The National Academies of Science, Engineering, and Medicine is a private, nonprofit NGO established more than 150 years ago to advise the nation on scientific matters. It’s considered one of the gold standard institutions of science. In 1999, largely in reaction to California’s legalization of medical cannabis, the Institute of Medicine (the medical research arm of the National Academies) was tasked by the White House Office of National Drug Control Policy (ONDCP) to conduct a systematic review of the scientific evidence pertinent to the health risks and benefits of cannabis and cannabinoids. To the White House’s surprise, the institute came back with a report that cautiously supported the idea that cannabis could have beneficial medicinal effects.
Almost two decades later, the National Academies put together this follow-up report to see what the science of the past 18 years has further revealed about cannabis and medicine.
The committee reached nearly 100 research conclusions based on consideration of more than 10,000 research articles. They gave more weight to articles published since 1999 report. From this information, each specific research conclusion was assigned to one of five “levels of evidence”: conclusive, substantial, moderate, limited, and no/insufficient evidence. Importantly, the committee focused exclusively on the human literature, and did not consider basic research conducted using animal models.
You can read the full report, highlights, and public release slides here.
The committee found three medical applications for cannabis use supported by conclusive evidence (as opposed to substantial, moderate, limited, or insufficient evidence):
There’s a lot more to dive into (see below), but let’s first consider some important caveats to the study’s conclusions.
One thing the report emphasizes is how much we don’t know—and why we don’t know it. It provides us with four conclusions about research barriers:
This is a key take away point. If there’s one thing that everyone can agree on when it comes to cannabis, it should be that we need to do the research necessary to inform ourselves about its health effects. But this is very difficult to do in practice because of cannabis’ absurd designation as a Schedule I controlled substance. Because this report identifies multiple medical applications supported by conclusive evidence, it directly contradicts the Schedule I designation the federal government places on cannabis.
There are three big caveats I think we should keep in mind when reading this report:
Caveat 1: When reviewing human studies, members of the committee looked for statistical associations between cannabis use and health outcomes, but they did not attempt to evaluate whether significant associations were due to cannabis use causing a specific health outcome or whether cannabis use and that outcome were associated for some other reason, such as a common underlying cause. This is a key weakness when we only consider human studies involving a Schedule I substance: the findings are usually correlational and thus prevent us from drawing conclusions about cause-and-effect.
Caveat 2: Many human studies rely on self-reporting of cannabis usage. This is a huge caveat for many human studies, as any conclusions drawn about the effects of light, moderate, or heavy cannabis use rest on the assumption that subjects are accurately reporting their consumption.
Caveat 3: The committee explicitly decided to not consider basic research studies. That helped simplify the gargantuan task of evaluating tens of thousands of research abstracts. There are many thousands of basic research studies out there (both test tube and animal studies), and considering these would have been incredibly time- and labor-intensive. While that basic research isn’t conducted in humans, it does allow us to dig deeper into mechanisms of action and establish cause-and-effect relationships. There’s a lot of interesting and compelling basic research out there that should inform the direction of human clinical research. This represents a huge knowledge gap in the report.
Below, I’ll walk us through the chapter highlights of the 11 chapters of the report. The basic conclusions reached by the committee are listed as bullet points, with my own commentary below them, including what we should take away and any important caveats we should keep in mind. Each heading below corresponds to one chapter of the report focused on a specific health concern.
Not much new here. These effects have been widely known for some time. Again, these are the areas supported by evidence deemed “conclusive” by the committee. There are many more things for which “substantial,” “moderate,” or less convincing levels of evidence exists. For me, the second bullet point is arguably the most important, as it could have huge implications for the country’s ongoing opioid epidemic.
Not a whole lot to add here. To be safe, people with cardiovascular issues should be extremely careful, as THC’s action through CB1 receptors in the brain can cause an acute (temporary) increase in pulse and blood pressure.
Not much new here, either. The weight of the evidence to date suggests that smoking cannabis, unlike smoking cigarettes, does not increase lung cancer risk. Apparently there is modest evidence for an association between cannabis use and a specific form of testicular cancer. The important caveat for these conclusions is that they’re based on finding (or failing to find) statistical associations between cannabis use and a specific cancer outcome. Controlling for confounding variables, such as tobacco smoking, is critical for interpreting the results.
The association between cannabis and one subtype of testicular cancer is considered limited because the studies finding a link suffered from one or more of the following: they relied on self-reported data, response rates were low, or potential confounding variables were not controlled for.
Again, not much new here. If you smoke cannabis all the time, there’s a good chance you’ll be coughing up some phlegm. We recommend regular smokers consider vaping instead.
This is a section where the authors concluded that limited or insufficient evidence exists across the board. But it’s also an area where a lot of basic research has been done that should be used to guide human clinical research. For example, we know that THC is a more potent anti-inflammatory that aspirin and hydrocortisone, and we know the endocannabinoid system has an important role in regulating the immune system’s inflammatory response. This makes it plausible that there are useful anti-inflammatory applications of cannabis-based therapies, and we should be pursuing human studies to investigate what these might be.
This one is pretty straightforward. Please don’t smoke anything if you’re pregnant. It’s the only safe and reasonable strategy to take. The report contradicts a 2016 review that Leafly covered, which found that cannabis use was not linked to negative birth weight of preterm delivery outcomes (at least when used without tobacco or other illicit substances). However, the current report points out that there’s reason to think that non-cannabinoid byproducts of combustion that are found in smoke (including carbon monoxide), can impair fetal growth. The only reasonable conclusion here is that pregnant woman should avoid all forms of smoke inhalation.
“Problem cannabis use” here means cannabis use disorder. While cannabis does not have nearly as much habit-forming potential as substances like alcohol, nicotine, or opioids, it is possible to develop a habit. What’s a habit? If you have trouble voluntarily taking a break, and especially if doing so gives you withdrawal symptoms, then you’ve got a habit. Using cannabis very frequently, especially if you start at a young age, increases the odds of habit formation. This is important but also nothing new.
This one jumped out at me. If you look at this chapter of the report in detail, things get confusing. The authors reach these three conclusions in the chapter of the full report:
I found this trio of conclusions confusing. The first two conclusions are that there’s limited evidence for an association between cannabis and both initiation of tobacco use and changes in the rate or pattern of use of other illicit substances. But then we’re told that there’s moderate evidence for an association between cannabis use and the development of dependence for other substances. Wouldn’t the development of dependence be considered a change in the pattern of use? And if the chapter summary (the bullet point above) is based on these three conclusions, what justifies the definitive-sounding statement, “Cannabis use is likely to…”?
It isn’t clear me why they separated the studies that were considered for points (2) and (3) above. Point (2) is about changes in patterns of usage, and point (3) is about the development of dependence, which is itself a change in the pattern of usage.
While some of the studies considered had large sample sizes and controlled a variety of confounds, most or all seem to have relied on self-reported data around consumption of cannabis and other substances. These types of studies also don’t include any biological data that might tell us whether someone is generally predisposed to developing a substance use disorder.
If lifetime use of one substance is associated with increased use of another, how do we know that using the first substance was what increased the risk of using the second? How do we rule out the existence of a biological predisposition that makes one more likely to develop a dependence on any intoxicating substance? We can’t, at least not from these types of studies. Thus, I find the conclusion given in the chapter highlights, that cannabis “is likely to increase the risk for developing substance dependence,” to be specious and inappropriately phrased given the content of this chapter.
Without diving into the report in detail I think we can come up with some common-sense conclusions about the first two points: You shouldn’t operate a motor vehicle while under the influence of any psychoactive substance, and you need to be extremely careful about storing your cannabis products if there are children around. That latter point is especially important for cannabis edibles, which can allow unsuspecting individuals to mistakenly consume uncomfortably large amounts of THC.
The last point should come as no surprise, since we know why you can’t die from cannabis consumption. People in states with legal medical cannabis also don’t appear to be getting stoned before heading into work. In fact, we’re seeing lower rates of workplace absenteeism in states with legal medical cannabis.
The first two points are straightforward. The acute effects of THC intoxication involve impairments in cognition (e.g. short-term memory, attention), and there is limited evidence that such impairments can persist after people stop consuming cannabis. The last point about academic impairments associated with adolescent use is true based on the studies considered in this report, but we also highlighted a 2016 study too recent to be considered in this report that didn’t find this type of association (but only after controlling for confounding variables, namely tobacco use). Because childhood and adolescence are critical periods of nervous system development, the use of cannabis or any psychoactive substance should be avoided before adulthood.
OK, the first two points beg for a double-take. Cannabis use can increase the risk of schizophrenia, but those with both schizophrenia and a history of cannabis use show better performance on learning and memory tasks? What? Box 12-1, titled “Co-Morbidity in Substance Abuse and Mental Illness,” helps us start to digest this. Here are its three main points:
The report follows with a statement about why the relationship between mental health and substance abuse is so difficult. Substance abuse can affect mental health, mental health can affect substance abuse, and other variables can affect both. In their words:
Although the precise explanation is still unclear, it is reasonable to assume that co-morbidity between substance abuse and mental health disorders may occur due to a mixture of proposed scenarios. With this context in mind, however, it is important to note that the issue of co-morbidity directly affects the ability to determine causality and/or directionality in associations between substance use and mental health outcomes. This is a complex issue, one that certainly warrants further investigation.
Again we see the emphasis on more research. The nature of the link between cannabis use and schizophrenia was debated by scientists in the journal Nature in 2015 (look here and here for opposing views from scientists).
It will take some time to dissect this 400-page report in more detail. This report looks at a lot of human health issues and how they may potentially be affected by cannabis use. While I don’t envy anyone tasked with such an enormous undertaking, I was somewhat disappointed to see that the report didn’t consider any basic research findings and instead relied on only human studies. While this allowed the committee to focus on studies directly related to human health, a large proportion of those studies are based on self-reported data that are correlational in nature. Occasionally, some of the language used to summarize their conclusions doesn’t adequately capture these important caveats.
A major emphasis of the report is that we need much more research. Unfortunately, doing the types of well controlled, large-scale clinical studies that we need to be doing is very difficult in the United States today. Given that this report, conducted by a cautious set of researchers, finds conclusive evidence that cannabis has legitimate medical applications, the federal government’s classification of cannabis as a Schedule I Controlled Substance, with “no currently accepted medical use,” must be considered untenable and inappropriate.