Lots of people think a glass of wine or beer at dinner can help them have a longer and healthier life. But a new study suggests that much of the evidence in favor of moderate drinking may be shaky at best.

Scientists took a closer look at findings from 87 previously published studies on drinking and death from all causes and found all but 13 of these experiments had a critical flaw.

Most of the studies compared moderate drinkers - people who had one or two drinks a day - with current abstainers. The problem is the studies didn't account for medical reasons that may have driven abstainers to avoid alcohol, potentially exaggerating the health benefits seen with moderate drinking.

After taking this so-called abstainer bias into account, "our study found no net benefits overall," said lead study author Tim Stockwell, director of the University of Victoria's Center for Addictions Research in British Columbia, Canada.

"People should not drink for health - health benefits can be obtained in many ways and drinking in any pattern is not a reliable means to this end," Stockwell added by email. "Most of us enjoy alcohol, and drinking lightly and occasionally presents the least risk."

Thirteen studies did account for abstainer bias, and none of them found health benefits associated with moderate drinking, Stockwell and colleagues report in the Journal of Studies on Alcohol and Drugs.

The analysis didn't look at whether certain types of alcohol such as red wine might be tied to longer life, the authors note. It also focused on mortality from all causes, which means it's still possible that drinking might be associated with a longer life for people with certain conditions such as heart disease, Stockwell said.

Even so, the evidence to date suggests that the safest level of alcohol consumption is none at all, said Jennie Connor, chair in preventive and social medicine at the University of Otago Medical School in New Zealand and author of an accompanying editorial.

More on this...

Studies looking for health benefits of drinking often rely on participants to accurately recall and report how much alcohol they consume, which can be an unreliable way to assess whether the habit is good for people, Connor said by email.

"Fairly poor measures of consumption are a feature of these kinds of studies," Connor said.

One way to prompt better recollection in study subjects is to ask them to name specific brands of beer or vintages of wine they drink, said Tom Greenfield, scientific director of the Alcohol Research Group in Emeryville, California, and author of a separate editorial.

There are also low-risk levels of drinking that people can follow to minimize harm even in the absence of research proving alcohol is beneficial, Greenfield added by email.

"Never binge drinking and drinking small amounts when drinking - with meals - is a good rule of thumb," Greenfield said.

Jurgen Rehm, chair of addiction policy at the Dalla Lana School of Public Health at the University of Toronto, echoed this sentiment in a third editorial.

"People should not drink for health reasons," Rehm said by email. "If they choose to drink - usually for reasons other than health - they should not exceed two drinks on any given day to minimize their health risks."