The

n e w e ng l a n d j o u r na l

of

m e dic i n e

spectively (P = 0.09); the group assigned to candesartan also had significantly lower rates of a first hospital admission for worsening heart failure than did the group assigned to placebo (24.2% vs. 28.0%, P = 0.02) and fewer hospital admissions for heart failure overall (P = 0.002). Of course, careful monitoring of biochemical levels is essential when two blockers of the reninangiotensin aldosterone system are used in combination.5 I am aware of Szabo and colleagues interesting findings. It will be important to replicate the apparent benefits of iron replacement shown in their study and to determine the effect of this treatment on the rates of hospitalization and death among patients with heart failure. John J.V. McMurray, M.D.
University of Glasgow Glasgow, United Kingdom j.mcmurray@bio.gla.ac.uk

Since publication of his article, the author reports no further potential conflict of interest.
1. Juurlink DN, Mamdani MM, Lee DS, et al. Rates of hyper-

kalemia after publication of the Randomized Aldactone Evaluation Study. N Engl J Med 2004;351:543-51. 2. Cohn JN, Tognoni G. A randomized trial of the angiotensinreceptor blocker valsartan in chronic heart failure. N Engl J Med 2001;345:1667-75. 3. McMurray JJ, Ostergren J, Swedberg K, et al. Effects of candesartan in patients with chronic heart failure and reduced leftventricular systolic function taking angiotensin-convertingenzyme inhibitors: the CHARM-Added trial. Lancet 2003;362: 767-71. 4. 4. Desai AS, Swedberg K, McMurray JJ, et al. Incidence and predictors of hyperkalemia in patients with heart failure: an analysis of the CHARM Program. J Am Coll Cardiol 2007;50:195966. 5. McMurray J, Cohen-Solal A, Dietz R, et al. Practical recommendations for the use of ACE inhibitors, beta-blockers, aldosterone antagonists and angiotensin receptor blockers in heart failure: putting guidelines into practice. Eur J Heart Fail 2005;7: 710-21.

Health Care System Rankings

To the Editor: In their Perspective article (Jan. 14 issue),1 Murray and Frenk review a number of indicators of the relatively poor state of the populations health in the United States. Most, if not all, of this information is well known to readers of the Journal, and the authors use of it is not objectionable. However, Murray and Frenk begin their discussion by referring to the World Health Report 2000, Health Systems: Improving Performance, from the World Health Organization (WHO), which ranked the U.S. health care system 37th in the world, and this is objectionable. (I was editorin-chief of the World Health Report 2000 but had no control over the rankings of health systems.) Fully 61% of the numbers that went into that ranking exercise were not observed but simply imputed from regressions based on as few as 30 actual estimates from among the 191 WHO member countries. Where the United States is concerned, data were available only for life expectancy and child survival, which together account for only 50% of the attainment measure. Moreover, the responsiveness component of attainment cannot be compared across countries, and the estimates of responsiveness for some countries were manipulated. This is not simply a problem of incomplete, inaccurate, or noncomparable data; there are also sound reasons to mistrust the conceptual framework behind the estimates, since it presupposes a production function for health system outcomes
1546
n engl j med 362;16

that depends only on a countrys expenditure on health and its level of schooling, ignoring all cultural, geographic, and historical factors.2 The number 37 is meaningless, but it continues to be cited, for four reasons. First, people would like to trust the WHO and presume that the organization must know what it is talking about. Second, very few people are aware of the reason why in this case that trust is misplaced, partly because the explanation was published 3 years after the report containing the ranking. Third, numbers confer a spurious precision, appealing even to people who have no idea where the numbers came from. Finally, those persons responsible for the number continue to peddle it anyway. To quote Wolfgang Paulis dismissal of a theory opposed to quantum mechanics, Not only is it not right, its not even wrong! Analyzing the failings of health systems can be valuable; making up rankings among them is not. It is long past time for this zombie number to disappear from circulation. Philip Musgrove, Ph.D.
Health Affairs Bethesda, MD No potential conflict of interest relevant to this letter was reported.
1. Murray CJL, Frenk J. Ranking 37th measuring the perfor-

mance of the U.S. health care system. N Engl J Med 2010;362: 98-9.
april 22, 2010

nejm.org

The New England Journal of Medicine Downloaded from nejm.org on August 6, 2012. For personal use only. No other uses without permission. Copyright 2010 Massachusetts Medical Society. All rights reserved.

correspondence
2. Musgrove P. Judging health systems: reflections on WHOs

methods. Lancet 2003;361:1817-20.

The Authors Reply: We welcome vigorous debate about all aspects of assessment of health system performance, with the ultimate goal of improving the health of populations. The World Health Report 20001 generated a rich discussion that advanced this field of study enormously, leadChristopher J.L. Murray, M.D., D.Phil. ing to a 927-page volume published by the WHO University of Washington in 20032 and feeding into health care reform ini- Seattle, WA tiatives in China, Mexico, Iran, and elsewhere. Julio Frenk, M.D., Ph.D., M.P.H. Musgrove criticizes our Perspective article for Harvard School of Public Health citing rankings from the 2000 report. It is impor- Boston, MA tant to note that he was one of the reports lead Since publication of their article, the authors report no furauthors and that 3 years after its publication, he ther potential conflict of interest. took the unusual step of critiquing his own re- 1. World health report 2000 health systems: improving perport.3 At that time, we published a thorough re- formance. Geneva: World Health Organization, 2000. 2. Murray CJL, Evans D, eds. Health system performance assessbuttal.4 ment: debates, new methods and new empiricism. Geneva: World The United States spends more on health care Health Organization, 2003. and yet has worse rates of death and a higher dis- 3. Musgrove P. Judging health systems: reflections on WHOs methods. Lancet 2003;361:1817-20. ease burden than countries that spend far less. 4. Brundtland GH, Frenk J, Murray CJL. WHO assessment of Apologists for the U.S. system tend to ignore these health system performance. Lancet 2003;361:2155.

facts and attempt to distract observers from the real challenges the country faces. If we do not build a strong evaluation component into reform, we will miss opportunities for learning through implementation, correcting the course if needed, promoting accountability, and mustering public support. Unfortunately, on these critical points, Musgrove is silent.

Failure to Validate Association between 12p13 Variants and Ischemic Stroke

To the Editor: Recently published data from a meta-analysis of genomewide association studies showed that two common single-nucleotide polymorphisms (SNPs) near the NINJ2 and WNK1 genes on chromosome 12p13 were associated with ischemic and, in particular, atherothrombotic strokes.1 To validate these results, we conducted a meta-analysis in a combined sample of 8637 cases and 8733 controls of European ancestry, as well as one population-based genomewide cohort study of 278 ischemic strokes among 22,054 participants. Both SNPs from the original report (rs12425791 and rs11833579) were tested for association with ischemic stroke, as well as incident stroke, recurrent stroke, and stroke subtypes (according to the Trial of Org 10172 in Acute Stroke Treatment [TOAST] criteria2). We also conducted similar analyses in cases and controls of African-American ancestry and in samples from Chinese and Pakistani subjects. Furthermore, we evaluated gene expression for NINJ2 and WNK1 in 132 cases with ischemic stroke and in 80 controls, as well as the association of the chromosome 12p13 variants
n engl j med 362;16

with the risk of hemorrhagic stroke. Details regarding the study design and extended results are available in the Supplementary Appendix, available with the full text of this letter at NEJM.org. We found no association between rs12425791 and ischemic stroke (odds ratio, 0.97; 95% confidence interval [CI], 0.91 to 1.04; P = 0.41) or between rs11833579 and ischemic stroke (odds ratio, 1.02; 95% CI, 0.95 to 1.10; P = 0.55) in persons of European ancestry (Fig. 1). We found no association between either of the SNPs and atherothrombotic stroke in 2235 cases (P>0.10) (Fig. 2) and no association between either of the SNPs and incident ischemic stroke, recurrent ischemic stroke, and ischemic stroke subtypes (P>0.10 for all comparisons). The power to detect an association at an effect size below the lower limit of the 95% confidence interval surrounding the original estimate exceeded 99% for ischemic as well as atherothrombotic stroke. We observed no differences in gene expression between cases and controls and no association between either of the SNPs and a risk of hemorrhagic stroke. Heterogeneity in meta-analyses of our studies was
april 22, 2010

nejm.org

1547

The New England Journal of Medicine Downloaded from nejm.org on August 6, 2012. For personal use only. No other uses without permission. Copyright 2010 Massachusetts Medical Society. All rights reserved.