Novelty and idiosyncrasy: the clay feet of rare disease descriptions

Medical understanding of rare diseases is hindered by incomplete initial descriptions. A related but often overlooked question is how many patients are required to define a phenotype. We evaluated the phenome coverage of 10 recently published gene-disease associations using the Human Phenotype Ontology (HPO) system as a template and calculated the sample sizes required to significantly associate the prevalence of a given trait with particular phenotype. The phenome coverage of the studied descriptions was below 50% in all cases (range 4 to 43%, median, 26%) and the number of patients described (median, 3; range 1 to 17 patients) was insufficient to establish all but very strong associations (relative risks > 5). These results highlight the importance of thorough phenotyping on the one hand, and of continued publication of case reports on the other, even after a disease is thought to be well known. • Descriptions of new diseases cover less than 50% of the human phenome. • Patient numbers are insufficient to identify potentially characteristic symptoms associations • Required sample sizes vary non-linearly with relative risk and underlying prevalence