Addressing challenges in selecting dietary risk factor-outcome pairs for health impact assessments: Developing an evidence database and novel approach to dose-response extraction

Dietary health impact assessments are valuable decision-support tools for action on population-wide dietary shifts. These assessments rely on epidemiological evidence to estimate associations between dietary risk factors and health outcomes. However, the review and selection of evidence poses several challenges including a lack of systematic approaches to risk-outcome pair selection and ability to extract harmonised dose-response data from scientific papers. To address these challenges, this work aimed to 1) develop a database of dietary risk factor dose-response curves by updating the Nordic Nutrition Recommendations 2023 (NNR2023) evidence base and 2) propose a novel extraction method for non-linear dose-response curves. The NNR2023 scoping review methodology was replicated and risk-outcome pairs were selected according to modified NNR 2023 criteria. Non-linear relationships were estimated by extracting data points with an open-source graph reader and fitting piecewise constant functions to the extracted points. From the selected updated studies and original NNR2023 evidence, 101 risk-outcome associations were included, of which 44 were non-linear, and one quarter included beverages. Around half of the associations reported levels of at least mid-range evidence certainty. The provided database and methodology contribute to increased transparency and provide a standard approach to evidence selection and use in health impact assessments.