Admittedly I haven't read the whole study. But how do they account for the confirmation bias in their review – the fact that null results typically aren't published. Studies that look at an intervention (Tylenol) and fail to find an association are less likely to include it as a variable in their model or less likely to publish a null finding.