Multiple 24-Hour Urinary Sodium and Potassium and Cardiovascular Disease

Abstract

Background

The relation between sodium intake and cardiovascular disease (CVD) remains controversial, partially due to inaccurate assessment of sodium intake. 24-hour urinary excretion over multiple days is considered the optimal method.

Methods

We included individual participant data from 6 prospective cohorts of generally healthy adults in which sodium and potassium excretion were assessed by at least two 24-hour urine collections. Of 10,709 participants (54.2% women; mean [SD] age, 51.5 [12.6] years), 571 incident CVD events (myocardial infarction, coronary revascularization, and stroke) were ascertained during a median follow-up of 8.8 years (incidence rate: 5.9 per 1000 person-years). We analyzed each cohort using consistent methods and combined results using random-effects meta-analysis.

Results

Median 24-hour urinary sodium (10^th-90^th percentile) was 3,270 (2,099-4,899) mg. Higher sodium, lower potassium excretion and higher sodium-to-potassium ratio were all associated with higher hazard of CVD after controlling for confounding factors (all P-values ≤0.005). The hazard ratio [HR] comparing top with bottom quartiles was 1.60 [95% confidence interval [CI]: 1.19-2.14] for sodium, 0.69 [0.51-0.91] for potassium and 1.62 [1.25-2.10] for sodium-to-potassium ratio. Each 1,000 mg/day increment in sodium excretion was associated with an 18% increase in CVD hazard (95%CI: 8%-29%) and each 1,000 mg/day increment in potassium excretion was associated with 18% lower hazard (95%CI: 6%-28%). These findings may support reducing sodium from current intakes and increasing potassium intakes to lower CVD risk.

Conclusions

Higher sodium and lower potassium intakes, measured in multiple 24-hour urine samples, were associated with higher risk of CVD in a dose-response manner.

High sodium consumption, a major cause of hypertension, is considered a leading dietary risk factor for cardiovascular disease (CVD) globally.^1,2 Population-wide reduction in sodium intake has been suggested as a cost-effective strategy to prevent hypertension and CVD.^3,4 However, several cohort studies have linked both estimated lower (<3,000 mg/d) and higher sodium intakes (>5,000 mg/d), mainly from spot urines, to increased CVD risk.^5–8 Methodological limitations, including the inaccurate estimates of individual sodium intakes and reverse causation related to pre-existing comorbidities or changes in health status, may play a major part in these controversial findings.⁹ Accumulating evidence suggests that daily sodium intake estimated from spot urines leads to a spuriously increased CVD risk at lower estimated sodium intake levels due to inherent systematic error in the estimating formulas and is also subject to substantial random error.^9–11 Further, a single 24-hour urine is not sufficiently representative of an individual’s usual sodium intake because of the large day-to-day variations in sodium intake. Additionally, participants may reduce sodium intake because of pre-existing illness or frail status preceding major diseases while still at an elevated CVD risk, resulting in a spurious association between low sodium intake and increased disease risk (i.e., reverse causation).⁹ Given the interrelationship of sodium and potassium, it is also important to consider their associations jointly.¹² We investigated the relationship between sodium and potassium intakes and CVD risk by combining individual-level data from studies in healthy populations with multiple 24-hour urine collections, the most accurate method to assess sodium intake.

Methods

Study Population

This study combined individual data from six cohorts: the Health Professionals Follow-up Study (HPFS), the Nurses’ Health Study (NHS), the Nurses’ Health Study II (NHSII), the Prevention of REnal and Vascular ENd-stage Disease (PREVEND) study, and the Trials of Hypertension Prevention (TOHP I and TOHP II) Follow-up Studies. More details on the study design are provided in Text 1-2 and Figure S1 in the Supplementary Appendix. The HPFS, NHS and NHSII are ongoing prospective cohort studies among U.S. health professionals and registered nurses. In the current analysis, we included a subset of 2,289 participants with two 24-hour urine samples collected between 2003-2007 and an additional subset of 1,218 with four 24-hour urine collections between 2010-2013. Incident CVD and death were followed up to June 2017 and June 2020, respectively. The PREVEND study is a prospective, population-based cohort of 8,592 adults, where participants collected two 24-hour urine samples between 1997-1998 and another two 24-hour urine samples between 2001-2003.¹³ Incident CVD and death was followed up to January 2011 and January 2017, respectively. The present study excluded 1,908 participants with preexisting chronic kidney disease. The TOHP Follow-up Studies are observational follow-up extensions of the TOHP I and TOHP II trials in which the effect of dietary supplements and lifestyle interventions, including weight loss and sodium reduction, were evaluated among 2182 and 2382 participants in TOHP I and TOHP II respectively.^14,15 There were 5 to 7 24-h urine collections in TOHP I (1987-1990) and 3 to 5 24-h urine collections in TOHP II (1990-1995). Post-trial follow-up for CVD and death beginning at the end of each trial through early 2005 for CVD and December 2013 for death.¹⁶ Individuals randomized to the sodium reduction group were excluded from the current analysis.

For all these studies, the baseline was defined as the period when 24-hour urine collections were completed. The primary analysis excluded the first year of follow-up as dietary information collected close to the diagnosis of CVD events or death may not reflect habitual sodium intakes. Out of 11,435 participants completing ≥two 24h urine collections at baseline, 626 (5.5%) did not respond to the follow-up for CVD events, and another 100 had less than one year of follow-up. Ultimately, a total of 10,709 participants were included in the primary analysis. All cohorts obtained institutional review board approval and informed consent from participants.

24-Hour Urinary Sodium and Potassium Excretion

Our primary exposure measures were individuals’ sodium and potassium excretion assessed by averaging the excretions from all available 24-hour urine samples per person. To reduce measurement error arising from under- or over-collection of 24-hour urine, we used the following exclusion criteria: 1) 24-hour urine volume <500 or >5000ml or 2) duration <20 or >28 hours when the start and end time of urine collection were available or 3) urine volume loss of >100 ml if estimated volume lost was reported. Alternative exclusion criteria were used in the sensitivity analyses.

Outcome Ascertainment

The primary outcome was CVD, defined as a composite of fatal and nonfatal coronary heart disease (CHD, including myocardial infarction [MI], coronary artery bypass grafting [CABG], and percutaneous transluminal coronary angioplasty [PTCA]) and stroke. The secondary outcomes were CHD, stroke, and all-cause mortality. The criteria used to ascertain these outcomes have demonstrated high validity in these individual studies. (see Text 3 in the Supplementary Appendix).

Statistical Analysis

Our primary analysis evaluated the associations of continuous urinary sodium and potassium excretion and sodium-to-potassium ratio with subsequent risk of CVD. We assessed the relationships for each study independently using consistent analytical methods and then synthesized the results from each study using random-effects meta-analysis (i.e., two-stage pooled analyses). For each study, hazard ratios were estimated by Cox proportional hazards models, with adjustment for potential confounding factors including age, sex, and race/ethnicity in a partially adjusted model (Model 1) and additional adjustment for educational attainment, height, baseline values of body mass index, smoking, alcohol drinking, physical activity, and history of diabetes, elevated cholesterol status, family history of CVD, as well as average 24-hour urinary potassium (for sodium as the exposure) and sodium (for potassium as the exposure) in a fully adjusted model, which is the primary analytic model (Model 2). Missing covariate values (<1% of the participants) were handled using missing indicators. Person-time was calculated from one year following completion of all 24-hour urine collections until the occurrence of CVD, death or the end of follow-up, whichever occurred first. The proportional hazard assumptions were tested and verified by including an interaction term with time in the model. Results from individual studies were pooled using random-effects meta-analysis to account for the design differences between these studies. Between-study heterogeneity was statistically assessed by I². Although our primary analyses were based on a priori hypotheses, to account for the hypothesis tests for three exposures, we present P values corrected for multiple comparisons using the Holm–Bonferroni procedure.¹⁷ All confidence intervals presented for secondary outcomes and sensitivity analyses have not been corrected for multiplicity and should not be used to infer statistical significance.

In secondary analyses, we repeated the analyses for study-specific quartile-based categorical exposures. We also investigated the associations combining all individual data in the same models stratified by study (i.e., simple pooled analysis). To assess the deviation from linearity of the associations, we used penalized splines for the exposures in the simple pooled analysis.¹⁸ We also conducted pre-defined stratified analyses and assessed the association with secondary outcomes including total CHD, total stroke, and all-cause mortality separately. The stratified analyses and analyses for sub-endpoints (i.e., CHD and stroke) were conducted using a simple pooled analysis because of small sample sizes in individual studies. In a post-hoc analysis, we assessed the association with CVD and non-CVD mortality separately using a simple pooled analysis. In sensitivity analyses, we assessed the impact of measurement error, other potential confounding factors, missing covariate values, non-response and competing events on the association (See Text 4 in the Supplementary Appendix). Yuan Ma and Frank B. Hu vouch for the data and the analysis.

Results

Characteristics of the 10,709 participants are presented by cohort in Table 1 and by study-specific quartiles of 24-hour urinary sodium excretion in Table S1. The overall median (10^th-90^th percentile range) of sodium excretion assessed by 37,896 24-hour urine samples was 3,270 (2,099-4,899) mg. Assuming non-renal loss of 7% for sodium¹⁹ and 23% for potassium²⁰, the estimated overall daily sodium and potassium intakes were 3,516 (2,257-5,268) mg and 3,292 (2,162-4,700) mg respectively. During a median follow-up of 8.8 years (interquartile range [IQR], 7.6-10.9), there were a total of 571 incident CVD events (445 CHD events including 232 with MI and 213 with coronary revascularization, 136 stroke events, 22 with both CHD and stroke, and 12 other CVD deaths; incidence rate: 5.9 per 1000 person-years). There were 1,100 deaths recorded during a median follow-up of 14.8 years (IQR, 13.7-16.7).

Table 1.

Baseline characteristics of the study population

Characteristic*	Overall (n=10709)	Included cohorts
		HPFS (n=1028)	NHS (n=1064)	NHS2 (n=1305)	PREVEND (n=5131)	TOHP I (n=1334)	TOHP II (n=847)
Calendar years of baseline measurements	-	2003-2007 or 2010-2013	2003-2007 or 2010-2013	2003-2007 or 2010-2013	1997-2003	1987-1990	1990-1995
Age (years) at baseline	51.5 ± 12.6	68.1 ± 6.1	69.2 ± 5.8	52.1 ± 5.5	47.8 ± 11.5	43.3 ± 6.3	43.5 ± 6.2
Women, %	54.2	0	100	100	54.1	28.8	32.5
Race							-
White, %	92.4	94.3	95.5	92.2	95.3	85.0	81.5
Black, %	3.9	0.5	1.5	3.7	0.9	12.7	15.2
Asian/Pacific Islander, %	1.3	0.4	0.2	0.5	2.0	1.3	1.2
Hispanic, %	0.3	0	0.1	0.5	-	0.8	1.5
Height, cm	172.0 ± 9.6	179.0 ± 6.1	164.0 ± 6.0	165.0 ± 6.5	173.1 ± 9.5	174.3 ± 9.3	174.0 ± 9.1
Body mass index, kg/m2	26.6 ± 4.5	26.1 ± 3.6	26.5 ± 5.0	27.1± 6.2	25.6 ± 4.0	27.5 ± 3.6	30.8 ± 3.1
Educational attainment
Below high school %	19.1	0	0	0	39.5	1.3	1.1
High school, %	20.0	0	0	0	26.6	35.2	37.2
College or above, %	60.8	100	100	100	34.0	63.4	61.7
Smoking status
Never smoker, %	45.5	52.7	48.5	70.7	32.4	57.4	54.8
Past smoker, %	35.9	44.3	45.6	24.3	36.1	31.8	36.8
Current smoker, %	18.6	3.0	5.8	5.0	31.6	10.8	8.4
Alcohol consumption
≤1 drink per week, %	44.1	24.1	49.0	50.5	38.9	55.2	66.8
2-6 drinks per week, %	33.1	36.2	31.4	37.9	35.7	26.4	18.5
≥7 drinks per weeks, %	22.8	39.7	19.6	11.5	25.4	18.4	14.6
Physical Activity †
<1 per week, %	25.6	9.3	25.8	28.4	25.9	31.2	32.0
1-2 per week, %	27.5	17.1	30.1	27.9	74.1	28.3	35.3
≥3 per week, %	46.9	73.5	44.1	43.8	-	40.5	32.7
Diabetes, %	2.4	4.3	7.0	4.7	1.6	0	0
Hypercholesterolemia, %	29.8	59.8	66.9	42.3	20.6	0	-
Hypertension, %	23.6	38.3	52.1	24.4	24.5	0	0
Family history of CVD, %	16.8	13.1	27.2	20.1	17.9	8.1	9.5
Number of 24h urine collections per individual	2-7	2-4	2-4	2-4	2-4	2-7	2-5
24h urinary sodium, mg	3270 (2615-4042)	3647 (2936-4542)	2843 (2254-3519)	3186 (2599-3956)	3157 (2544-3862)	3450 (2768-4252)	3961 (3200-4783)
(men)	3690 (3013-4486)	3647 (2936-4542)	-	-	3584 (2959-4300)	3689 (3038-4451)	4290 (3578-5073)
(women)	2944 (2385-3610)	-	2843 (2254-3519)	3186 (2599-3956)	2820 (2312-3414)	2902 (2357-3548)	3342 (2770-3915)
24h urinary potassium, mg	2535 (2067-3069)	2882 (2367-3500)	2301 (1831-2779)	2184 (1755-2672)	2679 (2247-3196)	2334 (1877-2899)	2368 (1968-2891)
(men)	2808 (2323-3371)	2882 (2367-3500)	-	-	2959 (2488-3461)	2511 (2064-3067)	2573 (2142-3073)
(women)	2340 (1911-2798)	-	2301 (1831-2779)	2184 (1755-2672)	2479 (2102-2931)	1962 (1595-2353)	2005 (1589-2390)
Sodium/potassium (mmol/mmol)	2.2 (1.8-2.9)	2.2 (1.7-2.8)	2.2 (1.7-2.8)	2.5 (2.0-3.4)	2.0 (1.7-2.5)	2.7 (2.1-3.3)	3.0 (2.4-3.6)
24h urinary creatinine, mg	1344 (1110-1648)	1581 (1379-1807)	1030 (909-1186)	1235 (1085-1407)	1346 (1119-1630)	1460 (1156-1761)	1725 (1382-2070)
24h urine volume, ml	1620 (1283-2028)	1705 (1307-2175)	1796 (1400-2210)	1785 (1326-2345)	1588 (1290-1940)	1459 (1159-1875)	1597 (1246-2055)
Incident CVD
No. of events	571	80	56	20	242	115	58
Follow-up, y	8.8 (7.6-10.9)	8.4 (2.8-10.7)	12.1 (5.8-13.0)	11.3 (5.8-12.5)	8.3 (7.8-8.9)	14.7 (12.4-14.7)	9.5 (7.2-9.5)

Values are means ± SD, median (25th percentile-75th percentile) or percentages unless otherwise specified; 24h=24-hour; CVD=cardiovascular disease. HPFS=Health Professionals Follow-up Study, NHS=Nurses’ Health Study, NHSII=Nurses’ Health Study II, PREVEND=the Prevention of REnal and Vascular ENd-stage Disease study, TOHP I=Trial of Hypertension Prevention I Follow-up Study, TOHP II=Trial of Hypertension Prevention II Follow-up Study.

^*Proportion of missing data: height (0.48%), body mass index (0.49%), smoking habits (0.31%), alcohol consumption (0.43%), and physical activity (0.65%).

^†Physical Activity in the PREVEND study was classified as ≤1 per week (25.9%) or >1 per week (74.1%), and was not included in the overall descriptive data.

Sodium Excretion, Potassium Excretion, and CVD Risk

Table 2 shows a graded association between sodium excretion and incident CVD (HR comparing the top with bottom quartiles in the fully adjusted model: 1.60 [95% CI: 1.19, 2.14]). Each additional 1,000 mg of daily sodium excretion was associated with a 18% increase in the hazard of CVD (adjusted HR, 1.18 [95% CI:1.08, 1.29], Figure 1). The spline analysis using the pooled data further supports a linear association over the range of sodium excretion within this population (P for linearity<0.001; Figure 2).

Table 2.

Association of sodium and potassium excretion (measured by multiple 24-hour urine samples) with CVD risk

Variables	HR for CVD risk across the quartiles of urinary biomarkers				HR (continuous exposures)	P-value
	Q1	Q2	Q3	Q4
24-hour Sodium Excretion
Sodium (mg/d, median)	2212	2942	3588	4692	per 1000 mg/d	-
CVD events/Total	111/2677	139/2677	145/2679	176/2676	-	-
HRs (model 1)	1 (ref)	1.15 [0.89, 1.49]	1.25 [0.86, 1.80]	1.40 [1.07, 1.82]	1.12 [1.04, 1.21]	-
HRs (model 2)	1 (ref)	1.25 [0.96, 1.63]	1.44 [0.93, 2.23]	1.60 [1.19, 2.14]	1.18 [1.08, 1.29]	<0.001
24-hour Potassium Excretion
Potassium (mg/d, median)	1755	2336	2784	3501	per 1000 mg/d	-
CVD events/Total	148/2682	156/2674	140/2677	127/2676	-	-
HRs (model 1)	1 (ref)	1.02 [0.81, 1.29]	0.88 [0.69, 1.11]	0.74 [0.57, 0.95]	0.85 [0.76, 0.96]	-
HRs (model 2)	1 (ref)	1.00 [0.78, 1.26]	0.86 [0.67, 1.11]	0.69 [0.51, 0.91]	0.82 [0.72, 0.94]	0.005
Sodium-to-potassium ratio
Ratio (mmol/mmol, median)	1.5	2.0	2.4	3.4	per unit	-
CVD events/Total	113/2676	116/2678	152/2679	190/2676	-	-
HRs (model 1)	1 (ref)	1.02 [0.79, 1.33]	1.43 [1.01, 2.01]	1.76 [1.37, 2.24]	1.27 [1.16, 1.40]	-
HRs (model 2)	1 (ref)	1.02 [0.78, 1.33]	1.40 [0.99, 1.99]	1.62 [1.25, 2.10]	1.24 [1.12, 1.37]	<0.001

Model 1: with adjustment for age, sex, race, and urine collection cycles (only for Health Professionals Follow-up Study, Nurses’ Health Study, and Nurses’ Health Study II).

Model 2 (primary analytic model): model 1+ educational attainment, height, body mass index, smoking habits, alcohol consumption, history of diabetes and elevated cholesterol status, family history of cardiovascular disease, average 24-hour urinary potassium (for sodium as the exposure) and sodium (for potassium as the exposure) excretions, as well as total energy intake and modified DASH diet quality score for HPFS, NHS, and NHS2 studies.

The confidence intervals reported in the table have not been adjusted for multiplicity.

Figure 1.

Forest plots for the association of sodium (left), potassium (middle) and sodium-to-potassium ratio (right) with CVD risk

* HR=hazard ratio, HPFS=Health Professionals Follow-up Study, NHS=Nurses’ Health Study, NHS II=Nurses’ Health Study II, PREVEND=the Prevention of REnal and Vascular ENd-stage Disease study, TOHP I=Trial of Hypertension Prevention I Follow-up Study, TOHP II=Trial of Hypertension Prevention II Follow-up Study.

Figure 2.

Spline plots for the association of sodium (left), potassium (middle) and sodium-to-potassium ratio (right) with CVD risk

*Hazard ratios were estimated from Cox models stratified by cohort with adjustment for age, sex, race, education attainment, height, body mass index, alcohol consumption, smoking habits, physical activity, history of diabetes and elevated cholesterol status, family history of cardiovascular disease and mutual adjustment for 24-h urinary potassium and sodium excretions. The boxplots demonstrate the distributions of exposure, with lower and upper hinges corresponding to the first and third quartiles. The whiskers extend from the hinge to values no further than 1.5*IQR from the hinge.

Higher potassium excretion was associated with lower CVD hazard (adjusted HR comparing top with bottom quartile=0.69 [95% CI: 0.51, 0.91], Table 2). Each additional 1,000 mg of daily potassium excretion was associated with a 18% reduction in CVD hazard (HR, 0.82 [95% CI:0.72, 0.94], Figure 1). The spline plot also shows a linear trend (Figure 2). Consistently, there was a 24% increase in the hazard of CVD per unit (mmol/mmol) increase in sodium-to-potassium ratio (HR=1.24, [95% CI:1.12, 1.37]). The associations were consistent across subgroups defined by age, sex, race, history of hypertension, overweight, and duration of follow-up (Figure 3). Sex-specific associations were further presented in Figure S2.

Figure 3.

Subgroup plots for the association of sodium (left), potassium (middle) and sodium-to-potassium ratio (right) with CVD risk

Secondary Analyses

Similar associations with sodium, potassium and their ratio were also observed for CHD (Table S2). The estimates for stroke showed similar linear trend but with wider confidence internals probably due to the small number of cases (Table S3). The secondary analysis in which one single Cox model was built using the combined data stratified by study yielded estimates almost identical to those from the meta-analysis for incident CVD (Table S4). Sodium excretion was not associated with all-cause mortality (HR per 1,000 mg increase=1.02 [95%CI: 0.95 to 1.10]), while both higher potassium and lower sodium-to-potassium ratio were associated with lower risk of all-causes mortality (Table S5, Figure S3). The linear trend of the association of sodium excretion with CVD mortality was consistent with the results on CVD incidence (Tables S6, Figure S4).

Sensitivity Analyses

The association of sodium excretion with CVD risk was attenuated (HR per 1,000 mg increase=1.10 [1.03, 1.17]) when only one 24-hour urine sample was used, although the association was attenuated to a lesser degree when only urine samples that were deemed complete were used (Table S7). The linear association of sodium with CVD risk remained essentially unchanged in the following sensitivity analyses: (1) additionally adjusting for baseline hypertension status, urinary creatinine excretion, other dietary factors, and neighborhood socioeconomic status separately; (2) excluding participants with diabetes and underweight respectively or excluding coronary revascularization from incident CVD endpoints and (3) imputing missing covariates using multiple imputation, accounting for non-response to follow-up using inverse probability weighting, and assessing competing risk of death using subdistribution hazard models (Table S8). The results from sensitivity analysis for potassium and sodium-to-potassium ratio were also similar to those from the primary results (Tables S9-10).

Discussion

Among 10,709 adults pooled from 6 prospective cohort studies across the United States and Europe with a median follow-up of 8.8 years, higher sodium intake measured by multiple 24-hour urine samples was significantly associated with higher risk of incident CVD in a dose-response manner from a daily sodium intake of approximately 2,000 mg to 5,000 mg. Lower potassium intake and higher sodium-to-potassium ratio were also associated with higher risk of CVD.

Several meta-analyses of prospective cohort studies and randomized trials have reported a linear relationship between sodium intake and the risk of CVD,^21–25 while a few other meta-analyses and more recent cohort studies found that both low and high sodium intakes were associated with higher risk, i.e. a J-shaped relationship ^5–8,26 A key limitation of previous studies is the assessment of sodium intake using methods that are prone to measurement errors, such as questionnaires, spot urine and single 24-hour urine samples.²⁷ The J-shaped association is likely due to confounding variables used in the equations, e.g., age, sex, body weight, urinary creatinine, all of which are related to CVD risk and may partly contribute to the increased CVD risk associated with lower sodium intake in observational studies.²⁸ Another methodological issue of previous studies is the inclusion of participants with existing chronic diseases such as heart failure,²⁹ resulting in the potential for reverse causation bias. To overcome these limitations, we combined 6 prospective cohorts with generally healthy participant populations that had obtained multiple 24-hour urine samples from the participants. There was no evidence of heterogeneity in the results across individual studies, although the estimates were weaker in the PREVEND study. Of note, the PREVEND study participants had lower sodium and higher potassium intakes than those from other cohorts. Also, the PREVEND study oversampled participants with albuminuria even though we excluded participants with chronic kidney disease. Although we found a strong and consistent association of sodium excretion with CVD events and the association with CVD mortality showed consistent linear trend, sodium excretion was not associated with all-cause mortality, which we speculate may be due to non-CVD deaths diluting the relation between sodium and mortality. In contrast, the SSaSS trial³⁰ showed that salt substitutes containing reduced sodium and increased potassium significantly reduced both CVD incidence and total mortality in a high-risk population in China in which CVD mortality accounted for nearly 60% of the deaths whereas the proportion of deaths from CVD was 22% in our study.

Our study offers new insights concerning current methodological challenges in sodium assessment. For example, when one single 24-hour urine sample was used, the magnitude of linear association with CVD was attenuated-- and it was further attenuated when the potential under-collected and over-collected urine samples were not excluded, consistent with a previous report.³¹ A single measurement is not sufficient to reflect a person’s usual sodium intake because of large day-to-day variations in sodium consumption and excretion.^32,33 When potassium was considered together with sodium, both lower sodium and higher potassium excretion were associated with a lower risk of CVD, suggesting an additive role of lower potassium and higher sodium in increasing CVD risk.¹² By combining individual data from cohort studies with multiple 24-hour urines, our study constitutes a very large study on CVD risk and sodium intake assessed by the most reliable method. By studying generally healthy populations we also reduced confounding and potential reverse causation bias. The use of individual-level data and the same methods for covariate adjustments and modeling renders the findings more robust.

Several limitations must be noted. First, although we were able to reduce the degree of heterogeneity through the use of consistent analytical approaches in individual studies, between-study heterogeneity inherent in the design of original studies exists, for example, health profiles of participants, number and intervals of 24-hour urine collections, and available information on covariates. Second, this observational study, does not allow us to rule out potential residual confounding, such as other dietary factors, energy intake, physical activity, and socioeconomic status, for which data were available in only some of the included studies. Additionally, given within-person variation in sodium excretion, the observed association may have underestimated the true association. Our study was based upon study populations mainly comprised of generally healthy white participants (see Table S11 for further descriptions), hence the generalizability of our results to other racial/ethnic populations and patients with specific conditions is not possible.

In sum, our study demonstrated a significant linear association between sodium intake measured by multiple 24-hour urine samples and CVD risk in a dose-response manner from a daily sodium intake of approximately 2,000 mg to 5,000 mg. Higher potassium intake was associated with lower CVD risk.