original adicciones vol. xx, nº x · 2021 Abstract Resumen Methodology used to estimate alcohol- attributable mortality in Spain, 2001-2017 Metodología utilizada para estimar la mortalidad atribuible a alcohol en España, 2001-2017 Marta Donat*,**, Luis Sordo**,***, Juan Miguel Guerras**,****, Julieta Politi****, José Pulido**,***, Gregorio Barrio *,**. * Escuela Nacional de Sanidad. Instituto de Salud Carlos III, Madrid. ** Centro de Investigación Biomédica en Red de Epidemiología y Salud Pública (CIBERESP). *** Departamento de Salud Pública y Materno-infantil, Facultad de Medicina, Universidad Complutense de Madrid. **** Centro Nacional de Epidemiología. Instituto de Salud Carlos III, Madrid. Received: August 2020; Accepted: February 2021. Send correspondence to: Luis Sordo. Departamento de Salud Pública y Materno-Infantil. Facultad de Medicina. Universidad Complutense de Madrid. E-mail: lsordo@ucm.es El objetivo es describir y discutir los métodos y asunciones para esti- mar la mortalidad atribuible a alcohol en España en 2001-2017. Se estimó el nº medio anual de muertes atribuibles a alcohol (MAAs) basándose en 19 grupos de causas de muerte relacionadas con alco- hol (18 parcialmente atribuibles y uno directamente atribuible), y 20 fracciones atribuibles poblacionales al alcohol (FAPs) para cada gru- po de causas, resultantes de combinar sexo, 5 grupos de edad, y los períodos 2001-2009 y 2010-2017. Las muertes por causa se obtuvieron del Instituto Nacional de Estadística. Para las causas parcialmente atri- buibles se calcularon FAPs específicas para España, usando la fórmula de Levin con datos de exposición al alcohol procedentes de encuestas de salud y estadísticas de ventas, y riesgos relativos procedentes de metanálisis internacionales. Se consideraron las prevalencias anuales de exbebedores y de siete niveles de consumo diario de alcohol. Se corrigió la subestimación del consumo medio diario autoinformado con respecto a las estadísticas de venta, multiplicando por un factor de 1,58-3,18, dependiendo del año-calendario. Se calcularon tasas de MAA y porcentajes de la mortalidad general atribuibles a alcohol es- tandarizados por edad, según sexo, grupo de edad, periodo-calenda- rio, tipo de bebedor y comunidad autónoma. Se realizaron análisis de sensibilidad observando cómo cambiaban las estimaciones de MAA al hacerlo algunas opciones metodológicas, como el criterio de exbebe- dor o la introducción de un período de latencia. Palabras clave: alcohol, mortalidad atribuible, España, metodología. The objective is to describe and discuss methods and assumptions to estimate the mortality attributable to alcohol in Spain in 2001-2017. The annual mean number of deaths attributable to alcohol (DAAs) was estimated based on 19 groups of alcohol-related causes of death (18 partially attributable and one directly attributable), and 20 al- cohol population-attributable fractions (PAFs), resulting from com- bining sex, 5 age groups, and the periods 2001-2009 and 2010-2017, for each cause group. Deaths from causes were obtained from the Spanish National Institute of Statistics. For partially attributable caus- es, Spain-specific PAFs were calculated using the Levin formula with alcohol exposure data from health surveys and sales statistics, and relative risks from international meta-analyses. Annual prevalences of ex-drinkers and seven levels of daily alcohol consumption were con- sidered. The underestimation of self-reported daily average consump- tion with respect to the sales statistics was corrected by multiplying by a factor of 1.58-3.18, depending on the calendar year. DAA rates stan- dardized by age and standardized proportions of general mortality attributable to alcohol, according to sex, age group, calendar period, type of drinker and autonomous community were calculated. Sensitiv- ity analyses were performed to assess how the DAA estimates changed when changing some methodological options, such as the ex-drinker criterion or the introduction of a latency period. Key words: alcohol, attributable mortality, Spain, methodology. ADICCIONES, 2021 · VOL. xx NO. x · PAGES xx-xx Methodology used to estimate alcohol-attributable mortality in Spain, 2001-2017 Alcohol use is one of the main preventable risk factors for morbidity, mortality and disability in many countries (Global Burden of Disease [GBD], 2018a; Institute for Health Metrics and Evaluation [IHME], 2020). People who drink alco- hol, especially high-risk drinkers or people with alcohol use disorder, have a much higher mortality risk than the general population (Roerecke & Rehm, 2013). However, alcohol use begins to raise the risk of death long before levels of consumption considered to be high risk are actu- ally reached, or before an alcohol use disorder develops (Stockwell et al., 2016). In estimating deaths attributable to alcohol (DAA), es- sential synthetic indicators are used to compare the impact of alcohol consumption across regions, periods and sub- groups, to determine the need for public health interven- tions, to assess them and allocate resources (World Health Organization [WHO], 2009). The World Health Organi- zation defines DAA as the algebraic sum of deaths caused and preceded by alcohol use that would not have occurred in a counterfactual scenario without historical consump- tion of this substance. In theory, they should be estimated by comparing the actual risk of death with the hypothetical risk in a counterfactual scenario without historical alcohol consumption (WHO, 2018a). However, such estimates require information from mul- tiple sources, which often have considerable gaps, so that different authors tend to make methodological choices and assumptions which do not always match, and/or turn to data that are not always possible to extrapolate from al- cohol use in other times or regions. Among other aspects, methodological choices may be made based on the in- clusion of alcohol-related causes of death, studies or me- ta-analyses providing risk functions relating the amount of alcohol consumed to mortality from each cause, whether or not risk of DAA is considered in ex-drinkers or that as- sociated with binge drinking and the definitions of these terms, the number of subgroups for which the alcohol-re- lated population attributable fractions (PAFs) used in the calculations are obtained, whether or not a latency period is considered between alcohol use and death, and whether or not the underestimation of average self-reported drink- ing in the surveys is corrected with respect to more valid sources. For these reasons, quite dissimilar estimates are some- times published for the same country, producing a great deal of uncertainty when the results are used in deci- sion-making. Spain is also affected by this situation, with estimates by authors or national or international insti- tutions of the annual number of DAA in the population aged 15 years and over ranging from 8,558 in 1999-2004 (Fierro, Ochoa, Yánez, Valderrama & Álvarez, 2008) to 37,000 in 2016 (GBD, 2018b), and the proportion of gen- eral mortality attributable to alcohol from 2.1% to 9.0%, respectively. However, the lowest estimate was only partial- ly based on empirical consumption data obtained from Spanish surveys, and these data were not corrected for underestimation of self-reported average consumption in population-based surveys (Fierro et al., 2008). Regarding estimates by foreign or international institutions (GBD, 2018a; IHME, 2020; WHO, 2020), these use internation- al data sources, so their data are based on the population distribution of average consumption and alcohol con- sumption patterns non-specific to Spain. In addition, some methodological details are opaque or difficult to compre- hend, which makes it impossible to apply them to obtain different results from those published, for example, with reference to different subgroups of interest. It therefore seems necessary that a country like Spain should have detailed procedures to allow the estimation of DAA based on data on the population distribution of alcohol consumption collected directly from the Spanish population. This would allow routine estimates of the risk of DAA to be made rapidly, including series on changes over time and inter-regional comparisons (for example, between autonomous communities) and between socio- demographic subgroups. Beyond the global estimates that may exist, many countries, especially those with an An- glo-Saxon tradition, have their own methodologies of this type for making such routine estimates. The objective of this article is thus to describe and dis- cuss a methodology specific to Spain with which to estimate its alcohol-attributable mortality, which will be applied in a later article to obtain estimates during the period 2001- 2017. Description of methodology General methodological approach Deaths caused and preceded by alcohol use that would not have occurred in a counterfactual scenario without his- torical use of the substance are considered DAA. To esti- mate the total number of DAA, the specific cause approach is used. For this purpose, a series of causes or groups of causes considered to be related to alcohol use are select- ed, deaths attributable to alcohol for each of these causes (DAAc) are estimated and the results of all those selected are added together. DAA =∑DAAc The number of DAAc is estimated by multiplying the number of deaths from this cause (Nc), extracted from mortality statistics, by its corresponding PAF (PAFc), which is the proportion of deaths from this cause attributable to exposure to alcohol and which could be avoided if the pop- ulation stopped consuming this substance completely, and expresses the proportional contribution of alcohol use to population mortality from this cause. The PAF is obtained through an algorithm incorporating the relative risks (RR) ADICCIONES, 2021 · VOL. xx NO. x Marta Donat, Luis Sordo, Juan Miguel Guerras, Julieta Politi, José Pulido, Gregorio Barrio with respect to abstainers and the population prevalences of different categories of alcohol use. DAAc = (Nc) (PAFc) For each cause, the process is stratified for different population subgroups of sex, age and calendar-period, with the aim of achieving a balance that allows PAFc to be obtained which are specific enough to give greater valid- ity to the results while maintaining a reasonable level of precision. Specifically, for each cause, PAFc are obtained for the 20 subgroups based on combining both sexes, five age groups (15-24, 25-44, 45-64, 65-74 and ≥75) and two calendar periods (2001-2009 and 2010-2017). This means that independent estimates are obtained for each of the 20 subgroups and subsequently summed. The estimation applies to the population aged 15 years and over. Thus, the total number of deaths attributable to alcohol is the sum of the deaths attributed to different causes. The process therefore involves: (1) identifying and quantifying the different related causes of death, (2) establishing an attributable fraction for each cause and population sub- group, and (3) adding the results for absolute totals and rates. Causes of death related to alcohol Two types of alcohol-related causes of death are select- ed: 1) causes directly or completely attributable to alco- hol, such as alcohol use disorder, where alcohol is always considered a necessary cause, and 2) other causes of alco- hol-related death, in which alcohol is a contributing but not the sole factor; these comprise 18 groups, for example oesophageal cancer. The selection of causes is based on the most recent reviews and meta-analyses focused on assess- ing the risk of developing or dying from certain diseases associated with alcohol use (Corrao, Bagnardi, Zambon & Arico, 1999; Rehm et al., 2017; Samokhvalov, Irving & Rehm, 2010; Sherk, Stockwell, Rehm, Dorocicz & Shield, 2017), and in the selections made in the estimates of other countries (Connor, Kydd, Rehm & Shield, 2013; Jones & Bellis, 2013; Marmet, Rehm, Gmel, Frick & Gmel, 2014; Rey, Boniol & Jougla, 2010). Table 1 shows the codes of the International Classification of Diseases, tenth edition (ICD-10), corresponding to the 19 groups of selected caus- es. To avoid duplication, the codes corresponding to caus- es which are part of broader categories already included in a group of partially attributable causes are excluded from the list of codes for the group of causes directly attributable to alcohol. This is the case of alcoholic liver disease (K70), already included in cirrhosis/chronic liver disease, and in- voluntary alcohol poisoning (X45) or intentional alcohol self-poisoning (X65), included as external causes, among others (Table 1). Not included as partially attributable causes are a broad group of diseases or health problems probably related to alcohol, such as certain cancers (for example, stomach and pancreatic cancers) since the evidence is considered insuf- Table 1. Groups of causes of death selected to estimate alcohol-attributable mortality in Spain 2001-2017. Groups of causes ICD-103 codes Causes partially attributable to alcohol 1. Tuberculosis A15-A19, B90, K67.3, P37.0 2. Lower respiratory infection/pneumonia A48.1, A70, J09-J15.8, J16, J20-J21, P23.0-P23.4 3. Cancer of the mouth and pharynx1 C00-C13 4. Cancer of the esophagus1 C15 5. Colorectal cancer1 C18-C21 6. Liver cancer1 C22 7. Laryngeal cancer1 C32 8. Breast cancer (women)1 C50 9. Diabetes mellitus E10.0-E10.1, E10.3-E11.1, E11.3-E12.1, E12.3-E13.1, E13.3-E14.1, E14.3-E14.9, P70.0-P70.2, R73 10. Epilepsy G40, G41 11. Hypertensive heart disease I11 12. Ischemic heart disease I20-I25 13. Atrial fibrillation/flutter I48 14. Ischemic stroke G45, I63, I67.2-I67.3, I67.5-I67.6, I69.3 15. Non-ischemic stroke I60-I62, I69.0-I69.2, I67.0-I67.1 16. Cirrhosis/chronic liver disease B18, I85, I98.2,K70, K71.3–K71.5, K71.7, K72.1–K74.6, K75.8–K76.0, K76.6–K76.7, K76.9 17. Pancreatitis K85-K86 18. External cause V01-Y98 19. Causes directly attributable to alcohol2 E24.4, F10, G31.2, G62.1, G72.1, I42.6, K29.2, O35.4, R78.0 Note. 1 D codes (carcinoma in situ, benign tumour, and tumours of uncertain or unknown behaviour) were not included because a fourth digit is often not available. 2 Causes of death directly attributable to alcohol part of broader categories included in groups of partially attributable causes were not considered, including alcoholic liver disease (K70), involuntary alcohol poisoning (X45), intentional alcohol self-poisoning (X65), intoxication by alcohol with undetermined intention (Y15) and evidence of involvement of alcohol (Y90-Y91). 3 ICD-10: International Statistical Classification of Diseases and Related Health Problems, tenth review. ADICCIONES, 2021 · VOL. xx NO. x Methodology used to estimate alcohol-attributable mortality in Spain, 2001-2017 ficient to establish the causality of alcohol or to quantify the relative risk (Rehm et al., 2017; WHO, 2018b). HIV disease is excluded, although it is included in some coun- tries (Gmel, Shield & Rehm, 2011; Marmet, Rehm & Gmel, 2016). The number of deaths from the selected causes is sourced from the mortality register of the Spanish National Institute of Statistics [INE] (INE, 2020). Population fractions attributable to alcohol The population fraction attributable to alcohol for a cause or group of causes of death (PAFc) expresses the relative contribution of alcohol use to total mortality from this cause. In this way, an attributable fraction of one was applied (PAFc=1) for causes of death directly attributable to alcohol, while for causes partially attributable to alcohol, specific PAFc were calculated for Spain. For the calculation, the Levin formula for categorical data was used, combin- ing prevalence data of different levels of exposure to alco- hol with the relative risks of mortality from a given cause (Rehm, Klotsche & Patra, 2007), and to which a term to include the fraction corresponding to ex-drinkers has been added. The formula is shown below: In this formula, Pexd is the prevalence of ex-drinkers, Pi the prevalence for category i of the average amount of alcohol consumed daily during the last 12 months in the population, RRexd is the relative risk of ex-drinkers versus unexposed individuals (abstainers), RRi the relative risk of the exposed versus abstainers for this cause of death, and category i the amount of alcohol consumed daily. The number of categories i used in the calculations was seven, corresponding to the intervals ≤19, 20-39, 40-49, 50-59, 60- 79, 80-99 and ≥100 grammes of pure alcohol/day. For each of the 18 groups of causes of death partially attributable to alcohol, 20 PAFs were calculated, the result of combining sex, five age groups, and the periods 2001-2009 and 2010- 2017. It was decided to calculate PAFs for two multi-year periods rather than doing so for each calendar year, which would have provided more valid estimates, because the an- nual consumption prevalences included in the calculation formula would have been subject to greater variability due to randomness, especially considering that it would have been necessary to estimate them simultaneously by age stratum, sex and amount of alcohol consumed. The calcu- lated PAFs are shown in Table 2. The above formula makes it possible to segment the numerator in order to study different groups of drinkers (Sherk et al., 2017), for example, ex-drinkers and two groups of current drinkers, with high risk and medium-low risk, defined by certain values of i . Ex-drinkers are those who have not consumed alcoholic beverages in the last year, but have drunk them at least 12 times in some year of their life, following the definition of the United States National Health Survey (Villarroel, Clarke & Schoenborn, 2016). People who have drunk alcohol less than 12 times in any year of their life and have not drunk in the last year are abstainers. Current high-risk drinkers are men (or women) who have consumed an average of ≥60 (or 40) g of pure al- cohol/day during the last year, which corresponds to 6 (or 4) standard drink units (SDUs). This criterion was adopt- ed following earlier studies (Marmet et al., 2014; Rehm et al., 2017; Rehm, Rehm, Shield, Gmel & Gual, 2013; Rehm, Shield, Rehm, Gmel & Frick, 2012), including high and very high-risk drinkers according to classifications by the WHO (WHO, 2000) and the European Medicines Agen- cy [EMA] (EMA, 2010; Mann, Aubin & Witkiewitz, 2017). Other people who have drunk alcohol in the last year are considered current low-risk drinkers. When distributing DAA by type of drinker, it is assumed that all deaths di- rectly attributable to alcohol (PAFc = 1) occur in current high-risk drinkers and that there are no deaths from exter- nal causes attributable to alcohol among former regular drinkers. The meaning of the different parameters in the Levin formula used to calculate the PAFs and the way in which they are obtained is detailed below. Relative risks for different categories of population exposure to alcohol Relative risk (RR) is a measure of association strength between alcohol exposure and death from a certain cause in a given group, place, and time compared to another group. The RRi used to calculate the PAFc can be consult- ed in Table 3. As reference or counterfactual scenario (RR = 1) they have abstainers, and for current drinkers they are calculated using continuous RR functions from different international meta-analyses, most of which were includ- ed in a recent review by Rehm et al. (Corrao et al., 1999; Rehm et al., 2017; Samokhvalov et al., 2010). As a representative RRi for each exposure interval (average daily amount of alcohol consumed) for current drinkers of < 100 grams of pure alcohol/day, the figure i corresponding to the midpoint of each interval is used (10, 30, 45, 55, 70 and 90). For current drinkers of ≥ 100 grams of alcohol/day, however, 130 grams of pure alcohol/day is used as representative RRi , this being the median amount of alcohol consumed by the drinkers of that group includ- ed in the population health survey samples in both periods 2001-09 and 2010-17. In addition, in the case of some caus- es of death, sex-specific RRs were used (diabetes mellitus, hypertensive heart disease, ischemic heart disease, isch- emic stroke, non-ischemic stroke, cirrhosis/chronic liver disease, and pancreatitis). ADICCIONES, 2021 · VOL. xx NO. x Marta Donat, Luis Sordo, Juan Miguel Guerras, Julieta Politi, José Pulido, Gregorio Barrio In the case of ex-drinkers, there are no published RRs for all the selected causes of death. For specific causes with available information, RRs are taken from a 2010 meta-anal- ysis (Rehm et al., 2010a), and for those without informa- tion, the all-cause mortality RR of 1.38 from a Stockwell et al. meta-analysis is applied, corresponding to the joint and completely adjusted model (Stockwell et al., 2016). Table 2. Population alcohol-attributable fractions for the selected causes, by sex, period and age. Spain, 2001-2017. Men 2001-2009 2010-2017 15-24 25-44 45-64 65-74 ≥75 15-24 25-44 45-64 65-74 ≥75 Cause of Death 1. Tuberculosis 0.427 0.563 0.628 0.576 0.511 0.341 0.440 0.573 0.601 0.474 2. Lower respiratory infection/pneumonia 0.111 0.157 0.188 0.171 0.159 0.084 0.119 0.167 0.185 0.155 3. Cancer of the mouth and pharynx1 0.506 0.633 0.691 0.644 0.581 0.414 0.518 0.642 0.668 0.549 4. Cancer of the esophagus1 0.529 0.629 0.677 0.635 0.579 0.447 0.540 0.638 0.658 0.558 5. Colorectal cancer1 0.139 0.200 0.239 0.215 0.195 0.105 0.149 0.210 0.231 0.186 6. Liver cancer1 0.369 0.586 0.670 0.607 0.528 0.286 0.394 0.592 0.631 0.451 7. Laryngeal cancer1 0.483 0.497 0.501 0.493 0.460 0.240 0.318 0.418 0.443 0.348 8. Breast cancer (women)1 - - - - - - - - - - 9. Diabetes mellitus -0.056 -0.042 -0.021 -0.018 -0.008 -0.060 -0.058 -0.033 -0.016 -0.011 10. Epilepsy 0.290 0.396 0.455 0.409 0.357 0.225 0.302 0.407 0.434 0.335 11. Hypertensive heart disease 0.210 0.292 0.342 0.305 0.269 0.161 0.221 0.303 0.327 0.254 12. Ischemic heart disease -0.061 -0.044 -0.026 -0.026 -0.016 -0.049 -0.058 -0.036 -0.017 -0.016 13. Atrial fibrillation/flutter 0.133 0.188 0.224 0.201 0.184 0.101 0.142 0.198 0.218 0.177 14. Ischemic stroke -0.015 0.040 0.090 0.076 0.074 -0.045 -0.015 0.056 0.087 0.065 15. Non-ischemic stroke 0.157 0.222 0.263 0.234 0.208 0.119 0.166 0.232 0.252 0.198 16. Cirrhosis/chronic liver disease 0.666 0.799 0.844 0.808 0.755 0.576 0.680 0.803 0.824 0.711 17. Pancreatitis 0.411 0.545 0.610 0.558 0.494 0.327 0.424 0.555 0.584 0.459 18. External cause 0.191 0.256 0.297 0.267 0.239 0.144 0.200 0.267 0.288 0.233 Women 2001-2009 2010-2017 15-24 25-44 45-64 65-74 ≥75 15-24 25-44 45-64 65-74 ≥75 Cause of Death 1. Tuberculosis 0.245 0.251 0.246 0.197 0.144 0.220 0.222 0.257 0.219 0.158 2. Lower respiratory infection/pneumonia 0.063 0.070 0.068 0.057 0.051 0.059 0.066 0.073 0.066 0.056 3. Cancer of the mouth and pharynx1 0.309 0.315 0.311 0.250 0.186 0.282 0.283 0.324 0.279 0.204 4. Cancer of the esophagus1 0.360 0.364 0.355 0.286 0.222 0.339 0.342 0.373 0.325 0.244 5. Colorectal cancer1 0.077 0.085 0.083 0.069 0.059 0.072 0.079 0.088 0.079 0.065 6. Liver cancer1 0.156 0.163 0.161 0.140 0.079 0.117 0.110 0.152 0.123 0.079 7. Laryngeal cancer1 0.425 0.427 0.418 0.375 0.347 0.163 0.169 0.190 0.163 0.121 8. Breast cancer (women)1 0.132 0.138 0.135 0.107 0.085 0.121 0.127 0.144 0.124 0.094 9. Diabetes mellitus -0.243 -0.222 -0.177 -0.107 -0.070 -0.241 -0.230 -0.202 -0.136 -0.081 10. Epilepsy 0.163 0.170 0.166 0.131 0.101 0.149 0.154 0.175 0.150 0.111 11. Hypertensive heart disease 0.191 0.205 0.219 0.185 0.122 0.143 0.143 0.219 0.192 0.132 12. Ischemic heart disease -0.052 -0.032 -0.014 0.009 0.020 -0.059 -0.042 -0.019 0.004 0.020 13. Atrial fibrillation/flutter 0.075 0.082 0.080 0.066 0.057 0.070 0.077 0.086 0.076 0.062 14. Ischemic stroke -0.154 -0.136 -0.095 -0.045 -0.044 -0.186 -0.183 -0.122 -0.081 -0.054 15. Non-ischemic stroke 0.191 0.194 0.190 0.147 0.105 0.172 0.172 0.199 0.167 0.116 16. Cirrhosis/chronic liver disease 0.430 0.434 0.429 0.363 0.249 0.380 0.370 0.433 0.373 0.264 17. Pancreatitis 0.234 0.240 0.236 0.189 0.139 0.211 0.213 0.246 0.210 0.152 18. External cause 0.107 0.114 0.112 0.089 0.074 0.100 0.107 0.120 0.105 0.082 Note. The population attributable fractions (PAFs) for the group of causes of death directly attributable to alcohol are always 1 and were not included. To calculate the PAF, the corrected prevalences of consumption were used, raising the average consumption of each participant in the surveys to 80% of the average per capita consumption estimated from the sales statistics. For a more accurate calculation of deaths attributable to alcohol for a given territory or subgroup, PAFs with 5 decimal places are often required, so it is recommended to request the file from the authors. ADICCIONES, 2021 · VOL. xx NO. x Methodology used to estimate alcohol-attributable mortality in Spain, 2001-2017 Prevalences of population exposure to alcohol The formula for calculating PAFs involves the prev- alence of ex-drinkers and the annual prevalences of dif- ferent average daily amounts of alcohol consumed. These prevalences are shown in Table 4 by age, sex, and period. Data sources and how they were obtained are described below. Ex-drinker prevalences For the period 2010-2017, these are estimated from fig- ures in the Spanish National Health Survey (ENS) and the European Health Survey in Spain (EESE) corresponding to the period. As the studies providing RRs for mortality in ex-drinkers most likely exclude infrequent ex-drinkers or those just trying out alcohol, it is advisable to correct the prevalences by eliminating this subgroup. To make this correction, the distribution of infrequent and regular ex-drinkers from the United States National Health Survey in the period 2011-14 (Villarroel et al., 2016) is used since there are no data for this from Spanish sources. For the period 2001-09, the prevalence figures for ex-drinkers obtained from ENS and EESE are very low compared to 2010-17 and to those obtained from other sources for some age subgroups and were thus not con- sidered reliable; instead, they were estimated from the 2010-17 corrected prevalences, assuming a relative change between periods similar to that observed in the Spanish Household Survey on Alcohol and Drugs (Encuesta Domi- ciliaria sobre Alcohol y Drogas en España, EDADES) (Del- egación del Gobierno para el Plan Nacional sobre Drogas [DGPNSD], 2018). Annual prevalences of average daily consumption of different amounts of alcohol These are obtained from the individual files of ENS 2001, 2006, 2011 and 2017, and EESE 2009 and 2014 (INE, 2019a; Ministerio de Sanidad, Consumo y Bienestar Social [MSCBS], 2019). To this end, the average self-reported dai- ly amount of alcohol consumed by each participant during the last year is first corrected for underestimation, given the well-known fact that self-reports of alcohol use strong- ly underestimate real consumption (Sordo et al., 2016). The categories of consumption used for stratifying annual consumption prevalence, in grams of pure alcohol, corre- spond to the seven categories i mentioned in the PAFc cal- culation. The prevalences for the intermediate years with- out the survey are estimated by linear interpolation. The algorithm used to correct for underestimation is detailed below. For each participant in the survey, average daily consumption in a given year in grams of pure alcohol (Ac), is obtained by multiplying the self-reported average daily consumption during the 12 months prior to the sur- vey (As), by an elevation factor (Ef). This factor is calculat- ed in turn by dividing the best estimate of the average daily Table 3. Relative risks of death from the selected causes, by sex and level of exposure to alcohol. Spain, 2001-2017. Men Women Average amount of alcohol consumed (g pure alcohol/day) Exd1. 10 30 45 55 70 90 130 Exd. 10 30 45 55 70 90 130 Cause of Death 1. Tuberculosis 1.38 1.20 1.71 2.24 2.69 3.52 5.04 10.34 1.38 1.20 1.71 2.24 2.69 3.52 5.04 10.34 2. Lower respiratory infection/pneumonia 1.38 1.05 1.15 1.24 1.30 1.40 1.54 1.86 1.38 1.05 1.15 1.24 1.30 1.40 1.54 1.86 3. Cancer of the mouth and pharynx1 1.38 1.28 2.03 2.81 3.45 4.65 6.70 12.68 1.38 1.28 2.03 2.81 3.45 4.65 6.70 12.68 4. Cancer of the esophagus1 1.38 1.46 2.39 3.21 3.81 4.80 6.29 9.76 1.38 1.46 2.39 3.21 3.81 4.80 6.29 9.76 5. Colorectal cancer1 1.38 1.06 1.21 1.23 1.41 1.55 1.76 2.26 1.38 1.06 1.21 1.23 1.41 1.55 1.76 2.26 6. Liver cancer1 1.38 1.01 1.16 1.40 1.66 2.29 3.94 17.55 1.38 1.01 1.16 1.40 1.66 2.29 3.94 17.55 7. Laryngeal cancer1 1.38 1.16 1.52 1.85 2.10 2.52 3.17 4.77 1.38 1.16 1.52 1.85 2.10 2.52 3.17 4.77 8. Breast cancer (women)1 - - - - - - - - 1.38 1.11 1.36 1.58 1.75 2.04 2.50 3.76 9. Diabetes mellitus 1.18 0.90 0.88 0.93 0.97 1.07 1.16 1.16 1.14 0.68 0.62 0.86 1.18 1.18 1.18 1.18 10. Epilepsy 1.38 1.14 1.45 1.75 1.98 2.38 3.04 4.97 1.38 1.14 1.45 1.75 1.98 2.38 3.04 4.97 11. Hypertensive heart disease 1.38 1.10 1.31 1.50 1.65 1.89 2.26 3.25 1.38 0.89 1.60 2.43 3.16 4.59 7.38 17.76 12. Ischemic heart disease 1.25 0.95 0.84 0.74 0.67 1.00 1.00 1.43 1.54 0.84 0.99 1.15 1.26 1.45 1.74 2.53 13. Atrial fibrillation/flutter 1.38 1.06 1.19 1.30 1.37 1.50 1.68 2.11 1.38 1.06 1.19 1.30 1.37 1.50 1.68 2.11 14. Ischemic stroke 1.33 0.85 0.96 1.07 1.14 1.26 1.43 1.81 1.15 0.65 0.77 1.00 1.22 1.70 2.75 7.81 15. Non-ischemic stroke 1.33 1.07 1.23 1.36 1.46 1.62 1.86 2.45 1.15 1.16 1.55 1.93 2.24 2.79 3.74 6.73 16. Cirrhosis/chronic liver disease 1.31 1.33 2.32 3.53 4.67 7.10 12.42 37.95 6.50 2.82 5.97 8.90 11.20 15.24 21.93 40.87 17. Pancreatitis 1.38 1.19 1.68 2.18 2.60 3.37 4.76 9.53 1.38 0.77 0.66 1.53 3.01 4.80 10.26 10.26 18. External cause 1.38 1.09 1.30 1.48 1.62 1.85 2.21 2.42 1.38 1.09 1.30 1.48 1.62 1.85 2.21 2.42 Exd1: ex-drinkers. Note. The references used to obtain the RRs for each cause of death are Corrao, Bagnardi, Zambon and Arico (1999); Samokhvalov, Irving and Rehm (2010), and Rehm et al. (2017), in the case of current drinkers. For ex-drinkers, references are Rehm et al. (2010a), and Stockwell et al. (2016). ADICCIONES, 2021 · VOL. xx NO. x Marta Donat, Luis Sordo, Juan Miguel Guerras, Julieta Politi, José Pulido, Gregorio Barrio population consumption per capita from multiple sources, mainly sales statistics (Ar), by As and multiplying the result by the degree of correction of the desired underestimation (C). Table 4. Corrected prevalence of alcohol consumption1 in the population aged 15 years and over2, by age, sex, average daily amount of alcohol consumed and period (%). Spain, 2001-2017. Prevalence (%) Sample size Drinker status Abstainer3 Ex-drinker4 Drinker in the previous year Grams alcohol/day5 0 0 1-19 20-39 40-49 50-59 60-79 80-99 ≥100 2001-2009 Age Sex 15-24 Men 22.8 1.8 41.0 17.1 5.5 3.0 4.2 2.3 2.3 3368 Women 32.2 2.0 51.6 9.5 1.8 0.9 0.7 0.7 0.7 3365 Total 27.5 1.9 46.3 13.3 3.7 2.0 2.6 1.5 1.5 6733 25-44 Men 14.3 3.2 37.1 19.6 5.3 4.8 5.6 3.1 7.0 11240 Women 32.4 4.2 48.8 9.7 1.3 1.4 1.2 0.5 0.7 13344 Total 24.1 3.7 43.4 14.2 3.1 3.0 3.2 1.7 3.6 11111 45-64 Men 15.2 4.9 27.0 19.9 4.7 7.5 6.8 3.9 10.3 9033 Women 41.4 4.2 38.0 10.8 1.0 2.3 1.2 0.4 0.7 11800 Total 30.0 4.5 33.2 14.7 2.6 4.6 3.6 1.9 4.7 11199 65-74 Men 20.7 7.8 25.1 19.0 2.4 8.7 6.1 2.4 7.6 3614 Women 58.6 5.3 23.4 9.1 0.4 1.8 0.7 0.1 0.6 5489 Total 43.5 6.3 24.1 13.0 1.2 4.5 2.9 1.0 3.4 9103 >=75 Men 26.2 13.2 20.3 18.7 1.5 8.8 3.7 1.6 5.3 2938 Women 67.8 6.7 15.3 7.7 0.3 1.5 0.3 0.1 0.2 5104 Total 52.6 9.1 17.1 11.7 0.7 4.2 1.5 0.6 2.1 8042 Total Men 19.8 6.2 30.1 18.9 3.9 6.6 5.3 2.7 6.5 30193 Women 46.5 4.5 35.4 9.4 1.0 1.6 0.8 0.4 0.6 39102 Total 34.9 5.2 33.1 13.5 2.2 3.7 2.8 1.4 3.2 69295 2010-2017 Age Sex 15-24 Men 27.4 1.9 48.3 12.1 3.6 2.0 2.1 0.8 1.8 2195 Women 32.8 2.2 52.8 8.0 1.6 1.2 0.7 0.4 0.4 2202 Total 30.1 2.1 50.5 10.1 2.6 1.5 1.4 0.6 1.1 4397 25-44 Men 15.7 3.5 46.0 17.1 5.4 3.5 4.5 1.5 2.8 9749 Women 32.0 4.5 49.9 9.3 2.1 0.9 0.9 0.2 0.3 10312 Total 24.1 4.0 48.0 13.1 3.7 2.2 2.6 0.8 1.5 20061 45-64 Men 13.9 5.2 33.5 19.5 5.8 4.5 8.0 2.8 7.0 10753 Women 35.4 4.5 42.3 11.4 2.3 1.3 1.9 0.4 0.5 11529 Total 25.0 4.8 38.1 15.3 4.0 2.8 4.8 1.6 3.6 22282 65-74 Men 14.9 8.4 26.0 20.4 5.1 4.6 9.2 3.2 8.4 4051 Women 49.5 5.7 29.1 10.4 1.6 1.2 2.0 0.1 0.3 5148 Total 34.3 6.9 27.7 14.8 3.1 2.7 5.2 1.5 3.9 9199 >=75 Men 24.5 14.9 19.8 21.8 4.0 3.8 6.1 1.7 3.4 3597 Women 64.1 7.2 17.5 8.5 0.6 0.9 1.1 0.1 0.1 6546 Total 50.1 9.9 18.3 13.2 1.8 1.9 2.9 0.7 1.3 10143 Total Men 19.3 6.8 34.7 18.2 4.8 3.7 6.0 2.0 4.7 30345 Women 42.8 4.8 38.3 9.5 1.6 1.1 1.3 0.2 0.3 35737 Total 32.0 5.7 36.7 13.5 3.1 2.3 3.5 1.0 2.3 66082 Note. 1 Corrected prevalence of alcohol consumption: Prevalences obtained after correcting the underestimation of self-reported consumption in population surveys with respect to alcohol consumption sales statistics. The correction was made by applying an elevation factor to the average daily quantity consumed by each individual participant in the survey up to 80% of the average daily quantity per capita estimated from the sales statistics. 2Population aged 15 year and older: In reality, the consumption prevalences correspond to the population aged 16 years and over, which is the reference population of the population surveys included in this study. However, the estimates of sales statistics from international organizations usually apply to the population aged 15 years and over; the fractions attributable to alcohol in the population and the estimates of attributable mortality were thus applied to the 15-and-older population. 3Abstainer: Person who has never drunk alcoholic beverages in their life. 4Ex-drinker: Person who has not drunk alcoholic beverages in the previous year and has drunk such beverages less than 12 times in any other year of their life. The prevalence was estimated as explained in the Methodology section. 5Alcohol consumed (g/day): Refers to the average amount of pure alcohol consumed daily in grams. ADICCIONES, 2021 · VOL. xx NO. x Methodology used to estimate alcohol-attributable mortality in Spain, 2001-2017 , donde According to international recommendations (Kehoe, Gmel, Shield, Gmel & Rehm, 2012; Rehm et al., 2010b; Stockwell et al., 2018), C is 0.8; that is, As is corrected up to 80% of Ar. The justification for this procedure is some- what extensive and is included in Appendix Table 1. The estimates of As, Ar and Ef by calendar year are included in Table 5. As can be seen in this table, the Ef has increased in the most recent years, highlighting the increasing diffi- culty in capturing real consumption in surveys and further justifying the need to correct self-reported consumption. A specific Ef was calculated for each year with survey re- sults and was applied to all the participants in the survey who had drunk alcohol during the last year, irrespective of their sociodemographic profile and their consumption patterns, because the annual Ar could only be estimated for the country as a whole, without the possibility of strati- fication by sociodemographic variables. The methodology for estimating Ar and As has been published (Sordo et al., 2016) and is summarized below. Estimation of average self-reported alcohol consumption (As) The surveys used for estimating As can be considered representative of the non-institutionalized population aged ≥ 15 years in Spain. Its characteristics can be seen in Appendix Table 2. The amount of alcohol consumed (As) in litres of pure alcohol (lpa) per person-year (py) is es- timated following the classical approach (Dawson, 2003) with the algorithm: , where the subscript i rep- resents the different categories of beverages, such as wine/ cava, beer/cider, aperitifs/intermediate products (drinks with alcohol content of 1.2-22% ABV, other than fermented ones such as vermouth, sherry, port and pale sherries or amontillados) and spirits/distilled beverages (including mixes made with spirits), Di is the annual number of days or times each drink is consumed, SDi the number of SDUs con- sumed each day or each time, Vi the volume of each SDU in litres, Ci the proportion of alcohol over the total volume of the drink, and SS the effective sample size. Given the great scarcity of empirical data (Rodríguez-Martos, Gual & Llopis Llácer, 1999), the SDU volume of each drink, Vi , is assigned using the upper limits of the volumes reflected in some clinical and public health guidelines (Organización Médica Colegial y Ministerio de Sanidad y Consumo [OMC-MSC], 2006; Sociedad Española de Medicina de Familia y Comuni- taria [SEMFYC], 2005): wine (125 ml), beer/cider (250 ml), aperitifs (100 ml), spirits (60 ml), and Ci are applied to each drink on the basis of the Spanish Tax Agency guidelines (5.5%, 11.5%, 15% and 35%, respectively) (Agencia Trib- utaria [AT], 2015; 2019). In this way, the amounts in grams of pure alcohol per SDU of each drink are: 11.36 (wine), 10.86 (beer/cider), 11.85 (aperitifs), and 16.59 (spirits). For a vague category, such as “regional drinks”, it is assumed that an SDU contains 10 g of pure alcohol. Table 5. Elevation factors used for correcting underestimated self-reported annual alcohol consumption. Spain, 2001-2017. Registered alcohol consumption (Ar)1 Self-reported alcohol alcohol consumption (As)2 Elevation factor (Ef)3 2001 12.5 6.3 1.58 2002 10.8 2003 11.7 2004 12.0 2005 11.3 2006 11.4 3.8 2.38 2007 11.1 2008 10.3 2009 9.9 3.0 2.61 2010 9.6 2011 9.4 2.6 2.83 2012 9.5 2013 9.8 2014 9.6 2.9 2.64 2015 9.6 2016 9.7 2017 9.6 2.4 3.18 Note. 1 Registered alcohol consumption (Ar): Average annual per capita consumption in litres of pure alcohol/year by residents in Spain aged 15 years and over, estimated from sales statistics. 2 Self-reported alcohol consumption (As): Average annual per capita consumption of pure alcohol/year in litres by residents in Spain aged 15 years and over, self-reported in population surveys. 3 Elevation factor (Fe): Ef=(Ar/As)*0.8. The multiplication factor used to correct the underestimation of self-reported alcohol use in the population statistics with respect to the estimates from sales statistics. The correction was made only up to 80% of the sales statistics consumption estimate, as recommended (Rehm et al., 2010a; Stockwell et al., 2018; Kehoe et al., 2012). It was applied to the average consumption of each individual participating in the surveys so that the population prevalences of consumption according to the average daily amount consumed could then be obtained. ADICCIONES, 2021 · VOL. xx NO. x Marta Donat, Luis Sordo, Juan Miguel Guerras, Julieta Politi, José Pulido, Gregorio Barrio The results of the surveys were weighted to adjust for the imbalance of the sample according to sex, age group, province, size of household, and response rate. Estimated average registered alcohol consumption (Ar) The main aggregated data for estimating the average alcohol use registered per capita (Ar) refer to legal sales or supplies of drinks intended for human consumption within Spain with an alcohol content by volume (ABV) of > 1.2%. These figures are corrected to account for unreg- istered alcohol, consumption/purchase by international travellers, and post-sale alcoholic drink loss. The data for estimating the alcohol consumed/purchased by foreign visitors in Spain and Spanish visitors abroad are obtained from the Spanish Tourism Institute (TURESPAÑA, 2019), INE (INE, 2019b), WHO (registered per capita alcohol consumption by country) (WHO, 2020), Eurostat (Eu- rostat, 2019) and World Bank (World-Bank, 2019) and the scientific literature (Sordo et al., 2016). The remaining data necessary for correcting the sales statistics is obtained from the scientific literature, including the consumption of contraband beverages or those made with alcohol for other purposes, drinks from unregistered sales or pro- duction (for example, self-use), products with an alco- holic content greater than zero and ≤ 1.2% ABV, loss of alcoholic beverages (spilled, spoiled or wasted/unfinished drinks) and beverages used for cooking or for purposes other than direct human consumption (Boniface & Shel- ton, 2013; Landberg & Norstrom, 2011; Meier et al., 2013; Norstrom & Skog, 2001; Rehm et al., 2010b; Rehm et al., 2007; Trolldal, 2001; WHO, 2020). The characteristics of the main routine sources providing useful aggregate data for estimating per capita alcohol consumption in Spain are included in Appendix Table 3. Alcohol quantity is expressed in litres of pure alcohol per person-year (lpa/py), using the population resident in Spain each year as the denominator. To obtain figures for the real consumption of alcohol, a multistage process was followed starting with the availability of alcohol for con- sumption published by the Spanish Tax Agency (TA). The TA calculates availability by adding to the alcohol from legal sales of beverages subject to excise duty on alcohol (beer, spirits and snacks) the alcohol from wine purchases which are self-reported to the Spanish Food Consumption Panel (Panel de Consumo Alimentario). This availability does not include cider, and it has been noted that self-re- ported wine purchases are underestimated compared to sales (as in the case of beer, where underestimation reach- es 45%). A multi-source availability indicator was therefore created by replacing the wine component in TA availability with the Eurostat wine supply statistic and adding the FAO figure for cider supply. Finally, the real average per capi- ta consumption was obtained by adding the alcohol from drinks consumed/purchased abroad by Spanish residents and the remaining unregistered alcohol, and subtracting alcohol loss and alcohol from drinks consumed/bought in Spain by foreign visitors. The specific calculation algo- rithms can be consulted in a previous study (Sordo et al., 2016). Alcohol-attributable mortality indicators The main indicators used to express the results of alco- hol-attributable mortality estimates in Spain are: absolute number of DAA and various age-standardized indicators such as DAA rate (RDAA), and the proportional contribu- tion of alcohol to general mortality risk (CAGM), propor- tional contribution to the total risk of alcohol-attributable mortality of different causes of death (CMAAC), and from high-risk consumption (CMAAR). RDAA is an indicator of the absolute risk or probability of dying from alcohol use in a given population subgroup. Population rates are expressed per 100,000 person-years (py) and standardized using the direct method with the age structure weights of the 2013 European Standard Pop- ulation (Eurostat, 2013). The standardized proportional contributions are expressed as percentages and are the result of dividing the RDAA by the all-cause mortality rate (CAGM), the specific RDAA for each cause by the RDAA for all causes (CMAAC), and the RDAA in current high- risk drinkers by the sum of RDAAs in high-risk and medi- um-low-risk drinkers (CMAAR). Most of the indicators were calculated by sex, age group (15-34, 35-54, 55-74 years and >= 75 years), autonomous community, calendar period (2001-09 and 2010-17) and type of drinker (ex-regular drinkers, medium-low risk drinkers and high-risk drinkers). The populations by age group, sex, autonomous com- munity and calendar year for the calculation of the indi- cators were obtained from the population figures of the Spanish National Statistics Institute (INE, 2019c). To com- pare the risk of DAA between groups (for example, men and women) or periods (2001-09 and 2010-17), the ratio and difference of age-standardized rates were applied, transforming the rate ratio into a percentage of change for temporal comparisons. Sensitivity analysis Several sensitivity analyses were carried out to observe how the change of some methodological options affected the estimation of the total number of DAA in 2017 (Ap- pendix Table 4). These included introducing prevalenc- es without correction for underestimated consumption, excluding the DAA occurring in ex-drinkers from the calculations, adopting a broader definition of ex-drink- er which included both “regular ex-drinkers” and “infre- quent ex-drinkers”, using the alcohol use prevalences of ADICCIONES, 2021 · VOL. xx NO. x Methodology used to estimate alcohol-attributable mortality in Spain, 2001-2017 the period 2001-09 instead of 2010-2017 figures, which is equivalent to introducing a latency period of approximate- ly 12 years with respect to 2017, and estimating the DAA in traffic accidents taking as the PAFc the proportion of drivers and pedestrians who died in such accidents with a blood alcohol level above 0.8 grams/litre during 2015-17, calculated with data extracted from the annual reports of the National Institute of Toxicology and Forensic Sciences during 2001-17 (1,870 drivers and 489 pedestrians) (Insti- tuto Nacional de Toxicología y Ciencias Forenses [INTCF], 2019). Methodology discussion This article sets out a detailed methodology for estimat- ing alcohol-attributable mortality in Spain, following as far as possible the most common and updated international recommendations and procedures in the field (IHME, 2020; Rehm et al., 2017; Rehm et al., 2009; Sherk et al., 2017; WHO, 2000, 2018b, 2020). Its application to the pe- riod 2001-2017 has made it possible to obtain results by sex, age, calendar period, autonomous community, cause of death and type of drinker, which are quite consistent with other indicators of alcohol problems. In the future, it will in turn allow comparative results to be obtained for other subgroups of interest, such as those defined by educational level or socioeconomic position based on the mortality co- horts from the 2001 and 2011 censuses of the KharonXXI Platform for National Longitudinal Mortality Studies that is being established within the CIBERESP framework. The two main strengths of the methodology are that the PAFs were calculated from empirical data on population exposure to alcohol from almost all Spanish sources, and that individual survey data on average self-reported alco- hol use were corrected for underestimation in relation to other sources of greater validity such as statistics on the sale of alcoholic drinks, following a very painstaking pro- cess (Sordo et al., 2016). The methodology also has other strengths, such as the inclusion of DAA linked to ex-drink- ers. Finally, the PAFc obtained are quite robust because the distribution of population exposure to alcohol is based on the joint analysis of surveys with a significant sample size (n = 66,082 in 2010-2017 and 69,295 in 2001-2009). This makes it possible to obtain PAFc for the 20 population sub- groups mentioned above and to segment the PAFc for each cause of death by type of drinker (ex-drinker, low-risk and high-risk drinker). Despite its strengths and the fact that great pains were taken to collect all the relevant data available, it is undoubt- edly still a work in progress which will require reviewing and improving in the future. In this regard, a number of methodological options may be considered for the pro- posed model. A specific cause approach was chosen in preference of all causes because it allows better control of confounding factors, as well as making it possible to estimate the relative weights of different causes or groups of causes of death in the total mortality attributable to alcohol, an aspect of in- terest to clinical and public health practice (Corrao, Rub- biati, Zambon & Arico, 2002). To design more appropriate prevention and treatment interventions, it is not only a matter of knowing the number of DAA but also the specific causes. Beyond this general approach, conservative method- ological strategies were adopted to avoid overestimating the number of DAA, which makes it highly likely that the figures obtained underestimate the mortality attributable to alcohol in Spain. Among these strategies, the follow- ing are worth noting: 1) Only those causes of death for which there is clear evidence linking them to alcohol use and which have valid RR estimates were included (Table 1). This excludes a broad group of diseases with a proba- ble relationship to alcohol. 2) In calculating the amount of alcohol consumed, an alcohol content of 11.5% ABV was applied to wine, which is probably low in the current Span- ish context (Alston, Lapsley, Soleas & Tumber, 2013). 3) A restrictive definition of ex-drinker was applied, including only the fraction of these individuals (regular ex-drinkers), which may offer a better fit with the profile of ex-drinkers referred to in the RRs obtained in epidemiological studies. This option, which has probably not been adopted in other studies (Rehm et al., 2010a; Stockwell et al., 2016), consid- erably reduces estimates of the number of DAA because ex-drinkers have RRs close to high-risk drinkers (Rehm et al., 2010a; Stockwell et al., 2016) given that many stopped drinking due to problems caused by alcohol. In any case, it must be noted that the data to make the correction came from the United States, where the epidemiological situa- tion of alcohol use may not be equivalent to Spain’s. For example, alcohol use in Spain has clearly declined during the 21st century, something that has not happened in the United States (Breslow, Castle, Chen & Graubard, 2017; Centers for Disease Control and Prevention [CDC], 2020). The effect of including only regular ex-drinkers or all ex-drinkers in the calculations can be seen in Appendix Ta- ble 4. 4) Binge drinking was not specifically addressed due to the lack of corresponding information; however, it is considered that this consumption pattern was already tak- en into account in the algorithm used to obtain the PAFs in a general way. Nevertheless, the changes in the estimate as a result of considering the effect of this consumption pattern can be seen in Appendix Table 4. 5) The RRi of some diseases are probably underestimated given that they are calculated in comparison with abstainers, a group that in some studies may also include former regular drinkers (Rehm et al., 2017; Rehm, Shield, Roerecke & Gmel, 2016; Stockwell et al., 2016). However, a sensitivity analysis was ADICCIONES, 2021 · VOL. xx NO. x Marta Donat, Luis Sordo, Juan Miguel Guerras, Julieta Politi, José Pulido, Gregorio Barrio carried out for 2017 which considered a latency period of approximately 12 years and the estimate of the number of DAA was only 1.03 times greater than that considered valid in this study (Appendix Table 4). In addition, there are some general limitations that affect the estimates of alcohol-attributable mortality worldwide. Among them we must mention the scarcity and low quality of the RRi estimates published for certain causes of death, age groups, sex and average daily consumption. Thus, the RRi for some severe health problems, for example, traffic accidents, can vary widely by age, so the application of the same RRi at all ages could result in an underestimation of these deaths in young people and an overestimation in old- er people (Jones, Bellis, Dedman, Sumnal & Tocke, 2008; Rehm, Patra & Popova, 2006). The DAA estimates for the oldest age groups and some causes of death, such as cir- culatory diseases, are particularly dependent on the meth- odological options regarding exposure to alcohol or the RRi chosen to cmeasure them (Marmet et al., 2016; Sherk, Thomas, Churchill & Stockwell, 2020; Trias-Llimós, Mar- tikainen, Mäkelä & Janssen, 2018). Furthermore, the RRi used would most likely reflect the strength of the associa- tion between average consumption and mortality by cause in countries and time periods with different consumption patterns from those prevalent in Spain in the study period. Finally, it should be noted in this regard that when linking consumption to incidence, RRs do not normally differen- tiate between morbidity and mortality, so that sometimes reference is made to incidence of disease and not necessar- ily mortality. However, there is no alternative way to obtain these RRi . A futher limitation is related to the practice of correcting for underestimation homogeneously across all participants in population surveys. If, as is likely, there are differences in the degree of underestimation according to sociodemographic factors or level of average daily alcohol use, this could bias the estimates of the total number of DAA, as well as its distribution by sociodemographic fac- tors, cause of death and type of drinker. As indicated, the purpose of this methodology is to serve as the basis for establishing a routine alcohol-attributable mortality indicator in Spain to help monitor the issue, starting with an analysis of the limitations of existing gen- eral health information systems. In this regard, statistics on alcohol consumption in population health surveys, as well as alcoholic beverage sales, need to be made more consis- tent and maintained over time. Frequent changes in the questionnaires of the National Health Survey during the period 2001-2017 in terms of the content and format of the questions on alcohol use, or the uncertainty associat- ed with the data on sales of alcoholic beverages not sub- ject to special tax (wine and cider) published by the TA or international institutions, made it extremely difficult to estimate the population’s exposure to alcohol during this period. Adopting international instruments has not neces- sarily improved the situation because they are often not well adapted to the consumption patterns and types of bev- erages prevalent in Spain. Likewise, it would be helpful to determine the changes in the practices of certification and codification of causes of death in Spain, especially as re- gards circulatory diseases, in order to better interpret the findings on the changes in mortality attributable to alcohol by cause. Acknowledgments This article is a product of work carried out within the framework of a research project funded by the Government Delegation for the National Plan on Drugs (Delegación del Gobierno para el Plan Nacional sobre Drogas, DGPNSD) [Nº Exp: 2015I040] and the support of the human resourc- es department of the Carlos III Health Institute (Instituto de Salud Carlos III) (ISCIII-PFIS contract, Nº Exp: ENPY- 397/18-PFIS). We also wish to thank Cristina Ortiz from the ISCIII for her collaboration in carrying out the anal- ysis of alcohol consumption from some of the population surveys used for the estimates, and the Alcohol Group of the Spanish Epidemiology Society for their support in the completion of this study. Conflict of interests The authors declare no conflict of interest related to as- pects discussed in this article. References Agencia Europea de Medicamentos. (2010). Guideline on the development of medicinal products for the treatment of al- cohol dependence. Retrieved at https://www.ema.europa. eu/en/documents/scientific-guideline/guideline-de- velopment-medicinal-products-treatment-alcohol-de- pendence_en.pdf. Agencia Tributaria. (2015). Impuestos especiales. Retrieved at http://www.agenciatributaria.es/AEAT.internet/Ini- cio/Aduanas_e_Impuestos_Especiales/Impuestos_Es- peciales/Impuestos_Especiales.shtml. Agencia Tributaria. (2019). Impuestos sobre el alcohol y las bebi- das alcohólicas, 2015. Retrieved at https://www.agencia- tributaria.es/AEAT.internet/Inicio/La_Agencia_Tribu- taria/Memorias_y_estadisticas_tributarias/Estadisticas/ Estadisticas_por_impuesto/Impuestos_especiales/Im- puestos_especiales.shtml. Alston, J. M. F., Lapsley, J.T., Soleas, G. & Tumber K.P. (2013). The high and rising alcohol content of wine. In. E. P. Giraud-Héraud, M.C. (Ed.), Wine Economics. Lon- don: Palgrave Macmillan. ADICCIONES, 2021 · VOL. xx NO. x Methodology used to estimate alcohol-attributable mortality in Spain, 2001-2017 Boniface, S. & Shelton, N. (2013). How is alcohol con- sumption affected if we account for under-reporting? A hypothetical scenario. European Journal of Public Health, 23, 1076-1081. doi:10.1093/eurpub/ckt016. Breslow, R. A., Castle, I. P., Chen, C. M. & Graubard, B. I. (2017). Trends in alcohol consumption among old- er americans: national health interview surveys, 1997 to 2014. Alcoholism: Clinical and Experimental Research, 41, 976-986. doi:10.1111/acer.13365. Centers for Disease Control and Prevention. (2020). Alco- hol use. Retrieved at https://www.cdc.gov/nchs/fastats/ alcohol.htm. Connor, J., Kydd, R., Rehm, J. & Shield, K. (2013). Alco- hol-attributable burden of disease and injury in New Zealand: 2004 and 2007. Research report commissioned by the Health Promotion Agency. Retrieved at https://www.hpa.org.nz/ sites/default/files/Attributable%20fractions%20Final. pdf. Corrao, G., Bagnardi, V., Zambon, A. & Arico, S. (1999). Exploring the dose-response relationship between alco- hol consumption and the risk of several alcohol-relat- ed conditions: a meta-analysis. Addiction, 94, 1551-1573. doi:10.1046/j.1360-0443.1999.9410155111.x. Corrao, G., Rubbiati, L., Zambon, A. & Arico, S. (2002). Al- cohol-attributable and alcohol-preventable mortality in Italy. A balance in 1983 and 1996. European Journal of Pub- lic Health, 12, 214-223. doi:10.1093/eurpub/12.3.214. Dawson, D. A. (2003). Methodological issues in measuring alcohol use. Alcohol Research & Health, 27, 18-29. Delegación del Gobierno para el Plan Nacional sobre Dro- gas. (2018). EDADES 2017. Encuesta sobre alcohol y otras dro- gas en España (EDADES), 1995-2017. Retrieved at http:// www.pnsd.mscbs.gob.es/profesionales/sistemasInfor- macion/sistemaInformacion/pdf/EDADES_2017_In- forme.pdf. Eurostat. (2013). Revision of the European Standard Population. Report of Eurostat’s task force. Retrieved at https://ec.eu- ropa.eu/eurostat/documents/3859598/5926869/ KS-RA-13-028-EN.PDF/e713fa79-1add-44e8-b23d- 5e8fa09b3f8f. Eurostat. (2019). Purchasing power parities (PPPs), price level indices and real expenditures for ESA 2010 aggregates. Re- trieved at https://ec.europa.eu/eurostat/web/prod- ucts-datasets/-/prc_ppp_ind&lang=en. Fierro, I., Ochoa, R., Yánez, J. L., Valderrama, J. C. & Ál- varez, F. J. (2008). Mortalidad y mortalidad premat- ura relacionadas con el consumo de alcohol en Es- paña entre 1999 y 2004. Medicina Clínica, 131, 10-13. doi:10.1157/13123036. Global Burden of Disease. (2018a). Alcohol use and burden for 195 countries and territories, 1990-2016: a systematic analysis for the Global Burden of Disease Study 2016. Lancet, 392, 1015-1035. doi:10.1016/S0140- 6736(18)31310-2. Global Burden of Disease. (2018b). Global, 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 Dis- ease Study 2017. Lancet, 392, 1923-1994. doi:10.1016/ S0140-6736(18)32225-6. Gmel, G., Shield, K. D. & Rehm, J. (2011). Developing a method to derive alcohol-attributable fractions for HIV/ AIDS mortality based on alcohol’s impact on adherence to antiretroviral medication. Population Health Metrics 9, 5. doi:10.1186/1478-7954-9-5. Institute for Health Metrics and Evaluation. (2020). GBD Results Tool. Retrieved at http://ghdx.healthdata.org/ gbd-results-tool. Instituto Nacional de Estadística. (2019a). Encuesta Na- cional de Salud. Microdatos. Retrieved at https://www. ine.es/dyngs/INEbase/es/operacion.htm?c=Estadisti- ca_C&cid=1254736176783&menu=resultados&secc=12 54736195295&idp=1254735573175. Instituto Nacional de Estadística. (2019b). INEbase. Ser- vicios. Hosteleria y turismo. Retrieved at https://www. ine.es/dyngs/INEbase/es/categoria.htm?c=Estadisti- ca_P&cid=1254735576863. Instituto Nacional de Estadística. (2019c). INEbase. De- mografía y población. Cifras de población y censos demográf- icos. Cifras de población. Retrieved at https://www. ine.es/dyngs/INEbase/es/operacion.htm?c=Esta- dis t ica_C&cid=1254736176951&menu=ult iDa- tos&idp=1254735572981. Instituto Nacional de Estadística. (2020). INEbase Socie- dad. Salud. Estadística de defunciones según la causa de muerte. Resultados. 2020. Retrieved at https://www.ine. es/dyngs/INEbase/es/operacion.htm?c=Estadisti- ca_C&cid=1254736176780&menu=resultados&idp=125 4735573175#!tabs-1254736194710. Instituto Nacional de Toxicología y Ciencias Forenses. (2019). Memorias. Víctimas mortales en accidentes de tráfico. Retrieved at https://www.mjusticia.gob.es/cs/Satellite/ Portal/va/ministerio/organismos-ministerio-justicia/ instituto-nacional/memorias. Jones, L. & Bellis, M. A. (2013). Updating England-Specific Alcohol-Attributable Fractions. Retrieved at http://www. cph.org.uk/wp-content/uploads/2014/03/24892-AL- COHOL-FRACTIONS-REPORT-A4-singles-24.3.14.pdf. Jones, L., Bellis, M. A., Dedman, D., Sumnal, H. & Tocke, C. (2008). Alcohol-attributable fractions for England. Alco- hol-attributable mortality and hospital admissions. Retrieved at http://www.nwph.net/nwpho/publications/alcoho- lattributablefractions.pdf. Kehoe, T., Gmel, G., Shield, K. D., Gmel, G. & Rehm, J. (2012). Determining the best population-level alcohol consumption model and its impact on estimates of alco- ADICCIONES, 2021 · VOL. xx NO. x Marta Donat, Luis Sordo, Juan Miguel Guerras, Julieta Politi, José Pulido, Gregorio Barrio hol-attributable harms. Population Health Metrics, 10, 6. doi:10.1186/1478-7954-10-6. Landberg, J. & Norstrom, T. (2011). Alcohol and homicide in Russia and the United States: a comparative analy- sis. Journal of Studies on Alcohol and Drugs, 72, 723-730. doi:10.15288/jsad.2011.72.723. Mann, K., Aubin, H. J. & Witkiewitz, K. (2017). Reduced drinking in alcohol dependence treatment, what is the evidence? European Addiction Research, 23, 219-230. doi:10.1159/000481348. Marmet, S., Rehm, J. & Gmel, G. (2016). The importance of age groups in estimates of alcohol-attributable mortal- ity: Impact on trends in Switzerland between 1997 and 2011. Addiction, 111, 255-262. doi:10.1111/add.13164. Marmet, S., Rehm, J., Gmel, G., Frick, H. & Gmel, G. (2014). Alcohol-attributable mortality in Switzerland in 2011--age-specific causes of death and impact of heavy versus non-heavy drinking. Swiss Medical Weekly, 144, w13947. doi:10.4414/smw.2014.13947. Meier, P. S., Meng, Y., Holmes, J., Baumberg, B., Purshouse, R., Hill-McManus, D. & Brennan, A. (2013). Adjusting for unrecorded consumption in survey and per capi- ta sales data: quantification of impact on gender- and age-specific alcohol-attributable fractions for oral and pharyngeal cancers in Great Britain. Alcohol and Alcohol- ism, 48, 241-249. doi:10.1093/alcalc/agt001. Ministerio de Sanidad, Consumo & Bienestar Social. (2019). Banco de datos. Retrieved at https://www.mscbs. gob.es/estadEstudios/estadisticas/bancoDatos.htm. Norstrom, T. & Skog, O. J. (2001). Alcohol and mor- tality: Methodological and analytical issues in ag- gregate analyses. Addiction, 96 (Suppl 1), S5-S17. doi:10.1080/09652140020021143. Organización Médica Colegial y Ministerio de Sanidad y Consumo. (2006). Guía de buena práctica clínica en uso y abuso de alcohol. Retrieved at http://www.msssi.gob.es/ alcoholJovenes/docs/guiaBPCusoabusoAlcohol07.pdf. Rehm, J., Baliunas, D., Borges, G. L., Graham, K., Irving, H., Kehoe, T., ...Taylor, B. (2010a). The relation be- tween different dimensions of alcohol consumption and burden of disease: an overview. Addiction, 105, 817-843. doi:10.1111/j.1360-0443.2010.02899.x. Rehm, J., Gmel, G. E., Sr., Gmel, G., Hasan, O. S. M., Imti- az, S., Popova, S., ...Shuper, P. A. (2017). The relation- ship between different dimensions of alcohol use and the burden of disease-an update. Addiction, 112, 968- 1001. doi:10.1111/add.13757. Rehm, J., Kehoe, T., Gmel, G., Stinson, F., Grant, B. & Gmel, G. (2010b). Statistical modeling of volume of al- cohol exposure for epidemiological studies of popula- tion health: the US example. Population Health Metrics, 8, 3. doi:10.1186/1478-7954-8-3. Rehm, J., Klotsche, J. & Patra, J. (2007). Comparative quan- tification of alcohol exposure as risk factor for global burden of disease. International Journal of Methods in Psy- chiatric Research, 16, 66-76. doi:10.1002/mpr.204. Rehm, J., Mathers, C., Popova, S., Thavorncharoensap, M., Teerawattananon, Y. & Patra, J. (2009). Global burden of disease and injury and economic cost attributable to alcohol use and alcohol-use disorders. Lancet, 373, 2223- 2233. doi:10.1016/S0140-6736(09)60746-7. Rehm, J., Patra, J. & Popova, S. (2006). Alcohol-attribut- able mortality and potential years of life lost in Canada 2001: implications for prevention and policy. Addiction, 101, 373-384. doi:10.1111/j.1360-0443.2005.01338.x. Rehm, J., Rehm, M. X., Shield, K. D., Gmel, G. & Gual, A. (2013). Alcohol consumption, alcohol dependence and related harms in Spain, and the effect of treat- ment-based interventions on alcohol dependence. Adic- ciones, 25, 11-18. Rehm, J., Shield, K. D., Rehm, M. X., Gmel, G. & Frick, U. (2012). Alcohol consumption, alcohol dependence and attrib- utable burden of disease in Europe. Potential gains from effective interventions for alcohol dependence. Retrieved at http:// amphoraproject.net/w2box/data/AMPHORA%20Re- ports/CAMH_Alcohol_Report_Europe_2012.pdf. Rehm, J., Shield, K. D., Roerecke, M. & Gmel, G. (2016). Modelling the impact of alcohol consumption on car- diovascular disease mortality for comparative risk as- sessments: an overview. BMC. Public Health, 16, 363. doi:10.1186/s12889-016-3026-9. Rey, G., Boniol, M. & Jougla, E. (2010). Estimating the num- ber of alcohol-attributable deaths: methodological issues and illustration with French data for 2006. Addiction, 105, 1018-1029. doi:10.1111/j.1360-0443.2010.02910.x. Rodríguez-Martos, D. A., Gual, S. A. & Llopis Llácer, J. J. (1999). The “standard drink unit” as a simplified record of alcoholic drink consumption and its measurement in Spain. Medicina Clínica, 112, 446-450. Roerecke, M. & Rehm, J. (2013). Alcohol use disorders and mortality: A systematic review and meta-analysis. Addic- tion, 108, 1562-1578. doi:10.1111/add.12231. Samokhvalov, A. V., Irving, H. M. & Rehm, J. (2010). Alco- hol consumption as a risk factor for atrial fibrillation: a systematic review and meta-analysis. European Journal of Cardiovascular Prevention & Rehabilitation, 17, 706-712. doi:10.1097/HJR.0b013e32833a1947. Sherk, A., Stockwell, T., Rehm, J., Dorocicz, J. & Shield, K. D. (2017). The International Model of Alcohol Harms and Policies (InterMAHP). Version 1.0: December 2017. Re- trieved at www.intermahp.cisur.ca. Sherk, A., Thomas, G., Churchill, S. & Stockwell, T. (2020). Does drinking within low-risk guidelines prevent harm? Implications for high-income countries using the inter- national model of alcohol harms and policies. Journal of Studies on Alcohol and Drugs, 81, 352-361. Sociedad Española de Medicina de Familia y Comunitaria. (2005). Guía rápida para abordar el abuso de alcohol. Re- ADICCIONES, 2021 · VOL. xx NO. x Methodology used to estimate alcohol-attributable mortality in Spain, 2001-2017 trieved at https://www.semfyc.es/pfw_files/cma/Infor- macion/modulo/documentos/abuso-alcohol.pdf. Sordo, L., Barrio, G., Bravo, M. J., Villalbí, J. R., Espelt, A., Neira, M. & Regidor, E. (2016). Estimating average alcohol consumption in the population using multiple sources: The case of Spain. Population Health Metrics, 14, 21. doi:10.1186/s12963-016-0090-4. Stockwell, T., Zhao, J., Panwar, S., Roemer, A., Naimi, T. & Chikritzhs, T. (2016). Do “moderate” drinkers have reduced mortality risk? A systematic review and me- ta-analysis of alcohol consumption and all-cause mortal- ity. Journal of Studies on Alcohol and Drugs, 77, 185-198. doi:10.15288/jsad.2016.77.185. Stockwell, T., Zhao, J., Sherk, A., Rehm, J., Shield, K. & Naimi, T. (2018). Underestimation of alcohol consump- tion in cohort studies and implications for alcohol’s con- tribution to the global burden of disease. Addiction, 113, 2245-2249. doi:10.1111/add.14392. Trias-Llimós, S., Martikainen, P., Mäkelä, P. & Janssen, F. (2018). Comparison of differents 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. Trolldal, B. (2001). Alcohol sales figures in 15 Europe- an countries: Corrected for consumption abroad and tax-free purchases. Nordic Studies on Alcohol, 18, 71-81. doi:10.1177/145507250101801S02. TURESPAÑA. (2019). Estadísticas turísticas. Retrieved at http://estadisticas.tourspain.es/es-ES/turismobase/Pa- ginas/default.aspx. Villarroel, M. A., Clarke, T. C. & Schoenborn, C. A. (2016). Tables of adult health behaviors, alcohol use: National Health Interview Survey, 2011–2014., from National Center for Health Statistics. Center for Disease Control and Pre- vention (CDC). Retrieved at https://ftp.cdc.gov/pub/ Health_Statistics/NCHs/NHIS/SHS/2011-2014_AHB_ Table_ALC-1.pdf. World-Bank. (2019). International Comparison Program. ICP 2011. Retrieved at https://datacatalog.worldbank.org/ dataset/international-comparison-program-2011. World Health Organization. (2000). International guide for monitoring alcohol consumption and related harm. WHO/ MSD/MSB/00.4. Retrieved at http://whqlibdoc.who. int/hq/2000/who_msd_msb_00.4.pdf. World Health Organization. (2009). Global health risks. Mor- tality and burden of disease attributable to selected major risks. WA 105. Retrieved at http://www.who.int/healthinfo/ global_burden_disease/GlobalHealthRisks_report_full. pdf. World Health Organization. (2018a). Methods and data sources for global burden of disease estimates, 2000-2016. Re- trieved at https://www.who.int/healthinfo/global_bur- den_disease/GlobalDALY_method_2000_2016.pdf. World Health Organization. (2018b). Global status report on alcohol and health. Retrieved at https://www.who.int/ substance_abuse/publications/alcohol/en/. World Health Organization. (2020). Global Information Sys- tem on Alcohol and Health (GISAH). Retrieved at http:// apps.who.int/gho/data/node.gisah.GISAH?showon- ly=GISAH. ADICCIONES, 2021 · VOL. xx NO. x