Discussion
The aim of our scoping literature review was to compile studies that assessed the BoD attributable to risk factors performed in Europe since 1990, with a main focus on CRA. We extracted data on the data input sources and methodological choices needed to compute YLL, YLD or DALY attributable to one or more risk factors. A total of 113 papers were identified. Within them, 107 used the CRA approach and were categorized as either independent BoD or GBD-linked studies. Our results showed that the methods used to perform CRA varied substantially across independent European BoD studies. While there were some methodological choices that are more common than others, we did not observe patterns in terms of country, year or risk factor. All the different methodological choices could affect the comparability of estimates between and within countries and/or risk factors, since they might significantly influence the quantification of the attributable burden.
In general, our review showed a propensity in Europe to explore behavioural risk factors more than others. Tobacco use, air pollution and alcohol use are the most analysed risk factors. This might be due to the fact that causal relationships for these risk factors are more studied, together with availability of better exposure data (e.g. tobacco and cancer). These risk factors are obviously important, but there are other risk factors that while equally important, have not been widely studied in a European population. Particular attention could be drawn to the relationship between dietary risks/low physical activity and cardiovascular disease, highly prevalent in European countries. Independent studies analysed risk factors not included in GBD, like low socioeconomic status and depressive disorders. The definition of risk factor might also be a point of discussion when some diseases can also be regarded as risk factors. In this study, we defined a risk factor as every individual behavioural choice or environmental factor that affects the risk associated to a disease outcome. We used the categorization adopted in the GBD studies but also included risk factors that were not regarded as such in the GBD framework. An example is type 2 diabetes which could increase the risk of cancer. This can be problematic since it is difficult to establish if the risk factor triggered a chronic diseases or if it was triggered by a chronic disease. For example, depressive disorders can develop in the setting of chronic disease such as dementia but can also result in the development of chronic disease [
34]. That is why the investigation of causality is a keystone for the estimation of the attributable burden.
From 2000 onwards, the most common methodology was the use of PAF with RR and a counterfactual value set to the category with the lowest risk, as performed in the GBD study [
14]. Deviations from the GBD framework included even differences in basic concepts, like the definition of risk factor or the terminology used to refer to PAF. Even though the diversity in terminology does not affect the final results, it hampers comparison of results across studies and adds to confusion in the interpretation of the methods and results.
Among independent BoD studies, many differences were observed for the exposure–response function, in its definition and source. The latter particularly varies across studies, with less than half of the studies using meta-analysis and literature reviews. Different sources and definitions can lead to different estimations of the attributable burden within the same risk factor. Another important choice in the CRA framework is the selection of the TMREL. The great majority of the studies decided to set it to the category at the lowest risk, which is implicit in the CRA methodology. However, this was rarely specified in the papers, affecting the comparability across countries, diseases and/or risk factors. Exploring different optimal scenarios was not uncommon in our review and was used to assess the impact of different interventions, most commonly in air pollution, alcohol and smoking. This is often referred to as health impact assessment (HIA), where the CRA methodology and HIA go hand in hand but for which the difference in purposes is often neglected.
The remaining six studies used an approach other than CRA, which was more common for some types of risk factors and outcomes. These included environmental and occupational risk factors, which more commonly employ bottom-up approaches and health outcomes where disease envelopes may not be available. This could be explained by the fact that traditionally this category of risk factors may have been included in a toxicological risk assessment of chemicals-approach, where for example exposure–response functions were derived with another purpose than a quantitative estimate of number of incident cases.
Our review highlighted some gaps in the uncertainty analyses and the investigation of causal relationships of BoD studies. Half of the independent BoD studies did not perform uncertainty analysis, and half of those that took into account parameter uncertainty did not report important methodological information like the method used for the analysis. Uncertainty on exposure–response functions was more frequently propagated than other inputs, such as exposure assessments. Less than half of the studies investigated or discussed causality in their CRA. Although the gold standard for concluding on causality is often considered to be a randomized controlled trial, in practice researchers often must rely on the strength of evidence that is brought by a variety of studies. It is therefore important to discuss causality in a risk factor assessment exercise [
35]. On the other hand, causality might be very difficult to prove for certain hazards and restricting inclusion of health effects to only those where causality is proven might underestimate the true burden. For many diseases causality is multifactorial leading to a difficulty to clearly disentangle the burden of each risk factor on a determinate disease, as well as how different risk factors may further interact with each other. In addition, randomized controlled trials are often not feasible due to numerous reasons, e.g. resources availability, ethical controversies. Scenario analysis was undertaken in very few independent studies but represents an essential tool for exploring the impact of different methodological choices and inclusion or exclusion of health outcomes of varying degree of causality.
The detected lack of consistency in terminology and methods makes comparisons and interpretation of results more challenging. Well-established guidelines that can be used in future studies estimating risk factor attributable burden could be achieved by publication of handbooks, manuals, protocols, etc. While heterogeneity is inevitable, it is important to make assumptions and methodological choices explicit, and to discuss possible limitations or develop alternative scenarios to quantify the associated uncertainties.
Strengths and limitations
Our scoping literature review brings together existing risk factor BoD studies undertaken in Europe. We comprehensively reviewed the methodological choices and assumptions used to calculate the BoD attributable to risk factors in terms of YLL, YLD, and DALY within CRA studies. This literature review used a variety of literature databases and search engines, as well as the consultation with European experts that work in the field of BoD in their respective countries. Nevertheless, our search may be limited by the nature of the grey literature searched and the national public health websites targeted, where some BoD studies may have been missed. In contrast to what is commonly done in systematic literature reviews, we did not perform a quality assessment of the included studies. Considering that no estimates were extracted but only methodological information, we did not consider a bias assessment relevant for the objectives of this literature review. In addition, to the best of our knowledge, there is not a tool that was specifically develop for evaluating the quality of BoD studies and CRA studies. For this reason, the results of this review will be used to feed future developments of these kind of bias assessment tools. Within our study we focused on the methodological choices of CRA studies. Considering the importance of other methods, we decided to also include the non-CRA studies identified by our search string. Nevertheless, since the search strategy focused on CRA, some risk assessment studies might have been disregarded. This limitation might be mitigate by the access to a wide international network that helped us finding and translating independent and/or GBD-linked BoD studies.
Research implications
This review is part of a series of reviews [
5,
9] that aims to compile BoD studies in Europe and to summarize methodological choices in the estimation of DALY. Each review focuses on the assessment of methodological design choices that were used in studies assessing the burden of non-communicable, injuries, infectious diseases, and risk factors. One of the main aims of the
burden-eu network is to provide a standardized statement for reporting DALY calculations in BoD studies. The development and use of key standardized guidelines for reporting BoD methodological choices may help to have more accessible BoD estimates. Our literature review serves as a critical input for such developments since we underlined the necessity for transparency and uniformity in risk factor BoD studies.
Acknowledgements
The authors wish to thank Maarten Engel from the Erasmus MC Library for developing and updating the search strategies. The authors would like to acknowledge the networking support from COST Action CA18218 (European Burden of Disease Network;
www.burden-eu.net), supported by COST (European Cooperation in Science and Technology;
www.cost.eu).
The COST Action CA18218 participants and affiliations:
Gunn Marit Aasvang 19, Balázs Ádám 20, Ala’a Alkerwi 21, Jalal Arabloo 22, Ana Lúcia Baltazar 23, Hilal Bektas Uysal 24, Boris Bikbov 25, Anette Kocbach Bolling 19, Maria Borrell-Pages 27, Giulia Carreras 28, Giulio Castelpietra 29, José Chen-Xu 30, Šeila Cilović Lagarija 31, Barbara Corso 32, Sarah Cuschieri 33, Robby De Pauw 1, Sonia Dhaouadi 35, Klara Dokova 36, Keren Dopelt 37, Mary Economou 38, Theophilus I. Emeto 39, Peter Fantke 5, Florian Fischer 41, Alberto Freitas 42, Lucia Galluzzo 43, Juan Manuel García-González 44, Federica Gazzelloni 45, Mika Gissler 46, Artemis Gkitakou 4, Sezgin Gubes 48, Irina Guseva Canu 49, Cesar A. Hincapié 50, Paul Hynds 51, Irena Ilic 52, Milena Ilic 53, Gaetano Isola 54, Zubair Kabir 55, Pavel Kolkhir 56, Naime Meriç Konar 57, Mirjam Kretzschmar 58, Mukhtar Kulimbet 59, Carlo La Vecchia 60,Carina Ladeira 61, Brian Lassen 5, Paolo Lauriola 63, Heli Lehtomäki 46,Miriam Levi 65, Marjeta Majer 52, Scott A. McDonald 58, Enkeleint A. Mechili 68, Janis Misins 69, Lorenzo Monasta 70, Javier Muñoz Laguna 50, Sónia Namorado 72, Evangelia Nena 73, Edmond S.W. Ng 74, Paul Nguewa 75, Vikram Niranjan 76, Iskra Alexandra Nola 77, Marija Obradović 78, Rónán O’Caoimh 55, Nazife Öztürk 80, M. Ramiro Pastorinho 81, Panagiotis Petrou 82, Mariana Peyroteo 83, Miguel Reina Ortiz 84, Silvia Riva 85, João Rocha-Gomes 42, Cornelia Melinda Adi Santoso 87, Tugce Schmitt 88, Rajesh Shigdel89, Rannveig Sigurvinsdottir 90, Joan B. Soriano 12, Ana Catarina Sousa 81, Maximilian Sprügel 93, Paschalis Steiropoulos 73, Fimka Tozija 95, Brigid Unim 43, Bram Vandeninden 1, Orsolya Varga 87, Milena Vasic 99, Susana Viegas 83, Rafael Vieira 42, Francesco S. Violante 102, Grant M.A. Wyper 103, Vahit Yigit 104, Jelka Zaletel 105, 19Norwegian Institute of Public Health, Norway, 20United Arab Emirates University, United Arab Emirates; University of Debrecen, Hungary, 21Ministère de la Santé; Direction de la santé, Luxembourg, 22Iran University of Medical Sciences, Iran, 23Polythecnical Institute of Coimbra, Coimbra Health School, Portugal, 24Aydin Adnan Menderes University, Turkey, 25Istituto di Ricerche Farmacologiche IRCCS Mario Negri, Italy, 27Sant Pau Institute for Biomedical Research, Spain, 28Oncologic network, prevention and research institute (ISPRO), Italy, 29Central Health Directorate, Inpatient and Outpatient Care Service, Italy, 30Universidade NOVA de Lisboa, Portugal, 31Institute of Public Health for FBiH, Bosnia and Herzegovina, 32National Research Council, Italy, 33University of Malta, Malta, 35Faculty of Medicine of Tunis, Tunisia, 36Medical University "Prof. Dr Paraskev Stoyanov", Bulgaria, 37Ashkelon Academic College, Israel; Ben Gurion University of the Negev, Israel, 38Cyprus University of Technology, Cyprus, 39James Cook University, Australia, 41Charité—Universitätsmedizin, Germany, 42University of Porto, Portugal, 43National Institute of Health ISS, Italy, 44Universidad Pablo de Olavide, Spain, 45Institute and Faculty of Actuaries, United Kingdom, 46Finnish Institute for Health and Welfare, Finland, 48Ondokuz Mayis University, Turkey, 49Instute for Work and Health, Switzerland, 50University of Zurich and Balgrist University Hospital, Switzerland, 51Technological University Dublin, Ireland, 52University of Belgrade, Serbia, 53University of Kragujevac, Kragujevac, Serbia, 54University of Catania, Italy, 55University College Cork, Ireland, 56Berlin Institute of Health, Germany, 57Kirsehir Ahi Evran University, Turkey, 58Center for Infectious Disease Control, National Institute of Public Health and the Environment (RIVM), The Netherlands, 59Al Farabi Kazakh National University, Kazakhstan, 60Università degli Studi di Milano, Italy, 61Instituto Politécnico de Lisboa, Portugal, 63ISDE, Italy, 65Central Tuscany Local Healthcare Authority, Italy, 77University of Zagreb, Croatia, 68University of Crete, Greece, 69Centre for Disease Prevention and Control, Latvia, 70Institute for Maternal and Child Health IRCCS Burlo Garofolo, Italy, 72National Institute of Health Dr. Ricardo Jorge, Portugal, 73Democritus University of Thrace, Greece, 74The London School of Hygiene & Tropical Medicine, England, 75University of Navarra, IdiSNA, Spain, 76University College Dublin, Ireland, 78University of Banja Luka, Bosnia and Herzegovina, 80Antalya Training and Research Hospital, Turkey, 81University of Évora, Portugal, 82University of Nicosia, Cyprus, 83NOVA Medical School, Portugal, 84University of South Florida, USA, 85St Mary's University, London, 87University of Debrecen, Hungary, 88Maastricht University, The Netherlands, 89University of Bergen, Norway, 90Reykjavik University, Iceland, 93Friedrich-Alexander-Universität, Germany, 95Saints Cyril and Methodius University of Skopje, North Macedonia, 99Institute of Public Health of Serbia "Dr Milan Jovanovic Batut", Serbia, 102University of Bologna, Italy, 103Public Health Scotland, Scotland, 104Suleyman Demirel University, Turkey, 105National Public Health Institute Slovenia and University Medical Centre Ljubljana, Slovenia