To minimize health risks, the optimal amount of alcohol someone should consume is none. That’s the simple, surprising conclusion of a massive study, co-authored by 512 researchers from 243 institutions, published Thursday in the Lancet.

The researchers built a database of more than a thousand alcohol studies and data sources, as well as death and disability records from 195 countries and territories between 1990 and 2016. The goal was to estimate how much alcohol affects the risk of 23 health problems. The number that jumped out, in the end, was zero. Anything more than that was associated with health risks.

“What has been underappreciated, what’s surprising, is that no amount of drinking is good for you,” said Emmanuela Gakidou, a professor of global health at the University of Washington and the senior author of the report.

“People should no longer think that a drink or two a day is good for you. What’s best for you is to not drink at all,” she said.

The report found that 2.8 million people across the globe died in 2016 of alcohol-related causes, which is about the same proportionally as the 2.0 million who died in 1990. For people ages 15 to 49, alcohol is the leading risk factor for experiencing a negative health outcome.

This is a sobering report for the roughly 2 billion people who drink alcohol. The report challenges the controversial hypothesis that moderate drinking provides a clear health benefit. That notion took hold in the 1990s after news reports on the “French paradox”: The French have relatively low rates of heart disease despite a fatty diet. Some researchers pointed to red wine consumption among the French as potentially protective.

Numerous peer-reviewed studies found evidence that people who have a drink or two a day are less likely to have heart disease than people who abstain or drink excessively.

But the new study, while noting the lower risks of heart disease from moderate drinking, as well as a dip in the diabetes rate in women, found that many other health risks offset and overwhelm the health benefits. That includes the risk of breast cancer, larynx cancer, stroke, cirrhosis, tuberculosis, interpersonal violence, self-harm and transportation accidents.

Drinkers may take some reassurance from the fact that the new Lancet report focuses not on individuals but on populations. It estimates risks of alcohol-related diseases and disabilities per 100,000 people as a function of alcohol consumption. The authors do not suggest that there is significant danger in having a sip of alcohol. The risks spike dramatically with heavy drinking.

U.S. dietary guidelines define low-risk drinking as one drink a day for women and two a day for men. Robert Brewer, who directs the alcohol program at the U.S. Centers for Disease Control and Prevention, said that if people stick to the guidelines, “the risk of harms across the board is going to be low. It’s not going to be zero. But it’s going to be low.”