Risk of opioid overdose associated with concurrent use of hydrocodone and CYP3A4-inhibiting calcium channel blockers: a population-based cohort study

Abstract

Nondihydropyridine calcium channel blockers (CCBs), including diltiazem and verapamil, inhibit cytochrome P450 3A4 (CYP3A4), an enzyme involved in the metabolism of hydrocodone, the most commonly used opioid in the United States. This study evaluated whether concomitant use of hydrocodone with CYP3A4-inhibiting CCBs increases the risk of opioid overdose compared to use of hydrocodone with amlodipine, a CCB that does not inhibit CYP3A4. Using 3 US databases (2000-2021), two cohorts were identified: (1) hydrocodone initiation while on CCB; and (2) CCB initiation while on hydrocodone. The outcome was hospitalization or emergency department visits for opioid overdose. Propensity score matching weights were applied to balance confounders, and Cox regression estimated HRs, pooled using random-effects meta-analysis. In hydrocodone initiation cohort (mean age 61.2 years; 53.8% female), weighted incidence rates were 2.8 and 2.6 per 1000 person-years, with a weighted HR of 1.07 (95% CI, 0.90-1.29). In CCB initiation cohort (mean age 55.2 years; 59.9% female), weighted incidence rates were 6.5 and 6.0, yielding an HR of 1.08 (95% CI, 0.88-1.32). The pooled HR was 1.07 (95% CI, 0.94-1.23). Concomitant use of hydrocodone with CYP3A4-inhibiting CCBs was not associated with an increased risk of opioid overdose relative to amlodipine.

Trial registration:

Real World Evidence Registry- https://osf.io/ybdt5

Introduction

Opioid prescribing is a major public health issue, with a substantial impact of opioids on mortality.^1-3 In 2023, approximately 125 million opioid prescriptions were issued in the United States,⁴ with a prescribing rate of 37.5 prescriptions per 100 people.⁵ Due to stricter prescribing regulations,⁶ monitoring,⁷ and improved access to treatment and naloxone,⁸ prescription opioid-related overdose deaths declined from 3.4 to 2.9 per 100 000 population between 2018 and 2023, ref lecting a 23.7% reduction.⁹ However, prescription opioids still accounted for about 13% of all opioid-related deaths.¹⁰ Opioids have the potential to interact pharmacokinetically and pharmacodynamically with various other medications.^11-13 Previous studies have explored drug-drug interactions involving opioids and confirmed their contribution to overdose risk.^14-18 Nonetheless, not all pharmacologically identified interactions lead to clinically meaningful outcomes. Given the narrow therapeutic index of opioids,¹³ and their susceptibility to pharmacologic interactions with a wide range of medications, further research is warranted to clarify the clinical impact of specific drug combinations.^19,20

Calcium channel blockers (CCBs) represent a cornerstone in the pharmacological management of hypertension and are also commonly recommended for arrhythmia management.^21,22 The high prevalence of cardiovascular comorbidities such as hypertension and arrhythmia, combined with the frequent use of opioids for pain management in these patients, has led to the common co-administration of CCBs and opioids.^23-25 Certain CCBs, namely diltiazem and verapamil, inhibit key hepatic enzymes involved in drug metabolism, including the cytochrome P450 (CYP) 3A4 enzyme.^26-29 Although direct evidence on interactions between hydrocodone and CYP3A4-inhibiting CCBs is limited, their concomitant use raises pharmacologic concerns due to the metabolic pathway of hydrocodone. CYP3A4 primarily metabolizes hydrocodone into norhydrocodone, an inactive metabolite,³⁰ and its inhibition may lead to elevated hydrocodone exposure and an increased risk of opioid-related adverse events. As hydrocodone is one of the most frequently used opioids, understanding the clinical implications of this potential interaction is essential to improve patient safety and guide prescribing practice for these commonly prescribed medications.^31,32

Thus, in this population-based cohort study, we aimed to assess the comparative risk of opioid overdose in patients with concomitant use of hydrocodone and CYP3A4-inhibiting CCBs. To minimize bias, an active comparator was selected from the same therapeutic class, namely amlodipine, a CCB which has minimal CYP3A4 inhibitory activity.³³

Methods

Data source

We analyzed data from three large US healthcare claims databases covering individuals enrolled in commercial and public health insurance plans between 2000 and 2021. Specifically, we retrieved deidentified individual patient-level data from Merative MarketScan (2003-2021), Medicare fee-for-service (2016-2019), and Medicaid (2000-2018). These databases contain longitudinal administrative records of pharmacy and medical claims for commercially insured individuals (Merative MarketScan), publicly insured individuals with low income (Medicaid), and individuals aged 65 years and older in Medicare fee-for-service (Medicare) across the United States.

Study population and design

To reduce bias in observational studies of medications, we employed an active-comparator, new-user cohort design.^34-36 As part of this approach, we constructed 2 separate cohorts by identifying patients who initiated hydrocodone while receiving a CCB and those who initiated a CCB while receiving hydrocodone (Figure S1). This approach is recommended for the evaluation of drug-drug interactions, since even if the biological effect of a potential drug-drug interaction may not depend on the sequence of drug initiation, the order in which the drugs are initiated can reflect different clinical contexts and prescribing behaviors, and may be associated with differences in patient characteristics and confounding patterns.^34,35 The hydrocodone initiation cohort included individuals who initiated hydrocodone while on a CCB, following at least 180 days in the insurance plan with no hydrocodone use. The date of hydrocodone initiation was set as the index date. The CCB initiation cohort included individuals who initiated a study CCB (diltiazem, verapamil, or amlodipine) while on hydrocodone, following at least 180 days without prior CCB use. The CCB initiation date was also set as the index date. In both cohorts, individuals were required to be at least 18 years of age on the index date and have no record of an opioid overdose during the 180 days before the index (baseline period). We excluded individuals with missing age or sex, and residents in US territories such as Guam, the Mariana Islands, Puerto Rico, and the Virgin Islands, which have different healthcare environments.³⁷ To ensure reliable claims and prescription data, we restricted the Medicaid population to individuals younger than 65 years at the index date, as Medicaid enrollees typically become eligible for Medicare upon turning 65.

Patients who initiated both hydrocodone and CCB on the same day were assigned to the hydrocodone initiation cohort, since opioid overdose risk was expected to be mainly influenced by hydrocodone initiation. The protocol of this study has been pre-published in the real-world evidence registry (https://osf.io/ybdt5) for reproducibility and transparency.

Exposures

We evaluated drug initiation and continuing use based on pharmacy records of dispensation fills, which provide information on the dispensing date, dispensed quantity, and pharmacist-recorded days’ supply. Concomitant use was defined based on overlapping days’ supply with 14-day grace periods added to account for minor nonadherence.

In both cohorts, patients were assigned to treatment groups based on CCBs they received (hydrocodone initiation cohort) or initiated (CCB initiation cohort) on the index date (Figure S1). Diltiazem and verapamil, recognized as moderate CYP3A4 inhibitors, were considered CYP3A4-inhibiting CCBs.^38,39 Amlodipine was chosen as the active comparator because it has minimal potential for inhibiting the CYP3A4 enzyme and is a commonly used CCB.³³

Outcome definition and follow-up

The outcome was opioid overdose, defined as hospitalizations or emergency department (ED) visits with International Classification of Diseases, 9th revision (ICD-9), or corresponding 10th revision (ICD-10) codes, indicating opioid poisoning (Table S1). The ICD-9 codes used to define opioid overdose demonstrated a positive predictive value (PPV) of 84.6%,⁴⁰ while ICD-10-based definitions showed comparable performance in emergency department settings, with positive predictive values of 79.4% for opioid poisonings, excluding heroin.⁴¹

In the hydrocodone initiation cohort, follow-up began on the index date to capture early events potentially attributable to hydrocodone initiation. Follow-up continued until the occurrence of the outcome, end of insurance enrollment, death, end of data availability within each database, switching to the other CCB exposure group, or discontinuation of either hydrocodone or the index CCB. In the CCB initiation cohort, follow-up began on the day after the index date to reflect some delay in outcome occurrence due to CYP3A4 inhibition and to ensure temporality. Censoring criteria were the same as described above.

Covariates

To mitigate the impact of nonrandom allocation of study individuals to exposure groups, we adjusted for 167 covariates that were considered potential confounders or proxy indicators of confounders. These variables included calendar year of index date and demographic variables such as age, sex, race/ethnicity (available in Medicare and Medicaid data only), and state of residence. Additional variables included detailed measures of opioid use on the index date and during the baseline period, as well as comorbidities, other medications, and healthcare utilization, measured during the 180 days prior to index (baseline). Morphine milligram equivalents (MME) were used to estimate total opioid dosage dispensed.⁴² For the index hydrocodone dispensing, we assessed total MME for all dispensed hydrocodone pills. In patients who had other opioids dispensed during the 180-day baseline period prior to the index date, we assessed the type of opioid and total MMEs dispensed, stratified by proximity to the index date (ie, within 60 days before the index date, and within 61-180 days before the index date) to capture more recent exposure, as well as overall exposure over the baseline period. Measured comorbidities included a broad range of conditions such as pain conditions, cancer, diabetes, liver disease, kidney dysfunction, opioid abuse or dependence, comorbidity score, and frailty.^43,44 Since CCBs represent a pharmacologically diverse class, we adjusted for a broad range of cardiovascular conditions, including arrhythmia and heart failure, to account for potential differences in clinical use. All variables are listed in Table S2.

Statistical analyses

Analyses were conducted separately in the hydrocodone initiation cohort and the CCB initiation cohort. Within each database and cohort, we estimated a propensity score (PS) using a multivariable logistic regression model with all the baseline variables included as covariates and the probability of receiving a CYP3A4-inhibiting CCB (diltiazem/verapamil) vs amlodipine as the dependent variable. To adjust for confounding, we used PS matching weights, an approach that approximates PS matching without discarding study individuals.^45,46 Patients were assigned a weight that was determined by the ratio of the minimum of the 2 predicted probabilities (diltiazem/verapamil [PS] or amlodipine [1-PS]) and the predicted probability of the treatment actually received.⁴⁶ The PS and matching weights were calculated within each database. We assessed covariate balance between the groups, within each database-specific and the overall population pooled across databases, using standardized differences. Imbalances of potential significance were defined as absolute mean standardized differences (SMD) exceeding 0.1.

Cumulative incidence curves were used to assess the cumulative incidence over time for each treatment group. We also calculated both crude (unweighted) and weighted overdose incidence rates per 1000 person-years. We used a Cox proportional hazards regression model, stratified by database, to estimate cohort-specific unweighted and weighted HR with 95% CI, comparing the risk associated with the use of hydrocodone with CYP3A4-inhibiting CCBs versus amlodipine. Cohort-specific estimates were pooled for the overall effect estimate using a random-effect meta-analysis.⁴⁷

Several sensitivity analyses were conducted to test the robustness of the findings. First, to mitigate the impact of informative censoring and since exposure misclassification is possible, we conducted a 60-day intention-to-treat analysis, in which patients were followed for 60 days regardless of treatment changes and censored only at the end of data, death, or end of insurance enrollment. Given the typically short duration of opioid use, we chose a 60-day follow-up period for the intention-to-treat analysis to capture outcomes attributable to exposure while minimizing exposure misclassification due to hydrocodone discontinuation or treatment switch. Second, since a substantial proportion of hydrocodone dispensing in our cohorts were for fewer than 14 days, we conducted a sensitivity analysis applying a shorter, 7-day grace period (end of days’ supply without subsequent refill within 7 days). Third, since not all overdoses may be coded as poisoning and there are no hydrocodone-specific codes, we implemented 2 alternative outcome definitions: (1) excluding poisoning codes specific to heroin, opium, and methadone from the outcome definition; (2) adding ICD-9 and ICD-10 codes for unspecified adverse effects of opioids (Table S1). Fourth, given the higher likelihood that diltiazem and verapamil are prescribed for arrhythmic conditions relative to amlodipine, we conducted a sensitivity analysis excluding patients with baseline evidence of arrhythmia. Specifically, we excluded individuals with diagnosis codes for atrial fibrillation, arrhythmias, or prescriptions for antiarrhythmic agents. Lastly, because opioid treatment may differ in certain populations, we further excluded patients with (1) cancer or (2) substance abuse or dependence, defined as opioid dependence or abuse, other drug poisonings, other or unspecified drug abuse, alcohol abuse, or use of buprenorphine.

In subgroup analyses, we stratified the study population based on individual CYP3A4-inhibiting CCB (diltiazem, verapamil), age (18-44, 45-64, 65+ years), sex (male, female), cardiovascular disease (defined as atrial fibrillation, arrhythmia, cerebrovascular disease, coronary artery disease, heart failure, or peripheral artery disease), and respiratory disease (defined as asthma, chronic obstructive pulmonary disease, home oxygen use, and prior respiratory distress), and chronic opioid use (prescriptions for more than 90 days at baseline). For each subgroup analysis, the PS was reestimated, patients were reweighted, and cohort-specific estimates were calculated as described earlier.

The study was approved by the Mass General Brigham Institutional Review Board (No. 2020P002865). All statistical analyses were conducted using SAS version 9.4 (SAS Institute, NC).

Results

Baseline characteristics

Overall, there were 4 264 415 eligible individuals who initiated hydrocodone while on a CCB and 680 519 eligible individuals who initiated a CCB while on hydrocodone during the study period (2000-2021). Patients in the CCB initiation cohort were younger than patients in the hydrocodone-initiation cohort, but more likely to have chronic pain conditions, and prior use of other opioids and CNS depressants (Table 1).

Table 1.

Selected baseline characteristics for study cohort.

Variable	Individuals who initiated hydrocodone while on CCB				Individuals who initiated CCB while on hydrocodone
	Unweighted			Weighted	Unweighted			Weighted
	Diltiazem/ Verapamil	Amlodipine (Reference)	Std diff	Std diff	Diltiazem/ Verapamil	Amlodipine (Reference)	Std diff	Std diff
Number of patients	1011 683	3 252 732			142 463	538 056
Demographics
Age, years, mean (SD)	62.0 (14.3)	61.0 (13.8)	0.07	0.01	53.9 (15.7)	55.6 (14.1)	0.11	0.01
Sex, n (%)
Female	616 875 (61.0)	1677 548 (51.6)	0.19	0.00	94 496 (66.3)	312 991 (58.2)	0.17	0.00
Index hydrocodone and prior opioid use
Index hydrocodone, MME, mean (SD)	469.7 (1309.1)	493 (1653.6)	0.02	0.00
Hydrocodone for most recent prescription before or on the index date, MME, mean (SD)					1959.4 (11 313.7)	2723.6 (15 262.4)	0.06	0.00
Continuous hydrocodone exposure for at least 90 days, n (%)					38 316 (26.9)	152 931 (28.4)	0.03	0.00
Prior use of other opioids, n (%)	270 057 (26.7)	827 132 (25.4)	0.03	0.00	54 320 (38.1)	185 746 (34.5)	0.08	0.00
Total opioids use in 60 days prior to index, MME, mean (SD)	310.0 (1120.8)	251.5 (1017.5)	0.06	0.00	4811.1 (24 511.1)	6311.4 (32 475)	0.05	0.00
Pain conditions, n (%)
Abdominal pain	167 946 (16.6)	517 993 (15.9)	0.02	0.00	32 901 (23.1)	115 687 (21.5)	0.04	0.00
Back and neck pain	290 860 (28.8)	918 653 (28.2)	0.01	0.00	64 787 (45.5)	256 185 (47.6)	0.04	0.00
Diabetic neuropathy	42 761 (4.2)	171 051 (5.3)	0.05	0.00	7060 (5.0)	37 842 (7.0)	0.09	0.01
Fibromyalgia	32 637 (3.2)	78 326 (2.4)	0.05	0.00	10 445 (7.3)	29 961 (5.6)	0.07	0.00
Headache or migraine	107 801 (10.7)	246 625 (7.6)	0.11	0.00	38 283 (26.9)	81 812 (15.2)	0.29	0.01
Osteoarthritis	205 743 (20.3)	659 406 (20.3)	0.00	0.00	29 014 (20.4)	122 412 (22.8)	0.06	0.00
Musculoskeletal injuiy	89 919 (8.9)	288 311 (8.9)	0.00	0.00	12 752 (9.0)	47 322 (8.8)	0.01	0.00
Other comorbidities, n (%)
Combined comorbidity score, mean (SD)	1.0 (2.3)	0.7 (2.2)	0.13	0.01	1.6 (2.7)	1 (2.5)	0.22	0.01
Cardiovascular disease	398 226 (39.4)	963 071 (29.6)	0.21	0.00	63 610 (44.7)	170 812 (31.7)	0.27	0.01
Atrial fibrillation	179 730 (17.8)	168 186 (5.2)	0.40	0.00	28 649 (20.1)	22 960 (4.3)	0.50	0.00
Stable angina	32 073 (3.2)	91 544 (2.8)	0.02	0.00	6452 (4.5)	19 791 (3.7)	0.04	0.01
Heart failure	88 950 (8.8)	202 095 (6.2)	0.10	0.00	19 656 (13.8)	48 696 (9.1)	0.15	0.01
Ischemic heart disease	176 398 (17.4)	534 541 (16.4)	0.03	0.01	27 685 (19.4)	91 396 (17.0)	0.06	0.01
Other arrhythmia	121 406 (12.0)	255 555 (7.9)	0.14	0.00	22 319 (15.7)	37 933 (7.1)	0.27	0.00
Respiratory disease	396 581 (39.2)	998 708 (30.7)	0.18	0.00	76 065 (53.4)	232 368 (43.2)	0.21	0.01
Home oxygen use	52 484 (5.2)	80 356 (2.5)	0.14	0.00	12 382 (8.7)	22 027 (4.1)	0.19	0.00
Prior respiratory distress, hospitalization	73 038 (7.2)	175 628 (5.4)	0.08	0.00	23 247 (16.3)	54 479 (10.1)	0.18	0.00
Diabetes	245 204 (24.2)	867 605 (26.7)	0.06	0.01	31 006 (21.8)	135 646 (25.2)	0.08	0.01
Depression	119 814 (11.8)	360 801 (11.1)	0.02	0.00	28 190 (19.8)	92 646 (17.2)	0.07	0.00
Anxiety	90 310 (8.9)	283 446 (8.7)	0.01	0.00	23 671 (16.6)	81 469 (15.1)	0.04	0.00
Cancer	95 565 (9.4)	286 491 (8.8)	0.02	0.00	11 561 (8.1)	37 808 (7.0)	0.04	0.00
CKD stage 3 or dialysis	54 410 (5.4)	248 630 (7.6)	0.09	0.01	7011 (4.9)	38 981 (7.2)	0.10	0.00
Liver disease	47 469 (4.7)	178 862 (5.5)	0.04	0.00	9681 (6.8)	40 266 (7.5)	0.03	0.00
Sleep apnea	93 868 (9.3)	292 977 (9.0)	0.01	0.00	11 150 (7.8)	36 896 (6.9)	0.04	0.00
Tobacco use	52 856 (5.2)	216 839 (6.7)	0.06	0.00	7274 (5.1)	24 109 (4.5)	0.03	0.00
Opioid dependence	4581 (0.5)	20 666 (0.6)	0.03	0.00	2542 (1.8)	13 134 (2.4)	0.05	0.00
Other medication use, n (%)
Beta blockers	290 722 (28.7)	1294 329 (39.8)	0.24	0.02	51 138 (35.9)	208 583 (38.8)	0.06	0.00
COPD or asthma medications	254 536 (25.2)	574 108 (17.7)	0.18	0.00	49 645 (34.8)	141 626 (26.3)	0.19	0.01
Anticonvulsants	72 379 (7.2)	166 074 (5.1)	0.09	0.00	19 198 (13.5)	43 315 (8.1)	0.18	0.00
Atypical antipsychotic	41 747 (4.1)	128 076 (3.9)	0.01	0.00	11 884 (8.3)	40 595 (7.5)	0.03	0.00
Anxiolytics	54 940 (5.4)	159 688 (4.9)	0.02	0.00	14 105 (9.9)	48 512 (9.0)	0.03	0.00
Benzodiazepines	222 880 (22.0)	567 175 (17.4)	0.12	0.00	55 988 (39.3)	176 600 (32.8)	0.14	0.01
CNS stimulants	14 412 (1.4)	42 172 (1.3)	0.11	0.00	4033 (2.8)	11 071 (2.1)	0.05	0.00
Sedatives or hypnotics	84 105 (8.3)	232 955 (7.2)	0.04	0.00	19 988 (14.0)	65 408 (12.2)	0.06	0.00
SSRI	208 083 (20.6)	572 009 (17.6)	0.08	0.00	37 676 (26.4)	122 173 (22.7)	0.09	0.00
Tricyclic antidepressant	62 070 (6.1)	128 490 (4.0)	0.10	0.00	16 638 (11.7)	40 711 (7.6)	0.14	0.00
CYP3A4 inducers	61 687 (6.1)	174 139 (5.4)	0.03	0.01	9662 (6.8)	31 179 (5.8)	0.04	0.01
CYP3A4 inhibitors	279 465 (27.6)	804 047 (24.7)	0.07	0.02	47 647 (33.4)	159 282 (29.6)	0.08	0.01
CYP2D6 inhibitors	417 319 (41.2)	1 155 942 (35.5)	0.12	0.01	77 922 (54.7)	258 198 (48.0)	0.14	0.01
Gabapentinoids	119 895 (11.9)	394 011 (12.1)	0.01	0.00	30 008 (21.1)	121 854 (22.6)	0.04	0.00
Healthcare utilization, mean (SD)
Number of distinct generics	11.8 (5.8)	10.9 (5.4)	0.17	0.02	14.2 (7.2)	12.9 (6.8)	0.18	0.02
Number of physician visits	8.7 (10.7)	7.8 (9.7)	0.09	0.01	13.4 (16.3)	11.5 (14.3)	0.13	0.01
Number of hospitalizations	0.3 (0.7)	0.2 (0.6)	0.06	0.01	0.5 (1.0)	0.4 (0.9)	0.13	0.01
Number of emergency room visits	0.8 (1.7)	0.7 (1.5)	0.03	0.00	1.7 (3.2)	1.5 (2.8)	0.09	0.00

Abbreviations: CKD, chronic kidney disease; CNS, central nervous system; COPD, chronic obstructive pulmonary disease; CYP, Cytochrome P450; MME, morphine milligram equivalents; NSAIDs, nonsteroidal anti-inflammatory drugs; SSRI, selective serotonin reuptake inhibitors; Std diff, standardized difference.

In the hydrocodone initiation cohort, 1 011 683 patients were on diltiazem/verapamil (diltiazem: 637 765 [63.0%]; verapamil: 373 918 [37.0%]), and 3 252 732 on amlodipine (Figure 1). Before weighting, the mean age was similar between the diltiazem/verapamil group (62.0 years [SD 14.3]) and the amlodipine group (61.0 years [SD 13.8]), but a higher percentage of females was observed in the diltiazem/verapamil group (diltiazem/verapamil: 61.0% vs amlodipine: 51.6%). Both groups showed a similar prevalence of mental health-related disorders, prior nonhydrocodone opioid use, and hydrocodone MME on the index date. However, more patients in the diltiazem/verapamil group had atrial fibrillation or other arrhythmia, as well as respiratory disease. After applying PS weighting, the baseline characteristics were well balanced between the 2 treatment groups, with all SMD below 0.1 (Table 1, Tables S2-S5, and Figure S2).

Figure 1.

Flow of study cohort selection. Abbreviations: CCB, calcium channel blockers; CYP, cytochrome P450.

In the CCB initiation cohort, 142 463 patients initiated diltiazem/verapamil (diltiazem: 87 267 [61.3%]; verapamil: 55 196 [38.7%]), and 538 056 initiated amlodipine while on hydrocodone (Figure 1). As in the hydrocodone-initiation cohort, the mean age was comparable between the diltiazem/verapamil group (53.9 years [SD 15.7]) and the amlodipine group (55.6 years [SD 14.1]), with a higher percentage of females in the diltiazem/verapamil group (diltiazem/verapamil: 66.3% vs amlodipine: 58.2%). Also, in line with the hydrocodone-initiation cohort, patients initiating diltiazem/verapamil were more likely to have atrial fibrillation, other arrhythmia, and respiratory disease. Almost 30% of patients in the CCB-initiation cohort had used hydrocodone for at least 90 days before the index date (Table 1). After PS weighting, baseline characteristics between the 2 groups were well balanced, with all SMD being less than 0.1 (Table 1, Tables S2-S5, and Figure S3).

Risk of opioid overdose

In the hydrocodone initiation cohort, we observed a total of 771 opioid overdose events over a mean follow-up of 24.3 days (SD 34.0), with 208 events occurring in the diltiazem/verapamil group and 563 events in the amlodipine group. Most patients were censored due to hydrocodone discontinuation (Table S6). Weighted incidence rates were 2.8 per 1000 person-years (95% CI, 2.4-3.3) in the diltiazem/verapamil group and 2.6 per 1000 person-years (95% CI, 2.2-3.1) in the amlodipine group, yielding a weighted HR of 1.07 (95% CI, 0.90-1.29) (Table 2, Figure 2, and Figure S4). Detailed results, before and after weighting, from the overall cohort and individual databases are provided in Tables S7-S10.

Table 2.

Concomitant use of hydrocodone and CYP3A4-inhibiting calcium channel blockers (diltiazem/verapamil) vs amlodipine and the risk of opioid overdose, before and after weighting.

Analysis/Exposure group	CYP3A4-inhibiting CCB (Diltiazem or Verapamil)				CYP3A4 noninhibiting CCB (Amlodipine)				HR (95% CI)
	Number of individuals	Number of events	Incidence rate per 1000 person-years (95% CI)		Number of individuals	Number of events	Incidence rate per 1000 person-years (95% CI)		Unweighted	Weighted	Pooleda
			Unweighted	Weighted			Unweighted	Weighted
Individuals who initiated hydrocodone while on CCB (hydrocodone initiation cohort)
Diltiazem/verapamil vs amlodipine	1011 683	208	2.9 (2.5-3.3)	2.8 (2.4-3.3)	3 252 732	563	2.5 (2.3-2.7)	2.6 (2.2-3.1)	1.22 (1.04-1.44)	1.07 (0.90-1.29)	1.07 (0.94-1.23)
Individuals who initiated CCB while on hydrocodone (CCB initiation cohort)
Diltiazem/verapamil vs amlodipine	142 463	164	7.0 (6.0-8.2)	6.5 (5.5-7.8)	538 056	531	5.3 (4.8-5.7)	6.0 (5.0-7.2)	1.33 (1.12-1.59)	1.08 (0.88-1.32)

^aMeta-analysis with a random-effect model was conducted using estimates from 2 cohorts.

Figure 2.

Weighted cumulative incidence curve of opioid overdose.

In the CCB-initiation cohort, we observed 695 opioid overdose events over a mean follow-up of 65.7 days (SD 118.6), with 164 events occurring in the diltiazem/verapamil group and 531 in the amlodipine group. The weighted incidence rate was 6.5 per 1000 person-years (95% CI, 5.5-7.8) in the diltiazem/verapamil group, while the amlodipine group had a weighted incidence rate of 6.0 per 1000 person-years (95% CI, 5.0-7.2). The weighted HR was 1.08 (95% CI, 0.88-1.32) (Table 2, Figure 2, and Figure S4). Reasons for censoring and detailed findings from the overall cohort and individual databases are available in Tables S6-S10.

A random-effects meta-analysis of the 2 cohorts (the hydrocodone-initiation cohort, and the CCB-initiation cohort) resulted in a weighted HR of 1.07 (95% CI, 0.94-1.23) (Table 2).

The results from the sensitivity analyses aligned with the primary findings, though a potential small increase in overdose risk was suggested when a shorter gap between hydrocodone prescriptions was applied (weighted HR 1.15 [95%, CI, 0.99-1.34]) (Table S11-S17). Subgroup analyses were largely consistent with the primary finding. However, elevated HRs were observed in specific subgroups: young adults (HR 1.39 [95% CI, 0.89-2.17]) and those receiving verapamil (HR 1.30 [95% CI, 1.00-1.70]) in the hydrocodone initiation cohort, as well as older adults (HR 1.58 [95% CI, 1.02-2.46]) in the CCB initiation cohort. In the pooled analysis of 2 cohorts, increased HRs were also noted among adults aged 65 years or older (weighted HR 1.24 [95% CI, 0.83-1.86]) and patients with chronic opioid use (weighted HR 1.18 [95% CI, 0.98-1.42]) (Figure 3 and Table S18).

Figure 3.

Drug-specific and subgroup results of concomitant use of hydrocodone and CYP3A4-inhibiting CCBs (diltiazem/verapamil) vs amlodipine and the risk of opioid overdose. *Meta-analysis with a random-effect model was conducted using data from 2 cohorts.

Discussion

In this population-based cohort study, the concomitant use of hydrocodone and CYP3A4-inhibiting CCBs, diltiazem and verapamil, was not associated with a higher risk of opioid overdose compared with the concomitant use of hydrocodone and amlodipine. Overall, the incidence of opioid overdose was low in this population; however, it was twice as high among those who initiated CCBs while on hydrocodone compared with those who initiated hydrocodone while on CCBs, suggesting a higher risk in patients on chronic hydrocodone use. Nevertheless, the lack of a substantial association between CYP3A4-inhibiting CCBs and opioid overdose was consistent across both cohorts and in sensitivity analyses.

While opioids, particularly hydrocodone, are commonly used for pain management in the United States, evidence of clinical harm due to drug-drug interactions between opioids and CYP-inhibiting medications is limited and inconsistent. At least 1 experimental study using a mouse model reported that verapamil enhances the potentiation of morphine,⁴⁸ however, another study found that verapamil reduced the respiratory depressant effects of morphine.⁴⁹ A small randomized crossover trial showed an increased analgesic effect of morphine combined with verapamil, compared to either treatment alone,⁵⁰ whereas another small trial found that CCBs (diltiazem, verapamil, and nimodipine) had no effect on pain outcomes when used with morphine.⁵¹ Importantly, until now, no studies have evaluated the effects of adding CYP-inhibiting CCBs to hydrocodone, the most commonly used opioid in the United States. While our findings should be reassuring to patients who need to take hydrocodone concomitantly with either diltiazem or verapamil, caution should be taken when extrapolating our results to hydrocodone drug-drug interactions with other CYP inhibitors. Both diltiazem and verapamil are known as moderate CYP3A4 inhibitors; it is possible that strong inhibitors, when added to hydrocodone, may produce clinically relevant outcomes.

In subgroup analyses, we observed a modestly increased risk of opioid overdose among hydrocodone initiators receiving verapamil. Verapamil, a stronger in vivo CYP3A inhibitor than diltiazem or amlodipine, may reduce hydrocodone metabolism and increase systemic exposure.⁵² Prior observational data also indicate higher interaction potential for verapamil, including elevated bleeding risk compared to other CCBs in anticoagulant users.⁵³ We also observed an increased overdose risk among individuals aged 65 or older, who may have reduced hepatic clearance, heightened opioid sensitivity,⁵⁴ and more frequent polypharmacy.⁵⁵ Given the exploratory nature of these analyses, a cautious interpretation is warranted and future studies should be undertaken to further confirm or refute these findings.

In addition, while the relative risks associated with CYP3A4-inhibiting CCBs were almost identical across patients who initiated hydrocodone while on CCB and patients who initiated CCB while on hydrocodone, we observed different incidence rates in these 2 populations. More specifically, patients who were receiving hydrocodone at the time of CCB initiation experienced opioid overdose at a twice higher rate than patients who initiated hydrocodone. Several factors could explain this finding. Patients on existing hydrocodone therapy at the time of the coexposure to both agents were taking higher doses of hydrocodone than patients who were initiating hydrocodone. Moreover, they were also more likely to stay on hydrocodone longer than patients who initiated hydrocodone. Thus, despite the large number of patients censored for hydrocodone discontinuation in both cohorts, followup was twice as long in patients who were already on hydrocodone than in those who initiated hydrocodone. Differences in medication utilization patterns should also be considered. Patients initiating hydrocodone, particularly when prescribed for temporary pain conditions, are less likely to fully adhere to their prescription. In contrast, patients already receiving hydrocodone before initiating CCB therapy are more likely to continue prescribed hydrocodone as chronic users. The differences between these 2 patient populations were the main reason we evaluated them separately. In addition, evaluating these 2 cohorts separately allowed us to have better control of confounding and other sources of bias, including immortal person-time.³⁴

Some additional limitations should be noted. First, while we considered a wide range of covariates, the potential for residual confounding should be acknowledged. Important risk factors for opioid overdose, including socioeconomic status and illicit opioid use, were not available in our data.⁵⁶ Moreover, although we adjusted for measured confounders, residual confounding may persist due to the limited ability of claims data to capture clinical severity. For example, administrative codes do not distinguish between mild and severe forms of arrhythmia, which could influence the selection of specific CCBs. Nonetheless, the active-comparator design with PS-matching weights most likely mitigated confounding, and measured confounders have been shown to serve as proxies for unmeasured confounders in prior research.^57,58 Second, while prescription data include accurate information on medication dispensing, we lack information on patient adherence to prescribed medications. Patients might have stopped their CCB therapy earlier than what was assumed based on days’ supply information, or they might not have taken all their prescribed hydrocodone. In sensitivity analyses using a shorter gap between hydrocodone prescriptions, a small increase in risk was observed; however, the results were generally consistent with those of the primary analysis. We would not expect such potential exposure misclassification to differ substantially between the 2 treatment groups. Although nondifferential exposure misclassification typically biases effect estimates toward the null when independent of other errors or covariates, this assumption is derived from single-drug exposure settings and may not hold in the context of drug combinations.⁵⁹ Third, our study only captured outcomes for individuals who reached the hospital, potentially leading to an underestimation of the true incidence of opioid overdose. However, this outcome misclassification is likely to be nondifferential. Given the high specificity of opioid overdose diagnoses (99.9%),⁴⁰ the relative risks are expected to be unbiased.⁶⁰ Lastly, we could not distinguish between prescription opioid overdose and opioid overdose due to illicit drug use. However, the impact of the difference in the cause of opioid overdose on the results was limited, as we observed consistent results in sensitivity analyses using both a stricter definition of opioid overdose, excluding heroin, methadone, and opium poisoning, as well as a broader definition including opioid overdose-related adverse effects.

Conclusion

In this study of US health insurance claims data covering the 2000-2021 period, concomitant use of hydrocodone and CYP3A4-inhibiting CCBs was not associated with an increased risk of opioid overdose in comparison with concomitant use of hydrocodone and amlodipine. Sensitivity analyses using a shorter grace period yielded similar findings, although a small increase in risk could not be excluded. Overall, the incidence of opioid overdose was low but higher in patients who added a CCB to hydrocodone than in those who initiated hydrocodone with prior CCB exposure. In these 2 clinical scenarios involving a potential drug-drug interaction, initiating hydrocodone while taking CCBs and initiating CCBs while taking hydrocodone, this study provides reassurance to both patients and clinicians regarding the safe use of hydrocodone with either diltiazem or verapamil.

Supplementary Material

Supplementary material is available at American Journal of Epidemiology online.

Data availability

This study used deidentified data accessed under a data use agreement. In accordance with contractual terms and applicable privacy regulations, the data cannot be made publicly available.