Résumé: There is an increasing need to understand community-level or whole-ecosystem responses to multiple stressors since the impacts of multiple stressors on marine systems depend not only on species-level responses, but also on species interactions and ecosystem structure. In this study, we used a multi-model ecosystem simulation approach to explore the combined effects of fishing and primary productivity on different components of the food-web across a suite of ecosystems and a range of model types. Simulations were carried out under different levels of primary productivity and various fishing scenarios (targeting different trophic levels). Previous work exploring the effects of multiple stressors often assumed that the combined effects of stressors are additive, synergistic or antagonistic. In this study, we included a fourth category “dampened”, which refers to less negative or to less positive impacts on a given ecosystem component compared to additive effects, and in contrast to previous studies, we explicitly considered the direction of the combined effects (positive or negative). We focused on two specific combined effects (negative synergism and positive dampened) associated with the ecological risk of resultant lower fish biomass than expected under additive effects. Through a meta-analysis of the multi-models’ simulation results, we found that (i) the risk of negative synergism was generally higher for low-trophic-level (LTL) taxa, implying that following an increase of fishing pressure on a given LTL stock, the subsequent decrease of biomass under low primary productivity would be higher than expected when fishing is the sole driver and (ii) the risk of positive dampened effects was generally higher for high-trophic-level (HTL) taxa, implying that given a management measure aimed at reducing the impact of fishing on HTL stocks, the subsequent rebuilding of these stocks would be slower than expected if only fishing were considered. Our approach to categorizing and exploring cumulative risk can be applied to evaluate other community properties and indicators and our findings could provide guidance in fisheries management.