Genetic association studies have focused on testing additive models in cohorts with European ancestry. Little is known about recessive effects on common diseases, specifically for non-European ancestry. Genes & Health is a cohort of British Pakistani and Bangladeshi individuals with elevated rates of consanguinity and endogamy, making it suitable to study recessive effects. We imputed variants into 44,190 genotyped individuals, using two imputation panels: a set of 4,982 whole-exome-sequences from within the cohort, and the TOPMed-r2 panel. We performed association testing with 898 diseases from electronic health records. We identified 185 independent loci that reached standard genome-wide significance (p<5x10-8) under the recessive model and had p-values more significant than under the additive model. 140 loci demonstrated nominally-significant (p<0.05) dominance deviation p-values, confirming a recessive association pattern. Sixteen loci in three clusters were significant at a Bonferroni threshold accounting for multiple phenotypes tested (p<5.5x10-12). In FinnGen, we replicated 44% of the expected number of Bonferroni-significant loci we were powered to replicate, at least one from each cluster, including an intronic variant in PNPLA3 (rs66812091) and non-alcoholic fatty liver disease, a previously reported additive association. We present novel evidence suggesting that the association is recessive instead (OR=1.3, recessive p=2x10-12, additive p=2x10-11, dominance deviation p=3x10-2, FinnGen recessive OR=1.3 and p=6x10-12). We identified a novel protective recessive association between a missense variant in SGLT4 (rs61746559), a sodium-glucose transporter with a possible role in the renin-angiotensin-aldosterone system, and hypertension (OR=0.2, p=3x10-8, dominance deviation p=7x10-6). These results motivate interrogating recessive effects on common diseases more widely.