A Proposal to Make Biomedical Research into Alzheimer’s Disease More Democratic Following an International Survey with Researchers

Ethical approval

The project received ethical approval from the Research Ethics Community of Université Paris Descartes and the data analysis complied with the French Commission nationale de l’informatique et des libertés (CNIL) guidelines. All the data were anonymous and were analyzed in aggregate form. All of the raw survey data are available as Supplementary Material.

Survey design and promotion

Questions and responses comprised two categories: research (on theories and treatments of AD) versus personal (age, profession, country of primary affiliation, clinical versus academic researcher). All questions were optional and multiple choice, based on extensive literature review and consultation with colleagues. The participant filled out the form by following the URL to the Google Forms sheet.

We used Twitter (the account of The Alzheimer’s Association International Society to Advance Alzheimer’s Research and Treatment “@ISTAART,” and T.D.’s personal account, “@PhilAlz”) and a poster at The Alzheimer’s Association International Conference (AAIC) 2019 to promote the link to the Google Forms to recruit survey participants between January 1, 2019 and December 31, 2019. A second wave of promotion was undertaken in January 2021, but with less than 20 responders, thus only data from 2019 are analyzed herein.

Hypothesis testing and statistical analysis

Three hypotheses were tested concerning participant responses. Statistical analyses were performed using R version 3.6.1 (R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/). Participant characteristics were compared between the pro-ACH and non-ACH groups using Wilcoxon-Mann-Whitney test for continuous variables and Fisher’s exact test for categorical variables. Discrepancies in participant response numbers were due to the optional nature of every question.

Hypothesis 1: Pro-ACH responders will account for the majority of the participants

When participants were asked which categories of drugs gave them hope for a treatment of AD, if they answered “BACE inhibitors” and/or “Anti-Aβ antibodies” (which we combine in Results as “ACH drugs”) and also considered “Aβ physiology (production, clearance, etc.)” to be the number 1 priority in pre-clinical, early, or late stage AD, then they were considered to be “pro-ACH.”

Hypothesis 2: The ACH-group will have different constitutive characteristics as compared to the non-ACH group

In order to describe the profiles of pro-ACH and non-ACH survey participants, we performed the Classification And Regression Tree (CART) algorithm. The CART algorithm, also known as a “decision tree”, is a non-parametric supervised technique that combines variables in such a way as to best discriminate between two groups. We trained a decision tree of depth 5 through entropy minimization with characteristics such as age higher than 60; gender; country: USA; number of publications; clinical researcher (versus academic); key opinion leader (KOL); has received money from the pharmaceutical industry; whether or not the researcher thinks that “there is problematic adherence to the ACH from either industry, academia, associations or funding bodies”. We used the term adherence so as to insist upon the ACH’s ability to guide research.

Hypothesis 3: There will be gender differences in the popular vote towards treatments of AD

We investigated the top three therapeutic targets at pre-clinical, early-stage, and established AD according to popular vote of all the survey participants, pro-ACH and non-ACH taken together. Furthermore, if there are gender differences to be found in the pro-ACH/non-ACH groups, we might expect to find gender differences in the popular vote. Only participants identifying as M/F were included in the gender differences so as to use comparable group sizes for significance testing (n = 7 of “trans/prefer not to say/other”).

Study limitations

Firstly, 173 Twitter-using researchers represent a small minority of AD researchers (for example, AAIC in Los Angeles in 2019 alone counted 5,700 researchers). And this small sample may have been biased: only those with a strong opinion responding and giving theirs. Thus, the generalizability of our findings may be low. Forcing the research community into polarized groups (“pro-ACH” versus the rest) may not reflect the nuance in opinions that researchers have toward theories which can be studied thanks to other methods, such as bibliometrics [15]. This polarization is aggravated by the fact that quotes taken out of context from the scientific and lay literature (e.g., from Rudolph Tanzi) were used as sources of survey questions.

Concerning self-identification of individuals, our gender categories were highly limited, and our relatively small sample did not allow us to undertake statistical analysis on the contributions of non-traditional or non-conforming gender identities to the popular vote on treatments for AD. It is clear that there is need for greater work on “accountability, justice and representation” for gender minorities in STEM [16]. Furthermore, we did not ask questions on ethnicity, which other STEM researchers are indeed asking so as to “boost diversity in science” [17]. Finally, we did not offer an explicit definition of a “key opinion leader,” an ambiguous term whose value to these results is debatable because of the fact we let participants self-identify as KOL or not.

Moreover, as regards the ACH, just as our results suggest that there is some diversity of opinion within the pro-ACH group (e.g., their view of possible problematic adherence to the ACH), it is also clear that non-ACH opinions are not of one kind: some researchers are vocally in favor of “rejecting the amyloid cascade hypothesis” [14] and would be more aptly described as “anti-ACH.”

Finally, these results are time-sensitive: as different results from clinical trials and other studies are published, so do opinions change toward theories and treatments. The fact that our final round of survey promotion was unsuccessful warrants further analysis into researchers’ susceptibility to change their opinion on a scientific topic over a short period of time. The lower participation in 2021 could be due to current events in the field (see Conclusion), complications due to the Covid-19 pandemic, or simply the same participant population not wishing to undertake another similar survey. However, any such explanatory hypothesis would be highly speculative, and the issues being discussed in this paper (i.e., the possibility that there might be publishing and funding advantages of supporting the ACH) are worthy of further discussion and investigation. Limitations on the speed with which such research can be designed, ethically approved, undertaken, and published, should be taken into account in further studies with similar objectives.

A proposal to make biomedical research into AD more democratic

It is well-known that biomedical science, as a complex social activity, is guided by non-scientific factors, such as economic interests [18]. Reiss and Kitcher [19] argue that well-ordered biomedical science should follow the “fair-share principle,” where the amount of global funds spent on different diseases should be proportional to the suffering caused by them on a global scale. By analogy, we might ask: within the study of a single disease, how should resources best be dedicated to testing hypotheses and developing therapies based on them according to a “fair-share principle”? In other words, how can we make sure that promising theories of AD get their fair share of study and funding?

Solving this incredibly difficult problem is well beyond the scope of this article, but we will offer a sketch of a pro-democracy argument based on “crowd wisdom,” the empirical finding that informed collectives outperform individuals in estimating true values of different variables, before underlining two tragedies if the AD community does not succeed in organizing science better.

Kitcher [20, 21] argues in favor of a democratic deliberation process: taking the points of view of different segments of the community and attempting to guide research according to them. This does not have to mean a majority vote, but the phenomenon known as the wisdom of the crowd [22] suggests that the average value of multiple estimates tends to be more accurate than any one single estimate. Therefore, listening to the popular vote of researchers––at fora such as the yearly AAIC, and pooling a certain percentage of available funding towards the therapeutic leads suggested by popular vote––would mean drawing on many thousands of collective years of experience and perspective, which could lead to more accurate estimates of the causes of AD, and the best treatments to pursue. There is also increasing research being done with dementia patients in a co-research role in gerontology research (for example [23]); there is also much unexplored scope for including the patient community in deliberation processes concerning curative and preventive research into AD, and popular vote could also be used here.

The major idea defended here is that projects should be funded in a way that better represents the plurality of therapeutic leads offered by the research community. A yearly popular vote could be one step in that direction. But this leaves many questions open which we cannot definitively answer in one article, including, but not limited to the following:

  How could we improve representation on funding bodies and editorial boards, including a role for the patient community?

  Upon what kinds of evidence should publication and funding decisions be based so that both scientific pluralism and plausibility are guaranteed in AD research according to a fair-share principle?

  What kind of funding model would be most suited to a more democratic approach: private and/or public ventures?

  Are there some domains and methods within biomedical science which might be particularly under the influence of monopolized ways of thinking? (e.g., at the level of pre-clinical or clinical research?) Could publication and funding quotas be used to make monopolized domains more inclusive?

  How, and to what extent, could the themes of calls for contributions and projects by publishers and funders be broadened on a long-term, community-wide scale?

Furthermore, a more democratic model itself would not be perfect, particularly if it were taken to the extreme of eroding individual expertise, which is and should remain a cornerstone of rigorous science. Instead, the model we propose serves to reduce monopoly, and thereby take any possible institutionalized brakes off the contributions of individual scientists.

In any case, if research cannot become better organized, we anticipate two major tragedies. The first concerns the survival of the fragile economic model underlying therapeutic research into AD, without which patients will never receive disease-modifying treatment. The second concerns science itself.

Indeed, the current high-risk model encourages the opposition between patient need and return on investment for innovators. Bringing an AD drug to market is estimated to cost $5.6 billion [24]. The developer of the first monopolized disease-modifying treatment of AD would stand to gain an astronomical return on their major investment. Conversely, when a clinical trial of a much-anticipated AD treatment fails, the market value of the pharmaceutical company that developed it loses as much as 40% overnight, as in the case of Eli Lilly and solanezumab [13]. In January 2018, the pharmaceutical company Pfizer decided to stop its research on AD and Parkinson’s disease by laying off 300 researchers due to numerous drug failures amid a dismal context for research on neurodegenerative diseases: pulling out was part of “an exercise to reallocate spend across our portfolio,” according to the company [25]. It is not clear what the future of AD research looks like, but it is fragile and, in its current state, mostly dependent on amyloid being a viable target, with millions of patients and families living in hope. By ensuring that other promising theories are funded, at a community-wide level, it would allow bets to be hedged against the possibility that the ACH does not deliver on its promises.

The second tragedy, done against science itself and those individuals who defend it, is “epistemic injustice,” a term coined by philosopher Miranda Fricker as “wrong done to someone specifically in their capacity as a knower” [26]. Fricker draws on examples from literature and history where factors such as race and gender have led to points of view being ignored and condemned. Within AD research, there are surely examples of intellectually honest researchers defending controversial hypotheses of AD who have struggled to get data published, receive funding, and retain their place within academia. In other words, certain hypotheses might be rejected not because of scientific argument but rather the social structure of the field of biomedical research. We finish by noting that the results from our small sample tentatively suggest that the majority of women do not support the ACH and may therefore be particularly vulnerable to the negative consequences of a community gatekeeping bias. Taking the example of hypotheses concerning the role of microbes in AD [3], Fig. 1 from Liu et al. (2019) showed that “ ...  up to 2019  ...   0.5% of trials tested the virus hypothesis” [1]. Concerning this “fringe theory  ...   now, researchers are taking it seriously” [27], but the fact that up to 2019 only 1 in 200 clinical trials were dedicated to testing a direct viral contribution to AD, does beg the question: are theories of AD being funded according to fair-share principle? Ruth Itzhaki, first author on the previously cited [3] editorial on microbes in AD, has described “a series of battles  ...   awful problems getting [research] published” [28]. This example does point to the possibility of epistemic injustice in AD research and suggests the existence of perspectives whose contribution to improving the lives of AD patients has not yet been fully taken into account. This seems like community-wide oversight, since the perspectives of marginalized individuals in institutionalized social structures may offer particularly insightful contributions to research, since they may recognize patterns in the world that those in more dominant groups may be blinded to [29].