Response by Olde Engberink et al to Letter Regarding Article, “Use of a Single Baseline Versus Multiyear 24-Hour Urine Collection for Estimation of Long-Term Sodium Intake and Associated Cardiovascular and Renal Risk”

In Response:

We thank Graudal and Mente for their interest in our article.¹ We would like to clarify the authors’ comment that the confidence intervals indicate no outcome difference between baseline and follow-up measurements. The overlapping confidence intervals do not indicate absence of significance between both regression lines because they reflect confidence intervals that accompany the hazard ratios compared with a 2-g sodium intake. Moreover, we want to stress that both regression lines are based on the same dependent variable (ie, estimated sodium intake) and do not represent 2 different interventions. Therefore, these data seem to be a logical consequence of the large changes in sodium excretion between baseline and follow-up.

We agree that the hazard ratios in Table 2, which are based on categorical data, are not identical to the U-shaped curve of the natural spline analysis that is based on a continuous variable. We included both comparisons to enable comparison of our data with previous studies that have used both categorical and continuous data but recognize that the analysis of continuous data is the most valuable. It is important to note that these data demonstrate inconsistencies between baseline and follow-up data in both the categorical and continuous data analyses.

Our outcome analysis is based on the landmark approach that excludes subjects who had an event or were lost to follow-up before estimation of sodium intake was completed. This approach eliminates potential bias that is induced by patients who had developed an event or were lost to follow-up quickly after inclusion. Such patients may represent a relatively sick population in which sodium intake may reflect their health status rather than long-term intake. We agree with the authors that this analysis may limit power. Yet for the 1- and 5-year analyses, we had to exclude 6 to 12 and 87 to 123 subjects, respectively, depending on the outcome of interest. When including all patients, we nonetheless observed similar inconsistencies in sodium intake, underlining the robustness of the results.

The authors comment that the inaccurateness of a single estimate of sodium intake is caused by random error. This statement is highly questionable because the magnitude of the inconsistency in sodium intake is far beyond random error. We found that the observed coefficient of variation of individual sodium intake was 26% in our study, which is much higher than data reported for a traditional risk factor such as plasma total cholesterol (3%).² Also, the authors state that none of the hazard rates reverse direction, indicating that change in hazard rate is of minor importance. We disagree with this remark. When analyzing 5-year data instead of baseline data, hazard rates for cardiovascular outcome increase from 1.09 (confidence interval, 0.61–1.95) to 1.80 (confidence interval, 1.03–3.13), which represent a significant and clinically relevant change.

In conclusion, a single measurement is not sufficient for the estimation of individual sodium intake and associated cardiovascular risk. Misclassification of sodium intake by the use of a wobbly parameter as a single 24-hour urinary sodium excretion induces changes in hazard rates that cannot be explained solely by random error.