How to interpret cannabis research and separate science from speculation

The volume of research on cannabis has never been higher, but caution is needed: not all studies are equal. Weekly, new claims emerge about cannabinoids relieving pain, aiding sleep, or reducing anxiety. However, the quality of these findings varies drastically.

Recent evidence confirms that this pattern extends beyond the academic sphere. A 2021 cross-sectional analysis, published in the Journal of General Internal Medicine, examined over 100 online articles with high engagement.

The study found that over 80% of claims about the plant's health benefits were not supported by clinical evidence. Only 4.9% were considered true and 8.6% partially true.

The most common unsubstantiated claims involved pain, anxiety, and cancer treatment. These are precisely the same areas most promoted in both marketing and media coverage (Lau et al., 2021).

The gap between science and media

Studies analyzing how medical cannabis is reported show that the line between science and speculation often becomes blurred. An analysis of discourse in Swedish newspapers found that journalists frequently recontextualized initial or anecdotal findings as "solid science."

At the same time, they gave equal weight to patient testimonials and commercial advocacy (Abalo, 2021). This reinforces uncertainty about what the evidence actually demonstrates.

This communication gap fuels misinformation. Weak or preliminary evidence is amplified as certainties by the media and online comments.

For clinicians, investors, and policymakers, knowing how to interpret a scientific article is a professional necessity. Poorly designed cannabis research can lead to misguided policies, while well-conducted studies underpin clinical practice and responsible investments.

The hierarchy of evidence

Before analyzing the methods, identify the type of study. The literature ranges from single case reports to meta-analyses of randomized clinical trials (RCTs).

At the top of the pyramid are systematic reviews and meta-analyses, which gather data from multiple trials to assess the consistency of evidence. Below them are RCTs, the gold standard for testing if a formulation causes a measurable effect.

Further down are cohort and case-control studies, which reveal associations but do not confirm causality. At the base are case reports and anecdotal observations, useful for generating hypotheses but fragile for informing prescriptions or regulations.

This hierarchy does not suggest that every question requires an RCT. As statisticians often note, we do not need a randomized trial to know that parachutes prevent death when jumping from a plane.

However, problems arise when evidence from the base of the pyramid is presented as if it were at the top. Most viral claims stem from these lower levels of evidence.

Methods: where good science begins

The methods section is the backbone of any article. In cannabis research, where regulatory barriers and small samples are common, inadequate methods can lead to misleading findings.

Validity questions whether a study measures what it claims to measure, while reproducibility questions if it would produce the same result under the same conditions. In the sector, low construct validity is common.

Using an individual's self-report of "well-being improvement" as an indicator of pharmacological efficacy says little about the mechanism of the effect. High-quality studies employ validated tools and precise laboratory measures.

Sample size and statistical power

The sample size determines if findings reflect real effects or random variation. Many clinical trials with cannabis recruit fewer than 30 participants but draw broad conclusions.

These studies often fall into two traps:

- Type I error (false positive): finding an effect that is not real.

- Type II error (false negative): not detecting a real effect due to lack of statistical power.

Without adequate statistical power, even well-designed studies risk producing results that cannot be replicated. Replication, not novelty, is the hallmark of reliable science.

Interpreting results and biases

Numbers can be impressive, but without context, they lead to error. The p-value indicates the probability of data, but does not prove that a treatment works.

It is crucial to differentiate statistical significance from clinical relevance. A "significant" reduction in pain score may be numerically real but too small for the patient to perceive.

Additionally, biases must be considered. In cannabis research, selection bias (participants who already believe in the plant) and confirmation bias are common.

Funding also plays a central role. Industry-sponsored studies are not inherently bad but require full transparency about conflicts of interest.

As cannabis science matures, the focus should shift from producing more studies to producing better studies. Good science relies on cumulative verification, not just headlines.

With information from Businessofcannabis