In the March 28th, 2013 issue of the NEJM, a review of sorts
entitled "Salt in Health and Disease - A Delicate Balance" by Kotchen et al can be
found. My interest in this topic stems
from my interest in the question of association versus causation, my personal predilection for salt, my observation that I lose a good deal of sodium in outdoor activities
in the American Southwest, and my concern for bias in the generation of and especially
the implementation of evidence in medicine as public policy.
This is an important topic, especially because sweeping
policy changes regarding the sodium content of food are proposed, but it is a
nettlesome topic to study, rife with hobgoblins. First we need
a well-defined research question: does reduction
in dietary sodium intake: a.) reduce
blood pressure in hypertensive people? in
all people? b.) does this reduction in
hypertension lead to improved outcomes (hypertension is in some ways a
surrogate marker)? In a utopian world,
we would randomize thousands of participants to diets low in sodium and "normal"
in sodium, we would measure sodium intake carefully, and we would follow the
participants for changes in blood pressure and clinical outcomes for a
protracted period. But alas, this has
not been done, and it will not likely be done because of cost and logistics,
among other obstacles (including ideology).
Even if we could conduct such a study, it may not be
definitive because some of our unavoidable assumptions would be unfounded. In almost all of the studies referenced in
the NEJM review, sodium intake was inferred from sodium excretion in
the urine, and this is a potentially highly flawed assumption. Here's a thought experiment. Suppose that a group of people enrolled in
the utopian study are from Salt Lake City, Utah and they are avid bikers and
backcountry skiers, and that they consume a good deal of salt, say 7 grams per
day of sodium chloride. Suppose also
that they sweat out, during vigorous exercise, 3 grams per day. Their urinary sodium excretion will show them
to be low sodium consumers, but they are not.
The surrogate measure of sodium consumption has misclassified them. Moreover, guess what their outcomes will
be? Presumably they will be good, but because
of exercise and all the associated lifestyle choices - not because of low
sodium intake. Sodium intake is actually
high! But the conclusion will be that the (mistakenly) low sodium intake was associated with the good outcomes (which were actually due to exercise.) So, in my opinion, all of these
studies are fundamentally and fatally flawed because of inability to measure
sweat sodium losses, which means that vigorous exercise confounds the proper
classification of sodium intake and at the same time influences outcomes of
interest. Moreover, almost all (all?)
of the cited studies used sodium excretion at a single point or at intervals as
the surrogate for sodium intake. This
carries the obvious assumption that the surrogate was stable over time.
With that in mind, I'll guide us through the review to see
if other premises, taken for granted, are vulnerable to challenge.
The authors state (page 1229, paragraph 4) that
"Studies across populations provide more convincing evidence than
within-population studies of the association of salt intake with both blood
pressure..." Why might that
be? Several obvious explanations. First, ecological fallacy. The people consuming salt and driving up the
population mean may not be the people having the events. But there's a more compelling (and equally
obvious) explanation - differences between populations have a good deal more to
do with differences between populations in factors other than salt intake,
which are not or cannot be measured. The
jewel of the references in this section of the review is reference 7, the Intersalt study, a herculean effort by investigators more than a quarter a
century ago to catalog differences in sodium excretion and blood pressure in
participants in 33 countries and 52 centers across the world. The Intersalt report is rife with data, and
one could spend hours poring over it (I did).
Suffice it to say that it is an observational dataset, and no matter how
intriguing its findings, how prolific the data collected, it cannot answer the
fundamental question of causation versus association.
In paragraph 5 which continues on page 1230, the authors
emphasize several meta-analyses of studies of dietary sodium reduction on blood
pressure measurements. The most notable
among these, the Midgley report in JAMA, 1996, (reference 10) found: evidence of publication bias; that the
regression analyses crossed the y-intercept suggesting that blood pressures
declined after enrollment irrespective of sodium reduction; that the effect of
sodium reduction on blood pressure was modest;
that trials generally were poor at reporting other variables known to
affect blood pressure (with potential for confounding); that the effect of sodium reduction on blood
pressure was (much) greater in hypertensive participants; that trials involving
"free range" participants (those not institutionalized or whose meals
were prepared for them) showed no effect of sodium reduction on BP. The authors concluded in the last sentence of
their report:
This analysis, in part because of the large number of normotensive trials now published, does not support one of the goals of the Nutrition Labeling and Education Act (1990), that of lowering blood pressure in the normotensive population at large, and questions the wisdom of universal dietary sodium restriction without better evidence of the long-term benefits and safety of such an intervention.
A similar conclusion was reached by the authors of reference
12 (Graudal et al, JAMA, 1998):
"These results do not support a general recommendation to reduce sodium intake. Reduced sodium intake may be used as a supplementary treatment in hypertension. Further long-term studies of the effects of high reduction of sodium intake on blood pressure and metabolic variables may clarify the disagreements as to the role of reduced sodium intake, but ideally trials with hard end points such as morbidity and survival should end the controversy."
It seems odd that these two studies should figure so
prominently as part of the basis of the authors' arguments (and in Table 1). They state (paragraph 5) that "clinical
trials provide definitive evidence of a direct cause-and-effect relationship
between salt consumption and blood pressure." Not so, says Dr. Scott. A utopian clinical trial would provide definitive
evidence, but these trials are not utopian and thus not definitive. (They don't control all the variables [diet,
exercise], they don't account for all the variables, their surrogate for sodium
intake is flawed, etc.) If they were, we would not be combining them
into a meta-analysis, and we would not have the authors of the meta-analyses
hedging their conclusions as they do.
The use of references 10 and 12 to support this statement is duplicitous.
A subsequent premise described in the review, namely that
sodium reduction on the population level will reduce clinical disease on the
population level, is quite curiously attempted to be supported by a 2010 NEJM
article (Reference 31: Bibbins-Domingo et al, NEJM, 2010) using complex modeling to estimate the reduction in cardiovascular
disease events that might be achieved, (with
all the assumptions inherent in the model) if sodium intake were reduced by
certain levels. Their optimistic
estimate, reduction in salt by 3 grams per day (which equates to 1.2 grams of
sodium or about 52 mmol), is greater than what was achieved in a strictly
controlled, monitoring-intensive study of fewer than 1000 hypertensive subjects
followed for a relatively short 30 months (see Whelton et al, JAMA, 1998). To expect that results superior to that may
be achieved on the level of the entire population for extended periods is
optimistic indeed. In any case,
projections based on premises do not support the premises themselves. The next reference in support of this premise
is the meta-analysis by Strazullo (reference 32) which did show that among observational studies, cardiovascular events, especially
stroke, were associated with lower
sodium intake. So we've advanced the
association from sodium and hypertension to sodium and clinical events, but we
still don't have evidence of causality and all the associational caveats of
observational studies apply (also the GIGO - "garbage in, garbage
out" - caveat of meta-analyses).
Enter the clinical trials references (39-41). (Unfortunately, because I don't have an
academic affiliation, I cannot access many of the studies in journals to which
I do not subscribe. I'm a victim of
academia's proprietary monopoly on [mostly] publicly funded data. That's right, the taxpayer pays for it, but can't
access it. Even when it's used to
promote a public health agenda in which the taxpayer is a major
stakeholder. But that's a story for
another post.) Reference 39 (Cook et al, BMJ, 2007) is not really a clinical trial after all. It's a long[er]-term observational follow-up of
patients enrolled in sodium reduction trials, and the results showing improved
cardiovascular outcomes with sodium reduction were barely statistically
significant, and then only after statistical adjustment. Technically, a it's a "FAIL" by rigorous
standards, and certainly not a "long-term, prospective, randomized
clinical trial" as touted by Kotchen et al. Reference 41 (Whelton et al, JAMA, 1998) is
indeed an RCT of sodium reduction in elderly hypertensive patients that shows a
statistically significant reduction in a composite outcome with sodium
reduction as the intervention. A careful look at Table 3
(page 845 of the article) shows that all of the benefit of sodium reduction on
the composite outcome had to do with the ability to withdraw antihypertensive
medications, and there were no reductions in cardiovascular events in the sodium
reduction group. If you want to show that sodium reduction does something to "clinical outcomes", this is a very clever design indeed. But for patients, the operative question remains: would I rather restrict my sodium intake, or take an antihypertensive medication?
The review goes on, and so could this post, but I'll stop
here to spare the reader the drivel. (Nor will I attempt to entertain the reader with links to pages all over the web where proponents and detractors of sodium reduction battle furiously over their positions, along with the evidence of bias and conspiracy alleged by both sides.) The
above paragraphs summarize the main data upon which the purported benefits of
sodium reduction are predicated. A perusal of these data shows that the premises, the very foundations of the
arguments proffered, are precarious.
Here's what we can say with some confidence:
- Most people consume far more sodium than is necessary for survival.
- All studies of sodium consumption and reduction thereof are subject to severe methodological and logistical limitations.
- There are associations between sodium intake and blood pressure, and between blood pressure and clinical outcomes of interest.
- Reduction of sodium intake in patients with hypertension lowers blood pressure.
- None of this establishes a causal link between sodium intake or reduction thereof and clinical outcomes of interest other than hypertension in hypertensive persons.
So, what does this mean for patients and public policy? For patients, if you don't have hypertension,
you probably don't need to worry about sodium intake - for now. If you do have hypertension, you can consider
sodium intake reduction among several alternative options for reducing your blood
pressure. Some may prefer to take a
pill, some may prefer to limit sodium intake, some may decide to lose weight or exercise or reduce alcohol consumption (or initiate it - it's Mediterranean!), some may elect to (or be forced
to) do a combination of these things.
As for public policy (the nuances and philosophy and ethics
and morals and politics thereof are far too many to discuss in this post), for me it's
clear. Before the true-believers (that
is, those guided by faith rather than by reason; the devout; the ideological - as
opposed to the reasoned; the scientific; the logical) - before they begin to
tinker with the food that people eat, they had better have robust data of
causation as the logical foundation of their enterprise, lest they rush us
headlong into another ill-begotten catastrophe like hormone replacement therapy. Behold how quickly the lessons of history are
forgotten, especially by the zealous.
I don't like feeling that somebody is trying to hoodwink me (or tell me what to eat). So the true-believers and
public-policy-tinkerers can go pound salt.
The rest of us can keep eating it.
Added 6/7/2013: See this NYT article about the recent IOM report on reducing sodium intake.
Added 2/10/15: The PURE study was published in the NEJM in August, 2014. I wrote this letter to the editor, which I thought had a high likelihood of publication, but alas it did not. Suffice it to say that the PURE study suffers from some of the same limitations described above:
Added 6/7/2013: See this NYT article about the recent IOM report on reducing sodium intake.
Added 2/10/15: The PURE study was published in the NEJM in August, 2014. I wrote this letter to the editor, which I thought had a high likelihood of publication, but alas it did not. Suffice it to say that the PURE study suffers from some of the same limitations described above:
Two important limitations could have profound implications for the interpretation of the PURE Investigators' results [1, 2]. Firstly, the use of a single urinary sodium measurement assumes that sodium balance is in a steady state at the time of measurement. This assumption may be false if a sodium retaining state such as heart failure is developing. This would lead to misclassification of a person with incident cardiovascular disease as having lower than actual sodium intake. Secondly, the urinary sodium surrogate ignores the substantial sodium lost in sweat (on the order of 1 gram sodium per liter sweat)[3]. This will lead to misclassification of lower sodium intake in those who exercise vigorously. If they also have low blood pressure and low mortality, it will be attributed to the spuriously low sodium excretion rather than vigorous exercise. The analysis included self-reported physical activity but perceptions of physical activity are known to be inaccurate and the temperature and humidity in which the activity takes place affect the amount of sweat and its sodium content [3, 4].Reference List1. Mente A, O'Donnell MJ, Rangarajan S, McQueen MJ, Poirier P, Wielgosz A, et al. Association of Urinary Sodium and Potassium Excretion with Blood Pressure. New England Journal of Medicine 2014; 371(7):601-611.2. O'Donnell M, Mente A, Rangarajan S, McQueen MJ, Wang X, Liu L, et al. Urinary Sodium and Potassium Excretion, Mortality, and Cardiovascular Events. New England Journal of Medicine 2014; 371(7):612-623.3. Bates G, Miller V. Sweat rate and sodium loss during work in the heat. Journal of Occupational Medicine and Toxicology 2008; 3(1):4.4. Canning KL, Brown RE, Jamnik VK, Salmon A, Ardern CI, Kuk JL. Individuals Underestimate Moderate and Vigorous Intensity Physical Activity. PLoS ONE 2014; 9(5):e97927.
No comments:
Post a Comment