Betting on Biotech: Kalshi’s Clinical-Trial Prediction Markets

For years, online prediction markets were largely associated with sports and political forecasting. Today, however, this speculative architecture has expanded into healthcare, allowing traders to buy and sell contracts based on clinical trial outcomes and FDA decisions

Kalshi (a CFTC-regulated prediction-market exchange) has launched a pilot program for event contracts on late-stage drug trials and FDA outcomes. Each contract is a binary “Yes/No” bet tied to a specific public event (e.g. “Will Drug X meet its Phase 3 endpoint by date Y?”), with settlement based on published data (ClinicalTrials.gov records, FDA letters, advisory votes, etc.). Kalshi’s rules require that no market opens until after a trial’s enrollment closes, and that all traders verify employment and have no material nonpublic information (MNPI). The partners (Kalshi and AppliedXL) define clear resolution criteria before trading begins and review all evidence from primary sources to determine outcomes.

These markets aggregate dispersed expectations about trial outcomes into a public probability. Advocates argue they can reveal hidden insights in drug development and “democratize forecasting” by giving stakeholders (investors, smaller companies, patients) a shared “odds” number. Critics warn of manipulation risks, ethical issues, and unintended effects on trials. For example, bioethicist Jonathan Kimmelman notes that market signals might corrupt trials themselves by influencing patient enrollment, investigator behavior, or dropout patterns. Both academic commentary and industry voices highlight pros (efficient aggregation, speed) and cons (insider trading, feedback loops, patient recruitment impacts) of these markets.

We compare these markets to (a) statistical trial simulations (power analyses and Monte Carlo), (b) Wall Street/analyst success-probability models, and (c) informal FDA-approval forecasts. These approaches differ in inputs (e.g. design parameters vs expert judgment vs historical rates), outputs (market price vs model probability vs policy statement), transparency (public vs private), users (traders vs sponsors vs regulators), and demonstrated accuracy. A summary table contrasts these attributes.

The STAT Readout Loud podcast (July 16 2026) featuring Kimmelman underscored that prediction markets, by making expectations public, create a paradox: “the very market that aggregates expert knowledge… can also nudge the patients, doctors, and researchers inside a clinical trial toward the outcome that the market predicts”. Kimmelman argued the most socially valuable markets would bet on broad scientific paradigms (e.g. the amyloid vs tau hypotheses) rather than individual trials.

Prediction markets offer novel, real-time signals about trials, but carry risks. Regulators and sponsors should proceed cautiously. Stakeholders might benefit from markets that aggregate evidence, but must guard trial integrity and privacy. We recommend strict safeguards (as Kalshi is starting) and focus on post-recruitment or paradigm-level markets to avoid distorting ongoing trials. Broadening transparent data sharing (while respecting confidentiality) and encouraging regulatory guidance will help realize potential benefits (better forecasting, informed decisions) without undermining ethical trial conduct.

---

1. How Kalshi’s Clinical-Trial Markets Work

Kalshi’s new markets are structured as CFTC-regulated event contracts. Each contract asks a binary question about a named trial or FDA event (e.g. “Will Company A’s Phase 3 trial meet its registered primary endpoint by date D?” or “Will the FDA approve Drug X by date D?”). The contract has a price from $0 to $1, interpretable as the market-implied probability of the outcome. For example, a $0.72 price implies ~72% probability of “Yes”. Contracts are settled in cash: if the event occurs (“Yes”), contracts pay $1; if not (“No”), they pay $0.

Market Creation and Eligibility: Kalshi (not AppliedXL) decides which markets to list. To protect trial integrity, Kalshi and AppliedXL only list late-stage (Phase 3) trials by large companies, and only after enrollment completes. This limits insider influence and enrollment effects. Traders must pass employment verification (to ensure U.S. residency) and Kalshi bars anyone with nonpublic insider information. These KYC and MNPI rules mirror those in finance.

Contract Terms & Resolution: Each contract’s terms specify exactly how the outcome will be determined. The resolution source is a specific public document or record: e.g. the posted results on ClinicalTrials.gov, the FDA’s published approval letter, or the transcript of an advisory committee vote. AppliedXL predefines a logic for interpreting that source (e.g. what constitutes “meeting” the endpoint) before trading starts. If multiple data sources conflict or are delayed, the contract’s resolution rules (and Kalshi’s exchange rules) dictate how to proceed.

In practice, once results emerge, AppliedXL gathers the evidence (registries, filings, publications) and maps it to the contract criteria. A human reviewer checks for ambiguities. Kalshi then independently reviews the compiled record and makes the final settlement call. In short, Kalshi is the ultimate arbiter: its published exchange rules govern any dispute. AppliedXL does not bind the outcome; it only provides an “auditable evidence package” to support resolution.

Regulatory Status: Kalshi operates as a CFTC-designated contract market (DCM). This means U.S. federal rules for derivative trading (similar to futures) apply. Kalshi’s exchange complies with transparency and surveillance rules (including MNPI prohibitions) common to financial markets. In sum, the markets are legally structured as regulated binary futures on factual, objectively verifiable drug-development events.

flowchart LR
    TrialStart[Trial Initiated] --> EnrollEnd[Enrollment Closes]
    EnrollEnd --> MarketOpen[Contract Listed on Kalshi]
    MarketOpen --> Trading{Trading Period}
    Trading --> DataPub[Results Published (e.g. CT.gov, FDA letter)]
    DataPub --> ApplyGather[AppliedXL Gathers Evidence]
    ApplyGather --> KalshiReview[Kalshi Reviews & Adjudicates]
    KalshiReview --> Settled[Contract Settles $0 or $1]
    Settled --> MarketClose[Market Closes]

This flowchart shows the resolution process: only after enrollment ends does Kalshi list the market. Traders then bet until results are published. AppliedXL compiles relevant evidence, and Kalshi officially determines the outcome and pays out.

2. Pros and Cons of Prediction Markets for Clinical Outcomes

Pros – Information Aggregation and Efficiency: Prediction markets are praised for efficiently pooling dispersed information. In theory, many participants (scientists, investors, even well-informed patients) each contribute “private signals” by trading, and the market price continuously reflects the aggregate wisdom. As one commentator noted, markets draw on people “touching different parts of the elephant” of knowledge to reach the truth. By trading, insiders and outsiders alike reveal their beliefs. This can produce a real-time probability that updates with every new data point, unlike sporadic polls or reports.

FierceBiotech quotes Endpoint Arena’s CEO arguing that markets “democratize” trials and “incentivize” people to forecast scientific outcomes, potentially speeding up discovery. Patients and doctors, Fischer claimed, could use markets as one more tool to gauge trial success alongside medical advice. Indeed, markets make explicit what is often implicit: they turn qualitative analyses and rumors into a single number. And unlike a stock price (which lumps all assets), a prediction market isolates the specific event (one trial or drug).

Cons – Manipulation, Bias, and Ethics: Many experts warn that betting on trials risks distorting them. A constant concern is insider trading: participants with nonpublic knowledge (e.g. biotech execs, CRO employees, trial investigators) could trade on early data. While Kalshi bans MNPI and job-related insiders, enforcement is challenging. A small, thinly traded contract is vulnerable: even a modest bet could swing the odds, misleading outsiders. For example, a drug’s sponsor might (illegally) bet on success to move prices, or its competitor might bet on failure. The Science/AAAS policy forum warns modern markets can be “designed to leverage legal ambiguities” and can be “gamed by a single firm” in thin markets.

Critics also highlight feedback loops: public odds might influence patient behavior and investigator judgments. As Windisch observes, clinical trial recruitment is sensitive to perceived risks. If a market price dips low, potential volunteers or their doctors might decide not to enroll, making failure more likely. Conversely, if odds are high, control-arm patients might drop out or report fewer side effects (to get the active drug), biasing results. Kimmelman’s analysis formalizes this: he lists channels by which market signals could bias trials (e.g. enrollment reluctance, investigator drift, differential dropout).

The ethical stakes run deep: if a prediction market becomes sufficiently accurate, Kimmelman argues, it implies trials are collecting data we largely already know—raising a “Moneyball paradox” for pharma. In his view, “we ought not to know the answer… if the trial is going to be ethical”. In other words, profoundly predictable trials might violate the principle of clinical equipoise. There are also patient confidentiality concerns: open markets could tempt leaks of trial details (or patient-level experiences) into public view.

Finally, social and legal downsides echo broader gambling risks. The Science/AAAS warning notes that large commercial prediction platforms prioritize engagement and profit, potentially normalizing gambling-like behavior. Prediction markets could encourage addictive speculation or distract from evidence-based decision-making.

In summary, the benefits (aggregating knowledge, transparent signals) must be weighed against serious risks: insider abuse, changes in trial conduct, and ethical objections. Many recommend strict limits (as Kalshi is doing) or even pausing markets during active recruitment.

3. Impacts on Clinical-Trial Operations

Introducing betting on trials could alter many aspects of how trials are run:

• Trial Design: Sponsors may rethink blinding and endpoints. Knowing a market will be watching, companies might favor harder endpoints (e.g. overall survival) over subjective measures, or avoid early-phase exploratory designs. Conversely, some worry sponsors could game designs to influence markets (e.g. choosing broad endpoints that are harder to “bet down”). Kalshi’s pilot avoids this by focusing on agreed-upon primary endpoints and late trials.
• Recruitment & Enrollment: The clearest impact is on patient enrollment. If public odds suggest a low chance of success, fewer patients may volunteer. Melinda Chu (oncologist) warned that a small biotech could fail to meet enrolment targets if a market is pessimistic. The Stat/SensibleMed article models this: a patient Googling a trial’s odds and seeing them low “may decide not to participate”. This could slow enrollment or force sponsors to extend trials. Even after closing enrollment, participants already enrolled might drop out if they perceive an arm as failing. Kalshi’s requirement to list after enrollment closes is meant to blunt this effect, but as Kimmelman notes, markets can still influence behavior during follow-up.
• Investigator Behavior: Knowledge of market odds could bias clinical assessments. Investigators might (consciously or unconsciously) grade patient outcomes more favorably for the arm predicted to win, or be extra vigilant for side effects in the underdog arm. Windisch notes that even supposedly “blinded” trials are often effectively unblinded by side effects. Thus outcome adjudication could drift to align with expectations, compromising data integrity.
• Reporting and Sponsor Actions: Sponsors watch markets too. If negative odds rise, a sponsor might hasten a press release or request an earlier interim analysis. Mike Abrams (Numerof) speculates that poor market pricing “invites those with access to non-public information to compromise their integrity”. In extreme cases, a sponsor might consider altering trial conduct (e.g. unblinding early or stopping a trial) in response to market pressure. Transparency moves by regulators (like releasing complete response letters) have already strained companies; adding prediction markets could amplify investor and media pressure around each trial result.
• Data Integrity and Privacy: Public odds effectively publish a piece of interim information. The FDA and trialists guard interim data carefully (Data Monitoring Committees are kept blind). A real-time market price is a novel public signal. As Windisch warns, this breaks the traditional data firewall. Moreover, if individuals feed inside info (e.g. one site’s data) into trades, patient confidentiality could be indirectly breached.
• Investigator Incentives: While most discussion focuses on patients/sponsors, even trial investigators have stakes. They cannot legally trade on MNPI, but they will see market odds. A clinician who built their career on a drug might feel pressure if the odds are low, or pride if they are high. It’s unclear how that social factor might subtly affect trial conduct.

In sum, markets introduce new dynamics into trials. Kalshi’s safeguards (no early-phase bets, post-enrollment launch, MNPI bans) are aimed at minimizing distortion. But evidence (from social science and medical ethics literature) suggests even later-phase, after-enrollment bets could influence participation and conduct. Regulators and trial sponsors will need to monitor for these effects carefully if such markets grow.

4. Comparison: Forecasting Methods for Clinical Success

We compare four approaches to forecasting trial outcomes:

• Prediction Markets (Kalshi) – Inputs: collective trader info (public news, science publications, unofficial reports). Outputs: live market price (probability). Transparency: prices and contract terms are public (but trader identities are anonymized). Incentives: financial profit; participants have skin in the game. Users: investors, analysts, smaller companies, patient advocates, curious public. Accuracy Evidence: Empirical data is sparse for biotech markets specifically, but in other domains (politics, sports) markets often match or beat polls/experts. (For example, markets famously predicted elections better than most polls.) Accuracy depends on liquidity and broad participation.
• Statistical Simulations (Power Analysis / Monte Carlo) – Inputs: known trial design parameters (sample size, event rates, effect size assumptions, dropout rates), historical data. Outputs: probability of detecting an effect (power), expected distribution of outcomes under hypotheses. Transparency: usually private or technical; sponsors and regulators see them, but models are rarely public. Incentives: none of profit – used for trial planning and regulatory justification. Users: trial statisticians, CROs, regulatory reviewers. Accuracy Evidence: These models are “accurate” only to the extent the assumptions (e.g. effect size) are correct. They help design the trial (e.g. set sample size) but are not predictive of actual real-world outcome beyond those assumptions.
• Analyst/Model Predictions (Wall Street) – Inputs: company disclosures, clinical data, published literature, historical success rates, competitive landscape, sometimes proprietary databases. Outputs: often a “probability of success (POS)” or recommendation (“buy/hold/sell”) but usually qualitative. Transparency: low. Each firm’s model is proprietary, and analysts typically do not publish their probability models; investors only see summaries or final forecasts. Incentives: institutional profit, reputation. Analysts may overestimate success to maintain stock coverage. Users: biotech equity investors, pharma strategists. Accuracy Evidence: Historically mixed; analysts incorporate many factors, but can be swayed by hype or ignore failures. AppliedXL’s research shows traditional models often miss trial-specific execution issues. (For example, analysts grossly overestimated some high-profile trials that ultimately failed.)
• FDA/Historical Forecasts – Inputs: broad historical success rates, class-effect knowledge, advisory committee input. Outputs: not formally published, but the FDA effectively uses internal probabilities to guide decisions (e.g. pre-specified approval benchmarks). Some outside groups infer likely outcomes from trial context. Transparency: limited. FDA guidelines and public data give clues (e.g. FDA reports average success rates: historically ~48–50% for Phase 3 to approval. Incentives: regulatory (public safety and efficacy). Users: regulators, large pharma (for portfolio planning). Accuracy Evidence: Reflects aggregate historical truth: if “FDA forecast” means just using the baseline approval rate, it’s about 50% for a Phase 3 trial. But FDA also evaluates each trial's data rigorously, so final decisions are tailored, not fixed by formula.

Below is a comparative summary table:

Attribute	Prediction Markets (Kalshi)	Trial Simulations (Statistical)	Analyst Probability Models	FDA/Historical Forecasts
Inputs	Public signals, news, expert tips (and sometimes private info) from many traders	Assumed effect size, variance, enrollment rates, epidemiology; mathematical models	Published data (trials, disclosures), historical success rates, expert judgment	Historical approval rates; broad clinical and drug-class knowledge (e.g. “Amyloid vs Tau”)
Outputs	Market price (real-time probability)	Statistical power curves, simulated outcome distributions	Percentage likelihood or ordinal scores (often unpublished)	None formal; general heuristic (“~50% if phase3”) often inferred
Transparency	High: contract terms and prices visible to all	Low/Medium: typically internal to sponsor/regulator; not public	Low: models proprietary; only sometimes summary data or odds are shared	Medium: FDA shares some data (success rates reports); but no real-time “forecast” given
Incentives	Monetary gain from correct prediction; crowdsourced wisdom	None (academic/regulatory goal of sound trial design)	Career/reputation of analysts; investment profits	Regulatory mandate; patient safety and public trust
Typical Users	Investors/traders, biopharma analysts, journalists, patient advocates	Sponsors’ statisticians, clinical trial designers, FDA reviewers	Institutional investors, biotech equity funds	FDA reviewers, big pharma R&D planners
Accuracy Evidence	Empirically good in other domains (e.g. politics); untested in pharma. Dependent on liquidity and diverse participation.	Relies on quality of assumptions; can accurately predict power given assumptions, but no real “success rate” metric.	Mixed: many documented misses. (E.g. analysts missed NKTR-214 failure.)	Tied to history: ~50% of Phase 3 trials succeed. Provides baseline odds but not trial-specific factors.

This table highlights that prediction markets provide a real-time public signal tied narrowly to a defined event, whereas simulations are forward-looking design tools, analysts rely on patchwork models, and FDA forecasting is mostly implicit and aggregate. Markets score high on transparency but raise unique incentive issues.

5. Insights from STAT’s “Predicting Biotech Clinical Trials” Podcast

STAT’s Readout Loud podcast (July 16, 2026) discussed these markets with bioethicist Jonathan Kimmelman. Key points included:

• The Moneyball Metaphor: Kimmelman described drug development as a “Moneyball problem” for pharma. Traditional decision-making pools opinions of a few insiders, whereas markets could synthesize broader expertise. He noted that markets can “do a pretty good job getting us as close as possible to the truth” by aggregating dispersed knowledge.
• Risk of “Infecting” Trials: He warned that trading on trials “effectively bet[s] on the behavior of human beings” – patients, doctors, researchers. He outlined four bias channels (enrollment reluctance, assessment drift, patient-reported distortions, asymmetric dropouts) through which public odds can influence trial data. Even blinding doesn’t fully protect, since patients infer their assignment over time.
• Ethical Paradox: Kimmelman’s standout argument: “If… we can predict the outcomes of… trials, it suggests we know too much at the point where we’re running clinical trials”. In other words, if the market price is consistently accurate, one must ask why randomization is ethical at all. He concluded “if the trial is going to be ethical, we ought not to know the answer to that trial in advance”.
• Stock Market vs. Prediction Market: When asked why not just use stock prices, Kimmelman replied stocks are “clumsy predictors” (they mix all company factors). A pure contract “focuses the lens on one particular question”, isolating the trial from corporate noise. Still, he cautioned that prediction markets “will not eliminate the biases” inside companies – the market is only one input among many.
• Better Targets – Scientific Paradigms: Kimmelman argued the most valuable bets are “paradigm-level”: e.g. “Will anti-tau therapy meaningfully alter Alzheimer’s progression by year X?”. Such questions aggregate across trials and cannot be gamed by one company’s data. They would help allocate R&D resources more broadly, unlike narrow Phase 3 bets whose “development decision has already been made”. He noted, however, that designing and resolving such long-horizon bets is complex and not part of Kalshi’s initial launch.

These perspectives reinforce caution: even STAT’s hosts and guests recognized the dual nature of the tool. Fischer (Endpoint Arena CEO) remained optimistic about motivation and speed, but Kimmelman and others stressed potential trial impacts. As the podcast summarized, one must guard “whether traders, patients, and trial investigators can resist the impulse to let the odds shape the outcome.”

6. Conclusion & Recommendations

Kalshi’s clinical-trial markets mark a novel experiment in biopharma transparency. On one hand, they potentially unlock “a public probability” for drug success that can inform investors, competitors, and patient communities. On the other, they introduce new complexities for trial ethics and conduct.

Balance & Guardrails: Early experience suggests markets should remain tightly constrained. Kalshi’s pilot wisely restricts participation (Phase 3 only, post-enrollment, large sponsors, verified traders). Regulators may consider formal guidance on prediction markets (e.g. echoing FDA advice on interim data and DMCs). Patient-trial recruitment might require monitoring if markets proliferate.

Focus on Wider Signals: Stakeholders may find more value in markets that avoid these pitfalls. Kimmelman’s idea of “paradigm bets” suggests regulators or public institutions could sponsor markets on broad questions (disease mechanisms or multiple-trial outcomes). These would aggregate expertise without tying to one active trial.

Transparency and Education: Markets are not a substitute for data. It is vital that physicians, patients, and media understand these odds as probabilistic estimates, not guidance to skip trials. Clear disclosures (like Kalshi’s disclaimers) are needed to prevent misinterpretation of prices as scientific verdicts.

Research and Oversight: Finally, both academic and industry groups should study these markets’ behavior. Empirical research (like comparing market prices to actual outcomes over time) will reveal their predictive power and any unintended consequences. The Kalshi–AppliedXL report itself calls for ongoing evaluation and ethical review.

Recommendations by Stakeholder:

• Regulators (FDA): Monitor whether prediction markets affect trial integrity. Issue guidance on data sharing and physician communications in the context of public betting markets. Encourage clinical trial transparency and timely reporting to reduce opaque information gaps that markets try to fill.
• Sponsors/Investigators: Advise trial sites and ethics committees about the existence of relevant prediction markets. Ensure robust blinding and patient education. Consider the potential PR impact of market odds when planning trial communications.
• Investors/Analysts: Use market probabilities as one input among many, with awareness of their limitations (thin liquidity, potential distortions). Until verified, treat early market prices as speculative signals rather than facts.
• Patient Advocates: Engage with these markets carefully. They could offer clues about trial success, but patients should continue relying on physicians and trial protocols. Markets also spotlight which trials lack public information, bolstering calls for better reporting.

In conclusion, Kalshi’s markets bring innovative tools to biotech forecasting and clinical trial prediction. If managed responsibly—with strict rules, transparency, and ethical oversight—they could complement traditional models. Yet stakeholders must heed warnings: a misused market might do more harm than good. The safest path is cautious experimentation, rigorous study, and a focus on long-run benefits (like faster knowledge sharing) while guarding core trial integrity.