Mendelian randomisation (MR) is an epidemiological design which uses the allocation of genes from parents to children as a random source of variation in exposures of interest. To perform MR exactly, we need genetic data on children and one or both of their parents. Due to ease of data collection, however, MR has typically been performed using data from unrelated individuals. As family data becomes more widely available, there has been renewed interest in a better within-family method for MR. Our project involves the development of an (almost) exact test for MR which is explicitly based on the randomisation of genes from parents to children. We are using ALSPAC to demonstrate our new method using real data. In particular, we aim to explore the effect of childhood BMI on systolic blood pressure and risk of diabetes in adulthood, which was recently explored using the UK Biobank cohort (Richardson et al, 2020).