Document resource
The letter by Dr Campbell1 repeats criticism of the Kawasaki formula for estimation of sodium intakes to which we have responded in detail earlier. Dr Campbell argues that our results showing a U-shaped association between sodium intake and premature ventricular contractions are invalid for two reasons. First, the estimation of sodium intake using the Kawasaki formula is subject to random error. However, random error would only lead to reduced statistical power and thus a weakened, but not substantially different, association between estimated sodium intake and premature ventricular contractions. Second, Dr Campbell suggests that systematic error in sodium intake estimation could produce a U-shaped association regardless of urine sodium input values. This claim is based on studies where factual urine sodium values have been substituted with a constant set at the sex-specific mean to generate sodium excretion estimations that are not based on factual urine sodium measurements. However, as we have shown in a recent study where we reported a J-shaped association between sodium intake and atrial fibrillation risk,2 this results in non-factual sodium intake estimates that are highly correlated with the factual measurements, and therefore it is not surprising that these non-factual estimates produce similar associations with outcomes to the factual ones. When we used randomly generated urine sodium values in the ONTARGET/TRANSCEND cohort, we found weaker correlations between the non-factual and factual sodium intake measurements and no association between the non-factual sodium intake estimates and atrial fibrillation. These results call into question the validity of the analyses with non-factual urine sodium measurements.