Last week, the largest randomized controlled trial ever conducted on a multi-cancer early detection (MCED) diagnostic test, Grail’s National Health Service England (NHS)-Galleri trial, reported topline results that did not meet the primary endpoint: the test did not reduce the number of cancers first diagnosed in Stage III and IV. In response, Grail’s stock dropped roughly 50%.

With a blood test, also known as a liquid biopsy, Galleri detects DNA shed by tumors into the bloodstream and identifies cancer signals that predict their tissue of origin. Predicting the tissue of origin helps direct follow-up diagnostic tests. Grail designed its test for a false positive rate below 1% for population-scale use. In a large screening program, even a modestly higher false positive rate could cause unnecessary follow-up procedures.¹ In turn, the cost of that constraint has been sensitivity. Currently, Galleri detects only 40–50% of all cancers and ~75% of the 12 cancer types that account for the majority of cancer deaths, most of which are not screen-able today.²

The NHS-Galleri trial enrolled 142,000 participants aged 50–77 across England, randomizing them between standard NHS cancer screening and standard screening plus annual Galleri testing over three years.³ The primary endpoint was a statistically significant reduction in combined Stage III and IV cancer diagnoses. Grail’s study did not meet that endpoint. Stage IV diagnoses declined meaningfully in the intervention arm, but Stage III diagnoses actually increased, as did Stage I and II detections. The net trend moved in the expected direction but did not reach statistical significance.

The scientific rationale of the trial is that Galleri would diagnose cancer early, before it progresses to a later stage in the control arm. Stage IV cancers traditionally are considered incurable, while Stage III cancers are at high risk of transitioning to Stage IV despite treatment. Reducing the number of cancers that transition to later stages could reduce the number of patients succumbing to cancer.

That said, the transition from one stage to another takes time and can vary considerably by cancer type. Pancreatic cancer, for example, can advance from Stage I to Stage IV within months, whereas colorectal or prostate cancer can progress over a decade. Three years might not be enough time to capture the transition between stages. In the aftermath of this trial, Grail intends to extend the follow-up window to assess whether stronger signals emerge over time and to submit detailed results for peer review in mid-2026.⁴

Sensitivity rates also are challenges for this kind of trial. Currently, the detection rates for Stage I and Stage II cancers are only ~17% and ~40%, respectively, suggesting that tests miss the majority of early-stage cancers.⁵ Improving sensitivity without compromising specificity is the core technical challenge for the next generation of MCED.

The NHS-Galleri results neither refute nor validate MCED in its current form. Instead, they demonstrate the difficulty of translating a biologically plausible hypothesis into population health outcomes within the constraints of a clinical trial. Questions about the trial results suggest some next steps. Risk-stratified screening—deploying MCED tools in populations with elevated cancer probability based on age, behavior, or other biomarkers—could be more productive than universal population testing. Embedding MCED in existing screening workflows, which is Guardant Health’s path, would sidestep some of the early-stage sensitivity and reimbursement challenges that Galleri is facing.

Diagnosing a wide range of cancers in early stages with routine blood tests seems feasible scientifically and commercially. According to management, Grail is committed to pursuing FDA approval for its Galleri test. Moreover, on a patient-pay basis, the company already has generated ~$150 million revenue in the US, and management believes that a rollout in the UK is still possible.

Last week, U.S. Food and Drug Administration (FDA) Commissioner Marty Makary and Center for Biologics Evaluation and Research (CBER) Director Vinay Prasad announced in the New England Journal of Medicine that the agency's default standard for drug approval will shift from two pivotal clinical trials to one, supplemented by confirmatory evidence.⁶ The announcement marks one of the most significant changes to the FDA's evidentiary framework in decades, albeit a direction toward which the agency has been moving for years.

The Old Standard
The FDA's current evidentiary standard traces back to the thalidomide crisis of the late 1950s and early 1960s, when a sedative used for morning sickness caused severe birth defects in thousands of children.⁷ In response, the US Congress passed the 1962 Kefauver-Harris Amendments, requiring for the first time that drug manufacturers demonstrate efficacy—not just safety—through substantial evidence from adequate and well-controlled investigations. The FDA interpreted that language as requiring two or more studies, each one convincing on its own—the “two-trial dogma” standard.⁸

In 1997, Congress softened the requirement in the FDA Modernization Act. The FDA issued implementation guidance the following year, clarifying that "data from one adequate and well-controlled clinical investigation and confirmatory evidence" could establish effectiveness.⁹ Initially, single pivotal trial approvals were the exception, focused on oncology and rare diseases with limited patient populations and acute unmet need.¹⁰ Increasingly they became the norm, so much so that, in 2024, 66% of novel molecular entities (NMEs)—drugs with active ingredients not approved previously—relied on a single pivotal trial. Indeed, in 2020, the median number of pivotal trials per approval dropped from two to one.¹¹ Among single-trial approvals, ~70% received Orphan Drug Designation, granting incentives to sponsors developing drugs treating rare diseases that impact fewer than 200,000 people in the US.¹²

What Will Change Now?
Now, Makary and Prasad are formalizing that the FDA will require only one pivotal trial, which is not a relaxation of standards but a recognition that science has advanced. The historical rationale for requiring two trials—reducing the probability of a false positive from roughly 1 in 40 to 6 in 10,000—was designed for an era with less precision in evaluating whether a drug works.¹³

Today, drug development builds a body of evidence in several ways, among others determining a product’s mechanism of action, assessing its effect on biomarkers, and incorporating intermediate endpoints that collectively build a more complete picture of the way a drug works. Confirmatory evidence, in lieu of a second trial, includes mechanistic data, findings from related indications, animal models, evidence from drugs of the same class, and/or real-world data.¹⁴

Implications For Drug Development Economics
The downstream implications for drug development could be meaningful. While the cost of a single pivotal trial can vary by therapeutic class and trial design, removing a second pivotal trial could save up to $350 million, per program.¹⁵ While many drug sponsors run pivotal trials in parallel or in short succession, removing the second trial could shave years from development timelines and lower operational complexity.

While oncology and rare diseases already have benefited from greater regulatory flexibility, the FDA’s shift on pivotal trials is likely to impact drugs that target common conditions—like cardiovascular and respiratory diseases—in which two trials have been not only the norm but also often the largest and most expensive.¹⁶

That said, the FDA will retain full discretion to require additional studies, particularly for interventions without a clear mechanism of action or those relying on indirect markers instead of direct patient outcomes like survival.¹⁷

A Broader Shift Ahead?
In our view, the one-pivotal trial announcement could signal broader changes to come in the clinical development paradigm. Makary has described a vision for "continuous trials" in which the traditional sequential phase structure—Phase 1 for safety, Phase 2 for proof of concept, Phase 3 for large-scale “pivotal” efficacy and safety data—gives way not only to continuous trial designs producing data that reviewers can access in the cloud but also to the end of administrative pauses after each Phase during which drug sponsors resubmit applications.¹⁸ Seamless and adaptive trial designs already exist, particularly in oncology, but Makary's vision could extend the concept more broadly across drug development. If implemented, such reforms could compress timelines further and reduce costs, helping expedite the delivery of new medicines and shore up the competitive position of the US in biotech drug development.

Source: CGTN 2026.¹⁹ For informational purposes only and should not be considered investment advice or a recommendation to buy, sell, or hold any particular security.

Last week, humanoid robots stole the show during China’s annual Spring Festival Gala. Globally, Chinese humanoid robot companies have taken the lead in sales and now command ~98% of the global market, as shown below. Given its status as the epicenter of global manufacturing and supply chains, as well as strong government support, China’s lead is not surprising.

Note: “RoW” = Rest of World. ARK Investment Management LLC, 2026, based on data from 36Kr European Central Station 2025²⁰ and company reports as of 2/22/26. For informational purposes only and should not be considered investment advice or a recommendation to buy, sell, or hold any particular security. Past performance is not indicative of future results.

The quality and functional utility of China’s robots—particularly their autonomy—is the subject of much debate. In research and development stages, many Chinese humanoid robots can wave at customers and strut around factories, but skeptics wonder if their movements are pre-programmed? In our view, real progress will depend on the amount of data gathered from live manufacturing environments.

ARK’s Big Ideas 2026 report outlined both the timeline and the complexity involved in developing humanoid robots. In our view, a general-purpose humanoid robot will be ~200,000x more difficult to develop than a special purpose robot like a robotaxi.²¹ Importantly, the cumulative compute investment necessary to reach human-level proficiency is likely to be ~$55 billion, a crossover point that could take place in 2028. Stay tuned for our upcoming blog on this topic.