Unintentional discrimination in clinical research: Why the small decisions matter

by Arthur T. Ryan, M.A. and Elaine F. Walker, Ph.D.



Arthur Ryan is a graduate student in clinical psychology at Emory University. His research focuses on understanding the etiology and neuropathology underlying severe mental illness.



Elaine Walker is a Professor of Psychology and Neuroscience in the Department of Psychology at Emory University and is the Director of the Development and Mental Health Research Program, which is supported by the National Institute of Mental Health. Her research is focused on child and adolescent development and the brain changes that are associated with adolescence. She is also a member of the AJOB Neuroscience editorial board.

Arthur Ryan, M.A.

Over the past several decades, there has been a significant effort to minimize bias against individuals based on ethnicity and other demographic factors through the creation of seemingly impartial and objective criteria across a host of domains. For example, when the United States Federal Sentencing Guidelines were created in the 1980’s, one of their primary goals was to alleviate “...unwarranted disparity among offenders with similar characteristics convicted of similar criminal conduct” [1]. Unfortunately, even well-intentioned efforts such as this one can still have a disparate negative impact upon historically marginalized groups, such as the well-documented disproportionate sentencing of black individuals due to differing rules governing offenses committed with crack vs. powdered cocaine [2]. Concerns about such inadvertent bias are not limited to the legal domain. Agencies that fund clinical investigations are paying greater attention to demographic representativeness and access to participation in health-related research.

Let us consider a hypothetical example, drawn from the authors’ own field of research in a US context, of how seemingly objective research design choices can results in biases in access to research participation. There is increasing evidence that inflammatory factors may play an important role in the etiology of schizophrenia and other psychotic illnesses [3]. One thing researchers do when attempting to understand a complex system like the human brain is to minimize external sources of variance. One readily identifiable correlate of inflammatory markers is body mass index (BMI) [4]. Schizophrenia itself is also correlated with BMI, such that patients tend to have a higher BMI than healthy individuals [5]. So a hypothetical researcher might reasonably say to herself, “Let me compare inflammatory markers in individuals with and without schizophrenia who have a BMI below 25 (BMIs of 25 or greater are considered to be medically overweight). That way, if I find a difference between the groups, I can more strongly conclude that the difference has to do with schizophrenia’s underlying pathology and was not due to individuals with schizophrenia being medically overweight.”

Elaine Walker, Ph.D.

Our hypothetical researcher’s experimental design choice is scientifically defensible and seemingly innocuous, but her decision may have unintended negative consequences. In a study of patients at the Grady Medical Clinic, an Atlanta primary care clinic serving inner city residents, 80% of black American women had a BMI of ≥ 25 [6]. Of particular note, the Grady Health System and its patients regularly participate in mental health research, including our own clinical studies, and the vast majority of those patients— around 85% [7]— are African American. So if our hypothetical researcher had unknowingly gone ahead with her BMI exclusion criterion and was recruiting from a similar population, she would effectively be excluding four out of five black American women from participating in her research, despite the disproportionately high number of black patients served. This would be more than a minor unfairness or lost opportunity for the individual women who could not participate: given the growing research literature showing that various biological and genetic risk markers have differential predictive utility across racial and ethnic groups, this hypothetical study might produce findings that are invalid for minority individuals. Because rates of schizophrenia seem to be similar across various racial groups and nationalities when measured by well-controlled studies, such an omission is not acceptable. If the study yielded results valid largely for one racial group, our hypothetical researcher would also be compromising one of her ethical obligations described under the Belmont Report, which requires that researchers avoid creating unjust patterns in the “...overall distribution of the burdens and benefits of research” [8]. The report explicitly extends this principle to research involving racial minorities and other historically exploited groups.

To reiterate, the preceding example was hypothetical and no such exclusionary criterion was employed in our own work. However, the possibility of research design choices having a discriminatory impact is no hypothetical hazard. In a 2006 review of randomized controlled drug trials, only 24% of participants were women [9], while a 2008 review of trials funded by the National Heart, Lung, and Blood Institute showed a mean female participation rate of 27% [10]. Even non-human females are underrepresented in research, with male-animal-only studies outnumbering female-animal-only studies at a roughly 5 : 1 ratio in neuroscience and pharmacology [11]. Such exclusion is even more notable considering that some conditions may be experienced at higher rates by those left out of such trials. For example, major depression affects women at an approximately 2:1 ratio [12]. So, theoretically, it would be particularly egregious to exclude women from antidepressant medication trials—and any research that sought to create a representative sample should include women at that same 2:1 ratio.

So why do male-only studies still predominate? Because it is cheaper and easier to conduct male-only studies. With less variance among individuals, experimental effects are easier to detect. In addition, male hormones fluctuate less over time and including women in drug trials necessitates extra experimental protections to prevent harm should one of the participants become pregnant. Again, this practice is not simply an unfairness to the individual women who would otherwise want to participate in medical research. FDA studies have shown that drug concentrations in blood and tissue can vary by as much as 40% between men and women, with similar variations in side effect profiles [13]. This shouldn’t be surprising as gene expression may vary between males and females by more than 50% in liver, fat, and muscle tissue [14]. It is clear that such a systematic policy of excluding women from research, even if it lacked any conscious discriminatory intent, could have serious, even life-threatening, consequences for women receiving medical treatment for years to come.

There is no way to completely prevent unintentional discriminatory sequelae of research design choices. And it would be naïve to believe that there will never be genuine tradeoffs that need to be weighed when designing research studies. Sometimes researchers will need to decide between sample representativeness and experimental control. The important point here is that if researchers make their decisions in a reflective and intentional manner, always considering the downstream consequences of their study design choices, they are more likely to identify and mitigate secondary negative consequences of their work. In doing so, they are increasing the scientific value of their work, as well as fulfilling their ethical obligations to promote beneficence and justice with their research.

References



1. United States Sentencing Commission. An Overview of the United States Sentencing Commission.

2. NPR. High Court Rules on Drug Sentencing Disparities.

3. Miller, B. J., Buckley, P., Seabolt, W., Mellor, A. & Kirkpatrick, B. Meta-Analysis of Cytokine Alterations in Schizophrenia: Clinical Status and Antipsychotic Effects. Biol. Psychiatry 70, 663–671 (2011).

4. Festa, A. et al. The relation of body fat mass and distribution to markers of chronic inflammation. Int. J. Obes. 25, 1407–1415 (2001).

5. Homel, P., Casey, D. & Allison, D. B. Changes in body mass index for individuals with and without schizophrenia, 1987–1996. Schizophr. Res. 55, 277–284 (2002).

6. Jacobson, T. A., Morton, F., Jacobson, K. L., Sharma, S. & Garcia, D. C. An assessment of obesity among African-American women in an inner city primary care clinic. J. Natl. Med. Assoc. 94, 1049–1057 (2002).

7. Saunders, S. P. & Campbell, C. L. The Word on the Street: Performing the Scriptures in the Urban Context., (Wipf and Stock Publishers, pp.23, 2006).

8. The National Commission for the Protection of Human Subjects of Biomedical and Behavioral Research. The Belmont Report: Ethical principles and guidelines for the protection of human subjects of research. (1979)

9. Geller, S. E., Adams, M. G. & Carnes, M. Adherence to Federal Guidelines for Reporting of Sex and Race/Ethnicity in Clinical Trials. J. Womens Health 15, 1123–1131 (2006).

10. Kim, E. S. H., Carrigan, T. P. & Menon, V. Enrollment of Women in National Heart, Lung, and Blood Institute-Funded Cardiovascular Randomized Controlled Trials Fails to Meet Current Federal Mandates for Inclusion. J. Am. Coll. Cardiol. 52, 672–673 (2008).

11. Beery, A. K. & Zucker, I. Sex Bias in Neuroscience and Biomedical Research. Neurosci. Biobehav. Rev. 35, 565–572 (2011).

12. Kessler, R. C. Epidemiology of women and depression. J. Affect. Disord. 74, 5–13 (2003).

13. Anderson, G. D. Sex and racial differences in pharmacological response: where is the evidence? Pharmacogenetics, pharmacokinetics, and pharmacodynamics. J. Womens Health 2002 14, 19–29 (2005).

14. Yang, X. et al. Tissue-specific expression and regulation of sexually dimorphic genes in mice. Genome Res. 16, 995–1004 (2006).





Want to cite this post?

Ryan, A and Walker, E. (2015). Unintentional discrimination in clinical research: Why the small decisions matter. The Neuroethics Blog. Retrieved on , from http://www.theneuroethicsblog.com/2015/09/unintentional-discrimination-in.html.

Experimental Neuroethics

By Peter Reiner, VMD, PhD



Dr. Reiner is Professor in the National Core for Neuroethics, a member of the Kinsmen Laboratory of Neurological Research, Department of Psychiatry and the Brain Research Centre at the University of British Columbia, and a member of the AJOB Neuroscience Editorial Board.



Four years ago, Neil Levy gave the concluding lecture at the first Brain
Matters conference in Halifax. He alerted the audience of neuroethicists to the
fact that the field of philosophy was undergoing a revolution – rather than muse
from their armchairs in the ivory tower, a group of renegade philosophers were
carrying out real experiments, asking people what their intuitions were about central
issues in philosophy. Dubbed experimental
philosophy, the new initiative was met with more than passing resistance
from traditional philosophers. The apostate experimental philosophers responded
by developing a logo of a burning armchair.

Photo credit: Timothy Epp, Shutterstock

The landmark experiment was carried out by Josh Knobe, and its findings
subsequently became known as the Knobe effect (you can watch a great recreation
of the phenomenon in this
YouTube video). Essentially, what Josh did was repurpose an old method from
social psychology called the contrastive vignette technique (CVT)¹. At its simplest, the CVT
involves designing a pair of vignettes that carefully describe a particular situation
(in the case of experimental philosophy, one that is often morally charged) but
crucially differ in one detail, hence the term contrastive. Respondents see one
and only one version of the vignette, and are then asked questions about what
they have just read, with responses commonly recorded as a numerical rating on
a Likert scale. By comparing the averaged responses between separate groups of
people who have read the vignettes, the experimenter can systematically investigate
the effects of small changes (of which the respondents are entirely unaware)
upon attitudes towards nearly any topic. The experimental philosophers tend to
use the technique to explore the meaning of concepts. Neil Levy pointed out
that this same approach could, in principle, be applied to the full range of
issues in neuroethics.




Neil’s presentation struck me like a thunderbolt. I had come to the field of neuroethics with a background in cellular and molecular biology, and had spent much of my career as a card-carrying reductionist: as a graduate student in the 1980’s, I championed the then-novel technique of recording from single neurons in freely moving animals, and as a postdoc I moved on to the better controlled (if less naturalistic) technique of patch clamp analysis of identified neurons in slices of brains. My subsequent rise through the ranks of academia was one in which I applied quantitative rigor to every question that I asked, and in the circles in which I traveled, this was lauded as the ultimate way to provide reproducible (and by inference, meaningful) results. I saw at once that the CVT opened the door towards doing something similar in the field of neuroethics.




My research group at the National Core for Neuroethics has embraced the use of contrastive vignettes wholeheartedly, and with a nod to the experimental philosophy camp, we call the approach Experimental Neuroethics. The team is applying the technique to a range of issues in contemporary neuroethics, probably best exemplified by our recent publications exploring public attitudes towards cognitive enhancement² as well as the acceptability of overt and covert nudges³.




If the vignettes appear simple, I can assure you that properly crafting them is hard work. We begin with a carefully considered hypothesis and regularly find that the hypothesis morphs substantially (usually into something much more insightful) as the process unfolds. We then compose two or more contrastive vignettes, working hard to have the vignettes as minimally contrastive as possible (one word differences between vignettes is the ultimate goal, but this is often not feasible). Finally, we develop questions; we like to have the wording of the questions always be identical irrespective of the contrastive nature of the vignette.

Filming a vignette (Source)

Then the real fun begins. After a day or two, we assemble as a team and attack our previous work. Inevitably, we find it wanting in some respect. Sometimes, embarrassingly so. We find it best to begin by asking whether the vignette and the questions directly address the hypothesis. Sometimes that means that the hypothesis changes. Nearly always, that means that the vignette changes. This process is repeated again and again, over days and weeks and sometimes months (yes, and even sometimes years!) until we have a set of vignettes that get to the heart of the matter.




At some point late in the process we carry out cognitive pre-testing. This involves sharing the vignette and the questions with someone who has no particular expert knowledge (friends of friends are likely culprits), and debriefing them about what they read. We are sometimes amazed to find that what we intended for people to glean from a vignette is at odds with their reading of the vignette. That sends us back to the drawing board.




We also run some metrics to determine whether the words we have used are
understandable by a general audience. We use online readability tests such as this
one to establish the educational level required for understanding the
vignette; our goal is that no more than a high school education is required. Finally,
we launch the survey, recruiting respondents from amongst the thousands of
people who have signed up on Amazon’s Mechanical Turk – they’re more
representative of the real population and aren’t as blatantly WEIRD as typical
undergraduate samples. And then we hold our breath.



Once the data is analyzed, we get mired once again in deep discussion. For it is not just the quantitative aspect of Experimental Neuroethics that is satisfying (to me), but also that the data gives us an entirely new benchmark for engaging in the process of wide reflective equilibrium. Throughout this process we remain aware that an ought can not derive from is, but having the data at hand, our version of ought is very much informed by the is. Ultimately, our data emerge in concert with our normative insights, and then one more advantage of Experimental Neuroethics is realized: it is easy for others to replicate our experiments, or even to improve them by taking our vignettes and modifying them to further test their own. This iterative process of replication, critique, and systematic modification has proven to be a robust strategy for advancing insights into the nature of biological and physical phenomena. Only time will tell whether Experimental Neuroethics catches fire in our discipline as it has in the field of philosophy (where it remains controversial). If it does, we can trace it back to Neil’s presentation in Halifax….





Cross posted on Neuroethics at the Core

References



1. Burstin K, Doughtie E, Raphaeli A. Contrastive Vignette Technique: An indirect Methodology Designed to Address Reactive Social Attitude Measurement. Journal of Applied Social Psychology. 1980;10(2):147–65.



2. Fitz NS, Nadler R, Manogaran P, Chong EWJ, Reiner PB. Public attitudes toward cognitive enhancement. Neuroethics. 2013 doi: 10.1007/s12152-013-9190-z.



3.Felsen G, Castelo N, Reiner PB. Decisional enhancement and autonomy: public attitudes towards overt and covert nudges. Judgment and Decision Making. 2013;8(3):202–13.





Want to cite this post?



Reiner, P. (2013). Experimental Neuroethics. The Neuroethics Blog. Retrieved on
, from http://www.theneuroethicsblog.com/2013/11/experimental-neuroethics_2.html.

Experimental Ethics: An Even Greater Challenge to the Doctrine of Double Effect

In his article Neuroethics: A New Way of Doing Ethics, Neil Levy (2011) argues that “experimental results from the sciences of the mind suggest that appeal to [the Doctrine of Double Effect] might be question-begging.” As Levy frames the Doctrine, the Doctrine is a moral principal that is meant to ground the intuitive moral difference between effects that are brought about intentionally versus those that are merely foreseen. More specifically, the Doctrine is supposed to ground the intuition that, when certain conditions are met, it is morally permissible to bring about a bad outcome that is merely foreseen, but, under these same conditions, it would not be morally permissible to bring about a bad outcome intentionally. Or, another way to put this, the Doctrine claims that it takes more to justify causing harm intentionally than it takes to justify causing harm as a merely foreseen side effect (Sinnott-Armstrong, Mallon, McCoy, & Hull, 2008).

The intellectual roots of the Doctrine of Double Effect begin with St. Aquinas and St. Augustine. The Doctrine has since played a central part in moral theorizing within both the Catholic Church and within secular moral theorizing.




Intuitive illustrations of the Doctrine include (adapted from (McIntyre, 2011)):



1. In a military campaign, it is typically judged impermissible to target civilians, but it is often judged permissible to target a legitimate military target (e.g., a WMD factory) even if the attack on the military target is foreseen to lead to civilian causalities.



2. Someone who thinks abortion is wrong, even in circumstances that would save the mother’s life, might nevertheless consistently believe that it is permissible to perform a hysterectomy on a pregnant woman with cancer, even if it is foreseen that the hysterectomy will lead to the death of the fetus.



However, as Levy points out, there is a great deal of evidence suggesting that if one judges an effect to be morally bad, it is more likely that one will judge the act to have been brought about intentionally (and, thus, not merely foreseen). Much of this evidence comes from a series of studies conducted by Joshua Knobe and others (for an overview see: Knobe, 2010). In the earliest of these studies, Knobe (2003)randomly assigned participants to read one of two stories about a chairman of a company who instituted a new profit-generating program. The only difference between the two stories was that the foreseen side effect of the program would either harm the environment (a morally negative outcome) or help the environment (a morally positive outcome).



The harm version read as follows:

The vice president of a company went to the chairman of the board and said, “We are thinking of starting a new program. It will help us increase profits, but it will also harm the environment.”



The chairman of the board answered, “I don’t care at all about harming the environment. I just want to make as much profit as I can. Let’s start the new program.”



They started the new program. Sure enough, the environment was harmed.

The “help” version was identical to the “harm” version except ‘harm’ was replaced with ‘help’.





After reading the stories, participants were asked if the chairman intentionally harmed [helped] the environment. What Knobe found was that the vast majority of participants were willing to say that the chairman intentionally harmed the environment, but very few were willing to say that the chairman intentionally helped the environment. These findings have led Knobe and others to claim that judgments of intentionality are sensitive to moral considerations. This pattern of asymmetrical attribution of intentionality (and other state-of-mind attributions) based (ostensibly) on manipulations of moral considerations has become known as the side-effect effect, or the Knobe effect.

When is it OK to harm the environment in the name of economic growth? Well, according to one reading of the Doctrine, under certain conditions, it may be permissible to harm the environment if the harm was merely foreseen (and, thus, non-intentional). But when is harming the environment construed as merely foreseen? According to the Knobe effect, probably not often.






If Levy’s construal of the Doctrine are correct, and if people’s judgments of intentionality are sensitive to moral considerations, then the Doctrine of Double Effect is circular and would be an unreliable guide for grounding judgments about the permissibility of actions.



For example, if one already thought bringing about civilian causalities was impermissible, then one would be more likely to judge that the foreseen bringing about of civilian deaths was intentional, even in cases where the bringing about of civilian deaths was a side effect of attacking a legitimate military target. Since the civilian deaths would be judged to be intentional, according to Levy’s construal of the Doctrine of Double Effect it would be impermissible to attack the legitimate military target. Impermissibility judgments feed into intentionality judgments which feed back into impermissibility judgments. Thus, the Doctrine is circular and question-begging!



However, the Doctrine of Double Effect is not always primarily construed as depending on the distinction between intentionally bringing about an outcome versus merely foreseeing an outcome will be brought about. Rather, under many traditional formulations of the Doctrine of Double Effect, the morally relevant distinction depends primarily on whether an act or outcome is a side effect or a non-side effect (i.e., means or a goal).[1]



To build an even stronger case against the Doctrine of Double Effect, one would also need evidence that people more readily construe bad outcomes as non-side effects. To the best of my knowledge, there is currently no published study that shows that moral considerations can affect people’s classification of an outcome as being a side effect or a non-side effect. However, my lab is currently exploring just this possibility, and the early results are in: it looks like moral considerations do have an impact upon whether people classify an outcome as a side effect or a non-side effect.



For example, we find that people overwhelming (about 82%) classify HELPING THE ENVIRONMENT in Knobe’s helping the environment version of the chairman case (see above) as a SIDE EFFECT. However, only 42 percent of people classify HARMING THE ENVIRONMENT in Knobe’s harming the environment version of the chairman case as a side effect.



To make sure this finding was not simply a consequence of the exact details of Knobe’s chairman cases, we also tested seven other cases that were modeled on Knobe’s chairman cases.



For example, in one of the cases, a scientist (instead of a chairman) is deciding whether to implement a new methodology (instead of a new program) that would help her get the results she wanted (instead of generating more profits). In one version of the story, the new methodology would also violate ethical guidelines (instead of harm the environment), and in the other version of the story, the new methodology would also conform to ethical guidelines (instead of help the environment).



In another example, a ship captain is deciding whether to take a new route that would help her arrive at the destination quicker. In one version of the story, the new route was a dangerous route and thus would put the crew into extreme danger, and in the other version of the story, the new route was a very safe route, allowing the safety of the crew to be ensured. Same basic structure of Knobe’s original chairman case, but the actor and outcomes were varied.



What we found was that in six of the eight cases (including Knobe’s chairman case) people were more willing to say that the bad outcome was not a side effect than they were willing to say that the good outcome was not a side effect. To put this differently, when the outcome was good (e.g., helping the environment, conforming to ethical guidelines, ensuring the crew’s safety), people overwhelming judged the outcome to be a side effect. But when the outcome was bad (e.g., harming the environment, violating ethical guidelines, putting the crew into extreme danger), people tended to be split on whether the outcome was a side effect or a non-side effect.



Thus, our initial evidence suggests that people’s classification of outcomes as side effects or non-side effects may, in part, depend on moral considerations. [2] If this is right, then even when construing the Doctrine as being primarily concerned with the distinction between side effects and non-side-effects, the evidence suggests that the Doctrine of Double Effect is circular and would be an unreliable guide for grounding judgments about the permissibility of actions.







Want to cite this post?

Shepard, J. (2012). Experimental Ethics: An Even Greater Challenge to the Doctrine of Double Effect. The Neuroethics Blog. Retrieved on
, from http://www.theneuroethicsblog.com/2012/08/experimental-ethics-even-greater.html



_____________________________________________________________________________

Notes



[1] While it is true that talk of intentions is present in almost all discussions of the Doctrine of Double Effect (even those that construe the primary distinction as side effects versus non-side effects), I take it that discussion of intentions play a role in the Doctrine in so far as intentions are a guide to distinguishing what outcomes should count as side effects versus non-side effects.



[2] In my view, technically, people’s asymmetric classification of outcomes as side effects/non-side effects is not dependent on moral considerations, but rather is dependent on a non-moral considerations that typically (though not always) correlate with the evaluative valence of an outcome. Getting into the details of my view is beyond the scope of the post and is unimportant for the particular point at hand. (It would turn out on my view that the Doctrine would be circular for a large portion of cases due to the nature of the correlation between the non-moral considerations and evaluative outcomes.)



Works cited



Knobe, J. (2003). Intentional action and side effects in ordinary language. Analysis, 63(3), 190–194.



Knobe, J. (2010). Person as scientist, person as moralist. The Behavioral and brain sciences, 33(4), 315–29; discussion 329–65.



Levy, N. (2011). Neuroethics: A new way of going ethics. AJOB neuroscience, 2(2), 3–9.



McIntyre, A. (2011). Doctrine of double effect. Stanford Encyclopedia of Philosophy. Retrieved from http://plato.stanford.edu/entries/double-effect/



Sinnott-Armstrong, W., Mallon, R., McCoy, T., & Hull, J. G. (2008). Intention, temporal order, and moral judgments. Mind & Language, 23(1), 90–106.