Sodium intake and mortality: a case-cohort study in a prospective cohort of 150,000 adults in the Northern Netherlands

Raised blood pressure is an important cause of cardiovascular disease events and mortality.2 Findings from both randomized trials and large cohort studies suggest that sodium intake and blood pressure have a causal inverse – and importantly – linear (or, at a minimum, monotonic3) relationship.4,5 That is, on average, the lower the sodium intake is, the lower is the blood pressure. This body of evidence has led to recommendations for decreasing sodium intake (generally to below 2.3 g/day) in both blood pressure treatment and public health guidelines.6–9 The World Health Organization recommends that populations consume less than 2g of sodium per day.9 

Conflicting with the recommendation of these guidelines, several large prospective cohort studies have been published in recent years that suggest that sodium intake in the range of 2.3 to 4.6 g/day is optimal, with intakes both above and below this range being associated with a higher risk of mortality.10–14 These findings of a “J-shaped” association of mortality with sodium intake have led to controversy that was covered extensively in both the scientific and lay press.15–21 This debate has also led several prominent scientists to propose a randomized clinical trial in a controlled environment “to end the salt wars”.18 However, because of concerns around the logistical and ethical feasibility of such a trial,22 along with the required large sample size and long follow-up time,18,23 the public health community will for the foreseeable future continue to be faced with uncertainty around which level of salt intake is optimal. 

A key criticism of the cohort studies that showed a J-shaped association of mortality with sodium intake has been their imprecise and possibly biased measurement of sodium intake.24–27 Specifically, these studies have relied on equations to predict 24-hour urinary sodium excretion from spot or morning urine samples.10–14 This approach has been criticized as not only suffering from a high degree of random measurement error but also bias, with overestimation at lower sodium intake levels and underestimation at higher levels.25,28,29 A recent analysis of follow-up data from the Trials of Hypertension Prevention (TOHP) provides suggestive evidence that the J-shaped association may indeed stem from the use of such equations to estimate 24-hour urinary sodium excretion.25 However, the study was limited by a small sample size (272 deaths among 2,974 individuals in total), which did not allow for precise estimates of the shape of the association, particularly at lower levels of sodium excretion. In addition, because TOHP did not collect spot or morning urine samples, it had to assume that the sodium concentrations in the 24-hour urine samples are equivalent to those that would have been obtained in a spot or morning urine sample.

The Lifelines cohort provides a unique opportunity to help settle the controversy around the relationship between sodium intake and mortality at low to moderate levels of sodium intake for three key reasons. First, with around 150,000 adults enrolled and approximately 4,000 deaths since the baseline assessment, Lifelines is among the largest cohort studies worldwide. In fact, both the total number of participants and number of deaths are larger in Lifelines than in the PURE study, which has been a key cohort in advancing the hypothesis of a J-shaped association of mortality with urinary sodium excretion.11–13,30 Given that few individuals consume low levels of sodium and that mortality is an uncommon outcome in these relatively new cohorts, a large sample size is essential for this research. Second, Lifelines is unique for a cohort of its size to have collected a 24-hour urine sample from all participants. The PURE study and the UK Biobank cohort, the only other relevant cohorts that are of similar size to Lifelines, have had to rely on a random spot (UK Biobank) or morning urine (PURE study) sample.11,13,14,30 This study is, thus, the first to benefit from both a sufficient sample size and measured (rather than estimated) 24-hour urinary sodium excretion to determine the shape of the mortality-to-sodium association. Third, Lifelines has additionally collected a morning urine sample from all participants. Thus, a key advantage of this study is that we are able to directly determine in the same cohort whether the J-shape of the mortality-to-sodium association is due to the estimation of 24-hour urinary sodium excretion from a spot rather than 24-hour urine sample, as has been widely hypothesized.24–27 

This study has the potential to convincingly settle the controversy around the J-shaped association of mortality with urinary sodium excretion. If we demonstrate a J-shaped association when estimating 24-hour urinary sodium excretion from the morning urine samples but find an inverse linear association when using the measured 24-hour excretion, this will be strong evidence that the J-shape reported in other cohort studies is a result of biased estimation of sodium excretion. These findings would have important policy implications because they imply that observational studies on the mortality-to-sodium association do, in fact, not conflict with those from randomized trials on the linear associations between sodium intake and blood pressure, as well as blood pressure and mortality. This agreement across study types would allay concerns that reducing salt intake at the population level will lead to harm among those who currently consume low to moderate amounts of sodium. On the other hand, if we find a J-shaped association when using the 24-hour urine samples, our study will add to the urgency with which a randomized trial on the subject needs to be conducted.