The case for single-patient studies