Vaporized cannabis versus placebo for acute migraine: A randomized, double‐blind, placebo‐controlled crossover trial

Abstract

Objective

To assess the efficacy of cannabis for the treatment of acute migraine.

Background

Preclinical and retrospective studies suggest cannabinoids may be effective in migraine treatment. However, there have been no randomized clinical trials examining the efficacy of cannabinoids for acute migraine.

Methods

In this randomized, double‐blind, placebo‐controlled, crossover trial, adults with migraine treated up to four separate migraine attacks, one each with vaporized (1) 6% Δ9‐tetrahydrocannabinol (THC) (THC‐dominant), (2) 11% cannabidiol (CBD) (CBD‐dominant), (3) 6% THC + 11% CBD, and (4) placebo cannabis flower in a randomized order. Washout period between treated migraine attacks was ≥1 week. The primary endpoint was pain relief, and secondary endpoints were pain freedom and most bothersome symptom freedom, all assessed at 2‐h post‐vaporization.

Results

Ninety‐two participants were enrolled and randomized, and 247 migraine attacks were treated. THC + CBD was superior to placebo at achieving pain relief (67.2% vs. 46.6%, odds ratio [95% confidence interval] 2.85 [1.22, 6.65], p = 0.016), pain freedom (34.5% vs. 15.5%, 3.30 [1.24, 8.80], p = 0.017), and most bothersome symptom freedom (60.3% vs. 34.5%, 3.32 [1.45, 7.64], p = 0.005) at 2 h, as well as sustained pain freedom at 24 h and sustained most bothersome symptom freedom at 24 and 48 h. THC‐dominant was superior to placebo for pain relief (68.9% vs. 46.6%, 3.14 [1.35, 7.30], p = 0.008) but not pain freedom or most bothersome symptom freedom at 2 h. CBD‐dominant was not superior to placebo for pain relief, pain freedom, or most bothersome symptom freedom at 2 h. There were no serious adverse events.

Conclusion

Acute migraine treatment with 6% THC + 11% CBD was superior to placebo at 2‐h post‐treatment with sustained benefits at 24 and 48 h.

Plain Language Summary

Many people with migraine self‐treat with cannabinoids or are interested in using cannabinoids to treat migraine. In this double‐blind study, people with migraine treated up to 4 migraine attacks, 1 attack was treated with each of 3 vaporized cannabis flower treatments (THC 6%, CBD 11%, and THC 6% + CBD 11%) or placebo cannabis flower without THC or CBD, within the first 4 h of migraine attack onset. Four puffs of cannabis flower containing THC 6% + CBD 11% was superior to placebo at treating migraine attacks, though the study did not examine the long‐term effects of frequent use.

Abbreviations

BI

blinding index

CBD

cannabidiol

CGRP

calcitonin gene‐related peptide

CI

confidence interval

DSP

Drug Supply Program

FDA

Food and Drug Administration

FUPP

Foltin Uniform Puff Procedure

GLMM

generalized linear mixed model

ICHD‐3

International Classification of Headache Disorders, 3rd Edition

IHS

International Headache Society

IRB

Institutional Review Board

ITT

intention‐to‐treat

MBS

most bothersome symptom

mITT

modified intention‐to‐treat

NIDA

National Institute on Drug Abuse

RCT

randomized controlled trial

THC

Δ9‐tetrahydrocannabinol

UCSD

University of California, San Diego

INTRODUCTION

Migraine is the second leading cause of years lived with disability worldwide. It affects over a billion people worldwide, including 38 million Americans. ¹ Migraine treatments are classified for acute and/or preventive use. ² Nearly everyone with migraine uses acute treatments for migraine attacks; however, rates of treatment optimization with traditional acute therapies are relatively low and rates of discontinuation are high. ³ , ⁴ There is significant interest in and use of cannabinoids for acute migraine treatment. Migraine is among the most common medicinal uses of cannabinoids, with 35.5% of 1429 medical cannabis users reporting use for headache/migraine. Inhalation was the most common route of administration (81.4%). ⁵

More than 125 identified phytocannabinoids are naturally found in the cannabis plant, including Δ9‐tetrahydrocannabinol (THC) and cannabidiol (CBD). Although the US Food and Drug Administration (FDA) has approved cannabinoid‐based medications for specific conditions, none have been approved for migraine. At the time of this writing, THC was legal for medical and/or recreational use in 38 US states and the District of Columbia. Since 2018, hemp‐derived CBD (<0.3% THC) is not a controlled substance in the United States. ⁶

Patients often ask healthcare professionals about cannabinoids for migraine, but data to inform medical advice is lacking. ⁷ Preclinical evidence suggests that cannabinoids may have effects on migraine pathogenesis through mechanisms including inhibiting calcitonin gene‐related peptide (CGRP) release, inhibiting CGRP‐induced nitric oxide release, inhibiting trigeminovascular neurons, and inhibiting cortical spreading depression—and surveys and retrospective studies suggest that cannabinoids may have benefit as acute and/or preventive treatments for migraine. ⁸ , ⁹ , ¹⁰ , ¹¹ , ¹² , ¹³ , ¹⁴ , ¹⁵ , ¹⁶ , ¹⁷ , ¹⁸ , ¹⁹ , ²⁰ , ²¹ However, the efficacy of cannabis for acute treatment of migraine had not been previously studied via a randomized, controlled trial (RCT). ¹⁰ , ¹² , ²² Therefore, this RCT was conducted to assess the efficacy of vaporized cannabis against placebo cannabis. We hypothesized that cannabis flower containing active cannabinoids (THC and/or CBD) would show superior efficacy over a placebo by achieving significantly greater pain relief at 2‐h post‐vaporization among adults experiencing a migraine attack.

METHODS

Trial design

This randomized, double‐blind, placebo‐controlled, crossover trial was conducted at the University of California, San Diego (UCSD). The study was approved by the UCSD Institutional Review Board (IRB #181944) and the FDA. The study was prospectively registered at ClinicalTrials.gov (NCT04360044). The authors take full responsibility for the data and the accuracy and integrity of the publication. No writing assistance was provided by outside parties. The study was performed in accordance with the principles of the Declaration of Helsinki. All participants provided written informed consent prior to enrollment.

Sample size

The sample size was determined based on our primary outcome of pain relief at 2 h, assuming a response of 68% for treatment and 45% for placebo as seen in an intranasal sumatriptan RCT. ²³ We used a two‐sided type I error level alpha = 0.05 to estimate the sample size to achieve 80% power based on the two‐sample proportion test for a binary outcome to accommodate the crossover study design. This effect size required 72 participants to detect a significant difference with 80%. Assuming a 20% dropout rate, we planned to enroll 90 participants.

Study drug

Cannabis flower was obtained from the National Institute on Drug Abuse (NIDA) Drug Supply Program (DSP) and consisted of four different treatments. The most closely matched batches available from the NIDA DSP at study initiation were selected. They were: (1) 5.62% THC + 0.03% CBD (referred to in this study as 6% THC or THC‐dominant), (2) 11.27% CBD + 0.35% THC (referred to as 11% CBD or CBD‐dominant), (3) 6.16% THC + 10.77% CBD (referred to as 6% THC + 11% CBD or THC + CBD), and (4) <0.025% THC + 0.14% CBD (referred to as placebo cannabis flower or placebo). All treatments contained <1% minor cannabinoids and were devoid of terpenes. The studied THC potency was based on previous pain studies, ²⁴ , ²⁵ , ²⁶ and the studied CBD was based on the goal of studying THC and CBD in an approximately 1:2 ratio.

Participants

Key inclusion criteria were: aged 21 to 65 years; migraine according to ICHD‐3 criteria ²⁷ ; 2 to 23 headache days and 2 to 23 migraine days per month; agree not to use cannabis (outside of the study drug), opioids, or barbiturates. Key exclusion criteria were: screening visit urine drug test positive for THC, barbiturates, opioids, oxycodone, or methadone; pregnant; breastfeeding; known cognitive impairment; current moderate–severe or severe depression; history of bipolar disorder, schizophrenia, or psychosis; history of substance use disorder; and active pulmonary disease or other severe medical illnesses at the discretion of the researchers. Inclusion criteria initially required THC use within the prior 2 years due to concern that cannabis‐naïve patients may be at risk of experiencing dysphoria; this was removed March 18, 2021, due to slow recruitment and no serious adverse events at that time. Full criteria are available at ClinicalTrials.gov.

Participants agreed not to use any other acute migraine treatments prior to or within 2 h following study drug administration.

Enrollment

Participants were enrolled at UCSD November 20, 2020 to November 4, 2022, with follow‐up completed February 23, 2023. A headache neurologist experienced in ICHD‐3 criteria (NMS) confirmed participant eligibility.

After written informed consent, baseline characteristics were captured using Research Electronic Data Capture. ²⁸ Then, participants were trained in the use of the Storz & Bickel (Tuttlingen, Germany) Mighty Medic handheld vaporizer and the Foltin Uniform Puff Procedure (FUPP), a validated cannabis vaporization procedure, and a smartphone application was installed on participants' smartphones to guide participants through the study and collect data. ²⁹

Randomization

A research pharmacist randomized participants to receive the four different treatments using simple 1:1:1:1 assignment (24 possible orders) using Microsoft Excel's (Redmond, WA, USA) random number function.

Blinding

Participants, research coordinators, investigators, and statisticians were blinded until after the initial statistical analysis of primary and key secondary endpoints was completed; only the research pharmacists were unblinded during the study. A research pharmacist loosely loaded the four treatments into identical Storz & Bickel Filling Set vaporization capsules and placed them into sealed bags. The research pharmacist affixed labels to the sealed bags stating the order in which the participant would use the four treatments (“migraine 1” through “migraine 4”). The key linking the treatments to their identifying number was stored on a password‐protected computer in the locked pharmacy available only to the research pharmacists.

Participant blinding was promoted by using placebo cannabis flower from which the THC and CBD had been extracted via a chemical process by the NIDA DSP and by framing, including educating patients that they might experience a “placebo high” from the CBD‐dominant and placebo treatments and that they might not experience a “high” from the THC‐dominant and THC + CBD treatments because the study potencies were lower than “recreational” potencies. For each migraine attack, after answering 2‐h efficacy and safety questions, participants were asked which treatment they thought they had received.

Treatments

Upon each migraine attack onset, the participant accessed the smartphone application. If ≥7 days had elapsed since the prior cannabis administration (ensuring ≥7 days washout period between cannabis administrations), the application asked the participant questions to establish whether the attack met treatment criteria of: (1) headache <4 h from onset, (2) headache pain moderate or severe in intensity, (3) associated with photophobia and phonophobia and/or with nausea, (4) no acute treatments used since attack onset. If the attack met all criteria, the application instructed the participant to vaporize study cannabis at 180°C using the Mighty Medic handheld vaporizer. The application guided the participant through the timed FUPP to standardize vaporization procedure across treated attacks and across participants. FUPP consists of 5 s of inhalation, followed by a 10‐s breath hold, exhalation, and 45‐s waiting period before repeating the process. ²⁹ For each attack, participants performed the FUPP four times under continuous application guidance. The application pushed surveys at 1, 2, 24, and 48 h.

Outcomes

Outcomes were based on International Headache Society (IHS) guidelines. ³⁰ The prespecified primary outcome was pain relief at 2‐h post‐vaporization ³⁰ based on prior evidence of pain relief without prior evidence of pain freedom or most bothersome symptom (MBS) freedom with cannabinoids for migraine ¹⁹ and a concern that MBS freedom rates could be superior to placebo due to THC's known antiemetic effects rather than due to antimigraine effects. ³¹ The prespecified secondary outcomes were 2‐h pain freedom and 2‐h MBS freedom. ³⁰ Other prespecified outcomes included freedom from pain, photophobia, phonophobia, nausea, vomiting, and rescue medication use, with data collected at 1, 2, 24, and 48 h. ³⁰ Pain at time 0 and each subsequent timepoint was reported as none, mild, moderate, or severe. ³⁰ Pain relief was defined as reduction of pain from moderate or severe to mild or none. ³⁰ Pain freedom was defined as absence of pain. Sustained pain freedom was defined as pain freedom at 2 h without return of pain or rescue medication use. ³⁰ MBS (selected from photophobia, phonophobia, or nausea) was identified by the participant during each attack prior to vaporization. MBS freedom was defined as absence of the MBS at each subsequent timepoint. Sustained MBS freedom was defined as MBS freedom at 2 h without return of the MBS or rescue medication use at each subsequent timepoint. ³⁰ Primary, secondary, and tertiary outcomes were prospectively registered (NCT04360044).

Analyses

Continuous and categorical variables were reported as mean (standard deviation) and count (percentage), respectively. The intention‐to‐treat (ITT) analysis analyzed all treated migraine attacks, counting treated attacks without recorded 2‐h data as treatment failures as per IHS guidelines. ³⁰ The modified intention‐to‐treat analysis (mITT) only analyzed attacks with recorded 2‐h data. During the trial, some participants filled out surveys early or late. We performed a sensitivity analysis analyzing only attacks with 2‐h data time stamped 1.5‐ to 3‐h post‐vaporization, a time window selected to retain as much data as possible without increasing possible exposure to retention bias. ³² , ³³ , ³⁴ As per IHS guidelines, for all analyses, attacks for which the participant used rescue medication before the 2‐h assessment were counted as treatment failures. ³⁰ The differences in pain relief, pain freedom, and MBS freedom among the four treatments were assessed using a generalized linear mixed model (GLMM) with a logit link for binary outcomes. A random intercept structure was included to account for the cluster effect of participants going through the same trial multiple times. Secondary and other analyses at 1, 24, and 48 h were conducted using similar methods. Blinding effectiveness was assessed using Bang's Blinding Index (BI) for each treatment period, calculated for each active treatments and placebo with two‐sided 95% confidence intervals (CIs) without multiple comparison adjustment. The BI ranges from −1 to 1, with values greater than 0.2 indicating more correct guesses than expected by chance, values between −0.2 and 0.2 suggesting random guessing, and values less than −0.2 indicating more incorrect guesses than expected by chance. ³⁵ Significant unblinding was considered present when the 95% CI did not include zero. Model diagnostics were performed to assess uniformity, overall fit, dispersion, and the presence of outliers, confirming that the model assumptions were adequately met. ³⁶ Statistical analyses were performed using R 4.1.2 (R Foundation for Statistical Computing, Vienna, Austria) in the RStudio 2024.09.1 + 394 “Cranberry Hibiscus” environment (Posit PBC, Boston, MA, USA), with the lme4 and DHARMa, and BI packages used for model development, assessment, and BI calculation. ³⁷ , ³⁸ , ³⁹ All tests were two‐sided, with p < 0.05 indicating statistical significance.

RESULTS

A total of 678 people were screened for eligibility; of those, 92 were enrolled (Figure 1). Participants had a median age of 41 years, and 83% (76/92) were female (Table 1).

FIGURE 1.

Enrollment, randomization, and follow‐up. The intention‐to‐treat (ITT) analysis analyzed all treated attacks. The modified intention‐to‐treat analysis analyzed all attacks with 2‐h data. The sensitivity analysis analyzed all attacks with 2‐hour data time stamped 1.5‐3 hours post‐vaporization.

ITT, intention‐to‐treat; mITT, modified intention‐to‐treat.

TABLE 1.

Characteristics of patient population.

Characteristic	All patients (n = 92)
Age—median years (quartiles 1 to 3)	41 (34 to 54)
Birth sex—no. (%)
Female	76 (83%)
Male	16 (17%)
Gender identity—no. (%)
Women	77 (84%)
Men	15 (16%)
Other	0 (0%)
BMI—median (quartiles 1 to 3)	26.0 (22.8 to 32.9)
Ethnicity
Hispanic or Latino	15 (16%)
Not Hispanic or Latino	77 (84%)
Race or ethnic group—no. (%)
White	71 (77%)
Black	8 (9%)
Asian American	7 (8%)
Mixed/Other	2 (2%)
Middle Eastern	1 (1%)
Pacific Islander	1 (1%)
Declined to state	2 (2%)
Primary language—no. (%)
English	83 (90%)
Spanish	6 (7%)
Other	3 (3%)
Migraine history
Age at onset—median years (quartiles 1 to 3)	20 (14 to 28)
Headache days per month—median (quartiles 1 to 3)	15 (8 to 16)
Migraine days per month—median (quartiles 1 to 3)	6 (4 to 10)
Chronic migraine a —no. (%)	28/92 (30%)
Most bothersome symptom, treated attack (N = 234)
Photophobia	148 (63%)
Phonophobia	38 (16%)
Nausea	48 (21%)
Medical comorbidities
Anxiety	27 (29%)
Depression	20 (22%)
Insomnia	18 (20%)
Low back or lumbar spine problems	16 (17%)
Sinus problems	11 (12%)
Asthma	8 (8.7%)
Thyroid disease	5 (5.4%)
Head injury	5 (5.4%)
Hypertension	6 (6.5%)
Heart disease	4 (4.3%)
Gastrointestinal disease	4 (4.3%)
Arthritis	4 (4.3%)
Fibromyalgia	2 (2.2%)
Irritable bowel syndrome	2 (2.2%)
Neck or cervical spine problems	2 (2.2%)
Last cannabis use
Never	33 (36%)
More than 1 year ago	21 (23%)
Used within prior year	37 (40%)
Incomplete data	1 (1%)
Frequency of cannabis use
N/A	46 (50%)
Less than once a year	16 (17%)
Less than once a month	10 (11%)
Monthly	4 (4%)
Weekly	10 (11%)
Daily	5 (5%)
Incomplete data	1 (1%)

Abbreviations: BMI, body mass index; N/A, not applicable.

^a Chronic migraine: At least 15 headache days and at least eight migraine days per month.

The ITT analysis included all 247 migraine attacks treated with vaporized cannabis from 73 participants. The mITT analysis included 234 attacks from 71 patients. The sensitivity analysis included 202 attacks from 70 patients (Figure 1). The results were similar across the ITT, mITT, and sensitivity analyses for the primary endpoint of 2‐h pain relief and secondary endpoints of 2‐h pain freedom and 2‐h MBS freedom (Supporting Information S1). Where not otherwise specified, mITT results are reported because they neither used imputed data (as in the ITT) nor excluded data (as in the sensitivity analysis).

Primary endpoint: 2‐h pain relief

For the primary outcome of 2‐h pain relief, THC + CBD and THC‐dominant were superior to placebo but CBD‐dominant was not in all three analyses. In the ITT analysis, 2‐h pain relief responder rates were 64% (39/61) with THC + CBD, 68% (42/62) with THC‐dominant, 50% (30/60) with CBD‐dominant, and 42% (27/64) with placebo. In the mITT analysis, 2‐h pain relief was achieved by 67% (39/58) with THC + CBD (OR [95%] 2.85 [1.22 to 6.65], p = 0.016), 69% (42/61) with THC‐dominant (3.140 [1.35 to 7.30], p = 0.008), 53% (30/57) with CBD‐dominant (p > 0.05), and 47% (27/58) with placebo. In the sensitivity analysis, 2‐h pain relief responder rates were 64% (32/50) with THC + CBD, 70% (35/50) with THC‐dominant, 53% (26/49) with CBD‐dominant, and 45% (24/53) with placebo.

Co‐secondary endpoints: 2‐h pain freedom and 2‐h MBS freedom

For the secondary outcomes, THC + CBD was superior to placebo in all three analyses (Supporting Information S1) for 2‐h pain freedom (mITT 35% (20/58) vs. 16% (9/58), 3.30 [1.24 to 8.80], p = 0.017), and 2‐h MBS freedom (mITT 60% (35/58) vs. 34% (20/58), 3.32 [1.45 to 7.64], p = 0.005), but THC‐dominant and CBD‐dominant were not (Figures 2 and 3).

FIGURE 2.

Bar and line plots of efficacy data. (A) Efficacy outcomes at 2 h. p‐values were based on GLMM. (B) Pain freedom rates at 1 and 2 h and sustained pain freedom rates at 24 and 48 h. (C) MBS freedom rates at 1 and 2 h and sustained MBS freedom rates at 24 and 48 h (*p < 0.05). GLMM, generalized linear mixed model; MBS, most bothersome symptom.

FIGURE 3.

Forest plots of efficacy data. Forest plots of primary, secondary, and tertiary endpoints. (A) CBD versus placebo. (B) THC versus placebo. (C) THC + CBD versus placebo. CBD, cannabidiol; CI, confidence interval; MBS, most bothersome symptom; OR, odds ratio; THC, Δ9‐tetrahydrocannabinol.

THC + CBD was superior to placebo at 2 h with regard to freedom from photophobia and phonophobia but not nausea or vomiting, whereas THC‐dominant and CBD‐dominant were not superior to placebo for any of these (Figure 3).

1‐h endpoints

At 1 h, pain relief responder rate for THC + CBD (54% (30/56), 2.56 [1.05 to 6.25], p = 0.039), THC‐dominant (66% (38/58), 4.72 [1.89 to 11.81], p < 0.001), and CBD‐dominant (59% (33/56), 3.20 [1.31 to 7.82], p = 0.011) were all superior to the 37% (22/60) responder rate with placebo. One‐hour pain freedom was superior only for THC‐dominant versus placebo (17% (10/58) vs. 5% (3/60), 4.90 [1.12 to 21.34], p = 0.034), and 1‐h MBS freedom was superior for THC‐dominant (38% (22/58), 2.68 [1.06 to 6.79], p = 0.038) and CBD‐dominant (41% (23/56), 3.10 [1.21 to 7.91], p = 0.018), but not for THC + CBD (34% (19/56) vs. 2.19 [0.85 to 5.64], p = 0.103) vs. placebo (22% (13/60)).

24‐ and 48‐h endpoints

Only THC + CBD had superior sustained benefit versus placebo (Figure 2B,C). THC + CBD was superior to placebo regarding 24‐h sustained pain freedom (28% (14/49) vs. 11% (6/54), 3.45 [1.14 to 10.50], p = 0.029), 24‐h sustained MBS freedom (46% (23/50) vs. 25% (14/56), 2.83 [1.10 to 7.26], p = 0.031), and 48‐h sustained MBS freedom (40% (19/47) vs. 18% (9/49), 3.39 [1.24 to 9.32], p = 0.018), but the difference for 48‐h sustained pain freedom was nonsignificant (23% (11/48) vs. 10% (5/49), 2.77 [0.84 to 9.12], p = 0.094). THC‐dominant and CBD‐dominant were not different from placebo regarding 24‐ or 48‐h sustained pain freedom or sustained MBS freedom.

Adverse events

Adverse events are reported in Table 2. Mean subjective highness on a 0 to 10 scale at 1 h was greatest with THC‐dominant at 3.5 (SD 3.0), followed by 2.4 (SD 2.5) with THC + CBD, 1.5 (SD 2.0) with CBD‐dominant, and 0.6 (SD 1.20) with placebo. At 2 h, subjective highness was reduced to 2.4 (SD 2.8) with THC‐dominant, 1.3 (SD 1.9) with THC + CBD, 0.9 (SD 1.6) with CBD‐dominant, and 0.4 (SD 1.0) with placebo. THC‐dominant had the greatest euphoria, cognitive impairment, and subjective highness, followed by THC + CBD, then CBD‐dominant, and least with placebo. Across all four treatments, there were no serious adverse events (Table 3).

TABLE 2.

Rates of queried adverse events.

	Placebo	CBD	THC + CBD	THC
1 h (N)	N = 60	N = 56	N = 56	N = 58
Sleepiness	16 (27%)	21 (38%)	25 (45%)	24 (41%)
Euphoria	4 (7%)	5 (9.0%)	16 (29%)	21 (36%)
Cognitive impairment	4 (7%)	8 (14%)	12 (21%)	20 (34%)
Highness (0 to 10) Mean (SD)	0.6 (1.2)	1.5 (2.0)	2.4 (2.5)	3.5 (3.0)
2 h (N)	N = 58	N = 57	N = 58	N = 61
Sleepiness	22 (38%)	29 (51%)	18 (31%)	26 (43%)
Euphoria	1 (2%)	2 (4%)	6 (10%)	17 (28%)
Cognitive impairment	3 (5%)	4 (7%)	7 (12%)	17 (28%)
Highness (0 to 10) Mean (SD)	0.4 (1.0)	0.9 (1.6)	1.3 (1.9)	2.4 (2.8)

Abbreviations: CBD, cannabidiol; SD, standard deviation; THC, Δ9‐tetrahydrocannabinol.

TABLE 3.

Categorized free response adverse events.

	Placebo	CBD	THC + CBD	THC
Participants reporting any adverse events	3/60 (5%)	11/56 (20%)	11/56 (20%)	18/58 (31%)
Adverse events reported at 1 h	0	7	12	13
Sedation	0	2	3	0
Slowness	0	1	0	4
Paresthesias	0	0	1	2
Throat irritation	0	3	2	0
Increased appetite	0	2	1	1
Impaired cognition	0	0	2	1
Very high	0	0	0	1
Impaired focus	0	0	2	2
Dizziness	0	0	0	2
Dry mouth	0	2	1	0
Adverse events reported at 2 h	3/58 (5%)	4/57 (7%)	4/58 (7%)	11/61 (18%)
Sedation	0	3	2	2
Impaired focus	1	0	0	1
Slowness	0	0	0	3
Dizziness	0	0	2	2
Throat irritation	0	1	1	0
Dry mouth	2	0	1	0
Paresthesias	0	0	0	1
Increased appetite	0	0	0	1
Impaired cognition	0	0	1	1
Very high	0	0	0	0
Serious adverse events	0	0	0	0

Abbreviations: CBD, cannabidiol; THC, Δ9‐tetrahydrocannabinol.

The adverse events of subjective highness and euphoria were added as covariates to the GLMM to assess whether 2‐h pain relief, pain freedom, and MBS freedom remained significant. The statistical significance of the treatment type variables remained unchanged after adjusting for the covariates of euphoria and highness (Supporting Information S2A–F).

Other analyses

In this crossover study, the potential effects of treatment order and session number on the primary and key secondary outcomes were tested. Adjusting for treatment session number in the GLMM did not alter the results.

Exploratory outcomes included return to usual activity. Rates of return to usual activity at 2 h were THC + CBD 65% (37/57), THC 62% (38/61), placebo 57% (33/58), and CBD 47% (27/57). Rates of sleeping at 2 h by treatment were THC 15% (9/61), THC + CBD 12% (7/57), placebo 14% (8/58), and CBD 11% (6/57).

Post hoc subgroup analyses should be interpreted with caution because their analyses were not prespecified and sample sizes were small. Examining response rates with treatment at 0 to 2 h from migraine onset (early treatment) versus 2 to 4 h from migraine onset (late treatment) with THC + CBD; 2‐h pain relief was 67% (24/36) with early treatment versus 68% (15/22) with late treatment; 2‐h pain freedom was 39% (14/36) with early treatment versus 27% (6/22) with late treatment; and 2‐h MBS freedom was 67% (24/36) with early treatment versus 50% (11/22) with late treatment. Response rates were overall greater among attacks without allodynia at time of treatment versus those with allodynia at time of treatment; with THC + CBD, 2‐h pain relief was 74% (28/38) in attacks without allodynia versus 55% (11/20) in attacks with allodynia; 2‐h pain freedom was 40% (15/38) in attacks without allodynia versus 25% (5/20) in attacks with allodynia; and 2‐h MBS freedom was 68%% (26/38) in attacks without allodynia versus 45% (9/20) in attacks with allodynia. Response rates were similar in patients with episodic and chronic migraine; in attacks treated with THC + CBD, rates of 2‐h pain freedom were 34% (13/38) in people with episodic migraine versus 37% (7/19) in people with chronic migraine, and rates of 2‐h MBS freedom were 63% (24/38) in people with episodic migraine versus 58% (11/19) in people with chronic migraine.

Bang's BI was used to assess blinding but should be interpreted with caution because the sample sizes were small. For THC + CBD, BI was 0.11 (95% CI −0.28, 0.49) for treatment 1 (n = 19), −0.20 (95% CI −0.66, 0.26) for treatment 2 (n = 15), −0.15 (95% CI −0.62, 0.32) for treatment 3 (n = 13), and 0.36 (−0.09, 0.82) for treatment 4 (n = 11), within the limits for random guesses for treatments 1 to 3 and a nonsignificant tendency toward correct guessing for treatment 4. For placebo, BI was 0.31 (95%, CI −0.14, 0.75) for treatment 1 (n = 13), 0.00 (95% CI −0.43, 0.43) for treatment 2 (n = 13), 0.18 (95% CI −0.20, 0.55) for treatment 3 (n = 17), and 0.47 (95% CI 0.06, 0.87) for treatment 4 (n = 15), within the limits for random guesses for treatments 2 and 3, a nonsignificant tendency toward correct guessing for treatments 1 and significant correct guessing for treatment 4. For CBD, BI was 0.13 (95% CI −0.31, 0.58) for treatment 1 (n = 15), 0.00 (95% CI −0.43, 0.43) for treatment 2 (n = 17), −0.25 (95% CI −0.72, 0.22) for treatment 3 (n = 12), and −0.08 (95% CI −0.47, 0.32) for treatment 4 (n = 13), within the limits for random guesses for treatments 1, 2, and 4 and a nonsignificant tendency toward correct guessing for treatment 3. For THC, BI was −0.41 (95% CI −0.71, −0.11) for treatment 1 (n = 22), −0.07 (95% CI −0.54, 0.40) for treatment 2 (n = 15), −0.31 (95% CI −0.80, 0.19) for treatment 3 (n = 13), and 0.18 (95% CI −0.31, 0.67) for treatment 4 (n = 11), within the limits for random guesses for treatments 2 and 4, demonstrating significant incorrect guessing for treatment 1, and a nonsignificant tendency toward incorrect guessing for treatment 3 (Supporting Information S3).

DISCUSSION

In this study, the first randomized, double‐blind, placebo‐controlled trial testing the efficacy of cannabinoids for acute migraine, 6% THC + 11% CBD was superior to placebo for pain relief, pain freedom, and MBS at 2 h, as well as freedom from photophobia and phonophobia at 2 h and 24‐h sustained pain freedom and sustained MBS freedom and 48‐h sustained MBS freedom. BI indicated successful blinding with THC + CBD, and adjusting for psychoactive side effective did not change the results of the GLMM, suggesting that this study's findings are not explainable by expectation effects or unmasking due to psychoactive side effects.

Whereas 6% THC‐dominant was superior to placebo for the primary outcome of pain relief at 2 h, it was not superior to placebo for the key secondary outcomes of pain freedom or MBS freedom at 2 h, did not demonstrate superior sustained pain freedom or MBS freedom versus placebo at 24 or 48 h, and had higher rates of adverse events than THC + CBD.

THC potencies (6%) were lower than typical with cannabis available via US dispensaries, and subjective highness (2.4 out of 10 at 1 h) with THC + CBD was roughly half of that reported by research participants using cannabis ad libitum, ⁴⁰ bolstering evidence that higher potencies and titrating to highness are unnecessary for medicinal benefit. THC + CBD had less euphoria, subjective cognitive impairment, and subjective highness than THC‐dominant. This is explainable by CBD decreasing the psychoactive effects of THC through its action as a noncompetitive, negative allosteric modulator of the CB1 receptor. ⁴¹ , ⁴² , ⁴³

THC + CBD's efficacy for reducing photophobia and phonophobia but not nausea at 2 h in this study may be explainable by lower time 0 rates of nausea (59.9%) than photophobia (91.9%) and phonophobia (87.0%) and high rates of freedom from nausea at 2 h with both the THC + CBD (75.9%) and placebo (70.7%) treatments in this study. This may be indicative of a large placebo effect on nausea in this study. Rates of vomiting were < 10% across all treatments, and differences were not significant. That THC + CBD reduced photophobia and phonophobia but not nausea or vomiting in this study demonstrates that THC + CBD's superior 2‐h MBS freedom versus placebo is due to antimigraine effects rather than THC's established antiemetic effects ³¹ (Figure 3).

In this study, Bang's BI demonstrated overall satisfactory blinding with significant correct guessing in only one of 16 cohorts of studied migraine attacks—those attacks treated with placebo for treatment 4 (participants correctly guessing that they had received placebo or CBD‐dominant). There was also significant incorrect guessing in one of 16 cohorts of studied migraine attacks—those attacks treated with THC for treatment 1 (participants incorrectly guessing that they had received placebo or CBD‐dominant). BI in this study was limited by small sample sizes due to the crossover design of this study and the necessity of categorizing participants' guesses into THC‐containing (THC‐dominant or THC + CBD) versus non‐THC‐containing (placebo or CBD‐dominant). Focusing on the BI results of the THC + CBD treatment, it is important to note that blinding was overall satisfactory in attacks treated with THC + CBD, with a nonsignificant tendency toward correct guessing in migraine attacks treated with THC + CBD as treatment 4.

Strengths of the study include that it was a randomized, double‐blind, placebo‐controlled study conducted with foundation funding and cannabis from the NIDA DSP and without industry involvement. However, this also limits the generalizability of the study's findings to commercial products used by patients in real‐world settings. A 2 × 2 study design (THC, CBD, THC + CBD, placebo) was used to study the benefits of THC and CBD separately, as well as their synergistic effect, within the same study population. To our knowledge, this is the first RCT studying an acute migraine treatment to include patients with both episodic and chronic migraine frequencies and to include subgroup analyses of response rates in both populations. This increases the generalizability of its findings to neurology and headache subspecialty clinics. However, given the study's high screen fail rate and demanding research protocol, the study many not be fully representative of the general migraine population. A strength and weakness of this study is that it only examined THC and CBD; minor cannabinoids were <1% and the treatments were devoid of terpenes. This minimizes confounding variables; however, future studies will be necessary to further study the possible role of minor cannabinoids and terpenes for the treatment of migraine. Other limitations of this study include that it only studied single potencies of THC of CBD and a single THC:CBD ratio. This study included patients who were not cannabis‐naïve, which may have impacted placebo rates and unblinding. Also, due to the limited number of observations, our statistical models included only random intercepts, so potential heterogeneity in treatment effects across participants could not be fully accounted. Another limitation is that pain relief at 2 h was the primary outcome, which differs from coprimary outcomes of pain freedom at 2 h and MBS freedom at 2 h in most acute trials. We prospectively selected pain relief at 2 h as the primary outcome and pain freedom and MBS freedom at 2 h as our two co‐secondary outcomes based on prior evidence of pain relief without prior evidence of pain freedom or MBS freedom with cannabinoids for migraine ¹⁹ and a concern that MBS freedom rates could be superior to placebo due to THC's known antiemetic effects rather than due to antimigraine effects. ³¹ We were concerned that if the results would be positive for MBS freedom due to the known antiemetic effects of THC, an otherwise negative study might be interpreted by some as a positive study for having achieved one of its coprimary outcomes. Therefore, we prospectively selected 2‐h pain relief as our sole primary outcome. Future studies should use pain freedom at 2 h and MBS freedom at 2 h as coprimary outcomes.

Typical of acute migraine RCTs, this study did not assess repeated administrations or regular, long‐term treatments to assess possible risks (including rates of development of medication overuse headache, cannabis use disorder, cannabis hyperemesis syndrome, cognitive impairment, and psychiatric sequela) or benefits (including possible migraine preventive effects). ³⁰ , ⁴⁴ A small body of literature shows improvements on patient‐reported outcomes when using cannabis‐based medicinal products on a regular, longer term or preventive basis. ²¹ , ⁴⁵ More research is needed to evaluate repeated administrations and regular, long‐term use of cannabinoids for migraine.

CONCLUSION

In this first randomized, double‐blind, placebo‐controlled trial testing the efficacy of cannabinoids for the acute treatment of migraine, vaporized 6% THC + 11% CBD cannabis flower was superior to placebo for pain relief, pain freedom, and MBS freedom at 2 h as well as 24‐h sustained pain freedom and sustained MBS freedom and 48‐h sustained MBS freedom. Future research should include multicenter RCTs and long‐term studies of benefits and risks with repeated use.

AUTHOR CONTRIBUTIONS

Nathaniel M. Schuster: Conceptualization; investigation; funding acquisition; writing – original draft; methodology; supervision; data curation; formal analysis; visualization; project administration; software. Mark S. Wallace: Resources; writing – review and editing; investigation; project administration. Thomas D. Marcotte: Resources; writing – review and editing; project administration. Dawn C. Buse: Writing – original draft; writing – review and editing; methodology. Euyhyun Lee: Writing – review and editing; writing – original draft; data curation; formal analysis; visualization. Lin Liu: Writing – review and editing; data curation; supervision; formal analysis; visualization; writing – original draft. Michelle Sexton: Writing – review and editing; writing – original draft; methodology; software.

FUNDING INFORMATION

This study was sponsored by the Migraine Research Foundation, now administered by Hartford HealthCare. The project described was partially supported by the University of California San Diego (UCSD) Academic Senate, the National Institutes of Health (NIH) Clinical and Translational Science Awards (CTSA) grant UL1TR001442, and the generous support of Leonard and Marilyn Berger. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.

CONFLICT OF INTEREST STATEMENT

Nathaniel M. Schuster has received research funding from Migraine Research Foundation, ShiraTronics, National Institute of Neurological Disorders and Stroke (NINDS) U24 NS115714, NINDS 3OT2NS122680‐01S3, National Center for Complementary and Integrative Health 5R00AT009466‐04, National Institutes of Health CTSA grant UL1TR000100, UCSD Academic Senate, and UCSD Department of Anesthesiology Research Advisory Group; has served on advisory boards for Lundbeck, Rapport, and Vertex; and has consulted for Averitas, Delphian, Eli Lilly, Lohocla, Salvia, Syqe and Vertex. He serves on the editorial board of Pain Medicine (Oxford University Press) and Interventional Pain Medicine (Elsevier). Dawn C. Buse has been a consultant for Abbvie, Amgen, Biohaven, Lilly, Lundbeck, Teva, and Theranica. She has received grant support from the US Food and Drug Association and National Headache Foundation. She is an editor for Current Pain and Headache Reports. Mark S. Wallace, Thomas D. Marcotte, Euyhyun Lee, Lin Liu, and Michelle Sexton have no conflicts to report.

AI USE DISCLOSURE

Artificial intelligence tools were used solely to improve clarity and conciseness in limited portions of the article.

Supporting information

Table S1. Modified intention‐to‐treat, intention‐to‐treat, and sensitivity analyses for primary and secondary endpoints.

Table S2. Table S2A: Treatment model for pain relief adjusting for feeling euphoric at 2 h.

Table S2B. Treatment model for pain freedom adjusting for feeling euphoric at 2 h.

Table S2C. Treatment model for most bothersome symptom freedom adjusting for feeling euphoric at 2 h.

Table S2D. Treatment model for pain relief adjusting for feeling high at 2 h.

Table S2E. Treatment model for pain freedom adjusting for feeling high at 2 h.

Table S2F. Treatment model for most bothersome symptom freedom adjusting for feeling high at 2 h.

Table S3. Bang’s blinding index.