Letter From Cohen and Alderman Regarding Article, “Lower Levels of Sodium Intake and Reduced Cardiovascular Risk”
To the Editor:
We read with interest “Lower Levels of Sodium Intake and Reduced Cardiovascular Risk”1 but are concerned that the validity of the results and conclusions of this article may be questioned by the methodological approaches used. For example, a very large proportion of participants were excluded—all those who were in the salt restriction arms of the studies, including about 15% of Trial of Hypertension Prevention, Phase I (TOHP I)2 and 50% of Trial of Hypertension Prevention, Phase II (TOHP II).3 The stated justification was that the sodium measures would not reflect long-term “usual values.” Maybe so, but in a previous article, the authors maintained that “questionnaire data support the long-term effects of the intervention”4 to justify why their follow-up results were attributed to long-term lower sodium intake. Thus, because the authors interpret their previous results as evidence that salt restriction is beneficial, we question why they have excluded all of those who were most likely to have sodium values in the range of 2300 mg or lower.
Also excluded were outcome events that occurred during the trial. For TOHP I, this would be about 18 months, and for TOHP II at least 3 years. In this time, there were 15 deaths, 7 of which were cardiovascular disease (CVD).
The authors have never revealed how many nonmortal CVD events took place within the TOHP trials, even though, to our knowledge, most studies report outcome events from the time the study starts rather than only after it ends. In the period after the trial, there were about 2.5 total CVD events per 1 all-cause death. If that ratio occurred during the study period, as many as 35 CVD events might have been excluded.
We believe that it would be important to see the results when the 1518 sodium restriction participants (40% of sample) and all the excluded in-trial CVD events (perhaps 15% of the total) are included.
Lastly, the linear association reported with a spline analysis, and for sodium as a continuous variable, obscures what appears from Figure 2 to be the substantial influence of sodium values >5500 mg. It is well known (see Anscombe’s quartet)5 that even a small proportion of influential values can give a false estimation of statistical linearity. Because there is no controversy about the very high end of sodium intake, and the main inference the authors wish to convey relates to low values (eg, <2300 mg) readers need to know the result of a sensitivity analysis of linearity confined to subjects consuming <5500 mg.
In sum, it is our opinion that the borderline significant findings from an analysis of only a subset of the whole data are inadequate for the very important topic that Cook et al have addressed. We suggest the authors provide a more complete disclosure and analysis of all the data so that readers can draw more informed conclusions, especially with respect to the lower levels of sodium intake.
Hillel W. Cohen, DrPH, MPH
Michael H. Alderman, MD
Department of Epidemiology and Population Health
Albert Einstein College of Medicine
New York, NY
- © 2014 American Heart Association, Inc.
- Cook NR,
- Appel LJ,
- Whelton PK
- 3.↵The Trials of Hypertension Prevention Collaborative Research Group. Effects of weight loss and sodium reduction intervention on blood pressure and hypertension incidence in overweight people with high-normal blood pressure. The Trials of Hypertension Prevention, Phase II. Arch Intern Med. 1997;157:657–667.
- Cook NR,
- Cutler JA,
- Obazanek E,
- Buring JE,
- Rexrode KM,
- Kumanyika SK,
- Appel LJ,
- Whelton PK