The paper is a rigorous analysis compiling the best available epidemiological information. It includes only prospective studies (studies that are less biased than other methodological designs), and assesses their quality, especially in relation to the usual biases of studies on alcohol consumption: the bias of occasional drinkers and ex-drinkers. In many cases, ex-drinkers have stopped drinking as a consequence of a related health problem, and their inclusion in the category of non-users 'contaminates' this group, which is often considered as the reference group for comparison, since ex-drinkers may have a higher risk of mortality and this may be related precisely to previous alcohol consumption. 

The results confirm previous evidence of a harmful effect of high alcohol consumption, while suggesting neither a protective nor a risk effect in moderate consumption. Low or moderate alcohol consumption is not associated with a higher or lower risk of overall mortality, so it is not beneficial, but neither is it risky. The fact that no higher or lower risk is observed does not mean that it does not exist, but that it could not be proven in the present study. 

The results compile the available evidence and include studies from different countries, including Spain. However, as these are not representative populations but groups of voluntary participants, the generalisation of the results should be made with caution. In addition, the countries where most research is carried out (generally North America and Europe) tend to be over-represented. This may mean that if well-designed studies had been carried out in countries with very different levels of consumption, and could have been included in the meta-analysis, it cannot be ruled out that the results would have been different. 

This is evidence from observational studies, and this entails known biases and limitations: errors in the calculation of alcohol intake and the possible existence of unknown or unmeasured relevant factors. 

These aspects should always be taken into account as limitations - although, having included prospective cohort studies, the biases and limitations are smaller than in other epidemiological study designs (surveys, case-control, retrospective cohort), but larger than in controlled clinical trials. However, the latter type of design cannot be performed due to ethical issues, with prospective cohorts being the most suitable for this type of study. 

The article helps to reinforce the available evidence on the harmful effect of excessive alcohol consumption. It should be understood that these results should not encourage consumption, but discourage the abuse of alcoholic beverages. Regarding low or moderate consumption, controversy remains, although other authors and studies support that any consumption could be a health risk.

This is a systematic review and meta-analysis that jointly analyses data from a large number of studies to find out the effects of drinking different amounts of alcohol on the risk of death. It follows the relevant methodological recommendations and can therefore be considered of good quality. 

There is now no doubt that heavy drinking is clearly detrimental to health, but controversy remains about the effects of moderate drinking, although studies with more rigorous methodologies are increasingly appearing that show no benefit from drinking small amounts. The results of this study are along these lines, as they do not find a lower risk of death for low or moderate consumption, but a higher risk for high consumption. Moreover, in women the risk starts to increase with lower amounts of alcohol, something that is already known and is reflected in lower risk limits set for women than for men. The results could be perfectly applicable to our environment, since the work includes numerous studies carried out in several countries, including our own. 

Limitations of the paper include possible errors in the amounts participants say they drink, and in the classification of non-drinkers, since in some studies it is difficult to know whether they have never really drank or have stopped drinking because of health problems. 

As a result of this work, alcohol consumption should not be recommended to improve health, even in small quantities, and it should be clear that, if you drink, the less the better.