Salt and Cardiovascular Disease: Insufficient Evidence to Recommend Low Sodium intake

Martin O'Donnell; Andrew Mente; Michael H. Alderman; Adrian J.B. Brady; Rafael Diaz; Rajeev Gupta; Patricio López-Jaramillo; Friedrich C. Luft; Thomas F. Lüscher; Giuseppe Mancia; Johannes F.E. Mann; David McCarron; Martin McKee; Franz H. Messerli; Lynn L. Moore; Jagat Narula; Suzanne Oparil; Milton Packer; Dorairaj Prabhakaran; Alta Schutte; Karen Sliwa; Jan A. Staessen; Clyde Yancy; Salim Yusuf


Eur Heart J. 2020;41(35):3363-3373. 

In This Article

Can we Measure Sodium (Salt) Intake Adequately?

A key obstacle to making specific, individual-level recommendations on sodium intake (e.g. <2.3 g/day) is the lack of a valid and reliable method to objectively quantify sodium intake in individuals. The use of multiple 24-h urine collections is the accepted reference method, but this approach is impractical in clinical practice and for very large population studies.[33] Dietary assessment tools, including food frequency questionnaire (FFQs), 24-h dietary recalls, and dietary records, while more convenient, are time-consuming and difficult to complete with a high degree of accuracy.[34] Dietary records and country-specific validated 24-h dietary recall measures have been reported to provide a reasonably good estimate of sodium intake and appear to be superior to FFQs which provide poor estimates of individual intake.[35,36] Finally, subjective self-categorization of individuals into high, medium, and low sodium intake is unreliable.[37] All dietary questionnaires, especially FFQs, require validation in different populations with distinct dietary patterns and different sources of sodium intake.

The absence of a convenient and valid method to estimate sodium intake in individuals has implications for current recommendations. Unlike other risk factor recommendations (e.g. blood pressure, glucose, or cholesterol), members of the public are unable to access valid and reliable estimates of their sodium intake, and are therefore, unable to interpret or implement a recommendation to consume specific levels of salt per day.

For population-level studies, large sample sizes are essential to minimize random error and obtain sufficiently precise estimates of sodium intake. In this context, obtaining even single 24-h urine collections can be challenging, especially for generalizable national and international population studies. In one single country volunteer study,[38] the non-completion rate for 24-h urine collection was ~18%. Given the importance of developing a simple, reliable, and valid measure of sodium intake, considerable attention has been given to formula-derived estimates using single measurements of urinary sodium (conceptually similar to estimating eGFR).[39]

A number of formula-based approaches have been published, including one for use with a fasting early morning sample and others for use with random (or spot) urines.[40,41] While none of these approaches is appropriate for individual-level estimates of sodium intake, they have been used for population-level estimates in large epidemiologic studies.[29,32] For example, the PURE study examined sodium intake in over 100 000 individuals from 17 countries based on formula-derived estimates of 24-h urinary excretion of sodium from a single fasting urine sample (as a surrogate for intake).[29,42] Intraclass correlation (ICC) for single urine-formula-derived estimates of 24-h urinary sodium excretion vs. actual 24-h collections are good (e.g. 0.71 in the PURE validation study).[42]

Use of the Kawasaki formula (estimated from fasting urine estimate of sodium and clinical variables) to calculate 24-h sodium intake from a fasting measurement has been criticized.[5,40,43,44] However, these criticisms rely heavily on studies that misused the formula by not using a fasting urine sample, using an incorrect timing of urine collections or not following protocol for the reference method (e.g. did not check for or exclude those with incomplete 24-h urine collections), resulting in biased estimates due to incorrect methodology.[40,43,44] The World Health Organisation (WHO) considers spot/fasting urine assessment of sodium intake for population studies to be appropriate for measuring mean sodium intake.[45]

To demonstrate associations between an essential nutrient and a health outcome, the method of measurement needs to categorize individuals into different groups of mean sodium intake with methods suitable for large generalizable studies. Given that the rates of cardiovascular events and mortality in healthy populations are low (<1% per year), very large sample sizes are required to detect differences between categories of sodium intake, especially since the relative differences in outcome rates between those with high, moderate, or low sodium intake are expected to be modest (e.g. relative risk of 10% to 20%). Table 2 illustrates approximate sample sizes within sodium intake categories that would be required to detect different relative risks in primary and secondary cardiovascular prevention populations. For example, a study with over 10 000 participants per sodium intake group followed for 5 or 10 years would be needed to detect a 15% relative risk reduction in a primary prevention population. Therefore, measurement approaches with adequate validity and reliability that are sufficiently practical and feasible to use in large samples size are needed to provide much needed evidence.

We employed the designation of low, moderate, and high sodium intake (Table 1), which are based on current recommended intake (low), average intake (moderate), and above average intake (high), and selected categories that equate to teaspoons of salt, for ease of practical interpretation.


  1. The majority of the world consumes 2.3–4.6 g/day of sodium with fewer than 5–10% consuming <2.3 g/day, so there is limited human experience on the long-term associations of very low sodium intake with health outcomes.

  2. A specific sodium intake target of <2.3 g/day cannot be implemented at an individual level, as there is no feasible, valid method for estimating sodium intake in individuals.

  3. Estimating population-level mean intakes is feasible with validated formula-derived estimates with single urine samples, when used appropriately (e.g. Kawasaki formula with fasting morning sample).

  4. Dietary (24 h) recalls and records may be used to estimate mean intake but tend to underestimate sodium intake (compared with 24-h urinary collections) and are impractical in large studies. All self-report methods are subject to reporting bias.