Dose-dependent Effects of Smoked Cannabis on Capsaicin-induced Pain and Hyperalgesia in Healthy Volunteers

The institutional review board of the University of California at San Diego approved the study. All individuals gave informed, written consent before participating in the research. Recruitment of healthy volunteers was conducted by advertisement in local print media and word of mouth. Inclusion criteria were men and women aged 18 yr or older, in good health, English speaking, literate, and able to understand the study procedures and communicate with the research team. Exclusion criteria included (1) active acute or chronic medical illness or pain problems; (2) current or past cannabis abuse or dependence, current other psychoactive substance use disorder, or major mental disorder (e.g. , major depression or psychosis) as determined by Diagnostic and Statistical Manual of Mental Disorders , Fourth Edition criteria; (3) pulmonary disease; (4) lack of use of cannabis within the past 6 months; (5) pregnancy; and (6) allergy to the study drug. These criteria were intended to identify persons likely to be able to successfully complete the brief research protocol, based on evidence of having some experience with smoking cannabis within the past 6 months and thus being more likely to tolerate the delivery method of smoking. Subjects were asked to abstain from smoking cannabis for at least 30 days before administration of the study drug.

A randomized, double-blinded, placebo-controlled, crossover design was used. Before the blinded phase, a blood sample was drawn to screen for the absence of tetrahydrocannabinol (THC); then, all subjects participated in a high-dose training session, using the standardized cued-dosing procedure described below. If subjects were unable to tolerate the highest dose, they were excluded from the study. Eligible subjects participated in four dose-randomized sessions separated by at least 1 week. At each session, subjects were exposed to placebo or a low (2%), medium (4%), or high (8%) dose of cannabis (see Cannabis Assignment and Dosing).

At each visit, before study drug administration, a blood sample was taken for plasma assay of THC and metabolites (see below for detail plasma assay of THC and other test procedures) and the following data were collected: (1) neurosensory testing (thermal sensation, thermal pain, touch, and mechanical pain) on the volar aspect of both forearms; (2) neurocognitive evaluation; and (3) blood pressure, heart rate, respiratory rate, and temperature. The study drug was then administered to the subjects. At 5 min after the dose, a sample of venous blood was taken to quantify acute cannabis exposure, and the following assessments were performed: (1) neurosensory testing on the right forearm, (2) neurocognitive testing, and (3) a subjective rating of “highness.” After completing the testing, capsaicin (10 μl, 10 mg/ml) was injected intradermally on the volar aspect of the right forearm. Pain scores, blood pressure, heart rate, and respiratory rate were measured at the time of injection and every 2.5 min for 10 min. A McGill Pain Questionnaire (MPQ) was administered at the time of capsaicin injection only. Ten minutes after the capsaicin injection, the hyperalgesic area was established to von Frey hair and stroking, the flare response was outlined, and neurosensory testing was performed halfway between the edge of this defined area and the capsaicin injection site. Forty minutes after cannabis dose, a final blood sample was taken from the right antecubital vein for plasma assay of THC and metabolites, and the neurosensory testing (left forearm), neurocognitive testing, and subjective “highness” ratings were repeated. After completing the testing, capsaicin (10 μl, 10 mg/ml) was injected intradermally on the volar aspect of the left forearm, and the methods of testing described for the right forearm were repeated. Figure 1illustrates the schedule of assessments.

Each subject received placebo and three doses of cannabis, given in random order as determined by a computerized random number generator held by the University of California, San Diego Research Pharmacy, which alone had access to the assignment scheme and order. Standardized cannabis and placebo cigarettes were provided by the National Institute on Drug Abuse and were constructed from similar base material using THC concentrations of 0%, 2%, 4%, and 8% by weight. Because of the variable nature of plant material, actual cigarette concentrations varied within nominal dose ranges as follows: low dose (2%), 1.76–2.03%; medium dose (4%), 3.3–3.96%; and high dose (8%), 6.3–7.95%. Placebo cigarettes were prepared from ethanol-extracted cannabis, which reduces the cannabinoid content of the plant material to trace amounts, 0.009% THC in this study. These placebo cigarettes are visually indistinguishable from cigarettes in the experimental arm and maintain the distinctive taste and odor of the active material.

Dosing levels were controlled by using a standardized cued-smoking procedure.24Study treatments were administered under direct observation by a study nurse who instructed the participant to light the cigarette and, once lit, inhale for 5 s. The subject was then instructed to remove the cigarette from the lips and hold the inhalation for 10 s (if possible) before exhaling fully. The process was repeated three more times with the subject being given a 40-s resting period between smoking sessions. Smoked cigarettes were collected after each session for weighing and, to ensure proper custody, were stored under locked conditions in the Research Pharmacy.

Three pain intensities were measured after capsaicin injection: (1) spontaneous pain (visual analog scale of pain intensity [VASPI]), (2) pain resulting from gently stroking the injected area (brush), and (3) pain resulting from application of a 5.18 von Frey hair to the painful area (von Frey). Pain scores were measured using a visual analog scale consisting of a 100-mm line with “no pain” written at one end and “worst imaginable pain” written at the other end. Subjects were asked to place a mark along the line in a location corresponding with the intensity of their pain. The distance, in millimeters, from the “no pain” end to the location of the mark yields a measurement of the pain intensity.

The McGill Pain Questionnaire Short Form (MPQ-SF)25was used to assess the quality of the pain experience. This instrument consists of 15 pain descriptors, assessing both affective (e.g. , tiring, punishing) and sensory (e.g. , throbbing, shooting) dimensions of pain. The MPQ is generally used in chronic clinical pain, but it has a long history of application in experimentally induced and acute postoperative and dental pain research with demonstrated sensitivity to change in loading on sensory-discriminative and motivational-affective dimensions of the pain experience.26,27Given the known psychotropic effects of cannabis, we included the MPQ as a secondary outcome measure of the affective dimension of pain reports.

Four neurosensory tests were performed: (1) warm and cold sensation, (2) warm and cold pain, (3) touch, and (4) mechanical pain. The neurosensory tests were ordered from least noxious to most noxious (warm/cold thresholds, hot/cold pain thresholds, mechanical thresholds, and mechanical pain). Warm and cold sensation was measured using a Thermal Sensory Analyzer (Medoc Advanced Medical Systems, Minneapolis, MN). A 2 × 2-cm probe was used with a rate of temperature change of 1°C/s for warm and cool and 1.5°C/s for warm and cold pain. Warm and cold pain measurements used the same instrument to obtain an endpoint of pain rather than temperature change sensation. The subject was given four trials with the warm and cold stimulus and three trials with the warm and cold pain stimulus. The average of each stimulus was used to determine the threshold. Touch was measured using calibrated von Frey hair filaments of varying size. The filaments were selected at random, and three successive stimuli were applied for 2 s at 5-s intervals per filament, applied in an ascending pattern. The patient was instructed to report if the stimulus was felt. Thresholds are expressed in millinewtons. Mechanical pain was also measured using the von Frey hair filaments but with the endpoint being pain.

One hundred milligrams capsaicin (8-methyl N-vanillyl 6-nonamide), dissolved in 10 ml of a 20% cyclodextrin vehicle to achieve a concentration of 10 μg/μl, was prepared after aseptic precautions. A volume of 10 μl was then injected intradermally with a sterile tuberculin syringe. The region of secondary hyperalgesia was established with a 5.18 von Frey hair and foam brush gently stroked on the skin. These stimuli began away from the injection site in an area of skin that did not produce pain and were repeated tangentially to the injection site at a progressively closer radius until the subject reported pain or tenderness. That site was marked on the skin with a felt-tip pen and a new series was started from the periphery at a different angle until eight determinations of the borders of secondary hyperalgesia were outlined on the skin. These borders, as well as the flare response, were outlined onto a transparency for area determination (cm²).

Cannabinoid concentrations in plasma were ascertained by gas chromatography mass spectrometry for (1) δ-9-tetrahydrocannabinol (δ-9-THC), the principal active compound; (2) 11-nor-9-carboxy-tetrahydrocannabinol (11-nor-THC), the principal inactive metabolite of δ-9-THC; (3) 11-hydroxyl-tetrahydrocannabinol (11-OH-THC), the principal active metabolite of δ-9-THC; (4) cannabinol, a secondary active cannabinoid; and (5) cannabidiol, a secondary active cannabinoid.

At enrollment, subjects underwent a directed physical examination, vital signs, and an electrocardiogram and/or chest x-ray. They were given information about the subjective effects of cannabis and instructed in relaxation techniques, should those effects become disturbing. Vital signs were monitored throughout the protocol, and subjects remained in the laboratory under direct observation by staff for 2 h after the cannabis procedures were completed. Before release from the clinic, a final vital sign and self-report status check was made, and the subject was transported from the clinic by taxicab or prearranged transportation.

Measures of cognitive performance were obtained for safety, and to assess for potential confounding of pain reports:

This is a measure of psychomotor speed, attention, and cognitive sequencing that requires subjects to connect a series of randomly arranged circles in a designated sequential order, based on alternating numbers and letters (i.e. , 1 to A to 2 to B, and so forth) Scores are equal to time taken to complete the task in seconds.28 

In this test,29a set of randomized digits was serially presented via  tape recording. Subjects were asked to add the current number to the number that preceded it and respond with the sum. Thus, after each new digit was presented, a new total was achieved. Scores were equal to the number of correct responses out of 50 items presented. This test yielded a measure of speed of information processing.

Subjects were asked to rate their feeling of “high” on a 10-point scale from 0 =“not high at all” to 10 =“the highest you’ve ever been.”

The Beck Depression Inventory II was administered once at each visit before exposure to cannabis. The Beck Depression Inventory II30consists of 21 questions, each graded on a four-point scale ranging from 0 to 3. Statements are ordered to show increasing severity of the cognitive and somatic dimension of depressed mood. We used the Beck primarily to exclude smoking for subjects who might be depressed or suicidal—and thus vulnerable to the known adverse psychotropic effects of cannabis.

The software used for the analysis was R: Version 2.3.1 (2006-06-01) (R Foundation for Statistical Computing, Vienna, Austria). Mixed-effects, repeated-measures linear regression was used to address the main question about the relation of pain and cannabis. Pain scores were modeled as a function of cannabis dose and time. The pain score (brush, VASPI, or von Frey, considered separately) was the dependent variable in regression, whereas cannabis dose was the main predictor of interest. The decrease of pain over time was expected to be nonlinear, so a quadratic time component was modeled, as well as a random (subject-specific) intercept. The early and late capsaicin injections were modeled separately. The assumption that the dose curves were parallel over time was checked by modeling the interaction between time and dose. In addition, an analysis as described above was performed using a linear combination of the three pain scores (first principal component) instead of a single pain score as the outcome. In addition, the first pain score (brush, VASPI, or von Frey) obtained after each of the two (early and late) capsaicin injections was modeled as the function of the amount of each of cannabinoids/metabolites measured during the blood draw closest to the time of the cannabis dose. This model was as similar as possible to the model used to assess the main pain–cannabis dose relation, although there were some design-imposed departures. In addition to the subject-specific intercept, binary indicators of the medium and high doses were included as a conservative measure. The question became “Over and above the gross dose effects established in the main analysis, was there further explanatory power in the assayed plasma concentration?” The difference in time between blood draw and capsaicin injection was also included as a covariate. The effect of cannabis on heart rate, respiratory rate, and blood pressure was examined in a repeated-measures model similar to the one described above, where heart rate and blood pressure were modeled as functions of cannabis dose and time since smoking. Beck scores were examined over the four visits. The difference in the baseline and postcannabis neurocognitive test performance (at 5 and 40 min past smoking) was assessed in a series of paired Wilcoxon rank sum tests.

Of 29 subjects screened, 4 were ineligible because one or more exclusion criteria were met. Twenty-five subjects entered the high-dose training session, of which 2 dropped out because of time constraints, 2 were lost to follow-up, 1 dropped out because of fear of the blood draws, and 1 was excluded because of anxiety after cannabis exposure. Nineteen subjects were randomized in the blinded phase of the study. Participants in the study had a mean age of 28.9 yr (SD = 10.9 yr) and were 58% male (n = 11), 95% unmarried (n = 18), and 37% white (n = 7). After randomization, 2 subjects dropped out because of time constraints, 1 was lost to follow-up, and 1 dropped out because of dizziness. Fifteen subjects completed the protocol.

Intradermal capsaicin injections induced spontaneous pain and elicited pain (stroking and von Frey stimulation) in all subjects. Repeated-measures analysis of subjective pain scores showed no difference in pain perception between any of the cannabis doses and placebo during the early time course (right arm) on any measure, (table 1), and the low dose did not differ from placebo at any time point (tables 1 and 2). However, during the late time course (left arm), both medium and high doses differed significantly from placebo, albeit in opposite directions (table 2). At the medium dose, subjects reported decreased pain sensation, above and beyond the decrease to be expected as a function of time elapsed after capsaicin (P  values of the medium dose coefficients for the subjective pain scores ranged between 0.011 and 0.027; overall decrease in pain score between medium dose and placebo ranged between 6.2 and 6.7 points). At the high dose, subjects reported increased perception of pain (P  values of the high dose coefficients for the subjective pain scores ranged between 0.009 and 0.002; overall increase in pain score between medium dose and placebo ranged between 7.1 and 8.7 points).

The results were similar for all three pain measures: brush, VASPI, and von Frey. On further examination, these tests were found to be highly correlated (Pearson r = 0.985 between brush and VASPI and between VASPI and von Frey; Pearson r = 0.974 between brush and von Frey) and were subsequently combined in a first principal component analysis. The plots of raw and fitted data for all pain measures and the resultant composite pain scores are presented in figures 2 and 3. The assumption of the parallel dose curves present in the random effects model was tested by modeling time–dose interactions. The interactions were not significant, which supports the parallel dose curves assumption.

There was no effect of cannabis on the pain quality as measured by the MPQ-SF. Total MPQ-SF scores for the early and late time period were as follows: placebo, 8.3 ± 3.0 and 8.8 ± 3.0; low dose, 9.9 ± 3.1 and 8.3 ± 2.5; medium dose, 9.7 ± 3.7 and 11 ± 3.9; and high dose, 9.6 ± 3.1 and 8.6 ± 2.7. Neither sensory nor affective dimensions were affected.

Capsaicin produced a secondary hyperalgesia to stroking and von Frey hair stimulation in 14 of 15 subjects and a flare response in all subjects. Twelve of 15 subjects had heat hyperalgesia as evident by decrease in the hot pain thresholds after capsaicin injection. Cannabis did not attenuate the heat hyperalgesia at any dose (fig. 4). Furthermore, there was no effect of any dose of inhaled cannabis on the secondary hyperalgesia or flare response in either the early or the late time course (fig. 5). Baseline neurosensory thresholds did not differ significantly for any of the measures recorded across the placebo and cannabis test days. There was no significant effect of cannabis on any neurosensory threshold, painful or nonpainful (table 3).

The percent weight (±SD) of the cannabis cigarette smoked for the low, medium, and high dose was 49 ± 11, 44 ± 9, and 41 ± 13, respectively. Mixed-effects regression analysis of the assayed concentration of cannabinoids/metabolites in the blood as predictors of pain perception showed that δ-9-THC and 11-OH-THC were significantly negatively associated with subjective pain scores both at the early and the late capsaicin injection (P  values of corresponding regression coefficients ranging from 0.009 to 0.05; partial correlations ranging from −0.52 to −0.43). At the late injection, the high-dose indicator showed borderline significance in increasing pain perception. 11-Nor-THC was significantly negatively correlated with estimated amount of pain at the early capsaicin injection but not at the late injection (P  values of early injection regression coefficient ranging from 0.007 to 0.07; partial correlations ranging from −0.52 to −0.36). Cannabinol and cannabidiol did not show any association with the pain scores obtained at early or late capsaicin injection, although this was expected, given that neither cannabinol nor cannabidiol were present at levels above 0.25% in the cannabis used in this study. Table 4summarizes the assayed plasma levels of the THC and other cannabinoids and metabolites.

Adverse events are summarized in table 5. Mild to moderate side effects occurred in 7 of 19 randomized subjects, primarily at the highest dose of cannabis. No serious adverse events occurred. Repeated-measures analysis of the vital signs data showed that all doses of cannabis increased heart rate compared with placebo. Specifically, the low, medium, and high doses resulted in heart rate increases of 7.9, 7.5, and 12.0 beats/min, respectively, consistent with previous reports of the cardiovascular effects of cannabis. In the case of respiratory rate, only the high dose of cannabis was significantly different from placebo (+0.79 respirations), whereas other cannabis doses showed no effect. Low-dose cannabis resulted in lower systolic blood pressure (−3.0 mmHg), although other doses did not replicate this effect. None of the cannabis strengths tested showed any effect on diastolic blood pressure. There was no difference in Beck scores at the four visits, and the summary of scores was at the low end of the Beck scale (Kruskal-Wallis test P  value = 0.82). On measures of neuropsychological functioning, there was no significant difference in performance in Paced Auditory Serial Attention Test total correct, Trails B time to complete before and after cannabis exposure. There was a slight (not reaching significance) worsening on medium and high doses between baseline and 5 min after exposure, which stabilized at 40 min after exposure (fig. 6). However, subjects did report a dose-dependent increase in their sensation of “high,” as ascertained by their rating of “highness” on a scale from 1 to 10 (fig. 7). This effect persisted into the late time course.

This study demonstrated different effects of three doses of smoked cannabis on spontaneous and elicited pain secondary to intradermal capsaicin injection. The medium cannabis dose, 4% THC by weight, produced delayed analgesia, whereas the high dose, 8% THC cannabis, produced a delayed increase in pain. The low dose had no analgesic effect. There was a significant correlation between plasma levels of THC and metabolites with decrease in pain; however, there was no correlation between the high-dose plasma levels and increase in pain. This suggests that there may be another compound within the cannabis leaf that we did not measure that may be leading to the increased pain at the high dose. It is known that the cannabis leaf contains more than 400 compounds, of which 60 are called the cannabinoids.31We only measured plasma levels of three compounds.

The delayed onset of analgesia is surprising considering that the subjective ratings of “high” peaked early, suggesting early central nervous system penetration. However, there may be a dissociation between analgesia and side effects, a phenomenon seen with other analgesics such as the opioids. The increase in pain observed with the high dose is consistent with previous analgesic studies with cannabinoids. For example, chronic delivery of cannabinoids has been shown to cause thermal hyperalgesia,32although the mechanism of this pronociception is unclear. Another possible explanation for the increased pain at the high dose seen in our study may be that the emotional effects produced by cannabinoids, e.g. , dysphoria, could counteract the analgesic effect. However, there was no effect observed in the affective score on the MPQ to support this hypothesis.

In contrast to the effects on spontaneous and elicited pain, we found no effect of inhaled cannabis on acute painful and nonpainful heat, cold, and mechanical thresholds. This finding is in conflict with preclinical studies on the effects of cannabinoids on acute nociceptive processing, demonstrating that administration of cannabinoids to normal animals produces both thermal1,6,7,33and mechanical34–36antinociception via  the CB¹receptor. Two previous studies have demonstrated that acute delivery of oral THC22and smoked cannabis37resulted in a decreased pain response to radiant noxious heat. Both studies resulted in significant psychomotor side effects that suggest that the dose of cannabis required to affect acute nociception may lead to psychomotor side effects. Likewise, our study showed no effects on acute nociceptive processing at doses that had minimal psychomotor effects.

It has been noted that CB¹receptors are located on the periaquaductal gray, raphe nucleus, and forebrain, which are known to process nociceptive input.3,5,7However, the site of analgesic action of the cannabinoids is unknown. It has been suggested that human experimental pain can be used to evaluate analgesic site of action. Although far from conclusive, there are components of the intradermal capsaicin response that may be used to evaluate the site of action. Intradermal capsaicin triggers selective and transient (<20–30 min) activation of C fibers, resulting in a rapid onset of pain, secondary hyperalgesia, and a flare response.18The spontaneous pain is mediated by both peripheral and central mechanisms, the secondary hyperalgesia is mediated by spinal mechanisms, and the flare response represents antidromic invasion of the axon collaterals and the subsequent release of neuropeptides, which is a peripheral mechanism.38,39Therefore, a review of the effects of analgesics with known mechanisms may suggest the site of action of the cannabinoids. For example, previous studies with systemically mediated opioids and N -methyl-d-aspartate antagonists show an effect on capsaicin-induced pain and secondary hyperalgesia but no effect on the flare response.40,41In addition, intravenous lidocaine has been shown to block the flare response of intradermal capsaicin, whereas the opioids have no effect.42 

It is difficult to draw firm conclusions on the predictive value of human experimental pain for drug efficacy in clinical pain because studies involving experimental pain often test drug efficacy with a single dose of drug and often administer the drug before the initiating stimulus. Nonetheless, it has been suggested that such models may provide a link between preclinical animal pain models and clinical trials in patients with chronic neuropathic pain.41,43,44Comparing our results with the human experimental pain literature on agents with known clinical efficacy has yielded two comparisons of interest. First, cannabinoids behave similarly to opioids in models of facilitated pain (i.e. , intradermal capsaicin, heat-capsaicin sensitization, first-degree burn). These protocols result in a brief report of intense pain followed by a longer lasting area of secondary hyperalgesia. Both the pain and the hyperalgesia of these models have shown consistent responses to opioids40,41and inconsistent responses to nonopioids.44–46Our results indicate that inhaled cannabis decreases the pain of intradermal capsaicin (within a therapeutic window) but has no antihyperalgesic effects. Second effects of cannabinoids resemble the actions of nonopioids in acute pain models (i.e. , thermal and mechanical) involving a brief report of pain that quickly resolves when the stimulus is removed. These models are consistently sensitive to opioids and resistant to nonopioids.41–43,47Our results suggest that the cannabinoids act more like the nonopioids on acute nociception.

Limitations on the generalizability of the current study for human experimental or clinical research include the small sample size and use of only healthy volunteers. In addition, only subjects who were experienced cannabis users and who were able to tolerate the highest study dose of cannabis were randomized. It is possible that clinically ill samples, especially cannabis-naive subjects, would have a different analgesic response and incidence of side effects when exposed to the effective dose found in this study (4% THC).

Results of this study may, however, raise interesting questions of relevance to the design of human experimental pain models—or perhaps of clinical trials assessing the potential therapeutic use of cannabinoids. We identified a potential narrow “therapeutic window” for analgesic efficacy. Future studies might include a more comprehensive pharmacokinetic assessment, with the goal of elaborating the time course of analgesia. Analgesic effects were delayed, were modest, and may or may not be translated into the clinical arena. The biphasic response (analgesia at medium dose and hyperalgesia at high dose) is of concern. It is unknown whether this property is unique to cannabis or unique to the cannabinoids as a class.

Finally, because more information is needed regarding abuse potential, tolerance, efficacy in neuropathic pain, and safety issues of inhaled cannabis, we cannot advocate a place for using cannabis in the treatment armamentarium at this time. A concern over the clinical use of inhaled cannabis is health-related issues that result from the delivery method. Long-term use of inhaled cannabis has been shown to be associated with increased respiratory symptoms suggestive of obstructive lung disease; however, short-term use of inhaled cannabis does not seem to be associated with respiratory complications.48Long-term use of inhaled cannabis has not been associated with increased aerodigestive cancers as is seen with tobacco use.49Another safety issue with cannabis relates to the psychotropic effects of cannabis, and its known “paradoxical” effects (e.g. , dysphoria, dejection, depressed mood). Such effects must be carefully considered in work addressing the future clinical application of cannabinoids. As for neurocognitive effects of long-term cannabis use, a large meta-analysis showed no effects on memory, recall, speeded information processing, or executive function.50 

In summary, in this model of human experimental pain, smoked cannabis was demonstrated to have a delayed biphasic effect on pain scores induced by intradermal capsaicin. The low dose had no effect, the medium dose significantly reduced the pain, and the high dose significantly increased the pain. There was no effect on capsaicin-induced secondary hyperalgesia, acute sensory thresholds, or neurocognitive assessments. There was a significant correlation between plasma levels of THC and metabolites with decrease in pain; however, there was no correlation between the high-dose plasma levels and increase in pain. No conclusions on the analgesic efficacy of smoked cannabis on clinical pain states can be made from this study because the relation between analgesic effects in experimental pain and clinical pain states is unknown.