Salt intake, stroke, and cardiovascular disease: meta-analysis of prospective studies
BMJ 2009; 339 doi: https://doi.org/10.1136/bmj.b4567 (Published 25 November 2009) Cite this as: BMJ 2009;339:b4567- Pasquale Strazzullo, professor of medicine1,
- Lanfranco D’Elia, clinical lecturer in medicine1,
- Ngianga-Bakwin Kandala, principal research fellow in medical statistics2,
- Francesco P Cappuccio, professor of cardiovascular medicine and epidemiology2
- 1Department of Clinical and Experimental Medicine, “Federico II” University of Naples Medical School, Naples, Italy
- 2University of Warwick, WHO Collaborating Centre for Nutrition, Warwick Medical School, Clinical Sciences Research Institute, Coventry CV2 2DX
- Correspondence to: P Strazzullo strazzul{at}unina.it, FP Cappuccio f.p.cappuccio{at}warwick.ac.uk
- Accepted 22 October 2009
Abstract
Objective To assess the relation between the level of habitual salt intake and stroke or total cardiovascular disease outcome.
Design Systematic review and meta-analysis of prospective studies published 1966-2008.
Data sources Medline (1966-2008), Embase (from 1988), AMED (from 1985), CINAHL (from 1982), Psychinfo (from 1985), and the Cochrane Library.
Review methods For each study, relative risks and 95% confidence intervals were extracted and pooled with a random effect model, weighting for the inverse of the variance. Heterogeneity, publication bias, subgroup, and meta-regression analyses were performed. Criteria for inclusion were prospective adult population study, assessment of salt intake as baseline exposure, assessment of either stroke or total cardiovascular disease as outcome, follow-up of at least three years, indication of number of participants exposed and number of events across different salt intake categories.
Results There were 19 independent cohort samples from 13 studies, with 177 025 participants (follow-up 3.5-19 years) and over 11 000 vascular events. Higher salt intake was associated with greater risk of stroke (pooled relative risk 1.23, 95% confidence interval 1.06 to 1.43; P=0.007) and cardiovascular disease (1.14, 0.99 to 1.32; P=0.07), with no significant evidence of publication bias. For cardiovascular disease, sensitivity analysis showed that the exclusion of a single study led to a pooled estimate of 1.17 (1.02 to 1.34; P=0.02). The associations observed were greater the larger the difference in sodium intake and the longer the follow-up.
Conclusions High salt intake is associated with significantly increased risk of stroke and total cardiovascular disease. Because of imprecision in measurement of salt intake, these effect sizes are likely to be underestimated. These results support the role of a substantial population reduction in salt intake for the prevention of cardiovascular disease.
Introduction
During the past century, the evidence for the risks imposed on human health by excess salt consumption has become compelling. The causal relation between habitual dietary salt intake and blood pressure has been established through experimental, epidemiological, migration, and intervention studies. Most adult populations around the world have average daily salt intakes higher than 6 g, and for many in eastern Europe and Asia higher than 12 g. International recommendations suggest that average population salt intake should be less than 5-6 g per day. Population based intervention studies and randomised controlled clinical trials have shown that it is possible to achieve significant reductions in blood pressure with reduced salt intake in people with and without hypertension.1 Based on the effects of high salt intake on blood pressure and on the prominent role of high blood pressure in promoting cardiovascular diseases, it has been suggested that a population-wide reduction in salt intake could substantially reduce the incidence of cardiovascular disease.2 On the basis of the results of a meta-analysis of randomised controlled trials of salt reduction,3 it was estimated that a reduction in habitual dietary salt intake of 6 g a day would be associated with reductions in systolic/diastolic blood pressure of 7/4 mm Hg in people with hypertension and 4/2 mm Hg in those without hypertension. At the population level these reductions in blood pressure could predict an average lower rate of 24% for stroke and 18% for coronary heart disease.4 Validation of these predictions by a randomised controlled trial of the effects of long term reduction in dietary salt on morbidity and mortality from cardiovascular disease would provide definite proof. At present, a study of this kind is not available and, in fact, it is extremely unlikely that it will ever be performed because of practical difficulties, the long duration required, and high costs. Nevertheless, prospective cohort studies performed in the past three decades that measured the levels of dietary salt intake at baseline and recorded the incidence of vascular events have provided important indirect evidence. Most of these studies found evidence of such relation, although few had enough power to attain statistical significance.
We performed a systematic review and meta-analysis of the prospective studies of habitual dietary salt intake and incidence of stroke and total cardiovascular disease using strictly predetermined criteria for inclusion or exclusion. We assessed whether or not the overall evidence in prospective studies supports the presence of a relation between levels of dietary salt intake and both stroke and cardiovascular outcomes and calculated an estimate of the risk.
Methods
Data sources and searches
We performed a systematic search for publications using Medline (1966-2008), Embase (from 1988), AMED (from 1985), CINAHL (from 1982), Psychinfo (from 1985), and the Cochrane Library. Search strategies used subject headings and key words with no language restrictions. Further information was retrieved through a manual search of references from recent reviews and relevant published original studies. We examined reference lists of the relevant reviews, identified studies, and reviewed the cited literature.5
Study selection
Two reviewers (LD and N-BK) independently extracted the data. Discrepancies about inclusion of studies and interpretation of data were resolved by arbitration (PS or FPC), and consensus was reached after discussion. In the case of missing data for potentially suitable studies, we contacted authors and asked them to provide the necessary information. To be included in the meta-analysis a published study had to be an original article published from January 1966 to December 2008, be a prospective population study, assess salt intake as baseline exposure, determine either stroke or total cardiovascular disease prospectively as the outcome, follow participants for at least three years, include an adult population, and indicate the number of participants exposed and the rate or number of events in different categories of salt intake.
Of the 3246 publications retrieved, we identified 15 studies that met the inclusion criteria. One was a duplicate analysis of a single cohort previously described by the same authors6 7 and another8 referred to the same cohort (national health and nutrition examination survey (NHANES) I) analysed by other authors with more stringent criteria.9 We therefore included 13 studies in the meta-analysis that provided suitable data on 19 population samples6 10 11 12 13 14 15 16 17 18 19 20 21 (tables 1⇓ and 2⇓).
Data extraction
From the identified studies and respective populations we recorded publication reference, total number of participants, country, sex, age (mean, median, or range), recruitment time, follow-up (years), outcome reported (stroke, cardiovascular disease) and method of outcome assessment, number (rate) of events, method of assessing salt intake, and level of salt intake in different categories.
Categorisation of salt intake differed among studies. Some reported the number of subjects exposed and the rate (number) of events across the distribution of salt intake; others reported differences in the event rate for a 100 mmol/day difference in sodium intake, as in the studies by He et al9 and Tuomilehto et al.13 In the last two cases we used the relative risk or hazard ratio reported by the authors for the analysis. In all the cases in which categorisation of the study participants by level of salt intake was available, we calculated the relative risk of higher versus lower salt intake by comparing the event rate in the two categories with a difference in average salt intake closest to 100 mmol of sodium or about 6 g of salt a day.
Statistical analysis
We evaluated the quality of the studies included in the meta-analysis with the Downs and Black score system.21 We extracted relative risks or hazard ratios from the selected publications and calculated their standard errors from the respective confidence intervals. The value from each study and the corresponding standard error were transformed into their natural logarithms to stabilise the variances and to normalise their distribution. The pooled relative risk (and 95% confidence interval) was estimated with a random effect model, weighting for the inverse of the variance.22 The heterogeneity among studies was tested by Q statistic and quantified by H statistic and I2 statistic.23 The influence of individual studies, from which the meta-analysis estimates are derived, was examined by omitting one study at a time to see the extent to which inferences depend on a particular study or group of studies (sensitivity analysis). Subgroup or meta-regression analyses were used to identify associations between risk of stroke or cardiovascular disease and relevant study characteristics (age and sex of participants, year of publication, duration of follow-up, method of assessment of sodium intake, difference in sodium level, control for baseline blood pressure) as possible sources of heterogeneity. We used funnel plot asymmetry to detect publication bias and applied Egger’s regression test to measure any asymmetry.24 25 All statistical analyses were performed with MIX software version 1.726 and Stata software for meta-regression analysis.
Results
Characteristics of the study cohorts
We included in the meta-analysis 13 studies reporting on 19 independent cohorts (table 1).⇑ There were 177 025 participants from six different countries (six studies from the United States, two each from Finland and Japan, one each from the Netherlands, Scotland, and Taiwan). Eleven studies recruited both male and female participants, while two studies included only men. Follow-up ranged from 3.5 to 19 years. Four studies reported only stroke events (either total stroke rate or stroke deaths), three only cardiovascular disease (total cardiovascular disease rate or cardiovascular disease deaths), and six reported both. Salt intake was assessed by 24 hour dietary recall (n=4), food frequency questionnaire (n=4), 24 hour urine excretion (n=4), and questionnaire (n=1). In total there were 5346 strokes reported and 5161 total cardiovascular disease events. Of the 11 studies that included both men and women, five reported outcomes separately.6 9 13 14 15 The TOHP study included two different cohorts (I and II)17 and the study by He and coworkers provided separate findings for men and women or, alternatively, for normal weight and overweight participants.9 Overall, data on the relation between salt consumption and stroke were available from 14 cohorts and on the relation between salt intake and cardiovascular disease from 14 cohorts.
The overall study quality, evaluated by the Downs and Black score, averaged 15.5 (range 12-18) on a scale of 19 (table 1).⇑
Salt intake and risk of stroke
Table 2 provides data on the relation between salt intake and risk of stroke in each of the 14 cohorts included in our study.⇑ Figure 1 shows the results of the pooled analysis⇓. In the pooled analysis, higher salt intake was associated with greater risk of stroke (relative risk 1.23, 95% confidence interval 1.06 to 1.43; P=0.007). There was significant heterogeneity between studies (P=0.04; I2=61%). The funnel plot did not show asymmetry, thus excluding publication bias (Egger’s test P=0.26; see appendix on bmj.com). As shown in figure 1 for the individual cohorts included in the analysis, we found a trend towards a direct association between salt intake and risk of stroke in nine cohorts, which was significant in four. We observed a non-significant inverse trend in three cohorts.
Sensitivity analysis showed that the pooled estimate of the effect of salt intake on risk of stroke did not vary substantially with the exclusion of any one study; in particular, the exclusion of the study by Umesawa et al,19 which accounted for about 40% of all participants in the meta-analysis and nearly 20% of all strokes, resulted in a pooled relative risk of 1.19 (1.03 to 1.39), P=0.022.
Salt intake and risk of cardiovascular disease
Table 2 provides data on the association between salt intake and the risk of cardiovascular disease in 14 cohorts⇑. In the pooled analysis, there was an association between higher salt intake and risk of cardiovascular disease (1.14, 0.99 to 1.32; P=0.07) (fig 2⇓). The heterogeneity between studies was significant (P<0.01; I2=80%), but the funnel plot did not show asymmetry, thus excluding publication bias (Egger’s test: P=0.39; see appendix on bmj.com). The evaluation of individual studies showed a trend towards a direct association between salt intake and risk of cardiovascular disease in 10 cohorts, with significantly higher relative risk in six. An inverse trend was observed in four cohorts and was significant in one. Sensitivity analysis showed that the exclusion of the only study showing a significant inverse trend6 led to a pooled estimate of relative risk of 1.17 (1.02 to 1.34), P=0.02 (fig 2). Further exclusion of the study by Umesawa et al,19 which accounted for over 50% of all participants and about 40% of all cardiovascular disease events led to a pooled relative risk of 1.14 (0.99 to 1.31), P=0.06.
Sources of heterogeneity
Age—Meta-regression analyses indicated no association between mean age of study participants and effect of sodium intake on the risk of stroke: exp(b)=1.01 (0.99 to 1.03). Likewise, meta-regression showed no association between age and effect of sodium intake on the risk of cardiovascular disease: exp(b)=0.99 (0.97 to 1.02).
Sex—Three studies reported data for men and women separately for incidence of stroke.6 13 14 The pooled estimates from these three studies were 1.30 (0.64 to 2.65; P=0.47) and 1.56 (1.14 to 2.13; P<0.01), respectively. Three studies reported data for men and women separately for incidence of cardiovascular disease.9 12 13 The pooled estimates were 1.31 (0.97 to 1.77; P=0.08) and 1.27 (1.05 to 1.55; P=0.01), respectively.
Method of assessment of sodium intake—In nine cohorts that used food frequency questionnaires or dietary recall for the evaluation of habitual sodium intake the pooled risk estimate for stroke was 1.25 (1.03 to 1.51; P=0.02).10 11 12 14 15 18 19 In five cohorts that used 24 hour urine collection the pooled risk estimate was 1.16 (0.94 to 1.44; P=0.17).6 13 16 In five cohorts that used food frequency questionnaires or dietary recall the pooled estimate for cardiovascular disease was 1.21 (0.92 to 1.59; P=0.17),9 15 19 20 and in nine cohorts that used 24 hour urine collection the pooled risk estimate for cardiovascular disease was 1.10 (0.92 to 1.31; P=0.32).6 12 13 16 17
Baseline blood pressure or hypertension status—In the studies that provided relative risk estimates adjusted for baseline blood pressure or hypertension status, the pooled relative risk was 1.22 (1.02 to 1.45; P=0.03) for stroke (nine cohorts) and 1.25 (0.99 to 1.57; P=0.06) for cardiovascular disease (seven cohorts).
Baseline body mass index (BMI) or body weight—In the studies that provided relative risk estimates adjusted for baseline BMI or body weight, the pooled relative risk was 1.20 (1.02 to 1.40; P=0.02) for stroke (10 cohorts) and 1.22 (1.00 to 1.49; P=0.05) for cardiovascular disease (10 cohorts).
Length of follow-up—Meta-regression analysis showed a significant association between duration of follow-up and the effect of sodium on the risk of stroke. The log relative risk was estimated to increase by 0.07 per increase of one year of follow-up: exp(b)=1.07 (1.04 to 1.10). The estimated variance between studies (heterogeneity) was reduced from 0.05 to 0.02. In contrast, however, we found no association between duration of follow-up and effect of sodium on the risk of cardiovascular disease: exp(b)=0.98 (0.95 to 1.02).
Dose-response analysis—Variance weighted least squares regression of the risk of stroke on the study specific difference between higher and lower categories of sodium intake (differences in sodium intake values reported in table 2) provided evidence of a significant direct association (exp(b)=1.06 (1.03 to 1.10)), indicating a 6% increase in the rate of stroke for every 50 mmol/day difference in sodium intake. There was a similar trend for the risk of cardiovascular disease (exp(b)=1.19 (0.69 to 2.07)), that was not significant.
Time trend (year of publication)—Starting with the first published study10 we calculated the cumulative pooled relative risk by stepwise addition of the results of the other available studies up to the last one published in July 2008.19 Figure 3⇓ shows the results of these cumulative meta-analyses. The pooled relative risk for stroke stabilised early in the 1.20-1.30 interval and achieved significance starting in 2001. Similar results were obtained in the analysis of cardiovascular disease, for which the pooled relative risk estimate also stabilised early and close to the final value, achieving significance starting in 1999.
Discussion
This meta-analysis shows unequivocally that higher salt intake is associated with a greater incidence of strokes and total cardiovascular events. Our systematic review identified 13 relevant and suitable studies published from 1996 to 2008. These studies provided evidence from 170 000 people contributing overall more than 10 000 vascular events.
Cardiovascular diseases are the major cause of death among people aged over 60 and second among those aged 15-59. According to the World Health Organization, 62% of all strokes and 49% of coronary heart disease events are attributable to high blood pressure.27 The direct causal relation between levels of dietary salt intake and blood pressure at the population level has also been recognised.1 28 29 Given the graded causal relation between blood pressure and cardiovascular disease, beginning at around 115 mm Hg systolic pressure,30 it is reasonable to expect considerable benefit on the rate of cardiovascular disease from a reduction in salt intake.
Association between salt intake, stroke, and cardiovascular disease
The results of this meta-analysis provide evidence of a direct association between high dietary salt intake and risk of stroke. Despite the considerable heterogeneity between the 14 cohorts available for the analysis, the results are strengthened by the lack of major publication bias and by the observation of a significant association in four individual cohorts included in the analysis, whereas in none was an inverse statistical association apparent. The pooled relative risk indicates a 23% greater risk of stroke for an average difference in sodium intake (weighted for the population size of each study) of 86 mmol (equivalent to about 5 g of salt a day). Sensitivity analysis with the exclusion of a single study, on the basis of its particular weight with regard to both number of participants and events, only moderately reduced the difference in risk (from 23% to 19%), which remained significant.
Likewise, the pooled analysis of the 12 cohorts for which data on cardiovascular disease outcome were available (after the exclusion of a single outlier) showed a direct association between higher salt intake and risk of cardiovascular disease, with a pooled relative risk of 1.17.4 A trend in this direction occurred in as many as nine of the 12 cohorts and was significant in six. There was an inverse trend in three cohorts.15 16 20 The study by Alderman et al,6 showing a relative risk in men of 0.37, has been challenged because of the low number of events recorded and several methodological inadequacies, the most important being the evaluation of habitual salt consumption on the basis of 24 hour urine collection obtained shortly after the study participants had been instructed to reduce their usual level of sodium intake.31 The results of sensitivity analysis indicate that the exclusion of this single study from our meta-analysis strengthens the estimate. The additional exclusion of a large Japanese cohort providing a high proportion of participants and events overall19 only slightly reduced the pooled relative risk estimate (from 1.17 to 1.14) and the level of significance (to 0.06).
Evaluation of main sources of heterogeneity
We used subgroup and meta-regression analyses to assess the influence of several factors on the association between habitual sodium intake and risk of stroke or cardiovascular disease. For both stroke and cardiovascular disease outcomes, separate analyses of the male and female cohorts suggest that the associations are consistent and not significantly different between the sexes. Similar results were obtained with respect to the method of assessment of habitual sodium intake used in the various studies.
Eight studies provided data adjusted for baseline blood pressure or hypertension status. Separate evaluations of these studies provided relative risk estimates for both stroke and cardiovascular disease similar to those obtained for the total number of studies included in the meta-analysis. This finding seems at variance with the hypothesis that the effect of salt on cardiovascular risk is substantially mediated by its unfavourable action on blood pressure. Adjustment for baseline blood pressure or hypertension status only partially corrects for the overall influence of blood pressure on the study results in as much as it does not account for changes in blood pressure occurring during the observation period, a problem more relevant the longer the follow-up period. Part of the association observed, however, might be mediated by factors other than blood pressure, and there is evidence in the literature of deleterious effects of high salt intake on left ventricular mass,32 33 34 arterial stiffness,35 and renal function,36 37 which are not totally explained by its effect on blood pressure.
Overweight and obesity are often associated with high blood pressure and are causally involved in the development of hypertension.38 Nine out of 13 studies included in the meta-analysis provided relative risk estimates adjusted for BMI or body weight at entry into the study.9 14 15 16 17 18 19 20 21 Therefore, as for blood pressure, the association between habitual sodium intake and risk of stroke and cardiovascular disease seems partly independent from the influence of excess body weight. Two studies, however, reported a significant interaction between overweight and habitual sodium intake on the risk of cardiovascular events.9 13 This finding is consistent with the description of alterations in renal tubular sodium handling in obese individuals, making them particularly sensitive to the effects of high salt intake.39
Study limitations
The studies included in our meta-analysis were heterogeneous regarding sample size, number of events, and duration of follow-up, with a few cohorts having small numbers. In the calculation of the pooled relative risk we weighted the results of the individual studies for sample size but did not account for the duration of follow-up. Our meta-regression analysis indicated that the longer the follow-up the greater the effect of habitual sodium intake on the risk of stroke but not, apparently, on the risk of total cardiovascular events. Possible explanations for this discrepancy are the higher mean age at occurrence of stroke, which would increase the chances of an event the longer the follow-up, and the closer relation of high blood pressure to stroke compared with other types of vascular events.
The estimate of the baseline population salt intake in each study was based on a single measurement (whether through 24 hour urine collection or dietary assessment). We were therefore unable to correct for regression dilution bias. Because of the large day to day variability within people in salt consumption and the consequent diluting effect imposed on the average estimate of exposure, our estimates of risk are probably underestimated.
Categorisation of salt intake was also heterogeneous: some studies stratified the population by categories of sodium intake and compared cardiovascular outcomes across categories, other studies gave a difference in outcome for a given difference (for example, 100 mmol/24 h) in sodium intake or excretion. To standardise our comparison between higher and lower salt consumption we sought to refer to a difference as close as possible to 100 mmol or 6 g a day between high and low salt intake. Nevertheless, there remained appreciable differences in this respect between studies. We tried to overcome the problem with meta-regression analysis, which provided evidence of a highly significant dose-response relation between the difference in sodium intake and the increase in risk of both stroke and cardiovascular disease.
Implications
The habitual salt intake in most Western countries is close to 10 g a day (and much higher in many Eastern European and Asian countries), and we calculated that the average difference between higher and lower salt intake across the study cohorts included in our meta-analysis was 5 g a day. Given this approach, we believe that, despite the inherent inaccuracies, the results of our meta-analysis are applicable to real life conditions. A reduction of 5 g (about one teaspoon) of salt would bring consumption close to the WHO recommended level (5 g a day at the population level).
According to a recent report of the World Heart Federation there are over 5.5 million deaths a year from stroke throughout the world and close to 17.5 million deaths a year from cardiovascular disease.40 Given that the case fatality rates for stroke is estimated at one in three and those for total cardiovascular disease at one in five, a 23% reduction in the rate of stroke and a 17% overall reduction in the rate of cardiovascular disease attributable to a reduction in population salt intake could avert some one and a quarter million deaths from stroke and almost three million deaths from cardiovascular disease each year. Many studies have also shown that a reduction in salt intake is cost effective,41 42 43 arguing for the more widespread introduction of national programmes to reduce dietary salt consumption. In recent years, a few countries have made some progress towards reduction of habitual salt intake through a voluntary approach1 or by regulation, as in Finland,42 but levels of salt consumption are still far from the WHO recommended targets. There are many reasons for these delays. One barrier to a more effective implementation of public health policies has been the historical opposition of the food industry,44 based on the arguments that the available evidence does not show significant benefits on hard end points at a population level from a moderate reduction in salt intake. Our study now clearly addresses those doubts. Some progress has been made in the past few years by closer collaboration between governments, public health bodies, and some sectors of the industry on a “voluntary” basis, as in the UK, with the reformulation of many food items towards a lower salt content and proposals of improved labelling. These efforts have led to a reduction of 0.9 g a day (or about 10%) in population salt intake in four years (from 9.5 to 8.6 g a day), still far from the recommended 6 g a day initial targets that were set in the UK. While the voluntary approach is the preferred choice for many governments, the “regulatory” approach has advantages,45 sometimes being the most efficient, effective, and cost effective way of achieving public health targets.41 42 43 For population salt intake to approach the recommended targets within a reasonable time frame, an “upstream” approach is now necessary alongside the traditional “downstream” public health approach based on health promotion and behavioural changes.
What is already known on this topic
Experimental, epidemiological, migration, and intervention studies have shown a causal relation between habitual dietary salt intake and blood pressure
Population based intervention studies and meta-analyses of randomised controlled trials have shown that it is possible to achieve significant reductions in blood pressure with reduced salt intake in both hypertensive and normotensive individuals
What this study adds
Higher salt intake is associated with significantly greater incidence of strokes and total cardiovascular events, with a dose dependent association
A difference of 5 g a day in habitual salt intake is associated with a 23% difference in the rate of stroke and 17% difference in the rate of total cardiovascular disease
Each year a 5 g reduction in daily salt intake at the population level could avert some one and a quarter million deaths from stroke and almost three million deaths from cardiovascular disease worldwide
Notes
Cite this as: BMJ 2009;339:b4567
Footnotes
Contributors: PS and FPC conceived the study aims and design, contributed to the systematic review and data extraction, performed the analysis, interpreted the results, and drafted the manuscript. LD’E and N-BK contributed to the data extraction, interpretation of results, and revision of the manuscript. PS is guarantor.
Funding: This study was funded in part by an EC Grant (FP7-HEALTH- 2007-201550). The publication does not necessarily represent the decisions or the stated policy of WHO and the designations employed and the presentation of the material do not imply the expression of any opinion on the part of WHO.
Competing interests: None declared.
Ethical approval: Not required.
Data sharing: Search strategy and flow chart are available from the corresponding authors by e-mail.
This is an open-access article distributed under the terms of the Creative Commons Attribution Non-commercial License, which permits use, distribution, and reproduction in any medium, provided the original work is properly cited, the use is non commercial and is otherwise in compliance with the license. See: http://creativecommons.org/licenses/by-nc/2.0/ and http://creativecommons.org/licenses/by-nc/2.0/legalcode.