The World Health Organization (WHO) recommends fiscal measures (e.g. taxation of certain food items/groups) to prevent non-communicable diseases. Some primary studies assess taxation effects on food consumption patterns or sales data in real-world scenarios, but they face challenges such as seasonality of consumption patterns, substitution with other food items or sales/consumption data of different individuals before/after the implementation. Study designs investigating these effects may be common in econometrics and the social sciences, but they are uncommon in epidemiology and clinical medicine, and are rarely included in Cochrane Reviews. The current public debate on implementing food taxes does not sufficiently consider whether causality can be inferred from such study designs. At the same time, terms used for these uncommon study designs by study authors or the Cochrane Effective Practice and Organisation of Care (EPOC) study design terminology may fall short of differentiating important design features and their implications for quality of evidence.
The primary goal of this methodological case study is to compare the performance of two different study design classification systems in a sample of Cochrane Reviews: 1) the EPOC study design classification and 2) the Reeves et al 2017 classification - a taxonomy without labels.
Our case study will be based on included studies from three ongoing Cochrane Reviews on food taxation - namely the taxation of sugar-sweetened beverages, sugar and fat. Studies were detected through searches of 20 electronic databases.
We will present the major methodological advantages and shortcomings of both classification systems based on the studies included in our upcoming Cochrane Reviews on food taxation.
Public discussion on food tax implementation is substantially informed by evidence based on primary studies with uncommon types of study designs. Thus, these uncommon study designs need to be classified (and assessed) correctly by Cochrane authors. This case study of classifying primary studies according to two different classification systems will illustrate which major study design aspects are worth noting to understand how the results of individual studies can be interpreted by end-users in a meaningful way and will show room for improvement for study design classification systems in general.