To what extent does the subconscious of FDA advisory committee members influence their votes? Writing as "Scrying Biotech," Michael Webb asserted here that (1) under certain conditions, AdCom members are influenced, (2) these conditions are in place for the April 1 AdCom panel meeting reviewing Afrezza, the inhaled insulin developed by MannKind (NASDAQ:MNKD), and (3) the nature of this influence will reduce the chances that the panel will recommend approval for Afrezza.
As a psychologist who studies how the actual, imagined, or applied presence of others influences how we feel, think, and act, I agree with some of what Webb says, but disagree with his conclusions. Below I describe Webb's argument and why I both agree and disagree with him. My counterclaim is based on research findings that demonstrate that at every moment, the way that each of us feels, thinks, and acts is affected by two simultaneously operating self-regulation systems, one that operates mostly subconsciously, and one that operates mostly consciously. I describe how both systems operate, spell out the conditions that determine which is dominant, and explain how all of this applies to the April 1 AdCom hearing.
Michael Webb's Three Key Assertions
- The subconscious of individual FDA Advisory Committee members may exercise undue influence on their conscious deliberation.
- Such undue influence was most likely under four conditions: when (1) the market for the drug being assessed was very large, (2) the drug's developer was a small pharmaceutical company, (3) one or more large-cap pharmaceutical companies were marketing competing drugs, and (4) the small company had no large-cap pharmaceutical partner.
- In particular, the FDA AdCom members' efforts to objectively assess the evidence of the drug's safety and efficacy would be subconsciously undermined by their long-standing relationships with staff of companies whose level of success is threatened by the drug, especially since the force of these relationships was not being offset by an opposing force stemming from long-standing relationships with staff from a big pharmaceutical partner.
Webb and I agree with the following four propositions:
- Decisions are made in a political context.
- Emotions affect behavior. In a fascinating book, Jonathan Haidt offers a cogent metaphor: emotions are the elephant and our reasoning is the rider. He asserts that when we make moral judgments, our emotions signal our conclusion, and frequently, our efforts at reasoning work to rationalize (make sense out of) what we have already determined. The large elephant determines where we will go, but we have the illusion that the elephant is responding to our guidance. (What is especially interesting about this book is his explanation as to why liberals don't understand conservatives.)
- Long-standing relationships trigger emotions.
- Tugs toward maintaining a given set of relationships will have more force if they aren't counterbalanced by opposite tugs from another set of relationships.
Nevertheless, life is more complicated. In the next section, I explain why.
Our Two Simultaneously Operating Self-Regulation Systems
I begin with a note about my mode of communication. I believe that compared to straight prose, carefully constructed tables are more effective in helping you grasp main ideas. Accordingly, when you get to the three tables below, please linger a little on them until you determine that you have digested their basic messages. Now to the substance.
The feelings, thoughts, and actions of all of us (whether AdCom panel member, investor, dictator, or grape picker) is influenced by two fundamentally different but simultaneously operating self-regulation systems: a reactive, self-protective, and scripted system and a deliberative, meaning-seeking, and creative system. Table 1 summarizes four important differences between these two systems. The first three rows describe how the two systems strive to meet different goals, and how they operate. The last row describes four factors that collectively determine for any given moment, which system exerts more influence. You can read more about these two systems in this information-packed book by Nobel Laureate Daniel Kahneman.
One of the factors listed in the last row of Table 1 (affecting which of the two systems exerts greater influence) is the level of situational stress. Table 2 describes how the level of stress exerts its influence. Note that under high stress, we draw on the fundamental features of our reactive system, whereas when we feel safe, our purpose and creative energy can take advantage of the features of our deliberative system.
Whether we end up relying primarily upon our deliberative self-regulation system will depend upon the other three factors mentioned in the bottom row of Table 1. Table 3 fills in details by describing how the nature of the task and social setting promote rational deliberation.
The routine to be followed by AdCom panels is well established, and their tasks are to assess whether the evidence demonstrates that the drug being reviewed is safe and efficacious. The panel members face public accountability in part because via the internet, the general public can watch and listen to the proceedings in real time. Please take special note of the traditions and social norms that draw heavily upon the panel members' deliberative self-regulation systems.
Am I suggesting that politics and emotion play no role? No, again it depends upon still other factors. However, the most important of these is clarity: both regarding the specificity of the questions to be answered and the persuasiveness of the evidence. If the questions are open to multiple interpretations and the evidence is ambiguous, politics and emotion will increase the influence of panel members' reactive (and subconscious) self-regulation systems.
One more point: Panel members are invited to serve on subsequent panels if their actions are consistent with FDA traditions and norms. In other words, the FDA leadership is one very specialized constituency to which panel members are accountable, thereby increasing the influence of their deliberative self-regulation systems.
The next section briefly reviews the questions to be answered and the nature of the evidence so that you can make your own assessment as to which self-regulation system will be most dominant for panel members.
Implications for Afrezza
This section will be brief because the information has been presented elsewhere, for example, in George Rho's excellent article. Its purpose is to remind MannKind investors whose confidence may have been shaken by Webb's article where to focus their attention when assessing his points.
Specificity of the Questions the AdCom Must Answer
This submission has a history. Afrezza has been turned down twice by the FDA. To skeptics, this means that the medicine is cursed. Why throw good money after bad, etc.? However, if our goal is to evaluate the clarity of the questions this panel must answer, we can see that the history has helpfully sharpened the FDA's focus.
As we now know, the FDA was on the verge of approving Afrezza when its staff was confronted with the fact that FDA rules required MannKind to present evidence demonstrating that the Gen2 inhaler had a lung safety profile that was equivalent to the MedTone. Even though the news was delivered at the last minute and in a way that some people consider underhanded, the fact is that the FDA needs to follow its own rules. In my judgment, it was the right decision.
Review what the FDA asked for in its January 18, 2011 Complete Response Letter. The following excerpt comes from Slide 16 here: "You should conduct two randomized controlled phase 3 trials with the Gen2 device, one in patients with type 1 diabetes and the other in patients with type 2 diabetes. At least one of these trials should include a treatment group using the MedTone inhaler so that we can obtain a head-to-head comparison of the pulmonary safety data for the two devices." As you can see, the lung safety profile of the MedTone was never an issue.
Furthermore, as discussions continued, the FDA expressed interest in expanding the population of subjects being studied. This same Slide 16 contains an excerpt from the May 4, 2011 End of Review meeting: "Possible AFREZZA users may be patients with type 2 diabetes who have failed oral antidiabetic medications and who prefer to add an inhaled insulin product with the goal of delaying the need for injectable antidiabetic therapy. Overall, the type 2 diabetes Study should focus on the most likely users of AFREZZA among type 2 patients so that results can be the most generalizable." In other words, the FDA saw that Afrezza was sufficiently promising to encourage MannKind to seek approval for a larger population.
There were three main questions:
- In patients with Type 1 diabetes, was Afrezza "non-inferior" to a Rapidly Acting Analogue with respect to lowering A1c levels? This question did not leave room for judgment. Specifically, the FDA said that to be non inferior, if Afrezza didn't lower A1c levels as much at the RAA, the difference between the reductions had to be smaller than .40. (Elsewhere, I have argued that A1c is less appropriate than other metrics, but the FDA chose to use A1c, so A1c reductions are what the panel will examine.)
- Was the Gen2 lung safety profile equivalent to the MedTone lung safety profile? The issue here is whether there is a clinically significant difference between the two profiles.
- In patients with Type II diabetes, was Afrezza superior to a placebo in lowering A1c levels? This question also did not leave room for judgment. The FDA did not set a fixed target for the absolute difference between A1c reduction levels. Rather, its question is whether the difference was statistically significant, which means that the difference in how much more Afrezza reduced subject A1c levels compared to the placebo had to be large enough so that the chance that the difference was merely due to having randomly drawn a favorable sample was less than one out of 20.
Persuasiveness of the Evidence
Let's review the data, one question at a time.
- The difference in reductions had to be smaller than .40, and it was .21. In these studies, the computed difference is an estimate for the full population difference, which could have been larger or smaller. The upper end of the 95 percent confidence interval is still below .40.
- The lung safety profiles for the MedTone and Gen2 inhalers are virtually identical. The differences between them are clearly neither statistically or clinically significant.
- Remember that the difference in how much more Afrezza reduced subject A1c levels compared to the placebo had to be large enough so that the chance that the difference was merely due to having randomly drawn a favorable sample was less than one out of 20. In fact, the difference was large enough so that the chance that the difference was merely due to having randomly drawn a favorable sample was less than one out of 10,000. This is an unusually high level of statistical significance. Afrezza was clearly superior to the placebo.
It is hard to call this anything but an open and shut case. Nevertheless, Webb raised the possibility that the FDA could reject Afrezza on the basis that there is no way to know whether long-term use might lead to permanent lung problems, including cancer. On the one hand, no one can rule out this possibility. But there are several things that can be said. First, this is not one of the questions to which the FDA is currently seeking answers.
Second, the very small number of cancers that occurred in Exubera, which had been marketed by Pfizer (NYSE:PFE), was larger than that for the control group, but (1) the difference wasn't statistically significant, and (2) the control group had an unusually low rate of patients developing lung cancer. If the rate of developing lung cancer had been compared to the normal base rate, it would have done better, not worse. Because of these facts on the ground, the FDA did not remove Exubera from the market. Instead, it only added information about the finding on the label.
Third, Afrezza stays in the lung for a substantially shorter time than Exubera did, making it harder to envision why it would fare any worse than Exubera. In fact, when MannKind did a two-year study with Afrezza using the MedTone, there were only a couple of cases of lung cancer, which was no more than expected, and one of those had metastasized from another part of the body. Thus, there is absolutely no basis for panel members to argue that the very remote possibility that there could be long-term lung problems should prevent the FDA from approving Afrezza.
The bottom line is that all the conditions are in place for the deliberative self-regulation systems of AdCom panel members to be dominant. This means that social, emotional, and political considerations will be trumped by the nature of the FDA's task, its traditions and norms, the narrow focus of the questions to be addressed, and the compellingness of the evidence.
So I am sticking with the thesis of my previous article: First, Afrezza will be approved by the FDA. Second, that because of its attractive characteristics (how it performs on more appropriate metrics), MannKind will negotiate an attractive deal with a large-cap pharmaceutical partner, and Afrezza will prove a blockbuster commercial success.
Disclosure: I am long MNKD. I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.