Risky drinking or risky governance?

The National Board of Health and Welfare (NBHW) in Sweden (Socialstyrelsen, 2023) recently updated its threshold for risky drinking, to be considered by healthcare staff in patient work. This is part of their national guidelines for the treatment and prevention of unhealthy lifestyles. The scientific basis for NBHW's threshold consists of recent guidelines published in other countries and two additional systematic reviews identified through literature searches. According to the new threshold, women and men who drink 10 or more standard units (referred to as “standard glasses”) of alcohol during a week are drinking risky (“riskbruk”). Drinking four or more standard units on one occasion at least once a month is also considered risky. As a reference, a standard unit of alcohol in Sweden is defined as 12 g of pure ethanol, corresponding to, for example, 330 ml of strong beer, or 120–150 ml of wine. Below, we raise some concerns with the threshold.

Guidelines for low-risk drinking play an important role in alcohol policy and are considered legitimate as they essentially do not violate people's freedom (Lovatt et al., 2015). The guidelines are said to provide more or less objective information that individuals can utilise to decide about their drinking. There are different approaches to the development of guidelines, but typically a baseline risk is set that is considered acceptable, and then the highest level of drinking compatible with this baseline is identified. Drinking levels that equal this baseline risk are considered low risk. In some cases, the baseline risk is drawn from that of abstainers and in other cases it is an estimate of a risk that is assumed to be acceptable by the public (e.g., Holmes et al., 2019). Regarding the latter, when the outcome is alcohol-attributable mortality, the acceptable risk is typically set at either one death per 1000 individuals or one death per 100 individuals. The basic assumption is that people can accept drinking at a level where the lifetime risk of death attributable to alcohol consumption is 1‰ or 1%, respectively (Holmes et al., 2019). The same logic applies when using a measure such as years of life lost (YLL) due to alcohol consumption. For instance, in the recent Canadian low-risk drinking guidelines, a YLL of 17.5 was used based on the average number of years lost per death in Canada (Paradis et al., 2023). Thus, in this case, a 1 ‰ lifetime risk of death would correspond to 17.5 YLLs per 1000 people and a 1% lifetime risk would correspond to 17.5 YLLs per 100 people (Paradis et al., 2023).

The only risk curves shown by the NBHW are the Canadian YLL estimates. For the average reader (including healthcare staff), these estimates are probably difficult to comprehend. To take the most extreme example included in the figure (Socialstyrelsen, 2023, p. 6), which pertains to men, drinking 35 standard Canadian units per week corresponds to approximately 2500 YLLs per 1000 individuals. Converting these estimates back to a more intuitively understandable metric, 143 out of 1000 men (i.e., 14.3%) drinking 35 units a week would die from alcohol during their lifetime (2500/17.5) whereas 85.7% would not die from this alcohol consumption. We believe that this information should be much easier to understand and relate to for most people. It also appears less dramatic compared to YLL figures.

An important difference between the threshold provided by the Swedish NBHW and recent guidelines published in other countries, such as Canada, Denmark and Australia, mentioned by the board itself, is that the former should be used for making recommendations to healthcare. For people exceeding the low-risk threshold, Swedish healthcare is recommended to provide counselling or other support. This difference should be of crucial importance for the public legitimacy of the threshold. While this new healthcare practice in Sweden might identify individuals benefiting from counselling, it will probably also intrude on the integrity and freedom of those who consider themselves as ordinary, non-problematic drinkers. Low-risk drinking guidelines already lack credibility among some drinking groups (Lovatt et al., 2015). The NBHW's recommendation is therefore likely to give the impression that people without signs of alcohol-related problems also need to seek help. This appears as an unintended and (hopefully) unwanted consequence of the initiative, which also might fuel underreporting of drinking among those who seek help for health issues not related to alcohol.

The inherent limits of low-risk thresholds are quite obvious and should be acknowledged – certainly so when used in a healthcare setting. Here, we are not only thinking about the fact that guideline panels make quite arbitrary cut-offs on smooth risk curves (Room & Rehm, 2012), or that the 1‰ or 1% risk levels are far from given (Livingston, 2023). We, as others, also believe that any guideline should allow people to make decisions based on their own risk tolerance (Livingston, 2023). This issue is important for healthcare staff to understand in order to avoid blanket recommendations that patients might find non-credible or even misplaced. Estimates from the UK, based on the Sheffield Alcohol Policy Model (Holmes et al., 2016, p. 6), show that in absolute terms, many drinking patterns do not differ substantially from the 1% threshold. For instance, men drinking an average of 14 alcohol units a weak, spread over three days have an absolute lifetime mortality risk of 1.78%. Stated differently, their chance of not dying from this level of alcohol consumption is 98.22%. For women, the exact same drinking pattern is estimated to yield a risk that is even lower than 1% (0.6%). We generally do not believe that drinking at this level would warrant counselling and support.

When the Sheffield Alcohol Policy Model is applied to Australian data, the lifetime risk for the same drinking pattern is 2.1% for men and 4.1% for women (Angus et al., 2019, p. 5). While the figures thus are similar for men in the UK and Australia, in relative terms they differ notably for women. Women with this drinking pattern in Australia have 6.83 times (or 583%) higher risk of dying from a specific level of alcohol consumption compared to their UK counterparts (0.041/0.006). This highlights that whether risks look large or small often depends on how they are presented and in what context. Often, relative risks appear large and absolute risks appear small. An infamous example is the “Pill scare” in the UK during the 1990s, where estimates showed that women using third-generation contraceptive pills had a 100% higher risk of thrombosis (blood cloths) compared to those using second-generation pills (e.g., Bodemer et al., 2014). When looking at the absolute risks, however, the picture changes. The absolute risk of the former was 0.028% (2 in 7000 women) and for the latter it was 0.014% (1 in 7000 women) (Bodemer et al., 2014). These figures are extreme but they illustrate the confidence tricks that can be played by presenting risks in different ways.

We understand the public health logic behind low-risk drinking guidelines and the large impact reduced drinking may have at the population level. At the same time, advice is given to unique individuals who most likely care more about their own drinking patterns and well-being than about the total consumption of the population. The NBHW holds, with reference to Swedish law, that “healthcare is appointed with the task to prioritise interventions to those with the greatest needs…” (Socialstyrelsen, 2023, p. 5). We consider this statement a perfect argument against their newly updated risky drinking recommendations. Scarce resources may be better used elsewhere within healthcare. Finally, we suggest that future drinking guidelines should present absolute risks along various drinking levels, and clearly describe what the absolute risks at different levels actually mean. This would truly inform people about the statistical risks associated with their behaviour.