Login with Google
151 Effects of low sodium diet versus high sodium diet on blood pressure, renin, aldosterone, catecholamines, cholesterol, and triglyceride
Authors
1. Niels Albert Graudal
2. Thorbjørn Hubeck-Graudal
3. Gesche Jurgens
Affiliations:
1. Department of Rheumatology VRR4242, Copenhagen University Hospital Rigshospitalet, Copenhagen, Denmark (N.A.G.)
2. Department of Nuclear Medicine & PET-Centre, Aarhus University Hospital, Aarhus N, Denmark (T.H.-G.)
3. Clinical Pharmacology Unit, Roskilde Hospital, Roskilde, Denmark (G.J.)
Contact
Email: graudal@dadlnet.dk
Summary
This study aimed to evaluate the impact of lowering sodium on blood pressure and potential side effects, including hormonal changes and lipid levels.
The researchers reviewed randomized controlled trials from various databases up to April 2018, with additional searches in March 2020. They included studies that compared low-sodium and high-sodium diets, focusing on outcomes like blood pressure and hormone levels.
Since 2003, the number of studies reviewed has grown significantly. The effects of sodium reduction on blood pressure varied by race:
- White Participants:
- Normal BP: Systolic BP (SBP) decreased by 1.14 mmHg; Diastolic BP (DBP) had no significant change.
- Hypertensive: SBP decreased by 5.71 mmHg; DBP decreased by 2.87 mmHg.
- Black Participants:
- Normal BP: SBP decreased by 4.02 mmHg; DBP decreased by 2.01 mmHg.
- Hypertensive: SBP decreased by 6.64 mmHg; DBP decreased by 2.91 mmHg.
-Asian Participants:
- Normal BP: SBP decreased by 1.50 mmHg; DBP decreased by 1.06 mmHg.
- Hypertensive: SBP decreased by 7.75 mmHg; DBP decreased by 2.68 mmHg.
The study also found that sodium reduction led to increases in various hormones and lipid levels, indicating potential side effects. Overall, the evidence was mostly of high quality, except for some aspects related to adrenaline.
Results
Effect of Sodium Reduction on Blood Pressure in White Participants
- In participants with normal blood pressure, sodium reduction from 203 to 65 mmol/day resulted in a decrease of 1.14 mmHg in systolic blood pressure (SBP) and no change in diastolic blood pressure (DBP).
- In participants with hypertension, sodium reduction from 203 to 65 mmol/day resulted in a decrease of 5.71 mmHg in SBP and 2.87 mmHg in DBP.
Effect of Sodium Reduction on Blood Pressure in Black Participants
- In participants with normal blood pressure, sodium reduction from 195 to 66 mmol/day resulted in a decrease of 4.02 mmHg in SBP and 2.01 mmHg in DBP.
- In participants with hypertension, sodium reduction from 195 to 66 mmol/day resulted in a decrease of 6.64 mmHg in SBP and 2.91 mmHg in DBP.
Effect of Sodium Reduction on Blood Pressure in Asian Participants
- In participants with normal blood pressure, sodium reduction from 217 to 103 mmol/day resulted in a decrease of 1.50 mmHg in SBP and 1.06 mmHg in DBP.
- In participants with hypertension, sodium reduction from 217 to 103 mmol/day resulted in a decrease of 7.75 mmHg in SBP and 2.68 mmHg in DBP.
Effect of Sodium Reduction on Hormones and Lipids
During sodium reduction, the study found increases in renin, aldosterone, noradrenalin, adrenalin, cholesterol, triglyceride, and LDL, while HDL remained unchanged.
Conclusion
In white participants, sodium reduction in accordance with public recommendations resulted in a mean arterial pressure (MAP) decrease of about 0.4 mmHg in participants with normal blood pressure and a MAP decrease of about 4 mmHg in participants with hypertension. The effects may be slightly greater in black and Asian participants, but the evidence is weaker. The effects of sodium reduction on hormones and lipids were more consistent than the effect on blood pressure, especially in people with normal blood pressure.
Variables
1. Blood Pressure (BP)
- Systolic Blood Pressure (SBP)
- Diastolic Blood Pressure (DBP)
2. Hormonal Factors
- Renin: A hormone that regulates blood pressure and fluid balance.
- Aldosterone: A hormone that increases sodium reabsorption in the kidneys.
- Catecholamines:
- Noradrenalin (Norepinephrine): A hormone and neurotransmitter involved in the body's fight-or-flight response.
- Adrenalin (Epinephrine): Another hormone involved in the stress response.
3. Lipid Profiles
- Cholesterol: Total cholesterol levels.
- Low-Density Lipoprotein (LDL): Often referred to as "bad" cholesterol.
- High-Density Lipoprotein (HDL): Often referred to as "good" cholesterol.
- *riglycerides: A type of fat found in the blood.
4. Demographic Factors
- Race: The study stratifies results based on race (e.g., white, black, Asian), which can influence blood pressure responses to sodium intake.
5. Duration of Intervention
- The length of time participants were on low or high sodium diets was also considered, as it may affect the outcomes.
In total, the study measures even primary variables (blood pressure, renin, aldosterone, noradrenalin, adrenalin, cholesterol, and triglycerides) while also considering demographic factors and the duration of intervention as influential factors affecting the outcomes.
Full Study
Abstract
Background
Recent cohort studies show that salt intake below 6 g is associated with increased mortality. These findings have not changed public recommendations to lower salt intake below 6 g, which are based on assumed blood pressure (BP) effects and no side‐effects.
Objectives
To assess the effects of sodium reduction on BP, and on potential side‐effects (hormones and lipids)
Search methods
The Cochrane Hypertension Information Specialist searched the following databases for randomized controlled trials up to April 2018 and a top‐up search in March 2020: the Cochrane Hypertension Specialised Register, the Cochrane Central Register of Controlled Trials (CENTRAL), MEDLINE (from 1946), Embase (from 1974), the World Health Organization International Clinical Trials Registry Platform, and ClinicalTrials.gov. We also contacted authors of relevant papers regarding further published and unpublished work. The searches had no language restrictions. The top‐up search articles are recorded under "awaiting assessment."
Selection criteria
Studies randomizing persons to low‐sodium and high‐sodium diets were included if they evaluated at least one of the outcome parameters (BP, renin, aldosterone, noradrenalin, adrenalin, cholesterol, high‐density lipoprotein, low‐density lipoprotein and triglyceride,.
Data collection and analysis
Two review authors independently collected data, which were analysed with Review Manager 5.3. Certainty of evidence was assessed using GRADE.
Main results
Since the first review in 2003 the number of included references has increased from 96 to 195 (174 were in white participants). As a previous study found different BP outcomes in black and white study populations, we stratified the BP outcomes by race.
The effect of sodium reduction (from 203 to 65 mmol/day) on BP in white participants was as follows: Normal blood pressure: SBP: mean difference (MD) ‐1.14 mmHg (95% confidence interval (CI): ‐1.65 to ‐0.63), 5982 participants, 95 trials; DBP: MD + 0.01 mmHg (95% CI: ‐0.37 to 0.39), 6276 participants, 96 trials. Hypertension: SBP: MD ‐5.71 mmHg (95% CI: ‐6.67 to ‐4.74), 3998 participants,88 trials; DBP: MD ‐2.87 mmHg (95% CI: ‐3.41 to ‐2.32), 4032 participants, 89 trials (all high‐quality evidence).
The largest bias contrast across studies was recorded for the detection bias element. A comparison of detection bias low‐risk studies versus high/unclear risk studies showed no differences.
The effect of sodium reduction (from 195 to 66 mmol/day) on BP in black participants was as follows: Normal blood pressure: SBP: mean difference (MD) ‐4.02 mmHg (95% CI:‐7.37 to ‐0.68); DBP: MD ‐2.01 mmHg (95% CI:‐4.37, 0.35), 253 participants, 7 trials. Hypertension: SBP: MD ‐6.64 mmHg (95% CI:‐9.00, ‐4.27); DBP: MD ‐2.91 mmHg (95% CI:‐4.52, ‐1.30), 398 participants, 8 trials (low‐quality evidence).
The effect of sodium reduction (from 217 to 103 mmol/day) on BP in Asian participants was as follows: Normal blood pressure: SBP: mean difference (MD) ‐1.50 mmHg (95% CI: ‐3.09, 0.10); DBP: MD ‐1.06 mmHg (95% CI:‐2.53 to 0.41), 950 participants, 5 trials. Hypertension: SBP: MD ‐7.75 mmHg (95% CI:‐11.44, ‐4.07); DBP: MD ‐2.68 mmHg (95% CI: ‐4.21 to ‐1.15), 254 participants, 8 trials (moderate‐low‐quality evidence).
During sodium reduction renin increased 1.56 ng/mL/hour (95%CI:1.39, 1.73) in 2904 participants (82 trials); aldosterone increased 104 pg/mL (95%CI:88.4,119.7) in 2506 participants (66 trials); noradrenalin increased 62.3 pg/mL: (95%CI: 41.9, 82.8) in 878 participants (35 trials); adrenalin increased 7.55 pg/mL (95%CI: 0.85, 14.26) in 331 participants (15 trials); cholesterol increased 5.19 mg/dL (95%CI:2.1, 8.3) in 917 participants (27 trials); triglyceride increased 7.10 mg/dL (95%CI: 3.1,11.1) in 712 participants (20 trials); LDL tended to increase 2.46 mg/dl (95%CI: ‐1, 5.9) in 696 participants (18 trials); HDL was unchanged ‐0.3 mg/dl (95%CI: ‐1.66,1.05) in 738 participants (20 trials) (All high‐quality evidence except the evidence for adrenalin).
Authors' conclusions
In white participants, sodium reduction in accordance with the public recommendations resulted in mean arterial pressure (MAP) decrease of about 0.4 mmHg in participants with normal blood pressure and a MAP decrease of about 4 mmHg in participants with hypertension. Weak evidence indicated that these effects may be a little greater in black and Asian participants. The effects of sodium reduction on potential side effects (hormones and lipids) were more consistent than the effect on BP, especially in people with normal BP.
The effect of a low salt diet on blood pressure and some hormones and lipids in people with normal and elevated blood pressure
Review question
In this 4th Cochrane update since 2003, studies in which participants were distributed by chance into groups with high and low salt intake were analysed to investigate the effect of reduced salt intake on blood pressure (BP) and potential side effects of salt reduction on some hormones and lipids.
Background
As a reduction in salt intake decreases blood pressure (BP) in individuals with elevated BP, we are commonly advised to cut down on salt, assuming that this will reduce mortality. However, the effect of salt reduction on BP in people with normal BP has been questioned. Furthermore, several studies have shown that salt reduction activates the salt conserving hormonal system (renin and aldosterone), the stress hormones (adrenalin and noradrenalin) and increases fatty substances (cholesterol and triglyceride) in the blood. Finally, recent observations in general populations indicate that a low salt intake is associated with increased mortality
Search date
The present evidence is current to April 2018.
Study characteristics
One hundred and ninety‐five intervention studies of 12296 individuals lasting three to 1100 days were included, which evaluated at least one of the effect measures. Participants were healthy or had elevated blood pressure. Longitudinal studies have shown that the effect of reduced salt intake on BP is stable after at maximum seven days and population studies have shown that very few people eat more than 14.5 g salt per day. Therefore, we also performed subgroup analyses of 131 studies with a duration of at least seven days and a salt intake of maximum 14.5 g.
Study funding sources
Only six studies were supported by food industry organisations.
Key results
The mean salt intake was reduced from 11.5 g per day to 3.8 g per day. The reduction in SBP/DBP in people with normal blood pressure was 1.1/0 mmHg (about 0.3%) , and in people with hypertension 5.7/2.9 mmHg (about 3%). In contrast, the effect on hormones and lipids were similar in people with normotension and hypertension. Renin increased 55%; aldosterone increased 127%; adrenalin increased 14%; noradrenalin increased 27%; cholesterol increased 2.9%; and triglyceride increased 6.3%.
Quality of evidence
Only randomised controlled trials were included and the grade of evidence was therefore considered to be high, although downgraded in some of the smaller analyses.
Authors' conclusions
Low sodium intake compared with high sodium intake for hormones | ||||
Patient or population: Participants with normal or elevated blood pressure, but otherwise healthy Settings: Hospital units Intervention: Low sodium intake Comparison: High sodium intake | ||||
Outcomes | Mean difference | No of Participants | Quality of the evidence | Comments |
|---|---|---|---|---|
Renin ng/ml/hour | 1.56 [1.39 to 1.73] N*: 1.83 [1.56 to 2.10] H*: 1.26 [1.04 to 1.49] | 2904 | ⊕⊕⊕⊕ | |
Aldosterone pg/mL | 102.4 [86.9 to 117.8] N*: 123.8 [99.5 to 148.1 ] H*: 67.4 [53.5 to 81.3] | 2506 | ⊕⊕⊕⊕ | |
Noradrenaline pg/mL | 62.3 [41.9 to 82.8] N*: 61.5 [40.4 to 82.6] H*: 54.0 [1.8 to 106.2] | 878 | ⊕⊕⊕⊕ | |
Adrenaline pg/mL | 7.55 [0.85 to 14.26] N*:9.1 [-1.2 to 19.4] H*:4.21 [-2.7 to 11.11] | 331 | ⊕⊕⊕⊝ | |
GRADE Working Group grades of evidence Very low quality: We are very uncertain about the estimate. SMD: standardised mean difference | ||||
N*: Study populations with mean SBP < 140 mmHg and mean DBP < 90 mmHg H*:Study populations with mean SBP > 140 mmHg and/or mean DBP > 90 mmHg 1. Downgraded due to the wide confidence interval and few studies | ||||
Summary of findings 5. Summary of findings: Low sodium intake compared with high sodium intake for lipids
Patient or population: Participants with normal or elevated blood pressure, but otherwise healthy Settings: Hospital units Intervention: Low sodium intake Comparison: High sodium intake | ||||
Outcomes | Mean difference | No of Participants | Quality of the evidence | Comments |
|---|---|---|---|---|
Cholesterol mg/dL | 5.19 [2.1 to 8.3] N*:5.98 (2.0 to 10.0) H*:3.95 (-1.1, 9.0) | 917 | ⊕⊕⊕⊝ | |
Trigyceride mg/dL | 7.10 [3.1 to 11.1] N*: 7.1 (2.9 to 11.4) H*: 7 (-3.7 to 17.8) | 712 | ⊕⊕⊕⊝ | |
High‐density lipoprotein (HDL) mg/dL | ‐0.3 [‐1.66, 1.05] N*: 0 (-1.6 to 1.6) H*: -1 (-3.4 to 1.4) | 738 | ⊕⊕⊕⊝ high | |
Low‐density lipoprotein (LDL) mg/dL | 2.46 [‐1.0 to 5.9] N*: 2.6 (-1.5 to 6.7) H*: 2.2 (-4.0 to 8.4) | 696 | ⊕⊕⊕⊝ | |
GRADE Working Group grades of evidence | ||||
Background
Description of the condition
Sodium is essential for life. Man can survive on a very low sodium diet of about 7‐8 mmol/d as exemplified by the Yanamamo Indians in the Brazilian jungle (Oliver 1975), but also has a very large capacity to eat sodium of about 0,4 g/kg bodyweight, i.e. about 30 g/d (1500 mmol/d). It is not precisely defined, which sodium intake is the optimal for the general health, but the present usual sodium intake (100‐200 mmol/d) is in the low end of this tolerable interval (7‐1500 mmol) .
Some health institutions (WHO 2012), and dietary recommendations (DGA 2015), assume that reduction in sodium intake from "high" to "low" levels is associated with reduction in systolic and diastolic blood pressure (SBP and DBP), which might result in a decrease in mortality. However, the definitions of “high”, “normal” and “low” sodium intake are unclear. The present usual sodium intake indicates that an intake in the interval 109 mmol/day to 209 mmol/day (McCarron 2013; Powles 2013, Table 1) would be “normal”, a high sodium intake would be above 209 mmol/day and a low sodium intake would be below 109 mmol/day, but according to the health institutions a “normal” sodium intake is below 100 mmol/day (DGA 2015), or below 87 mmol/day (WHO 2012), and a sodium intake above 100 mmol/day is “high”, whereas a “low” sodium intake is not defined. The confusion is strengthened by the use of different terms to describe salt (salt (sodium chloride) and sodium) and different units for salt/sodium intake (mg/day or mmol/day). To reduce the confusion we have shown the different definitions and units for salt and sodium intake in Table 1. In the present review, which represents a fourth update of the first meta‐analysis that includes an analysis of hormones and lipids in addition to blood pressure (Graudal 1998), updated in 2003 (Jürgens 2003),2011 (Graudal 2011), and 2017 (Graudal 2017), we use the term "sodium" and the unit "mmol".
Table 1. Sodium intake in populations
Reference | Recommended upper level* | World, lower range* | World, lower 2.5%* | World, mean* | World, Upper 97.5%* | World, upper range* |
1001 (2300)2 (5800)3 | ||||||
871 (2000)2 (5046)3 | ||||||
901 (2070)2 (5220)3 | 1091 (2500)2 (6320)3 | 1591 (3660)2 (9220)3 | 2091 (4810)2 (12120)3 | 2481 (5700)2 (14400)3 | ||
951 (2200)2 (5510)3 | 1721 (3950)2 (10000)3 | 2401 (5520)2 (13920)3 |
1mmol; 2mg sodium; 3 mg sodium chloride
Blood pressure is associated with mortality (Collins 1990).The hypothesis that a reduced sodium intake (sodium reduction) will reduce blood pressure (BP) and subsequently reduce morbidity and mortality was raised in 1904 on the basis of individual patient cases (Ambard 1904). Subsequently in 1907, these results were opposed (Löwenstein 1907). The clinical and physiological effects of salt published in studies during the first half of the 20th century were reviewed in 1949 (Chapman 1949). Consequently, scientific studies have been performed for almost 70 years before modern standard scientific randomised controlled trials (RCTs) (1000 Parijs 1973 (H)) and observational studies (Kagan 1985) were performed in humans. Despite the non‐existence of RCTs and population studies, sodium‐reduction for hypertension was introduced as a national priority in USA in 1969 (White House conference 1969‐70) and for the general population in 1977 (Dietary Goals US 1977). The subsequent RCTs and observational studies have been interpreted differently (Taubes 1998, Graudal 2005, Bayer 2012). While health institutions (IOM 2005, WHO 2012, DGA 2015) support sodium reduction below 100 mmol/day, sceptics have claimed that this recommended upper limit (UL) for sodium intake is based on a biased selection of evidence (Folkow 2011, Graudal and Jürgens 2018), and is inconsistent with Institute of Medicine’s definition of an adequate nutrient intake, which is “the approximate intake found in apparently healthy populations" (IOM 2006; Heaney 2013). For sodium "the approximate intake in apparently healthy populations" is between 90 mmol/day and 248 mmol/day (Table 1).
The present 4th updated Cochrane review is based on a meta‐analysis published in 1998 (Graudal 1998). In 1998, the usual sodium intake was known in some populations, but it was not well‐defined worldwide until recently (Table 1). Furthermore, the significance of the duration of sodium reduction was not established. In 1998, we therefore included all available randomised studies, irrespective of sodium intake and duration of intervention, assuming that the average values of multiple studies would be relevant for the general population. We separated study populations in a group of populations with normal BP to investigate the potential effect of sodium reduction in the general population and in a group of hypertensive populations to investigate the potential effect of sodium reduction as a treatment for hypertensive individuals. In a cross‐sectional multiple regression analysis including many co‐variates we found that the duration of the sodium reduction intervention had no impact on the effect of sodium reduction on BP (Graudal 1998). In addition to this cross‐sectional meta‐regression analysis, a recent meta‐analysis of longitudinal studies measuring the BP‐effect of sodium reduction several times during the observation period showed that there was no difference in SBP effect or DBP effect between week one and week six, thus estimating the time point for maximal efficacy to be one week (Graudal 2015). These results are shown in Table 2. In the Graudal 1998 analysis, the average sodium intake in the non‐reduced group was 203 mmol/day and in the reduced group it was 62 mmol/day. In the two following updates of the review, the corresponding sodium reductions were from 205 mmol/day to 64 mmol/day (Jürgens 2003) and from 202 mmol/day to 67 mmol/day (Graudal 2011). We now know (McCarron 2013; Powles 2013) that this reduction corresponds to a reduction from a high usual level to the present recommended levels below 100 mmol (IOM 2005) (WHO 2012) i.e. the present review is relevant in the context of evaluating the consequences of the present recommendations to reduce sodium intake to a level below 100 mmol/day.
Table 2. Differences in BP effects of reduced sodium intake at different time points in longitudinal studies
Data from Graudal 2015
Description of the intervention
As in the previous meta‐analyses, RCTs are included, which allocate participants to two diets with a different content of salt (sodium chloride) or to either salt tablets or placebo tablets. The compliance in the RCTs is ensured by measurement of sodium excretion in the urine, which is accepted to be a reliable surrogate for the measuring of sodium intake. The sodium content of the “high” and “low” sodium diets were not defined according to the recommendations or the usual sodium intake, but just to describe the relative content of the two randomised study populations.
How the intervention might work
Extracellular fluid volume (ECFV) is determined by the balance between sodium intake and renal excretion of sodium. A steady state exists whereby sodium intake equals output, while ECFV is expanded during salt loads and shrunken during salt restriction (Palmer 2008). Thus, the idea behind sodium reduction is to shrink ECFV in order to decrease BP. The precondition for this idea is that the smaller ECFV associated with the decrease in BP has no counteracting effects on health outcomes that could outweigh the BP‐effect.
Why it is important to do this review
A verification of the hypothetical sodium‐BP relationship would support continuous attempts to lower sodium intake in order to reduce mortality. In this context it is important to define the correct UL for a healthy sodium intake, which would have a significant impact on the strategy to lower sodium intake. For instance if 100 mmol/day is the correct UL, more than 95% of the World’s populations should reduce sodium intake, but if the UL is 250 mmol/day, only about 5% should reduce sodium intake. In the latter case, a strategy to lower sodium intake in the general population would not be necessary, which would save significant efforts and costs. The same would be the case if the sodium‐BP relationship could be denied, as indicated by many RCTs of participants with normal BP (Graudal 2017). Worst case scenario is that sodium reduction could lead to side effects, which might trump the potential BP effect and result in increased mortality, as indicated by longitudinal observational studies (Alderman 2010, Pfister 2014, O'Donnell 2014, Graudal 2014; Mente 2016). Consequently, it is important to investigate the effect of sodium reduction not only on BP, but also on potential surrogate markers for clinical side effects. In our first meta‐analysis (Graudal 1998) different races were mixed. Due to the clinical observation in the DASH study (DASH 2001) that black participants had a greater response to sodium reduction than white participants, we decided to examine this possibility. Accordingly, we stratified blood pressure outcomes by race in the first Cochrane version, and because this stratification indicated racial differences, (Jürgens 2003), we have maintained this stratification of blood pressure outcomes in white, black and Asian populations in later versions including the present one. The first Cochrane version was invited by Cochrane and based on the 1998 version. Therefore there is no protocol.
Objectives
The purpose of the present review was to estimate the influence of low‐ versus high‐dietary sodium intake on systolic blood pressure (SBP) and diastolic blood pressure (DBP), and blood concentrations of renin, aldosterone, catecholamines, cholesterol, high‐density lipoprotein (HDL), low‐density lipoprotein (LDL) and triglyceride to contribute to the evaluation of the possible suitability of sodium reduction as a prophylaxis initiative and treatment of hypertension.
Methods
Criteria for considering studies for this review
Types of studies
Randomised controlled trials (RCTs), double‐blind or open, parallel or cross‐over, allocating participants to diets with different sodium contents, the lowest defined as “low” and the highest defined as “high”, and in which the sodium intake was estimated by the 24‐hour urinary sodium excretion (either measured on the basis of a 24‐hour urine collection, or estimated from a sample of at least eight hours).
Types of participants
Study populations were included irrespective of sex and age and stratified by race (black/white/Asian populations) and blood pressure (normotension/hypertension). If race was not defined, the study population was defined according to the predominant race in the study country. If data were not reported separately by race in mixed populations, the mixed population was classified according to the predominant race in the study. Hypertension was defined as SBP ≥ 140 mmHg and/or DBP ≥ 90 mmHg. Study populations in which participants were treated with antihypertensive treatment were defined as hypertensive irrespective of baseline BP. Studies systematically investigating participants with comorbidities, for instance diabetes or heart failure, were excluded. Studies, which tested for sodium sensitivity and excluded participants, who were either sodium sensitive or sodium resistant, were also excluded.
Types of interventions
The intervention was a change in sodium intake, the study populations randomly being divided into a group eating a “low” sodium diet or a "high" sodium diet. As "low" and "high" were not specifically defined in relation to the usual intake or the definitions of the health institutions (Table 1), both diets could contain any amount of sodium, the assumption being that in most studies a "low" sodium diet would contain sodium within the low range (< 100 mmol)/day or usual range (100 mmol to 250 mmol/day) and the “high” sodium diet would contain sodium within the usual range (100 mmol to 250 mmol/day) or above the usual range (≥ 250 mmol/day). Confounding was not allowed, i.e. studies treating persons with a concomitant intervention such as an antihypertensive medication, potassium supplementation or weight reduction were only included if the concomitant intervention was identical during the low and the high‐sodium diet.
Types of outcome measures
All outcomes were considered primary outcomes.
Primary outcomes
Outcome measures were effects on SBP, DBP, renin, aldosterone, adrenaline, noradrenaline, triglyceride, cholesterol, LDL and HDL. In studies reporting BP only as mean arterial pressure (MAP), SBP was estimated from SBP = 1.3 MAP + 1.4, and DBP was estimated from DBP = 0.83 MAP – 0.7 (Tozawa 2002). Separate meta‐analyses were performed for each outcome measure. Concerning blood pressure, outcomes were stratified by race (white, black and Asian populations) and according to level of blood pressure (hypertension or normotension). All other outcome variables were stratified according to level of BP (normotension/hypertension) but not by race. A minimum duration of intervention before time of measurement was not defined, but additional analyses were performed on studies with a duration of at least 7 days.
Secondary outcomes
None.
Search methods for identification of studies
Electronic searches
The Cochrane Hypertension Information Specialist conducted systematic searches in the following databases for randomised controlled trials without language, publication year or publication status restrictions:
the Cochrane Hypertension Specialised Register via the Cochrane Register of Studies (CRS‐Web) (searched 11 April 2018);
the Cochrane Central Register of Controlled Trials (CENTRAL) (Issue 3, 2018) via the Cochrane Register of Studies (CRS‐Web) (searched 11 April 2018);
MEDLINE Ovid, MEDLINE Ovid Epub Ahead of Print, and MEDLINE Ovid In‐Process & Other Non‐Indexed Citations (searched 11 April 2018);
Embase Ovid (searched 11 April 2018);
ClinicalTrials.gov (www.clinicaltrials.gov) searched 11 April 2018);
World Health Organization International Clinical Trials Registry Platform (www.who.int/trialsearch) searched 11 April 2018).
The Hypertension Group Specialised Register includes controlled trials from searches of CAB Abstracts & Global Health, CINAHL, Cochrane Central Register of Controlled Trials, Embase, MEDLINE, ProQuest Dissertations & Theses, PsycINFO, and Web of Science.
The Information Specialist modelled subject strategies for databases on the search strategy designed for MEDLINE. Where appropriate, they were combined with subject strategy adaptations of the sensitivity and precision‐maximising search strategy designed by Cochrane for identifying randomised controlled (as described in the Cochrane Handbook for Systematic Reviews of Interventions Version 6 (Handbook 2019)). We present search strategies for major databases in Appendix 1.
Searching other resources
The Cochrane Hypertension Information Specialist searched the Hypertension Specialised Register segment (which includes searches of MEDLINE, Embase and Epistemonikos for systematic reviews) to retrieve existing systematic reviews relevant to this systematic review, so that we could scan their reference lists for additional trials.
We checked the bibliographies of included studies and any relevant systematic reviews identified for further references to relevant trials.
Where necessary, we contacted authors of key papers and abstracts to request additional information about their trials.
Searches carried out for previous versions of this review
Trial search: Parijs and colleagues published the first RCT of the effect of sodium reduction on BP in 1973 (1000 Parijs 1973 (H)). In our first meta‐analysis (Graudal 1998), a literature search in MEDLINE (1966‐through December 1997) was performed using the following combinations of search terms: 1) salt or sodium, 2) restriction or dietary, 3) blood pressure or hypertension, 4) randomized or random. We combined 1, 2, 3 and 4 and found 291 references. Of these, 76 randomised trials from 60 references met the inclusion criteria. From the reference lists of these articles and from four previous meta‐analyses (Grobbee 1986, Law 1991, Cutler 1991, Midgley 1996), an additional 23 references reporting on 39 trials were identified, resulting in a total of 83 references.
Similar searches were made for hormones and lipids changing the third search term (blood pressure or hypertension) with the hormone or lipid term resulting in additional five sub‐studies dealing with hormones and lipids (Jula‐Karanko 1992, Jula‐Mäki 19921026 Koolen 1984(2), 1104 Overlack 1993, Ruppert 1994). Of these 88 references, three dealing exclusively with diabetes patients were excluded in the 2003 update (Dodson 1989, Mühlhauser 1996, Miller 1997).
In January 2002, a repeated search was performed through December 2001, revealing an additional 12 references, of which one was excluded because it only included patients with diabetes (Imanishi 2001). Accordingly, the 2004 updated review included a total of 96 references.
In December 2009, a literature search for the 2011 update was performed from 1950 through December 2009. This search revealed a total of 511 references in Ovid MEDLINE, 282 in Ovid EMBASE and 1428 in Cochrane CENTRAL. Headlines and abstracts were read and 44 articles from MEDLINE (26 included), eight from Embase (one included) and 129 from CENTRAL (45 included) were retrieved as full‐text papers for further review. A total of 72 new references investigating at least one of the effect variables met the inclusion criteria for this review. The search was not limited to English language studies. Two studies in Italian were identified and included. During the present revision, we discovered that in a few of the previously included studies, some subgroup data were published in two papers. To avoid duplication due to including subgroup data from several papers, we included them from the main paper only. As a result, three previously included references were excluded (Steegers 1991, Ruppert 1991, Ruppert 1994). The most recent search was performed on July 21, 2011, revealing 293 additional references. After screening of titles and abstracts, four full‐text papers were retrieved, of which two contained data to be included. Consequently a total of 167 studies were supposed to be included in the 2011 updated version of this systematic review. However, in connection with the present update, a recount revealed a counting error, as the number of references in reality was 166.
During the 2017 update, we identified two studies with duplicate data, which were subsequently excluded (Jula‐Karanko 1992;Jula‐Mäki 1992), as all data could be extracted from a later paper (1110 Jula 1994 (H)).
In September 2014, a literature search for the 2017 update was performed as described in "Search methods for identification of studies". The de‐duplicated results from the searches revealed 626 articles. On the basis of titles, 549 were excluded. Seventy‐seven abstracts were read and 27 full‐text articles obtained, of which, nine fulfilled the inclusion criteria. In a supplementary search in April 15 2015, an additional 102 references were identified. Six articles were obtained, of which three fulfilled the inclusion criteria.The last updated search was performed on 7 March 2016. The de‐duplicated results from the searches revealed 994 articles. During the primary screening, 687 were excluded and on the basis of titles and abstracts, a further 236 articles were eliminated. Seventy‐one abstracts were read in detail and 29 full‐text articles obtained, of which, seven fulfilled the inclusion criteria. Additionally, two articles were identified from a reference list of a review article. A WHO International Clinical Trials Registry Platform search using the search term “diet and sodium” revealed 141 trials, but none were included.
A total of 185 references (164 from the 2011 review plus 9 + 3 + 9 new references) were thus included in the updated 2017 version.
Data collection and analysis
Selection of studies
See Search methods for identification of studies.
Review author NG performed the study selection for the 1998 version (Graudal 1998) and the 2003 version (Jürgens 2003). Review authors NG and GJ independently performed the supplementary study selection for the 2011 version (Graudal 2011. NG and THG independently performed the supplementary study selection for the 2017 version and the current version. Discrepancies were resolved by agreement.
Data extraction and management
Two authors independently recorded the following data from each trial:
the sample size (N);
the mean age of participants;
the fraction of females and males; White, Black and Asian participants;
the duration of the intervention;
the sodium reduction measured as the difference between 24‐hour urinary sodium excretion during low‐sodium and high‐sodium diets and standard deviation (SD);
SBP (SD) and DBP (SD) before and after intervention;
difference between changes in SBP and DBP obtained during low‐sodium and high‐sodium diets and the SD of these differences;
for cross‐over studies, when possible, the overall effect estimate and standard error (SE);
levels of hormones and lipids in the blood and their standard deviations during low‐sodium and high‐sodium diets. Concerning lipids, cholesterol units of mmol/L were transformed to mg/dL by means of the factor 38.6 and triglyceride units of mmol/L were transformed to mg/dL by means of the factor 88.4. Other renin units than ng/mL/hour were when possible transformed to ng/mL/hour, and units of aldosterone, noradrenalin and adrenalin other than pg/mL were transformed to pg/mL by means of the molecular weights.
Assessment of risk of bias data
If there were discrepancies between review authors they looked at the data together and came to an agreement.
Assessment of risk of bias in included studies
This was performed using the Cochrane 'Risk of bias' tool described in Chapter 8 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011) including recording of allocation, blinding, incomplete outcome data and selective reporting. Blinding was recorded as defined by the study authors (blinding) and as assessed by the present review authors (performance bias and detection bias).
Measures of treatment effect
No outcomes were dichotomous, but all were continuous. Consequently, the outcome measure was defined as the mean difference (MD) with 95% confidence intervals (CI) between the changes from baseline to end of treatment during low‐ and high‐sodium diets.
Unit of analysis issues
Blood pressure (BP)
Combined analyses were performed including both parallel and cross‐over studies. The generic inverse variance data type was used to analyse the effect in order to ensure that the weight of the cross‐over studies was not underestimated compared with the parallel studies. For parallel studies, the SE was calculated in the usual way as follows: SE (diff) = sqrt SE12 + SE22. For cross‐over studies the given SE (difference) was used. In case the SE (difference) was not reported, it was estimated by a linear regression equation linking the given SE to the calculated SE (sqrt SE12 + SE22) by means of the studies which reported both SE (difference) and SE on BP during both intervention periods. This regression equation was used to transform all SEs calculated by (sqrt SE12 + SE22) to estimated “true” SEs (difference) in cross‐over studies that did not report SE (difference). In this way, it was ensured that cross‐over studies were attributed proper weight compared with the parallel studies. There were not enough studies to calculate separate equations for Black and Asian populations and therefore the equations calculated in the white populations were used to transform these SEs when necessary.
Hormones and lipids
The very few parallel studies were excluded and the large fraction of cross‐over studies were analysed separately. As the large majority of cross‐over studies reported separate data for each intervention period instead of overall estimates of effect, the continuous data type was used in the separate analyses of the cross‐over studies.
Dealing with missing data
If the SD was not reported it was calculated from a given SE, 95% confidence interval (CI), P value or t value, estimated from a figure or imputed from the formula SD (change) = sq root (SD1sq + SD2sq), SD1 is SD on blood pressure before intervention and SD2 is SD on blood pressure after intervention.
Assessment of heterogeneity
A Chi2 test included in the forest plot was used to assess whether observed differences in results are compatible with chance alone. A low P value (or a large Chi2 statistic relative to its degree of freedom) provides evidence of heterogeneity of intervention effects (variation in effect estimates beyond chance). The Chi2 statistic can be transformed to a statistic (I2), which describes the percentage of the variability in effect estimates that is due to heterogeneity rather than sampling error (Higgins 2011). The I2 value can be interpreted as less important (0‐40%), moderate 30‐60%, substantial 50‐90% and considerable (75‐100%) (Higgins 2011). The I2 value was used to describe the percentage of heterogeneity in the individual analyses.
Assessment of reporting biases
For studies reporting SBP and DBP we evaluated whether the studies did not report other variables, which might have been mentioned in the methods or a protocol, i.e. whether there was selective reporting of outcome variables in the individual studies.
To analyse the possibility of publication bias, Funnel plots were assessed for asymmetry. Higgins states that as a rule of thumb, tests for funnel plot asymmetry should be used only when there are at least 10 studies included in the meta‐analysis, because when there are fewer studies the power of the tests is too low to distinguish chance from real asymmetry (Higgins 2011). Therefore, we did not assess funnel plots for black and Asian study populations
Data synthesis
Individual study data defined before randomisation based on race and state of hypertension were included in separate meta‐analyses, whereas sodium sensitivity subgroups, which were defined by the authors of the individual studies after they had analysed the data, were combined by the present authors, and subsequently the combined data were included in the meta‐analyses.
The mean difference (MD) was calculated for outcome measures with identical units in the included studies (BP without transformation of data (all measured as mmHg), renin, adrenaline, aldosterone, noradrenalin and lipids, after transformation). In this version we transformed the outcomes of 14 studies which reported renin as concentrations (mU/L) to renin as activities (ng/ml/h) by division with 8.2 (the regression coefficient assessed from the regression line associating renin concentration (mU/L) with renin activity (ng/ml/h) in figure 2 in De Bruin 2004). After this transformation only three studies was left in which there was no information to decide the unit of the outcome. These studies were excluded from the statistical analysis (1080 Huggins 1992 (increase from 6.1 to 8.6 during sodium reduction);1146 Herlitz 1998 (H)(increase from 1.7 to 19.6 during sodium reduction); 1234 Twist 2016 (H) (increase from 15.2 to 18.6 during sodium reduction)). Thus renin‐outcome in this version is reported as MD and not as SMD.
As we accumulated data from a series of studies that had been performed by researchers operating independently, and as the goal of the analysis was to extrapolate to other populations, we used a random‐effects model in our primary analysis to estimate the summary measure as the mean of a distribution of effects.
Level of significance: In case of multiple independent comparisons, it is important to avoid coincidental significance. Ten meta‐analyses were performed. However, the SBP and DBP comparisons are not independent of each other and BP depends on renin and aldosterone as well as catecholamines. Concerning lipids, these are mutually dependent, whereas the dependency on BP and hormones is not obvious. Consequently, the 10 meta‐analyses could be sub‐classified into a group of meta‐analyses of mutually dependent BP and hormones and an independent group of meta‐analyses of mutually dependent lipid fractions. Consequently, the level of significance was reduced by means of the formula 1‐0.95 x 1/N = 1‐0.95 x 1/2 = 0.025, (N = number of independent investigations = 2).
Subgroup analysis and investigation of heterogeneity
We prevented some heterogeneity by stratifying by race and BP. Furthermore, study duration might explain some heterogeneity. There is reasonable evidence to determine the time of maximal efficacy to be one week (Table 2). Therefore, there is a risk that studies lasting for less than one week may underestimate the effect of sodium reduction. Furthermore, evidence has appeared to indicate that all of the world’s populations have a mean sodium intake below 250 mmol/day (Table 1), and as dose‐response studies have indicated that sodium reductions from very high levels have bigger effects than reductions from usual levels (Graudal 2015), such studies may contribute to overestimate the effect. We therefore performed a subgroup analysis intending to eliminate these potential biases on SBP and DBP (stratified according to normal BP or hypertension) and renin, aldosterone, noradrenalin, adrenalin, cholesterol triglyceride, HDL and LDL by exclusion of studies with a duration of less than seven days and sodium intake above 250 mmol/day. The subgroup analysis on SBP and DBP was performed in studies of white participants only. As all studies in Asian and black participants except 3 lasted at least 7 days it was not meaningful to make a sub‐analysis in this small group of studies.
Sources of bias: To investigate possible heterogeneity due to methodological diversity, subgroup analyses of the primary analysis of SBP were performed for contrasting sources of bias appearing from the 'Risk of bias' analysis.
Sensitivity analysis
Sensitivity analyses were performed 1) excluding studies with extreme positive and negative outcomes in order to analyse the effect of reducing heterogeneity, and 2) including studies, which were excluded from the primary analysis due to undefined outcome units or study design differences (parallel versus cross‐over). In order not to underestimate the effect of the excluded parallel studies, these were included in the sensitivity analyses without adjusting the weight of the study to that of a cross‐over study of similar size.
Summary of findings and assessment of the certainty of the evidence
Outcomes (effect on SBP and DBP in normotensive and hypertensive white, black and Asian populations, effect on hormones and on lipids) is reported in summary of findings tables, which will show the mean differences (95% CI), number of participants and the certainty of the effect sizes estimated by the GRADE system.
Results
Description of studies
Forty‐six studies did not mention support. One hundred and twenty‐nine studies were supported by public foundations. Twelve studies were supported by the pharmaceutical industry. Six studies were supported by food industry organisations. Two studies were supported by companies unassociated with the pharmaceutical or food industry. Nineteen mixed populations were classified as white and 3 as black.
Results of the search
April 11th 2018, a literature search for the present update was performed as described in "Search methods for identification of studies". The de‐duplicated results from the searches revealed 877 articles. During the primary screening 561 were excluded. 316 abstracts were read, and 39 full‐text articles obtained, of which 10 fulfilled the inclusion criteria. A total of 195 references (185 from the 2017 review plus 10 new references) were thus included in the present updated version. These 195 references included 225 study populations. A top‐up search covered the period from March 9, 2016 to March 18, 2020 and revealed 2222 references and after de‐duplication 1448 remained. Compared versus previous searches 605 were duplicates and were removed. Thus, compared with the complete April 11, 2018 search, the top‐up search identified 843 new references. After a primary screening 271 were eliminated. Based on a screening of abstracts 538 of the remaining 572 articles were excluded. Thus 34 full‐text articles were reviewed and of these 12 fulfilled criteria for inclusion. These 12 references are not included in the present version but are placed under “studies awaiting classification”
Included studies
See Characteristics of included studies and Table 3 (Acronyms used in table "Characteristics of included studies")
A | Adrenaline |
Aldo | Aldosterone |
Chol | Cholesterol |
CO | Cross‐Over |
DBP | Diastolic blood pressure |
Dur | Duration |
HDL | High density lipoprotein |
Hyp | Hypertension |
IT | Intention to treat |
LDL | Low density lipoprotein |
LoFo | Lost to follow‐up |
MAP | Mean arterial pressure |
M/F | Male/Female |
N | Number |
Norm | Normotension |
NE | Noradrenaline |
Op | Open |
P | Parallel |
S | Single blind |
SBP | Systolic blood pressure |
SR | Sodium reduction |
TG | Triglyceride |
W/B/A: | Number of white participants/ black participants/ Asian participants |
One hundred and ninety‐five references were included in the review. Eight included only data on hormones and lipids, whereas 187 included BP data, as well as hormone and lipid data in a significant number of these. The total number of study populations with BP outcomes included in the primary analysis was 217. The median of the mean ages was 43 years (range: 12 to 73), which is a little higher than the median age of most populations (typically 35 years) and the mean sodium intake in the high‐sodium group was 203 mmol/day (SD: 66) and in the low‐sodium group was 65 mmol/day (SD: 40), corresponding to a mean sodium reduction of 138 mmol/day.The median of the mean ages of the study's 131 white populations included in the subgroup analysis (duration of at least seven days, a sodium intake of less than 250 mmol/day) was 45 years (range: 13 to 73) the mean sodium intake in the high‐sodium group was 178 mmol/day (SD: 35) and in the low‐sodium group was 68 mmol/day (SD: 36), corresponding to a mean sodium reduction of 110 mmol/day. The mean BP in the normotensive study populations was 119/71 mmHg, which is close to the population mean of the USA population (119/71 mmHg) (Wright 2011), and a little higher than the mean of the normotensive fraction of the USA population (115/70 mmHg) (Wright 2011). The mean BP in the untreated hypertensive study populations was 151/94 mmHg and in the treated hypertensive study populations was 143/88 mmHg, both of which are higher than corresponding pressures in the USA population (146/84 mmHg and 131/72 mmHg) (Wright 2011).
In 88 study populations including 7383 participants, there was information of the baseline 24‐hour sodium excretion, not influenced by diets. This was 158,7 mmol/24‐hour (range: 90‐274 mmol) (10‐90 percentiles: 123‐192 mmol).
Excluded studies
See Characteristics of excluded studies.
Risk of bias in included studies
See Characteristics of included studies and Figure 1
'Risk of bias' graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.
The obligatory trial quality criterion was randomisation. Double‐blind, single‐blind or open studies with a parallel or a cross‐over design were accepted. A study was defined as single‐blind if an investigator measured BP without knowledge of the diet or by a computerised manometer, and as open if precautions to decrease observer bias were not mentioned.
We found two important contrasts: general blinding and blinding of outcome detection (Figure 1). We performed subgroup analyses of BP in both normotensive and hypertensive white populations, but not in the black and Asian populations due to the small numbers of trials. We did not perform subgroup analyses on the biochemical outcomes (hormones and lipids) as they are supposed to be performed blindly in 100% of cases.
Allocation
Only 16 studies (1034 Watt 1985; 1078 Egan 1991 (H); 1081 TOHP I 1992; 1107 MacFadyen 1994;1135 TOHP II 1997; 1136 van Buul 1997;1142 Knuist 1998; 1195 Jessani 2008; 1197 Dickinson 2009; 1198 He 2009 W (H); 1206 Graffe 2012; 1208 Todd 2012; 1217 Markota 2015 (H); 1225 Gijsbers 2015 (H); 1229 He 2015; 1232 Nielsen 2016), either partly or sufficiently explained the allocation sequence generation and concealment. Consequently, there is a general significant risk that allocation could be biased.
Blinding
Seventy‐two study populations were reported to be double‐blind by the authors (general blinding), and in 140 study populations we assessed the risk of detection bias to be low (Figure 1). Separate analyses were performed on studies with low and high risks of general blinding and outcome detection in white populations.
Incomplete outcome data
Based on the information given in the individual articles, the risk of reporting participants with incomplete outcome data generally was small, about 20% being unclear or high (Figure 1). However, only a few studies showed flow charts of the fate of the participants. Therefore, this bias may be significant.
Selective reporting
Based on the information given in the individual articles, selective outcome reporting bias on the study level was small, about 10% being unclear or high (Figure 1). However, as protocols did not exist for the vast majority of studies, this evaluation may be imprecise.
Potential publication bias assessed by inspection of Funnel plots is described below for each outcome.
Other potential sources of bias
The effect of an intervention on BP may depend on factors such as baseline BP and race. Therefore, a biased distribution of such factors in the included study populations compared with the general population may bias the effect of the intervention found in the meta‐analysis to be different from the potential effect in the general population. We therefore performed separate analyses for hypertensive and normotensive individuals and for different races.
Effects of interventions
See: Summary of findings 1 Summary of findings: Low sodium intake compared with high sodium intake for blood pressure in white participants; Summary of findings 2 Summary of findings: Low sodium intake compared with high sodium intake for blood pressure in black participants; Summary of findings 3 Summary of findings: Low sodium intake compared with high sodium intake for blood pressure in Asian participants; Summary of findings 4 Summary of findings: Low sodium intake compared with high sodium intake for hormones; Summary of findings 5 Summary of findings: Low sodium intake compared with high sodium intake for lipids
See Data and analyses.
Blood pressure in white participants
See Summary of findings table 1
In the meta‐analyses of trials of white participants with normal blood pressure (BP), the mean difference (MD) was a change in systolic blood pressure (SBP) of ‐1.14 mmHg (95% CI: ‐1.65 to ‐0.63) (P = 0.0001) (95 trials, 9077 participant measurements (5982 participants)) (Analysis 1.1; Figure 2), and in diastolic blood pressure (DBP) of + 0.01 mmHg (95% CI: ‐0.37 to 0.39) (P = 0.96) (96 trials, (9341 participant measurements,(6276 participants)) (Analysis 1.2; Figure 3) (high‐quality evidence).
Forest plot of comparison: 1 Effect of salt reduction on systolic blood pressure (SBP) and diastolic blood pressure (DBP) in white participants, outcome: 1.2 White participants, normotensive, SBP.
Forest plot of comparison: 1 Effect of salt reduction on systolic blood pressure (SBP) and diastolic blood pressure (DBP) in white participants, outcome: 1.2 White participants, normotensive, DBP.
In subgroup meta‐analyses of trials with a duration of at least one week and a sodium intake of a maximal 250 mmol/day, the MD showed a decrease in SBP of ‐1.38 mmHg (‐1.87 to ‐0.89) (P = 0.00001) (63 trials, 7433 participant measurements,5186 participants) (Analysis 4.1) and in DBP of ‐0.37 mmHg (95% CI: ‐0.78, 0.04) (P = 0.08) (64 trials, 7697 participant measurements,5480 participants) (Analysis 4.2).
In all analyses (1.1, 1.2, 4.1 and 4.2) the I2 value decreased, when the four most extreme negative and positive effects were eliminated from the analysis without changing the overall effect: 1.1: From 69% to 40% (Effect from ‐1.14 mmHg to ‐1.32 mmHg), 1.2: From 60% to 40% (Effect from 0.04mmHg to ‐0.11mmHg), 4.1: From 55% to 29% (Effect from ‐1.38 mmHg to ‐1.26 mmHg), 4.2: From 52% to 17% (Effect from ‐0.37 mmHg to ‐0.40 mmHg).
In the trials of white people with elevated BP, MD showed a decrease in SBP of ‐5.71 mmHg (95% CI: ‐6.67 to ‐4.74) (P < 0.00001) (88 trials, 6116 participant measurements (3998 participants)) (Analysis 1.3), and in DBP of ‐2.87 mmHg (95% CI: ‐3.41 to ‐2.32) (P < 0.00001) (89 trials, 6140 participant measurements (4032 participants) (Analysis 1.4) (high‐quality evidence).
In subgroup meta‐analyses of trials with a duration of at least one week and a sodium intake of a maximal 250 mmol/day, MD showed a decrease in SBP of ‐5.32 mmHg (‐6.36 to ‐4.28) (P < 0.00001) (66 trials, 5274 participant measurements, 3645 participants) ( Analysis 4.3) and in DBP of ‐2.76 mmHg (95% CI: ‐3.38 to ‐2.13) (P < 0.00001) (67 trials, 5298 participant measurements, 3657 participants ) (Analysis 4.4).
Elimination of four extreme positive and negative effects also reduced I2 but to a lesser degree degree without changing the outcomes: 1.3: From 77% to 65% (Effect from ‐5.71 mmHg to ‐5.34 mmHg), 1.4: From 66% to 50% (Effect from ‐2.87 mmHg to ‐2.71 mmHg), 4.3: From 72% to 53% (Effect from ‐5.32 mmHg to ‐4.79 mmHg), 4.2: From 68% to 46% (Effect from ‐2.76 mmHg to ‐2.52 mmHg).
The BP funnel plots were asymmetric with missing values in the lower left corner for normotensives and in the lower right corner for hypertensive studies.
Bias analysis: Comparing low bias risk versus high/unclear bias risk of general blinding and blinding of outcome detection for SBP‐outcomes in white people with normotension and hypertension showed no important differences (Data and analyses: 9 Bias analyses).
Blood pressure in black participants
See Summary of findings table 2
In the meta‐analyses of seven cross‐over trials involving 253 black participants (506 participant measurements) with normal BP, MD showed a decrease in SBP of ‐4.02 mmHg (95% CI:‐7.37 to ‐0.68) (P = 0.02) (Analysis 2.1) and in DBP of ‐2.01 mmHg (95% CI:‐4.37, 0.35) (P = 0.09) (Analysis 2.2) (low‐quality evidence).
In the meta‐analyses of six cross‐over and two parallel trials of 398 black participants with elevated BP (619 participant measurements), MD showed a decrease in SBP of ‐6.64 mmHg (95% CI:‐9.00, ‐4.27) (P = 0.00001) (Analysis 2.3) and in DBP of ‐2.91 mmHg (95% CI:‐4.52, ‐1.30) (P = 0.0004) (Analysis 2.4) (low‐quality evidence).
The results are inconsistent as each of these groups could be separated in two subgroups with large and small effects, each having very low I2 values, whereas the combined group had very high I2values.
Blood pressure in Asian participants
See Summary of findings table 3
In the meta‐analyses of three parallel and two cross‐over trials involving 950 Asian participants (973 participant measurements) with normal BP, MD showed a decrease in SBP of ‐1.50 mmHg (95% CI: ‐3.09, 0.10) (P = 0.07) (Analysis 3.1) and in DBP of ‐1.06 mmHg (95% CI:‐2.53 to 0.41) (P= 0.16) (Analysis 3.2) (moderate‐quality evidence).)
In the meta‐analyses of six cross‐over and two parallel studies involving 254 Asian participants (501 participant measurements) with elevated BP, MD showed a decrease in SBP of of ‐7.75 mmHg (95% CI:‐11.44, ‐4.07) (P < 0.0001) (Analysis 3.3) and in DBP of ‐2.68 mmHg (95% CI: ‐4.21 to ‐1.15)(P = 0.0006) (Analysis 3.4) (low‐quality evidence).
The outcomes of the normotensive study populations were homogenous (I2 =0), but the outcomes of the hypertensive populations were inconsistent as each of these groups could be separated in two subgroups with large and small effects, each having very low I2 values, whereas the combined group had very high I2values.
Renin
See Summary of findings table 4
Four parallel trials were excluded (1110 Jula 1994 (H) (renin increase from 0.7 to 1.1 ng/ml/h during sodium reduction), 1155 Heer 2000 (renin increase from 20,4 to 76.1 mikroU/ml during sodium reduction);1226 Cavka 2015 (renin increase from 0.87 to 1.99 ng/ml/h during sodium reduction); 1228 Jablonski 2013 (H) (renin increase from 0.84 to 1.48 ng/ml/h during sodium reduction)).
In the remaining 82 cross‐over trials (2904 participants, 5808 participant measurements) of measurement of renin (including 91 comparisons reported in the Data & analyses), MD showed a mean increase of 1.56 ng/ml/h (95% CI: [1.39 to 1.73]) (P < 0.00001) (Analysis 5.1) (high‐quality evidence).
In comparisons with a duration of at least seven days and a sodium intake of less than 250 mmol/day (46 trials, 1819 participants), MD was 1.29 ng/ml/hour (95% CI: 1.07 to 1.51), (P < 0.00001) (Analysis 6.1).
The increase in normotensive participants was higher than in hypertensive participants (Summary of findings table 4).
A visual inspection of the Forest plot reveals that almost all studies showed an increase in renin, when sodium intake was reduced, indicating a high degree of consistency. Despite this the I2 value was very high (97%). The funnel plot was asymmetrical with missing values in the lower right corner.
Sensitivity analysis: Inclusion of four parallel trials with full weight contributions and three studies with undefined units using SMD as outcome measure did not change the outcome.
Aldosterone
See Summary of findings table 4
Five parallel trials were excluded (1110 Jula 1994 (H) (aldosterone increase from 297 to 363 pg/ml during sodium reduction); 1111 Howe 1994 (H)(aldosterone increased by 95 pg/ml during sodium reduction); 1155 Heer 2000 (aldosterone increase from 80.7 to 418.1 pg/ml during sodium reduction); 1226 Cavka 2015 (aldosterone increase from 214 to 424 pg/ml during sodium reduction); 1228 Jablonski 2013 (H) (aldosterone was unchanged from 41 to 57 pg/ml during sodium reduction)).
In the remaining 66 cross‐over trials (2506 participants, 5012 participant measurements) of measurement of aldosterone, MD was 104.02 pg/mL (95% CI: 18.4 to 119.69) (P < 0.00001) (Figure 4, Analysis 5.2) (high‐quality evidence). In comparisons with duration of at least one week and sodium intake of less than 250 mmol/day (35 trials, 1612 participants), MD was 96.64 pg/mL (95% CI: 75.35 to 117.94), P = 0.00001 (Analysis 6.2).The effect in normotensive participants was significantly higher than in hypertensive participants (Summary of findings table 4).
Forest plot of comparison: 5 Effect of salt reduction on hormones, outcome: 5.2 Aldosterone (pg/mL).
A visual inspection of the Forest plot reveals that almost all studies showed an increase in aldosterone, when sodium intake was reduced, indicating a high degree of consistency. Despite this the I2 value was very high (96%). The funnel plot was asymmetrical with missing values in the lower right corner.
Sensitivity analysis: Inclusion of five parallel trials with full weight contributions did not change the outcome.
Noradrenaline
See Summary of findings table 4
Two parallel trials were excluded (1110 Jula 1994 (H) (noradrenaline was unchanged from 610 to 520 pg/ml during sodium reduction); 1228 Jablonski 2013 (H) (noradrenaline was unchanged from 392 to 400 pg/ml during sodium reduction)).
In the remaining 35 cross‐over trials (878 participants, 1756 participant measurements) of measurement of noradrenaline (including 37 comparisons reported in the Data & analyses), MD was 62.33 pg/mL (95% CI: 41.90 to 82.76), (P = 0.00001) (Figure 5, Analysis 5.3) (high‐quality evidence). In comparisons with duration of at least one week and a sodium intake of less than 250 mmol/day (23 studies, 482participants) MD was 48.7 pg/mL (95% CI: 28.9 to 68.4), P = 0.00001 (Analysis 6.3). There was no difference between normotensive participants and hypertensive participants (Summary of findings table 4).
Forest plot of comparison: 5 Effect of salt reduction on hormones, outcome: 5.3 Noradrenaline (pg/mL).
A visual inspection of the Forest plot reveals that most studies showed an increase in noradrenalin, when sodium intake was reduced, indicating a high degree of consistency. Despite this the I2 value was high (74%). The funnel plot was symmetrical
Sensitivity analysis: Inclusion of two parallel trials with full weight contributions did not change the outcome.
Adrenaline
See Summary of findings table 4
One parallel trial was excluded (1110 Jula 1994 (H) (adrenaline was reduced by 10 pg/ml during sodium reduction).
In the remaining 15 cross‐over trials (331 participants, 662 participant measurements) of measurement of adrenaline (including 16 comparisons reported in the Data & analyses), MD was 7.55 pg/mL (95% CI: 0.85 to 14.26), (P = 0.03) (Analysis 5.4) (moderate‐quality evidence). In comparisons with duration of at least one week and sodium intake of less than 250 mmol/day (11 studies, 12 comparisons, 243 participants,) MD was 7.79 pg/mL (95% CI: 0.31 to 15.28), P = 0.04 (Analysis 6.4). There was no difference between normotensive participants and hypertensive participants (Summary of findings table 4).
The I2 value was moderate (58%). The funnel plot was symmetrical
Sensitivity analysis: Inclusion of one parallel trial with full weight contribution did not change the outcome (MD = 6.91 pg/ml).
Cholesterol
See Summary of findings table 5
Three parallel trials were excluded (1015 Bulpitt 1984 (H)(No changes during sodium reduction); 1085 Sciarrone 1992 (H)(Minor decrease during sodium reduction); 1199 Meland 2009 (H))(No changes during sodium reduction). In the remaining 27 cross‐over trials (917 participants, 1834 participant measurements) of measurement of cholesterol (including 28 comparisons reported in the Data & analyses), MD showed an increase of 5.19 mg/dL (95% CI: 2.07 to 8.32), P = 0.001 (Figure 6, Analysis 7.1) (high‐quality evidence). In comparisons with duration of at least one week and sodium intake of less than 250 mmol/day (21 trials, 1214 participants) MD was 4.31 mg/dL (95% CI: 0.71 to 7.91), P = 0.02 (Analysis 8.1). There was no difference between normotensive participants and hypertensive participants (Summary of findings table 5)
Forest plot of comparison: 6 Effect of salt reduction on lipids, outcome: 6.1 Cholesterol.
The I2 value was low (0 %). The funnel plot was symmetrical.
Sensitivity analysis: Inclusion of three parallel trials with full weight contributions did not change the outcome.
Triglyceride
See Summary of findings table 5
Two parallel trials were excluded (1085 Sciarrone 1992 (H)(Minor increase during sodium reduction ; 1199 Meland 2009 (H))(No changes during sodium reduction).
In the remaining 20 cross‐over trials (712 participants, 1424 participant measurements) of measurement of triglyceride, MD showed an increase of 7.10 mg/dL (95% CI: 3.14 to 11.07), P = 0.0004 (Analysis 7.2) (high‐quality evidence). In comparisons with duration of at least one week and sodium intake of less than 250 mmol/day (13 trials, 804 participants) the effect was 7.03 (mg/dL [95% CI: 2.01 to 12.04), P = 0.006 (Analysis 8.2). There was no difference between normotensive participants and hypertensive participants (Summary of findings table 5)
The I2 value was low (0 %). The funnel plot was symmetrical.
Sensitivity analysis: Inclusion of two parallel trials with full weight contributions did not change the outcome.
High‐density lipoprotein (HDL)
See Summary of findings table 5
Two parallel trials were excluded (1085 Sciarrone 1992 (H)(Minor decrease during sodium reduction); 1199 Meland 2009 (H))(No changes during sodium reduction).
In the remaining 20 cross‐over trials (738 participants, 1476 participant measurements) of measurement of HDL, there was no effect of sodium reduction on serum HDL: MD: ‐0.30 mg/dL (95% CI: ‐1.66 to 1.0) P = 0.66 (Analysis 7.3) (high‐quality evidence). This result did not change in comparisons with duration of at least one week and sodium intake of less than 250 mmol/day (‐0.68 mg/dL (‐2.17 to 0.81), P = 0.37 (15 trials, 982 participants)) (Analysis 8.3). There was no difference between normotensive participants and hypertensive participants (Summary of findings table 5)
The I2 value was low (0 %). The funnel plot was symmetrical.
Sensitivity analysis: Inclusion of two parallel trials with full weight contributions did not change the outcome.
Low‐density lipoprotein (LDL)
See Summary of findings table 5
One parallel trial was excluded (1085 Sciarrone 1992 (H))(Minor decrease during sodium reduction).
In the remaining 18 cross‐over trials (696 participants, 1392 participant measurements) of measurement of LDL, MD showed a non‐significant increase of 2.46 mg/dL (95% CI: ‐0.97, to, 5.9), P = 0.16 (Analysis 7.4)(moderate‐quality evidence). In comparisons with duration of at least one week and sodium intake of less than 250 mmol/day (13 trials, 898 participants), MD was 2.74 mgdL (95% CI: ‐1.18 to 6.66), P = 0.17 (Analysis 8.4). There was no difference between normotensive participants and hypertensive participants (Summary of findings table 5)
The I2 value was low (0 %). The funnel plot was symmetrical.
Sensitivity analysis: Inclusion of one parallel trial with full weight contribution did not change the outcome.
Discussion
Summary of main results
The GRADE of the evidence was generally high with the exception of the small blood pressure analyses of the few black and Asian study populations, which were assessed to be of low or moderate GRADE.
The effect of sodium reduction from an average high usual intake (203 mmol/day) to the recommended level (65 mmol/day) was small in study populations with normal blood pressure (BP) (‐1.14/+0.01 mmHg) corresponding to a mean arterial pressure effect of only ‐0.38 mmHg . In hypertensive study populations the effect was (‐5.71/‐2.87 mmHg)corresponding to a mean arterial pressure effect of about ‐3.8 mmHg. In a subgroup analysis intending to eliminate the potential bias of a very short intervention duration (< seven days) and very high sodium intake (> 250 mmol/day), the decrease in BP in study populations with a normal BP (‐1.38/‐0.37 mmHg) and hypertension (‐5.32/‐2.76 mmHg) was also small. The effect of sodium reduction on hormones and on lipids showed increases in renin, aldosterone, noradrenalin, cholesterol and triglyceride in the primary analysis, as well as in the subgroup analysis, whereas the increase in adrenalin was small. The increase in cholesterol in the low‐salt group seemed mainly to be due to an increase in low‐density lipoprotein (LDL). Sodium reduction had no effect on high‐density lipoprotein (HDL).
The analysis of black populations showed that the effect of sodium reduction in black people with normotension corresponded to the one found in black people with hypertension. This was in contrast to the analyses of white and Asian populations in whom the effect was smaller in those who were normotensive than in those who were hypertensive. However, compared with previous analyses (Graudal 1998; Jürgens 2003), the diverging results within the black populations and between the black and white populations are smaller. In a recent detailed analysis, we found that a significant fraction of the differences between the three racial groups could be ascribed to differences in baseline BP, age, and amount of sodium reduction. Furthermore there was no difference in BP outcome between racial groups investigated in the same study (Graudal 2015b) indicating that the differences found in the present meta‐analysis mainly may be due to confounders rather than racial differences.
The funnel plots indicated that the effect of sodium reduction for on BP may have been a little overestimated for normotensives and a little underestimated for hypertensives. The funnel plot asymmetry for renin and aldosterone probably does not reflect publication bias, but rather that large sodium reductions to low levels of sodium intake mainly was performed in small studies.
Overall completeness and applicability of evidence
In the primary analysis, population samples from the whole BP distribution of the populations were included. In this analysis, the intake of sodium in the “high” sodium group was in the interval 100 mmol/day to 795 mmol/day in 217 comparisons (99%), and below 100 mmol/day in one comparison, the mean level being 203 mmol/day. The intake of sodium in the low‐sodium group was below 100 mmol/day in 178 comparisons (82%) and above 100 mmol/day in 39 comparisons, the mean level being 66 mmol/day. In the subgroup analysis, the intake of sodium in the “high” sodium group was in the interval 109 mmol/day to 248 mmol/day in 131 comparisons (100%), and below 100 mmol/day in zero comparisons, the mean level being 177 mmol/day. The intake of sodium in the low‐sodium group was below 100 mmol/day in 106 comparisons (81%) and above 100 mmol/day in 25 comparisons, the mean level being 68 mmol/day. Consequently, this meta‐analysis in general compares the effects of a dietary sodium intake, which is lower than usual and in accordance with the recommendations to reduce sodium below 100 mmol/day with a sodium intake, which is within the present world‐wide usual range of sodium intake, the level in the primary analysis being in the high end of the usual intake and the level in the subgroup analysis being close to the world mean of 159 mmol/day (Table 1). The mean and the range of the baseline 24‐hour sodium excretion of the included populations before diet manipulation (159 mmol/24 hours (10 to 90 percentile: 123 to 194)) were almost identical with the usual range of sodium intake in the world's populations (McCarron 2013; Powles 2013). Thus, the present review shows the consequences of the recommendations of the health institutions, which is to reduce the usual sodium intake of the world's populations (90 mmol/day to 250 mmol/day) to a level below 87 mmol/day WHO 2012 or below 100 mmol/day (IOM 2005, DGA 2015).
Quality of the evidence
Only randomised controlled trials were included and the basic grade of evidence was therefore considered to be high, although the grade of evidence was downgraded in some of the smaller analyses. In general, the description of the randomisation procedure was insufficient, introducing a bias which could exaggerate the effects, but many of the studies were published in a period where it was not customary to report such descriptions. This bias could not be explored in a meaningful way due to the lack of contrasts between the number of studies with low and high risk of these biases (Figure 1). Almost all individual studies of participants with normal blood pressure (BP) show no significant effect of sodium reduction on BP, whereas a large number of studies in people with hypertension did show significant effect of sodium reduction on BP. Thus, there was a high grade of consistency between the outcomes of the individual studies and the outcomes of the meta‐analyses. Sensitivity analyses of studies lasting at least one week (the time of maximal efficacy) confirmed the primary analyses. Finally, the potential influence of commercial interests on the outcomes was negligible.
In general the I2 values indicated heterogeneity. There was significant clinical diversity in age, baseline BP and degree of sodium reduction across studies, which could explain a significant fraction of the heterogeneity. Our analysis of bias contrasts did not show obvious differences disclosing methodological diversity. If chance alone decides an effect of an intervention the distribution of effect sizes would be expected to be symmetrical around zero and the number of studies showing a positive effect would equal the number of studies showing a negative effect. Inspecting the forest plots, this is almost the case for the effect of sodium reduction on SBP and the case for DBP in normotensive white individuals indicating that most of the observed variation in the forest plots may be due to random variation. After elimination of the four most extreme positive and negative effects in each of these analyses the I2 values generally fell to 40% or less indicating that heterogeneity might not be important (Higgins 2011).
Concerning the hypertensive studies in white people, the I2 values indicated moderate heterogeneity. Thus a downgrade of the GRADE assessment could be considered, but the statistical strength of the outcome is very strong and it is not likely that additional studies will change these outcomes, as we have previously shown in a cumulative meta‐analysis, which showed that the outcome was robust already after the inclusion of about 10 studies (Graudal 1998). In conclusion we have assessed the quality of the analyses in white people to be of high grade. In Asian normotensive participants the outcome was similar to the outcome in white people and the I2 values for both SBP and DBP were small indicating that this outcome may be robust. Due to the few studies included we have downgraded the quality of this evidence to be moderate. The BP outcomes in the black study populations and Asian hypertensive study populations were all inconsistent as each of these groups could be separated in two subgroups with large and small effects, each having very low I2 values, whereas the combined groups had very high I2 values. Therefore, it is difficult to define a reliable effect, as the mean effect did not represent any of the two groups, indicating that further studies are needed to establish a robust result for these population groups. We have downgraded this evidence to be of low quality.
Concerning hormones there was consistency, as almost all studies showed the same trend, namely an increase in the hormone level, when sodium intake was reduced. Despite this the I2 values showed substantial (noradrenalin and adrenalin) or considerable heterogeneity (renin and aldosterone). As the association between sodium intake and hormones may follow an exponential curve as proven for renin and aldosterone (Brunner 1972), the explanation for the large heterogeneity probably is that a reduction of sodium intake from a high level to a moderate level induces a factor 2 increase in hormones whereas a reduction of sodium intake from a moderate level to a low level induces a factor 5 increase in hormones, as we have previously shown (Graudal 1998). Due to narrow confidence intervals and consistency in the direction of the outcomes we have graded the quality of the hormone analyses to be high except for adrenalin, which was downgraded due to wide confidence limits and few studies.
Concerning lipids all I2 values were 0. Concerning cholesterol and triglyceride, there was also a high degree of consistency as most studies showed an increase, when sodium intake was reduced. Accordingly, we have graded these analyses to of high quality. The neutral result on HDL also seemed robust with confidence limits almost symmetrical around zero. In contrast, the neutral result on LDL seemed less robust with the lowest confidence limit close to zero, indicating that further studies might change the result to be significant. Accordingly, we downgraded the LDL result to be of moderate quality.
The number of studies included in the BP analyses (n = 217) is substantial as is the number of participants (12232). This should allow robust conclusions. A weakness was that a large number of studies were not double‐blind. However, concerning this source of bias, there were no obvious trends towards different effects in the low‐risk blinded groups compared with the high‐risk open groups (Analysis 9.1; Analysis 9.2; Analysis 9.3; Analysis 9.4; Analysis 9.5; Analysis 9.6; Analysis 9.7; Analysis 9.8).
Potential biases in the review process
In the nineties journals did not generally demand protocols as a condition for publication and we did not publish a protocol before the 1998 version (Graudal 1998). Cochrane invited us to publish this finished paper as a Cochrane review. Consequently, there is no protocol. Registration of a protocol counteracts bias by revealing differences in methods or outcomes between the planned protocol and the published review. Thus, we cannot document the absence of such bias in our review. However, as multiple reviews (discussed below) with different inclusion criteria have found similar results, we assess that such potential bias have no impact on the present results.
The present review is the largest of the many existing meta‐analyses on sodium reduction, and other meta‐analyses have not identified studies, which were not identified by our search. Our analysis is the largest partly because our selection criteria were less restrict. Therefore, a fraction of the included studies had an experimental character investigating a sodium intake far beyond the sodium intake in the general population for only four to six days, which may not be relevant for the general population on long‐term sodium reduction. The fact that the subgroup analysis, which eliminated the potential short‐term intervention bias and very high sodium intake bias, showed similar results as the primary analysis, indicates that the inclusion of extreme studies had a minor impact on the mean of the outcome effects. Other meta‐analyses have extracted almost identical data in the individual studies indicating that our data extraction is unbiased.
In total 24 of 651 people in the defined black populations were white people (3.7%) and 642 of 10308 in the defined white populations were black people (6.2%). We assess that these small contributions have no impact on the outcomes.
Agreements and disagreements with other studies or reviews
The scientific evidence behind the sodium reduction recommendations is a series of studies and meta‐analyses, which are biased by high baseline blood pressure, high age and overweight (Graudal (3) 2016). The most prominent of these studies (DASH 2001), was additionally biased by a control group diet, which was designed to contain only half of the normal amount of potassium (Graudal and Jürgens 2018). Despite these studies being irrelevant as evidence for pubic health recommendations, the Food and Drug Administration (FDA) has released draft proposed voluntary guideline to encourage companies to steadily reduce sodium in processed foods (Frieden TR 2016). The main argument for this guideline was a dose‐response meta‐regression analysis of mixed normotensive and hypertensive study populations, which was biased because it included mainly studies with high blood pressure and inappropriately forced the dose‐response relationship through zero and thereby further doubled the postulated effect. In contrast, previous meta‐analyses of randomised controlled trials (RCTs) have shown similar results of sodium reduction on BP. In 1986, Grobbee and Hofman combined 13 studies of persons with normal and elevated BP in a meta‐analysis and found a significant hypotensive effect of reduced sodium intake on SBP of ‐3.6 mmHg and a non‐significant effect on DBP of ‐2.0 mmHg (Grobbee 1986). In 1991, a second meta‐analysis of 24 RCTs showed an effect of ‐4.0/‐2.5 mmHg for persons with elevated BP and ‐1.0/‐0.2 for persons with normal BP (Cutler 1991). This was verified in an update from 1997 (Cutler 1997). In 1996, a meta‐analysis of 53 RCTs showed an effect of ‐3.7/‐0.9 mmHg in persons with elevated BP and ‐1.0/‐0.1 in persons with normal BP (Midgley 1996). In an analysis of eight RCTs lasting for at least six months, the effect was ‐2.9/‐2.1 mmHg for persons with elevated BP and ‐1.3/‐ 0.8 mmHg for persons with normal BP (Ebrahim 1998). These results were confirmed in an update (Hooper 2002). All these similar results confirm that selection of RCTs based on magnitude of sodium difference or duration of the intervention does not significantly change the overall effect size estimate. These meta‐analyses indicate that major disagreements about this effect size no longer seem to exist. However, there is still significant disagreement regarding the relevance of the effect size and the relevance of potential side effects (Taubes 1998).
The effect of sodium reduction on BP in hypertensive and normotensive study populations in the present review matches the effects found in most of these previous reviews, although the effect of sodium reduction on BP in normotensives is marginally lower than in the meta‐analysis, which supports the WHO recommendations (Aburto 2013). In hypertensive study populations, there was no differences between the WHO review and our review. In normotensive study populations, the difference was small, the BP effect in the WHO review being ‐1.38/‐0.58 mmHg and in ours being ‐1.14/0.01 (‐1.38/‐0.37 in the subgroup analysis). This study differed from ours as it only included studies lasting at least four weeks. However, as duration has no impact on the BP effect (Table 2), a more reliable explanation for the difference between the WHO review and our review is that the study populations with normal BP in the WHO review generally have a high baseline BP in the upper 50% percentile of the population.
According to WHO, the small effect in normotensive study populations is sufficient to recommend sodium reduction for the whole population, the assumption being that the association between BP and mortality is consistent. This, however, may not be the case. For instance, beta‐blockers reduce BP in hypertensive individuals, but not mortality (Wiysonge 2012), and a recent meta‐analysis of patients with diabetes showed that antihypertensive treatment reduces the risk of mortality and cardiovascular morbidity in diabetes patients with SBP higher than 140 mm Hg, but if SBP is less than140 mm Hg further treatment is associated with an increased risk of cardiovascular death, with no observed benefit (Brunström 2016). Such studies indicate that it is not possible to extend the general association of BP with mortality (Collins 1990) to the effect of a BP‐reducing intervention on mortality. The reason for this inconsistency may be side effects of the intervention. However, while short duration has been suggested to underestimate the BP effect, it has concomitantly been suggested to overestimate possible adverse effects on hormones and lipids. This idea that the duration of the intervention tends to underestimate some physiological outcomes and overestimate others has not been documented, but still has been used to disregard side effects shown in studies lasting less than four weeks. Very few studies lasting more than four weeks have investigated side effects, and further more these studies do not reduce sodium to the recommended level, but to levels above 87 mmol/day, and therefore the side effects in these few studies may not be fully disclosed. In contrast, the present analysis shows that the adverse effects on hormones and lipids are significant, when the sodium intake is lowered from a high usual sodium intake to a level in accordance with the recommendations of the health institutions. In addition, we have just shown that sodium reduction results in an increase in heart rate of 2.4% (Graudal (2) 2016). This may be an important side effect as resting heart rate is directly associated with mortality (Ho 2014; Jensen 2012). The assumption that at least some of these effects may be persistent and not just temporary has been indicated in observational studies. Yanomamo Indians, who persistently ingest very small amounts of sodium, have a three times higher level of renin in the blood and a 10 times higher excretion of aldosterone in the urine than normal controls (Oliver 1975). Furthermore, renin and aldosterone rise slowly as long as the intake is above 100 mmol/day, but exponentially, when sodium intake is reduced to levels below 100 mmol/day (Brunner 1972. Thus, the present meta‐analysis provides a possible explanation for the small effect of reduced sodium intake on blood pressure: compensatory activation of the renin‐aldosterone system is proportional to the degree of sodium reduction. Furthermore, the increases in noradrenaline and adrenaline may contribute to this counter‐regulation (Warren 1980) and contribute to an increase in heart rate.
The very small effect of sodium reduction on BP in healthy individuals shown in the present review and other reviews including the WHO review, the risk of significant side effects shown in this review, and the possibility that an intervention to reduce BP may not reduce mortality (Wiysonge 2012), and even may increase mortality in some population groups with a normal BP (Brunström 2016) indicate that the BP‐effect is not sufficient as a basis for recommendations in the general population, but should be verified in studies directly relating sodium intake with morbidity and mortality. Unfortunately, RCTs of the effect of sodium reduction below 100 mmol/day on mortality in healthy individuals do not exist (Graudal (1) 2016). A recently updated meta‐analysis of eight RCTs with follow‐up data on morbidity and mortality found a non‐significant trend versus reduced cardiovascular (CV) morbidity, but could not demonstrate reduced all‐cause mortality in the low‐sodium group (Adler 2014). These trials were performed in overweight pre‐hypertensive or hypertensive individuals and the sodium reduction was not below 100 mmol/day, but down to 100 mmol/day.
The sodium‐mortality relationship has also been estimated by means of 27 observational studies (Alderman 2010; Mente 2016; O'Donnell 2014; Pfister 2014), which directly asses the relationship between sodium intake in the individual and mortality. Most of these studies were evaluated in an IOM report (IOM 2013). This IOM report did not confirm the 100 mmol/day upper level for sodium intake, which was defined in a previous IOM report (IOM 2005), but concluded that “Science was insufficient and inadequate to establish whether reducing sodium intake below 2300 mg/d (100 mmol) either decreases or increases CVD risk in the general population”. A later meta‐analysis of these population studies found that a sodium intake below 114 mmol/day was associated with increased mortality, as was a sodium intake above 214 mmol/day (Graudal 2014). Increased mortality with high sodium intake has also been shown in another meta‐analysis, which, however, did not investigate the effect of a low sodium intake (Strazzulo 2009). This U‐shaped relation between sodium intake and mortality has been identified in several individual population studies (O'Donnell 2011; O'Donnell 2014; Pfister 2014; Thomas 2011). The health institutions, however, generally do not accept this evidence from the observational studies (Gunn 2013; Whelton 2012; WHO 2012). In a recent paper, which discusses methodological issues of observational studies, representatives of the American Heart Association state that the association of low sodium intake with increased mortality observed in observational studies may reflect that sick people have a low sodium intake (reverse causality: sick people with a high mortality have a low sodium intake, it is not the low sodium intake, which increases the mortality) (Cobb 2014). This hypothesis is not directly supported by the observational studies, as the outcomes generally are adjusted for confounders such as cardiovascular and renal diseases and diabetes and show that the mortality associated with a low sodium intake is higher in healthy populations than in populations including sick individuals (Graudal 2014; O'Donnell 2014). Table 4 shows a meta‐analysis of the risk of all‐cause mortality in study populations within the usual sodium intake range versus a low sodium intake below 114 mmol/day (Graudal 2014) or below 130 mmol/day (O'Donnell 2014. The analysis is confined to include samples of individuals representative of the general populations and all individual study analyses are adjusted for multiple confounders such as cardiovascular disease, hypertension and diabetes. To further reduce the risk of reverse causality, the most healthy subgroup was included in the analysis, when results were given for subgroups, The possibility of reverse causality can never be completely excluded, but as a minimum there is no indication in population studies that sodium intake below 100 mmol/day has beneficial health effects in healthy individuals. In the NHANES I and III studies this was demonstrated by independent groups (Alderman 1998; Cohen 2008; He 1999; Yang 2011). Finally the recent analysis of NHANES V showed that sick individuals have exactly the same sodium intake as healthy individuals (Cogswell 2018) definitively confirming that the reverse causality hypothesis is unlikely (Jürgens 2018).
Study | Multiple adjustment* | Exclusion | N (LS) | N (US) | RR/OR (95% CI) |
Alderman 1998 (NHANES I) | Yes | None | 2837 | 8509 | 0.88 (0.80, to, 0.97) |
He 1999 (NHANES I) | Yes | Overweight (BMI > 27.3) | 1699 | 5098 | 0.98 (0.88 to 1.09) |
Yes | Males** | 634 | 311 | 0.91 (0.56 to 1.48) | |
Cohen 2006 (NHANES II) | Yes | None | 3711 | 3443 | 0.78 (0.67 to 0.91) |
Yes | CVD and HT | 392 | 392 | 1.12 (0.86 to 1.46) | |
Cohen 2008 (NHANES III) | Yes | None | 2175 | 4350 | 0.83 (0.73 to 0.94) |
Yang 2011 (NHANES III) | Yes | Overweight (BMI > 25) | 3067 | 6133 | 0.93 (0.73 to 1.18) |
Yes | None | 1250 | 1220 | 0.82 (0.62 to 1.08) | |
Yes | None | 1138 | 961 | 0.89 (0.74 to 1.07) | |
Pfister 2014 (Norfolk) | Yes | 0‐2 year events | 3070 | 9249 | 0.92 (0.82 to 1.02) |
O'Donnell 2014 (PURE) | Yes | CVD, Cancer, DM, smokers | 6162 | 38643 | 0.62 (0.54 to 0.71)] |
Total (95% CI)# | 21369 | 67078 | 0.84 (0.76 to 0.93) | ||
Total (95% CI)## | 21123 | 65450 | 0.87 (0.76 to 0.98) |
Only studies, which were representative for the general population and which adjusted for confounders were included.
If subgroup results were given, the results of the most healthy subgroup was used in the analysis to reduce
the possibility of reverse causation
#With primary NHANES analyses (Alderman 1998, Cohen 2008)
## With NHANES re‐analyses (He 1999, Yang 2011)
* Studies were generally adjusted for at least sex, age and CVD risk factors
** In the male group a low salt intake group could not be identified, as the salt intake
in the lowest salt intake quartile was up to 159 mmol.
BMI: body mass index; CVD: cardiovascular disease; DM: diabetes mellitus; HT: hypertension
In addition to reverse causality, inaccurate measurements of sodium intake has been used to explain the association of low sodium intake with increased mortality (He 2018), the claim being that more accurate multiple measurements would reverse the outcome. However, the individual sodium intake measurements are not inaccurate, they are imprecise. This is verified by some of the presented inconsistent data, which show that the outcome based on a single "inaccurate" estimate was almost identical with the outcome based on "accurate" multiple 24 hour measurements (He 2018). Thus, although imprecise, the individual sodium intake data are sufficiently accurate to classify most of the participants in population studies in the right percentile (typically tertile, quartile or quintile) (Olde Engberink 2017; Graudal and Mente 2018). Furthemore the actual misclassifications are not systematic, but random (Olde Engberink 2017; Graudal and Mente 2018), which contributes to an underestimation of the outcome, but do not cause a change of the direction of the outcome (Olde Engberink 2017; Fan 2014)
The BP effect of reduced sodium intake has been related to age. Freedman and Petitti analysed data from Intersalt (Intersalt 1988) and found the paradox that along with the significant association between increase in blood pressure with age and the salt excretion in urine, there was an inverse relationship between estimated BP and salt excretion in urine at age 20. Freedman stated that unless you preferred to conclude that salt should be eaten in high doses by youngsters and in reduced amounts by the elderly, the findings were probably due to uncontrolled confounding, not to variation in salt intake (Freedman 2001). Furthermore, it is now clear that the BP of different age cohorts in a cross‐sectional study like Intersalt is not representative of each other, verified by a study showing that recent birth cohorts attained lower BP than did earlier birth cohorts in the period 1887 to 1994 (Goff 2001). According to this study, based on data from more than 50,000 persons, it can be estimated that the median BP is about 15 mmHg lower in a 50‐year old person from a recent birth cohort compared with a 50‐year old from a birth cohort from the late 19th century. Consequently, there has been a dramatic fall in BP during the 20th century. In this context, the possible mean arterial pressure effect of sodium reduction of ‐0.7 mmHg in normotensive persons seems negligible. Finally, it has been difficult to maintain a significant sodium reduction in longer‐term studies, which should be taken into consideration, when recommending sodium reduction. One reason for this could be that the sodium intake is regulated by neuro‐physiological and hormonal mechanisms (Geerling 2008), and therefore difficult to diverge from.
The hypothetical consequences of the present findings are that people with normotension would have no benefit from sodium reduction, but may suffer from harms, because sodium reduction has a negligible effect on BP, but results in significant side effects. People with hypertension may benefit due to the effect on BP, but may also suffer from harms due to the side effects. This is exactly what was found in the most recent meta‐analysis of four population studies (133,000 individuals) in which the authors had access to individual participant data (Mente 2016). The conclusion was "Compared with moderate sodium intake, high sodium intake is associated with an increased risk of cardiovascular events and death in hypertensive populations (no association in normotensive population), while the association of low sodium intake with increased risk of cardiovascular events and death is observed in those with or without hypertension. These data suggest that lowering sodium intake is best targeted at populations with hypertension who consume high sodium diets", a conclusion, which matches perfectly with the results of the present meta‐analysis.
A recent critical review has indicated that the choice and methods of evidence to support dietary guidelines were insufficient (Teicholz 2015) Subsequently a National Academy of Medicine (NAM) report has expressed distrust in the process for the establishment of the dietary guidelines for Americans (DGA) (NAM 2017). The NAM report revealed that DGA was not based on a “a full body of evidence on a continuous basis over time”. The report concluded that “the process to update the DGA should be comprehensively redesigned to allow it to adapt to changes in needs, evidence, and strategic priorities”. Thus, fundamental problems in the process of the establishment of the DGA may in part explain the contrast between the public recommendations to lower sodium intake below 2300 mg and the evidence, which indicates that such a reduction is associated with only small effects on BP, a series of potentially harmful effects and increased mortality.
Intellectual Property Information
This system is subject to the following patents by Groom Ventures LLC:
- US 10,949,459 B2 Alternate Search Methodology, and
- US 9,911,132 B2 System and Method for Searching, Organizing, Exploring and Relating Online Content
Powered by Global Fast Fit
Login to Comment