Medical use of marijuana revisited

I described two experimental studies of the use of marijuana to reduce pain in Tools for Critical Thinking in Biology and discussed two technical reviews of the medical value of marijuana in a subsequent blog posting. The gist of these reviews was that there was limited but adequate evidence that marijuana could help patients suffering from chronic pain, but little evidence for other beneficial health effects. Additional research reinforces the evidence supporting use of marijuana for pain control, especially as an alternative to opioid drugs, which are widely abused in the US. This research has an important flaw, however, that was ignored in recent news stories.

In 2014, Michael Bachhuber and colleagues at the Philadelphia Veterans Affairs Medical Center compared death rates caused by overdoses of opioid painkillers such as oxycontin in US states with and without laws allowing medical use of marijuana. They used data through 2010 when 13 states had passed such legislation. Bachhuber’s group found that annual death rate due to opioid overdoses was 25% less in these states than in the remaining 37 states that didn’t allow medical use of marijuana.

In a follow-up study published in July 2016, Ashley Bradford and her father David, graduate student and professor respectively at the University of Georgia, compared Medicare data on prescription drug use in states with and without medical marijuana legislation. On average, individual doctors wrote almost 2000 fewer prescriptions for opioids and other typical painkillers in states in which medical marijuana was legal.

Bachhuber’s group and the Bradfords infer from their comparisons of states with and without legalization of medical marijuana that individuals might benefit by having marijuana available to control pain. For example, in a news story in Science magazine, Greg Miller quotes David Bradford describing his results as “suggestive evidence that medical marijuana might help divert people away from the path where they would start using [an opioid drug], and of course if they don’t start, they’re not on that path to misuse and potentially death”.

These analyses illustrate a classic breakdown of logic called the ecological fallacy: reaching conclusions about individuals from aggregated data for groups to which the individuals belong. The groups are states where individuals live; in Bachhuber’s study the data are death rates in each state due to opioid overdoses, and in the Bradfords’ study the data are numbers of prescriptions for opioid painkillers written by doctors in each state. Average death rates and opioid prescription rates are lower in US states with legal use of medical marijuana than in the remaining states. Paradoxically, individuals who use medical marijuana might also use more opioids or have a greater likelihood of dying than individuals who don’t use medical marijuana, despite the opposite effect in comparing average rates of opioid prescription and death between states with and without laws for medical use of marijuana.

Bachhuber’s group was fairly cautious in interpreting their results, although news reports such as “Could pot help solve the U.S. opioid epidemic?” in the November 4, 2016, issue of Science ignored the caveats discussed by Bachhuber and colleagues. For example, Bachhuber’s group mentioned the alternative hypotheses that “increased access to medical cannabis may reduce opioid analgesic use by patients with chronic pain” or act as a “’gateway’ or ‘stepping stone’ leading to further . . . opioid analgesic overdoses”. However, the researchers weren’t able to discriminate between these hypotheses because they only used statewide averages, not data for individual patients.

How could it be possible for medical marijuana to be a stepping stone for individuals to overdose on opioids and at the same time for states without legal access to medical marijuana to have more deaths from opioid overdoses? One reason is that other differences between states with and without laws allowing medical use of marijuana might account for greater death rates from opioid overdoses in the latter states. These other potential differences are called confounding variables. Bachhuber’s team considered unemployment rate as one possible confounding variable, on grounds that higher unemployment might cause more drug abuse, but they found no differences in unemployment rates between states with and without medical marijuana. Unemployment rate is only one of many possible confounding variables, however; states without legal access to medical marijuana might have older residents, with more pain, or younger residents, with more propensity to abuse drugs, or lower educational levels, or more military veterans suffering from PTSD, than states with legal access to medical marijuana. Indeed, one of the most dramatic differences between states that do and don’t allow medical use of marijuana is that all of Hillary Clinton’s 232 electoral votes in the 2016 presidential election came from states with laws allowing medical marijuana while only 23 of Donald Trump’s electoral votes came from such states. This certainly doesn’t mean that having medical marijuana available caused people to vote for Clinton and not having it available caused them to vote for Trump, but it illustrates the pitfalls of drawing conclusions about causation from purely correlational data. I illustrate this point in more detail in Chapter 5 of Tools for Critical Thinking in Biology.

Map of US states with and without legalization of medical marijuana.

US states with operational legalization of medical marijuana as of December 28, 2016. This excludes Arkansas, Florida, and North Dakota that legalized medical marijuana in November 2016 but haven’t yet established a process for patients to obtain marijuana. It also excludes Texas that legalized medical marijuana in 2015 with a doctor’s prescription, ignoring the fact that federal law forbids doctors from prescribing marijuana. Instead, doctors may issue a letter to the patient recommending marijuana for certain medical conditions, as occurs in the states where legalization is operational.

To test the alternative hypotheses proposed by Bachhuber’s group that medical marijuana may be a safer alternative to opioid drugs for treating pain or that medical marijuana may be a stepping stone to opioid abuse, we really need data on individual people who do and don’t have access to medical marijuana. Even better would be experimental data, although it’s hard to imagine an ethical study that could test these hypotheses. However, experiments are possible to test the efficacy of marijuana compared to opioids for pain control. For example, Colorado recently legalized marijuana for recreational use as well as medical use, and the state is using $9 million of the taxes collected from sales of marijuana to support experimental studies, much as California did in the early days of marijuana research (see Chapter 4 of Tools for Critical Thinking in Biology). Emily Lindley is in charge of one of the Colorado studies; her group will use 50 volunteer patients with back or neck pain who will be assigned randomly to receive either vaporized marijuana, an opioid drug, or a placebo on three visits to the hospital, after which they will report any pain relief they experience.

So, does legalization of marijuana for pain control reduce the likelihood of patients overdosing on opioids? We don’t know the answer at this time, so we shouldn’t use this possibility as an argument in support of legalization of marijuana. Uncertainties like this can be frustrating to those interested in promoting or opposing particular policies, but such uncertainties are embraced by scientists as challenges to be overcome. Be on the lookout for further research that may help answer this question!

This entry was posted in Causation, Correlation & Causation, Evaluating evidence, Experimentation, Medicine. Bookmark the permalink.