There is a consistent relationship between body mass index and sleep quality. However, the directionality and possible confounding factors of this relationship are unclear. Our aim is to confirm the association between sleep quality and body mass index, independent of possible genetic confounding, as well as to provide some indirect inferences about the directionality of this association. The co-twin study design was used to analyse the body mass index–sleep relationship in a sample of 2150 twins. We selected two parallel sub-samples of twins discordant for body mass index (n = 430 pairs), or discordant for sleep quality (n = 316 pairs). Sleep quality and body mass index showed an inverse relationship (b = 0.056, P = 0.032) in the global sample. When twins discordant for body mass index were selected, this association maintained a similar effect size and statistical significance, at all levels of the case–control analysis (all discordant pairs b = 0.173, P < 0.001; dizygotic twins b = 0.174, P = 0.002; monozygotic twins b = 0.173, P = 0.050). Nevertheless, when twin pairs were selected on the basis of their discordance for sleep quality, the association between body mass index and sleep quality appeared weaker and lost significance (b = 0.021, P = 0.508). The analyses including only dizygotic (b = 0.028, P = 0.526) or monozygotic (b = 0.001, P = 0.984) pairs produced similar non-significant results. Our results confirm the relationship between sleep quality and body mass index, even after applying high levels of control, including genetic factors. Moreover, this study suggests a possible directionality of this relationship, such that sleep quality would strongly affect body mass index, while the opposite would be less robust and consistent in non-clinical samples.