Defining the Phenotype and Interpreting Clinical Findings in Machine-Learning Studies of Diet and Restless Legs Syndrome

1

Hu et al. used machine-learning methods to associate micronutrient and whole-food profiles with restless legs syndrome (RLS) among individuals with coronary heart disease (CHD).
2

RLS was defined by reported use of RLS-specific medication in the preceding 30 days, not by symptom-based diagnostic criteria, potentially excluding untreated or undiagnosed cases.
3

CHD was defined by self-reported coronary heart disease, angina, or myocardial infarction rather than uniform clinical adjudication, limiting generalizability.
4

Diet was assessed using two nonconsecutive 24-hour recalls, which may incompletely capture habitual intake because of day-to-day variation and self-report error.
5

SHAP values describe contributions to model predictions, not causal effects; the findings concern dietary characteristics associated with model-predicted treated RLS, not RLS risk generally.

1