A widely reported new study purports to show that the pain-relieving properties of cannabis are primarily a result of the placebo effect. Its large statistical limitations, however, bring its findings into serious doubt.

The researchers at the Karolinska Institutet in Sweden conducted a meta-analysis of 20 different studies covering nearly 1,500 individuals in which a placebo was compared against a cannabinoid, generally THC and/or CBD. Changes in self-reported pain intensity before and after placebo or cannabinoid administration were compared across studies, first within and then between groups.

The authors concluded that placebo use alone was associated with a “moderate to large” effect in pain reduction and further that the “difference for active drug and placebo was not statistically significant”.

They also evaluated media reaction to these studies, which they found was “considerable” and which they believe may reinforce expectations of cannabis as a pain-reliever and so enhance its ability to function as a placebo.

However, among the hundreds of studies looking at the relationship between cannabinoids and pain, the ones included here were only those that reported changes in self-reported pain intensity for certain conditions in certain populations. This meant that 807 other studies were excluded raising questions about how representative of the field the 20 chosen studies really are. While overall the authors found a strong placebo effect, specific studies might tell a completely different story.


Two brief studies unbalance the findings


For several of the studies, the placebo effect was much less pronounced than the composite score would suggest. Indeed, just two studies appear to drive up the magnitude of the overall effect, and notably had study durations of just six and eight hours.

Studies with short durations are known to be much more prone to a placebo priming effect, as participants’ appreciation of treatment effectiveness is expected to diminish over time. The real finding here, therefore, might be that cannabis studies with short treatment duration are prone to strong placebo effects.

Subscribe to our Newsletter

Join in to hear about news, events, and podcasts in the sector

    See more

    By far the most technical, but consequential issue has to do with the authors’ calculation of the difference between placebo and cannabinoid use. The statistics behind randomised trials are powerful because individuals are assigned at random to either treatment or control arms, with outcomes then compared. This allows for the determination of causation as participants are essentially equally likely to be in either group: there is no reason other than the treatment for why the two groups differ.

    In this study the authors pool the effects across the placebo arm, then pool effects by treatment arms, and then compare the difference in the two composite estimates, an approach in meta-analysis known as subgroup analysis. This method ends up unpairing the comparisons that occur between randomised control and test groups in each study. While the inputs are all blinded randomised trials, the output is not. At best the results can be said to be observational, as statistically one cannot rule out the likelihood that the control groups for the placebo are inherently different from those for cannabinoids.


    Media exaggeration is a problem indeed


    Ultimately, by capturing a small sliver of research and then analysing the results in a problematic manner, the authors have not demonstrated that cannabis lacks therapeutic value in pain relief. Even if the statistical analysis had been better performed, it should be stated that placebo effects are common in pain studies.

    Pain is a complex phenomenon, embodying psychological components that are subjective and difficult to quantify or compare. The current standard for treatment for pain, opioids, have been observed to be only marginally better than placebos.

    The paper does have some merit regarding the point that media attention around cannabis may play some role in priming a placebo response. This is a fair warning against sensationalist and false reporting, but must be weighed against the value of disseminating evolving scientific research in a space of immense interest and therapeutic importance.

    Ironically, media coverage of this study has vastly overstated the extent and significance of its findings showing that indeed media restraint in this domain remains a challenge for both sides of the debate.

    Clayton Hale CannIntelligence contributing writer

    Photo: Gerd Altmann

    Author default picture


    This article was written by one of CannIntelligence’s international correspondents. We currently employ more than 40 reporters around the world to cover individual cannabis and cannabinoid markets. For a full list, please see our Who We Are page.