On the evening of January 10, 2016, Guillaume Molinet, a 49-year-old artist and father, was admitted to a hospital in Rennes, France, with the symptoms of a stroke. He died a week later. Molinet was participating in a Phase 1 clinical trial of a new drug designed to activate the cannabinoid system, with potential activity against pain, anxiety, and a variety of other symptoms. His death and the hospitalization of the other members of his dose cohort with severe adverse effects has re-opened concerns about safety in early phase first-in-human trials, concerns that were last debated after the death or hospitalization of participants in the TGN1412 trial in 2006.


The endogenous cannabinoid system was first identified as the target of marijuana. There are cannabinoid receptors in both the central and peripheral nervous systems, and they have effects on anxiety, mood, memory, and the perception of pain. The most prominent endogenous ligand is a molecule named N-arachidonoylethanolamine, or anandamide. Revealingly, the word anandamide is derived from the Sanskrit word ananda, which means “bliss” or “delight.” In the body, anandamide is rapidly degraded by membrane-bound fatty acid amide hydrolase (FAAH), which limits its action on its receptor.

Research into therapeutic manipulation of the endogenous cannabinoid system has focused on both synthetic ligands and on increasing the effect of anandamide by slowing its degradation. The latter approach is analogous to treating depression by using chemicals that block the reuptake of serotonin and other neurotransmitters to potentiate their activity. The drug that was being tested in the study in France, Bial 10-2474, was just such an inhibitor of FAAH.

Inhibition of FAAH is not a novel idea. Several major drug companies, including Janssen, Pfizer, Bristol Myers-Squibb and others, have developed such inhibitors, and several of these have been taken into clinical trials. At the pre-clinical level, viable knockout mice have been developed that lack endogenous FAAH; these mice display a phenotype consistent with increased sensitivity to endogenous and exogenous cannabinoids.[1]

Trial Design

The protocol for the Phase I trial of Bial 10-2474 (Bial 102474-101) was first leaked by Le Figaro and later released by France’s ANSM (L’Agence nationale de sécurité du médicament et des produits de santa).[2] The study is a comprehensive Phase I evaluation in healthy volunteers, starting with an escalation of single doses and proceeding through a series of dose escalations of the drug given daily for 10 days. One of the lessons of the TGN1412 trial was that individuals in a cohort should not be dosed at the same time, allowing the safety of a dose to be determined in the smallest number of individuals necessary before being expanded to the entire group. In Bial 102474-101, dosing within the lowest dose single administration cohort was staggered – later cohorts could be dosed simultaneously if no safety signals had been raised at lower doses. This design was presumably chosen because the mechanism of action of the drug was believed to be well characterized, and we have no record of objections being raised when the protocol was reviewed. The protocol has been criticized, after the fact, for lacking sufficient pharmacokinetic and preclinical data, but it is possible such data was disclosed in the investigator’s brochure, which the company has declined to release.

Timeline of events

The trial began enrolling participants in July 2015. What purports to be a leaked information sheet for participants suggests that compensation for the single dose administration part of the trial was 1900 € (around $2,115).[3] The single dose administration portion was completed, and 4 cohorts of the 10-day multiple dose portion also completed, with no toxicity that stopped the trial. According to the original protocol, there were to be 10 single dose escalations, from a starting dose of 0.25 mg to the highest dose of 100 mg. Each dose was tested in a cohort of 6, 4 of whom received the active drug and 2 of whom received placebo. The protocol does not specify the dose for the multiple administration cohort, as these were to be determined after completion of the single dose escalation, but information released after the trial was stopped disclosed that the first four multiple dose administrations occurred at doses of 2.5 mg, 5.0 mg, 10 mg, and 20 mg daily for 10 days. Each multiple-dose cohort consisted of 8 participants, 6 of whom received study drug and 2 of whom received placebo. Eighty-four participants received the study drug between the beginning of the first single dose cohort and the end of multiple dosing cohort 4.

Molinet was the 85th participant and the first member of the fifth multiple dose escalation cohort. He received his first dose on January 7, 2016. Three doses later he was admitted to the University Hospital in Rennes, with symptoms suspicious of a stroke. For reasons that remain in question, the clinical trial site administered the fifth dose to the remaining participants in the cohort the following morning, before the gravity of Molinet’s condition led the sponsor to stop the trial.

Imaging findings showed “deep necrotic and hemorrhagic lesions in the brain,” [4] and after the trial was stopped, the other 5 participants in the cohort who received the drug were reported to have similar findings on MRI. No such findings have been reported on any of the earlier participants.

Issues Raised

There is, as yet, no explanation for the severe toxicity observed in the 50 mg/day cohort. No such toxicity was observed over 10 days at 20 mg/day, nor at a single dose of 100 mg. There is nothing we know about the “on-target” mechanism of action of the study drug that would explain either the particulars of the findings or the narrow window between a tolerable and lethal dose.

Without an explanation related to FAAH inhibition, toxicity may have been the result of action of the drug at another, unexpected site, or the action of a metabolite. While it hasn’t been raised, adulteration or other problems in the production of the product for this particular cohort might also explain unexpected toxicity.

The protocol has been criticized for its lack of information on detailed pharmacokinetics and preclinical studies, but such information may have been available to the investigators. It has also been criticized for dosing all members of each cohort (except the first) simultaneously, a design that runs contrary to the EMA guidelines that arose from the TGN1412 trial; but sequential (rather than simultaneous) administration is explicitly recommended for “high risk” trials.[5]

The most pointed criticism has been directed at the conduct of the trial, specifically, that the remaining members of the cohort received their fifth dose after Molinet had been hospitalized, and that the fatal or life-threatening events were not reported to authorities for three days.

Comparison with TeGenero trial

Disastrous outcomes in Phase I trials are rare, and Bial 102474-101 invites comparisons with TGN1412. But the underlying issues are very different, and there is little to suggest that Bial made the same mistakes as TeGenero. In particular, toxicity in TGN1412 occurred at the first and lowest dose (and the first dose in the trial was given to the first group of participants simultaneously), dose selection was based on flawed extrapolation from pre-clinical data, and the observed toxicity was noted as possible in the protocol but its likelihood was minimized. In Bial 102474-101, the first dose cohort was dosed sequentially, not simultaneously; initial dose selection appears to have been appropriate (based on the completion of the first 84 participants), and the observed toxicity was neither expected nor as yet understood.


It is clear that lax oversight in the execution, not design, of Bial 102474-101 resulted in five subjects receiving a potentially lethal fifth dose. Sequential dosing of all cohorts would also have protected these five individuals, but it is hard to call for such a hard-and-fast design requirement in a protocol that was executed without unexpected toxicity through eight single dose and four multiple-dose cohorts. On the other hand, the trial site’s explanation for administering the next dose to the remaining members of the cohort after having sent one member to the hospital is suspect. The first subject appeared to be doing well, and his return to the trial site was “expected,” [6] but sites must have a low threshold for suspending drug administration in the face of adverse events, whether or not they appear severe or related, and particularly if they are unexpected and unexplained.

Something went very wrong in this trial, but it is too soon to know what it was. Lessons learned will need to wait until the full story is told and the full science understood. First-in-human research inherently carries risk. We should continue to try to minimize risks, particularly in the execution of trials, but we should also appreciate how Bial 102474-101 did not repeat the fundamental design errors of TGN1412.


TeGenero did not survive the disaster of TGN1412, but TGN1412 itself was acquired by the Russian company TheraMAB. Renamed TAB08, or theralizumab, the drug is now in Phase 2 trials for rheumatoid arthritis (at much lower doses).



[1]Cravatt BF et al. (2001) Supersensitivity to Anandamide and Enhanced Endogenous Cannabinoid Signaling in Mice Lacking Fatty Acid Amide Hydrolase. Proceedings of the National Academy of Sciences of the United States of America 98:9371–9376. Available at: http://www.jstor.org/stable/3056353.

[2]Clinical Study Protocol No BIA-102474-101 (2015) 1–96.

[3] Essai thérapeutique mortel à Biotrial Rennes. Le contenu du test du Bia 10 – 2474 de Bial révélé [exclusif] – Breizh-info.com, Actualité, Bretagne, information, politique (2016) Available at: http://www.breizh-info.com/2016/01/15/rennes-biotrial-bialessai-therapeutique/.

[4] More details emerge on fateful French drug trial (2016) Available at: http://www.sciencemag.org/news/2016/01/more-details-emerge-fateful-french-drug-trial.

[5] SWP C (2007) Guideline on Strategies to Identify and Mitigate Risks for First-In-Human Clinical Trials with Investigational Medicinal Products. 1–12.

[6] PRESS RELEASE – February 4th, 2016 (2016) 1–2. Available at: http://www.biotrial.com/news/press-release-february-4th-2016.html.