The analysis included 18,476 people. Such risk estimates can support research but are not a diagnosis or a precise prediction for an individual.