The Need for Balance in Evaluating the Evidence on Na and CVD Salim Yusuf Professor of Medicine, McMaster University Executive Director, Population Health Research Institute Vice-President Research, Hamilton Health Sciences President-Elect ( 2013 onwards) World Heart Federation I have no relationships that will influence my judgments or the conclusions of my work, no association with any food manufacturer or those committed to salt polices.
BMJ July 30, 1988: A tale of two study reports Urinary electrolyte excretion, alcohol consumption, and BP on the Scottish Heart Health Study Smith WC, Crombie IK, Tavendale RT, Gulland SK and Turnstall-Pedoe HD. Number of citations = 83 Intersalt: an international study of electrolyte excretion and BP: results for 24 hour urinary Na and K excretion. Number of citations = 622
Scottish Heart Study 24 hour urine in 7354 random individuals (40-59 yrs) from 22 districts in Scotland. Several significant correlations with both SBP and DBP but only the co-efficients for age, BMI and pulse rate were >0.1 -alcohol and sodium excretion showed a weak positive correlation with BP; and K+ excretion a weak inverse correlation with BP In multiple regression, age, pulse rate, BMI, alcohol and K+ excretion (but not sodium excretion) correlated with BP. The most likely explanation is that the relation between Na and BP in the population is weak, and that K+, alcohol (and weight) are of greater importance.
INTERSALT: BMJ 1988 10,079 individuals 20 to 59 from 52 centers and many countries -4 centres (e.g. Yanomamo Indians*) with very low Na showed low BP and no upward slope of BP with age. Across the remaining 48 centers : -Na was NOT significantly related to BP (prespecified analysis), but to an increase in BP (slope) with age (data-derived emphasis) BMI, alcohol intake and K+ excretion were correlated with BP
INTERSALT(BMJ 1988): Median Na vs BP
INTERSALT(BMJ 1988): Slope of age related BP change vs Na
INTERSALT Revisited: BMJ 1996 Estimated diff in BP (mmhg) for 100 mmol higher 24 hr urinary sodium SBP DBP With BMI Without BMI With BMI Without BMI Not corrected 1.0 (0.3) 2.1 (0.3) 0.04 (0.19) 0.9 (0.2) Corrected for RDB 3.1 (0.9) 6.0 (1.1) 0.1 (0.6) 2.5 (0.7 Elliott BMJ 1996
Impact of different assumptions in modelling of 100 mmols/day % reduction in mortality Diff in SBP Stroke CHD Total 1 mmhg (1) -3-2 -1.5 2 mmhg (2) -6-4 -3.0 3 mmhg (3) -8-5 -4.0 6 mmhg (4) -18-12 -9.0 Diff in DBP.04 mmhg (1) None.09 mmhg (2) Negligible Negligible.10 mmhg (3) 2.5 mmhg (4) -12-7 -6.0 (1) Not corrected + BMI adjusted (2) Not corrected + BMI not adjusted (3) Corrected + BMI adjusted (4) Corrected + BMI not adjusted
Characteristics of the Yanomamo Indians vs US participants Yanomamo Indians# US (Chicago) SBP 95.4 115.1 DBP 61.4 70.0 U. Na (mmol/24 hrs) 0.2 133.8 U. K (mmol/24 hrs) 57.9 52.4 BMI 21.3 25.3 % Drinkers 0.0 70.0 U. Cr (mmol/24 hrs) 4.0* 12.1 *Indicates substantial under-collection of urine # Life expectancy <40 yrs.
Summary re: Sodium and BP There is a modest association between Na intake and BP, which is weaker (BMI) than or similar to other lifestyle factors (Potassium consumption or alcohol) The estimates of differences in BP have a 5 fold uncertainity( but most likely that 100 mmol=2.5 mmhg). A balanced approach to moderate changes in weight (e.g. 2.0 kg), alcohol consumption, activity and increase in fruits and vegetables, and animal products and Na consumption will lead to a larger BP & CVD reductions vs focussing on large reductions in Na. Such an approach is likely more feasible.
Meta-analysis of salt reduction on BP (He and McGregor) 31 trials on 3022 subjects included. 112 trials excluded (<4 wks, concomitant intervention, children/preg women, U Na not measured, reduction in U Na of <40 mmol/24 hrs) Not clear about losses to follow-up Some suggestion of publication bias (asymmetry in funnel plots ).
Meta-analysis of salt reduction on BP (He & McGregor) Hypertensives: Duration of 5 weeks, Median BP 149/94 mmhg 24 hr U Na : 165 mmol (9.5 g/day of salt), Reduced to 87 mmol (5.1 g/day of salt) SBP 5.06 mmhg (CI: -5.81 to -4.31) DBP -2.70 mmhg (CI: -3.16 to -2.24) Non-hypertensives: Duration of 4 weeks, Median BP 127/78 mmhg 24 hr U Na : 154 mmol (9.1g/day of salt), Reduced to 82 mmol (4.8 g/day) SBP -2.03 mmhg (-2.56 to -1.50) DBP -0.99 mmhg (-1.40 to - 0.57) N.B. In the long-term TOHP, Na (-44,33 mmols)& BP reductions (-1.7,1.2/.8,.7) were much smaller( Cook et al,bmj 2007) NOTE MANY STUDIES WITH LARGE EFECTS ARE FEEDING STUDIES OF VERY SHORT DURATION.
Variable validity of surrogate (physiologic) outcomes in CVD Reliable surrogates: Smoking, LDL-cholesterol Variable reliability: BP (NB several recent trials show little or no benefit from lowering BP (-140 mmhg) Consistently disappointing: HDL raising, suppression of ventricular arrhythmias, vascular reactivity, glucose lowering,??wt loss NET CLINICAL BENEFIT OF AN INTERVENTION IS A BALANCE OF MANY EFFECTS, AND CAN NOT BE PREDICTED FROM SURROGATE OUTCOMES.
Observational data related Na Intake to CVD Few large studies with >1000 events. Multiple 24 hr urines not practical; so spot (single/multiple) collections can estimate group means of Na intake. Dietary methods have greater variability, difficult to use in multiethnic populations due to variable sources of Na and differing patterns of diet. Measures of Na intake needs to positively associated with BP (not observed in several studies using diet). Confounding (need to adjust for BMI, alcohol, phys activity and other dietary components that affect BP or risk not collected in most studies) Reverse causality (can test/adjust for this, but impossible to completely exclude)
Value of Observational Studies Large prospective studies with careful assessment of exposures/confounders and a large number of outcomes (minimum 1000, ideally 5000 to 10,000+) can provide useful insights that can inform design of trials and policy. -if several studies show a similar pattern of information then the inference is more robust.
Estimated 24-Hour Urinary Excretion of Sodium and Composite of Cardiovascular Death, Stroke, Myocardial Infarction, and Hospitalization for Congestive Heart Failure :O Donnell et al JAMA 2011
PHRI s ongoing programme into Na (& other lifestyle factors) and BP PURE-Sodium: 1600 healthy Canadians from 4 sites with multiple 24 hr & spot urine + BP + Diet ( 2014) INTERSTROKE: 12,500 people with stroke & 12,500 controls ( 2013) Main PURE study: >135,000 people (out of 154,000 people) with second AM urine (+ regional comparisons to 24 hr urine) and CVD (3 yrs 10 yrs)( 2013 to 2020).
Very little reliable data from RCTs that Na reduction will reduce CVD.
TOHP Long Term Follow-up (5 to 10 years post trial) for CVD events Na reduction (n=327) TOHP I (n=744) Died 6 (2%) 12 (3%) Usual care (n=417) Responded 225 (69%) 299 (72%) No address, unwilling, no response 96 (30%) 106 (25%) Na reduction (n=1191) TOHP II (n=2382) Died 16 (1.5%) 24 (2%) Usual care (n=1191) Responded 922 (77%) 911 (77%) No address, unwilling, no response 124 (22%) 129 (21%) About 25% of randomized participants not followed Of 297 events, records only obtained in 196( ie 66%) Cook et al BMJ 2007
TOHP FU: Results on CVD & Total Mortality CVD (among responders i.e. in <75% of randomized) Overall 88/1169 (7.5%) Total mortality (in nearly all) Overall 35/1518 (2.3%) 112/1246 (9.0%) 42/1608 (2.6%) P (CMH) Adjusted HR and (p) 0.21 0.75 (p=0.044) 0.64 0.81 (p=0.35)
Projections of events prevented based on high estimates of BP decrease may be substantially inflated if the underlying assumptions are exaggerated. Moreover,these projections do not take into consideration the inconsistent or J shaped association of Na intake vs CVD events
Comparisons for SBP reductions used by Bibbins- Domingo compared to INTERSALT SBP reduction (mmhg) Salt reduction of 1 gm/day Salt reduction of 3 gm/day Low estimate High estimate Low estimate High estimate B-Domingo 0.60 1.17 1.80 3.51 INTERSALT Uncorr + adj 0.18 0.5 Corr+ adj 0.50 1.5 1 gm salt = 400 mg Na = 16 mmol Na (i.e. 1/3 SD change in Na) 3 gm salt = 1200 mg Na = 50 mmol Na (i.e. 1 SD change in Na)
Estimated annual number of heart attacks (in 1000 s) prevented from population wide Na reduction (based entirely on SBP reduction & assuming linear relationship) Hi estimate* Lo estimate* Intersalt Sodium reduction (a) by 400 mg/day Sodium reduction by (a) 1200 mg/day -32000-20000 -7000-92000 -58000-21000 Smoking cessation (b) -92000 Statin(1o Prev) (c) -17000 Drugs for HTN (c) -100000 (a) Observational data not robust; RCTs unreliable (b) Robust epidemiologic data (c) Robust RCT data *Estimates from Bibbins-Domingo; NEJM 2010 From Elliott, BMJ Intersalt, 1996: using SBP and NOT DBP.
Conclusions 1. There is a moderate association between Na intake and BP. A large difference in Na intake of 100 mmol difference (or 2300 mg) equates a 2 to 3 mmhg difference in SBP( projects to a 4% to 6% reduction in CVD (assuming no non-bp related adverse effects).a more feasible target is modest Na reduction ( eg 35-40 mmol ) which will translate into a 1.5 to 2% RR in CVD. 2. The totality of the observational data do not suggest a lower rate of CVD with Na consumption below 3.5 or 4 gms/day 3. Current RCT data are inconclusive as to whether lowering Na reduces CVD because they were not designed for this purpose. 4. Public health policy should be evidence based and would suggest that the degree of Na reduction that is feasible( 35-40 mmol/d) will lead to at best a small reduction in CVD in N America.
Implications 1. Need much better data on Na and CVD: - observational epidemiology with studies of >100,000 people with reliable measures of Na intake and >1000 to 10,000 events -RCTs to demonstrate unbiasedly and reliably that lowering Na is safe and effective in reducing CVD. 2. In the meantime, public health policy should be cautious and balanced: -encourage moderate reductions in Na ( to 3.5 or 4.0 gms/day), physical activity, wt loss, alcohol, fruits & veg, animal fats -encourage other lifestyle changes : smoking cessation -targetted use of low cost drugs in combination to reduce BP and lipids Such measures will reduce CVD by >2/3 to 3/4 in the next two decades, and will be far more effective, efficient and feasible than a policy of extreme Na reduction.
Sit down before fact as a little child, be prepared to give up every preconceived notion, follow humbly wherever and to whatever abyss nature leads, or you shall learn nothing Thomas Huxley