Commentary on Costanzo et al. (2019): The need for secondary analyses of prospective cohort studies—creating a better understanding of alcohol consumption and hospitalization

Abstract

The risk relationsip between alcohol consumption and disease occurrence often differs depending on whether the end-point used is mortality or morbidity. More studies using morbidity as an end-point are required to avoid incorrect results when estimating the burden of alcohol-related harm. Cohort study networks offer opportunities for conducting such studies. Alcohol consumption is a leading risk factor for the burden of disease 1, 2, and a large proportion of the alcohol-attributable burden is due to disability and impaired functioning 1-3. However, the majority of cohort studies which examine the risk relationship (RR) between alcohol consumption and disease occurrence use mortality as an end-point 4 because mortality is easier to assess as its recording is often mandated by law, while the measurement of non-fatal end-points are more resource-intensive to collect and often are not mandated by law 4, 5. For example, studies using hospitalization as an end-point require a central collection and collation of hospitalization records, and for these records to be linked to a unique personal identifier. However, in many countries and regions, hospitalization services are administered by multiple companies with different data access and data organization policies. Under such circumstances, the resources needed to collect and harmonize hospitalization data from multiple sources are often cost- and resource-prohibitive. Consequently, for meta-analyses, data on the effects of alcohol on morbidity alone are not sufficiently available to produce accurate and precise RR estimates. Accordingly, for most disease conditions, meta-analyses combine studies which use mortality and morbidity as end-points 3, 4. Previous studies have found evidence that the alcohol RRs for various chronic diseases 6-9 may differ depending on whether mortality or morbidity is used as the end-point. In line with this, the Moli-sani study differs from mortality-based studies 3, 19 in that heavy alcohol consumers were found not to be at an increased risk of hospitalization from ischaemic heart disease and cerebrovascular accidents. These differences in the RRs by disease end-point are due to potentially confounding factors (e.g. the stage at which the diseases are diagnosed and treated, access to health care, heavy drinkers avoiding hospitals due to stigmatization by health-care workers 10-14 and/or factors affected by alcohol consumption (e.g. adherence to treatment regimens 15. Therefore, the combining of mortality and morbidity RRs may lead to incorrect results in studies which utilize both these RRs, such as the 2018 Global Status Report on Alcohol and Health 1, the 2017 Global Burden of Disease study 2 and other studies which have estimated hospitalizations and hospitalization costs attributable to alcohol 16-18. The data analysis of the Moli-sani study by Costanzo and colleagues 19 demonstrates the availability of opportunities to perform analyses of cohort studies which are linked to hospital discharge registries. As a result of the observation that there are different alcohol RRs for morbidity and mortality, there is a need for additional analyses of cohort studies linked to hospitalization records to examine the association of alcohol with specific alcohol-related diseases. Such analyses could take advantage of existing cohort study networks. An example of a large cohort study network is the European Prospective Investigation into Cancer and Nutrition (of which the Moli-sani study is a participant). Although this cohort study network was originally designed to examine the RRs for various risk factor exposures and the development of cancer, such a network can be used for the secondary purpose of assessing the RR for alcohol consumption and hospitalization. These sorts of analyses will lead to improvements in our understanding of how alcohol consumption leads to disease occurrence and, further, will improve the evidence base for studies which model how best to reduce morbidity, hospitalizations and hospital costs due to alcohol consumption. None. K.D.S. receives funding from the 2018 Canadian Institutes of Health Research - Institute of Population and Public Health Trailblazer Award in Population and Public Health Research.