409: Predicting biotech clinical trials & a new Alzheimer's drug controversy

The episode opens with discussion of three major biotech developments. First, Eli Lilly announced it is acquiring Atai Life Sciences for $2.8 billion upfront plus up to $1 billion in milestones, with the centerpiece being 5-MeO-DMT, a psychedelic treatment for treatment-resistant depression. Elaine Chen notes this is a riskier psychedelic compared to psilocybin or LSD due to its intense experience, though it offers advantages in terms of faster administration time (shorter clinic visits). The acquisition reflects Eli Lilly's financial capacity to make aggressive bets in emerging areas like psychedelics and neuropsych.

Second, Merck received FDA approval for Lipfendra, the first oral PCSK9 inhibitor for cholesterol management. This is significant because existing PCSK9 drugs require injections, which has limited patient adoption and revenue despite their efficacy. The hosts discuss whether this oral formulation will achieve better market penetration than injected alternatives given the prevalence of cheap, generic statins.

Third, Biogen presented phase two data for Dyrannosin, a tau-targeting Alzheimer's drug. The drug showed a 26% slowing of cognitive decline at its lowest tested dose, though there was an inverse relationship between tau reduction and clinical benefit—the lowest dose had the least tau reduction but most clinical benefit, while higher doses showed computational side effects. The hosts note this represents a different mechanism from current amyloid-targeting drugs like Lecanemab and Kisunla, potentially avoiding amyloid-related imaging abnormalities (ARIA). However, the confusing dose-response relationship and lack of clarity about the drug's true mechanism raise questions about whether Biogen should proceed to phase three, particularly given the company's past struggles with Aducanumab.

The main interview features Jonathan Kimmelman, a bioethicist at McGill University who has researched prediction markets in clinical trials. Kimmelman discusses both the potential and risks of Kalshi's new initiative to create prediction markets around phase three clinical trials and FDA decisions. He explains that prediction markets can aggregate dispersed expertise better than traditional decision-making approaches (the "Moneyball" problem in drug development). However, he raises significant ethical concerns: prediction markets could introduce bias by influencing patient enrollment decisions, investigator behavior, and patient adherence during the trial period. Even with Kalshi's mitigation strategy of launching markets only after patient enrollment closes, risks remain during the data collection period. Patients in comparator arms might drop out if market prices suggest the experimental drug is superior, and investigators might alter their assessment behavior based on market signals. Kimmelman argues that ideally, prediction markets should only launch after all data collection is complete, though this reduces their informational value. He also distinguishes between prediction markets on specific trial outcomes versus broader paradigm-level questions (like whether the amyloid hypothesis will succeed), arguing the latter may be more scientifically valuable. Regarding Adam's comparison to stock markets, Kimmelman contends that stock prices are crude predictors of trial outcomes because they reflect multiple factors beyond trial results. He also notes a fundamental tension: if prediction markets can accurately predict trial outcomes, it suggests researchers already know too much, raising ethical concerns about randomizing patients to potentially inferior treatments. True clinical equipoise requires genuine uncertainty about outcomes.