- Lee Hooper (), research fellow in evidence based care and systematic reviewa,
- Christopher Bartlett, research associate in health services researchb,
- George Davey Smith, professor of clinical epidemiologyb,
- Shah Ebrahim, professor in epidemiology of ageingb
- a MANDEC, University Dental Hospital of Manchester, Manchester M15 6FH
- b Department of Social Medicine, MRC Health Services Research Collaboration, University of Bristol, Bristol BS8 2PR
- Correspondence to: L Hooper
- Accepted 15 May 2002
Objective: To assess the long term effects of advice to restrict dietary sodium in adults with and without hypertension.
Design: Systematic review and meta-analysis of randomised controlled trials.
Data sources: Cochrane library, Medline, Embase, and bibliographies.
Study selection: Unconfounded randomised trials that aimed to reduce sodium intake in healthy adults over at least 6 months. Inclusion decisions, validity and data extraction were duplicated. Random effects meta-analysis, subgrouping, sensitivity analysis, and meta-regression were performed.
Outcomes: Mortality, cardiovascular events, blood pressure, urinary sodium excretion, quality of life, and use of antihypertensive drugs.
Results: Three trials in normotensive people (n=2326), five trials in those with untreated hypertension (n=387), and three trials in people being treated for hypertension (n=801) were included, with follow up from six months to seven years. The large high quality (and therefore most informative) studies used intensive behavioural interventions. Deaths and cardiovascular events were inconsistently defined and reported. There were 17 deaths, equally distributed between intervention and control groups. Systolic and diastolic blood pressures were reduced (systolic by 1.1 mm Hg, 95% confidence interval 1.8 to 0.4 mm Hg; diastolic by 0.6 mm Hg, 1.5 to 0.3 mm Hg) at 13 to 60 months, as was urinary 24 hour sodium excretion (by 35.5 mmol/24 hours, 47.2 to 23.9). Degree of reduction in sodium intake and change in blood pressure were not related.
Conclusions: Intensive interventions, unsuited to primary care or population prevention programmes, provide only small reductions in blood pressure and sodium excretion, and effects on deaths and cardiovascular events are unclear. Advice to reduce sodium intake may help people on antihypertensive drugs to stop their medication while maintaining good blood pressure control.
What is already known on this topic
Restricting sodium intake in people with hypertension reduces blood pressure
Long term effects (on blood pressure, mortality, and morbidity) of reduced salt intake in people with and without hypertension are unclear
What this study adds
Few deaths and cardiovascular events have been reported in salt reduction trials
Meta-analysis shows that blood pressure was reduced (systolic by 1.1 mm Hg, diastolic by 0.6 mm Hg) at 13 to 60 months, with a reduction in sodium excretion of almost a quarter (35.5 mmol/24 hours)
The interventions used were highly intensive and unsuited to primary care or population prevention programmes
Lower salt intake may help people on antihypertensive drugs to stop their medication while maintaining good control of blood pressure, but there are doubts about effects of sodium reduction on overall health
Several systematic reviews have reported that restricting sodium intake in people with hypertension reduces their blood pressure.1–5 However, most of the trials in these systematic reviews were short term and did not allow for complete adjustment of blood pressure to altered sodium intake or reduced motivation for following dietary restrictions over time. Also, some trials increased sodium intake in one arm and compared this with a reduced sodium intake in the other arm and so did not estimate likely effects of cutting down on sodium in a normal diet. 6 7 No review on long term outcomes has been carried out since 1998,7 although large relevant trials have been published.
The value of lowering blood pressure depends on its effects on cardiovascular events and deaths. The published systematic reviews on the effect of salt restriction on blood pressure and other risk factors have disagreed about the size of blood pressure changes8 and the effects on cardiovascular events and deaths. We assessed, in people with and without hypertension, the efficacy of advice to reduce dietary sodium intake over at least six months on mortality, cardiovascular events, blood pressure, urinary sodium excretion, quality of life, and use of antihypertensive medications.
A previous large scale search for dietary trials and cardiovascular disease covered the Cochrane library, Medline, Embase, CAB abstracts, CVRCT registry, and SIGLE to May 1998 plus bibliographies of collected papers and reviews.9 We carried out a further search, seeking trials on sodium restriction and blood pressure in Medline, Embase, and the Cochrane library (to July 2000). We checked bibliographies of systematic reviews and included trials; the searches were not limited by language.
We included trials in which randomisation was adequate, there was a usual or control diet group, the intervention aimed to reduce sodium intake, the intervention was not multifactorial, the participants were not children, acutely ill, pregnant, or institutionalised, follow up was at least 26 weeks, and data on any of the review outcomes were available.
For this review our primary outcomes were mortality and cardiovascular events, blood pressure, and urinary sodium excretion. We also collected data on quality of life and use of antihypertensive medication.
Two authors (LH and CB) assessed inclusion and validity and carried out data extraction independently in duplicate. Any differences were resolved by discussion and, when necessary, by a third reviewer (SE). For assessment of quality we collected data on randomisation procedure, allocation concealment, blinding of participants, providers of care, outcome assessors, and losses to follow up.10
For urinary sodium excretion and blood pressure we collected data on mean (SD) change from baseline for intervention and control groups at intermediate (latest data point from 6 to 12 months), late (13 to 60 months), and very late (after 60 months) follow up. Four trials provided baseline and follow up values, with SD or SE, but no SD for the change from baseline.11–14 We used three studies in which data were provided at baseline and follow up and mean differences given15–17 to calculate values for the correlations between baseline and follow up (for the control and experimental groups for systolic and diastolic blood pressure values but not for urinary sodium excretion).18 We used a conservative estimate (lowest correlation) to compute the SD of mean changes for four studies without this data. Correlations varied from −5.79 to 0.56.
In factorial trials of calorie and sodium reduction we used only data from the sodium reduction and control groups because, of three such factorial trials, 17 19 20 two showed definite17 or probable19 interaction effects. In one trial data on urinary sodium excretion were not available for sodium reduction groups alone but event and medication data were available and were used in analyses.20 Calorie reduction and calories plus sodium reduction arms were included in a sensitivity analysis.
We attempted to contact authors of all included trials for further information on trial characteristics, quality, and outcomes (including number and type of cardiovascular events, deaths, quality of life assessments, urinary sodium excretion, intake of other nutrients, blood pressure, and weight) as well as information on further published or unpublished trials.
Two trials were cluster randomised. In one small trial 19 general practitioners were randomised to deliver simple advice on low salt diets or no such advice to 77 patients.13 Patient numbers in the intervention and control groups were reduced to an effective sample size as described by Hauck,21 assuming the intraclass correlation (appropriate for nonfamilial clusters such as randomised practice units) to be 0.5.22 The other cluster randomised trial individually randomised “index” men and women and then included members of their families in the trial.23 We used only the “index” participants in our meta-analysis.
We checked the meta-analyses (weighted mean differences, random effects model, on Cochrane Collaboration Review Manager 4.1 software) for heterogeneity by visual inspection and by Cochran's test. We used sensitivity analysis to assess the robustness of the results to exclusion of the data with estimated SDs, or use of the largest correlations to estimate these SDs, exclusion of trials with unknown or inadequate allocation concealment, and addition of weight reduction arms. 10 24 We used the STATA metareg command25 for random effects meta-regression.26 We did not use funnel plots to investigate the presence of publication bias because the number of trials in each group was too small.
We used subgrouping of trials and meta-regression to examine the effects on blood pressure of length of follow up on sodium excretion and blood pressure, initial systolic blood pressure, presence or absence of hypertension, age, and change in sodium excretion.
We included three trials in people without hypertension (n=2326), 16 17 19 five in people with untreated hypertension (n=387), 11 13 15 23 27 and three in people with treated hypertension (n=801), 12 14 20 with follow up from six months to seven years. The people without hypertension were healthy (predominantly white men, mean age 40 years) with high normal blood pressure. The people with untreated hypertension were aged 16 to 64 years, while those with treated hypertension were aged 55 to 67 years. In trials on people with hypertension, sex and ethnic characteristics were generally poorly documented. All the trials in people without hypertension, but only one trial in people with treated hypertension,20 used a comprehensive behavioural change programme, whereas the others used varying types of advice or leaflets.
The quality of the trials, as judged by concealment of allocation, seemed higher in the trials in people without hypertension. Other aspects of quality that we assessed included blinding of outcome assessment and losses to follow up (table 1). There were different methods of dealing with missing data associated with losses to follow up. Most trials attempted to blind outcome assessors.
Mortality and cardiovascular events
Mortality and cardiovascular events were inconsistently reported. No differences in periods of admission to hospital were seen between intervention groups in the hypertension prevention trial (no further data were provided).19 Morgan et al reported that three participants in the control group and two participants on low sodium diets were treated for cardiac failure, with four cardiovascular deaths in the low sodium group and two in the control group. 15 28 The trial of non-pharmacological interventions in elderly people recorded a wide range of cardiovascular events: 57 in control participants and 44 in those on low sodium diets (relative risk 0.77, 95% confidence interval 0.41 to 1.14). 20 29 However, only nine of these events were due to stroke or myocardial infarction. Overall, the trials reported few deaths: nine in control groups and eight in low sodium groups.
Table 2 shows changes in blood pressure and urinary sodium excretion for each trial, and table 3 shows pooled changes at intermediate and late assessments (fig 2). Reductions in systolic and diastolic blood pressure were apparent at both intermediate (2.5 mm Hg, 3.8 to 1.2; 1.2 mm Hg, 1.8 to 0.7, respectively) and late follow up (1.1 mm Hg, 1.8 to 0.4; 0.6 mm Hg, 1.5 to 0.3). When we carried out sensitivity analyses excluding low quality trials, which included all trials on people with untreated hypertension, the statistical heterogeneity that had been apparent for systolic blood pressure at intermediate follow up and diastolic blood pressure at late follow up was no longer apparent. As these trials were small the effect on pooled estimates of change in blood pressure was minor. Sensitivity analyses including all arms of the factorial trials, 17 19 suggest that inclusion of weight reduction arms reduces the effect on blood pressure. Meta-regression of change in blood pressure up to 12 months that used all trials with relevant data showed no relation with change in urinary sodium excretion, baseline systolic blood pressure, or age (table 4).
Urinary sodium excretion
We found reductions in urinary 24 hour sodium excretion at both intermediate (48.9 mmol/24 hours, 65.4 to 32.5) and late follow up (35.5 mmol/24 hours, 47.2 to 23.9) (fig 3). We identified significant heterogeneity in both analyses that was not explained by trial quality. One trial in people without hypertension found that at seven years sodium excretion in a small subset of the original sample was similar in intervention and control groups.30
Quality of life
Information on quality of life was patchy, with no common outcome measures. The hypertension prevention trial asked participants whether they were having problems with their diets.31 Of those in the low sodium group, 69% reported problems such as inconvenience and difficulty with adherence when eating out at some time during the three years of the trial, and problems were reported at 42% of clinic visits.
The trials of hypertension prevention, phase I (TOHP I), reported psychological wellbeing scores, which improved significantly in participants in the low sodium groups at 18 months compared with the non-intervention control group.32 Thaler et al reported that participants did not find it difficult to stop adding salt at table, but many found cutting down on salt in cooking harder.23 Most found their low salt bread (salt cut from 2.1% to 1.0% dry weight) and salt-free butter acceptable. Only 13% of participants reported their salt restricted diet as unpleasant or worse. Overall dropout rates, a possible marker of quality of life on trial, were similar (relative risk 1.04; 0.86 to 1.25) in low sodium and control groups.
Antihypertensive medications used
Low salt diets seemed to allow people with hypertension to stop taking medication. In one small trial that compared 10 men in each group, six on low sodium diets had not restarted antihypertensive drugs at six months compared with only one in the control group (relative risk 0.44; 0.20 to 0.98).14 In a larger study of 975 participants, primary end points (a combination of high blood pressure at any visit, restarting antihypertensive medication, or any clinical cardiovascular disease) were less common in the low sodium group (relative risk 0.83, 0.75 to 0.92).20
Eleven long term randomised controlled trials of dietary salt reduction (including 3491 participants) provided few data on mortality (17 deaths in total), cardiovascular events, or quality of life but did show significant falls in systolic blood pressure (1.1 mm Hg, 1.8 to 0.4) and urinary sodium excretion (35.5 mmol/24 hours, 47.2 to 23.9) at 13 to 60 months after initial advice. Falls in diastolic blood pressure were smaller and were consistent with no effect (0.6 mm Hg, 1.5 to 0.3). A low salt diet may help people on antihypertensive drugs to stop their medication without loss of blood pressure control.
Limitations and strengths of review
Health promotion interventions involve several stages before any health outcome is seen. Firstly, the advice must result in changed behaviour (cutting down on salt in food) and, secondly, that behaviour must result in an improved health outcome (reduced cardiovascular illness, increased life expectancy). A major limitation of this review is that we were not able to assess the overall effect of advice to reduce dietary sodium on mortality or morbidity as too few events occurred. Instead we assessed several intermediate outcomes including urinary sodium excretion and blood pressure.
The observed sodium reduction of about a quarter of usual intake in US and UK populations may be an overestimate.33 Almost half the participants in one trial ate differently on food record days, eating less food and substituting simpler foods and also eating less salt.34 The completeness of urine samples is not known and it has been suggested that less salty foods were eaten on collection days in the trial of Thaler et al23 (O Simpson, personal communication, 2001).
While both urinary sodium excretion and blood pressure fell, the salt reduction may not have caused the fall in blood pressure. Alterations in diet aimed at reducing salt intake may systematically affect other dietary components (such as alcohol, potassium, or energy intake) that themselves alter blood pressure. This might explain why we found no relation between the degree of reduction in sodium excretion and change in blood pressure. However, the number of trials is small and relating a mean change in blood pressure to a mean change in urinary sodium is statistically weak. In previous meta-analyses an association was seen in some cases but not others (table 5). Data on individual participants are required to take this issue further.
Despite the importance of answering the question of the long term effects of dietary salt restriction, most of the many randomised controlled trials published have been of short duration and can show only that salt restriction is capable of reducing blood pressure but provide no useful information for primary care practice. As randomised controlled trials are available, we have not included population surveys, cohorts, or animal trials that are unable to estimate the unconfounded effects of salt restriction in human populations and are difficult to interpret.35
Is it realistic to ask people to alter their salt intake long term? Advice to reduce dietary salt is common in primary care and is a central part of the guidelines produced by the British Hypertension Society.36 Despite a great deal of ongoing encouragement and support used in the trials included in this review, it seems that salt reduction attenuates over time. In routine primary care the intervention is likely to be less intense and therefore of more limited impact.
Comparison with previous studies
It is unclear what effects a low sodium diet has on cardiovascular events and mortality. Lowering sodium intake may have adverse effects on vascular endothelium through stimulation of the renin-angiotensin system37 and on serum total and low density lipoprotein cholesterol concentrations.4 In cohort studies, lower salt intake in people with hypertension has been associated with higher levels of cardiovascular disease38 and in general populations with greater all cause mortality. 39 40 However, among obese people lower salt intake may be associated with a reduced risk of cardiovascular events. 41 42 These apparently contradictory findings may be explained by confounding or by differential sensitivity to salt intake but make it less clear that salt restriction is without hazards.
We expected that short duration trials would achieve larger reductions in blood pressure that would attenuate over time. As shown in table 5, short term trials of median length of eight days showed a greater reduction in urinary sodium excretion but a similar fall in systolic blood pressure to the findings from long term trials of median length 36 months in this review. The recent dietary approaches to stop hypertension (DASH) trial showed that over a 30 day period with intensive measures, which included provision of all food, systolic blood pressure can fall substantially (by 6.7 mm Hg, 5.4 to 8.0 mm Hg),43 but this finding is of little relevance to the issue of achieving long term reductions in blood pressure by practical means in primary care.
Long term maintenance of low sodium intake is difficult, even with a great deal of support, advice, and encouragement. A policy of reduction in salt intake for the entire population through cutting salt concentrations in processed foods,44 as recently announced by the UK chief medical officer,45 can achieve small reductions in blood pressure across the whole population for sustained periods of time. Individual reduction of risk would be small, but across a whole population the effects may be substantial. 46 47
However, raised blood pressure is only one risk factor for cardiovascular disease and overall clinical benefits (or harms) of a low sodium diet are unclear. Revisiting all participants of the large trials in people without hypertension some years later to assess long term effects of low sodium dietary advice on mortality and cardiovascular morbidity would be a cost effective and relatively rapid way to assess the clinical effectiveness of advice to reduce sodium intake. There is strong justification for a large scale, long term randomised controlled trial to explore the cost effectiveness of such advice if it is to remain a part of the strategy for prevention and treatment of hypertension.
On present evidence intensive interventions, unsuited to primary care or population prevention programmes, produce uncertain effects on mortality and cardiovascular events and only small reductions in blood pressure. However, advice to reduce sodium intake in the diet may help some people on antihypertensive drugs to stop their medication while maintaining good control of blood pressure.
This study was conducted as a Cochrane systematic review under the auspices of the Cochrane Hypertension and Heart Groups, whose assistance is gratefully acknowledged. We thank all the trialists and experts who kindly provided unpublished information, including Bruce Arroll (University of Auckland), Olaf Simpson (Dunedin), Susan Tonascia (Johns Hopkins University), Trefor Morgan (University of Melbourne), and Alexander Logan (Toronto), as well as Paul Durrington and Helen Worthington (PhD supervisors for LH).
Contributors: All authors were actively involved in the design of the review, checking the data, and critical revisions to the manuscript, which was drafted by LH. LH and CB independently searched, decided on trial inclusion or exclusion, extracted data, and assessed study quality. LH, CB, and SE performed and duplicated the statistical analyses. SE and GDS were primary advisers, guiding and interpreting the review. LH is the guarantor.
Funding North West Research and Development Training Fellowship (LH).
Competing interests LH owns 285 shares in West Indies Rum Distillery, Barbados.