A recent study challenges the role that legalizing medical marijuana might play in easing the opioid epidemic.
The paper, published in the Proceedings of the National Academy of Sciences in June 2019, indicates that a previously reported relationship between medical marijuana laws and declining opioid overdose deaths has not held up over time.
The paper uses the same research methods as a 2014 study published in JAMA Internal Medicine, which found that states that had legalized medical marijuana during the study period — 1999 to 2010 — had 24.8% lower annual opioid death rates, on average, compared with states that had not legalized medical marijuana.
The authors of the 2014 study found this association by examining opioid overdose death rates by state using data from the U.S. Centers for Disease Control and Prevention. The data included all overdose deaths in which an opioid was involved, including those that involved other drugs such as heroin. The authors used regression analysis to estimate the relationship between states enacting medical cannabis laws and opioid overdose death rates over time.
This new study replicated the original paper’s findings and extended the investigation through 2017, adding 32 states that legalized medical marijuana between 2010 and 2017 to the sample. When the authors examined the relationship over this longer period, they found that the relationship did not hold. In fact, it reversed direction — opioid overdose death rates increased by 22.7%, on average, in states that had legalized medical marijuana.
“To me, as a public health researcher working in drug policy and epidemiology, this idea that medical cannabis is a very important solution to the opioid overdose crisis is really salient,” explains Chelsea Shover, a postdoctoral research fellow in psychiatry at Stanford University who’s the lead author of the 2019 paper. “It’s just a pervasive idea, and it’s one that I, in the past, had found really compelling, even knowing the limitations that the authors of the first paper mentioned.”
The authors of the original study listed the following as limitations:
- The study is ecologic — that is, it’s concerned with trends at the population-level, such as at the level of the state or the country. This means that it does not and cannot account for what is happening at the level of the individual.
- The death certificate data used by the researchers could be inaccurate. Also, states may differ in how they classify opioid overdose deaths.
- Other factors that could explain the association might not have been accounted for in the model.
- The relationship between medical marijuana laws and opioid overdose death rates may change over time.
Despite this, the original article made a big — and lasting — splash. “That research received substantial attention in the scientiﬁc literature and popular press and served as a talking point for the cannabis industry and its advocates, despite caveats from the authors and others to exercise caution when using ecological correlations to draw causal, individual-level conclusions,” the authors of the follow-up study write.
In fact, the original article has been cited in over 400 scientific journal articles and 370 news articles.
Its widespread uptake stems from a few factors, Shover suggests. “There are a lot of people who really want this to be true,” she says. “It’s compelling because it just sort of makes sense on its face and it’s easy for people to understand.”
She maintains that there’s a problem with the way the research has been interpreted by academics and the media. Both studies are concerned with ecological-level associations — that is, population-level trends. These bigger picture associations don’t provide information about what’s happening from person to person. For example, the studies don’t shed light on whether, at the individual-level, people are (or aren’t) substituting medical marijuana for opioids.
Shover and her co-authors write in their paper that the relationship the original study found is likely “spurious.”
Chinazo Cunningham, an author on the 2014 paper, disagrees with using the term “spurious” to describe the association between medical marijuana legalization and decreases in opioid overdose mortality.
“To me, the findings are consistent,” Cunningham says. “The opioid epidemic has evolved. And what it was when we were looking at it, is now different,” she explains. Initially, people were overdosing on prescription opioids. In more recent years, they have died after using heroin and synthetic opioids like fentanyl.
“What we know is that medical cannabis helps with pain,” Cunningham continues. “And the prescription opioids were really around pain and pain management. And so as the epidemic has evolved, it’s become more around addiction. And there really are not data to support the use of medical cannabis to treat opioid addiction. I don’t expect that medical cannabis would treat opioid addiction. Ever. So that is really what I think the [original] findings are saying — medical cannabis helps with pain management, and there can be a reduction in opioid use, but it’s not going to treat necessarily addiction.”
However, Cunningham agrees that both studies are limited in that they cannot establish causality: “These studies also have limitations, definitely,” she adds.
Shover maintains that her work is agnostic on the role of marijuana in treating pain. “What does this study mean for how patients and families should treat their pain?” Shover asks. “The answer to that is nothing. It’s a study about policy; it explicitly is talking about population-level effects and not about the individual-level decisions.”
For now, it’s hard to study the individual-level due to federal marijuana policy that prevents national data collection and analysis, Cunningham points out. “The other really important thing here is, why are we talking about this? It’s about the data and the lack of causality. And that is because we can’t study cannabis in the way we want to study it,” she says. “If we could just change our federal policy to match more with what the states are doing, then we could actually do the studies and have the data that can inform this conversation in a much better way.”
But researchers might be able to get closer to understanding the crux of the issue through different kinds of studies, Shover says. To suggest how to understand the individual-level relationship between opioids and medical marijuana, Shover offers the hypothetical example of a study that looks at state-level medical marijuana registry data and corresponding medical records to see if people who had, prior to registering for medical marijuana, been prescribed opioids, were subsequently prescribed fewer opioids after joining the registry in comparison with people who received opioids for similar diagnoses but who hadn’t registered for medical marijuana. For now, Shover says it would be challenging to execute such a study due to issues with accessing the records involved.
Shover continues, “The relationship between two things can be really different on individual levels than on a population level, and so our work is not aimed at patients for making decisions about how to treat their pain. It’s not aimed at voters who are deciding whether to support decriminalization or legalization. It’s aimed at policy makers who are saying, ‘I could either spend my time trying to push for medical cannabis and then feel like I’ve really done something to address the overdose crisis, or I can say okay, medical cannabis is a separate issue, when I’m thinking of strategies about the overdose crisis I should be focusing on stuff we have better evidence for — like Narcan, like making it easier for people to get treatment and reforming incarceration policies to reduce vulnerability to overdose.’”
Lessons for journalists
This tale of two studies is a great example of why it’s important for journalists to stay close to the source material they’re covering — paying careful attention to the question the study is addressing, as well as the limitations of the research — and not extrapolate beyond the researchers’ findings.
Part of this has to do with framing the research accurately from the outset.
“Headlines help you stay in business, right?” Shover asks. “Well, whatever you go into in the actual article is really important. But it is undercut pretty easily by a headline that’s saying something that’s stronger than what the actual study would say.”
For example, these headlines overstate the 2014 study’s findings:
- “Marijuana Legalization Reduces Opioid Use, Studies Show”
- “Studies: Medical Marijuana Helps Ease Painkiller Addiction Crisis”
- “Despite the Skeptics, Legal Marijuana is Having a Positive Impact on the Opioid Crisis.”
Shover says some of the onus is on researchers, who must communicate their findings in a way that’s both accurate and understandable to the lay public.
Cunningham agrees. “I think that often, as researchers, we’re very specific in the ways that we talk about our findings, and not overselling it,” she says “But I think that gets lost by the media often because the media wants headlines.” But it’s not just on the media: “I do think there’s a dual responsibility of the researchers and the journalists,” she adds. “We have to discuss it in a way where the general public can understand the nuance. And I think that’s really a lot of where the challenge lies.”
From her own experience, Shover advises researchers to say yes to every interview. “Getting out in front of it, and being very forthcoming with reporters and being willing to talk to media has been helpful in that the content of the articles I see out there really do seem to capture what we found and to do a good job of communicating to what it does and doesn’t say,” she says.
Part of it, she says, has to do with putting a study into its proper context — as a contribution to a larger body of research, rather than the last word on a topic.
Shover suggests journalists bring in broader research perspectives beyond just the authors of the study on which they’re reporting. “Bring in basic scientists who are doing research on how it [cannabis] affects the brain, and then people who are doing clinical research, and just talking about it from angles and different levels like whenever possible,” she says. “Trying to get those perspectives is really helpful to understanding what you can and can’t learn from an individual study.”