Evaluating the Acute Effects of the Cannabinoid Dronabinol and the Opioid Hydromorphone Alone and in Combination: A Double-blind, Randomized, Placebo-controlled Trial in Knee Osteoarthritis

Abstract

Background:

This within-subject, double-blind, randomized, placebo-controlled study aimed to determine the acute analgesic and drug effects and the risk for extramedical use, of synthetic delta-9-tetrahydrocannabinol and hydromorphone, alone and in combination, in individuals with knee osteoarthritis.

Methods:

Participants (N = 21; 57% women; mean age = 63.4 ± 6.4 yr) with knee osteoarthritis received oral combinations of placebo, hydromorphone (2 mg), and dronabinol (10 mg). In the initial session, participants received hydromorphone + placebo, and the remaining sessions were randomized, with participants receiving placebo + placebo, dronabinol + placebo, or hydromorphone + dronabinol. Clinical and experimentally induced pain (quantitative sensory testing), physical and cognitive function, subjective drug ratings, and adverse events were evaluated at baseline and at 60, 120, 180, and 240 min after dosing.

Results:

For primary outcomes, hydromorphone produced greater pressure pain threshold analgesia than dronabinol (P = 0.029, ηp² = 0.074), greater capsaicin (P = 0.045, ηp² = 0.062), and noncapsaicin (P = 0.017, ηp² = 0.087) sensitized mechanical temporal summation analgesia than placebo. There were no significant drug-related differences for clinical pain severity (ηp² = 0.011), thermal threshold (ηp² = −0.025) or tolerance (ηp² = −0.008), temporal summation (ηp² = 0.009), cold pressor (ηp² = 0.056), conditioned pain modulation (ηp² = 0.038), capsaicin-induced thermal threshold (ηp² = −0.030), central sensitization (ηp² = 0.006), general pain sensitivity (ηp² = 0.021), or physical functioning (2-min walking distance [ηp² = 0.028], Timed Up and Go [ηp² = −0.027], and total stair climb time [ηp² = −0.005]; all P values > 0.05). For secondary outcomes, hydromorphone impaired working memory accuracy compared to all conditions and produced greater good effects than placebo (all P ≤ 0.005); hydromorphone + dronabinol impaired working memory reaction time and produced greater high ratings compared to placebo, greater drug effects than placebo and hydromorphone, and higher nausea than hydromorphone (all P < 0.05); and dronabinol had greater high ratings than hydromorphone (P = 0.001). There were no significant drug-related differences for fine motor movement, bad effects, drug liking, or adverse event occurrence or severity (all P > 0.05).

Conclusions:

Opioid and cannabinoid medications failed to produce robust analgesia in experimentally induced pain among patients with knee osteoarthritis. In contrast to preclinical studies, there was no evidence of synergistic analgesic effects by combining hydromorphone and dronabinol.

In people with knee osteoarthritis, hydromorphone and dronabinol—alone or combined—did not provide robust pain relief or show synergistic analgesic effects. Moreover, the combination did not yield clear benefits over opioid or cannabinoid alone, contrasting with preclinical findings and previous assumptions about their clinical potential.

Editor’s Perspective

What We Already Know about This Topic

• Previous preclinical and limited human studies have suggested that combining opioid (like hydromorphone) and cannabinoid (like delta-9-tetrahydrocannabinol/dronabinol) medications might enhance analgesia for conditions such as knee osteoarthritis, potentially letting patients use lower opioid doses

• Clinical evidence for any synergistic effect is mixed and mostly absent from rigorous, controlled trials

What This Article Tells Us That Is New

• In people with knee osteoarthritis, hydromorphone and dronabinol—alone or combined—did not provide robust pain relief or show synergistic analgesic effects

• Moreover, the combination did not yield clear benefits over opioid or cannabinoid alone, contrasting with preclinical findings and previous assumptions about their clinical potential

Pharmacologic treatments for knee osteoarthritis (e.g., joint injections and oral medications) often yield limited pain relief and benefit only a subset of people with knee osteoarthritis.¹ As interest grows in cannabis-based pain management in the context of the ongoing opioid crisis,² the combination of opioid and cannabinoid treatment has received increasing attention, fueled by both shifts in legal landscapes and the potential that these agents may produce synergistic effects on analgesia.³

Opioids remain a commonly used analgesic agent for a broad range of pain conditions. However, studies suggest that their effectiveness on chronic pain conditions in general, and knee osteoarthritis in particular, are small in magnitude and may incur a significant side-effect profile that includes respiratory depression, constipation, and nausea.^4,5 Additionally, many patients develop tolerance, putting them at risk for extramedical use and overdose,⁶ highlighting the necessity for alternative treatments that can enhance analgesia while minimizing opioid-induced burden.

Preclinical studies have suggested that coadministration of cannabinoids with opioids may enhance analgesia while allowing for lower opioid doses, therefore reducing the risks associated with opioid use.³ Moreover, it has been suggested that the endocannabinoid system involvement in osteoarthritis pain may indicate the utility of cannabis-based treatments.^7,8 Despite these findings, human studies evaluating the synergistic cannabinoid–opioid potential have been limited.⁹ Some studies have indicated the potential for pain reduction, improved quality of life, and reduced opioid consumption among individuals with knee osteoarthritis who use medical cannabis^10–13; however, the literature remains mixed and lacks high-quality evidence from controlled clinical trials at this time.¹⁴ Importantly, the synergistic potential of cannabinoids and opioids in clinical population remains poorly understood.

Our team recently conducted a randomized, double-blind, placebo-controlled trial combining synthetic delta-9-tetrahydrocannabinol cannabinoid (dronabinol, 10 mg) with a moderate dose of hydromorphone (4 mg, a classic full μ-opioid agonist) in individuals with knee osteoarthritis.¹⁵ We found minimal additive benefit for analgesia or physical function. Importantly, adverse events were frequently reported (35.1% of the entire laboratory sessions) particularly when participants received hydromorphone alone or hydromorphone + dronabinol, raising concerns about the tolerability of the drug pairing at the 4-mg hydromorphone dose.

To address this, the current study aimed to evaluate whether a lower but still therapeutic dose of hydromorphone (2 mg), when combined with 10 mg of dronabinol, would offer a more favorable balance of efficacy and tolerability. Using a rigorous, randomized, placebo-controlled laboratory design, we examined the effects of this combination on clinical and laboratory-induced pain, self-reported drug effects, physical and cognitive functioning, risk for extramedical use, and adverse events in individuals with knee osteoarthritis.

Materials and Methods

Participants

Individuals diagnosed with knee osteoarthritis were enrolled in the study from January 2021 to November 2023. Recruitment was conducted through a combination of local flyers, online platforms, and print/radio advertisements. A Consolidated Standards of Reporting Trials (CONSORT) flow diagram detailing participant enrollment can be found in figure 1. The study protocol received approval from the Johns Hopkins School of Medicine (Baltimore, Maryland) Institutional Review Board and was registered on ClinicalTrials.gov (NCT03098563; principal investigators: Kelly E. Dunn and Claudia M. Campbell; registered November 2017). All participants provided informed consent before their participation.

Fig. 1.

Study consort diagram. A/E, adverse event.

Study Design and Procedure

This phase II investigation employed a randomized, double-blind, placebo-controlled, within-subject design. Initial participant eligibility was assessed through both phone and in-person screenings.¹⁵ During the in-person screening, participants provided urine samples for drug testing and pregnancy testing for premenopausal women. Individuals with any reported opioid use in the past month were excluded. Eligibility criteria were further evaluated based on a review of medical history, a physical examination that included an electrocardiogram, and assessments of hepatic, hematologic, and biochemical function (detailed in supplemental table S1, https://links.lww.com/ALN/E385). A knee x-ray was also performed to determine the Kellgren−Lawrence score,¹⁶ a 5-point ordinal scale (ranging from 0 for no knee osteoarthritis to 4 for severe knee osteoarthritis) used to classify the severity of knee osteoarthritis. Additionally, participants completed self-report questionnaires and were familiarized with quantitative sensory testing (QST) during the screening process. Neither the participants nor the study staff were informed of the specific medications being tested. Instead, to maintain the double-blinding process, the informed consent process included a list of six possible study drugs.

Experimental Study Sessions

Eligible participants attended a total of four experimental study sessions, each spaced 7 or more days apart to allow for a washout period. Sessions began at approximately 8:00 am. Participants were instructed to avoid over-the-counter medications on the days of their sessions and to maintain consistent doses of any prescribed, noncontraindicated medications throughout the study period. Upon arrival, participants provided a urine sample, which was tested for drug use and pregnancy (when applicable). Participants were asked to consume a calorie- and fat-controlled breakfast (approximately 10 g of fat, participants provided with example menu) before arrival and to report details of this meal. Around 9:30 am, participants completed baseline assessments, including QST, physical and cognitive function evaluations, and self-report measures, which took about 60 min. Study medications, consisting of two oral capsules, were administered at approximately 11:00 am. After dosing, QST, function, and self-report assessments were repeated at +60, +120, +180, and +240 min.

Study Medications

Oral hydromorphone (2 mg; Sky Pharma, United Arab Emirates), dronabinol (10 mg; Akorn, USA), and placebo were overencapsulated in size 00 gelcaps to maintain blinding for both participants and experimenters regarding drug conditions. Hydromorphone was selected as a representative opioid due to its minimal involvement with differential CYP450 metabolism.¹⁷ The doses of hydromorphone and dronabinol were chosen because they fall within the U.S. Food and Drug Administration (FDA)–approved range for clinical use, making them suitable for prescription and hypothesized to produce analgesic effects.¹⁸ The study included one control condition: placebo + placebo (i.e., placebo) and three experimental conditions: hydromorphone + placebo (i.e., hydromorphone); dronabinol + placebo (i.e., dronabinol); and hydromorphone + dronabinol. The initial session was fixed to the hydromorphone + placebo condition to ensure participants could safely tolerate hydromorphone before combining it with dronabinol. The order of the remaining sessions was randomized using a random sequence generator, with randomization handled by a research pharmacist who had no direct involvement in the study.

Measures

QST Methods

We utilized various sensory pain assessments to quantify acute pain, including threshold responses, temporal summation, cold pressor testing, and conditioned pain modulation, as well as a model of chronic pain using 10% topical capsaicin cream.

Acute Pain Measures

Thermal and pressure pain tests were conducted over two trials to ensure consistency, with the results averaged for each measure. To evaluate thermal pain threshold and tolerance, a thermode (Medoc TSA II; Israel) was gradually heated from a baseline of 31°C at a rate of 0.5°C/s, using the ascending method of limits. Participants were instructed to press a button when the sensation first became painful, indicating thermal threshold, and when the pain became intolerable, indicating thermal tolerance. Pressure pain threshold was assessed using a pressure algometer (Somedic; Sweden), which applied steady pressure to the upper trapezius muscle or patella until either the maximum limit of 1,200 kPa was reached or the participant reported pain. The pressure pain threshold reported herein includes averages of both the patella and trapezius trials and therefore reflects overall pressure pain threshold. Additional analyses were run separately for pressure pain threshold of the trapezius and of the patella and are reported in the results as sensitivity analyses. Temporal summation was evaluated using repetitive thermal and punctate stimuli. For thermal temporal summation, the thermode delivered 10 heat pulses, each lasting 0.5 s, at temperatures of 49°C and 51°C, with 2.5-s intervals between pulses. Participants rated the sensation on a scale from 0 (no sensation) to 100 (intolerable pain). The thermal temporal summation score was calculated by averaging the difference between the initial and maximum pain ratings across pulses for each temperature. Mechanical temporal summation was assessed by applying weighted pinprick stimulators to the ventral forearm. Participants verbally rated their pain on a scale from 0 to 100 after a single stimulus with a force of either 256- or 512-mN probe, and again after a 10-stimulus sequence lasting 10 s. The mechanical temporal summation score was calculated by taking the average of two wind-up ratios, where 1 was added to each rating to avoid division by 0, and dividing the peak pain reported after the 10-stimulus sequence by the initial rating from the single stimulus. For the cold pressor tasks, participants submerged their hand in a circulating cold-water bath set to 5°C and rated their pain every 30 s. Both cold pain threshold, defined as the time to first pain sensation, and cold pain tolerance, defined as the time to hand withdrawal, were recorded. After the initial cold pressor task, two additional cold pressor tasks were conducted to assess conditioned pain modulation (CPM). The conditioning stimulus was cold pain applied to the participants’ hands, and the test stimuli involved pressure pain threshold and mechanical temporal summation. First, participants’ pressure pain threshold or peak mechanical temporal summation was assessed immediately before each additional cold pressor task. For each task, participants submerged their hands in cold water for 20 s, followed by reassessing either the pressure pain threshold or the mechanical temporal summation. To calculate CPM for each testing stimulus, the initial pressure pain threshold or peak mechanical temporal summation values were subtracted from those obtained during the combined cold pressor task. Last, we assessed after-sensation (residual pain) 15 s after the completion of the thermal and mechanical temporal summation tasks, as well as the cold pressor tasks.

Chronic Pain Measures

As described in previous studies,¹⁹ the chronic pain model was simulated using 10% topical capsaicin cream in combination with thermal stimuli. To safely allow for the absorption of the capsaicin cream, an open square adhesive frame was positioned around the application site (ventral forearm), where the cream remained for 30 min. After the removal of the cream, a 45°C thermal stimulus was applied to the treated area for 5 min, during which pain levels were recorded every minute on a 0 to 100 scale. The affected area was then evaluated for flare, secondary hyperalgesia, and mechanical temporal summation. At each QST session after drug administration, the treatment site was reheated to 45°C for 5 min. After this rekindling process, assessments were made for flare, secondary hyperalgesia, thermal threshold, and mechanical temporal summation. The primary outcomes reported in this article are thermal threshold and mechanical temporal summation within the capsaicin-sensitized area.

Global QST Outcomes

Two composite outcomes were derived from the full battery of QST tests: (1) central sensitization, calculated as the average z-scores of thermal and mechanical temporal summation, CPM, and after-sensation ratings, and (2) general sensitivity, calculated as the average z-scores of pressure and thermal thresholds, thermal tolerance, and cold pressor threshold and tolerance. Higher scores on these measures indicate greater levels of central sensitization and pain sensitivity, respectively.

Clinical Pain Severity

Clinical pain severity was assessed at each time point using a 0 to 100 visual analog scale (VAS)²⁰ that participants used to rate their current global pain. The scale consisted of a straight line, with one end labeled “no pain at all” and the other end labeled “the most intense pain imaginable.”

Self-reported Drug Effects

In accordance with FDA guidelines,²¹ participants’ ratings of drug effects (including drug effect, good effect, bad effect, high, “like the way I feel,” and nausea) were collected using a 0 to 100 VAS.

Risk for Extramedical Use

The primary risk for extramedical use was whether participants rated 60 or higher on the 0 to 100 VAS high scale after drug exposure.²¹ Additionally, participants provided ratings on whether they enjoyed the study medications (response options: “yes,” “no,” or “no effect”), the amount of money they would be willing to pay for the medication, and the likelihood of taking the medication again, using a 6-point Likert scale ranging from 0 (“not at all”) to 5 (“extremely”).

Physical Function Measures

Three objective measures of physical performance were administered²²: (1) a 2-min walking distance task²³; (2) the Timed Up and Go test, which measures the time taken to rise from a standard chair, walk 3 meters, and return to a seated position²³; and (3) stair climb total, which records the time to ascend and descend two-step stairs with hand rails.²²

Cognitive Function Measures

Cognitive function was assessed through three tasks^24–26: (1) psychomotor ability, measured by the percentage of correct responses on the Digit Symbol Substitution Task (DSST), in which participants used a keypad to replicate patterns shown on a computer screen; (2) working memory, evaluated by mean reaction time and accuracy on the Paced Serial Addition Task (PASAT), in which participants added sequentially presented numbers in rapid succession; and (3) fine motor movement, assessed by the maximum number of correct responses on a circular light test, in which participants repeated visual patterns displayed on a board within a 60-s timeframe.

Adverse Events

Throughout the session, participants were asked whether they experienced any side effects from the study medication. Reported adverse events were recorded and categorized based on severity (mild, moderate, or severe) and their relation to the study medication. The primary outcomes were the total number of study-related adverse events, aggregated across all severity levels, as well as the number of related adverse events classified as mild, moderate, or severe.

Power Analysis

The power analysis for this study was based on previous study examining the effects of oxycodone and smoked cannabis on cold pressor tests.²⁷ With a power of 0.8, an α level of 0.05, and an anticipated large effect size based upon this previous study (Cohen’s d = 1.15), it was determined that a sample size of 15 participants would be adequate to detect drug condition effects. However, we recruited 21 participants to account for potential attrition or incomplete data, ensuring sufficient statistical power for our analyses.

Statistical Analysis

The primary outcomes, specifically peak or trough ratings after drug administration, were analyzed using mixed-effects models for continuous variables, generalized estimating equations for dichotomous variables, and multinomial logistic regression for nominal categorical variables (e.g., the response options for medication enjoyment: “yes,” “no,” and “no effect”). In cases in which generalized estimating equation models failed to converge, chi-square analysis was employed. Comparisons between drug conditions were conducted using Tukey’s post hoc tests. The body mass index and sex were included as covariates. All primary data analyses were performed by an independent biostatistician who was not involved in outcome assessments. Analyses were conducted using SAS version 9.4, with an α level of 0.05 for two-tailed tests.

Results

Participant Characteristics

As shown in table 1, participants (N = 21; M_age = 63.38 ± 6.4 yr) were 57% female with an average Kellgren–Lawrence score = 2.42, were predominately white or Black, and were not of Hispanic origin. None of the participants from our previous study using a moderate dose of hydromorphone¹⁵ were included in the current study.

Table 1.

Participant Characteristics at Baseline

Characteristic	Value
Total (N)	21
Age, mean ± SD, yr	63.4 ± 6.4
Sex, No. (%)
Male	9 (42.9)
Female	12 (57.1)
Race, No. (%)
White	12 (57.1)
Black	7 (33.3)
Asian	1 (4.8)
American Indian/Alaska Native	0 (0)
Native Hawaiian/Pacific Islander	0 (0)
Multiracial	1 (4.8)
Prefer not to answer	0 (0)
Hispanic ethnicity, No. (%)	1 (4.8)
Relationship status, No. (%)
Never married	8 (38.1)
Married or remarried	6 (28.6)
Divorced, separated, or widowed	7 (33.3)
Education level, No. (%)
Less than high school	0 (0)
High school graduate	3 (14.3)
Some college, no degree	5 (23.8)
College or professional degree	11 (52.4)
Above college/doctorate	2 (9.5)
Yearly household income, No. (%)
$0 to $24,999	4 (19.0)
$25,000 to $49,999	6 (28.6)
$50,00 to $74,999	5 (23.8)
$75,000 to $99,999	3 (14.3)
$100,000 or more	3 (14.3)
Prefer not to answer	0 (0)
Disability, No. (%)	3 (14.3)
Currently receiving treatment for pain, No. (%)	2 (9.5)
Have been prescribed an opioid in lifetime, No. (%)	11 (52.4)
Body mass index, mean ± SD, kg/m2	33.8 ± 6.7
Average pain severity (0 to 10 scale from BPI), mean ± SD	4.8 ± 2.2

BPI, Brief Pain Inventory.

QST Outcomes

Acute Pain Outcomes

There was no evidence that dronabinol enhanced the analgesic effects of hydromorphone on acute pain QST outcomes (table 2). A significant main effect of drug condition was observed for pressure pain threshold [F(3,18) = 3.23; P = 0.029], with hydromorphone producing greater analgesia than dronabinol (P = 0.039; 95% CI, 24.56 to 156.34). Additionally, a significant main effect of drug condition was found for mechanical temporal summation [F(3,18) = 3.68; P = 0.017], indicating that hydromorphone provided greater analgesia than placebo (P = 0.020; 95% CI, 0.42 to 2.11). Figure 2 provides an overview of these pressure pain threshold and mechanical temporal summation findings. No significant drug-related differences were observed for thermal threshold and tolerance, thermal temporal summation, cold pressor outcomes, or CPM measures. Sensitivity analyses were run using pressure pain threshold for the patella and trapezius trials separately. A significant main effect was seen for pressure pain threshold trapezius [F(3,18) = 2.94; P = 0.041] with hydromorphone producing greater analgesia than dronabinol (P = 0.04; 95% CI, 21.31 to 143.59), but not for the patella pressure pain threshold [F(3,18) = 2.51; P = 0.07].

Table 2.

Summary of Primary Outcomes

Primary Outcomes	Placebo		Hydromorphone (2 mg, oral)		Dronabinol (10 mg, oral)		Hydromorphone (2 mg, oral) + Dronabinol (10 mg, oral)		P Value (Partial η2)
	Mean, %	SEM	Mean, %	SEM	Mean, %	SEM	Mean, %	SEM
Quantitative sensory testing
Acute pain model
Pressure pain threshold (0 to 1,200 kPa)	582.85	43.98	650.06a	48.84	566.46a	40.99	569.64	45.35	0.029 (0.074)
Heat pain threshold (°C)	44.63	0.67	44.55	0.68	44.37	0.67	44.84	0.71	0.809 (−0.025)
Heat pain tolerance (°C)	47.89	0.46	47.96	0.48	47.61	0.47	48.10	0.30	0.508 (−0.008)
Mechanical temporal summation	3.22a	0.55	1.96a	0.14	2.32	0.23	2.90	0.47	0.017 (0.087)
Thermal Temporal Summation	3.80	0.90	1.96	0.18	2.48	0.33	3.40	0.70	0.297 (0.009)
Cold pressor threshold (time), s	16.39	2.01	17.85	2.09	14.50	2.01	20.50	4.05	0.238 (0.016)
Cold pressor threshold severity rating (0 to 100 VAS)	72.73	5.28	78.13	5.66	68.33	7.36	70.00	8.47	0.725 (−0.020)
Cold pressor tolerance (time), s	93.32	16.51	97.96	15.38	80.33	15.52	72.51	14.70	0.056 (0.056)
Conditioned pain modulation
Mechanical temporal summation	−16.33	4.04	−13.00	3.46	−17.52	4.38	−13.57	3.18	0.762 (−0.022)
Pressure pain threshold	70.29	14.77	106.31	18.40	78.88	15.56	113.81	16.74	0.111 (0.038)
Chronic pain model
Capsaicin
Thermal threshold	41.49	0.53	41.30	0.44	41.36	0.38	41.60	0.48	0.913 (−0.030)
Mechanical temporal summation	2.43a	0.30	1.95a	0.22	2.09	0.30	2.16	0.49	0.045 (0.062)
Global QST measures
Central sensitization (z-score)	−0.04	0.11	−0.10	0.09	−0.04	0.09	−0.12	0.11	0.335 (0.006)
General pain sensitivity (z-score)	−0.21	0.10	−0.26	0.11	−0.13	0.13	−0.17	0.12	0.201 (0.021)
Clinical pain severity (0 to 100 VAS)	13.10	4.00	9.52	3.55	12.81	4.05	15.00	3.81	0.280 (0.011)
Physical functioning tests
2-min walking distance	322.86	22.72	325.24	21.86	319.05	18.58	310.50	17.37	0.156 (0.028)
Timed Up and Go time	10.59	0.75	10.60	0.81	10.32	0.65	10.44	0.87	0.847 (−0.027)
Total stair climb time	5.11	0.52	4.84	0.37	4.96	0.43	4.68	0.39	0.469 (−0.005)
Participant ratings (0 to 100 VAS)
Drug effect	29.43a	6.93	25.76b,c	6.09	45.29b	5.65	50.52a,c	6.51	< 0.001 (0.193)
Good effect	25.62a	6.86	52.67a	7.88	44.10	5.92	43.71	6.72	0.007 (0.111)
Bad effect	15.95	5.43	10.86	3.78	23.62	5.74	24.33	5.92	0.273 (0.012)
High	16.76a,b	5.41	18.38c,d	4.74	35.52a,c	5.87	32.57b,d	5.95	0.001 (0.173)
“Like the way I feel”	66.14	5.47	78.57	4.82	63.90	5.67	63.95	5.56	0.042 (0.064)
Nausea	4.38	1.96	1.38a	0.48	1.90	0.93	11.14a	3.90	0.027 (0.076)
Risk for extramedical use measures
Enjoyed medication (% yes)	38.9%		40.0%		50.0%		47.1%		0.634 (N/A)
Would take medication again (0 to 5)	1.28	0.29	1.55	0.31	1.33	0.27	1.06	0.26	0.672 (−0.017)
≥60 on “high” rating scale, %	4.2%		6.0%		7.4%		8.1%		0.535 (0.040)
Willingness to pay for medication, $	22.17	16.45	15.80	10.37	15.94	6.14	15.35	6.30	0.827 (−0.014)
Cognitive testing
Circular lights, maximum/min	64.25	2.86	64.25	3.67	67.25	3.71	64.38	3.22	0.056 (0.056)
DSST (proportion correct)	0.79	0.09	0.77	0.12	0.81	0.09	0.83	0.08	0.360 (0.004)
PASAT
Mean reaction time correct, s	1,762.13a	130.96	1741.21	54.20	1,749.60	93.72	1,938.64a	109.21	0.043 (0.068)
Correct, %	50.95a	11.05	37.30a,b,c	5.33	47.18b	12.66	49.05c	10.04	0.005 (0.121)

The outcomes represent mean peak ratings or percent participants for each condition (N = 37). Matching superscript letters indicate significant (P < 0.05) differences in post hoc comparisons. Partial η² effect sizes provided for significant results: small (0.01), medium (0.06), and large (0.14). Statistically significant P values are in bold.

DSST, Digit Symbol Substitution Task; PASAT, Paced Auditory Serial Addition Task; QST, quantitative sensory testing; SEM, standard error of the mean; VAS, visual analog scale.

Fig. 2.

Quantitative sensory testing (QST) outcomes. The data show pressure pain threshold (A), mechanical temporal summation (B), and capsaicin-sensitized mechanical temporal summation (C), as functions of study condition (x-axis). The medication conditions were placebo + placebo (Plc), 2 mg oral hydromorphone + placebo (Hydro), 10 mg oral dronabinol (Dmb), and 2 mg hydromorphone combined with 10 mg oral dronabinol (Hydro + Dmb). The brackets at the top that connect the bars represents a significant difference in post hoc comparison, and the error bars represent the standard error of the mean (SEM).

Chronic Pain Outcomes

QST outcomes based on the capsaicin-induced chronic pain model also provided no evidence that dronabinol enhanced the analgesic effects of hydromorphone (table 2). A significant main effect of drug condition was observed for mechanical temporal summation in the capsaicin-sensitized area [F(3,18) = 2.86; P = 0.045], with hydromorphone demonstrating significantly lower mechanical temporal summation compared to placebo (P = 0.035; 95% CI, 0.20 to 1.20; fig. 2). However, no significant differences were found between drug conditions for thermal threshold in the sensitized area.

Global QST Outcomes

No significant main effect of drug condition was observed for any of the global QST outcomes (table 2).

Clinical Pain Severity and Physical Functioning Outcomes

As shown in table 2, no significant main effects of drug condition were observed for clinical pain severity or physical functioning outcomes (i.e., 2-min walking distance, Timed Up and Go time, or total stair climb time).

Self-reported Drug Effect Outcomes

Overall, the drug conditions resulted in several subjective ratings that differed significantly from placebo (table 2; fig. 3). Drug effect significantly differed between drug conditions [F(3,18) = 7.68; P < 0.0001). Hydromorphone + dronabinol significantly increased ratings of drug effect compared to placebo (P = 0.022; 95% CI, −30.20 to −5.90) and hydromorphone (P = 0.001; 95% CI, −36.55 to −12.25), although it was not elevated relative to dronabinol (P = 0.979; 95% CI,[−14.55 to 9.75). Ratings for good effect showed a significant main effect of drug condition [F(3,18) = 4.49; P = 0.007], with hydromorphone producing significantly higher ratings than placebo (P = 0.004; 95% CI, −38.67 to −11.02). No significant main effect of drug condition was observed for bad effect (P = 0.273). Ratings for high [F(3,18) = 6.85; P = 0.001 for the main effect] were significantly greater than placebo for dronabinol (P = 0.008; 95% CI, −29.86 to −7.54) and hydromorphone + dronabinol (P = 0.022; 95% CI, −27.76 to −5.44) but not for hydromorphone (P = 1.00; 95% CI, −10.86 to 11.46). Additionally, ratings for high for dronabinol (P = 0.006; 95% CI, −30.16 to −7.84) and hydromorphone + dronabinol (P = 0.019; 95% CI, −28.06 to −5.74) were significantly greater than for hydromorphone. Although a significant main effect of drug condition was observed for ratings of “like the way I feel” [F(3,18) = 2.92; P = 0.042], no significant pairwise comparisons emerged. Last, ratings for nausea [F(3,18) = 3.30; P = 0.012 for the main effect] were significantly higher for hydromorphone + dronabinol compared to hydromorphone (P = 0.033; 95% CI, −13.27 to −2.23) but not compared to placebo (P = 0.189; 95% CI, −11.12 to −0.08).

Fig. 3.

Participant ratings. The data show results from participant ratings of drug, good, bad, and high effects, as functions of study conditions. The brackets at the top that connect the bars represents a significant difference in post hoc comparison, and the error bars represent the standard error of the mean (SEM).

Extramedical Risk Outcomes

None of the extramedical risk outcomes showed significant differences between drug conditions (table 2).

Cognitive Function

No significant differences between drug conditions were observed for the circular lights task, which assesses fine motor movement [F(3,18) = 2.67; P = 0.056], or the DSST, which measures psychomotor ability [F(3,18) = 1.10; P = 0.360]. However, a significant main effect of drug condition was found for mean reaction time in PASAT [F(3,18) = 3.05; P = 0.043], with hydromorphone + dronabinol significantly increasing reaction time compared to placebo (P = 0.029). Additionally, a significant main effect of drug condition was observed for percent correct responses in PASAT, which assesses working memory [F(3,18) = 4.87; P = 0.005]. Hydromorphone significantly impaired working memory compared to placebo (P = 0.013; 95% CI, 4.02 to 17.90), dronabinol (P = 0.034; 95% CI, −16.32 to −2.73), and hydromorphone + dronabinol (P = 0.011; 95% CI, −18.19 to −4.32). However, neither dronabinol nor hydromorphone + dronabinol significantly differed from placebo (P > 0.05).

Adverse Events

Study-related adverse events were documented in 24 sessions (28.6%). No serious adverse events occurred. Overall, there were no significant differences between drug conditions in the occurrence of study-related adverse events (P = 0.082). Additionally, no significant differences were observed in the frequency of mild (P = 0.564) or moderate (P = 0.530) adverse events across drug conditions. The frequency of study adverse events, categorized by each drug condition, is detailed in supplemental table S2 (https://links.lww.com/ALN/E385).

Discussion

This within-subject, double-blind, randomized, placebo-controlled trial examined the effects of combining a synthetic cannabinoid (dronabinol, 10 mg) and an opioid (hydromorphone, 2 mg) on acute and clinical pain, physical and cognitive functioning, self-reported drug effects, risk for extramedical use, and adverse events in patients with knee osteoarthritis. Consistent with our previous study with a higher dose of hydromorphone (4 mg),¹⁵ the current results do not support synergistic effects of combining 10 mg of dronabinol and hydromorphone. However, the lower hydromorphone dose appeared to show a better safety profile than our previous data: hydromorphone-only and hydromorphone + dronabinol administrations produced similar rates of adverse events as placebo.

We found no evidence of dronabinol-enhanced analgesia beyond the effects of hydromorphone on laboratory-induced pain outcomes. When only hydromorphone was administered, participants showed greater pressure pain threshold and mechanical temporal summation analgesia compared to dronabinol and placebo. Our earlier phase II trial in healthy individuals, which used a higher hydromorphone dose (4 mg) and varying dronabinol doses (2.5, 5, or 10 mg) without a dronabinol alone condition, also found limited evidence of dronabinol enhancement on QST outcomes.¹⁸ While dronabinol improved some QST measures compared to placebo, it did not outperform hydromorphone. Similarly, in our previous knee osteoarthritis study with 4 mg of hydromorphone, we found minimal additive effects of dronabinol for QST outcomes.¹⁵ Although hydromorphone + dronabinol increased cold pressor threshold and tolerance compared to placebo and dronabinol alone, it did not exceed hydromorphone alone. There were also no significant differences among drug conditions on clinical pain severity or physical functioning.

In the current study, the hydromorphone + dronabinol condition did not produce higher self-reported drug effects beyond dronabinol alone, but it was higher than placebo and hydromorphone. Nausea ratings were higher for hydromorphone + dronabinol compared to hydromorphone but not compared to placebo. Dronabinol, although sometimes used to reduce nausea, can cause paradoxical nausea.²⁸ It is unclear whether the combination of hydromorphone and dronabinol may increase the possibility of this side effect for those with knee osteoarthritis. There were no significant differences across drug conditions for ratings of bad effect or “like the way I feel,” and no hydromorphone + dronabinol or dronabinol differences for good effect. Importantly, no significant drug condition differences were observed on measures used to determine risk for extramedical use, in contrast to our earlier knee osteoarthritis trial with 4 mg of hydromorphone, in which there was a slightly increased risk for extramedical use of hydromorphone + dronabinol.¹⁵

In both of our previous investigations involving healthy individuals and a higher opioid dosage in those with knee osteoarthritis, only the hydromorphone condition negatively affected cognitive performance.^15,18 Similarly, the current study found impaired working memory for the hydromorphone condition compared to all other conditions. In addition, the current study also found increased working memory reaction time for the hydromorphone + dronabinol compared to placebo. These findings are consistent with our past work that indicates hydromorphone may impair cognitive functioning, while dronabinol may provide a protective effect. It should be noted that this is highly speculative at this time and warrants further investigation.

Despite its strengths, the current study has some limitations. While several previous reports of improved pain and reduced opioid use stem from inhaled, natural cannabis use, the current study investigated a synthetic, orally ingested cannabinoid.^29,30 The route of administration and form of cannabis (i.e., synthetic vs. natural) are important factors to consider, as there are substantially differing times to peak concentration (3 to 10 min inhaled vs. 1 to 2 h oral) and compound concentrations (more than 500 compounds possible in natural cannabis vs. pure tetrahydrocannabinol in synthetic).^31–34 These pharmacokinetic and compositional differences may affect study outcomes in ways not fully understood in knee osteoarthritis populations. Additionally, although the current study used FDA-approved therapeutic dosages for both hydromorphone and dronabinol, which provides clinical relevance for our findings, dronabinol may not represent the type of medical cannabis patients would typically acquire from dispensaries.³⁵ The current study’s assessment timepoints are reflective of dronabinol peak effects^15,36; future work may choose to include additional timepoints that capture the half-life of dronabinol. Moreover, a single dronabinol dose was used, limiting the ability to assess potential dose–response effects. In addition, for safety reasons—particularly given the older adult population with knee osteoarthritis—the hydromorphone condition was not randomized, which may have introduced order effects. Given the early-stage nature of this work and to reduce type II error that may obscure meaningful pharmacologic signals, the current study focused on the pattern, magnitude, and consistency of effects across related measures and domains to guide future hypothesis-driven research. The outcomes of interest represent conceptually distinct constructs rather than multiple tests of the same dependent variable; as such, we did not apply formal multiplicity corrections. However, as this literature expands, future work should consider multiple comparison test corrections, especially in the context of related outcomes. Last, this investigation cannot inform the impact of the long-term use of dronabinol and hydromorphone.

To our knowledge, this is one of the few rigorously controlled laboratory studies to evaluate the effects of dronabinol (10 mg) with hydromorphone (2 mg), both within clinically relevant therapeutic dose ranges, in individuals with knee osteoarthritis. The current findings build upon our previous work in both healthy adults and individuals with knee osteoarthritis receiving a higher dosage of hydromorphone. In line with our earlier work, the current findings do not provide support for the clinical utility of combining dronabinol and hydromorphone for managing knee osteoarthritis–related pain or improving physical function. To better determine the potential benefits and risks of opioid–cannabinoid co-use, future studies should explore different cannabis formulations and administration routes, employ dose-ranging designs, and incorporate long-term follow-up assessments.

Acknowledgments

The study team thanks Deborah Fashole-Luke, M.A. (George Mason University, Fairfax, Virginia), Bryan Herrera, B.A. (Department of Psychiatry and Behavioral Sciences, Johns Hopkins University School of Medicine, Baltimore, Maryland), Katie Smith, B.A. (Department of Psychiatry and Behavioral Sciences, Johns Hopkins University School of Medicine), Mike Sklar, M.A. (Department of Psychiatry and Behavioral Sciences, Johns Hopkins University School of Medicine), Jenna Pelly, M.P.P. (Mathematica, Washington D.C.), Sydney Jensen, C.M.H.C. (Rochester, Michigan), Savannah King, M.S.W. (Bel Air, Maryland), Alex Kearson, B.A. (School of Social Work, University of Maryland, Baltimore, Maryland), Jim Stone, B.A. (Department of Psychiatry and Behavioral Sciences, Johns Hopkins University School of Medicine), Sana Rehman, M.B.B.S. (Department of Neuroradiology, Johns Hopkins University, Baltimore, Maryland), Leticia Nanda, C.R.N.P. (Department of Psychiatry and Behavioral Sciences, Johns Hopkins University School of Medicine), and Annie Umbricht, M.D. (Department of Psychiatry and Behavioral Sciences, Johns Hopkins University School of Medicine) for assistance with study sessions and Paul Nuzzo, M.A. (University of Kentucky College of Medicine, Lexington, Kentucky) for assistance conducting data analyses.

Research Support

Supported by National Institutes of Health (Bethesda, Maryland) grant Nos. R01DA042751 (to K.E.D. and C.M.C.), R01DA040644 (to K.E.D. and C.M.C.), R01DA035246 (to K.E.D.), and F32DA04939302 (to C.J.M.) from the National Institute on Drug Abuse; T32NS070201 (to K.R.H.) from the National Institute of Neurological Disorders and Stroke; U01HL150568 (to K.R.H.) from the National Heart Lung and Blood Institute; and K24AR081143 (to C.M.C.) from the National Institute of Arthritis and Musculoskeletal and Skin Diseases.

Competing Interests

The investigators have no relevant competing interests to disclose. In the past 3 yr, Dr. Dunn has received honoraria for consulting with Cessation Therapeutics (San Jose, California) and participating on a study steering committee for Indivior (North Chesterfield, Virginia). Dr. Huhn has consulted for Gilgamesh Pharmaceuticals (New York, New York) and Rutgers University (New Brunswick, New Jersey), receives study medication and logistical/regulatory support from Merck Inc. (Rahway, New Jersey) for a clinical trial, and has received research funding from Ashley Addiction Treatment (Havre de Grace, Maryland) and Indivior Inc. (North Chesterfield, Virginia) through his university. Dr. Vandrey has been a paid consultant or received honoraria for advisory board work from Syqe Medical Ltd. (Tel Aviv-Yafo, Isreal), Charlotte’s Web (Littleton, Colorado), Schedule 1 Therapeutics (Chicago, Illinois), and WebMD (Newark, New Jersey). Dr. Bergeria has received grant funding from Canopy Growth Corporation (Ontario, Canada) and Pear Therapeutics (Boston, Massachusetts) and has consulted with Mind Med, Inc. (New York, New York) and Eli Lilly (Indianapolis, Indiana). Dr. Speed has received honoraria for consulting for Pain Care Labs (Atlanta, Georgia). The other authors declare no competing interests.

Reproducible Science

Full protocol available at: hamiltonk@ohio.edu. Raw data available at: hamiltonk@ohio.edu.

Supplemental Digital Content

Supplemental file, https://links.lww.com/ALN/E385

Table S1. Study Eligibility Criteria.

Table S2. Study-Related Adverse Events (AEs).

Supplementary Material

aln-144-1187-s001.pdf ^{(131.3KB, pdf)}