To help you make your own decisions about alcohol, I critically appraised the scientific research on alcohol and cancer risk – which sounds negative but is actually a formal, systematic approach to evaluate the quality of scientific research. Here are three key takeaways.

Several cancers – but not all of them – have been linked to drinking

Alcohol consumption has been linked to cancers of the oral cavity, throat, voice box, oesophagus, breast, colon and rectum, and liver. In the report, the surgeon general said that the risk for certain cancers may start to increase with one or fewer drinks a day.

One study that is cited in the advisory asked whether alcohol consumption increases risk of two types of breast cancer. The answer is “yes” to both – with a 35% increased risk for estrogen receptor-positive breast cancer and a 28% increased risk for estrogen receptor-negative breast cancer.

But it is much less clear whether increased alcohol intake is associated with some other cancers, such as leukaemia. So, when a patient asks me whether alcohol consumption causes cancer, the answer must go beyond “yes” or “no”. For some cancers, the answer is “no” or “we don’t think so,” but for others, the research clearly indicates the answer is “yes”.

Researchers can’t tell people to start drinking for the sake of science

The “gold standard” for study designs is randomised controlled trials, in which some people are randomly assigned to one intervention, others to another intervention, and then outcomes between the two groups are compared.

For most epidemiological research, this ideal study design is impossible and may be unethical. For example, in the study looking at alcohol and breast cancer risk, over 26,000 women with breast cancer were included from a variety of other studies looking at diet and health outcomes, in which alcohol intake was assessed.

It would have been unethical to design a study in which women were randomised to drink a lot of alcohol or none at all, to then determine who developed breast cancer, and it would take decades to answer the question.

So for this type of epidemiological question, a so-called cohort study, in which a group of people with a common characteristic (women who drink alcohol) is followed over time for a particular outcome (breast cancer), is the best path forward.

Multiple studies have reached the same conclusion. But unanswered questions remain

When critically appraising the quality of studies, epidemiologists analyse individual studies within the context of the larger scientific evidence. And when it comes to alcohol, what we do know – after multiple studies in hundreds of thousands of people reaching the same conclusion – is that alcohol is a risk factor for certain cancers.

But that doesn’t mean the data is perfect. One study cited in the Surgeon-General’s advisory included over 485,000 cancer cases from more than 500 other studies looking at several cancer types. If you take a closer look at the included studies, some information was collected in real time, but some was retrospective, which may be prone to errors.

Imagine if you had cancer and someone asked you, on average over your entire life, how much alcohol you drank each week. How accurate would your estimate be? Some might undercount their intake, while others might overcount.

Now let’s explore how the amount of drinking was defined. Light drinking was approximately one glass of beer, wine, or alcohol per day or less; moderate was less than four drinks daily; and heavy was more than four drinks daily. For some cancers, such as oesophageal and breast, the cancer risk was higher for light, moderate and heavy drinkers, compared to nondrinkers. But for others, such as colorectal, the risk was higher only for moderate and heavy drinkers.

Some of the included studies defined light, moderate and heavy drinking differently, so the investigators sometimes had to average alcohol intake reported in those studies to include them in their analyses. Additionally, there was no distinction between types of alcohol intake and cancer risks, leaving open the question of whether cancer risk is similar for people who drink wine and those who drink hard liquor.

Finally, let’s consider the scene. Do people tend to enjoy a glass of wine before dinner? Or do they drink beer at a bar while consuming a cheeseburger with fries? Or maybe smoke a pack of cigarettes with their whiskey? If it’s one of the latter two scenarios, the food and the smoking may contribute to cancer as much or more than the alcohol.

Even taking into consideration all of these factors affecting study quality, the totality of the data point to alcohol causing cancer. Less alcohol is better, and each of us must determine our own tolerance for cancer risk when deciding how much alcohol to drink.

Mikkael A. Sekeres is the chief of the division of haematology and professor of medicine at the Sylvester Comprehensive Cancer Centre, University of Miami. He is author of the books When Blood Breaks Down: Life Lessons from Leukaemia and Drugs and the FDA: Safety, Efficacy, and the Public’s Trust.