Editor's Choice | This Week in BMJ | Press releases
BMJ No 7106 Volume 315 Letters Saturday 23 August 1997 Intersalt dataReply to Rennolls's letterEditor,Keith Rennolls's letter was one of seven letters to the editor that the BMJ sent us for reply, after publication of our two Intersalt papers in the issue of 18 May 1996.(1,2) With no more than 500 words allowed us by the BMJ, we found ourselves with that quota already met by our comments on the first six letters, and we judged it essential to leave those intact. Therefore, in our response we noted that our reply to Rennolls would be available on the BMJ website. We note Rennolls's statements: "Research indicating the influence of dietary salt on blood pressure is convincing" and "I would not wish to support the analyses of Hanneman." Especially in the light of these judgements, we find his several qualms about the Intersalt analyses and findings puzzling, as discussed specifically below, in the sequence in which they appear in his letter. (1) Rennolls talks of the criss crossing of population trends - that is, of plots - for each of the 52 Intersalt population samples - of relationship of age to blood pressure, systolic and diastolic (a) Does he mean "criss crossing" of sample plots based on observed Intersalt data for people aged 20-59 or, alternatively, of plots extrapolated beyond the observed data - for example, down to age 0, the intercept (as per the Salt Institute's analyses by Hanneman (see (2) below)? (b) In particular, why would any criss crossing of plots across ages 20-59 be relevant or have meaning for the scientific matter of concern - that is, the relation in ecological analyses (n=52) of the population sample median (or mean) 24 hour sodium excretion to the slope of blood pressure with age? We find it of no relevance for this crucial matter that sample curves of the observed relation between age and blood pressure, whether linear or curvilinear, may cross each other. Our reasons for this conclusion are straightforward: Several lifestyle factors are known to influence the blood pressure of populations - that is, dietary salt and potassium intake, and ratio of salt to potassium; energy intake and physical activity, consequent energy balance and relative weight (leanness, obesity); and alcohol intake. Several others probably also do - for example, dietary magnesium and protein. Given the varied mix of such multiple traits across populations worldwide (as shown by Intersalt), it is to be expected that population sample median (or mean) blood pressures for young adults vary across populations, that slopes of blood pressure with age (from these young adult start points through middle age) vary across populations, and that some of these curves cross. For the Intersalt Steering and Editorial Committee, the relevant point is that, since slope of blood pressure from young adulthood through middle age is influenced by multiple factors, exploration of the effect of each trait requires control for other traits, as done by Intersalt. Beyond this, we know of no relevance of criss crossing (in so far as it occurs). Moreover, if Rennolls is referring to criss crossing based on extrapolations to age 0 (intercept), such extrapolations are methodologically erroneous and produce grossly misleading "findings" (for example, those of the Salt Institute in the 18 May issue of the BMJ) (see (2) below). (c) For the Intersalt Steering and Editorial Committee, a first point warranting emphasis is the consistent significant upward slope of both systolic blood pressure and diastolic blood pressure with age, from young adulthood into middle age (for example, age 25 to age 55), for 48 of the 52 Intersalt population samples - that is, for all samples except the four isolated samples with a low sodium intake (Xingu and Yanomamo Indians of Brazil, highlanders of Papua New Guinea, Kenyan farmers).(3) The optimal average blood pressure of the latter four samples, their virtual freedom from higher average blood pressure with age, their zero or very low prevalence of hypertension - along with similar findings for other isolated populations around the world, and data showing loss of these favourable blood pressure patterns when such peoples migrate and adopt modern lifestyles - all have great scientific portent. A rise in blood pressure with age from youth through middle age is not "man's fate," not an inevitable consequence of the human condition, not a built in outcome of effects of the human genome. Human evolution produced a predisposition for such rises on a mass scale, "loaded the gun" genetically, but whether the "trigger is pulled" - whether this predisposition actually becomes expressed extensively in the population - depends on lifestyles, environmental exposures, and nurture. The related lesson from the Intersalt data, as reaffirmed in the issue of 18 May, is that high population average sodium (salt) intake is one of the main widespread exposures that is significantly, independently, and strongly related to the occurrence of substantially higher blood pressure in most middle aged compared to young adult populations. In this regard, the concordance on this matter of data from Intersalt earlier and recent cross population analyses, the latter done at the request of the Salt Institute, speaks to their robustness, and hence merits emphasis. (The Salt Institute persists in its silence about these findings, generated in response to its requests.) In sum, these are the essential Intersalt findings; criss crossing is of no import. (2) Particularly since Rennolls took pains in his letter to dissociate himself from the analyses by Hanneman in the 18 May issue, we are puzzled by his two assertions defending use of the intercept - that is, extrapolation of the population sample blood pressure linear slopes from age 20-59 down to age 0. He infers that this manoeuvre in Hanneman's paper is not unsound statistically and not productive of biologically spurious data. But, as we noted in our commentary on Hanneman's article, statistical monographs dealing with regression analysis explicitly caution against such extrapolations. Thus, Rosner states: "In general, it is dangerous to make predictions from a regression line for values of x that are very far from x since they are likely to be very inaccurate."(4) Similarly, Neter et al state: "Another caution deals with inferences pertaining to levels of the independent variable which fall outside the range of observations. If the X level does not fall far beyond this range, one may have reasonable confidence in the application of the regression analysis. On the other hand, if the X level falls far beyond the range of past data, extreme caution should be exercised since one cannot be sure that the regression function which fits the past data is appropriate over the wider range of independent variable."(5) Also, explicit to the matter under consideration, Kleinbaum and Kupper state: "data are not usually gathered near the origin. For example, when dealing with age (X) and blood pressure (Y), we are not interested in knowing what happens at X=0 and we rarely choose values of X near 0."(6) As to the biological implausibility of "findings" from extrapolation of population sample systolic pressure or diastolic pressure to age 0 (the intercept) from Intersalt linear regression slopes of blood pressure on age across ages 20 through 59, examples - as used by Hanneman - virtually speak for themselves in documenting this. Thus, for the highlanders of Papua New Guinea, median systolic pressure for the whole sample aged 20-59 was 107.7 mm Hg; for those aged 20-29, 113.0 mm Hg; for those aged 50-59, 107.0 mm Hg - that is, a downward slope of systolic pressure with age,(3)(7-8), so that Hanneman's extrapolation of this slope down to the intercept (his fig 1)(1.9 yields the estimate of a systolic pressure at age 0 of about 113 mm Hg - a biologically most unlikely estimate. In sum, as Law and we noted,(2)(10) use of the statistically unsound procedure of extrapolating slopes of blood pressure on age for ages 20-59 down to the age 0 intercept yields bizarre, essentially spurious, findings. (3) We are also puzzled by Rennolls's next assertion and his basis for propounding it. He characterises as "problematic" Intersalt estimates of the 30 year effect on blood pressure of a 100 mmol/day difference in daily sodium intake (for example, 70 v 170 mmol from age 25 to age 55). These point estimates (blood pressure different by 9-11/5-6 mm Hg) were highly consistent with use of each of several statistical models to derive them (including all models proposed by the Salt Institute) and with control for multiple possible confounders (including blood pressure at age 25, as proposed by the Salt Institute). The standard errors of these estimates - computed for each model with use of all the data available, without exclusion - were also consistent (for systolic pressure 1.7-2.1 mm Hg, for diastolic pressure 0.9-1.1 mm Hg), as were consequently the 95% confidence intervals around the point estimates (for systolic pressure about 5 to 15 mm Hg, for diastolic pressure about 4 to 8 mm Hg). Frankly, we do not grasp Rennolls's argument that something about the standard errors makes these robust point estimates problematic. The statistical techniques used were standard regression or curve fitting procedures, used extensively to compute such point estimates and their standard errors. If these estimates are problematic, so must be most other ones similarly derived in biological research. (4) As to Rennolls's qualm about whether Intersalt analyses took into account "possible inhomogeneity of the variances about the regressions," no analyses involved variances about the regressions - neither analyses of the relation between age and blood pressure for each of the 52 samples nor analyses on the relation of sample median (or mean) 24 hour sodium excretion to sample linear slope of blood pressure with age or to sample difference in blood pressure with age (for example, blood pressure at age 55 minus blood pressure at age 25). It is not evident to us why this is relevant and what taking this into account might mean. (5) As to Rennolls's concern about whether the "prediction theory [[was]] correctly based on the errors in regressor variables model," no such model was used; again, it is not clear what he means here. How might such a model be meaningfully used in ecological analyses of the type done, and under discussion here? To reiterate, our prior ecological hypothesis in the area under discussion was: Across the 52 samples there is a significant independent positive relation of sample median (or mean) 24 hour sodium excretion to sample slope of systolic pressure (or diastolic pressure) with age. The dependent variable - sample slope of blood pressure with age-was assessed to be essentially unaffected by measurement error, since age was measured essentially without error. The independent variable - sample median (or mean) 24 hour sodium excretion - was also judged to be essentially unaffected by measurement error, since it is a value for the whole sample (n=about 200) and not for an individual, and therefore is not subject to intra-individual variability. We are unclear as to Rennolls's reasons for implying that the errors in the regressor variables model is relevant and useful in this situation. As to the other aspect of Rennolls's letter, it is for all researchers (ourselves included) a truism that "results should be reproducible by independent scientists." In fact, the Intersalt study was undertaken explicitly to test, with use of high quality standardised methods, the reproducibility of published positive findings on dietary salt and blood pressure from both cross population (Dahl; Gliebermann; Froment et al) and within-population investigations.(3)(10-14) Intersalt data confirmed and extended previous positive findings from these earlier epidemiological studies on the relation between salt and blood pressure - that is, showed such results to be highly reproducible. Furthermore, all these epidemiological findings are concordant with extensive data from other research disciplines (clinical pathophysiology, clinical therapeutics, randomised controlled trials, animal experimentation, anthropology, evolutionary biology). It is this totality of evidence that is the foundation for the conclusion, repeatedly set down by expert groups, that there is an aetiologically significant relation between dietary salt and blood pressure.(1-2)(15-23) Or, as Rennolls puts it in his first sentence, "Research indicating the influence of dietary salt on blood pressure is convincing." Given the extensive concordant data, including reproducible epidemiological data from multiple studies, why then does Rennolls imply the need "to reproduce data comparable to those of the Intersalt study"? Reproducibility has already been shown. In any case, this tested process of assessing the reproducibility of findings by separate repeated studies in varied samples and under varied circumstances is entirely different from Rennolls's proposal to have data from a given study turned over to other researchers for "confirmatory analyses." This has not been, and is not now, a component of the investigative process. Hence there is no factual foundation for Rennolls's assertion that this is "necessary [[for]] the medical-scientific community to achieve a shared understanding." On the contrary, as both the earlier and later history of progress in medical and public health amply document, shared understanding has been repeatedly achieved - based on reproducible data from many and varied studies - entirely without "confirmatory analyses." To cite a close-to-home and powerful example, note the progressive development by the medical-scientific community in recent decades of extensive knowledge and shared understanding on the causes of epidemic coronary-cardiovascular disease. This knowledge and understanding have been the scientific foundation for the role of the medical-scientific community in the development of effective public policy to turn the flank of this epidemic. Many have played and continue to play a part in all this, individually and through key societies, in keeping with well established traditions (scientific, social, and ethical) of the health professions. Members of the Intersalt Steering and Editorial Committee are among these colleagues. The great progress against epidemic coronary heart disease-cardiovascular disease has been achieved virtually without Rennolls's proposed "confirmatory analyses by other researchers." We support data sharing and practise it; in this regard, a few specifics are relevant. (A) Particularly the more senior among us have been involved in data sharing for decades - for example, in the US national cooperative pooling project(24); for US National Heart, Lung, and Blood Institute conference overviews(25-27); for other overviews(28); in the multiple risk factor intervention trial(29-30); and with the Salt Institute (see below). (B) Correspondingly, the more senior among us have been involved for years in studies that - at an appropriate juncture and in an appropriate way - made their data available to a data repository for general access - for example, the hypertension detection and follow-up programme and the systolic hypertension in the elderly programme. (C) On the basis of decades of research experience, we judge the recent novel idea that studies should make all their data available after a "primary analysis" to be fraught with complexities. Its productive and counterproductive aspects need careful delineation and thorough discussion, which are not amenable to pursuit here. (D) As to the Intersalt study, it is not a one-question project; it is a long term ongoing programme of investigation, with current funding in support of its work. It is not in the situation of having just one primary analysis. Thus, in its major second phase, pursued in the 1990s, it went beyond its original focus on sodium, potassium, and blood pressure, to study the relation of urinary markers of dietary protein to blood pressure and the relation of electrolytes to blood pressure in that context. This required more than three years of further analytical work by the Intersalt Central Laboratory, to measure total nitrogen, urea, and sulphate in 24 hour urine collections by the >10,000 Intersalt participants worldwide - laboratory work recently completed. A first definitive report of the findings on protein and blood pressure has only just been published.(31) Further analyses and reports of Intersalt are pending. It is therefore inappropriate for the Intersalt data files to be turned over to a data repository. (E) In Intersalt, as in other studies involving its senior colleagues (see above), the policy continues to be one of welcoming scientific proposals from other investigators for joint research based on the Intersalt dataset, given a scientifically sound set of projected investigative tasks and the feasibility of their accomplishment in a cooperative collegial endeavour. In view of this policy, we deem Rennolls's statement invalid that our approach to data sharing "forces dissidents into a limited framework of scientific discussion." (F) Intersalt responded positively to the Salt Institute's requests for further data analyses - to those assessed as sound scientifically (at least marginally). It carried them out at considerable cost in time and effort (without compensation); made the findings available to the Salt Institute for its use on a confidential basis pending publication; submitted a paper to the BMJ embodying the findings, which were ultimately published after revision in response to peer review comments.(1) All this sharing of data took place despite the fact that the request from the Salt Institute came not from "bona fide researchers" serving as consultants to that organisation or from its staff, but only from its lawyers, in an adversarial letter. Our primary reason for responding positively was to make clear to scientific colleagues our willingness to share data. The Salt Institute has to this date said nothing about the strong consistent significant findings on sodium and blood pressure in the tabulations prepared in response to its request. It was silent on these results in its use (misuse) of Intersalt data before the Advisory Committee on the US Dietary Guidelines(32-33) and in its piece in the BMJ.(2)(9) In Rennolls's judgment, does this experience have anything in common with data sharing with bona fide researchers so that the medical-scientific community can achieve shared understanding? Finally, in our judgment, Rennolls is not reflecting the true situation in regard to "the conflict" (his second paragraph): The conflict in this instance - in the US, Britain, and other countries - is between for - profit salt producers, their trade organisations (including the Salt Institute), and public policy. Their objective is to end public health recommendations to the population to eat less salt (1-2)(15-23) This is the meaning of the sustained attack by the Salt Institute on the Intersalt data and investigators. Having failed with respect to the fourth edition of the dietary guideline for Americans,(21)(34) the Salt Institute is currently seeking to nullify all reference to salt in the US Food and Drug Administration's regulations, including material on American food labels advising the American people to consume less than 2400 mg sodium/day (that is, less than about 100 mmol sodium/day, equivalent to less than about 6 g of salt/day). For all professional colleagues concerned with scientific findings and their application for prevention and control of major disease, it is important to do everything possible to resist and rebuff these current initiatives. So doing is in the tradition of successful struggles for public health against special interests, including many-sided struggles (remote and recent) for better nutrition. Each of us needs to be counted in this scientific, social, ethical endeavour.
Jeremiah Stamler
Professor emeritus
Department of Preventive Medicine, Hugo Kesteloot Professor
Department of Epidemiology, Michael Marmot Professor
Department of Epidemiology and Public Health, Paul Elliott Professor
Department of Epidemiology and Public Health, On behalf of the Intersalt Cooperative Research Group References 1 Elliott P, Stamler J, Nichols R, Dyer A R, Stamler R, Kesteloot H, et al for the Intersalt Cooperative Research Group. Intersalt revisited: further analysis of the 24 hour sodium excretion and blood pressure within and across populations. BMJ 1996;312:1249-53. 2 Stamler J, Elliott P, Dyer AR, Stamler R, Kesteloot H, Marmot M for the Intersalt Steering and Editorial Committee. Sodium and blood pressure in the Intersalt study and other studies - in reply to the Salt Institute. BMJ 1996;312:1285-7. [Commentary on article by R L Hanneman.] 3 Intersalt Cooperative Research Group. Intersalt: an international study of electrolyte excretion and blood pressure. Results for 24 hour urinary sodium and potassium excretion. BMJ 1988;297:319-28. 4 Rosner B. Fundamentals of biostatistics. Boston: Duxbury, 1982:368. 5 Neter J, Wasserman W, Kutner M. Applied linear regression models. Homewood, IL: Richard D Irwin, 1983:83. 6 Kleinbaum D G, Kupper L L. Applied regression analysis and other multivariable methods. North Soituate, MA: Duxbury, 1978:56. 7 INTERSALT Cooperative Research Group (Elliott P, guest editor). The INTERSALT study - an international cooperative study of electrolyte excretion and blood pressure - further results. J Hum Hypertens 1989;3:279-407. 8 Mancilha-Carvalho J J, Baruzzi R G, Howard P F, Poulter N, Alpers M P, Franco L J, et al. Blood pressure in four remote populations in the INTERSALT study. Hypertension 1989; 14:238-46. 9 Hanneman R L. Hypertension rise with age revisited. [With commentaries by M Law and J Stamler et al.] BMJ 1996; 312:1283-7. 10 Law M. Evidence on salt is consistent. BMJ 1996;312:1284-5. [Commentary on article by R L Hanneman.] 11 Dahl L. Possible role of salt intake in the development of hypertension. In: Cottier P, Bock K D, eds. Essential hypertension: an international symposium. Berlin: Springer-Verlag, 1960:53-65. 12 Gliebermann L. Blood pressure and dietary salt in human populations. Ecol Food Nutr 1973;2:143-56. 13 Froment A, Milon H, Gravier C. Relationship of sodium intake and arterial hypertension. Contribution of geographical epidemiology. Rev Epidemiol Sante Publique 1979;27:437-54. 14 Elliott P. Observational studies of salt and blood pressure. Hypertension 1991;17(suppl I):I3-8. 15 WHO Expert Committee on the Prevention of Coronary Heart Disease. Prevention of coronary heart disease. WHO Tech Rep Ser 1982; No 678. 16 WHO Expert Committee on Prevention in Childhood and Youth of Adult Cardiovascular Disease. Prevention in childhood and youth of adult cardiovascular diseases-time for action: report of a WHO expert committee. WHO Tech Rep Ser 1990; No 792. 17 WHO Study Group. Diet, nutrition, and the prevention of chronic diseases. WHO Tech Rep Ser 1990; No 797. 18 National Research Council, Committee on Diet and Health, Food and Nutrition Board, Commission on Life Sciences. Diet and health: implications for reducing chronic disease. Washington, DC: National Academy Press, 1989. 19 National High Blood Pressure Education Program. The fifth report of the Joint National Committee on Detection, Evaluation, and Treatment of High Blood Pressure (JNC V). Arch Intern Med 1993;153:154-83. 20 National High Blood Pressure Education Program Working Group. Report on primary prevention of hypertension. Arch Intern Med 1993;153:186-208. 21 Dietary Guidelines Advisory Committee. Report of the Dietary Guidelines Advisory Committee on the dietary guidelines for Americans. Washington, DC: US Government Printing Office, 1995. 22 Department of Health. Nutritional aspects of cardiovascular disease. Report of the cardiovascular review group, Committee on Medical Aspects of Food Policy. London: HMSO, 1994. (Report of health and social subjects No 46.) 23 Krauss R M, Deckelbaum R J, Ernst N, Fisher E, Howard B V, Knopp R H, et al. Dietary guidelines for healthy American adults. A statement for health professionals from the nutrition committee, American Heart Association. Circulation 1996;94:1795-800. 24 Pooling Project Research Group. Relationship of blood pressure, serum cholesterol, smoking habit, relative weight and ECG abnormalities to incidence of major coronary events: final report of the pooling project. J Chronic Dis 1978;1:201-306. 25 Dyer A R, Stamler J, Shekelle R B. Serum cholesterol and mortality from coronary heart disease in young, middle-aged, and older men and women from three Chicago epidemiologic studies. Ann Epidemiol 1992;2:51-7. 26 Manolio T A, Pearson T A, Wenger N K, Barrett-Connor E, Payne G H, Harlan W R. Cholesterol and heart disease in older persons and women. Review of an NHLBI workshop. Ann Epidemiol 1992;2:161-76. 27 Jacobs D, Blackburn H, Higgins M, Reed D, Iso H, McMillan G, et al. Report of the conference on low blood cholesterol: mortality associations. Circulation 1992;86:1046-60. 28 MacMahon S, Peto R, Cutler J, Collins R, Sorlie P, Neaton J, et al. Blood pressure, stroke, and coronary heart disease. Part 1. Prolonged differences in blood pressure: prospective observational studies corrected for the regression dilution bias. Lancet 1990;335:765-74. 29 Davey Smith G, Neaton J D, Wentworth D, Stamler R, Stamler J. Socioeconomic differentials in mortality risk among men screened for the multiple risk factor intervention trial. I. White men. Am J Public Health 1996;86:486-96. 30 Davey Smith G, Neaton J D, Wentworth D, Stamler R, Stamler J. Socioeconomic differentials in mortality risk among men screened for the multiple risk factor intervention trial. II. Black men. Am J Public Health 1996;86:497-504.
31 Stamler J, Elliott P, Kesteloot H, Nichols R, Claeys G, Dyer A R, et al for the Intersalt Cooperative Research Group. Inverse relation of dietary protein markers with blood pressure. Findings for 10
32 Hanneman R L. Letter to the Dietary Guidelines Advisory Committee, US Department of Agriculture and Department of Health and Human Services, January 31, 1995. Alexandria, VA: Salt Institute, 1995.
33 Stamler J. Letter regarding salt dietary guideline, to Dietary Guidelines Advisory Committee, US Department of Agriculture and Department of Health and Human Services [in reply to the letter of the Salt Institute, January 31, 1995.] Chicago, IL: Department of Preventive Medicine, Northwestern University Medical School, 27 February 1995.
34 US Department of Agriculture, US Department of Health and Human Services. Dietary guidelines for Americans. 4th ed. Washington, DC: US Government Printing Office, 1995.
|