Soapbox Post

“Faith based research is okay, shoddy research is common, but the two interact and end up ... in PNAS?”¹

Now that the debate over health care reform is beginning to heat up, expect to hear a loudening chorus of voices insisting that the key to the future health of Americans is more research funding for the National Institutes of Health.  An early salvo in this direction was published recently in the flagship publication of the National Academies (of science, engineering, and medicine), PNAS.

The article² is an extraordinary exercise in statistical distortion.  It’s basic points are these:  (1) Rising expenditures on NIH research correlate with rising indicators of health in America; (2) As Americans get (on average) older, the economic well-being of the nation increasingly will depend on them to lead economically productive lives; (3) this, in turn, will demand better health interventions for an aging population; therefore, (4) NIH budgets need to keep up with this economic imperative.  The paper concludes:  “the size of NIH expenditures relative to GDP should quadruple to about 1% (≈$120 billion) and be done sufficiently rapidly (10 years) to compensate for the slowing growth of the U.S. labor force.”

The paper includes four graphs (figure 2), each of which shows a curve of rising funding level over time for a particular NIH research institute, and a curve showing death rates from the diseases that each institute focuses on.  On three graphs (heart disease and the Heart, Lung and Blood Institute; stroke and the Institute of Neurological Disorders and Stroke; and cancer and the Cancer Institute) death rates show declines, and funding rates show increases.  A second set of graphs (figure 3) correlates “acceleration of mortality declines” with funding thresholds achieved at these three institutes.  An additional graph (figure 4) shows that total NIH funding trends and national death rate trends can be closely correlated, even to the extent that periods of rapid or slow growth in NIH budgets correlate with rapid or slow declines in overall death rates.

As aggravating (and common) as it is when scientists use illogical or unscientific arguments to promote science, it’s perhaps even more irritating when they employ bad or deceptive scientific arguments.  Let me just point out a few problems with this paper, and highlight a few issues that it raises.

There are many pathways to good health, many variables that contribute to good health, and complex, incompletely understood relations among these pathways and variables.  Standard of living, level of education, access to affordable medical care, levels of income equity, diet, climate, and other factors all have been shown to have a bearing on public health outcomes.  Claiming a direct causal relation between health outcomes and a single variable (in this case, NIH funding) without considering how other variables may be contributing to the outcome is inherently misleading.  To understand why this is so, imagine that the Clinton Administration’s efforts in the mid-1990s to implement meaningful health care reform had actually succeeded, so that in the ensuing 15 years, millions of more people had had access to affordable health care than has actually been the case.  The resulting improvements in health of the average American would have occurred independently of whatever was going on with the NIH budget.   But a similar analysis to the one in the PNAS paper would have shown the same strong correlations between NIH budget increases and the enhanced health outcomes; the key causal role of changing health policy would be invisible.

A variant of this hypothetical case is on display in countries that actually do make an effort to provide health care access for all citizens.  As recently summarized in an article in the June 25^th issue of The Economist, “Comparisons with other rich countries and within the United States show that America’s health-care system  . . . provides questionable value for money and dubious medical care.”³  The fact is that in many countries (such as Canada, Japan, and many European nations) with equitable, population-wide access to medical care, not only do people live longer, healthier lives than in the U.S., but less money (per capita) is spent on both health care and biomedical research.

The authors state that the fit between total NIH funding and death rate curves (their Figure 4) explains “98% of the variation of age-adjusted mortality rates.  Although [this] does not prove causation it makes the search for alternate explanatory variables of equal power difficult.”  Nonsense.  Given the authors don’t look at any other variables, they cannot test the real-world validity of their correlation.  This is an act of faith, not science; it is a classic formula for generating spurious correlations.  For example, given that budgets for pretty much everything have gone up during the last fifty years, and that budget trends across government programs tend to track one another, there are no doubt many other budget curves that could be nicely matched to the death-rate curve.  And in any case we have seen already that other countries can deliver better health to their citizens for less money and with less research—so even if the correlation had some validity, it would merely underscore the inequity and inefficiency of the U.S. system.

The paper further errs by attributing to NIH (and the NIH budget) activities and outcomes that in fact had little to do with NIH.  For example, the authors state that, in addition to medical interventions, “public health initiatives against smoking, and promoting screening for breast and colon cancers, led to the initiation of U.S. cancer mortality reductions in 1990,” but of course such crucial public health activities are largely outside the domain of NIH.  Moreover, the paper gets its history wrong when it notes that “Total cancer mortality rates did not decline until 1990, 25 years after the identification of the effect of smoking on lung and other cancers....”  Well, actually, it was more like 50 years, because the earliest studies to connect smoking and lung cancer were conducted not by NIH-funded scientists but by Nazi scientists in the run-up to World War II.⁴  By the logic of the PNAS paper, then, ought we to be crediting the Nazi health science agenda with whatever progress has been made on reducing lung cancer, rather than the incredibly protracted and difficult public health campaign (that, for the most part, NIH had nothing to do with) aimed at getting people to cut down on smoking?

Obviously my point is not that NIH does not contribute to the nation’s health in important ways, but that the contribution—one of many, many variables—cannot, in theory or practice, be teased out by discovering correlations between budget trends and health trends.  This sort of analysis contributes to the notion that funding policy for NIH amounts to health policy for the nation.  We’ve already tried that trick.  After the failure of health care reform during the Clinton Administration, the government’s fall-back policy was to double NIH’s budget between 1998 and 2003.  Surprise: health care costs continued to skyrocket, millions of more people became disenfranchised from an ever-more-unaffordable health care system, and more and more municipalities and corporations began to sink under the mounting obligation of providing unaffordable health care for their employees and pensioners.  How much healthier might the nation have been if these trends had been reversed (even if NIH funding had stayed flat!)?

One final point:  Imagine a publication in a prestigious journal claiming that pharmaceutical company revenues were strongly correlated with positive public health outcomes—that the more drugs the companies sold, the healthier the nation became.  And imagine that the authors concluded, based on their analysis, that government policies should therefore encourage pharmaceutical profits, e.g., by extending patent lives or providing tax credits to the industry.  And now finally imagine that the authors of the paper acknowledged that their research had been supported by millions of dollars of research funding from the pharmaceutical industry.  Would this paper have any credibility?  Could it even be published?

The PNAS article recommends a ridiculous four-fold NIH budget increase over the next decade.  The article also includes, on the bottom of the first page, in small print, this statement:  “The authors declare no conflict of interest.”  Yet the first author of the paper was described in an August 21, 2002 New York Times article⁵ as “among the 10 biggest recipients of National Institutes of Health grants,” and the research reported in the PNAS article was also NIH supported.  What’s the difference between the hypothetical case and the real one?

About the Author:  Daniel Sarewitz is the co-director of CSPO.

¹ Comment offered by a colleague who, having yet to achieve tenure, prefers to remain anonymous (which in itself raises the obvious question of how the tenure process is protecting freedom of expression—but that’s another post).  I thank this same invisible person for help with this Soapbox post.

² Manton, K., Gu, X-L, Lowrimore, G., Ullian, A., and Tolley, H.D., 2009, “NIH funding trajectories and their correlations with U.S. health dynamics from 1950 to 2004,” PNAS 106(27): 10981-10986.

³ “Heading for the Emergency Room”, 2009, The Economist, June 25, p. 75.

⁴ Proctor, R.  The Nazi War on Cancer, 2000, Princeton, NJ:  Princeton University Press.

⁵ Zernike, K., 2002, “Duke Repays $700,000 in Grant Money and Reports a Swindle, The New York Times (August 21).