■ site map
BLOG: April 2010
Salt studies: the latest score
Why is it that after thousands of studies on the subject of salt and hypertension we don't seem to be anywhere close to consensus on whether the present salt consumption makes people sick, or not?
Part of the answer is in the fact that all studies have limitations as to how directly their results can apply to the general population; another is in the fact that most of these studies used low-quality methodologies - particularly the older ones, and yet another is in frequent authors' bias in designing and interpretation, nearly as old as the controversy itself.
For several decades, the primary source of research data were experiments on animals (mainly rodents) and small-scale human trials involving hypertensive individuals with impaired renal function. Turning to a broader human population, the epidemiologic studies (i.e. those trying to establish rate of incidence of a disease in population) were gradually widening their scope from looking at hypertension rates solely in relation to salt intake, to the inclusion of other relevant factors.
These studies were either ecologic, analyzing data for any number of different populations, or intrapopulation studies, focused on a group of people within certain population. Most of them were observational studies, limited to collecting and analyzing data. Either type has its objective limitations; ecologic studies the difficulty of filtering out the effect of many variables related to different cultures, lifestyles and genotypes, while most intrapopulation studies had relatively small number of participants and short duration.
Given size and duration, intrapopulation study is more reliable than ecologic study, since the results are more directly applicable to that particular population.
Despite decades long commotion about salt danger, no large controlled intervention trial examining the effects of salt intake on health has taken place. The recipe for such study is simple: a three large groups of participants - normotensive, elevated and high blood pressure of otherwise similar characteristics - each are split into intervention (i.e. w/reduced salt intake) and control (regular salt intake) groups. It needs to have long enough duration - at least five years - since some health effects can show after a longer period of time.
A comparable in size to such study is Intersalt, an ecologic observational study funded by NHLBI, that was completed in 1988. The fact that similarly sized - or larger - controlled intervention trial hasn't taken place in more than two decades since, nor is it on NHLBI's agenda in the foreseeable future, suggests that
those in charge of the government's agency do not want it.
Chances are, they learned their lesson from the Intersalt: the facts are not likely to be friendly to their salt health policy, which they are stuck with.
Attempting to partly fill in for the absence of a large controlled clinical trial, an analysis of a selected number of studies - so called meta study - had been used more frequently in recent years. Unfortunately, with the existing heterogeneity of data from one study to another, it's been proven more than once that desired result with meta studies can be easily tailored with the "appropriate" study selection criteria, methodology applied and/or result interpretation.
In addition, the very data base for meta studies is unreliable; that is, most trials that they can use have low-quality methodologies and/or some form of bias (Ebrahim and Smith 1998, Swales 2000).
After this basic study primer, let's turn to the actual studies and their results.
In the absence of substantial evidence that the present-day salt consumption significantly increases rate of hypertension and related cardiovascular diseases, what emerges as a piece of information that can "make or brake" salt hypothesis is whether or not salt intake can be directly linked to a higher morbidity/mortality from these diseases. And that leads us to the core question:
what is specific evidence that salt consumption at present levels significantly increases the chances of cardiovascular disease and death,
both for individuals, and at the level of population?
Now comes the interesting part.
For no other reason than having a list of significant studies following Intersalt compiled at Salt Institute's Web page, we'll start from there. The main difference versus older studies, as well as Intersalt itself, is that these studies now look not only at the salt/hypertension link, but also at the
actual health outcome,
i.e. the rate of morbidity and/or mortality that should be affected by salt intake if it significantly changes blood pressure.
Here's a brief overview of these "health outcome" studies, focused on actual morbidity and mortality from cardiovascular (CV) causes that can be associated with salt intake (as opposed to studies merely looking to establish whether or not salt raises blood pressure) on the Salt Institute's Web site.
Four out of seventeen studies cited had health outcome better for those consuming more salt, nine of them found no health benefit from salt intake reduction, and only four showed benefit, at least for some sub-groups. In all,
nothing in this big picture supports the notion
The majority of studies are neutral, with the rest of them at least roughly balanced around this point.
But that, of course, is not the final word.
The most recent study that should be added to this list is a meta study of 13 selected trials focused on salt-intake-related health outcomes (among them are 10 studies from the above list: 1, 2, 3, 4, 5, 7, 8, 11, 16 and 17). Its conclusions are:
"High salt intake is associated with significantly increased risk of stroke and total cardiovascular disease."
(noting that these adverse effects are likely underestimated due to inaccuracies in measurements of salt intake), and
"These results support the role of a substantial population reduction in salt intake for the prevention of cardiovascular disease." (Salt intake, stroke, and cardiovascular disease: metaanalysis of prospective studies, Strazzullo et al, 2009).
This is a bit odd situation, having two diametrically opposite conclusions drawn from seemingly similar database. How did Strazzullo et al. come to the conclusion that the present salt intake is a serious health hazard for the general population?
The answer is found along these lines:
1 - study selection, and
2 - interpretation of the results
Looking at the study selection, it is immediately noticeable that it included all four studies from above table that found salt intake reduction beneficial, while only two of those that have found it detrimental, as well as four out of nine that were neutral.
In addition, all three remaining studies added to their selection (He et al. 1992, Larsson et al. 2008, and Umesawa et al. 2008) show higher salt intake having negative effect.
So there we have it: having seven studies showing higher salt intake detrimental and only two showing it beneficial, with the former outnumbering even beneficial and neutral-result studies combined. This certainly
doesn't reflect the real world proportion.
Not even close. Already at this stage, any result other than "confirmation" of the current salt intake being unhealthy was very unlikely.
After statistically digesting the data, Strazzullo et al. ended up with the following results: for the averaged daily intake differential of 86mmol, or 2 grams of sodium (equivalent of 5g of salt), the higher-intake population showed 23% and 17% higher risk for suffering stroke or general cardiovascular disease, respectively (with the respective 95% confidence intervals 1.06-1.43 and 1.02-1.32, which means that there is 1 in 20 probability - p=0.05 - that the actual result is not in this range).
These numbers, if correct, would implicate that elevated salt intake significantly increases risk for developing these diseases. On the level of populations, that would affect many individuals. If applicable to the U.S. death rates, it would imply over 100,000 deaths a year (some 25,000 from stroke and 90,000 from heart diseases, respectively) attributable to the average U.S. salt consumption of about 9 grams a day, as opposed to consuming 5g less.
But these numbers are
not applicable to the U.S. general population,
nor general population of most western countries. The reason? Three out of eight trials that showed elevated risk for the highest vs. lowest level of salt intake - and the three creating most of the pull toward increased risk - are for Asian populations with significantly higher salt intake.
For instance, in Hu et al. that top salt-consuming group of Taiwanese were those declaring to eat "very salty foods". Knowing that the average salt consumption in Taiwan is similar to that in the U.S. (Intersalt data indicates 8.1g/day, vs. 7.8g/day average for four U.S. locations), that could be double the average U.S. intake, or even more.
As a side note, this study has unreliable, sub-standard design and methodology, which is the reason why it is rarely cited, let alone used for database.
In Nagata et al, average sodium intake in the top salt consuming group was about double that in the U.S. population (7.2 and 6.5g a day for men and women, respectively; corresponds to 18 and 16.2g of salt). And in Umesawa et al, the estimated highest intake group average was 6.3g of sodium, or 15.8g of salt - nearly double the U.S. average.
If the basic assumption of Strazzullo et al. - and the anti-salt doctrine itself - that these adverse health effects are positively and significantly correlated with salt intake, is correct, then lowering salt intake by a similar amount at twice lower intake level, applicable to the Western countries,
would be significantly less beneficial.
What about the rest of five trials that had high salt intake associated with increased mortality? Well, four of them found it only for generally higher-risk subpopulations: overweight (He et al, Geleijnse et al.), obese (Tuomilheto et al.), and male smokers (Larsson et al.). Needless to say, we can only speculate how much of it is due to the detrimental effect of sodium alone - if any.
In the fifth study, Cook et al, the intervention group had been given "comprehensive education and counseling", which more likely than not has resulted in a healthier overall lifestyle. The latter may have more to do with their better numbers than the moderate reduction in sodium intake (less than 1g of sodium, or 2-2.5g of daily salt intake reduction).
And in Alderman et al, which was split between significantly lower incidence of stroke and cardiovascular disease in men on highest salt intake, and significantly higher incidence in the highest-intake women (both with mild to moderate hypertension), the latter had too low number of events (6 and 21, respectively) to be statistically reliable.
not one of these trials is directly applicable
They had either significantly higher sodium intake, different diets and lifestyle, insufficient sample size, or with some other confounding factor - or factors - present.
Consequently, the results obtained based on them cannot be used as the evidence that sodium intake reduction would be beneficial to the general U.S. population.
But there's more to it.
After the BMJ (British Medical Journal) published Strazzullo et al, Drs. Cohen and Alderman reported data errors in the study (BMJ Rapid Responses). One of them was in presenting He et al. results for the incidence of cardiovascular disease as applying to the entire population, while in fact they were for overweight men and women only, as well as omitting the fact that the study showed no association for the non-overweight majority.
The second error was in presenting negative association (i.e. benefit from higher salt intake) in Cohen et al. 2006 as statistically insignificant, while it was, in fact, statistically significant.
Strazzullo et al. lead author responded by stating that He et al. does not specify whether or not the positive association is limited to the overweight (odd statement, to say the least, with the trial results being published and quoted in their own reference section as:
"Dietary sodium intake and
subsequent risk of cardiovascular disease
and with the same trial being properly presented in the Strazzullo et al. stroke-alone analysis). With "let's see what would have happened" tone, Strazzullo presented recalculated data, which caused the pooled risk ratio for cardiovascular disease (CVD) for high-sodium-intake population to drop from 1.17 to 1.11 (17% and 11% higher risk, respectively, if we trust their calculation).
They admitted errors in presenting Cohen et al. 2006, but dimmed it unimportant, lowering the overall CVD risk ratio only from 1.17 to 1.16 Combined with the above, it indicates the overall CVD risk ratio in the study of 1.10, not much better than one half of the published 1.17 value.
Yet Strazzullo maintains that this does not change appreciably study's conclusions.
Cohen and Alderman also objected to the Strazzullo and al. methodology not addressing substantial differences in salt intake level from one trial to another. Strazzullo et al. response does not admit any methodology flaws, but does not address the specific objection.
Does the reluctance of study authors to face and discuss these objections have something to do with the fact that both senior authors, Dr. Strazzullo and Dr. Cappuccio
are members of WASH
(World Action on Salt and Health), a global organization for promoting the salt-hypertension link and, accordingly, advocating global reduction in salt intake?
This, by the way, was not disclosed in their study, which goes under "no conflict of interest" label.
One could tell that Strazzullo and al. does not hold water very well under closer scrutiny. Despite heavily biased study selection, they still couldn't manage to have more than marginally significant risk increase for cardiovascular disease, and statistically low level of significance in the risk increase for stroke.
If applied to the U.S. rates, the lifetime risk for suffering stroke due to high salt consumption would increase from about 3.2% to 3.9%. However, considering significantly higher salt intake levels in the study, and its selection bias, the risk increase would likely be significantly less - if any at all.
But how factual the study is may have been of secondary importance to their authors. Its results were widely publicized in support of ongoing or planned salt intake reduction programs in several countries - U.S. included - spurring headlines like:
▪ Meta-Analysis Reaffirms Salt's Link with Cardiovascular Disease
▪ High salt intake boosts stroke, CVD risk
▪ Direct Link Between High Salt Intake And Stroke And Cardiovascular ...
▪ Study prompts calls for Europe-wide salt legislation
and alike. Considering study's not hard to find shortcomings, its primary purpose may as well have been to help recover some lost grounds in the public relations war fought around recent universal salt intake reduction proposals.
The above study only adds to the pattern of bias and/or purposeful distortion of facts and evidence already noted in the three studies described earlier: Crystal-Spangler et al, Dahl's 1960 study and Intersalt. Another noted study supporting salt hypothesis, DASH (Effects on blood pressure of reduced dietary sodium and the dietary approaches to stop hypertension, Sacks et al. 2001), which hasn't been mentioned here because it is not a health outcome study, is not an exception.
The study concluded that both low-sodium intake and healthful Mediterranean diet, substantially lower blood pressure, and particularly when combined. Direct implications were that it is a viable public policy alternative for curbing hypertension and CVD.
However, in addition to the study being too small (412 participants total) and of short duration (30 days) for such implication, study's flaws quickly surfaced:
(1) selection bias (56% woman and 57% black, both statistically more likely to react to changes in salt intake, more so with the average BMI of 29.5 bordering obesity,
(2) ignoring unfavorable physiological effects, namely, over three times higher renin and aldosteron levels (each) - the consequence of lowering sodium intake by 2/3, down to 50 mmol/day - despite the evidence of higher aldosterone being positively associated with increased CVD risk,
(3) the decrease in systolic blood pressure was clinically significant only in black women with hypertension, thus even less applicable to the entire population,
(4) exclusion of the effect of salt intake on diastolic blood pressure, which is more significant with respect to CVD risk than systolic pressure (study authors not only excluded diastolic data from analysis, they declined requirement to present this data), and so on...
That, of course, didn't move the authors to change anything in their initial presentation of study results. They only agreed to the comment that longer term adherence to such a low sodium intake level would be difficult.
And this pattern of biased, unscientific "science" is not a coincidence. While this may not be representative of all research supporting salt hypothesis, the main purpose of quite a few of these studies seems to be
hypothesis at any cost,
not establishing scientific facts. It is a direct consequence of the body of evidence being generally unfavorable to salt hypothesis.
No single article sums it up better than Gary Taubes' award-winning "The (political) science of salt".
It is not that the other side, opposing such general "guilty" verdict for the common dietary salt intake levels is always objective. But reading articles of its most prominent advocates - for instance, the Salt Institute - does not reveal such selective use of facts coupled with "creative imagination". They don't oppose that excess intake of salt can be unhealthy for vulnerable sub-populations and individuals; only that it applies to the current intake level in Western countries, and the entire populations. That view seems to be better supported by the evidence, including that coming from studies sponsored by their greatest contender: NHLBI.
Researchers that found no significant link between salt, hypertension and cardiovascular outcomes are much less likely to be categorically positive in interpreting such result as proving salt hypothesis wrong, than those backing salt hypothesis to interpret just about anything they can manage to as addition to already "ample evidence" proving the salt-hypertension link.
The latter, of course, requires
ignoring or dismissing contradictory evidence.
A good example of this strategy is Scottish Heart Health Study, by far the largest intrapopulation trial on the effect of salt intake on blood pressure. At the end of 4-year study in 1988, as well as at the 10-year follow up in 1997, salt intake, after adjusting for potassium intake, wasn't among significant risk factors for cardiovascular disease.
The response from salt hypothesis camp? After offering speculative reasons why the study "failed" to detect "correct" outcome (such as that it was still "too small" not to be distorted by measurement errors), it was simply dismissed.
And so were scores of others that didn't fit, while some that did were embraced, no matter how flawed they evidently were (Intersalt Revisited, Law et al. 1991, and others).
With the scientific input about the link between salt, hypertension and health in general often having little to do with science, what is it that can be considered to be an actual data-set about salt metabolism and possible physiological effects of salt intake on blood pressure?
More on that in the following article.