IHME modeling on alcohol

Posted on 2022-07-19 by Karl Pettersson. Tags: epidemiology

Last week, a new IHME-coordinated Global Burden of Disease study was published, on health effects of alcohol, stratified by amount consumed, geography, age and sex (GBD 2020 Alcohol Collaborators 2022). The study models disability adjusted life-years (DALY), based on earlier studies on the effects of alcohol on different health outcomes, combined with GBD modeling of the prevalence of those outcomes among women and men in different parts of the world. From this, they have modeled a theoretical minimum risk exposure level (TMREL), i.e. the level of consumption where health risks are at their lowest, as well as a non-drinker equivalence (NDE), i.e. the level where health risks are the same as for people not consuming any alcohol. The latter measure may be interesting when TMREL>0, which holds for some age groups in different regions, because the modeling includes studies where some non-zero levels of alcohol consumption are correlated with lower morbidity or mortality from certain vascular causes, such as ischemic heart disease, stroke, and type 2 diabetes. With increasing age, these conditions increase relative to other negative health outcomes, where alcohol just increases risk, which means that both TMREL and NDE tend to be higher in older age groups.

There has been quite a lot of criticism against the study, not least the presentation of the results in press release and media (Angus 2022). When age groups below 40 have TMREL=0, this has been framed as alcohol being much more dangerous for young people than for old people, and that young people should not drink at all. But health effects are of course not the only relevant consideration when people decide whether or not they should drink alcohol, and the study does not offer any guidance when it comes to absolute risk increase for different consumption levels in different age groups. As Angus points out, also mortality in alcohol-related causes increases a lot with age, albeit not as fast as e.g. ischemic heart disease. Moreover, the protective effects of alcohol, seen in different observational studies, have often been questioned. If these effects are not real, or if people drink in excess of the levels where they could excess, a certain level of alcohol consumption will confer much larger net risk for bad health outcomes for older people, compared to younger.

A striking result in the study is that TMREL and NDE varies a lot between different regions, also for the same age and sex. Males in Eastern Europe aged ≥80 have highest TMREL of all age-sex-region combinations in the study, with 1.9 standard drinks (where 1 standard drink is 10 g alcohol) per day. For the corresponding group of females, TMREL is 1.7 drinks, but for e.g. women and men aged ≥80 in Western Europe, it is only 0.6 and 0.7 drinks. Similarly to the TMREL increase with age, this is a consequence of higher background risk of ischemic heart disease and stroke in Eastern Europe, relative to outcome where alcohol is assumed to have only negative effects. To do such modeling in a sane way, you must have reliable region- and age-specific data on both background risks for different health outcomes and how these are affected by alcohol.

As I have written about, e.g. in my 27 March post, comparing causes of death between different countries and time periods is hard, especially in advanced age. Fig. 1 and fig. 2 show life expectancy at birth in relation to the share of deaths caused by circulatory diseases for females and males the period 2014–19, for countries with such data available from WHO (2022) and University of California, Berkeley and Max Planck Institute for Demographic Research (2022).1

Figure 1: Circulatory deaths vs life expectancy females 2014–19.
Figure 2: Circulatory deaths vs life expectancy males 2014–19.

There is a rather clear clustering of countries, where one cluster has high proportion of circulatory deaths (often 50–60 percent among females, and somewhat lower among males), and another has relatively low proportion (often about 30 percent). The countries in the former group are often Eastern European countries, and tend to have shorter life expectancy than the countries in the latter group, but not always (e.g. female life expectancy is higher in Estonia than in countries such as the US and Scotland). There is no clear correlation between life expectancy and proportion of circulatory deaths within the clusters. For mortality in most low- and middle-income countries, as well as health outcomes besides mortality (included in the YLD part in DALY), available data are even more sparse or inconsistent.

For Eastern Europe, and especially Russia, periods with increased alcohol consumption have often been followed by increased total and circulatory mortality, e.g. after the collapse of the Soviet Union. Leon et al. (2010) could not find any correlation between high alcohol consumption and mortality from myocardial infarction and Russia, but a strong positive correlation with mortality from other types of ischemic heart disease. Similarly, Wood et al. (2018) have shown a negative correlation between alcohol consumption and non-fatal myocardial infarction, but a positive correlation with mortality from other ischemic heart disease. At the same time, Timonin et al. (2021) have shown that the distribution of ischemic heart disease deaths is different in Russia than in other countries, in that a lower proportion of these deaths in Russia are ascribed to myocardial infarction (e.g. just 12 percent in Russia 2005–17, compared to 63 percent in Norway 2005–16). From what I can see, the GBD modeling has not taken such factors into account, even though it is clear that they may be relevant for TMREL in regions such as Eastern Europe.

References

Angus, Colin. 2022. “Should we have lower drinking guidelines for younger people?” https://www.ias.org.uk/2022/07/15/should-we-have-lower-drinking-guidelines-for-younger-people/.
GBD 2020 Alcohol Collaborators. 2022. “Population-level risks of alcohol consumption by amount, geography, age, sex, and year: A systematic analysis for the global burden of disease study 2020.” The Lancet 400 (10347) (16 July): 185–235. doi:10.1016/S0140-6736(22)00847-9.
Leon, David A, Vladimir M Shkolnikov, Martin McKee, Nikolay Kiryanov and Evgueny Andreev. 2010. Alcohol increases circulatory disease mortality in Russia: acute and chronic effects or misattribution of cause? International Journal of Epidemiology 39 (5): 1279–1290. doi:10.1093/ije/dyq102.
Timonin, Sergey, Vladimir M Shkolnikov, Evgeny Andreev, Per Magnus and David A Leon. 2021. Evidence of large systematic differences between countries in assigning ischaemic heart disease deaths to myocardial infarction: the contrasting examples of Russia and Norway.” International Journal of Epidemiology 50 (6): 2082–2090. doi:10.1093/ije/dyab188.
University of California, Berkeley and Max Planck Institute for Demographic Research. 2022. Human Mortality Database.” https://www.mortality.org.
WHO. 2022. “WHO Mortality Database.” https://www.who.int/data/data-collection-tools/who-mortality-database.
Wood, Angela M., Stephen Kaptoge, Adam S. Butterworth, Peter Willeit, Samantha Warnakula, Thomas Bolton, Ellie Paige, et al. 2018. “Risk thresholds for alcohol consumption: Combined analysis of individual-participant data for 599 912 current drinkers in 83 prospective studies.” The Lancet 391 (10129): 1513–1523. doi:10.1016/S0140-6736(18)30134-X.

  1. The charts can be generated by cloning the blog repository, installing MortIntl with the relevant data files, as described in the documentation, and running circall_e0_1419.jl in the subdirectory postdata/2022-07-19-ihme.↩︎