NIH funding trajectories and their correlations with US health dynamics from 1950 to 2004

Results

The temporal relation of NIH funding and age adjusted death rates for each of the 4 causes is in Fig. 1. We used inflation adjusted annual expenditures for all NIH institutes beginning in 1938 with the creation of NCI. The national health trend data are rates (age-standardized to 2000) for total mortality and mortality specific to heart disease, stroke, cancer, and diabetes, for 1938 to 2004. Additionally, we examined national chronic disability trends from the 1982 to 2004 NLTCS, BLS labor force participation rates 1976 to 2006, and individually linked individual Medicare records 1982 to 2006 (1, 3).

Declines in heart disease and stroke began after large real increases in the NIH budget from 1956 to 1967. Total cancer mortality rates did not decline until 1990, 25 years after identification of the effect of smoking on lung and other cancers and 18 years after the start of the War on Cancer. The diabetes pattern is obscured by the health effects of increases in obesity prevalence, and changes in diabetes diagnostic and screening practices (18).

There is evidence of lagged negative correlations between the trajectory of NIH (and institute specific) funding and 3 of 4 causes of death representing 2/3 of US mortality 1950 to 2006. Dependent competing risks (19) and physiologic correlations of the 4 diseases exist at the individual level. Some increase in diabetes mortality may be due to biomedical and clinical advances preventing lethal circulatory events and delaying death for the individual to later ages where there is an increased direct mortality risk of diabetes.

In Fig. 2 we relate institute specific funding to each of 4 mortality trends. Fig. 2 A and B, for heart disease and stroke, are similar with large declines beginning in the late 1960s. To provide substance to the time lines the numerals indicate select research and therapeutic milestones.

Selected Milestones (Fig. 2A).

1. NHLBI founded (1948).

2. Framingham Study started (1950).

3. Male CVD mortality starts to decline (1968).

4. Innovation of angioplastic procedures (1977).

5. Improved treatment of congestive heart failure, for example, ACE inhibitors (1983).

6. Arterial stents (1986).

7. Introduction of tPA and streptokinase (1987).

8. Drug-eluting stents (2003).

9. Progenitor cell therapy (2004).

Selected Milestones (Fig. 2B).

1. Start of NINDS (1954).

2. Ultrasound use in medical diagnosis (1956).

3. Recognition of role of carotid atherosclerosis in stroke (1960).

4. CT scanning of strokes (1978).

5. Use of ultrasound to diagnosis arterial wall thickness (1986).

6. Platelet coagulation inhibitors (1995).

7. Cerebral thrombus dissolution using tPA and streptokinase (1996).

8. Identification of stem cell and neuronal replacement in brain (1999).

9. Approval of altace (ACE inhibitor) in hypertension control to reduce strokes (2000).

10. Use of MRI in stroke diagnosis (2002).

11. Use of warfarin in severe stroke (2006).

For NCI (Fig. 2C) the funding spike occurred in 1976. Declines in cancer mortality started after in 1990 and continued to 2006.

Selected Milestones (Fig. 2C).

1. Recognition nitrogen mustards could be used to treat cancer.

2. Folic acid inhibitors identified for ALL therapy (1950).

3. National Cancer Chemotherapy Service Center (1955).

4. Surgeon General reports on smoking (1962).

5. Adjuvant therapy (osteosarcoma) (1974).

6. Angiogenesis identified (1974) (17).

7. New cell lines used for drug screening (1986).

8. Cell signaling targets identified (1986).

9. Taxol effective in ovarian cancer (1987).

10. Amelioration of bone marrow toxicity (GM-CSF; G-CSF; 1991).

11. Cervical cancer (HPV) vaccine (2005).

For diabetes, after a post-World War II drop in mortality, there is an irregular pattern in which health improvements due to improved management of diabetes and related risk factors was hidden by surges in US obesity prevalence (2) and improved screening of previously undiagnosed adult diabetes (18). Diabetes mortality declined 1970 to 1990 due to reduction in co-morbid circulatory diseases and increased 1990 to 2005.

Selected Milestones (Fig. 2D).

1. Insulin discovered (1921).

2. Oral drugs to lower blood glucose (1955).

3. Two types of diabetes recognized (1959).

4. First pancreas transplant (1966).

5. Insulin pumps (1970).

6. Recombinant insulin (1978).

7. DCCT Report on stringent insulin control (1993).

8. Genetic engineering of insulin producing cells (2004).

9. Repair of insulin-producing islet cells (2007).

Although cause-specific mortality is an index of efficacy in changing the risk of specific diseases, the economic impact of cause-specific changes can also be summarized by their effect on the function of individuals at late ages where chronic disease has its greatest impact. This is relevant to human capital and may be a better index of health program efficacy in economically developed nations [e.g., measures of active life expectancy (20, 21)].

Estimates of the rate of disability changes before 1980 are from analyses of Civil War veterans and recruits by Fogel and Costa (13) that suggest a long term (from the end of the Civil War to the 1980s), slow decline in chronic disease and disability of 0.6% per annum occurred. Much of this improvement was due to health interventions at earlier ages which reduced individual chronic disease risks (e.g., preventing rheumatic heart disease at young ages prevented damage to the myocardium that would have produced congestive heart disease in survivors at later ages).

NIH funding did not manifest significant correlations with declines in the prevalence of disability in the elderly until 1984–1989—16 years after CVD and stroke mortality began to decline. For circulatory diseases, specific clinical and public health innovations have been shown to have large effects on mortality and disability and to be cost-effective (22). Labor force participation rates of persons 65+ (23) began to increase in the mid 1980s, for example, rates above 65 increased 1.5% per annum 1986 to 1996 and 3% per annum 1996 to 2006. Pressures to delay retirement will increase due to changes in pension programs and recent adverse economic events, such as the loss of equity in the US housing market (24).

More detailed analyses of US health changes would examine effects of biomedical interventions on intra-individual processes determining the progression of physiologic decline to frank morbidity, disability and death. This suggests the impact of the age-specific schedule of interventions should be studied (some diseases having higher risks in certain age ranges) as should diseases (and conditions like obesity) with significant functional impairment (e.g., osteoarthritis and peripheral neurological consequences of diabetes) but small, not immediate, mortality risk.

Many health policy, economic and aging researchers initially rejected the concept that functioning at later ages (e.g., 65+) could be significantly improved. In 1980, there was little US population evidence to indicate the age at onset of chronic disability increased in parallel to increases in life expectancy (25). The Greenspan Commission suggested, in 1982, unanticipated increases in US male life expectancy 1969 to 1980, be adjusted by an increase in the normal retirement age from 65 to 67. The 1982 to 2004 NLTCS and linked Medicare records are currently the primary longitudinal data source for documenting chronic disability declines in the elderly (3). Britain (to 69 years) and Japan (to 74 years) are now considering larger increases in the retirement age to respond to increased longevity. Manton et al. (3) found much of the recent increase in longevity is spent in physically active states.

Analysis of US labor force dynamics (23, 26) suggests that, for US economic growth to continue, improvements in human capital at late ages is not only desirable, but necessary (1). Increased investment in health research and improvement is necessary to enhance the human capital infrastructure necessary for current economic stimulus plans to succeed. This is because the rate of growth of the US labor force is projected to decrease, for example, from 1.2% 1996 to 2006 to 0.8% from 2006 to 2016. In addition the length of training for high technology occupations is increasing, and rapid rates of technological innovation require re-training for elements of the labor pool to switch to new high technology occupations. Such processes will affect economic growth in the US and its global competitiveness (4).

In addition to positive NIH budget growth effects on human capital the rate of increase in per capita Medicare expenditures may moderate as individual health is improved at later ages [e.g., ages 65 and 85 (3)]. The correlation of per capita NIH expenditures for Medicare enrollees with per capita Medicare expenditures (both in 1966 dollars) shows that the doubling of the NIH budget 1998 to 2003 is associated with declines in Medicare per capita expenditures 1998 to 2000 with little increase 2000 to 2004. Manton et al. (1) showed per capita, per annum Medicare cost decreases occurred after 1980 in the non-disabled US elderly population. Starting in 2004, the NIH budget has not kept pace with inflation.

In Figs. 2 A–D it is hard to discern the correlation of NIH expenditures with mortality declines. To make this transparent, in Figs. 3 A–D, this relation is plotted, after filtering for inflation, with investment averaged over a 10-year window and rates age-standardized to remove confounding effects of changes in population age structure. In Fig. 3 A–D we present the correlation of the age-adjusted mortality rate to 10-year aggregated institute-specific inflation adjusted budgets for 4 chronic diseases. To help relate this nonlinear correlation to calendar time we include dates for specific events on the trajectories Since expenditures are inflation adjusted, trajectories can “fold back” (i.e., “real” expenditures may decrease). A 10-year window was used to aggregate expenditures because research expenditures accumulated over such a time window better predict scientific advances than concurrent single year budgets. The 10-year window was selected based on the typical time (10 to 15 years) between major shifts in institute budgets and the emergence of health effects for specific diseases.

Three causes show an acceleration of mortality declines after passing a threshold 10-year budget level. Figs. 3 A and B show circulatory disease mortality was responsive to investment to establish a disease specific biotechnology, clinical, and basic science base. Instead of decreasing returns, disease specific benefits accrued at an increasing rate once the threshold for a disease was established. In Fig. 3C, despite increases in the inflation-adjusted NCI budget to 1986, cancer mortality rates increased. The NCI budget decline 1986 to 1990 was concurrent with continuing increases in cancer mortality. Despite continuing declines in inflation adjusted expenditures to 1995, cancer mortality stated to decline in 1990.

For diabetes (Fig. 3D), we may not have yet reached the level of scientific understanding necessary for rapid mortality reductions—as distinct from modification of the risk of related co-morbidity. Funding for diabetes (NIDDK) followed a trend similar to that for both CVD (NHLBI) and stroke (NINDS). Anticipating the same lag as for CVD and stroke, one could expect a downturn in diabetes mortality. Although a decline is evident 1970 to 1990, thereafter diabetes mortality increased. This may be because the most common risk factor for diabetes is obesity as measured by high body mass index (BMI). Diabetes is 3 times more prevalent for individuals with a BMI of 35+ relative to those with BMIs below 30. For BMIs of 30 to 35, there is a 50% to 100% increased prevalence relative to BMIs below 30 (see 27). In surveys (28) obesity was stable, 1960 to 1980, for females aged 60 to 74 and for the total population aged 20+. After 1980, obesity increased.

Although the age adjusted prevalence of obesity and diabetes increased, treatment and health outcomes for diabetic individuals improved significantly (29, 30), delaying population mortality increases. Analyses of mortality from CVD and stroke show a lag of 10 to 15 years from the time research funding increased until a mortality effect was observed. For both CVD and stroke, in the first decade after the effect was visible, mortality declined 2.5% to 3% per annum. Increases in obesity starting in 1980 appear mitigated by research, delaying diabetes mortality increases until 1990, that is, the 4.5% per annum increase in obesity prevalence is offset by an 2.5% decrease in mortality. After 1990, diabetes mortality should increase 2% per annum due to obesity. Fig. 3D shows a 2.5% per annum increase in mortality 1990 to 2004. Had obesity prevalence remained constant (a counter factual case) diabetes mortality would have dropped from 17 per 100,000 in 1980 to 9 in 2004—a proportionate decrease similar to that observed for CVD and stroke. The rise in obesity may reflect adverse effects of nutritional changes that earlier fueled many dimensions of positive population health gains, that is, in the competition between obesity increases, and improved clinical control of circulatory risk factors, biomedical research advances dominated health risk trends.