New booze guidelines mix science and opinion
The Canadian Centre for Substance Use and Addiction has issued its report Update of Canada’s Low-Risk Alcohol Drinking Guidelines: Final Report for Public Consultation. A serious failing of the report is that it blurs the distinction between science and opinion in its recommendations. This has contributed to confusion, so that public and media discussion of the report are a muddle of disbelief and exhortations to “follow the science.”
The public doesn’t appreciate the distinction between a scientific finding and an opinion. Polling by the Ontario Science Centre reveals that 43 per cent of respondents think that “scientific findings are a matter of opinion,” and 75 per cent think that “scientific findings can be used to support anything.” This conflation is evident in our responses to climate change and the COVID-19 pandemic, and persists even though rigorous scientific methods have been constructed over time to be reliable, reproducible and independent of the scientists who perform the work. Clear communication of this distinction is crucial if science is to contribute to addressing important societal questions.
There is excellent science in this report. The authors use scientific protocols with rigorous criteria to collect reliable clinical studies of the association between alcohol consumption and conditions such as cancer, heart disease and road accidents. They combine these results in a model calculation of the Years Lost in a Lifetime (YLL) for different levels of alcohol consumption in grams/day.
YLL is an unusual unit, but is directly related to the risk of dying prematurely. The results of their model are presented in the graph here, taken directly from the report (I have added a few vertical markers). The solid lines are the results of the calculation. The dashed lines running above and below the solid lines (one pair for males, the other for females) are the confidence limits of the calculation.
What do the dashed lines mean? Risk is a statistical probability, and alcohol consumption is just one of myriad factors that can shorten a person’s life. The study is trying to isolate with statistical confidence the effect of a small level of alcohol consumption (the “signal”) amid all the other possible factors (the “noise”). An analogy is trying to hear a friend across the room in a crowded restaurant. Between the upper and lower dashed lines, the noise is so loud you cannot tell what your friend is saying.
A rigorous, scientific interpretation of the graph is that the effect of alcohol consumption on lifespan lies somewhere between the upper and lower dashed lines with 95 per cent certainty, but we cannot tell where.
For a scientific finding that alcohol is reducing lifespan, the range between the dashed lines must exclude zero YLL — we must be confident your friend is at least in the restaurant. This occurs at 12 grams/ day for females or 16 grams/day for males. This consumption represents an average reduction in a lifespan anywhere between zero and six months (0 to 550 YLL per 1000 persons).
The authors have chosen four grams/day of alcohol as their recommendation for low risk. At this intake, because the lower dashed lines are not even close to zero YLL, the recommendation does not meet the scientific standard. It represents the opinion of the authors. Regrettably, this is not made clear, and the media and public have assumed that it is a recommendation based on a scientific finding.
The report blurs the distinction between science and opinion and causes confusion. It should clearly communicate both pieces of information: converted to standard drinks, science indicates a threshold of six to eight drinks/week, but the authors’ expert opinion is that two drinks/week is a better recommendation.
Individuals can then decide what risk they accept. Some may value the opinion of these experts; others may not. People have different opinions and trust different experts, but a common base of scientific findings can ground discussion. Undermining public understanding and confidence in scientific findings cannot serve any positive purpose.