This is part 2 my series where I describe pet peeves of mine that make me distrust a piece of media. Part 1 can be found here. There are some patterns in media that I can’t help but see as red flags. These patterns trigger my instinct to distrust what I’m reading.
Pet Peeve #2: The Number Zero
You might have see the following infographic a few years ago, or something like it:
This infographic lists the annual deaths associated with multiple drugs. Tobacco is at the very top with about 400,000 deaths annually, and Marijuana is at the bottom with 0 annual deaths. It even lists deaths from what we consider to be safe drugs, including 2000 deaths from caffeine, and 500 deaths from aspirin. At the very bottom of the list is Marijuana, which is listed at 0 deaths.
You might have heard that you can't die from an overdose on THC, which the active ingredient in marijuana. This is essentially true. There have been some rare reports of death from apparent THC overdose, but whether or not THC is to blame is almost always contested. One such example occurred when a coroner couldn't find any explanation other than THC for a woman's death.
St. John the Baptist Parish Coroner Christy Montegut said last week that toxicology results for a 39-year-old LaPlace woman who died in February showed that she was killed by an excess amount of THC, the main active ingredient in marijuana.
“It looked like it was all THC because her autopsy showed no physical disease or afflictions that were the cause of death. There was nothing else identified in the toxicology — no other drugs, no alcohol,” Montegut said. “There was nothing else.”
The woman's name was not released.
Montegut, who has served as the St. John coroner since 1988, believes this could be an index case in medicine, perhaps the first death on record solely as a result of THC exposure.
Some drug researchers and experts are skeptical.
Keith Humphreys, a former senior policy adviser at the White House Office of National Drug Control Policy, said that with the vast amounts of marijuana consumed in the U.S. every year, it's hard to imagine that more overdose deaths wouldn't be occurring if THC was toxic at consumable levels.
“We know from really good survey data that Americans use cannabis products billions of times a year, collectively. Not millions of times, but billions of times a year,” said Humphreys. “So, that means that if the risk of death was one in a million, we would have a couple thousand cannabis overdose deaths a year.”
Humphreys also said it's not uncommon for coroners to see a drug in the system, with no other sign for what might have caused an event leading to death, and so conclude that the drug was the cause.
So while it’s a bit contested, it is at least fair to say that there are 0 deaths from marijuana overdose. However, this infographic is about more than just overdose deaths. Most of the 400,000 deaths from tobacco aren’t from literal nicotine overdose. Most deaths from tobacco come from breathing in tar and other chemicals, which lead to long-term health problems, which reduce lifespan. Marijuana smoke has many of the same chemicals as cigarette smoke, and it also has more tar. Most of the 100,000 alcohol deaths also are not overdose. Only about 2200 of those deaths are from overdose. The rest are from long-term health effects and from accidents. I’m not sure where the 2000 caffeine deaths come from. As far as I can tell, there are only 92 deaths from Caffeine, ever. If anyone knows where that 2000 figure might come from, leave a comment down below. I’d love to see it.
When I see this infographic, I get a strong sense that the creator took the fact that you can’t die from marijuana overdose, and assumed that the annual deaths from marijuana can't possibly be anything except 0. It makes me think that they didn’t bother to check, especially not for deaths related to the tar and chemicals in marijuana smoke.
This is what I usually think when I see the number zero where a zero is not obvious. I assume by default that the creator of the work didn’t actually check the figure, and just assumed that it must be zero. It’s especially suspicious when there’s a reason to expect a non-zero figure to be there. In this case you might expect at least a few deaths due to the lung problems that marijuana causes.
It’s even more suspicious when the creator is clearly biased in favor of listing a low number like 0. We can tell that the creator of this infographic is biased because there’s a marijuana plant in the background, and because the line for marijuana is listed, in a bigger, differently colored font. It is explicitly a pro-marijuana piece.
Now I’ve been in favor of marijuana legalization for a long time. I’ve believed in legalization ever since I was 14. However, I’m not a fan of this infographic. The figures it contains make me question whether the author did their homework. It doesn’t pass the sniff test, and because of that I can’t rely on it.
I figured I might as well check to see if my intuition was right. I was unable to find an organization that actually tracks annual deaths due to marijuana, but I did find some literature about it. To the infographic’s credit, there is some literature that says that studies usually fail to find an impact from marijuana on all-cause mortality. However, even that literature doesn’t conclude that marijuana can’t cause death. Instead, they say that there is a lack of research on the topic.
From The Health Effects of Cannabis and Cannabinoids:
There is an overall dearth of cohort studies empirically assessing general population cannabis use and all-cause mortality. Although the available evidence suggests that cannabis use is not associated with an increased risk of all-cause mortality, the limited nature of that evidence makes it impossible to have confidence in these findings. These conclusions are not informed by the results of existing large-scale modeling studies that synthesized data from a variety of sources to estimate the burden of disease attributable to cannabis use (Degenhardt et al., 2013; Imtiaz et al., 2016). Although these studies were methodologically rigorous, their direct applicability to actual cannabis-related mortality rates in the United States is uncertain. Consequently, the committee chose not to include them in this review. Also excluded from review were studies of mortality among persons with known cannabis addiction or dependence, those who have been under medical treatment for these disorders, or those who were identified through a country's criminal justice system, due to presence in these populations of important and often inadequately controlled confounders such as concurrent mental illness and poly-substance abuse.
CONCLUSION 9-1 There is insufficient evidence to support or refute a statistical association between self-reported cannabis use and all-cause mortality.
So while they have something to point to in order to kinda sorta justify their “0” figure, it’s very weak, and the 0 figure still seems super suspect to me. I would bet that if they ever do track marijuana-related deaths like they do for tobacco, then they will find more deaths from marijuana smoke than was listed on the infographic for either caffeine or aspirin.
With this sort of media, when I see a low number other than 0, I’m usually less suspicious. Seeing a low non-zero number gives me a sense something was actually checked. If the figure actually is zero, then they might need to do a bit extra to convince me. Simply saying “yes, we checked” helps a little bit. At least it lets me know that they knew they were supposed to check. It’s even better when I can see what they checked, and where I can find it. At least that would let me know that they’re trying to be accurate.
Jump in the discussion.
No email address required.
Notes -
This is a good post!
I feel the same way pretty much any time statistics are used as the main argument. I can't think of many times I've actually dug into the numbers of a (popular) study and it wasn't garbage.
Often what I'll find is that the numbers look fine but the modeling is garbage. Either they didn't control for an obvious confounder, or they did control for an obvious mediator ... or at best they did try to control for the right thing but they couldn't do it well enough so they just published a wildly overstated conclusion anyway. "Past parachute use is associated with increased risk of future skydiving death, even when controlling for present altitude!"
And yeah, the selection bias here hurts. More surprising conclusions are less likely to be reached justifiably (surprising->surprisal even makes that a tautology!), yet less surprising conclusions aren't news (again...), so the news ought to include unjustified conclusions disproportionately.
More options
Context Copy link
More options
Context Copy link