Final Report Summary - ALCOHOLLIFECOURSE (Alcohol Consumption across the Life-course: Determinants and Consequences)
Alcohol consumption varies across the life course and there had been little research examining the effects of different drinking patterns over several decades on the risk of chronic disease. We obtained data from multiple cohort studies, achieving a sample size of over 100,000 individuals with repeat measures of alcohol over their life-span. Methods for obtaining information on alcohol consumption differed across participating studies. Mean weekly alcohol intake was derived for each cohort and harmonised into grams of alcohol. Likewise frequency of consumption was derived from each cohort and grouped accordingly.
Predicted mean volume of alcohol consumed as a function of age was estimated using multilevel models fitted to each cohort. In studies where age was consistent across individuals (birth cohorts), linear effects were assumed when only three measurement occasions were available while in studies with four or more measures, or a range of ages at each occasion, quadratic or cubic polynomial terms were used to describe non-linear trajectories. Models were fit for men and women separately and estimated using a maximum likelihood algorithm. Robust standard errors were calculated.
We then combined all cohorts into a single dataset and fitted a three-level multilevel model (observations nested within individuals nested in cohorts) to estimate volume of alcohol consumed as a function of age across the life course with adjustment for period (broadly defined using the decade in which the measurement took place). We used fractional polynomial terms to best describe the shape of the trajectory and centred age at 40 years. We also included an interaction between age and period. We found that for men, mean consumption rose during adolescence, peaked at around age 25 at 20 units (160g) per week and then declined and plateaued during mid-life, before declining from around age 65 years. A similar trajectory was seen for women, but with lower overall consumption (peak of around 8 units (64g) per week).
To our knowledge, this is the first attempt to synthesise information from overlapping large population based cohorts to represent alcohol consumption across the entire life span. The data were collected between 1981 and 2012 and the participants were born in different eras, therefore the analysis allows for inspection of period and cohort effects. In subsequent analyses looking at trajectories of alcohol over the life course, we found that drinking in early and mid-adulthood has a significant effect on biological markers, atherosclerosis and disease outcomes such as coronary heart disease, diabetes and liver functioning in later life.
Research findings were published in peer review academic journals. Presentations were made at specialist alcohol conferences by all the team to disseminate findings and also to publisize opportunities to use the data sources and methodological innovations developed during this project. We ran a project website (http://www.ucl.ac.uk/alcohol-lifecourse) which provides lay-summaries of our work to further facilitate engagement with non-academic audiences. We regularly disseminate our findings on twitter to further maximise the impact of our work. The Institute of Alcohol Studies and Alcohol Research UK agreed to act as non-academic partners on this project. These are non-governmental organisations with the aim of serving the public interest on policies linked to alcohol and health, and to do so they advocate the use of evidence-based practices. Their commitment is to ensure that the findings from this work will be disseminated widely and translated clearly so that those who are in a position to make decisions regarding alcohol policies are able to act upon them.
Our expertise in this area has resulted in us hosting three international research visitors in my group and led to collaborative outputs .
The ERC Grant enabled me to build a repository of data with longitudinal cohort studies with repeat measures of alcohol consumption from within the UK and elsewhere in the world. While this had immediate benefits for the ERC project; it also expands the opportunities for future work on the important topic of life-course alcohol consumption and other health related harms. More data will give greater power to look at effect modifications, such as social inequalities in alcohol-related harm. I will seek to make new linkages with routine data, such as Hospital Episode Statistics, to enhance the outcome analyses by looking at hospitalisations for alcohol-related conditions. With more harmonised cohorts we will have greater potential to study rarer endpoints that single cohort studies are unable to capture as well as examine the impact of alcohol intake on the progression of disease (i.e. prognosis), a topic that is currently hugely under-researched. I would very much like to expand our remit to analyse specific types of cancer, such as liver, breast, and oesophageal which are related to alcohol consumption, but the life course exposure risk is unknown.
Predicted mean volume of alcohol consumed as a function of age was estimated using multilevel models fitted to each cohort. In studies where age was consistent across individuals (birth cohorts), linear effects were assumed when only three measurement occasions were available while in studies with four or more measures, or a range of ages at each occasion, quadratic or cubic polynomial terms were used to describe non-linear trajectories. Models were fit for men and women separately and estimated using a maximum likelihood algorithm. Robust standard errors were calculated.
We then combined all cohorts into a single dataset and fitted a three-level multilevel model (observations nested within individuals nested in cohorts) to estimate volume of alcohol consumed as a function of age across the life course with adjustment for period (broadly defined using the decade in which the measurement took place). We used fractional polynomial terms to best describe the shape of the trajectory and centred age at 40 years. We also included an interaction between age and period. We found that for men, mean consumption rose during adolescence, peaked at around age 25 at 20 units (160g) per week and then declined and plateaued during mid-life, before declining from around age 65 years. A similar trajectory was seen for women, but with lower overall consumption (peak of around 8 units (64g) per week).
To our knowledge, this is the first attempt to synthesise information from overlapping large population based cohorts to represent alcohol consumption across the entire life span. The data were collected between 1981 and 2012 and the participants were born in different eras, therefore the analysis allows for inspection of period and cohort effects. In subsequent analyses looking at trajectories of alcohol over the life course, we found that drinking in early and mid-adulthood has a significant effect on biological markers, atherosclerosis and disease outcomes such as coronary heart disease, diabetes and liver functioning in later life.
Research findings were published in peer review academic journals. Presentations were made at specialist alcohol conferences by all the team to disseminate findings and also to publisize opportunities to use the data sources and methodological innovations developed during this project. We ran a project website (http://www.ucl.ac.uk/alcohol-lifecourse) which provides lay-summaries of our work to further facilitate engagement with non-academic audiences. We regularly disseminate our findings on twitter to further maximise the impact of our work. The Institute of Alcohol Studies and Alcohol Research UK agreed to act as non-academic partners on this project. These are non-governmental organisations with the aim of serving the public interest on policies linked to alcohol and health, and to do so they advocate the use of evidence-based practices. Their commitment is to ensure that the findings from this work will be disseminated widely and translated clearly so that those who are in a position to make decisions regarding alcohol policies are able to act upon them.
Our expertise in this area has resulted in us hosting three international research visitors in my group and led to collaborative outputs .
The ERC Grant enabled me to build a repository of data with longitudinal cohort studies with repeat measures of alcohol consumption from within the UK and elsewhere in the world. While this had immediate benefits for the ERC project; it also expands the opportunities for future work on the important topic of life-course alcohol consumption and other health related harms. More data will give greater power to look at effect modifications, such as social inequalities in alcohol-related harm. I will seek to make new linkages with routine data, such as Hospital Episode Statistics, to enhance the outcome analyses by looking at hospitalisations for alcohol-related conditions. With more harmonised cohorts we will have greater potential to study rarer endpoints that single cohort studies are unable to capture as well as examine the impact of alcohol intake on the progression of disease (i.e. prognosis), a topic that is currently hugely under-researched. I would very much like to expand our remit to analyse specific types of cancer, such as liver, breast, and oesophageal which are related to alcohol consumption, but the life course exposure risk is unknown.