Correspondence
3 Knott CS, Coombs N, Stamatakis E, control for education or income in alcohol. Second, a focus on current
Biddulph JP. All cause mortality and the case
for age specific alcohol consumption
their overall analysis. It would seem drinkers should reduce potential
guidelines: pooled analyses of up to plausible that the author’s main biases that are difficult to control in
10 population based cohorts. BMJ 2015; finding—mortality and morbidity observational studies (eg, reverse
350: h384.
4 Wood AM, Kaptoge S, Butterworth AS, et al. are higher in people who consume causality, residual confounding, and
Risk thresholds for alcohol consumption: 100–200 g alcohol per week than in unmeasured effect modification),
combined analysis of individual-participant
data for 599 912 current drinkers in
people who consume less alcohol—is because ex-drinkers might include
83 prospective studies. Lancet 2018; confounded by binge drinking. There people who have abstained from
391: 1513–23. are no binge drinkers in the group alcohol owing to poor health itself,2
5 Rimm EB, Giovannucci EL, Willett WC, et al.
Prospective study of alcohol consumption and that consume 0–100 g per week as well as people who have changed
risk of coronary disease in men. Lancet 1991; because the author’s definition of their habits to achieve a healthier
338: 464–68.
binge drinking is more than 100 g per lifestyle. Third, never-drinkers might
6 Pai JK, Mukamal KJ, Rimm EB. Long-term
alcohol consumption in relation to all-cause occasion. differ systematically from drinkers
and cardiovascular mortality among survivors I declare no competing interests. in ways that are difficult to measure,
of myocardial infarction: the Health
but might be relevant to disease
Professionals Follow-up Study. Eur Heart J Warren Thompson
2012; 33: 1598–605.
[email protected]
causation.3 For example, our study
7 Browne R. Just five alcoholic drinks a week found important differences in key
could shorten your life, study says. April 13, Mayo Clinic, Rochester, MN 55905, USA
2018. https://www.cnbc.com/2018/04/13/ baseline characteristics between
1 Wood AM, Kaptoge S, Butterworth AS, et al.
five-alcoholic-drinks-a-week-could-mean-a-
Risk thresholds for alcohol consumption: never-drinkers and current drinkers
shorter-life-study-says.html (accessed
May 18, 2018). combined analysis of individual-participant (eg, sex, ethnicity, smoking, diabetes
data for 599 912 current drinkers in
8 The Telegraph. Glass of wine or beer a day 83 prospective studies. Lancet 2018;
status), but other differences are also
reduces risk of an early death, says new study. 391: 1513–23. likely to exist for characteristics we
Aug 14, 2017. https://www.telegraph.co.uk/
news/2017/08/14/glass-wine-beer-day- did not record.4
reduces-risk-early-death-says-new-study/ Authors’ reply Our analysis therefore aimed to
(accessed May 18, 2018).
To define alcohol consumption minimise the scope for residual bias
thresholds associated with lowest by comparing different levels of
Angela M Wood and colleagues1 risk for all-cause mortality and alcohol consumption among current
conclude that all-cause mortality cardiovascular disease subtypes, we drinkers. A complementary approach
is higher in people consuming analysed individual-participant data to minimising such bias involves
100–200 g alcohol per week than in from 599 912 current drinkers without studying genetic proxies for drinking
people consuming 0–100 g alcohol previous cardiovascular disease.1 habits. For example, data from large
Atw Photography/Getty Images
per week. However, data on p38 of Among current drinkers, the threshold studies in east Asian populations,
the appendix suggest that this is for lowest risk of all-cause mortality where common genetic variants
only true for people who consume was about 100 g per week. For predict 20-fold differences in drinking
alcohol twice a week or less. For those cardiovascular disease subtypes other prevalence, have yielded results that
who consume alcohol more often, than myocardial infarction, there were are in agreement with those of our
consuming up to 200 g per week no clear risk-based thresholds below study.5
does not seem to increase mortality, which lower alcohol consumption In response to Warren Thompson,
as long as the individual does not stopped being associated with lower our consortium could not powerfully
engage in binge drinking behaviour. disease risk. These data challenge study binge drinking as a primary
People consuming alcohol more than the concept that moderate alcohol hypothesis because of limited
2 days a week had lower mortalities consumption is universally associated information available on the frequency
than those consuming alcohol 2 days with lower cardiovascular disease risk, of alcohol consumption (eg, fewer
a week or less, if they consumed the and suggest implications for low-risk than half of the recorded deaths had
same amount of alcohol. The authors limits in guidelines. this information). Furthermore, only
attribute these findings to smoking In response to Arne Astrup and partial information was available
and socio economic status, but colleagues and Eva Schernhammer, on important confounding factors
smoking was controlled for. There our paper stated the scientific (eg, socioeconomic characteristics) in
are some differences in education rationale behind focusing on current this subset of participants. We found
and income that might account for alcohol drinkers. First, guidelines wide confidence limits for hazard
these findings, as shown on p 19 in provide recommendations about ratios for association of binge drinking
the Article’s appendix. However, the low-risk limits only for drinkers; we and disease outcomes, reinforcing the
authors did not analyse these data to are unaware of any guidelines that need to interpret these exploratory
see if that was the case, and did not encourage non-drinkers to consume analyses with caution.
www.thelancet.com Vol 392 November 17, 2018 2167
