The National Institutes of Health (NIH) has recommended that research effort be compared to the societal burden of that disease [2]. Our analysis indicates that the amount of NIH funding for research on a disease is associated with the burden of the disease, but discrepancies exist. Different measures of the burden of disease yielded different conclusions about the degree to which some cancer types were over or underfunded; these general conclusions are in agreement with the study by Gross et al. [2] from 1999 but also indicate that more than a decade later these discrepancies persist.
For example, Figure 1 shows that stomach cancer receives the least amount of funding relative to its burden; stomach cancer receives less than 10% the amount of funding, per death and YLL, of breast cancer. Similarly, research funding to treat uterine cancer, per death and YLL, is 10% the funding given to treat testes cancer.
Figures 1, 2, 3, 4 and 6, 7, 8 allow us to identify the following cancers as funded at levels far below others which incur similar costs to society: bladder, oral, uterine, and stomach. A degree of underfunding also applies to esophageal, liver, and pancreatic cancers. Given their high YYLPI values, they represent promising areas for potential gain compared to cancers that have already achieved low mortality rates. Conversely, research on breast cancer, leukemia, and prostate cancer appear to be higher than justified relative to their burden on society as measured by YLL and economic costs. In summary, based on our data, we recommend that funding resources be directed away from breast cancer, prostate cancer, and leukemia toward bladder, esophageal, liver, oral, pancreatic, uterine, and stomach cancers.
To illustrate the type of redistribution we suggest, consider Figure 2. Reducing the funding for leukemia research to an equitable level with regard to YLL (from 11.70% of funding to 6.81%) would free more than enough resources to raise bladder and uterine cancer funding to equitable levels with regard to YLL (0.9% to 2.96% and 0.56% to 2.35% respectively), therefore creating a better overall distribution of funding relative to the societal burden measured by YLL. As a second example, plots 2 and 3 show that reducing breast cancer funding alone to an equitable level would provide more funding that could be used to raise several of these underfunded cancers up to parity.
The use of DALYs has received much attention in the prioritization of funding and studies have indicated that funding is more tightly correlated with DALY values than with incidence, mortality or YLL values [5]. Figure 3 illustrates a similar pattern of funding distribution relative to the different cancer types as seen in Figure 2 with the exception that liver cancer appears more equitably funded when using DALY values instead of YLL values. Use of DALY values is controversial for two main reasons [25, 26]. First, the weights used are determined by indirect methods rather than methods measuring impact of the disabilities directly. Second, the estimation and true cost of disability may vary in different regions and within different subpopulations. Wealthier regions and populations are better equipped to accurately estimate the factors that go into the DALY weighting values and may weight certain behavioral factors differently than less wealthy groups. The conclusions of the analyses using both the YLL and DALY values are similar therefore the controversial aspects of DALY valuations do not pose a problem for our overall recommendations.
We believe that considering the "Years of Life Lost Per Incidence" (YLLPI) is useful for making research decisions. Research to treat cancer is often concerned with improving the outcome of people afflicted. With the exception for special circumstances like the HPV vaccine (which reduces the risk of cervical cancer), most NCI funded research is aimed at reducing mortality instead of incidence. If the number of cases is mainly due to factors like behavior and genetics that are hard to influence, plots of research funding vs. YLLPI (Figure 4) may provide the clearest identification of situations where increased funding has the best potential to improve treatment. This comparison does not include the number of individuals that may be affected by improved treatment, but does estimate the "room for improvement" per treatment. Cancers that kill younger individuals and have high mortality rates tend to have a higher YLLPI. Marginal improvement in treatment for these conditions may be a more realistic goal and provide the same or greater benefit than improvement in conditions with low YLLPI because they already have low mortality rates or tend to affect the elderly.
The economic costs of cancer are also part of the overall consideration of research funding allocation. In Figures 6, 7, 8 we compare funding relative to several economic cost metrics and funding relative to YLL for each cancer. The first two of these metrics (Medicare payments and estimated medical care) measure direct medical care costs. The third metric (lost productivity) attempts to make a more comprehensive estimate of all costs associated with cancer. Cancers in the lower left of these plots appear to be underfunded according to both the economic and YLL metrics. In addition to arguments about reducing suffering, purely economic arguments would favor increasing funding for cancers in the lower left at the expense of decreased funding for cancers not in these regions. This analysis indicates that the most efficient changes in research allocation, from a purely financial perspective, involve reducing funding for leukemia, melanoma, breast, and prostate cancers while increasing funding for bladder, stomach, and uterine cancers.
Comparison of our results with previous similar studies reveals some similar conclusions and some differences.
Incidence and mortality values were used in a Canadian study by Branton [15] and in a strategic analysis document published by the UK National Cancer Research Institute [27]. For the Canadian data, colorectal and lung cancers were identified as underfunded relative to other cancers. The UK analysis identified leukemia, ovarian and cervical cancers as overfunded whereas lung, pancreatic, stomach, esophageal and bladder cancers were identified as underfunded. These results generally agree with ours which may reflect similar approaches to research science in the US, UK and Canada.
Years of Life Lost values were used in studies of Australian data. An analysis of overall Australian funding per YLL showed that breast, cervical, leukemia, melanoma and prostate cancers were funded at far higher levels than provided for bladder, brain, gall bladder, lung/mesothelioma, kidney, lymphoma and pancreatic cancers [10]. Interestingly a separate analysis of Western Australia showed similar results with the exception that a large number of studies of mesothelioma were funded [11], likely due to the history of asbestos mining in the region.
Burnet et al. [8] performed a study of cancer funding and societal burden as measured by YLL statistics Using data from the UK (life expectancy data from 1990 life tables and funding data from a 2002) to identify apparent cases of funding discrepancy.
In their AYLL ("Average Years of Life Lost" which is YLL per death) analysis they describe a "Cinderella" region with high AYLL and low funding to identify cancers they believe deserve additional funds. However, focusing on AYLL ignores the total number of cases and the per incidence mortality rate in the burden completely. For example, using their method, a cancer that kills a single child each year would be placed within their "Cinderella" region and be recommended for more funding. This approach would identify testes cancer as greatly underfunded even though it has the lowest rate of incidence and the lowest rate of mortality per incidence of the cancers we examined, resulting in only 350 deaths in 2010. By focusing on AYLL alone, two of the four cancer types they identify as underfunded (cervical cancer and melanoma) are misidentified and appear overfunded when using YLL as a metric. Furthermore, the six types of cancer we identify as underfunded do not appear in their "Cinderella" region of underfunded cancer types.
In their second analysis, comparing % overall funding relative to % overall YLL, Burnet et al. [8] addresses these issues and identifies several examples of funding discrepancies. Figure 9 presents our values of %funding/%YLL and those calculated using data in table 2 of Burnet et al. [8]. The general agreement of funding levels for different cancers (R2 = 0.57 for the data shown in Figure 9) based on two distinct data sets (older UK vs more recent US data) suggests that the conclusions we make from this data are likely to be robust. A number of cancers (bladder, esophageal, lung, pancreatic, stomach, and uterine) are identified as underfunded in both studies.
Various societal and cancer specific factors may account for some of the discrepancies observed.
Consideration of the data and discussions with medical professionals lead us to believe that the relatively high level of funding for breast cancer is due to the organized efforts of women’s groups and charitable organizations to raise awareness and concern about the burden caused by this cancer. Similar efforts by women's groups may account for the relatively high level of funding for cervical cancer. Given this, we are unsure why uterine cancer exhibits lower than expected rates of funding.
Funding for lung cancer is quite low given its cost, mainly due to a "blame the victim" attitude in which the personal choice to smoke is seen as the direct cause [15]. The levels of funding for liver cancer (2.6% of funding compared to 3.8% of deaths and 3.8% of YLL) and oral cancer (0.5% of funding compared to 1.6% of deaths and 1.8% of YLL) may also be influenced by this “blame the victim” prejudice (Hepatitus B infection and alcoholism contribute to liver cancer risk and chewing tobacco contributes to oral cancer risk). However, this rationale does not appear to be consistently applied. Cervical cancer is largely attributable to the HPV virus and the risk of skin cancer can be greatly reduced by the application of sunblock; in both cases, cancer patients are not typically blamed for their personal actions. In fact, the per-death research funding for cervical and skin cancers are both in the top 6 of our 21 examined cancers (Figure 1b) and cervical cancer has the highest per-mortality research spending rate, indicating a prioritization of research. Morally, if we decide that it is appropriate for cancer victims to receive blame for cancers that result as a consequence of their actions, funding for cervical and skin cancers should be reduced in lieu of increased funding for other cancers. Alternatively, if we decide that it is morally unacceptable to place blame on cancer victims for poor lifestyle choices, funding for lung, liver, and oral cancers should be increased to levels comparable to other cancers relative to their burden.
Although we examined the major source of research funding provided by government, our study did not examine the research effort conducted by commercial entities and is therefore a partial view of the overall funding landscape. The main reason we omitted research funding by private entities is because the reallocation decisions we suggest in this paper are guided by the goal of reducing societal burden; private research entities are motivated by financial concerns and the burdens analysed in this paper are therefore less relevant for their decision making processes. An interesting point arises however when considering the approach taken by private entities. Private research will tend to focus on the types of cancer with the largest potential profit and private funding will therefore be higher for more common cancers which create a larger market. Several of the cancers we identified as overfunded are just these more common types (breast, prostate, leukemia) and some of the ones we identified as underfunded are relatively rare (oral, uterine, stomach) and we expect that market forces will tend to exacerbate these discrepancies in the private sector. For this reason we feel that governmental priorities should lean on the side of favoring research for cancers that are more rare when societal burden is similar.
Another factor that may account for the degree of overfunding for certain cancers may be the existence positive feedback cycles. As progress is made on one type of cancer, successful researchers publish papers that attract new funds and inspire students to follow their footsteps. This progression leads to new discoveries and the preparation of more sophisticated proposals and studies [28]. For example, consider two hypothetical research communities where the first develops cancer treatments leading to high cure rates (e.g., 99%) and the second develops fewer treatments with lower cure rates (e.g., 40%). Clearly the potential for improvement is higher in the second community. However, the success rate of the first community may outshine the second and attract more funding. This misallocation of resources may have occurred in the case of leukemia research. Branton [15] notes that hematopoietic cancers attracted considerable attention and funding for treatment. These cancers now have among the highest survival rates. If the societal dynamic we just described is a potential cause of the non-equitable distribution of research funding, specific efforts to direct researchers away from highly studied cancers to less studied ones may be warranted.
Conversely, targeted focus on specific cancers may be beneficial depending on the nature of the relationship between research effort and results. In our study we assume a linear relationship between effort and result, but other relationships may exist (Figure 10). For example, if research efforts have a positive synergistic effect (where each additional dollar leads to a magnified result), the optimal strategy would be to concentrate funding in fewer cancers and shift the focus as each cancer gains improved treatments. Positive synergistic activities can come from increasing the diversity of research plans and funding high-risk high reward approaches that would otherwise not be tried. On the other hand, if a negative synergistic effect occurs where each additional dollar spent leads to proportionally smaller results, the optimal strategy would be to spread funding across cancers more equitably, regardless of societal burden. Negative synergistic effects can come from additional funding being allocated to projects that are similar to previous ones and offer marginal novelty or potential for benefit. In light of this, examining the relationship between funding and results to identify the type of synergistic relationship in practice may prove highly beneficial.
Currently, the NIH research funding decision process has two stages. The initial stage is the scientific review of the proposed study addressing significance, approach, and innovation. This stage generates ranking based on purely scientific factors. The second stage considers additional criteria, including public health priorities such as quality of life or health disparities. We believe studies like ours can enhance the second stage of this process by recognizing cancers that deserve more attention and improving the allocation of limited research resources.
Many may feel that reductionism to numbers in the realms of human suffering and death is cold and heartless and a number of researchers have discussed controversies associated with setting funding priorities based on measures of burden [5–7, 25, 26]. We strongly believe that funding decisions should take burden into account; since total funding is limited, only by objectively identifying disorders that receive non-optimal funding can we adjust our efforts to minimize the overall cost of disease. Our approach has led us to identify breast cancer, prostate cancer, and leukemia as funded at levels higher than other cancers relative to their societal burden and bladder, esophageal, liver, oral, pancreatic, stomach, and uterine cancers as relatively underfunded.