When a pediatrician tries to correct oncologists about cancer research
From time to time we see how the newspapers ring the alarm bells about the harms caused by our lifestyle choices. It is often very specific foods and habits that are targeted. Although healthy criticism always is appreciated, I think it is important to separate it from public health nihilism.
Unfortunately, I think Aaron E. Carroll in The New York Times is guilty of the latter in his article: “A Link Between Alcohol and Cancer? It’s Not Nearly as Scary as It Seems”
What is the background of my accusation? Although this isn’t news to most of us working with public health, The New York Times reported that The American Society of Clinical Oncology (ASCO) just released a new statement that calls to attention the link between alcohol and cancer. They emphasized that even light alcohol use affects cancer risk, especially breast and esophageal cancer.
In response to this, Aaron wrote a NYT blog where he criticized the angle taken in the original article as well as the statement from ASCO about their “new” announcement.
Logical fallacies in attempt to discard science
In my response I will try to correct Aaron on some points he made and also highlight some of his logical fallacies that he is responsible for. Let’s begin!
Citing freely from his article, “alcohol only accounts for 3.5% of all cancer deaths, which means that 96.5% of deaths due to cancer are not caused by alcohol”. Well, although this can be true in theory (with the exact number changing between countries and years) this is the simple nature of disease etiology, where many factors affect your risk of getting and dying of cancer. Even if we take tobacco, it “only” accounts for 8.7% of total mortality cases. This means that 90% of all deaths are not caused by tobacco! Does that mean we shouldn’t worry about tobacco use?
We must also remember that different fractions of the disease etiology only reflect certain practices at this exact moment. Thus, if people would use less alcohol, the number would decrease and if we (wrongfully) downplay the role of alcohol on our health it is probable that the total consumption of alcohol would increase and then become an even more important risk factor.
Relative and absolute risk
Then the difference between relative and absolute risk is mentioned. Once again, this is true in theory, a relative risk increase becomes a more tangible number if we know how many are affected by a certain disease.
But if we think in terms of the absolute risk for the individual of dying from different causes and diseases we are not in the domain of public health any more. As servants of public health we have to look at the bigger picture, like in this case: in the US alone 19,500 people die of cancers that are caused by alcohol. As public health advocates we work to decrease this number and try to keep the fire in check instead of further fueling it with ignorance or nihilism.
The same is true for other issue as well, like road traffic accidents and Sexually Transmitted Diseases(STDs): even if the risk could be quite small for the individual to die in a car accident or contract HIV, we should try to limit the damage by encouraging people to use seat belts and condoms.
Later on in his article the rhetoric becomes even muddier. On the one hand he advises to not lend excessive weight to observational data but at the same time Aaron says that the public should acknowledge the fact that alcohol also seems to be cancer protective for some types of cancer (which obviously is based on observational data). His conclusion is therefore that we must accept all suggestive evidence no matter the size of the effect nor the point of direction.
A crash-course in public health
I feel that a lot of conclusions made in the article are based on a flawed idea of the concept of scientific epistemology. I therefore think that a crash course in public health would be in its place.
The reasons we trust observational data that clearly is pointing to the fact that alcohol causes cancer is because there is coherence (alcohol affects several types of cancer growth) in the evidence and we have several plausible biological mechanisms for how this would occur. For example, the breakdown of alcohol is acetaldehyde which is a toxic chemical and a probable human carcinogen which can damage both DNA and proteins in the cells.
This is the exact same reason why we trust the evidence that smoking causes cancer. The only difference is that we have collected so much observational evidence that no one questions the connection any longer.
But why is it that we don’t lend the same weight to the (few) studies showing a marginal lower risk for non-Hodgkin lymphoma and renal cancer in heavy alcohol users? Because we interpret contradictory claims with caution. We don’t see a dose response relationship here and we don’t even have a plausible biological mechanism to explain this observation!
But why trust observational data when we have randomized controlled trials (RCT)? RCT is the gold standard for figuring out how humans are affected by different risk factors. The design of RCTs is strong indeed, but they have their limitations and their own place in epidemiological research. For example, some things are not possible to investigate through randomized controlled trials, which is also the reasons why we accept the effect of tobacco smoking: because this has never been proven in a RCT, it wouldn’t be ethical! The same is true for alcohol.
I therefore doubt that the RCT-design is the optimal tool for investigating the link between alcohol use and cancer.
Aaron thinks it is dangerous to inform the public when the scientific community finds evidence about the dangerous nature of one the most common drugs of the modern world, because people will throw their hands in their air and exclaim “everything causes cancer!” One thing I know for sure is that the public, and most importantly, the tobacco industry said the exact same thing when the Surgeon General’s Report came out in 1964. Although I agree with Aaron that people should decide for themselves when it comes to their lifestyle choices, in order to give people an informed choice we need to start to apply health information to alcohol, too – like with all other products for human consumption.
Let’s take a holistic view of alcohol
Another point I agree with Aaron on is that we should try to take a holistic view of alcohol when we talk about health, because it’s not just about the cancer. I don’t have time to go into the so-called evidence of alcohol and cardiovascular health but here are a few bullets points to chew on:
- Alcohol is the third most important risk factor for the loss of disability adjusted life years globally.
- Alcohol kills 3.3 million people worldwide every year. It means: Every 10 seconds a human being dies because of alcohol.
- Alcohol harm is a tremendous burden on the young people of the world: Alcohol consumption causes death and disability early in life – relative to other health hazards. In the age group 20 to 39 years of age about 25% of the total deaths are alcohol-attributable.
- There is a causal relationships between alcohol use and incidence of infectious diseases such as tuberculosis as well as the course of HIV/AIDS.
- Harm caused to others than the alcohol user him/ herself is an important aspect of the total burden of alcohol harm. Alcohol’s harm to others affects family members, friends, co-workers and strangers. Some examples are emergency room staff, police, taxi drivers, children of alcoholics, children born with fetal alcohol syndrome, road traffic fatalities, or alcohol-related violence – especially gender-based violence.
- And so on… all facts (and more) can be found here.
We must remember, alcohol is a packaged deal. So, we really need a holistic view. Alcohol comes with public health problems as wells as pervasive social harm and massive economic costs – which we all, from politicians to us citizens are paying for. Therefore, I think that informing about and addressing the true risks attributed to alcohol use is essential.