A simple methodological error meant that for years researchers considered drinking moderate amounts of alcohol to be healthy.