar expected effects on opioid metabolism.

In conclusion, we found that CYP3A4‐inhibiting macrolides were associated with a marked increase in the risk of opioid overdose and death among individuals treated with fentanyl, hydromorphone, or oxycodone. These findings likely extend to individuals receiving other potent CYP3A4 inhibitors and individuals exposed to illicitly manufactured fentanyl, which utilizes the same metabolic pathway. 26 , 27 No such risk was observed with azithromycin. Although the proportion of unmatched cases was high and our findings are based on a small number of exposed cases, clinicians should consider alternative antibiotics that do not inhibit CYP3A4, such as azithromycin, for patients receiving these opioids. When clarithromycin and other CYP3A4 inhibitors are required, close monitoring for signs of opioid toxicity is required.

FUNDING

This study was funded by the Canadian Institutes of Health Research. Tara Gomes holds a Canada Research Chair in Drug Policy and Substance Use.

CONFLICT OF INTEREST

Mina Tadrous has received consulting fees for unrelated work from Green Shield Canada and the Canadian Agency for Drugs and Technologies in Health. Tara Gomes has received funding from the Ontario MOH, the Ontario College of Pharmacists, and Canada's Drug Agency, and consulting fees/honoraria from Indigenous Services Canada and the Province of British Columbia for unrelated work. Joanne Ho has received funding from the Ontario MOH for unrelated work. All other authors declared no competing interests for this work.

AUTHOR CONTRIBUTIONS

All authors wrote the manuscript. TA, JY, DNJ, JMWH, MM, FW, MT, and TG designed the research. TA, JY, FW, and TG performed the research. JY and TA analyzed the data.

Supporting information

Table S1.

CPT-120-122-s001.docx (121.2KB, docx)

ACKNOWLEDGMENTS

This study was supported by ICES, which is funded by an annual grant from the Ontario Ministry of Health (MOH) and the Ministry of Long‐Term Care (MLTC). Parts of this material are based on data and information compiled and provided by the Ontario Ministry of Health, Ontario Health (OH) and the Canadian Institute for Health Information (CIHI). The analyses, conclusions, opinions, and statements expressed herein are solely those of the authors and do not reflect those of the funding or data sources; no endorsement is intended or should be inferred. This document used data adapted from the Statistics Canada Postal CodeOM Conversion File, which is based on data licensed from Canada Post Corporation, and/or data adapted from the Ontario Ministry of Health Postal Code Conversion File, which contains data copied under license from ©Canada Post Corporation and Statistics Canada. We thank IQVIA Solutions Canada Inc. for the use of their Drug Information File.

DATA AVAILABILITY STATEMENT

The dataset from this study is held securely in coded form at ICES. While legal data sharing agreements between ICES and data providers (e.g., healthcare organizations and government) prohibit ICES from making the dataset publicly available, access may be granted to those who meet prespecified criteria for confidential access, available at www.ices.on.ca/DAS.

References

  • 1. Substance‐related Overdose and Mortality Surveillance Task Group on behalf of the Council of Chief Medical Officers of Health . Opioid‐ and Stimulant‐Related Harms in Canada (Public Health Agency of Canada, Ottawa, 2025). <https://health‐infobase.canada.ca/substance‐related‐harms/opioids‐stimulants/>. Accessed August 31, 2025. [Google Scholar]
  • 2. Ledlie, S. , Juurlink, D.N. , Tadrous, M. , Mamdani, M. , Paterson, J.M. & Gomes, T. Opioid‐related deaths between 2019 and 2021 across 9 Canadian provinces and territories. CMAJ 196, E469–E476 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3. National Institute on Drug Abuse . Drug Overdose Deaths: Facts and Figures <https://nida.nih.gov/research‐topics/trends‐statistics/overdose‐death‐rates#Fig3>. Accessed August 31, 2025.
  • 4. Tanz, L.J. , Stewart, A. , Gladden, R.M. , Ko, J.Y. , Owens, L. & O'Donnell, J. Detection of illegally manufactured fentanyls and carfentanil in drug overdose deaths – United States, 2021–2024. MMWR Morb. Mortal. Wkly Rep. 73, 1099–1105 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5. Gomes, T. et al. Patterns of Medication and Healthcare Use Among People Who Died of an Opioid‐Related Toxicity during the COVID‐19 Pandemic in Ontario (Ontario Drug Policy Research Network, Toronto, 2022). [Google Scholar]
  • 6. Zedler, B.K. , Saunders, W.B. , Joyce, A.R. , Vick, C.C. & Murrelle, E.L. Validation of a screening risk index for serious prescription opioid‐induced respiratory depression or overdose in a US commercial health plan claims database. Pain Med. 19, 68–78 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7. Brady, J.E. , Giglio, R. , Keyes, K.M. , DiMaggio, C. & Li, G. Risk markers for fatal and non‐fatal prescription drug overdose: a meta‐analysis. Inj. Epidemiol. 4, 24 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8. Feierman, D.E. & Lasker, J.M. Metabolism of fentanyl, a synthetic opioid analgesic, by human liver microsomes: role of CYP3A4. Drug Metab. Dispos. 24, 932–939 (1996). [PubMed] [Google Scholar]
  • 9. Tateishi, T. , Krivoruk, Y. , Ueng, Y.F. , Wood, A.J. , Guengerich, F.P.  &  Wood, M. Identification of human liver cytochrome P‐450 3A4 as the enzyme responsible for fentanyl and sufentanil N‐dealkylation. Anesth. Analg. 82, 167–172 (1996). [DOI] [PubMed] [Google Scholar]
  • 10. Labroo, R.B. , Paine, M.F. , Thummel, K.E. & Kharasch, E.D. Fentanyl metabolism by human hepatic and intestinal cytochrome P450 3A4: implications for interindividual variability in disposition, efficacy, and drug interactions. Drug Metab. Dispos. 25, 1072–1080 (1997). [PubMed] [Google Scholar]
  • 11. Lalovic, B. , Phillips, B. , Risler, L.L. , Howald, W. & Shen, D.D. Quantitative contribution of CYP2D6 and CYP3A to oxycodone metabolism in human liver and intestinal microsomes. Drug Metab. Dispos. 32, 447–454 (2004). [DOI] [PubMed] [Google Scholar]
  • 12. Lalovic, B. , Kharasch, E. , Hoffer, C. , Risler, L. , Liu‐Chen, L.Y. & Shen, D.D. Pharmacokinetics and pharmacodynamics of oral oxycodone in healthy human subjects: role of circulating active metabolites. Clin. Pharmacol. Ther. 79, 461–479 (2006). [DOI] [PubMed] [Google Scholar]
  • 13. Benetton, S.A. , Borges, V.M. , Chang, T.K. & McErlane, K.M. Role of individual human cytochrome P450 enzymes in the in vitro metabolism of hydromorphone. Xenobiotica 34, 335–344 (2004). [DOI] [PubMed] [Google Scholar]
  • 14. de Wildt, S.N. , Kearns, G.L. , Leeder, J.S. & van den Anker, J.N. Cytochrome P450 3A: ontogeny and drug disposition. Clin. Pharmacokinet. 37, 485–505 (1999). [DOI] [PubMed] [Google Scholar]
  • 15. Samer, C.F. et al. The effects of CYP2D6 and CYP3A activities on the pharmacokinetics of immediate release oxycodone. Br. J. Pharmacol. 160, 907–918 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16. Samer, C.F. et al. Genetic polymorphisms and drug interactions modulating CYP2D6 and CYP3A activities have a major effect on oxycodone analgesic efficacy and safety. Br. J. Pharmacol. 160, 919–930 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17. Liukas, A. , Hagelberg, N.M. , Kuusniemi, K. , Neuvonen, P.J. & Olkkola, K.T. Inhibition of cytochrome P450 3A by clarithromycin uniformly affects the pharmacokinetics and pharmacodynamics of oxycodone in young and elderly volunteers. J. Clin. Psychopharmacol. 31, 302–308 (2011). [DOI] [PubMed] [Google Scholar]
  • 18. Nieminen, T.H. et al. Oxycodone concentrations are greatly increased by the concomitant use of ritonavir or lopinavir/ritonavir. Eur. J. Clin. Pharmacol. 66, 977–985 (2010). [DOI] [PubMed] [Google Scholar]
  • 19. Nieminen, T.H. et al. Grapefruit juice enhances the exposure to oral oxycodone. Basic Clin. Pharmacol. Toxicol. 107, 782–788 (2010). [DOI] [PubMed] [Google Scholar]
  • 20. Saari, T.I. et al. Effects of itraconazole on the pharmacokinetics and pharmacodynamics of intravenously and orally administered oxycodone. Eur. J. Clin. Pharmacol. 66, 387–397 (2010). [DOI] [PubMed] [Google Scholar]
  • 21. Hagelberg, N.M. et al. Voriconazole drastically increases exposure to oral oxycodone. Eur. J. Clin. Pharmacol. 65, 263–271 (2009). [DOI] [PubMed] [Google Scholar]
  • 22. Ziesenitz, V.C. , König, S.K. , Mahlke, N.S. , Skopp, G. , Haefeli, W.E. & Mikus, G. Pharmacokinetic interaction of intravenous fentanyl with ketoconazole. J. Clin. Pharmacol. 55, 708–717 (2015). [DOI] [PubMed] [Google Scholar]
  • 23. Saari, T.I. , Laine, K. , Neuvonen, M. , Neuvonen, P.J. & Olkkola, K.T. Effect of voriconazole and fluconazole on the pharmacokinetics of intravenous fentanyl. Eur. J. Clin. Pharmacol. 64, 25–30 (2008). [DOI] [PubMed] [Google Scholar]
  • 24. Olkkola, K.T. , Palkama, V.J. & Neuvonen, P.J. Ritonavir's role in reducing fentanyl clearance and prolonging its half‐life. Anesthesiology 91, 681–685 (1999). [DOI] [PubMed] [Google Scholar]
  • 25. Pergolizzi, J.V. Jr. et al. The prevalence of opioid‐related major potential drug‐drug interactions and their impact on health care costs in chronic pain patients. J. Manag. Care Spec. Pharm. 20, 467–476 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26. Kong, L. & Walz, A.J. Identification of human cytochrome P450 isozymes involved in the oxidative metabolism of carfentanil. Toxicol. Lett. 343, 28–33 (2021). [DOI] [PubMed] [Google Scholar]
  • 27. Wilde, M. et al. Metabolic pathways and potencies of new fentanyl analogs. Front. Pharmacol. 10, 238 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28. Gomes, T. et al. Comparing the contribution of prescribed opioids to opioid‐related hospitalizations across Canada: a multi‐jurisdictional cross‐sectional study. Drug Alcohol Depend. 191, 86–90 (2018). [DOI] [PubMed] [Google Scholar]
  • 29. Ontario Ministry of Health . Publicly Funded Drug Programs <https://www.health.gov.on.ca/en/pro/programs/drugs/funded_drug/fund_odbp.aspx>. Accessed August 31, 2025.
  • 30. Singh, S. , Martins, D. , Khuu, W. , Tadrous, M. , Gomes, T. & Juurlink, D. Characteristics of opioid‐related deaths in Ontario, Canada: leveraging the drug and drug/alcohol related death (DDARD) database. Int. J. Popul. Data Sci. 3, 4 (2018). [Google Scholar]
  • 31. Special Advisory Committee on the Epidemic of Opioid Overdoses . Opioid‐ and Stimulant‐Related Harms in Canada (Public Health Agency of Canada, Ottawa, 2021). <https://health‐infobase.canada.ca/substance‐related‐harms/opioids‐stimulants/>. Accessed August 31, 2025. [Google Scholar]
  • 32. Canadian Institute for Health Information . Opioid‐Related Harms in Canada, December 2018 (CIHI, Ottawa, 2018). [Google Scholar]
  • 33. Green, C.A. et al. Identifying and classifying opioid‐related overdoses: a validation study. Pharmacoepidemiol. Drug Saf. 28, 1127–1137 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34. Arbogast, P.G. , Kaltenbach, L. , Ding, H. & Ray, W.A. Adjustment for multiple cardiovascular risk factors using a summary risk score. Epidemiology 19, 30–37 (2008). [DOI] [PubMed] [Google Scholar]
  • 35. Yunusa, I. , Gagne, J.J. , Yoshida, K. & Bykov, K. Risk of opioid overdose associated with concomitant use of oxycodone and selective serotonin reuptake inhibitors. JAMA Netw. Open 5, e220194 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36. Westphal, J.F. Macrolide‐induced clinically relevant drug interactions with cytochrome P‐450A (CYP) 3A4: an update focused on clarithromycin, azithromycin and dirithromycin. Br. J. Clin. Pharmacol. 50, 285–295 (2000). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37. Austin, P.C. , Grootendorst, P. & Anderson, G.M. A comparison of the ability of different propensity score models to balance measured variables between treated and untreated subjects: a Monte Carlo study. Stat. Med. 26, 734–753 (2007). [DOI] [PubMed] [Google Scholar]
  • 38. VanderWeele, T.J. & Ding, P. Sensitivity analysis in observational research: introducing the E‐value. Ann. Intern. Med. 167, 268–274 (2017). [DOI] [PubMed] [Google Scholar]
  • 39. Lash, T.L. , Fox, M.P. , MacLehose, R.F. , Maldonado, G. , McCandless, L.C. & Greenland, S. Good practices for quantitative bias analysis. Int. J. Epidemiol. 43, 1969–1985 (2014). [DOI] [PubMed] [Google Scholar]
  • 40. Fox, M.P. , MacLehose, R.F. & Lash, T.L. Applying Quantitative Bias Analysis to Epidemiologic Data (Springer, New York, 2021). [Google Scholar]
  • 41. Bartlett, J.W. , Renner, E. , Mouland, E. , Barnes, G.D. , Kuo, L. & Ha, N.B. Clinical safety outcomes in patients with nonvalvular atrial fibrillation on rivaroxaban and diltiazem. Ann. Pharmacother. 53, 21–27 (2019). [DOI] [PubMed] [Google Scholar]
  • 42. Juurlink, D.N. , Mamdani, M.M. , Kopp, A. , Herrmann, N. & Laupacis, A. A population‐based assessment of the potential interaction between serotonin‐specific reuptake inhibitors and digoxin. Br. J. Clin. Pharmacol. 59, 102–107 (2005). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 43. Okumura, S.I. , Ishihara, Y. , Yoshino, K. & Echizen, H. Possible drug interaction between oxycodone and erythromycin: a case report and analyses of spontaneous reporting systems of adverse drug reactions. Rinsho yakuri/Japanese J. Clin. Pharmacol. Therap. 49, 205–209 (2018). [Google Scholar]
  • 44. Cronnolly, B. & Pegrum, H. Fentanyl–clarithromycin interaction. BMJ Case Rep. bcr0220125936 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 45. Moriya, A.S. & Fang, Z. Any Use and “Frequent Use” of Opioids among Adults Aged 18–64 in 2020–2021, by Socioeconomic Characteristics. Statistical Brief #552 (Agency for Healthcare Research and Quality, Rockville, MD, 2023). <https://meps.ahrq.gov/data_files/publications/st552/stat552.shtml>. Accessed August 31, 2025. [Google Scholar]
  • 46. Ontario Drug Policy Research Network . Ontario Opioid Indicator Tool (Ontario Drug Policy Research Network, Toronto, 2025). <https://odprn.ca/ontario‐opioid‐indicator‐tool/>. [Google Scholar]
  • 47. Cognitive Market Research and Consulting . Clarithromycin Market Trends and Future Opportunities <https://www.cognitivemarketresearch.com/articles/clarithromycin‐market‐trends‐and‐future‐opportunities>. Accessed March 3, 2026.
  • 48. 6W Research . Global Erythromycin Market (2025–2031) <https://www.6wresearch.com/industry‐report/global‐erythromycin‐market>. Accessed March 3, 2026.
  • 49. Larnder, A. et al. Variability in the unregulated opioid market in the context of extreme rates of overdose. Drug Alcohol Depend. 235, 109427 (2022). [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Table S1.

CPT-120-122-s001.docx (121.2KB, docx)

Data Availability Statement

The dataset from this study is held securely in coded form at ICES. While legal data sharing agreements between ICES and data providers (e.g., healthcare organizations and government) prohibit ICES from making the dataset publicly available, access may be granted to those who meet prespecified criteria for confidential access, available at www.ices.on.ca/DAS.

Articles from Clinical Pharmacology and Therapeutics are provided here courtesy of Wiley and American Society for Clinical Pharmacology and Therapeutics

RESOURCES