The mechanisms underlying the adverse cardiovascular effects of increased salt intake are incompletely understood, but parallel increases in serum sodium concentration may be of importance. The aim of this retrospective cohort study was to investigate the relationship between serum sodium, hypertension and incident cardiovascular disease (CVD). Routinely collected primary care data from the Royal College of General Practitioners Research and Surveillance Centre were analysed. A total of 231,545 individuals with a measurement of serum sodium concentration at baseline were included. Exclusion criteria were: age < 40 years; abnormal serum sodium; diabetes mellitus; prior CVD event; stage 5 chronic kidney disease; and liver cirrhosis. The primary outcome was incident CVD (myocardial infarction, acute coronary syndrome, coronary revascularisation, stroke, transient ischaemic attack or new heart failure diagnosis) over 5 years. There was a ‘J-shaped’ relationship between serum sodium concentration and primary cardiovascular events that was independent of established risk factors, medications and other serum electrolytes. The lowest cardiovascular risk was found with a serum sodium between 141 and 143 mmol/l. Higher serum sodium was associated with increased risk in hypertensive individuals, whereas lower concentrations were associated with increased risk in all individuals. Therefore, alterations in serum sodium concentration may be a useful indicator of CVD risk. Higher serum sodium could have a direct effect on the vasculature, particularly in hypertensive individuals. Lower serum sodium may be a reflection of complex volume and neuroendocrine changes.
Excessive dietary salt consumption increases blood pressure and is a risk factor for cardiovascular disease (CVD) [1, 2]. The mechanisms underlying this relationship are incompletely understood, but it has been proposed that changes in serum sodium concentration could be important [3, 4]. In randomised-controlled trials, higher salt intake is associated with small increases in serum sodium sustained over several days [3,4,5]. Furthermore, cross-sectional studies have demonstrated a positive association between serum sodium and blood pressure [6,7,8]. One hypothesis is that an increase in serum sodium elevates blood pressure by stimulating thirst and antidiuresis, thus expanding extracellular volume. However, studies in both animals and humans receiving dialysis have shown that there is a positive relationship between serum sodium and blood pressure, independent of extracellular volume status [9,10,11]. Additionally, a number of animal and in vitro studies suggest a direct effect of serum sodium concentration on the blood vessels: higher serum sodium stiffens the vascular endothelium by inhibiting nitric oxide release and promotes hypertrophy of vascular smooth muscle cells [12, 13].
Evidence for a potential relationship between serum sodium concentration and the development of CVD is limited. This is in stark contrast to the well-established association between hyponatraemia and adverse outcomes in those with existing CVD, in whom hyponatraemia is often of consequence of decompensated disease or pharmacological therapy [14,15,16,17]. Previous studies have demonstrated a positive association between increased serum sodium and estimated coronary heart disease risk, as well as surrogate markers of hypertensive heart disease in chronic kidney disease (CKD) [18, 19]. A small cohort study in older men has reported higher serum sodium to be predictive of cardiovascular events, as well as lower serum sodium . However, an important confounding factor may have been diuretic therapy. Therefore, the present study further investigated the relationship between serum sodium concentration and incident CVD in a large community-based cohort. The primary hypothesis was that increased serum sodium, within the normal physiological range, is a risk factor for primary cardiovascular events.
Subjects and methods
Study design and cohort
This was a retrospective cohort study using data from the Royal College of General Practitioners (RCGP) Research and Surveillance Centre (RSC) network. This is a database of coded primary care data from over 100 General Practitioner (GP) practices across England, comprising of around 1.8 million patients. The study protocol was approved by the Proportionate Review Sub-committee of the London—Stanmore Research Ethics Committee (REC reference: 15/LO/1865).
Routine data collected over a 10-year period between April 2005 and March 2015 were extracted using Read (version 2) and Egton Medical Information Systems (EMIS) codes. Data recorded between April 2005 and March 2010 were used to determine the baseline characteristics of study population, and the follow-up period was between April 2010 and March 2015. The study cohort comprised of a sub-population of individuals who were at least 40 years old at baseline, who had at least one measurement of serum sodium concentration during the baseline period and who were registered for the duration follow-up (unless they died). Exclusion criteria were: an abnormal serum sodium at baseline (defined as < 135 or > 146 mmol/l); diabetes mellitus or decompensated liver disease at any stage; stage 5 CKD or a cardiovascular event prior to the follow-up period. Individuals with diabetes were excluded due to the confounding effect of blood glucose on serum sodium concentration.
The primary outcome was incident CVD. This was a composite outcome comprising of the following cardiovascular events: myocardial infarction (MI); acute coronary syndrome (ACS); coronary revascularisation procedure; stroke; transient ischaemic attack (TIA); or a new diagnosis of heart failure. Cardiovascular deaths were not included as an outcome due to the limitations of mortality coding in this database.
Initial variable selection was based on traditional cardiovascular risk factors, associations with serum sodium identified by previous studies and other potential sources of confounding considered by the authors [20, 21]. Baseline data were established for each study participant: age, sex, race and smoking status were established using routinely collected registration data; deprivation quintile was assigned by matching postcodes to area scores of the United Kingdom Index of Multiple Deprivation (IMD) 2015 ; the most recent measurements of office blood pressure, height and weight for each individual were extracted and used to determine the baseline blood pressure and body mass index (BMI); prescription data were obtained from the latest prescription during the baseline period using electronic prescription data (EMIS codes); and hypertension was defined by diagnostic code or the prescription of an antihypertensive medication.
Laboratory data were derived from samples collected routinely in general practice, with the most recent results during the baseline period used for comparison. Extreme values beyond physiological limits were excluded as inputting errors. In order to limit the influence of acute alterations in serum sodium and creatinine, a penultimate result was also extracted. This was defined as the second most recent result during the baseline period that was at least 90 days prior to the first. The mean of the latest and penultimate serum sodium concentration was then calculated and rounded-up to the nearest integer. Creatinine values were used to calculate estimated glomerular filtration rate (eGFR) using the CKD Epidemiology Collaboration (CKD-EPI) formula, with the maximum estimation for each individual used as the baseline .
The study data were analysed using the statistical package R (version 3.4.1), with a P-value of < 0.05 being regarded as statistically significant. Continuous variables were analysed as grouped categorical variables to adjust for non-linear associations, with missing data included as unknown: equally spaced and clinically relevant ranges were used for categorisation, with the exception of BMI, eGFR and glycated haemoglobin (HbA1c) that were grouped using conventional cut-offs (Supplemental Table). Differences between groups at baseline were determined using the chi-squared test for categorical variables, one-way analysis of variance (ANOVA) for normally distributed continuous variables and the Kruskal–Wallis test for non-parametric data. A binary logistic regression model was then used to evaluate the association between serum sodium concentration and cardiovascular events after adjustment for confounding variables. A restricted cubic spline plot with four knots (at the 5th, 35th, 65th and 95th percentiles) was used to explore the relationship between serum sodium concentration, as a continuous variable, and the primary outcome . The cohort was subsequently categorised into serum sodium groups, based on 3 mmol/l increments, for further subgroup analysis.
The primary multivariable model included the following variables, employing a forced-entry method: age; sex; race; BMI; smoking status; IMD quintile; CKD stage; high-density and low-density lipoprotein (HDL and LDL) cholesterol; haemoglobin; packed cell volume (PCV); HbA1c; potassium; corrected calcium; phosphate; albumin; C-reactive peptide (CRP); and cardiovascular medications (anti-platelets agents, lipid-lowering therapy, diuretics, renin–angiotensin system (RAS) antagonists, beta-blockers, other antihypertensives). No significant multicollinearity was detected (variance inflation factor < 5 for all variables). Blood pressure was not included in the primary model because we hypothesised it to be a mediator of the relationship between serum sodium and cardiovascular outcomes. It was analysed separately using one-way ANOVA, to compare the mean blood pressure according to serum sodium, and linear regression models.
The distribution of serum sodium in the population showed a slight negative skew, and deviated from normal: the median concentration at baseline was 140 mmol/l (interquartile range (IQR) 139–142). The study cohort consisted of 231,545 individuals with a serum sodium within the normal physiological range (between 135 and 146 mmol/l). The median age was 58 years (IQR 49–67), 131,990 (57.0%) were female and 11,922 (5.1%) were identified as non-white race. Male patients, particularly those < 60 years old, were under-represented compared with the general population of England and Wales (Fig. 1). The median time from the most recent serum sodium measurement to the start of the follow-up period was 10 months (IQR 4–22).
In total, 97,616 (42.2%) individuals met the study criteria for hypertension. Of these, 47,535 (48.7%) were prescribed a diuretic and 51,423 (52.7%) were prescribed a RAS inhibitor. In all, 5852 (6.0%) individuals with a diagnostic code for hypertension were not prescribed an antihypertensive medication.
Table 1 outlines the baseline characteristics of the study population by serum sodium concentration. Due to the large sample size, testing for differences in baseline characteristics was statistically significant for all variables. Notably, lower serum sodium was associated with female gender and lower deprivation. These individuals were more likely to be prescribed cardiovascular medications, including diuretics. A higher serum sodium was associated with a higher BMI and non-white race: this was particularly the case for those of black race (n = 3925), in whom the sodium distribution curve was shifted to the right by 1 mmol/l. Also of note, positive linear relationships were observed between serum sodium and a number of laboratory variables at baseline: a 1 mmol/l increase in serum sodium was associated with a 0.64 g/l increase in haemoglobin (P < 0.001), a 0.23% increase in PCV (P < 0.001) and a 0.31 g/L increase in albumin (P < 0.001). In contrast, serum potassium was comparable between the groups.
Relationship between serum sodium and blood pressure
The unadjusted relationship between baseline serum sodium concentration and blood pressure is shown in Fig. 2. In normotensive individuals, there was a positive association between a serum sodium ≥ 137 mmol/l and blood pressure (0.28/0.07 mmHg higher for every 1 mmol/l increase in serum sodium, P < 0.001). However, no relationship was present after adjustment for age, sex, race, smoking, BMI and eGFR. In hypertensives, the majority of whom were prescribed antihypertensive medication, a serum sodium ≥ 137 mmol/l was significantly associated with systolic blood pressure only (0.10 mmHg higher for every 1 mmol/l increase in serum sodium, P < 0.001). However, this was no longer significant following multivariable adjustment. A serum sodium lower than 137 mmol/l was associated with higher systolic blood pressure in all individuals, and lower diastolic blood pressure in hypertensives.
The primary composite cardiovascular outcome occurred in 9070 (3.9%) individuals during the 5 years of follow-up: MI, ACS or a coronary revascularization was recorded in 3689 (1.6%) individuals; stroke or TIA occurred in 4118 (1.8%); and new heart failure was diagnosed in 1880 (0.8%). The mean time from time zero was 31.0 months (SD 17.5). Univariable associations with the primary outcome and the multivariable analysis are displayed in the Supplemental Table.
There was a ‘J-shaped’ relationship between serum sodium and primary cardiovascular events (Fig. 3). Individuals with a serum sodium concentration between 141 and 143 mmol/l had the lowest risk of a cardiovascular event and were used as the reference category for subgroup analysis (Table 2). After multivariable adjustment, a serum sodium ≤ 140 mmol/l and ≥ 144 mmol/l were both associated with an increased risk of the primary cardiovascular outcome. The relationship between lower and higher serum sodium and incident cardiovascular events remained significant after the exclusion of heart failure from the composite outcome.
On further subgroup analysis, the relationship between increased serum sodium and cardiovascular events was observed in hypertensives, but not in normotensives (Table 2). In the 97,616 individuals with hypertension, a serum sodium of ≥ 144 mmol/l was associated with a 16% higher risk of the primary outcome compared with a serum sodium between 141 and 143 mmol/l (OR 1.16; 95% CI 1.03, 1.30). In comparison, there was a significant inverse relationship between serum sodium concentrations ≤ 140 mmol/l and cardiovascular events in all individuals. This included the 133,929 individuals without hypertension, none of whom were prescribed antihypertensive medication at baseline (including diuretics). The inclusion of blood pressure in the primary model did not affect these results.
This study investigated the relationship between serum sodium concentration and primary cardiovascular events in a community-based cohort. We found there to be a significant ‘J-shaped’ association that was independent of established risk factors, medications and other serum electrolytes. Serum sodium concentrations ≥ 144 mmol/l were associated with greater cardiovascular risk in hypertensive individuals, whereas concentrations ≤ 140 mmol/l were associated with greater risk in hypertensive and normotensive individuals. Serum sodium is usually tightly controlled and so it is notable that one or two measurements in this analysis were predictive of events over 5 years in this analysis. Other studies have demonstrated there to be significant individuality in serum sodium over long periods, and so the importance of small inter-individual differences may have been underestimated [25, 26].
The cardiovascular risk associated with higher serum sodium
It has been proposed that the adverse cardiovascular effects of dietary salt intake could be mediated through small increases in serum sodium concentration [3, 4]. We found that higher serum sodium at baseline was associated with male sex, non-white race and greater deprivation. This is in keeping with cross-sectional data in England that has shown greater salt intake in these groups [27, 28]. It is also of interest that the relationship between higher serum sodium and cardiovascular events was significant only in those with hypertension. These individuals exhibit a greater blood pressure response to alterations in dietary salt intake compared with normotensive individuals, but it is unknown if there is also heterogeneity in the physiological response to serum sodium concentration .
However, although these findings support the primary hypothesis, we did not find convincing evidence for a relationship between serum sodium and blood pressure. It is possible that this was obscured by antihypertensive medication and the inaccuracy of single blood pressure readings in primary care . Alternatively, serum sodium concentration may influence cardiovascular risk independent of blood pressure: in mice, higher serum sodium is associated with endothelial activation, as well as the release of von Willebrand factor and inflammatory markers that promote thrombogenesis and atherogenesis [18, 30, 31]; although not observed in the present study, a positive association between serum sodium and cholesterol has been reported, with increased sodium promoting lipid accumulation in cultured adipocytes ; finally, higher serum sodium concentrations may be associated with aldosterone excess, which could moderate increased cardiovascular risk .
The cardiovascular risk associated with lower serum sodium
The clear demonstration of a significant inverse relationship between lower serum sodium and the risk of primary cardiovascular events is in keeping with a number of studies that have demonstrated a link between hyponatraemia and mortality in those with established CVD, especially heart failure [14,15,16,17]. It has been suggested that low serum sodium is a marker of ‘neuroendocrine activation’ including release of antidiuretic hormone, simulation of the sympathetic nervous system and activity of the RAS [32, 33]. For example, in heart failure, lower serum sodium is associated with increased RAS activity and predicts a better response to RAS inhibition . However, our study cohort was relatively healthy and excluded those with decompensated disease or prior cardiovascular events. Therefore, the higher cardiovascular risk associated with lower serum sodium concentrations (well within the normal physiological range) may warrant further explanation. Although some investigators have associated low dietary salt intake with adverse cardiovascular outcomes , lower serum sodium in this cohort is unlikely to be representative of low salt intake given that the average salt intake in the United Kingdom is 8 g/day . Experimental evidence implicates higher serum sodium with adverse vascular effects and so lower serum sodium concentration may occur secondary to other factors involved in the development of CVD.
A substantial strength of this study is that it involved a large community-based sample of patients. However, there is potential for selection bias because only individuals with a serum sodium test were included: it is unknown whether samples were taken as part of routine monitoring or due to ill health, which could have influenced the results. Other limitations include the observational nature of the analysis and the possibility of missing or inaccurately coded data. The duration of follow-up was also short and this may increase the possibility of reverse causality. Furthermore, variables not included in the multivariate model may be a cause of residual confounding bias. This includes serum chloride, which is infrequently measured in primary care but has been reported to be associated with CVD independently of serum sodium . In addition, although we included PCV to adjust for volume status, it is a crude estimate. Alterations in extracellular volume often accompany abnormalities of serum sodium and could have an effect on cardiovascular risk.
Although this analysis does not establish causation, it is in keeping with experimental evidence for direct effects of increased serum sodium on the vasculature. Whether lower serum sodium has a direct effect to moderate outcomes or is a marker for other factors, including neuroendocrine activity and volume status, is uncertain. Further investigation could determine the utility of serum sodium as a marker of cardiovascular risk, and whether it may enable individuals to be identified that would benefit from specific primary interventions.
What is known about this topic
Abnormalities of serum sodium concentration, particularly hyponatraemia, are associated with adverse outcomes in established cardiovascular disease.
In contrast, the relationship between serum sodium and primary cardiovascular events has yet to be established.
Experiments have shown that small increases in serum sodium raise blood pressure and have direct effects on the vasculature that could increase cardiovascular risk.
What this study adds
There is a ‘J-shaped’ relationship between serum sodium concentrations within the normal physiological range and incident cardiovascular events.
Higher serum sodium is associated with increased cardiovascular risk in hypertensive individuals, but not in normotensives.
Lower serum sodium is associated with increased cardiovascular risk in both normotensive and hypertensive individuals.
Patients and practices of the RCGP Research and Surveillance Centre who allow their data to be used for surveillance and research; EMIS, InPractice Computer Systems, TPP SystmOne and other computerised medical record system vendors; Apollo Medical Systems who facilitate data extraction; and Barbara Arrowsmith, SQL Developer, for support with data extraction.