We propose a statistical approach based on multivariate analysis and general linear models to test the consequences of removal of reserve status from the Abore Reef fish assemblage of the Abore Reef reserve, New Caledonia. Consequences on fish density were examined at the level of fish assemblage, accounting for habitat. Fish counts and habitat-related information were collected before and after the removal of reserve status, both inside and outside the reserve. Species groups were constructed on the basis of additional information on behavioural characteristics and taxonomy. The impact of the removal of reserve status was assessed by using 2 habitat proxies corresponding to 2 spatial scales and by considering several criteria for grouping species. Habitat appeared to be a determining factor in explaining density variations. Significant positive effects were found for several species groups such as macrocarnivores, piscivores, herbivores, Lethrinidae and Siganidae. A counterintuitive negative effect was observed for Acanthuridae. No significant differences between the area which remained closed to fishing and the area open to fishing were found for the other species groups. The approach allows for a synoptic diagnosis of the impact of changes in reserve status at the fish assemblage level. It could be used to develop and select potential indicators for monitoring such impacts on fish assemblages in coral reef ecosystems and in other contexts.