Cannabinoid Therapeutics in HIV: Strong Pain Trials, an Appetite Literature From Another Era, and Interaction Data That Has Not Been Updated

Two randomized placebo-controlled trials of smoked cannabis in HIV-associated sensory neuropathy produced the clearest results in this literature. In Abrams and colleagues’ trial in Neurology, 52 percent of the cannabis group achieved more than 30 percent pain reduction versus 24 percent on placebo. In Ellis and colleagues’ crossover trial in Neuropsychopharmacology, the corresponding proportions were 46 percent and 18 percent.

The appetite and weight literature is weaker than its reputation. In the pivotal 1995 dronabinol trial, appetite improved significantly, but weight change did not reach statistical significance, and the trial was conducted in a population defined by untreated infection.

People living with HIV use cannabis at higher rates than the general population, often for pain, nausea, appetite, and sleep. A clinician asked about it deserves to know which of those indications has trial support specific to HIV and which does not.

The interaction question is the one that has quietly aged out of usefulness. The reassurance most often cited comes from pharmacokinetic work with indinavir and nelfinavir, protease inhibitors that have largely been replaced. Whether that reassurance extends to contemporary regimens has not been tested.

Donald Abrams and colleagues at the University of California, San Francisco ran a prospective randomized placebo-controlled trial of smoked cannabis in painful HIV-associated sensory neuropathy, published in Neurology in 2007. Fifty patients completed the protocol, smoking either cannabis at 3.56 percent THC or identical cigarettes from which the cannabinoids had been extracted, three times daily for five days in an inpatient research unit. Daily pain fell by a median of 34 percent on cannabis versus 17 percent on placebo. More than 30 percent pain reduction, the conventional threshold for a clinically meaningful response, was reached by 52 percent of the cannabis group and 24 percent of the placebo group. The first cannabis cigarette alone reduced chronic pain by a median of 72 percent versus 15 percent.

Ronald Ellis and colleagues at the University of California, San Diego replicated the finding with a different design: a phase II double-blind placebo-controlled crossover trial in HIV-associated distal sensory polyneuropathy, published in Neuropsychopharmacology in 2009. Thirty-four patients enrolled and 28 completed both treatment arms, using cannabis ranging from 1 to 8 percent THC four times daily for five days during each of two treatment weeks separated by a two-week washout. Pain relief was greater with cannabis, with a median difference of 3.3 points on the Descriptor Differential Scale and an effect size of 0.60. The proportion achieving at least 30 percent relief was 0.46 with cannabis and 0.18 with placebo. Two participants experienced treatment-limiting toxicities.

Both trials enrolled patients whose pain had already failed other analgesic classes, and both continued patients on their existing regimens. That makes these add-on trials in refractory neuropathy, which is a demanding place to show an effect and a fair reflection of how the question presents clinically.

The limits are the sample sizes, the five-day exposure windows, and the route. Neither trial tested oral or vaporized preparations, and neither followed patients long enough to say anything about durability.

Dronabinol carries a US Food and Drug Administration indication for anorexia associated with weight loss in AIDS, and the trial behind it is worth reading carefully rather than citing from memory. Beal and colleagues randomized 139 patients with AIDS-related anorexia and at least 2.3 kilograms of weight loss to dronabinol 2.5 mg twice daily or placebo, publishing in the Journal of Pain and Symptom Management in 1995. Efficacy was evaluable in 88 of them.

Appetite improved by 38 percent above baseline on dronabinol versus 8 percent on placebo, a significant difference. Nausea decreased 20 percent versus 7 percent. Weight, however, was stable on dronabinol while placebo recipients lost a mean of 0.4 kilograms, and that difference did not reach statistical significance. The proportion gaining at least 2 kilograms was 22 percent versus 10.5 percent, also not statistically significant.

So the honest description of the pivotal wasting trial is that it demonstrated an appetite and nausea effect and did not demonstrate a weight effect. Pooled analyses have been kinder. The 2015 JAMA systematic review by Whiting and colleagues rated the evidence for weight gain in HIV infection as low quality, and the 2026 JAMA review reports a moderate pooled effect of cannabinoids on body weight in HIV and AIDS, with a standardized mean difference of 0.57 (95 percent CI 0.22 to 0.92).

The larger issue is who those trials studied. AIDS wasting syndrome was a manifestation of uncontrolled infection. In a person taking effective antiretroviral therapy with a suppressed viral load, involuntary weight loss is uncommon and, when it occurs, warrants a workup for a specific cause rather than an appetite stimulant. Weight gain on modern regimens is more often the clinical problem than weight loss. Citing 1995 wasting data as though it describes today’s patient is the most common error made on this topic.

Abrams and colleagues addressed the safety question directly in a randomized placebo-controlled 21-day inpatient trial in Annals of Internal Medicine in 2003. Sixty-seven patients with HIV-1 infection received smoked cannabis at 3.95 percent THC, dronabinol 2.5 mg, or placebo three times daily. Neither smoked nor oral cannabinoids adversely affected HIV RNA levels, CD4 counts, or CD8 counts over the treatment period. The adjusted average change in log10 viral load versus placebo was -0.07 for cannabis and -0.04 for dronabinol, both compatible with no effect.

The companion pharmacokinetic analysis by Kosel and colleagues in AIDS examined indinavir and nelfinavir levels in the same participants. Nelfinavir exposure over eight hours changed by -10.2 percent, not significant. Indinavir exposure changed by -14.5 percent and maximum concentration by -14.1 percent, the latter reaching significance. The authors concluded the changes were unlikely to have short-term clinical consequence.

That conclusion is probably still right in spirit, and it rests on drugs that contemporary practice has moved past. Indinavir and nelfinavir are not first-line agents. Modern regimens are anchored on integrase strand transfer inhibitors, and comparable controlled pharmacokinetic studies of cannabis or cannabidiol with those agents have not been published. The absence of a demonstrated interaction is not the same as a demonstration of no interaction.

The practical consequence is modest but real: a patient on a contemporary regimen who uses cannabis or high-dose cannabidiol should have their viral load monitored on the normal schedule and should tell their HIV clinician, and a clinician should not present the 2002 data as though it settles the question for drugs that did not exist then.

The claim that cannabinoids slow HIV disease progression traces to a single line of primate research, and that research did not replicate. Molina and colleagues reported in AIDS Research and Human Retroviruses in 2011 that chronic THC administration decreased early mortality in male rhesus macaques infected with simian immunodeficiency virus. The same group published in 2014 that in eight female macaques the protective effect was absent: no change in mortality, viral load, or CD4 to CD8 ratio, and less weight gain than controls.

An eight-animal study that contradicts an earlier eight-animal study in the other sex is not a foundation for a human claim. Anyone stating that cannabinoids slow HIV progression is extrapolating from primate work that its own authors describe as needing explanation.

The immunosuppression concern runs in the opposite direction and is similarly unresolved in humans. Cannabinoid receptors are expressed on immune cells, and the theoretical concern is legitimate, but the 21-day human trial found no adverse effect on CD4 or CD8 counts and no increase in viral load.

Cognition is where a measurable signal exists. A 2025 study in Neuropsychology examined 269 community participants stratified by cannabis use and HIV serostatus. Group comparisons of global and domain-specific neurocognitive performance were not significantly different, but participants who were both cannabis-using and HIV-positive were roughly three times more likely to be classified with a memory impairment than controls, and recent use interacted with serostatus on global and motor performance. The authors describe the effect as modest and call for weighing it against the severity of the symptoms being treated. That is the right framing.

The neuropathic pain evidence is the strongest in this literature and would be considered respectable in any therapeutic area: two independent randomized placebo-controlled trials, different designs, different institutions, concordant responder rates, in patients already refractory to other analgesics. The weaknesses are sample size, five-day exposure, and reliance on a smoked route that most clinicians would not recommend to a patient with any pulmonary vulnerability.

The appetite evidence is weaker and older than its regulatory status suggests, and the disease-modification evidence does not exist in humans at all. The safety data are reassuring for the era in which they were collected and silent about the drugs used today.

Both neuropathy trials were small and short. Fifty and 28 completers over five-day treatment periods can establish that an analgesic effect exists; they cannot describe what happens over months, whether tolerance develops, or how the benefit compares with gabapentinoids or duloxetine in a head-to-head design.

Blinding is a persistent problem in any trial of an intoxicating substance. Participants receiving active cannabis can often tell, which inflates subjective outcomes such as pain ratings. The Abrams trial used cannabinoid-extracted cigarettes as placebo, which is better than nothing and does not solve the problem.

The appetite literature carries a further issue: the pivotal trial’s weight outcome was not statistically significant, yet the regulatory indication and three decades of secondary citation have treated the wasting story as settled.

No human study shows that cannabinoids slow HIV disease progression, reduce viral load, or improve immune reconstitution. The primate work sometimes cited for this failed to replicate in female animals from the same laboratory.

No published trial establishes that cannabis is safe to combine with the integrase inhibitor regimens that anchor modern first-line therapy, and no trial has tested cannabinoids for weight loss in virologically suppressed patients, who are a different population from the 1995 trial participants.

HIV occupies an unusual place in cannabis medicine. It is one of the few indications where cannabinoids hold an actual FDA-approved status, and that status rests on trials whose clinical context has been transformed by effective antiretroviral therapy. The result is a literature that is frequently cited with more confidence than it earns.

The neuropathy finding has aged better. HIV-associated distal sensory polyneuropathy remains common in long-term survivors, is frequently refractory to standard agents, and is one of the few conditions where cannabis has been tested directly against placebo in the relevant population rather than extrapolated from general chronic pain trials.

The most useful thing a clinician can do with this evidence is to sort it by indication instead of by disease. For refractory HIV neuropathy there is direct trial support. For appetite in an untreated or poorly controlled patient there is limited support. For anything framed as slowing the illness there is none.

When a patient with HIV asks me about cannabis, the first thing I want to know is which symptom we are talking about, because the evidence splits sharply. Painful neuropathy in the feet that has not responded to gabapentin is a reasonable place to have this conversation. Weight loss in someone whose viral load is undetectable is a reason to look for a cause, not a reason to reach for an appetite stimulant.

The wasting literature gets quoted to me constantly, and it comes from a time when people were dying of untreated infection. That is not the patient in front of me. Treating a 1995 trial as current guidance for a virologically suppressed person in 2026 is a category error.

The interaction question is the one I am least comfortable waving off. I have seen the 2002 protease inhibitor data cited as general reassurance for regimens that did not exist when it was collected. I tell patients to keep their HIV clinician informed and to keep their monitoring schedule, which is a low-cost way to handle an unstudied question.

For refractory HIV-associated neuropathic pain, two randomized placebo-controlled trials support a real analgesic effect, with roughly half of treated patients reaching a clinically meaningful response versus a fifth to a quarter on placebo. For appetite, the evidence is limited and describes a pre-treatment era population. For disease progression, there is no human evidence. For antiretroviral interactions, the available reassurance concerns drugs that are no longer first-line, so disclosure and routine viral load monitoring remain the sensible approach.

Read this literature by indication and by date. The pain trials are directly relevant and still hold. The appetite trials describe a disease course that effective treatment has largely eliminated, and the safety trials describe a pharmacology that has moved on. Nothing here supports a claim that cannabinoids affect the course of HIV infection itself.