INTRODUCTION
- Several prospective cohort studies have reported that the
association between sodium consumption and cardiovascular disease or mortality
is U-shaped, with increased risk at both high and low sodium intake
- A meta-analysis of 23 epidemiological studies (n=274 683) also
reported a U-shaped relation (Graudal et al. Am J Hypertens 2014; 27:
1129-37)
- In view that increasing sodium intake is related to
increased blood pressure, and that this is steeper in those individuals with
hypertension compared with in those without hypertension, we hypothesised that
there might be differences in the association between sodium intake and
cardiovascular disease outcomes in individuals with hypertension compared with
in those without hypertension
OBJECTIVE:
- To explore whether the association between sodium intake and
cardiovascular disease events and all-cause mortality is modified by hypertension status.
- To compare the observed magnitude (and pattern) of association between sodium
intake and clinical events with the predicted hazard ratio (HR) derived from modelling the
association between sodium intake and blood pressure
[Assuming that all reductions in blood pressure should translate into cardiovascular disease reduction, with no other off -target effects (eg, activation of the renin system or increases in blood lipids)]
[Assuming that all reductions in blood pressure should translate into cardiovascular disease reduction, with no other off -target effects (eg, activation of the renin system or increases in blood lipids)]
METHODS
Study design and
participants
- Details of the studies’ designs and population characteristics
have been published before and are described in the appendix (pp 2-6).
- In brief:
1) The Prospective Urban Rural Epidemiological Study (PURE Study) is an ongoing large-scale
epidemiological cohort study that
has enrolled 156 424 individuals between 35 years and 70 years from the
population in 628 communities in 17 low-income, middle-income, and high-income
countries on five continents.
- The sampling strategy used in PURE ensures representation from
urban and rural communities from different geographical areas.
- For this analysis, we included 101 511 participants from
PURE who collected morning fasting urine samples suitable for analysis and with
baseline blood pressure measurements.
2) The EPIDREAM trial was a prospective cohort study of 17 453 individuals, aged 18-85 years, who were
screened for eligibility to enter the DREAM clinical trial (a randomised,
double-blind trial with a 2 × 2 factorial design that assigned participants at
high risk for type 2 diabetes to receive either ramipril [15 mg/day] vs placebo
or rosiglitazone [8 mg/day] vs placebo).
- The EPIDREAM cohort included participants who were
screened for the study and includes those who entered DREAM and those who were
not included in the trial and agreed to a long-term prospective follow-up
- For this analysis, to conserve power and at the same time
to be efficient on resources, we used a
case-cohort design to select all individuals who developed a cardiovascular
disease event (n=478) during the follow-up of the EPIDREAM cohort and a control
group comprised of a random sample of individuals (n=2372; five controls per
case) who did not develop a cardio vascular disease event.
3) ONTARGET was a randomised, double-blind, parallel trial
comparing the effects of ramipril (10 mg/day), telmisartan (80 mg/day), and
combination therapy of ramipril (10 mg/day) and telmisartan (80 mg/day) in 25
620 patients, aged 55 years or older, with vascular disease or high-risk
patients with diabetes.
4) TRANSCEND was a randomised controlled trial comparing
telmisartan (80 mg/day) with placebo in 5926 participants who were intolerant
to angiotensin converting enzyme (ACE) inhibitors
For this analysis, we included 28 757 participants from ONTARGET
and TRANSCEND with morning fasting urine samples and with baseline blood
pressure measures.
- All studies were coordinated by the Population Health
Research Institute, Hamilton Health Sciences and McMaster University, Hamilton, ON, Canada.
Procedures
Urine sample collection and analysis
- A morning fasting urine sample was collected from every participant
and shipped in ambient packaging (Saf-T-Pak) for analysis at the Clinical
Research and Clinical Trials
Laboratory at Hamilton General Hospital in Hamilton, ON,
Canada (the central laboratory), or the regional laboratory in Beijing,
Bangalore, India, or Kocaeli, Turkey, for analyses with the use of validated and
standardised methods.
- A description of the methods used for urinary analyses has
been described previously. (O’Donnell et al. N Engl J Med 2014; 371: 612–23; Mente et al., N Engl J Med 2014; 371: 601–11)
Validation of Kawasaki formula
- We used the Kawasaki
formula to estimate 24-h urinary excretion of sodium and potassium from a fasting
morning specimen and used these
estimates as surrogates for daily sodium and potassium intake (in grams).
- Previous studies have reported that this method provides a
reliable estimate of sodium intake in healthy Japanese participants (r=0.73) and
this was replicated later in Japanese participants with hypertension (r=0.69 in
those on blood pressure medication, r=0.66 in those not medicated), and more
recently in Chinese participants with hypertension (r=0.64).
- We did further validation
of the method in 1083 people from 11
countries, and showed that the estimated sodium excretion from the morning
urine specimen shows a good correlation with direct measures of sodium excretion
from the actual 24-h urine collection (intraclass correlation coefficient of 0.70
[95% CI 0.61-0.77]
among individuals with hypertension and 0.71 [0.61 to 0.78] among those without
hypertension). (Mente et al, 2014. J
Hypertens 32:1005-14)
- Further, the blood pressure change per g of sodium was 2.11/0.78
mm Hg, which is consistent with the results of a meta-analysis of sodium
lowering randomised controlled trials (appendix pp 7, 8)
Data Collection and Measurements
- Weight, height, and two recordings of blood pressure after
5 min of rest in a sitting position with the use of an Omron automatic digital
monitor (Omron HEM-757 used in all studies) were recorded in all participants.
- Individuals were considered
hypertensive if their untreated
baseline blood pressure was 140/90 mm Hg or greater or if they were prescribed
antihypertensive drugs at baseline.
- The information about study variables was collected with
similar approaches to measure risk factor variables and data collection forms
in each of the studies.
- Information about personal medical history and use of drugs
were recorded.
- Standardised case-report forms was used to capture data
for major cardiovascular disease events and death during follow-up.
- Events were classified according to the definitions used
in each study, but they were broadly similar.
- For this analysis, we included data from the PURE study
(which is ongoing) through to March, 2015, the complete data from ONTARGET/TRANSCEND,
and case-cohort data from EPIDREAM.
Statistical Analysis
- Mean estimated 24-h urinary excretion values of sodium were
computed overall and by hypertension status.
- Multivariable
linear regression was used to obtain estimates of the slope describing the relation between estimated sodium excretion
(exposure) and blood pressure measurements (outcome variable), within each subpopulation,
with adjustment for age, sex,
body-mass index, education, alcohol intake, current smoking, and geographical
region.
- We examined the
association between an estimated
so-called usual level of sodium excretion (ie, accounting for the degree of
correlation between sodium levels in urine when measured after 30 days and 90
days in 448 individuals; this also allows adjustment for regression dilution
bias) and blood pressure.
- Analysis of
covariance was done, with tests for linear trend, to compare the adjusted mean
blood pressure according to sodium excretion level.
- The primary outcome
was defined as the composite of death, myocardial infarction, stroke, and heart
failure.
- We used restricted
cubic-spline plots with four knots (at the 5th, 35th, 65th, and 95th
percentiles) to explore the shape of
the association between the estimated
sodium excretion and the outcomes.
- Participants were categorised
into urinary sodium excretion groups, based
on 1 g/day increments of excretion
- Because few individuals had excretion values less than 2
g/day or more than 8 g/day, we truncated
excretion values at less than 3 g/day and >7 g/day to avoid small numbers of
individuals at the extreme ends of the distribution (about 10% of
participants in the lowest and highest excretion categories within each
subgroup).
- We calculated HRs of time to event with Cox proportional hazards models, with
shared frailty models. The clustering variable was the study cohort. The
proportional hazards assumption was checked by visual inspection of log-log
plots.
- The primary model included age, sex, ethnicity, BMI, smoking
status, diabetes, educational level, alcohol consumption, physical activity,
past cardiovascular disease events, and treatment allocation (ramipril,
telmisartan, or both, and treatment with statins, β-blockers, diuretic therapy, calcium antagonist, and
antidiabetes medication), as in our previously published papers
- Separate analyses
were done that excluded those individuals
who had had previous cardiovascular
events.
- Interaction tests
were done to assess whether the slopes of the associations between estimated
sodium excretion level and blood pressure, cardiovascular disease events, or
deaths differed between those individuals with and without hypertension
Simulation Modelling (for objective number 2)
- We modelled the estimated effect of changes in sodium intake
on risk of incident cardiovascular disease events, based on the observed
associations between sodium excretion and systolic blood pressure, and between
systolic blood pressure and cardiovascular events (appendix p 9).
- For this modelling, we focused on 98 612 participants (3733 cardiovascular disease events;
median 3.7
years of follow-up (IQR 2.9-5.0) without
baseline cardiovascular disease, because this subcohort is comprised of
generally healthy people from the population among whom few were receiving
drugs.
- We compared
these simulated blood pressure-based estimates with directly observed HR of sodium excretion versus clinical outcomes
to assess the consistency between estimates derived overall and in those
individuals with and without hypertension.
- Cox regression
was used to calculate HRs and 95% CIs of cardiovascular disease events (total cardiovascular disease, stroke, and
myocardial infarction) per 1 mm Hg
increment in systolic blood pressure, within each subgroup of hypertension
status.
RESULTS
Participants’ Characteristics
- 133 118 individuals, (63 559 with hypertension and 69 559
without hypertension), were included in the study.
- 98 612 (74%) individuals were without previous cardiovascular
disease, and 118 232 (89%) were without diabetes.
- Baseline characteristics of the study participants are
shown in the appendix (pp 9, 10).
- The mean age was 58.6 years (SD 10.3) in individuals with hypertension
and 50.5 (10.7)
in those without hypertension.
- Individuals with hypertension were more likely to be men,
heavier, less physically active, and had more previous cardiovascular disease
and diabetes (appendix pp 10, 11).
Mean estimated sodium excretion
- 4956 g/day (SD
1747) in individuals with hypertension
and
- 4823 g/day (1647)
in those without hypertension (p<0.0001).
Among those individuals with hypertension:
- 7006 (11%) had
an estimated sodium excretion of <3.0
g/day and
- 15 126 (24% of
those with hypertension, or 11% of the overall population) had an estimated sodium excretion of ≥6
g/day
- 7060 (11% of those with hypertension, or 5%
of the overall population) had an estimated sodium excretion of ≥7g/day
- 41 427 (65%)
had an estimated sodium excretion between
3 g/day and 6 g/day.
In those individuals without hypertension,
- 7547 (11%) had
an estimated sodium excretion of < 3.0
g/day and
- 14 098 (20%)
had an estimated sodium excretion of ≥ 6 g/day
- 6271 (9%) had an estimated sodium excretion of ≥ 7 g/day
- 47 914 (69%) had an estimated sodium excretion between 3 g/day and 6 g/day
- After adjustment
for regression dilution bias, 3039 (<3%) participants had a sodium excretion
of < 3 g/day and 21 240 (16%) had an estimated sodium excretion of ≥ 6 g/day
(11 146 [18%] of those individuals with hypertension and 10 094 [15%] of those without hypertension; p<0.0001).
(11 146 [18%] of those individuals with hypertension and 10 094 [15%] of those without hypertension; p<0.0001).
Outcome
- 133 118 (96%)
participants had completed follow-up, with a median follow-up of 4.2 years (IQR 3.0-5.0).
- The primary
composite outcome of all-cause death or a major cardiovascular disease
event occurred in 6835 individuals (11%) with hypertension and 3021 (4%) without hypertension (table).
- Those participants
with 4-5 g of sodium excretion had the lowest risk and this was used
as the reference category.
Association between sodium excretion and primary
composite outcome
- The association between sodium excretion and the primary composite
outcome varied significantly by hypertension
status (pheterogeneity=0.0342; figure 1).
- In the hypertension
group, a U-shaped association
between sodium excretion and cardiovascular events and mortality was apparent.
- Compared with sodium excretion of 4-5 g/day (reference
category), sodium excretion of ≥7
g/day (HR 1.23 [95% CI 1.11-1.37];
p<0.0001)
and <3 g/day (1.34 [1.23-1.47];
p<0.0001)
were both associated with increased risk
of the composite outcome (table; figure 1).
- After adjustment
for blood pressure, the associations
between high sodium excretion and the composite outcome (HR 1.21
[95% CI 1.09–1.34]; p=0.0006),
and the association between low sodium excretion and the composite outcome were
unaltered (1.35 [1.23-1.49]; p<0・0001).
- In those individuals without
hypertension, compared with 4-5 g/day, sodium excretion of ≥
7 g/day was not associated with risk of the primary composite outcome
(HR 0.90 [95% CI 0.76–1.08]; p=0.2547), whereas
an excretion of < 3 g/day was
associated with a significantly increased risk (1.26 [1.10-1.45]; p=0.0009;
table; figure 1).
- After adjustment
for blood pressure, the association between low sodium excretion and the
composite outcome remained significant
(p=0.0011).
- Similar results
were noted for death from any cause (pheterogeneity=0.0135)
and major cardiovascular disease (pheterogeneity=0.0432;
table).
- The results described above of a U-shaped association in
those participants with hypertension were consistent in those participants with
and without vascular disease (appendix p 12).
- Among those participants without hypertension, an
increased risk with sodium excretion of < 3g/day compared with 4-5 g/day was
consistent in those with and without vascular disease, whereas a sodium
excretion of ≥7
g/day was associated with an increased risk only in those with known vascular disease
(appendix p 12).
Sensitivity Analysis
- When we exclude data from the EPIDREAM study from the
analysis (which is a case-cohort study of 2850 individuals), the results of the
study overall and by subgroup do not change and the estimates from the PURE
study report the same findings (appendix 8). - Further, the data from
the ONTARGET and TRANSCEND trials are consistent with the data from the two
observational studies
- Exclusion of those participants who had an event in the first
2 years of follow-up did not affect the estimates (table).
- Further, in those with hypertension, exclusion of 35 027
individuals who were taking antihypertensive medication did not change the findings
(table).
- Sodium excretion
was more strongly associated with
increased systolic blood pressure in
individuals with hypertension (2.08 mm
Hg increment in systolic pressure per g [95% CI 1.96-2.21]) than in those without hypertension
(1.22
mm Hg [1.13-1.30];
p<0.0001 for interaction; figure 2).
- Similar results were noted for diastolic blood pressure (0.72
mm Hg increment in diastolic pressure per g [95% CI 0.65-0.80]
and 0.52
mm Hg [0.46-0.58], respectively; p<0.0001 for interaction; figure 2).
Simulation models
- In the simulation models, in which we assumed that the effect
of sodium intake on cardiovascular disease events was solely related to its
association through systolic blood pressure, the projected HR of cardiovascular
disease events, stroke, and myocardial infarction increased in a graded
fashion.
- However, there was a greater increase in risk in
individuals with hypertension, and a more modest association in those without
hypertension (p<0.0001 for heterogeneity; figure 3; appendix p 18).
- The modelled
estimates differed from the observed HR of cardiovascular disease events
both in individuals with hypertension and those without hypertension.
- This discordance
was marked at lower levels of sodium excretion (ie, <3 g/day).
- The projection
model shows lower HR estimates with lower sodium excretion, whereas the
observed HR estimates show an increased risk of events with lower sodium
excretion.
- In individuals with
hypertension, the observed HR was similar to the modelled HR at average or
higher levels of sodium excretion (>4 g/day; figure 3; appendix 10).
DISCUSSION
- Significant heterogeneity was noted in the association
between sodium excretion and the composite outcome by hypertension status.
- In both individuals with or without hypertension, there is
an increased risk of cardiovascular disease events and deaths associated with
24-h urinary sodium excretion of <3 g/day
- However, an increase in risk of cardiovascular disease
associated with high sodium excretion (surrogate for intake) was only seen in individuals
with hypertension (which represents 24%
of those with hypertension but only about 10% of the overall populations
enrolled in the four cohorts included in this analysis), but not in those
without hypertension
Comparison with other studies
- Our results are consistent with another recently published
cohort study (PREVEND study; n=7543), which reported an association between
increased sodium intake and cardiovascular disease, that was confined to participants
with baseline hypertension (pinteraction=0.08) and in those with
baseline pro-BNP (brain natriuretic peptide) concentrations above the median.
- Other studies have not reported a significant modifying
eff ect of previous hypertension, but these studies have been smaller than our
study
- In our analysis, the association between low sodium
intake (<3 g/day) and increased cardiovascular disease and mortality was
consistent, irrespective of baseline hypertension status and after further
adjustment for blood pressure level indicating that mechanisms unrelated to
blood pressure might be operational
- Our findings are also in keeping with a previous
meta-analysis of prospective cohort studies showing a U-shaped association
between sodium intake and cardiovascular disease events, in both healthy and high-risk
populations (eg, those individuals with cardiovascular disease or diabetes),
with consistency across different methods of sodium estimation (Graudal et al. Am
J Hypertens 2014; 27: 1129–37)
- Although the meta-analysis included previous analyses from
the ONTARGET/TRANSCEND cohort, it did not include the PURE study and EPIDREAM
cohorts, and the PURE study accounts for most of the current study population
- Our findings replicate previous reports and extend these
observations to populations based on baseline hypertension status. Further, they
suggest that although there is a limit below which sodium intake would be
unsafe, the harm associated with high sodium consumption seems to be confined
to those individuals with hypertension.
- Only about 10% of the population in our study had both
hypertension and high sodium consumption (6 g/day or more).
- This argues against a population-wide approach to reduce
sodium intake in most countries, except in those where the mean sodium intake
is high (eg, some in central Asia or some parts of China).
- We noted that most
of the world’s population (about 95%) studied consumes more than 3 g/day of
sodium, regardless of hypertension status and only 22% consume 6 g/day or more of sodium—the threshold above which we note an
increase in mortality and cardiovascular disease risk
Mechanistic function of sodium
- Sodium is an essential cation and is crucial to the action
potential of all cells in the body.
- Sodium homoeostasis is under tight physiological
regulation.
- Further, emerging evidence suggests that inflammatory responses
with infections involve mobilising high concentrations of sodium to the local
tissues that are involved, and this ability might be part of an essential defence
mechanism to external infections.
- Sodium intake is governed by neural mechanisms that
regulate intake of sodium and related homoeostatic mechanisms, and so although
extreme reductions in sodium intake are possible in controlled settings for
short periods, this is unlikely to be sustainable in free living individuals in
the long term.
- Previous modelling studies that have estimated the effect
of reducing sodium intake globally on cardiovascular mortality are based on the
assumption that the blood pressure-lowering effects of sodium reduction seen in
short-term trials will translate into reductions in cardiovascular disease in
the long term. (Mozaffarian et al. N Engl J Med 2014; 371: 624–34; the previous paper shared in evidence-sharingblogspot)
- However, it is now known that whether lowering blood
pressure results in reductions in cardiovascular disease is dependent on the
baseline blood pressure of the population, the mechanism of blood pressure
lowering, and presence or absence of cardiovascular disease.
- Although the SPRINT trials (also shared in evidence-sharing blogspot) did report a reduction in
heart failure and cardiovascular death when lowering blood pressure to a mean
of 121 mm Hg systolic blood pressure in a primary prevention population,
several other randomised controlled trials have failed to show a benefit of
lowering systolic blood pressure below 130 mm Hg in a primary prevention
population (HOPE-3) and secondary prevention populations (ACCORD, SPS3,
PRoFESS), and some have shown harm.
- Three independent meta-analyses of large randomised trials
of blood pressure lowering with antihypertensive drugs in individuals with diabetes
show that the benefits of blood pressure lowering in reducing clinical events
is noted only in those with a systolic blood pressure of higher than 140 mm Hg.
- This finding is also supported by the results of the
recent HOPE 3 trial, which showed that reducing systolic blood pressure by 6 mm
Hg reduced cardiovascular disease risk by about 25% only in those with
increased baseline levels (systolic blood pressure >143 mm Hg), but not in
those with lower initial systolic blood pressure, despite similar reductions in
blood pressure.
- These data are consistent
with our finding that the association of high sodium intake and cardiovascular disease is confined to those with
baseline blood pressure higher than 140/90 mm Hg.
- The mechanism of blood pressure lowering is also important
and non-blood pressure effects might be beneficial or harmful
- Although high-risk people—eg,
those with previous myocardial infarction or stroke—have
benefited from ACE inhibitors or β
blockers, the benefits seem to be not exclusively due to blood pressure
lowering, and other drugs that lower blood pressure in high risk people might
not necessarily reduce cardiovascular disease events in such populations
- Further, some drugs that were shown to reduce blood
pressure to similar extents differed in their effect on cardiovascular disease
or its individual cardiovascular disease outcomes
- Our data suggest that although a persuasive case can be
made to reduce sodium intake in individuals with hypertension and high sodium
intake, it is unclear whether the remaining more than 90% of the population
will benefit from dietary sodium reduction.
Limitations of simulations models & possible
mechanistic explanation
- Our analyses indicate the limitations of estimates from modelled calculations based only on
projected changes in blood pressure from sodium lowering.
- This is apparent in view that the results differ compared with the directly observed data
relating sodium to cardiovascular disease events and supported by an absence of
cardiovascular disease reduction with blood pressure lowering in people without
cardiovascular disease.
- This suggests that
the effect of a given level of sodium intake on clinical outcomes is only
partly mediated through its effects on blood pressure and that other mechanisms
might also be at play.
- This is supported by observations of activation of the
renin system and of catecholamines with low sodium intake.
- High renin concentrations have been reported in studies of
the Yanomamo Indians who reportedly consume very little sodium.
- Several studies have shown that increases of renin,
aldosterone, and catecholamines are all associated with increased
cardiovascular disease events and mortality.
- Therefore, predicting the net clinical effect based on
only considering the effects of sodium on blood pressure might not provide a
comprehensive understanding of its effects on cardiovascular disease and
mortality, especially within the range of sodium intake that affects the renin
system (<4 g/day).
- We noted that the association of sodium intake with cardiovascular
disease remained strong even when adjusted for blood pressure levels.
- This indicates that the
association between sodium and cardiovascular disease might also be related to
non-blood pressure mechanisms, which requires mechanistic investigation in careful
physiological studies
- Randomised trials have shown that sodium lowering has only
a small effect on blood pressure in individuals without hypertension and that
such individuals might be less sensitive to the blood pressure effects of salt
consumption.
- Furthermore, understanding of why low sodium intake is
associated with increased event rates, despite slightly lower blood pressure,
is also of importance.
- As sodium is an essential cation, it should not be
surprising that there is an optimal range for its intake.
- This mirrors the situation of most biological systems and
it is only with external toxins (eg, tobacco or environmental pollutants) that
a linear association is likely
Limitations of study design and Recommendations
- Despite careful design, follow-up, and analyses, observational analyses cannot definitively
prove causality
- Therefore, ideally large and long-term randomised controlled
trials (RCTs) of sodium reduction to various levels to assess the effect on
clinical outcomes are essential to guide public policy
- In view of the absence of such RCTs, large prospective
observational studies (despite their inherent limitations) relating sodium
intake to cardiovascular disease should be considered the best available
evidence.
- Further, we have
initiated a pilot RCT to assess feasibility as a prelude to establishing a
larger and long-term study to definitively address this question (NCT02458248).
- In the absence of large definitive RCTs showing a clear
reduction in cardiovascular disease, the weight of the substantial
epidemiological studies describing a potentially adverse effect of low sodium should
urge caution in making broad public health recommendations.
- Further, the observation that high sodium intake is only
associated with increased cardiovascular disease in individuals with
hypertension raises questions whether public health policies targeted at reducing
sodium in the entire population are appropriate
- Therefore, until
new robust data emerge from large trials, it might be prudent to recommend
reduction in sodium intake only in those with high sodium intake and with
hypertension.
-Some might consider large randomised trials of sodium
reduction impractical to assess their effect on cardio vascular disease, but
they are essential to definitively resolve the controversy
Strength
- Large size, international cohorts, use of validated urinary
measure of sodium intake, standardised methods to measure a large number of
covariates, and careful and standardised measurement of blood pressure à
valid and generalizable
- Our analyses include participants with established
cardiovascular disease recruited into an RCT (ONTARGET/TRANSCEND) as well as
those without vascular disease identified from the population (PURE) or those
screened for a trial (EPIDREAM). This broad range of individuals from 49 countries
indicates that our findings are
widely applicable and robust because
similar findings were noted across all four studies.
Limitations of the study
- Although the collection
of one overnight urine sample to estimate the 24-h urinary sodium excretion
might be considered a limitation, it has been validated against 24-h urine
collections in previous studies of healthy individuals and those with hypertension
and in our international validation study, with correlations similar to that
noted with a blood pressure measured at a clinic visit versus 24-h ambulatory
monitoring.
- Further, our analyses take into account the day to day
variability of sodium intake in individuals by estimating the correlation of
two measures taken 30-90 days apart and then using statistical adjustments to
assess the degree of regression dilution.
- Adjustments for day to day variability and the absence of
perfect correlation with 24-h urinary estimates of sodium would steepen all the
associations (both at the low and high ends of sodium intake) and so would not
qualitatively affect the pattern of our results
- With our method, there is about a 10% overestimation of 24-h
sodium excretion, indicating that the
true intake range at which risk of cardiovascular events and death changes
might occur at a slightly lower level of sodium intake.
- Residual confounding cannot be completely ruled out in any
epidemiological study but extensive multivariable analyses did not change our
results.
- Further sensitivity analyses to minimise the potential for
reverse causality (in which sicker people reduce sodium intake) by excluding in
turn those with known cardiovascular disease, hypertension, or diabetes or by confining
analyses to events beyond 2 years, did not change the pattern of our findings.
- Therefore our results are robust to different forms of analyses.
CONCLUSION
- Association between low sodium intake (vs moderate
intake) and increased risk of clinical outcomes in those individuals with and without
hypertension, whereas high sodium intake (> 6 g/day) was associated with an
increased risk in individuals with hypertension.
- Our findings suggest that sodium reduction should be confined
to only those individuals with hypertension and high sodium intake; this
represents only about 10% of the population studied.
