You probably came across news recently on the Surgeon General Advisory on alcohol use and cancer.
I’ve had a few patients ask me about this, and I’ve similarly gotten a handful of worried questions from friends and family members.
When news breaks on a topic like this, I try to avoid drawing major conclusions until I’m able to read the actual report and take a look at some of the data directly.
A report like this makes news because it’s a large data synthesis by an official with a large public platform.
But when you’re interpreting how a report like this impacts your own life, it’s vital to remember the context that the report isn’t coming because of brand new information bringing unknown cancer risks to light.
My high level interpretation is that there’s literally zero new information here to update my priors and that my advice to patients will be the same after this report as it was before this report.
I also think that the report was alarmist and overstates the level of confidence that anyone has about a “safe” level of drinking.1
Of course, as with many things, there’s room for nuance and deeper understanding, so let’s look in more detail.
Why did the report say that alcohol causes cancer?
The report cites quite a bit of data that unambiguously demonstrates that people who drink alcohol have a higher risk of cancer than people who don’t drink alcohol.
But is that link causative or is it just an association?
It’s way harder than you might think to answer the simple question: does alcohol cause people to get cancer, or are people who drink more likely to get cancer because of some other reason?
Think about it this way: how would you design an experiment to test whether alcohol causes cancer?
The gold standard method of doing this would be to take two large groups of people, give half of them alcohol and half of them an alcohol placebo, make sure that no other behaviors were different, and then follow them for a few decades to see which group developed more cancer.2
And ideally, you’d really want a greater number of groups, so that you could give each of the groups with alcohol a different amount to drink so that you could see whether there’s a dose-response relationship.
In other words, it’s impossible to actually run the type of experiment that will answer the question that we want.
And so if you can’t design an experiment to answer your question, you’re going to be left with trying to draw conclusions from really limited data.
In the report, they reference the Bradford Hill Criteria as important for developing an understanding about why the link between alcohol and cancer is causal rather than associative.
These criteria require you to develop the evidence across multiple domains - the strength of the association matters, but so do things like biologic plausibility, a dose/response relationship, and the consistency of the link across populations.
We can feel pretty confident that the link between heavy drinking and cancer is a causative one
The epidemiology is clear - people who drink get more cancer than people who don’t and people who drink more get more cancer than those who drink a little bit.
This is a consistent finding across many populations.
We have plausible biologic mechanisms,3 which are clearly highlighted in the report:
There is experimental data and animal data to demonstrate that there is a clear pathway for alcohol to cause cancer to develop.
And so I think the evidence is pretty clear that high levels of alcohol use certainly do increase the risk of cancer.
Where this gets murky is whether small amount of alcohol also increase cancer risk.
The reason it gets murky with lower levels is the dose-response part of this relationship.
Take a look at this graphic studying the relative risk of different types of cancers stratified by people who are light, moderate, and heavy drinkers:
It’s pretty clear that heavy drinking is a different beast than moderate or light drinking.
It’s also clear that cancer risk related to alcohol exposure is probably different for different tissues in the body.
Notably, the areas of the body directly exposed to alcohol seem to have the more clear links - which provides additional support that alcohol is carcinogenic.
The breast cancer connection illustrates the challenge with just showing a mechanism and then assuming causation
One biologic plausibility issue from the report that doesn’t make sense to me is the way that the authors describe the role of estrogen in breast cancer.
You’ll see from the mechanism graphic above that one hypothesized mechanism for breast cancer development due to alcohol is through estrogen exposure.
If the alcohol-breast cancer link is mediated by estrogen, you would expect estrogen-receptor positive breast cancers to be more closely linked with alcohol use than estrogen-receptor negative breast cancers.
But the paper that the report cites on breast cancer and alcohol exposure says the opposite - there’s no difference in estrogen receptor positive and estrogen receptor negative breast cancer incidence when it comes to alcohol use.
My point here is not that alcohol does not cause breast cancer - it certainly might! - but rather that a biologic mechanism + epidemiology does not necessarily equal causation.
The impact of light and moderate drinking on cancer risk is truly impossible to ascertain
As I described earlier, it’s not possible to do the type of experiment that you’d want to do to answer the question of causality.
And it’s not fair to do what the Surgeon General’s report did: look at the epidemiology, link it with biologic mechanisms, add in some animal studies, and suggest that you have proven causation.
That type of case building is what the seed oil people do, too.
While a report like this is way more rigorous than most nutritional research that you’ll see, there’s still an element of sleight of hand and the transitive property being applied to suggest that we have a higher level of confidence in the conclusion than is warranted based on the data.
And so on one hand, you can make a compelling argument that says the link between cancer and heavy drinking is clear: we have a mechanism plus experimental evidence so the precautionary principle says that we should counsel people to avoid alcohol completely to lower their risk of cancer.
But are you really sure that there’s nothing different in the lifestyle behaviors between people who don’t drink and people who drink moderately that might explain a small increased risk of cancer?
Nothing about their diets, or their exercise habits, or their health consciousness, or their socioeconomic status, or their health literacy, or their risk aversion, or their sleep, or their visceral fat, or anything else?
Really?
You also need to keep in mind that the data on alcohol consumption is based on self reports, which introduces additional possibilities for data unreliability and may exacerbate the healthy user bias.4
The data that we have simply cannot tell us useful information about whether all alcohol exposure is carcinogenic and it’s dose dependent, whether there’s a threshold effect, or whether there’s complicated heterogeneity across people based on their alcohol dehydrogenase production.
We simply don’t know, and my argument is that we probably can’t know.
Part of the reason this is news: the link between alcohol and cancer is a knowledge gap for many people
One notable thing about the report: very few people know about the link between alcohol and cancer.
The report cites data showing that less than half of Americans are aware that alcohol use increases cancer risk, which interestingly, is lower than the number of people who are aware that obesity increases cancer risk.5
And so I suspect that one of the reasons that the evidence here was interpreted with the most alarmist perspective was to draw attention to an area that people don’t know as much about as we should.
So what are we supposed to conclude from this? The Surgeon General says that no amount of drinking is safe when it comes to cancer risk
In an interview with the New York Times, Vivek Murthy, the Surgeon General said:
“Many people out there assume that as long as they’re drinking at the limits or below the limits of current guidelines of one a day for women and two for men, that there is no risk to their health or well-being. The data does not bear that out for cancer risk.”
And while that’s not exactly incorrect, it is a bit misleading.
The data doesn’t bear out that light or moderate drinking is safe when it comes to cancer risk, but the report also doesn’t show that it isn’t.
We’re left now in the same position we were before the Surgeon General’s office synthesized all this data - uncertainty about how much alcohol increases cancer risk and whether there’s a safe amount to drink if you want to be cautious.
What I’ll tell my patients about this news
When I get asked about alcohol, I usually tell patients a handful of things:
You shouldn’t be under the misconception that alcohol use improves your health
Light drinking is probably fine
Heavy drinking is clearly bad for you
The line between ok and too much is tough to figure out and is different from person to person
Alcohol impacts health in other ways beyond the immediate impact of the substance - it changes your sleep, it changes your food choices, it has a significant calorie burden6
Every person likes alcohol a different amount, and alcohol plays different roles in all our lives, and I would advise you think about the role it’s playing in yours and have your drinking habits be intentional rather than accidental
And, as with many things, the advice of “don’t be stupid” applies here as well.
My other take home message here is that we’ll never know the answer to a lot of the seemingly simple questions that arise from a report like this.
It’s no different than a lot of the simple questions we’d like to answer about a healthy diet.
There are some universal aspects of the human existence where the impact on disease is literally unknowable based on the limits of the scientific method.
And so if you were not aware of the link between alcohol and cancer, then being exposed to this report is probably good for you.
I’m not sure whether the public health communication strategy of exaggerating the confidence we have about limited information is a productive one, but I certainly understand why that was the approach that was taken here.
Overstates it in either a good way or a bad way.
There’s a lot of other complexity you’d want to add to the process. Make sure that the calories in the placebo were the same as the calories in the alcohol but the placebo didn’t have any potentially toxic metabolites that would confound the results. Make sure that the alcohol group didn’t consume any extra calories after they had drank. Ensure that both groups had the same quality and duration of sleep (thought to be a chronic disease risk factor). Give them the same sun exposure, exercise regimen, macronutrient profile, consumption of processed meats, sun screen use, environmental pollution exposure, socioeconomic status, and a million other variables. In other words, it’s an impossible experiment to run.
As I have said before in this newsletter, the road to hell is paved with biologic plausibility. Don’t fall in love with a mechanism and think that you know everything about a complex problem like cancer development.
In other words, people who are more health conscious may report their drinking habits as differently than people who are less health conscious.
Knowledge about tobacco, radiation, and asbestos are all way higher.
The impact on sleep and food choices are also plausible ways through which alcohol could influence cancer risk.