July 26, 2004 

Kathryn McMurry
HHS Office of Disease Prevention and Health Promotion
Room 738-G
200 Independence Avenue, SW
Washington, D.C. 20201 

Dear Kathryn: 

When I wrote last week, I was still unaware that the “Bray paper” mentioned during discussion at the Advisory Committee’s last meeting had been published August 15.[1]  Since one of the co-authors is Dr. Appel, it now seems likely that Dr. Appel shared a copy of the unpublished manuscript with Committee members.  None of the audience at the last meeting, of course, knew the contents of the paper, yet Committee members seemed to be taking it quite seriously as a relevant input into your process.  In fact, it seemed from the conversation that Committee members were of the opinion that the Bray paper not only supported reducing dietary sodium, but also supported a 1,500 mg/day sodium intake level.  Now that the paper has been published, Committee members may wish to reconsider that characterization of the Bray paper.  I have appended a copy for your reference. 

Indeed, the Bray paper is relevant to understand the DASH-Sodium Trial results.   The relevance is more fundamental than the meager additional information beyond that published in the two previous articles.  Unfortunately, its relevance is not highlighted in the authors’ presentation of the data contained in the paper.   

The importance of the paper is to demonstrate that in six of eight reported subgroups, lowering dietary sodium in the diets of those consuming a quality diet (the DASH Diet), there was no statistically significant reduction in blood pressure.  Perhaps Dr. Appel was candid with the other members of the Advisory Committee behind closed doors, but this lack of association is contrary to the conclusions published in Bray et. Al.  The Bray paper shows conclusively that there is no statistical benefit by reducing dietary sodium for the vast majority in the DASH-Sodium Trial. 

Background 

Before addressing the science as reported by Bray and colleagues, it would seem appropriate to observe that this paper is only the latest of a series of efforts to mislead the public and nutrition policy-makers as to the true findings of the DASH-Sodium Trial.  Thus, the presentation of the data in Bray et al is consistent with a disturbing pattern of distortion. 

Bray et. al. took extraordinary license in reporting the data. This has been a repeated the concern of critics since the index report of the DASH-sodium trial in the NEJM.[2]  This derivative paper continues the authors’ unexplained reluctance to provide readers with the most basic of data from any well done randomized clinical trial. These include failing to provide, except at baseline, the means, standard deviations and Ns for systolic and diastolic blood pressures of the entire study population and the subgroups of interest. As with their index report the DASH-Sodium authors continue to report “delta” or “change in blood pressure” values rather than providing the only legitimate data the means, SD’s and sample sized for BP at the end of each intervention period.    

That this incomplete presentation has survived is due to the fact that Dr. Appel has had a lead role guiding the discussion of the Trial in each of the three major groups that have evaluated the DASH-Sodium Trial – the initial study team, the IOM/FNB panel and now the Dietary Guidelines Advisory Committee.  There has not yet been a truly objective critique by any of these groups.  Such an independent review may take place in the context of the judicial challenge scheduled for oral argument in federal court on August 13. 

Bray et. al. Deficiencies 

Beyond continuing the incomplete reporting of the two earlier DASH-Sodium papers,[3] the Bray paper introduces (at least) two new, extraordinary statistical maneuvers. 

1.      Whereas delta blood pressure is reported in lieu of valuable real numbers, the Bray paper does not even use actual deltas, but rather estimates based upon a model.  The authors state:   “Figures 1 and 2 display the estimated joint effects on systolic BP of the DASH diet and sodium reduction for 2 joint subgroups.   In each case, the value for the control diet, the higher sodium combination, is the raw observed mean, and all of the other points are estimated on the basis of the models as described in the methods.”  Thus, the data are theoretical numbers, not real numbers.  It is impossible for an independent expert to reconstruct and validate the model. 

2.      It gets worse.  The models themselves do not adjust for even the most obvious of confounders such as age, gender, BMI, race, BP status, etc.  The authors state:  “Each subgroup variable is examined separately, so that, for instance, the effects in hypertensive and nonhypertensive patients are not adjusted for race, gender, age, or obesity.”  Failure to adjust for these known confounders means that these factors to any changes contribute (unknown or undisclosed) as identified by the model.  Failure to adjust these analyses appropriately defies any reasonable standard of medical publishing of data reported from a clinical trial. Were such a report to be submitted for expert review to an NIH Study Section or FDA Expert Panel, this type of data analysis and presentation would, and should, be rejected. The Dietary Guidelines Committee should reject this analysis as reported and likely conveyed by Dr, Apple and insist upon an appropriate analysis before relying upon the Bray paper. 

Turning now to the BP results, I would direct your attention to Figures 1 and 2 in the paper. The eight panels (four in each figure) portray the change in BP for eight of the pre-defined subgroups.[4]  In each panel, three systolic and three diastolic pressures are recorded for subjects either on the Control Diet or the DASH Diet and one of three levels of sodium employed in each of the 30-day intervention periods. In each panel, as the sodium intake is progressively reduced moving horizontally left to right, a numerical value is recorded for the “estimated” BP change for each interval. Below that number, the graphs report the confidence intervals for each “estimated” change in BP. A statistically significant BP fall requires that the “sign” of the change in BP must be “negative” for any of these interval BP changes and that the confidence interval cannot “cross the zero point.” That means that for each interval reduction in sodium intake the “sign” before both numbers comprising the confidence interval must be “negative” for the effects of sodium reduction to be significant.  

It is readily apparent that this is not the case for the subjects on the DASH Diet, with one exception in each figure (Figure 1 African Americans with hypertension; Figure 2, over 45 years with hypertension).  With those two exceptions, once on the DASH Diet, no subgroup experienced a statistically significant reduction in its blood pressure when sodium intake was lowered from levels typical of the current sodium content of the American diet.  As noted in #2 above, the DASH-Sodium investigators’ failure to adjust each of these subgroup analyses for appropriate confounders and their use of estimates of the change in blood pressure likely further weakens the association and calls into question even the significance of the other two subgroups.  There is a distinct possibility that sodium reduction not only did not lower blood pressure in subjects on a normal diet, but that sodium restriction actually produced a significant BP increase in one or more of the subgroups.

Conclusion 

The DASH Diet is consistent with the Dietary Guidelines.   Though the two DASH trials were only 30 days, both were well done and produced exciting confirmation of the BP efficacy of improving the mineral density of the American diet through increased intakes of dairy products, fruits and vegetables.  The addition of a salt reduction Guideline adds nothing to that benefit and only distracts resources needed to achieve the DASH Diet.  The Bray paper offers no support for salt reduction; rather, it supports the notion that emphasis on the quality of the overall diet is what matters.  Sodium reduction from current levels provided no statistically reproducible benefit for the vast majority of the subjects in the DASH-Sodium Trial.  Surely, the Advisory Committee does not intend to recommend a diet similar to the control diet.  Lacking the nutrients added in the DASH Diet, the poor quality control diet invites salt sensitivity; but is the right answer ignoring the rest of the diet to pursue salt restriction or ignoring salt restriction and focusing on improving overall dietary quality? 

The DASH-Sodium Trial is the only human study where BP effects were studied in subjects eating a normal (control) or healthy (DASH) diet.   True, they are only two studies, but the verdict is clear:   quality diets make a difference; salt restriction doesn’t. 

We invite the Advisory Committee to have a full and open discussion on the DASH-Sodium Trial results at its August 11 meeting, particularly in light of the Bray paper and its evidence of no statistically significant population benefit of reducing dietary sodium.  Perhaps it would assist public understanding of the Dietary Guidelines process to include consideration of how these misrepresentations might have occurred. 

Sincerely,

Richard L. Hanneman
President

[About Salt Institute] [About salt] [About the salt industry] [News] [SI Member Business (password required] [E-Mail Salt Institute]

Search web site: