On 2016 Jun 12, John Duffy commented:

Kari Poikolainen and I sent a letter to the Lancet about this paper. It was not published. The content is as follows

The authors report no association between level of alcohol intake and a composite health measure. They write that "In summary, our study shows that current drinking is not associated with a net health benefit." What they fail to mention is that their study also "shows" that current drinking is not associated with net health harm. The interpretation, "sufficient commonalities to support global health strategies and national initiatives to reduce harmful alcohol use" does not follow from their study, which taken on its own suggests that there is no need for such initiatives. However there are a number of methodological problems with their data and analysis. The follow-up time (median 4.3 yrs) was short and the proportion of alcohol consumers (33%) small. There is a strong implication of overfitting, and in the stratified analyses in particular a shortage of degrees of freedom. Since a large body of earlier well-conducted studies shows evidence of both benefits and harms it would be inappropriate to base conclusions on this work.

On 2015 Sep 26, David Keller commented:

Dose-response relationship observed between concentration of ingested alcohol and cancer rate

This study reported that "the risk for cancer was 38% higher in wine drinkers than in never drinkers (HR, 1.38; 95% CI, 1.05 - 1.81), 69% higher in spirit drinkers (HR, 1.69; 95% CI, 1.26 - 2.26), and 20% higher in beer drinkers (HR, 1.20; 95% CI, 0.91 - 1.57)." These 3 data points demonstrate a dose-response curve relating the concentration of ingested alcohol with increased risk of cancer. For cardiovascular (CV) disease, it is the amount of alcohol consumed per day, not its concentration, which is known to be the relevant modifiable risk factor. The maximum CV benefits are associated with consumption of around 2 standard servings per day of alcohol of any concentration.

Note that the 95 percent confidence interval for the Harm Ratio due to drinking beer spanned 1.0, and thus beer did not raise the risk for cancer to a statistically significant degree.

5 percent Beer.....HR = 1.2 non-significant

12 percent Wine.....HR = 1.38

40 percent Spirits..HR = 1.69

Dose-Response Between Alcohol Concentration and Harm Ratio for Cancers

Thus, drinking beer, which, in moderation, is associated with significant cardiovascular (CV) benefits, is not associated with any statistically significant increase in cancers, as are wine and spirits, presumably due to beer's lower alcohol concentration.

This study did not report separate findings for high-volume and low-volume beer drinkers for its risk-benefit analyses. If beer drinkers who limited themselves to 2 servings per day had been analyzed separately, might the harm ratio for incident cancers have been even lower? Might a reduction in cancer incidence perhaps have been observed? At what volume of daily alcohol ingestion in the form of beer (the least carcinogenic alcoholic beverage) was the greatest improvement seen in the composite outcome of cancer and CV disease? Does the optimum amount of beer per day differ for the composite outcome versus the optimal 2 servings per day associated purely with optimal CV outcomes?

The main purpose of studies such as this should be to result in actionable information. Telling us that low economic level countries exhibit different outcomes associated with the consumption of alcohol does little good unless we are given clues as to what alcohol consumers in poor countries might be doing differently than their counterparts in richer nations. Are they drinking higher concentrations of alcohol, greater volumes, both, or is their nutritional status or some other factor implicated?

On 2015 Sep 26, David Keller commented:

Dose-response relationship observed between concentration of ingested alcohol and cancer rate

This study reported that "the risk for cancer was 38% higher in wine drinkers than in never drinkers (HR, 1.38; 95% CI, 1.05 - 1.81), 69% higher in spirit drinkers (HR, 1.69; 95% CI, 1.26 - 2.26), and 20% higher in beer drinkers (HR, 1.20; 95% CI, 0.91 - 1.57)." These 3 data points demonstrate a dose-response curve relating the concentration of ingested alcohol with increased risk of cancer. For cardiovascular (CV) disease, it is the amount of alcohol consumed per day, not its concentration, which is known to be the relevant modifiable risk factor. The maximum CV benefits are associated with consumption of around 2 standard servings per day of alcohol of any concentration.

Note that the 95 percent confidence interval for the Harm Ratio due to drinking beer spanned 1.0, and thus beer did not raise the risk for cancer to a statistically significant degree.

5 percent Beer.....HR = 1.2 non-significant

12 percent Wine.....HR = 1.38

40 percent Spirits..HR = 1.69

Dose-Response Between Alcohol Concentration and Harm Ratio for Cancers

Thus, drinking beer, which, in moderation, is associated with significant cardiovascular (CV) benefits, is not associated with any statistically significant increase in cancers, as are wine and spirits, presumably due to beer's lower alcohol concentration.

This study did not report separate findings for high-volume and low-volume beer drinkers for its risk-benefit analyses. If beer drinkers who limited themselves to 2 servings per day had been analyzed separately, might the harm ratio for incident cancers have been even lower? Might a reduction in cancer incidence perhaps have been observed? At what volume of daily alcohol ingestion in the form of beer (the least carcinogenic alcoholic beverage) was the greatest improvement seen in the composite outcome of cancer and CV disease? Does the optimum amount of beer per day differ for the composite outcome versus the optimal 2 servings per day associated purely with optimal CV outcomes?

The main purpose of studies such as this should be to result in actionable information. Telling us that low economic level countries exhibit different outcomes associated with the consumption of alcohol does little good unless we are given clues as to what alcohol consumers in poor countries might be doing differently than their counterparts in richer nations. Are they drinking higher concentrations of alcohol, greater volumes, both, or is their nutritional status or some other factor implicated?

On 2016 Jun 12, John Duffy commented:

Kari Poikolainen and I sent a letter to the Lancet about this paper. It was not published. The content is as follows

The authors report no association between level of alcohol intake and a composite health measure. They write that "In summary, our study shows that current drinking is not associated with a net health benefit." What they fail to mention is that their study also "shows" that current drinking is not associated with net health harm. The interpretation, "sufficient commonalities to support global health strategies and national initiatives to reduce harmful alcohol use" does not follow from their study, which taken on its own suggests that there is no need for such initiatives. However there are a number of methodological problems with their data and analysis. The follow-up time (median 4.3 yrs) was short and the proportion of alcohol consumers (33%) small. There is a strong implication of overfitting, and in the stratified analyses in particular a shortage of degrees of freedom. Since a large body of earlier well-conducted studies shows evidence of both benefits and harms it would be inappropriate to base conclusions on this work.