Invited Commentary: Can Estimation of Sodium Intake Be Improved by Borrowing Information From Other Variables?

Estimation of dietary sodium intake is problematic. The most accurate measure is average sodium excretion from multiple 24-hour urine collections, but such an approach is impractical. Using data from the Women's Health Initiative, Prentice et al. (Am J Epidemiol. 2017;186(9):1035-1043) assessed the relationship of calibrated estimates of sodium and potassium excretion with cardiovascular outcomes. The calibrated estimates were a function of self-reported sodium-to-potassium ratio from a food frequency questionnaire, age, body mass index, race, supplement use, smoking status, educational level, income, and aspirin use. In general, associations with outcomes using the calibrated estimates were in the expected direction: direct for the sodium-to-potassium ratio and sodium intake and indirect for potassium. The unexpected associations were an increased risk of hemorrhagic stroke with lower sodium-to-potassium ratio and sodium intake and increased risk with higher potassium intake, along with a null relationship of sodium intake with ischemic stroke. Overall, our assessment is that the authors have improved the estimation of mean dietary sodium and potassium intakes. However, more work is needed to show that calibrated estimates actually improve estimation of future clinical events. If this methodological issue can be successfully addressed, their approach has the potential to improve estimation of dietary sodium and potassium intakes in observational studies.