letter to the editor adicciones vol. xx, nº x · 2021 Considerations about population-level alcohol-attributable mortality estimates Reflexiones sobre las estimaciones de mortalidad atribuida al consumo de alcohol a nivel poblacional Sergi Trias-Llimós*,**. * Centre d’Estudis Demogràfics, Bellaterra. Spain. ** London School of Hygiene & Tropical Medicine, London. UK. P opulation level alcohol-attributable mortality within the family of attributable-fraction (AF) approach- estimates are essential metrics for public health es. The work done by Rehm and colleagues shed light on policy makers. Underlying causes of death pro- these estimates in an international perspective (Rehm et vide important insights on alcohol-attributable al., 2007), and was followed by the developments from the mortality in populations, but are highly underestimating Global Burden of Disease (Stanaway et al., 2018), and by the overall impact of alcohol on mortality (Rehm et al., other publications, including the recent update of esti- 2017; Trias-Llimós, Martikainen, Mäkelä & Janssen, 2018). mates for Spain (Donat, Sordo, Belza & Barrio, 2020). In Few causes of death are considered wholly-attributable to general terms, these approaches require two different data alcohol (e.g., alcoholic liver disease, alcohol use disorders), sources, namely: age- and sex-specific alcohol prevalence, which means that alcohol is necessary and that this death and relative risks (RR). would not have occurred in absence of alcohol use. Yet, for The combination of these different data sources may be a large group of causes alcohol is contributing to the dis- problematic as previously discussed elsewhere (Rehm, 2010; ease incidence and development, but it is not a necessary Rey & Jougla 2014). For example, the alcohol prevalence is component (e.g., ischaemic heart disease, several cancers) well known to be largely underreported in health surveys, (Rehm et al., 2017). Therefore, the impact of alcohol on and although corrections are applied the extent of the un- these diseases cannot be directly obtained from underlying derreported consumption varies across population groups cause of death data, and it is usually indirectly estimated. unpredictably, particularly regarding drinking patterns. The most popular methods to estimate population lev- Furthermore, the relative risks, retrieved from other anal- el alcohol-attributable mortality estimates are comprised yses, can also be problematic. First, they are often derived Received: September 2020; Accepted: December 2020. Send correspondence to: Sergi Trias-Llimós. Carrer Universitat Autònoma B, 08193 Bellaterra. E-mail: strias@ced.uab.cat. ADICCIONES, 2021 · VOL. xx NO. x · PAGES xx-xx Considerations about population-level alcohol-attributable mortality estimates from studies assessing disease incidence and not cause-spe- cause-specific interrelations. Furthermore, an important cific mortality. Second, RR are uncertain for certain causes number of MCOD data exists in several developed coun- and drinking categories. For example, several studies still tries, and these are gradually becoming available in other use cardioprotective RR whereas for cardiovascular diseases countries, including Spain. these effects are disputed and may be inexistent (Holmes In conclusion, AF approaches have important well- et al., 2014). Third, the RR are not often stratified by age known limitations that particularly apply to alcohol-related groups, which could play a major role, particularly at old estimates. Alternative methods using all the information ages. Furthermore, the RR for acute conditions (e.g., exter- in the death certificate offer new opportunities to partly nal causes) are known to have an inconsistent relationship overcome those limitations when estimating alcohol-re- with alcohol-related outcomes. Despite specific approaches lated deaths. Further studies should specifically assess the developed for acute causes, the estimation of the impact of strengths and limitations of MCOD approaches for estimat- alcohol on acute causes of death is complex and uncertain. ing cause-specific alcohol-attributable mortality. These main limitations have a substantial impact on age- and cause-specific estimates. For example, estimates based on AF approaches tend to potentially overestimate the im- Conflict of interests pact of alcohol on ischaemic heart disease at old ages due None. to the rising number of deaths at those ages and to the use of adult age RR for the old population. One alternative approach that can partly overcome some References main limitations of AF approaches is the one that direct- Donat, M., Sordo, L., Belza, M. J. & Barrio, G. (2020). Mor- ly estimates alcohol-related mortality using detailed data talidad atribuible al alcohol. En: Ministerio de Sanidad, from death certificates, multiple causes of death (MCOD). España (2001-2017). Metodología y Resultados. Retrieved This approach assumes that those deaths with a wholly-at- at https://pnsd.sanidad.gob.es/profesionales/pub- tributable to alcohol cause in the death certificate, either licaciones/catalogo/catalogoPNSD/publicaciones/ as underlying or as a contributory cause, are related to alco- pdf/2020_Mortalidad_atribuible_al_alcohol_en_ hol. Finnish studies were pioneers adopting this approach Espana_2001-2017.pdf. to estimate population level alcohol-attributable mortality, Holmes, M. V., Dale, C. E., Zuccolo, L., Silverwood, R. J., and this approach has been punctually adapted in other Guo, Y., Ye, Z., … on behalf of The InterAct Consortium. countries and regions (Martikainen, Mäkelä, Peltonen & (2014). Association between alcohol and cardiovascu- Myrskylä, 2014; Trias-Llimós et al., 2018). lar disease: Mendelian randomisation analysis based The comparison of AF and MCOD approaches can lead on individual participant data. BMJ 349, g4164–g4164. to interesting insights to identify strengths and limitations doi:10.1136/bmj.g4164. of both approaches. This comparison is an essential need Manthey, J. & Rehm, J. (2019). Mortality from Alcoholic for most populations, but the scarce related research sug- Cardiomyopathy: Exploring the gap between estimat- gested that AF approaches largely overestimate alcohol-at- ed and civil registry data. Journal of Clinical Medicine, 8, tributable mortality at older ages (Trias-Llimós et al., 2018), 1137. doi:10.3390/jcm8081137. particularly for cardiovascular causes (Manthey & Rehm, Martikainen, P., Mäkelä, P., Peltonen, R. & Myrskylä, M. 2019). Furthermore, alcohol-related estimates derived (2014). Income differences in life expectancy: The from MCOD approaches seem to account for a substantial changing contribution of harmful consumption of alco- share of deaths at relatively young ages in Finland, which hol and smoking. Epidemiology, 25, 182–190. doi:10.1097/ seems to indicate that this approach can capture the often EDE.0000000000000064. challenging to estimate, external causes of alcohol-related Rehm, J., Sulkowska, U., Manczuk, M., Boffetta, P., Powles, deaths (Martikainen et al., 2014). J. Popova, S. & Zatonski, W. (2007). Alcohol accounts Using all information in the death certificate to esti- for a high proportion of premature mortality in Central mate alcohol-attributable mortality (MCOD approaches) and Eastern Europe. International Journal of Epidemiology, requires facing challenges related to mortality coding 36, 458–467. doi:10.1093/ije/dyl294. practices, but also offers new opportunities for making Rehm, J. (2010). Commentary on Rey et al. (2010): How to comparisons and further improving estimates. The main improve estimates on alcohol-attributable burden? Ad- strength of this approach is that it does not require any diction, 105, 1030–1031. indirect estimation, and thus the uncertainty and potential Rehm, J., Gmel, G. E., Gmel, G., Hasan, O. S. M., Imtiaz, biases in the AF-related estimates do not apply in that case. S., Popova, S., … Shuper, P. A. (2017). The relationship MCOD approaches can offer new insights into mortality re- between different dimensions of alcohol use and the lated with alcohol at the age groups in which AF approach- burden of disease - an update: Alcohol and disease. Ad- es are more problematic, as well as shed new light in the diction, 112, 968–1001. doi:10.1111/add.13757. ADICCIONES, 2021 · VOL. xx NO. x Sergi Trias-Llimós Rey, G. & Jougla, E. (2014). Are alcohol-attributable mor- tality estimates reliable? The European Journal of Public Health, 24, 3–4. doi:10.1093/eurpub/ckt114. Stanaway, J., Ashkan Afshin, D., Gakidou, E., Lim, S. S., Abate, D. Hassen Abate, K., … Murray, C. (2018). Glob- al, regional, and national comparative risk assessment of 84 behavioural, environmental and occupational, and metabolic risks or clusters of risks for 195 countries and territories, 1990–2017: A systematic analysis for the Global Burden of Disease study 2017. Lancet, 392, 1923– 1994. doi:10.1016/S0140-6736(18)32225-6. Trias-Llimós, S., Martikainen, P., Mäkelä, P. & Janssen, F. (2018). Comparison of different approaches for esti- mating age-specific alcohol-attributable mortality: The cases of France and Finland. PLoS One, 13, e0194478. doi:10.1371/journal.pone.0194478. ADICCIONES, 2021 · VOL. xx NO. x