Extreme Variability in Urinary Sodium Excretion: Time to Stop Use of Spot Urines to Predict Clinical Outcomes

Accurate estimation of usual sodium intake is extremely difficult to accomplish in clinical and epidemiologic research (Table). The ‘gold’ standard is 24 hour urinary sodium excretion. From a single 24 hour urine collection, one can accurately estimate mean urinary sodium intake of a population but not the mean urinary sodium intake of individuals in that population. Because of high intra-individual variability, estimating average intake of individuals requires multiple 24 hour urines which are rarely collected. Importantly, the source of variability is not just variability in dietary sodium intake. In controlled feeding studies, in which sodium intake is held constant, there is tremendous intra-individual variability in urinary excretion, likely the result of physiological factors that regulate sodium excretion.^1,2 Nonetheless, investigators have repeatedly attempted to estimate usual intake of individuals from single collections, either spot or 24 hour, and use such estimates to predict clinical outcomes in prospective studies.^3–5

Table:

Methods to Estimate Usual Sodium Intake of Individuals

Optimal Multiple, high quality 24 hour urine collections Inaccurate Single 24 hour urine collection* Spot, overnight or timed urine Estimating equations (e.g. Kawasaki and INTERSALT equations) 24 hour dietary recall Food frequency questionnaire

^*A single, high quality 24 hour urine collection is adequate to estimate the mean intake of a population.

A large and robust body of evidence from randomized trials has documented that lowering sodium intake reduces blood pressure, the leading cause of preventable mortality worldwide.^6,7 However, the results of several studies based on spot or a single 24 hour urine collection have reported paradoxical findings that low intake of sodium is associated with a higher risk of clinical outcomes. Initially, the findings of these observational studies have shocked the research community – low sodium intake (or excretion) is not associated with cardiovascular disease and appears in some studies to be harmful! Subsequently, methodologic investigations have identified the apparent cause of the paradoxical findings,^8–11 which is typically, although not exclusively, related to measurement of sodium intake, usually a combination of suboptimal assessment method and an insufficient number of assessments. The latest assessment method, now in vogue, is use of spot urine collections in combination with estimating equations, often in large cohort studies.

The study by Re and colleagues, published in this issue of the journal, is a methodologic investigation to understand the relationship of spot urinary sodium concentration, and corresponding estimates of 24 hour excretion from common estimating equations (the INTERSALT and Kawasaki equations), with outcomes.¹² The investigators’ focus was within-person variability in spot urinary sodium excretion and its impact on associations with systolic blood pressure and risk of incident cardiovascular disease in the prospective UK Biobank study. Spearman correlation coefficients were used to assess within-person variability in spot urinary sodium between measurements at baseline and 4 years later. The investigators assessed the relationship of baseline sodium excretion with three outcomes: 1) baseline blood pressure in cross-sectional analyses, 2) future blood pressure measured nine years later, and 3) the risk of cardiovascular disease over nine years.

The within-person variability in urinary sodium was extreme, with a correlation coefficient of 0.35 between measurements collected four years apart. While spot urinary sodium measurements were associated cross-sectionally with blood pressure at baseline, associations of baseline sodium excretion with future blood pressure nine years later and with risk of cardiovascular disease over nine years were null. The authors attribute the null findings in their prospective analyses to the extreme within-person variability in urinary sodium excretion.

Other notable findings were the general concordance of results when the exposure was estimated 24hr sodium excretion using the INTERSALT and Kawasaki equations. Second, within-person variability differed considerably for exposures causally associated with BP and/or cardiovascular disease – extremely high for weight (0.95), body mass index (BMI; 0.92) and alcohol (0.86), yet extremely low for electrolyte concentrations of potassium (0.26) and sodium concentrations (0.35). Such findings highlight the challenges of attributing the fraction of hypertension to a specific cause – obesity will trump sodium intake, because of the high precision of BMI and low precision of sodium excretion.

This study has several important strengths, including the design of the UK Biobank with its long duration of follow-up, the large number of clinical outcomes, and importantly the resurvey conducted 4 years after baseline. The latter allows for adjustment for regression-dilution bias, as well as formal evaluation of the extent of intra-individual variability. Still, adjustment for regression-dilution bias does not mitigate the effects of systematic errors in urinary sodium excretion.

Limitations include the lack of 24 hour urines. While it is challenging to collect 24 hour urines in large cohort studies such as the UK Biobank, the optimal study design for understanding the implications of the use of spot urines in prospective analyses would be to compare risk prediction from spot and multiple 24 hour urines. It is noteworthy that with sufficient resources and commitment, it is possible to collect multiple 24 hour urine collections in large-scale observational studies and trials.^13–15 Future cohort studies should consider collection of multiple 24 hour urine collections, perhaps in sub-cohorts. A second issue, not specific to the UK Biobank, is the inability to understand the role of important factors that potentially confound the associations of sodium with outcomes (caloric intake and physical activity, both measured poorly in epidemiological studies) and factors that contribute to high variability in sodium excretion (components of the renin-angiotensin-aldosterone system and other physiological factors that regulate sodium excretion).

This study and its demonstration of the methodologic perils of using spot urines to examine the relationship between sodium and clinical outcomes has implications for researchers, editors, policy makers and media. Large studies with suboptimal methods garner considerable citations and media attention, even though their findings are likely spurious. It is important to avoid conflating huge sample size with scientific truth. In prospective observational studies, large sample sizes cannot make up for flawed data collection methods. Rather, policy makers should rely on the robust, direct association of sodium intake with blood pressure and avoid the trap of hedging their recommendations based on those observational studies with methodologic shortcomings.

In conclusion, we concur with Re and colleagues that the effects of extreme within-person variability in prospective cohort studies can lead to erroneous claims about hazards of sodium reduction or a lack of benefit, even if the sample size is large. At this time, it remains appropriate to base dietary sodium recommendations on the robust body of trial evidence that documents a direct, progressive relationship of sodium intake with blood pressure. Given the tremendous burden of BP-related cardiovascular disease worldwide, it is vitally important to proceed ahead with population-wide efforts to reduce sodium intake.