site banner

How to Make Me Instinctively Distrust You Part 2: The Number Zero

samschoenberg.substack.com

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:

https://substackcdn.com/image/fetch/w_1456,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F3ec2a783-c094-4923-918f-d16b6413f857_500x500.png

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.

https://samschoenberg.substack.com/publish/post/135983348

18
Jump in the discussion.

No email address required.

Something like "there is insufficient evidence to support or refute a statistical association between self-reported cannabis use and all-cause mortality" translated to human language means "no data points towards using of cannabis killing you". It could be that cannabis is killing you in some subtle and devious way, but so it would be with cucumbers and wearing shoes. Most people aren't afraid of eating cucumbers and wearing shoes because we could discover some weird way it's bad for you in the future. Same approach, I think, is warranted for cannabis, and that what 0 means here, and whether it's 0 or 0.0000001 is in practice of little importance, and there's no reason to demand from cannabis peddlers more rigor in distinguishing those than we demand from cucumber peddlers. As a side note: watermelons are deadly, at least in some countries - i.e. in Russia every year several people die from ingesting watermelons, as a result of bad agricultural safety controls, widespread tendencies of buying them from random providers, e.g. selling them off a truck in season, and virtual impossibility of checking if anything is wrong with it before eating.

Also, a "low non-zero number" may not be in any way more accurate than zero, but yes, psychologically it looks like "somebody did the job of calculating it" - even if the number if completely random and none of its digits is trustworthy. It's described in the "Peter principle" I think - the meeting set to 1pm is understood as "after lunch sometime" but 1:14pm is understood as "he's not wasting even a minute, and we shouldn't either!" But is that precision actually real? If they have data that suggests there may be anywhere from 0 to 100 deaths a year from cause X, and there's no way to make it more precise, then saying "it's 54.36" is kinda lying - there's nothing in the data that makes any of these digits mean anything. It looks precise but it's actually fake.

Except smoking tobacco also kill you in some subtle and devious ways, and it took us hundreds of years to actually get formal scientific data for that. It seems entirely reasonable that smoking marijuana would have similar effects which are significantly above zero but non-obvious due to the long-term nature of it.

Absent hard data pointing in either direction, it seems reasonable to ballpark guess that it's probably has similar effects to things most similar to it, of which tobacco cigarettes seem like the closest comparison (which are a thing people smoke and inhale), not watermelons (which are a fruit that people eat).

More importantly, if we have no idea, you shouldn't report it in a table of things that we do have an idea about. You should leave it off the table or say "we have no idea"

Yeah. I am thinking 'net benefit for a smallish number of people, net loss for most, catastrophic for one person in a couple hundred'. Cigarette smoke IIRC fucks you up because you're inhaling a bunch of burning shit, and that's not good. We might have only been able to tease that out once people weren't breathing in tons of woodsmoke from open indoor fires and shit tier chimneys on a regular basis plus not dying as often to disease in adulthood.

Doctors I know see lots of people with chronic conditions exacerbated by THC. I'm skeptical that it does not significantly increase all-cause mortality. This study finds a significant increase in hospital visits but not a significant one in mortality, while this one does find an increase in all-cause mortality, fwiw.

Given that medical marijuana is significantly more legalized than recreational marijuana, shouldn't one expect that the users of marijuana would have higher all cause mortality - because at least some of them would have medical conditions for which one can get medical marijuana prescription? That, of course, if we're talking about habitual users.

Also, the latter study compares "ever users" with "nonusers" - I don't think there's a serious basis to claim that using marijuana once in your life could kill you. If these results are real (which is not guaranteed, one study can be anything) I'd rather take it as a proxy for class, wealth, culture or something like that. It'd require extremely strong proof to make the idea that once-a-lifetime use of marijuana in high school raises your all-cause mortality 14% and your cardiovascular mortality 3x acceptable. It's much more plausible for me that we've got another reverse causality here. That of course if we assume these results are real and not just bad stats or wholesale fake, which is not that rare anymore.

Yeah those are great points, I just don't want to put in the work to find better studies haha. It does seem likely THC doesn't have a large effect on mortality.

WRT THC: I think even if you take the "smoking anything is bad for you" thing out of the equation and talk about, like, hash cookies or something (which have no delivery-related complications I'm aware of), it's probably still not totally free of mortality risk. It can trigger schizophrenia, which has at least two obvious pathways to dying (it can make you attack people, which can get you killed in self-defence or by police; it can also lead to homelessness, which can get you killed by murder/exposure/infection/etc.). Three "can"s in both cases means this is quite a low risk, but it's not zero.

Or, y'know, suicide from having your mind ruined. It can create/trigger other things too, like DP/DR, with pretty extreme subjective suffering. The people who say weed is some harmless wonder plant are annoying.

The people who say weed is some harmless wonder plant are annoying.

Agreed, but its worth considering that "weed is harmless" is a considerably better approximation of the truth than weeds classification as a schedule 1 nacotric in the US.

Fair enough; it's worse than cigarettes but not as bad as alcohol. Booze is legal; pot should also be legal.

DP/DR?

Depersonalization/Derealization. It can come in episodes or become chronic and last a long time. Severity varies too. The diagnosis for the chronic thing is called Depersonalization Disorder, last I checked.

That would be 99% of reddit, and I agree with you. Even my own limited experience with pot has given me pause. I think I may be more susceptible than most to its effects.

My guess at the 2000 would be something to do with long-term health effects of chronic abuse and/or making heart problems more likely in those already predisposed and/or some kind of weird drug interactions (caffeine normally is pretty clean on that front, but there's got to be something). This is not to say their numbers are accurate, but I could see triple or quadruple digits being plausible (note that the 92 are only cases of acute poisoning with no other contributing causes of death).

I figured it was probably something like that. The 92 was just all I found in the time I spent google searching for it.

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!"

(popular)

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.