1

ReportbytheTemporarySpecialistScientificCommittee(TSSC),"FAAH(FattyAcidAmideHydrolase)",onthecausesoftheaccidentduringaPhase1clinicaltrialinRennesinJanuary2016.1.ForewordThe"FAAHinhibitors"TSSCwassetupbytheDirectorGeneraloftheAgenceNationaledeSécuritéduMédicamentetdesproduitsdesanté(ANSM),followingtheaccidentthatoccurredduringthePhase1,first‐in‐humanclinicaltrialonthemoleculeBIA10‐2474,inRenneson10January2016.ThescientificmissionsoftheTSSC,were,onthebasisoftheavailabledataandexpertiseofitsmembers:

‐ Toanalysethemechanismsofactionandpotentialtoxicityofsubstanceswhich,like BIA 10‐2474, are presumed to have a direct or indirect effect via theendocannabinoidsystem.

‐ To put forward and, where possible, list hypotheses to be able to explain thetoxicity observed in several volunteers in the trial conducted in Rennes byBiotrial.

‐ Toenact,whereappropriate,generalrecommendationsaimingtotightensafetyforvolunteers,especiallyduringfirst‐in‐human(Phase1)trials.

TheTSSC,fromthetimeitwassetup(25January2016)toissueofitsreport(18April2016),workedaccordingtothreemethods:

‐ Individualexpertappraisalofthedocumentsprovidedtoeachmember.‐ Twoone‐day "open"meetings (15Februaryand24March2016)duringwhich

the expert appraisals were read. Another open meeting between Bial andmembersoftheTSSCalsotookplaceon18March2016.ThethreemeetingswereheldontheANSM'spremises,attendedbytwoinspectorsfromtheGeneralSocialWelfareInspectorate(Christined’AutumeandGillesDuhamel).Arepresentativefrom the EMA (EuropeanMedicines Agency) also attended themeetings of 15February(Hans‐GeorgEichler)and24March(JeanMarcVidal)asobserver;tworepresentatives from the Portuguese Medicines Agency (Ana Catarina FonsecaandIsabelVieira)alsoparticipatedasobserversinthelastmeeting.

‐ QuestionandanswerandgroupwritingsessionsforTSSCmembersonly,whichledtoapprovalofthetwoversions(intermediateandfinal)ofthepresentreport.

Thisorganisationmadeitpossibletoreservediscussionofthekeypointsoftheexpertappraisal,conclusionsandrecommendationsforTSSCmembersonly,independentlyofthepresenceoftheorganiser(ANSM)andobservers(IGASgeneralinspectors,Europeanand Portuguese Agency representatives). This closed phase represented by far thelargestpartoftheTSSC'swork.

2

ThemissionoftheTSSC'sexperts,althoughitrepresentedasignificantamountofexpertappraisalwork(estimatedatmorethan600hoursintotal),didnotservetoreplaceaninspectionunderanycircumstances.Therefore,theconclusionsofthisreportshallnotprejudgethoseoftheongoingadministrativeandlegalinvestigations.Concerning the source documents used, there is a foreword to the BIA 10‐2474Investigator Brochure, written by Bial, which contains incorrect translations andtranscription errors, especially in the tables and figures. This, in several places, givesrise to ambiguity and comprehension difficulties, includingwith respect to importantinformation (see chapter 6). This deserves to be highlighted as the InvestigatorBrochureisadocumentusedasreferenceduringpre‐approvalphasesofahealthcareproduct,asrecalledbyinternationalrulesandrecommendations.Finally, although theTSSCwassetupbydecisionof theANSM'sDirectorGeneralandreceived logistics support from the Agency, the Committee conducted its work andinvestigations fully independently during the two and half months of its existence,especiallywithregardtotheANSM,Bial,Biotrial,thevolunteersthatparticipatedinthetrialandtheirfamiliesanddefencelawyers.AllTSSCexpertsworkedonavoluntarybasisfortheirentiremission.VariousdraftversionsofthisreportweresubmittedtotheTSSC'sexpertsalone,andthenumerousdiscussionsrequiredtofinaliseitandtoreachaconsensusonthekeypointsofthecasetookplaceamongthoseexpertsaloneatalltimes.2.TSSCmembersBernard Bégaud (Medical Pharmacology. Bordeaux University and Teaching Hospital. CR INSERM1219),MarieGermaineBousser (LariboisièreTeachingHospital,AssistancePubliquedesHôpitauxdeParis,Paris‐DiderotUniversity),PascalCohen (InternalMedicine,CochinTeachingHospital,Paris),BertrandDiquet (MedicalPharmacologyandToxicology.MedicineDepartment,Healthresearchunit.Angers University and Teaching Hospital), Pierre Duprat (Veterinary doctor, Doctor of toxicology,EuropeanCollegeofVeterinaryPathologists)WalterJanssens(FederalMedicinesandHealthProductsAgency,Belgium),MichelMallaret (ClinicalPharmacology,RegionalPharmacovigilanceandMedicinalProduct Information Centre, Grenoble Teaching Hospital), Guy Mazué (Veterinary doctor), JoëlleMicallef (Medical Pharmacology, Aix Marseille University and Marseille Teaching Hospital, CNRSresearchunit7289NeurosciencesInstitute,TimoneHospital), ClaudeMonneret (EmeritusResearchDirector,CNRS,Chairmanof thePharmacyAcademy), Jean LouisMontastruc (MedicalandClinicalPharmacology. Toulouse Faculty of Medicine and Teaching Hospital), Laurent Venance(Interdisciplinary Biology Research Centre, College of France, INSERM U1050, CNRS UMR7241, LabexMemolife,Paris).

3

3.BackgroundThemoleculeBIA10‐2474,bytheBialpharmaceuticalcompany(PortelaCa,Portugal),belongs to the FAAH inhibitor family, an enzyme degrading anandamide, a biolipidactingasmediatorinwhatisknownastheendocannabinoidsystem.More than ten inhibitors of this type areorhave alreadybeendeveloped, nonebeingmarketedtodate;formanyduetoefficacyconsideredtobedisappointing.Intermsofstructural chemistry, these inhibitorsmainlybelong to two families:Moleculeswith aureafunctionandthosewithacarbamatefunction,.Research in the field of FAAH inhibitors has been driven by strong hopes and theprospect of highly varied therapeutic indications: pain, vomiting, anxiety, mooddisorders,Parkinson'sdisease,Huntington'schorea,variouscardiovascularindications,tonamebutafew.ForBial'sproduct,theInvestigatorBrochurestatesthatBIA10‐2474wasdeveloped"forthe treatmentofmedicalconditions inwhich there isadvantage inenhance the levelsofendogenousanandamide (AEA)and tonically increase thedriveof the endocannabinoidsystem(Sic)".Theindicationwhichappearedtohavebeenpreferred,atleastinitially,wasneuropathicpain;thiswasconfirmedbyBialduringitshearingon18March2016.First‐in‐human trials were entrusted to Biotrial Research in Rennes, a centrespecialising in investigations and research of this type for almost twenty years. Theaccident which occurred mid‐January 2016 led to the suspension of the clinicaldevelopment of BIA 10‐2474. Its severity and spectacular nature deeply affected thedrugs industry, scientists and the public, as much in France as around the World.Understandingthecircumstances,andifpossible,themechanismsofoccurrenceofthisunprecedentedaccident is thereforeacollectivepriorityandreasonbehindtheexpertappraisalworkconductedbytheTSSC.This expert report, after a reminder on the endocannabinoid system (prerequisite tointroduction of the discussion on the mechanisms of action of the molecule and thehypothesessurroundingitstoxicity)willanalysethemoleculeitself,itspharmacologicalproperties, followed by animal toxicity studies, the protocol used by Biotrial, thesymptoms observed in healthy trial volunteers, and pharmacodynamic andpharmacokineticdata.ThesecondpartwillexplorethehypotheseslikelytoexplaintheaccidentinRennes.AconclusionwillsummarizetheTSSC'sopinionsandpositionsastothekeypointsof thecase.Thereportwillconcludeonrecommendationsaffecting theconductof first‐in‐humantrials that theTSSCwishes tosee implementedatEuropeanandinternationallevel.4.ReminderontheendocannabinoidsystemBIA10‐2474 isanFAAHinhibitor,serinehydrolasedegradinganandamide,oneof themain mediators of what is known as the endocannabinoid system. This equivocally‐named system (it is in fact a lot broader andmore complex than cannabis derivativetargets) exists in a largenumberof species (vertebrates and invertebrates, except for

4

insects) and in mammals in particular. Knowledge is recent (the first receptor wasidentifiedbycloningin1990)andasyetincomplete.Therearetwotypesofreceptors(CB1andCB2),transmembraneandGprotein‐coupled(inhibitingadenylcyclase).

‐ CB1 is a highlyubiquitouspresynaptic receptor foundat the surfaceof severalcelltypes(neurons,astrocytes,pericytes,endothelialcells)andinalargenumberof cerebral sites (basal ganglions, cerebellum, hippocampus, cerebral cortex,olfactorybulb,etc.).CB1isoneoftheGprotein‐coupledreceptorsexpressedatthehighestlevelinthecentralnervoussystem,withthenoteworthyexceptionofthe brain stem. CB1 is also found in peripheral organs (lungs, bowel, testicles,uterus, etc.). The exogenous agonist specific to this receptor istetrahydrocannabinol(THC).

‐ CB2ismainlyfoundinimmunesystemcells(immunomodulatoreffects).

Eightendocannabinoidagonistshavebeen identified todate.Theyarebioactive lipidsactingbothasneurotransmittersandneuromodulatorsandproducedandreleased"ondemand", unlike conventional neurotransmitters which are released from storagevesicles.Thethreemainendocannabinoidsare:

‐ anandamide (AEA); this was the first endocannabinoid to be characterised(1992),

‐ 2‐arachidonylglycerol(2‐AG),arachidonicacidester,‐ 2‐AGether(arachidonicacidether).

LikeTHC,AEAhaspreferentialaffinityfortheCB1receptorandverylowaffinityfortheCB2 receptor. Conversely, 2‐AG has high affinity for both receptor types and it cantherefore be seen as the main endocannabinoid system mediator, whereas AEA hasalmostnoeffectonCB2andisabletointeractwithseveralothersystems.Also,2‐AGisfoundatlevels200to800timeshigherthananandamideinrodents.Unlike 2‐AG, anandamide is therefore little specific to the endocannabinoid system inthestrictsenseofthetermandcanalsobeconsideredtobeanendovanilloid.Itisableto activate TRPV1 (transient receptor potential vanilloid 1) which are non‐selectivecationchannelsfromtheTRPchannelsgroup.AEAalsoactsonothersystems:

‐ it is a good agonist for PPAR (peroxisomeproliferator‐activatedreceptor) alphaand gamma, nuclear receptors involved in the energy metabolism andinflammation,

‐ it interacts in NMDA (N‐methyl D aspartate) glutamate receptors, both asstimulatorbydirectactionandinhibitoractingindirectlyviaCB1,

‐ finally, like other endocannabinoids, it can lead to the activation of multipletranscriptionfactorsinvolvedinneuroprotectionphenomenabytheMAP‐kinasepathwayandachainreaction,whichisahighlypromisingresearchapproach.

The effects of endocannabinoid system stimulation are similar to those induced bycannabis derivatives. Low to moderate concentrations induce behavioural responsescombining stimulant and depressant effects, whereas at high doses, the effects arealwaysofthedepressanttype.Wethereforemainlyseethefollowinginanimals:

‐ antinociception,‐ hypothermia,

5

‐ hypolocomotion.Workingmemoryisaffected,withouteffectonreferencememory.Theeffectonlevelofanxietyisbiphasic:anxiolysisatlowdosesandanxiogenicathighdoses.In termsofsynaptic transmission,endocannabinoidsact inaretrograde fashion(fromtheneuronalpost‐synaptictopre‐synapticelement)andgenerallyreducetransmissionin the short (few seconds) or long‐term (several hours or days). Theymodulate bothexcitatory(glutamatergic)andinhibitory(GABAergic)transmission.After being produced and released by the postsynaptic compartment, AEA is usuallydegradedbyFAAH(membranehydrolase)whichalsopartlydegradesthe2‐AGbutalsoafairlylargenumberofotherbioactivelipids.Unlike in animals, two FAAH isoforms (FAAH‐1 and FAAH‐2) can exist in the humanspecies.Theprevalenceofcarriersof the two isoforms isestimatedataround38%inthegeneralpopulationandthatofcarriersofthelowactivityisoform(FAAH2)5%.Where there is inhibition of FAAH activity, AEA concentrations increase, however anadditional degradationpathway takes over: that of the cyclooxygenases.This leads tothe formation of eicosanoids: leukotrienes and prostanoids (prostaglandins,thromboxanes, prostacyclins) with the ability to act on apoptosis and vasomotricityphenomena; the vasoconstrictor effect of 20‐HETE (20‐hydroxyeicosatetraeinoic acid)inthebrainis,forexample,confirmed.5.ExaminationofmoleculeBIA10‐2474Examinationofthechemicalstructureofthismoleculedoesnottheoreticallyraiseanyspecific questions, especially concerning potential toxicity. The functional groups andchemical nuclei it contains are commonly found inmedicinal chemistry. For example,theN‐oxide function is found inchlordiazepoxide(benzodiazepinesedative),minoxidil(potassiumchannelagonistdevelopedasantihypertensiveagentandusedsecondarilytodevelophairgrowth),andinvariousantiretrovirals.TheoriginalityofBIA10‐2474is for theremainder,relative; itcanbeconsideredasa"variation"ofmoleculespreviouslydevelopedasFAAHinhibitors.Forexample,Pfizer'sPF‐3845 also contains a pyridine nucleus directly adjacent to the urea function. Thiscompound,effective invivoandselective,wasproventobeapowerfulFAAHinhibitor,well‐toleratedinPhaseIclinicaltrials,butwithoutsatisfactoryefficacyinPhase2trials.In the same way, the imidazole nucleus, common in pharmaceutical chemistry, iscontained in the compounds developed by Bristol‐Myers Squibb (carbamate typeinhibitors).However,inthecaseofBIA10‐2474,thisnucleusisinthepositionadjacenttothemolecule'selectrophilicsitewhich(seefurtheron)potentiallymakesita"leavinggroup".AllFAAHinhibitorsdevelopedarebasedontheformationofacovalentbondbetweenhydrolaseserine241and thecarbamateorureaelectrophiliccarbon.FAAH inhibitioncan therefore be considered to be irreversible. According to Bial, BIA 10‐2474 iseffectively covalently bound to FAAH (therefore irreversibly) in vitro but partiallyreversibly invivo. This has already been reported in the case of Janssen & Janssen'sinhibitor(JNJ‐42165279)withwhichpartialenzymeactivityisobservedafter8hours.

6

AsignificantdifferencebetweenBIA10‐2474andmostknown inhibitorsconcerns itslowspecificity for its target enzyme. Concentrations inhibiting FAAHactivity at 50%(IC50)range,onaverage,from1.7(1.5–1.9)micromolar(M)inmiceto1.1(0.9–1.3)M in rats. They are believed to be 100 times higher at most for the various otherenzymesagainstwhichBIAwastested,accordingtoBial.Bial thereforeonlytesteditscompound and one of its metabolites (BIA 10‐2445) on three serine hydrolases:monoacylglycerol lipase (MAGL), a carboxyl‐esterase and an acetylcholine‐esterase(selectivityof10forratFAAH,and50forhumanFAAH).Theotherenzymestestedweredopamine‐beta‐hydroxylase,glutamicaciddecarboxylase,monoamineoxidasesAandBandcholine‐acetyltransferase.This contrasts with the results with other compounds such as Pfizer's PF‐04457845(testedagainst68receptors)whichhasanIC50of7.2nanomolar(nM)forhumanFAAH(therefore240timeslowerthanthatofBIA10‐2474)andofover100Mforapanelofaround twenty hydrolases. The specificity ratio of Pfizer's compound is therefore nolonger100,butaround14,000.ThesameappliesforJNJ‐42165279byJanssen&Janssentested on 50 enzymes. The low affinity/specificity of BIA for its target enzyme willfurther lead us to envisage "parasite" binding to other serine hydrolases duringdiscussionofthetoxicitymechanismobservedinhumans.Itshouldberecalledthattheserine hydrolase superfamily counts around 300 members and that it is thereforerecommended developing inhibitorswith the highest possible affinity for the enzymetargeted.Proteomicscreeningwouldprobablyhaveprovidedusefulinformationinthisrespect.Nine presumedBIA 10‐2474metabolites have been synthesised (compounds BIA 10‐2639,10‐2583,10‐3258,10‐3827,10‐2445,10‐2631,10‐3844,10‐2580and10‐3764).Allhaveastructurethatisverysimilartothatofthemothermolecule.Theycorrespondeither to reduction ofN‐oxide, or to hydroxylation of the cyclohexanenucleus (whichleads to the formation of more hydrophilic compounds), or to demethylation of theamine function,or to twoconcomitant changes.Theoretically,nothing in the chemicalstructure of these metabolites portends potential toxicity. Three of them have thepotential to inhibit FAAH to a similar extent as that of the mother molecule. Thesemetabolitesaremainlyfoundinverysmallquantities,evenafter14days’administrationof BIA 10‐2474 to animals. BIA 10‐2631 (produced by N‐oxide reduction andcyclohexanehydroxylation),ishoweverfoundinlargerquantitiesinprimates.Duringpharmacokinetic studies inhumans (see furtheron), fourof thesemetaboliteswere identified, 2 were undetectable and 2 measured at much lower plasmaconcentrationsthanthoseofthemothermolecule(<3%).Unlesswepresumeacompletelydifferentmetaboliteexists,orthatthereissignificantaccumulationincerebraltissues(veryhightissueconcentration/plasmaconcentrationratio,explainingthelowcirculatingconcentrations),involvementofthesemetabolitesinthetoxicityobservedintheclinicaltrialinRennesappearstobelittlelikely.Inanotherrespect,itshouldberecalledthattheBIA10‐2474moleculeimidazolecycle,in adjacent position to the electrophilic carbon, site of binding to FAAH, can beconsideredtobea"leavinggroup»thatmayproduceanisocyanatetowhichmanybrainproteinsarelikelytobind.When in contact with hepatic microsomal enzymes, BIA 10‐2474, at least up to theconcentrationof30g/mL,only slightly inhibits cytochromesP4502D6and3A4anddoes not seem to inhibit cytochromesP4501A2, 2A6, 2B6, 2C8, 2C9, 2C19 or 2E1. It

7

doesnotappeartoactasaninducer,atleastoncytochromesP450fromseries2Band3A,withadoubtasto1A2.6.PreclinicalpharmacodynamicdataWithrespecttoFAAHinhibition(mechanismpresentedascentraltoitspharmacologicalactivity),wecanconsiderBIA10‐2474asacompound:

‐ withrelativelyweakactivity.50%FAAHinhibition invitrorequires, for instance,concentrations in the mid‐micromolar range whereas for most inhibitorsdevelopedtodate,theyareinthenanomolarrange.Asmentionedabove,theIC50ofBIA10‐2474forFAAHappearstobe240timeshigherthanthatofPfizer'sPF‐04457845,someinhibitorsbeingcharacterisedbyevenlowerIC50.

‐ that is little specific. Inhibition (again in vitro) of other enzymes occurs atconcentrations50to100timesthoseinhibitingFAAH.Itispossible(thishasnotbeen tested) that the ratio is even lower with other cerebral hydrolases. As acomparison,thisratiois,aswehaveseen,ofaround14,000forPF‐04457845.

‐ of little progressive effect. The dose‐effect curves (here inhibition of FAAHactivity) of BIA 10‐2474 show this in particular when we enter an unusuallynarrowconcentrationrange,goingfromabsenceoftoalmostcompleteinhibition.Even if it is fitting to take account of experimental variability, the dose‐effectcurveslopeappearstobehighandverysteep,ifwecompareittothoseofotherenzymeinhibitorsandmoregenerallytootherdrugs.

‐ that is long‐acting. Even if we consider that BIA 10‐2474 is not typically anirreversibleinhibitor,theFAAHinhibitionitinducesisextremelyprolonged.Itisstill almost complete after8hours. Inhumans, inhibitionpersistswell over24hours, whereas BIA plasma concentrations have fallen below the limit ofquantificationofthetestmethodused(thatistosaytheyarenon‐quantifiable).Concerningthedose fromwhichBIA inhibitsFAAHactivity, there isapparentlysignificantdiscordancebetweenwhatcouldbeextrapolatedfromanimalstudiesandthatwhichwasobservedinthevolunteersinthetrialinRennes.Animaldatashowsthatmaximumeffectisachievedfromadoseof0.3mg/Kginmonkeys;adose of 1 mg/Kg not making it possible to increase FAAH inhibition oranandamide concentrations. This (calculation not provided in this report butverifiedbytheTSSC)madeitpossibletopredictthatcompleteFAAHinhibitionwouldbeachievedinhumansforadoseofbetween10and40mg.However,inthevolunteersoftheBiotrialtrial,weseethataround50%inhibitionisachievedfora0.25mgdoseandalmost100%fora5mgdose.Thisisequivalenttoaratioof at least10between thedoseestimatedon thebasisof animaldataand thatmeasuredinhumans.As mentioned previously, the resulting inhibition is extremely prolonged asactivity recovery is not complete 72 hours after administration, whereas theproducthasalmostfullydisappearedfromtheplasma.

As for the analgesic activity of BIA 10‐2474 (therapeutic possibility theoreticallyfavoured), twotests,usuallyusedto test thispossibility(notspecially forneuropathicpain),havebeencarriedoutinmice:

8

‐ Formalinpawtest.Thistestconsistsof injectinga5%formalinsolutionintothetip of the paw of the hind leg. This causes persistent pain leading to repeatedreflex licking by the rodent. The efficacy of the investigational molecule isassessed by the reduction in the number of times the areas is licked (lickingscore) over a given period (here: 15 to 50 minutes) after the injection. Threedoses of BIA 10‐2474 (0.3 mg, 1 mg and 3 mg/Kg) were compared, eitheradministered alone or in combination with 5 mg/Kg AEA, also administeredalone.Gabapentin,atthedoseof300mg,wasusedasreferenceanalgesicforthiscomparison; indeed, this gamma‐aminobutyric acid derivative is used asanticonvulsantandtotreatneuropathicpain.Inthistest,theeffectofBIA10‐2474usedaloneappearedtobeprolongedbutofmoderateamplitude.Ineffect,forthethreeescalatingdoses,andcomparedtoaninactive solvent, the score decreased by 29%, 28% and 41% respectively. TheeffectofAEAalonewasofthesameorderofmagnitude(35%).However,weseea marked dose‐dependent effect for the BIA + AEA combination, the scoresdecreasing by 42% for the combination with 0.3 mg/Kg of BIA, 65% for thecombinationwith1mg/Kgand86%for thecombinationwith3mg/Kg. In thistest, gabapentin appeared to be clearly more effective than BIA since thevariationis76%comparedto41%forBIAaloneatitshighestdose.We note that in the Investigator Brochure the corresponding figure (4.6) isdoubly false compared to the source document provided by the investigator(Porsolt&Partners,Reportno.09.770/2,2010):theverticalaxisshowssecondswhereas they are score values and the column for the gabapentin score(neverthelesspresentedasreferenceforthiscomparativetest)hadbeendeletedfromthebarchart.

‐ Tail flicktest(reflex flick of a tail subject to heat stress). A significantly higherdoseofBIA10‐2474(10mg/Kg)wasusedforthistest.Theantinociceptiveeffectwas maximal in the eighth hour but, like in the previous test, of moderateamplitude(flicktimeincreasingfromaround4.8secondsonaveragetoalmost6,thereforea1.2 seconddifference)butprolonged (adifferencepersisting in the72ndhour).Onthebasisofthesametest,BIAandURB597(FAAHinhibitorfromthe carbamates family), both administered at the dose of 1 mg/Kg, werecomparedwithrespecttopotentiationoftheanalgesiceffectofAEA.PotentiationwasonlyobservedwithBial'scompound.

The doses used in these tests differ greatly (from 0.3 to 10mg/Kg),without it beingpossible to trace a dose‐effect curve or to estimate an effective dose 50 (which is asurprisingshortcoming).Wenotethat:

‐ TheantinociceptiveeffectofBIA10‐2474remainsmoderatewhenthemoleculeis administered alone (presumed conditions of future therapeutic use). BIAhoweverstronglypotentiatestheantinociceptiveeffectofAEA.

‐ The antinociceptive effect of BIA 10‐2474 increases little (score ranging from29%to41%)whenweincreasethedosefrom0.3to3mg/Kg, fortheformalintest, that is tosaywe increase thedosebya factorof10.Thiscouldmeanthatinhibitionof theenzyme involved in thiseffect isalmostcomplete fromthe0.3mg/Kg dose and therefore, that the range of doses tested, too narrow and/orbadlychosen,doesnotmakeitpossibletoaccuratelydetermineaneffectivedose.This would constitute another problematic shortcoming with respect to therationalchoiceofdosesinlaterdevelopmentstages.

9

‐ The potentiating activity of AEA by BIA 10‐2474 is, however, very clear andprolonged,regardlessofthedosetested.

TherelativelackofpharmacodynamicdatacomparedtootherpreclinicalstudieswouldjustifyarecommendationfromtheTSSCconcerningnewdrugdevelopment.7.Animaltoxicologydata7.1.OpeningremarkInterpretingtoxicologystudydataisalwayscomplex.Thesestudiesareconductedwithdoseswhichcanbeveryhigh,incommensuratewiththoseusedinhumans.Therefore,atthehighestdoses,varioussignsoftoxicity,oftenlittle‐specificorvisible(inmacroscopyormicroscopy)onlyaftersacrifice,areobservedinmostoftheanimals.WithanaccidentsuchasthatinRennes,thereisthereforeastrongprobabilityoffindingelementsconsistentwiththetypeoftoxicityidentified,inanimaldatalateron.Thisdoesnotinanywaymeanthattheseelementsconstitutedsignalspredictiveoftoxicityofthistype. To avoid this conventional interpretation bias, the TSSC closely examined theparticularlyextensivedossierofanimalstudiesconducted;thismustbe lookedatasawholeandinitscontext.7.2.ToxicologydossierStudies conducted with BIA 10‐2474 seem to have been conducted according tocurrently approved protocols (ICH recommendations especially) with a highly pureproduct (more than99.9%), identical to thatused for themanufactureof thecapsulesadministeredtothevolunteersattheBiotrialcentre.These studies covered, which is little common (especially for a molecule that is notparticularlyinnovative),fourdifferentspecies(rats,mice,dogsandmonkeys)andwereconductedintwocentresofsoundreputation:HarlanLaboratoriesSAinSpain(studiesinnon‐rodents)andAnaPathGmbHinSwitzerland(studiesinrodents).Anotherspecies (rabbits)wasalsoused forstudieson thepotentialeffectsofBIA10‐2474onfertilityandreproduction.During itshearingbytheTSSCon18March2016,Bialexplainedthat thisparticularlyextensive toxicology programme was related to a delay in the start of clinicaldevelopment, this meaning that additional toxicology studies were carried out orcontinued following first‐in‐human administration (i.e. carcinogenesis studies).Therefore,theTSSCdidnotfind,inanyofthedataitanalysed,anythingtosubstantiatethe hypothesis that this particularly comprehensive and costly toxicology programmewasconductedduetodoubtsastogoodtolerancetothemolecule.Onthebasisof thedata thatcouldbeanalysedtodate,andgenerally,up toveryhighdoses,wedonotobserveany toxicity fromBIA10‐2474 specifically targetinga givenorganandwhichcouldhavebeentakenasasignalprecludingclinicaldevelopment.Oneof the toxiceffects found themostoften in treatedanimals,as forseveralotherFAAHinhibitors,affectssemenandmoregenerally,gametes.Thispoint,withoutadoubt the

10

clearestofthecase,shouldbetakenintoaccountifBIA10‐2474weretobebroughttobeusedasadrug.WenotethatBial,unlikewhathasbeendoneforseveralotherFAAHinhibitors,didnotdefinetargetorgansinitstoxicologyprogramme.Thesensitivityoftheassaymethodsusedduringtoxicologystudiesonlyidentifiedfiveperipheral (plasma compartment) metabolites among the nine which BIA 10‐2474 isabletoproduce.Thesemetabolitesaretheoreticallyidenticaltothosefoundinhumansandalsoproducedinverysmallquantities(around1%oftheparentproduct),andthisfor the four species.Therefore, toxicity studies specifically for thesemetaboliteswerenotlegallycompulsoryandwerenotconducted.Wedonotobserveaccumulationof theproductorof itsmetabolites inmultipledosestudies(over13weeks).TheNOAEL(NoObservableAdverseEffectLevel)andNOEL(NoObservableEffectLevel)dosesseemtohavebeencorrectlydetermined.Theyvariedaccordingtospeciestestedandinterestinglytodurationofadministration,especiallyinmice.TheNOAELforthe4‐weekand3‐monthstudiesarethereforerespectively:

‐ 100and25mg/Kg/24hinmice,‐ 30and10mg/Kg/24hinrats,‐ 50and20mg/Kg/24hindogs,‐ 100and75mg/Kg/24hinmonkeys.

OnthebasesofthecalculatedNOAEL,andbyreferringtoFoodandDrugAdministration(FDA)procedures,itwasintheorylogicaltotestadoseofupto100mginhumans(96mgaccordingtotheTSSC'scalculation).Asinalltoxicologyprotocols,theorgansoftheanimalsprovidedforintheprotocol(40organs)were systematically submitted formacroscopic andmicroscopic examination,without,accordingtothestudycentres'reports,noteworthytoxicityofaspecificorgan,afortioricommontothefourspeciesstudied,beingobserved.Thisalsoappliestoboththecentralandperipheralnervoussystem,especiallyinprimates.However,inratsandmice,cerebraldamage,especiallyinthehippocampuswithgliosisand inflammatorycell infiltrationwereobserved in threeanimals treatedatveryhighdoses.Thisconcernedonemaleandonefemaleinthestudyonmiceat500mg/Kg/24hover4weeksandonerat in thestudyat150mg/Kg/24hover4weeks.Thisdamage,discussed by the TSSC given the context, seems to be fairly commonly observed inrodents in studies of this type, and was not theoretically likely to generate a signal,although such damage does not appear to have been observed with other FAAHinhibitors. For the three animals concerned, as for their fellow creatures in the samegroup, the observation reports do not mention any neurological or behaviouraldisorders.Inthesamewayinprimatesandrats,cerebraldamageandespeciallyoftheautonomicnervous systemwas observed in some animals treatedwith a highdose. Inmonkeys,during4‐weekstudiesconductedwith10,50and100mg/Kg/24hdoses respectively,medullaoblongata (spinal bulb) damage in the formofaxonaldystrophywasnoted insomeanimalsfromthe100mg/Kg/24hgroupandnotinthosereceivinglowerdoses.Itisdifficulttocommentontheprecisehistologicalnatureofthedamage,duetothefact

11

that thetwopathologistswho interpretedthesliceslidesonbehalfofBialdidnotusethesameterminology.Againinmonkeys,a clearly dose‐dependent toxicity (only occurring in the 100mg/Kg/24hgroup),andmoresignificantthanpreviously,concerneddamageintheformofoedemaoftheMeissnerplexuscellsofthedigestivetract.Inthegroupofdogs(beagles)treatedfor13weeks,theTSSC'sattentionwasdrawntothe presence of pulmonary changes clearly visible in macroscopy, confirmed inmicroscopy and termed as "bronchopneumonia/focal andmultifocal acute alveolitis".Thesesymptomsaresurprisingintheirfrequency.ThetoxicologyreportsubmittedbyBiallinkstheselesionstobronchialinhalationofBIA10‐2474.Thishypothesisseemedlittle plausible to themajority of the TSSC experts. In effect, the shell of the capsulesadministeredwas designed to resist gastric juices,making it little plausible that theywere opened and caused powder inhalation. The same reason, and the method ofadministration of the capsules and absence of suggestive symptoms also makeregurgitationfollowing"choking"littleplausible(i.e.explainedbyneurologicaltoxicityof BIA 10‐2474). No alternative hypothesis (infectious contamination, specificsusceptibilityof thisgroupofdogsorotherreason)wasfavouredbytheTSSCforthissymptom which is one of the highlights of the BIA 10‐2474 toxicology dossier. TherelationshipwiththeexistenceofhighCB1receptordensityinthelungs,intheabsenceoffurtherinvestigations,cannotbeconsideredasapossibleexplanation.Also,wenotethe absence of similar symptoms in the three other species (mice, rats andmonkeys)andapparentlysimilarsignsintoxicologystudiesconductedwithotherFAAHinhibitors,including indogsof thesameorigin.Theseclearlydose‐dependentsymptoms,notablyinterferedwith the study plan scheduled in dogs. It planned to test (versus controls)dosesof20,50and100mg/Kg/24hover4weeks.Duetothesignificantsymptoms(twodogs,onemaleandonefemalefromthehighdosegrouphadtobesacrificedbeforetheendofthestudy)andsignsofmotorincoordination,thedoseshadtobereducedinthe50 and 100 mg/Kg/24h groups to be able to complete the study. Therefore, the 20mg/Kg/24h dose was considered to be the NOAEL. As it was impossible to test thehighest doses this may explain why another species, the monkey, was used. Ittheoreticallyconstitutedamoreappropriatemodel for thestudyofanFAAHinhibitor(bettersensitivitytostimulationoftheendocannabinoidsystem)evenifitwasseennottobe thecaseon thedoseestimationplan. Itwouldhoweverseemthat thestudies indogsandmonkeysstartedatthesametimeVarious studieshave thereforebeen conducted inprimates (cynomolgus ormacaque).Maybedue to thatwhichwasobserved indogs, thesestudiescameafter (unlikewhatwasdoneforotherFAAHinhibitors)anup‐titrationphase.AccordingtoBial,thisaimedto achieve "stabilisation of the endocannabinoid system" by monitoring the onset ofsuggestivesignssuchashypothermia(examinationofthedossierhowevershowsthatitneverexceeded1degreeCelsius),hypolocomotionorreduceddietaryintake,etc.Nomortality was observed in the long‐term study (13 weeks at 75mg/Kg after up‐titrationbylevel).However,inanothergroup,onefemalediedafterup‐titrationover12days (10, 25 and 50 mg/Kg/24h) followed by 9 days' administration of BIA at 75mg/Kg/24h.Thedossierdoesnotsayanythingspecificaboutthisanimal.In the same way, several primates had to be put down for ethical reasonsduringascending dose studies to test tolerance to the product at very high doses: the twoanimalsfromGroup1onthefourthdayofthefinallevelat250mg/Kg,thetwoanimals

12

from Group 2 (125 mg/Kg/24h) and one female from Group 3 after threeadministrations at 60 mg/Kg/24h, the other animals having survived to the end ofescalation at 110 mg/Kg/24h. These premature deaths in primates neverthelessoccurred forveryhighmultipledoses,which inHumanEquivalentDoses (HED),wouldbe equivalent to 78 mg/Kg, 39 mg/Kg and 34 mg/Kg respectively. For purelyinformationpurposes,aHEDof78mg/Kgwouldbeequivalenttoaround100timesthehighestdosehavingbeentestedinhumansinmultipledoses(50mg).Duringthehearingon18March,Bialsaiditwasplanningalong‐termtoxicologystudyofBIA10‐2474inmonkeys(52weeks).TheanimalstudiesdossieronBIA10‐2474generallyappearstobeofgoodqualityandno aspects of the data that the TSSC has studied constituted a signal likely tocontraindicateadministrationinhumans.Thisisalsotrueforneurologicaltoxicitywithcentralnervoussystemandautonomousnervoussystemdamagehavingaffectedasmallnumberofanimalstreatedatthehighestdoses.Theinitiallynon‐alarmingnatureoftheneurological damage observed was confirmed by examination of the slices of tissueconcerned, by the TSSC's experts (photographs provided by Bial following the TSSC'srequeston18March2016).Severalcommentsdeservetoberaisedhowever:

‐ The reasons that may have led Bial to use four different species (including 2rodents), which is, for a case of this type, unusual (no other FAAH inhibitorappearstohavebeenstudiedonthisbasis)werediscussedatlengthbytheTSSC.It is however possible that thismay be the result of a change to the trial planduring development: switch from dogs to monkeys due to the poor toleranceobserved in the first species (dogandmonkeystudiesappearhowever tohavestarted on almost the same dates), postponement of the clinical developmentplan(officialresponsefromBial).Itisalsoplausiblethatthestudiesconductedinmice were in fact intended to determine the doses for the (long‐term)carcinogenesisstudies.

‐ Caution should be exercised as to the conclusions that could be drawn fromrelating the doses tested in animals to those administered to the Biotrial trialvolunteers(singleormultipledoses).Theplasmaconcentrationratiosandmorespecifically theareasunder the curve shouldbeused to compare theexposuredensities between animals and humans. Regardless, the doses used in animalsdid not give rise any reservations as to the validity of the preclinical dossier.According to the data provided by the company on the TSSC's request, thereports on the animal/human areas under the curve, where the values areavailable,arecloseto1,exceptinmonkeysinwhichthemarginsarewider(ratioofaround6).Thisconfirmsthatextrapolationsfromanimalstohumansintermsofexposuremarginsanddosestobetested,shouldconsiderallspeciesusedandnotonlythatappearingtobethebestpharmacologicalmodel.

‐ Evenifitcouldnotbeusedtohighlightaspecifictoxicitysignal,thesafetyofBIA10‐2474duringtoxicologystudies,generallyandclearlyappearstobeunderparcompared to previously developed FAAH inhibitors. Besides the specificproblemsoflungtoxicityindogs,theotherinhibitors(despitesometimeshigherdoseratiosthaninthecaseofBIA)didnotseemtoleadtoanytoxiceffectsworthnoting. In particular, damage to the Meissner plexus in monkeys was notobserved and no early sacrifices were necessary. The better tolerance is for

13

instanceconfirmedbythefactthatforoneoftheproducts,theabsenceoftoxiceffects observed led to the use of the highest dose administered in order todeterminetheNOAEL.

8.ClinicaltrialconductedinRennesbyBiotrialThe Phase 1, monocentric, First‐in‐Human (FIH) trial planned to include 128 healthymaleandfemalevolunteersintotal,aged18to55years,andinvolvedfourparts:

‐ singleascendingdose(SAD)study,‐ multipleascendingdose(MAD)study,‐ anfoodinteractionopenstudy,and‐ apharmacodynamicsstudy(notdone).

We see that dispersion of the ages of the volunteers recruited (18‐55 years) is high,some being relatively elderly, compared to what is usually seen in Phase 1, first‐in‐human administration trials. The ages of the six subjects hospitalised at RennesUniversity Hospital ranged for instance from 27 to 49 years. Furthermore, severalvolunteers considered to have a potential risk factor for certain drug‐related adverseeffects were included. Among others, we note a PR interval measured at over 240millisecondsonseveralpre‐doseelectrocardiogramsandbloodpressureofover140/90mmHgoverfourpre‐dosereadings.Thechoiceof the firstdoseadministered(0.25mg)wascareful for theSADpart,as itwasequivalenttoaround1/400thofthehighestdosewithnoobservableadverseeffectlevel(NOAEL)inanimals.TheSADpart1involved64volunteersineightcohortsof8volunteers(sixreceivingtheactivetreatmentandtwotheplacebo)forthe8doselevelstested(0.25mgto100mg);48 subjectswere therefore exposed to the active treatment. Two subjects (one activetreatmentandoneplacebo)weretestedbeforeadministrationtotheothersix.TheMADpart provided for six cohorts of 8 volunteers (six active treatment and twoplacebo),therefore48subjects.Thesixdosestobetestedwere:2.5mg;5mg;10mg;20mg;50mgand100mg.Eachdosewastobeadministeredfor10consecutivedays.The

1Asareminder,werecallthe2006recommendationsoftheFrenchMedicinesAgency(AFSSaPS)forfirst‐in‐humanadministration(page4):"In the same group: •number of volunteers receiving the new active substance simultaneously. It is necessary, except otherwise justified with arguments, to limit the number of volunteers receiving the new active substance simultaneously, according to the risk factors identified. •time between administration to one volunteer and administration to the next. A sufficiently long observation period should be provided for between administrations, especially depending on the product characteristics, the data available (pharmacokinetic, pharmacodynamic) and on the risk factors identified, •criteria for administration to the next volunteer, •criteria for discontinuation of administration to volunteers not yet treated"

14

subjects ineachcohortweretostayattheBiotrialcentrefor15days(and14nights).Fromthe10mgdose,administrationwasbasedonthepharmacokineticdatameasuredatn‐2(i.e.thatforthe10mgcohorttostartadministrationof50mg).AsistheruleinPhase1, thenextdose levelwasusedonly ifnoadverseeffectswereobserved in thevolunteersfromthepreviouslevel,followingthestudymonitoringcommittee'sopinion.Intheseconditions,themoleculewasadministeredatthesametimetoallvolunteersinthecohort.Accordingtotheprotocol,severeintoleranceobservedin4volunteersinthecohort(therefore50%oftotalnumbers)wastoleadtodiscontinuationofthetrial.DuetotheaccidentthatoccurredamongthevolunteersofMADcohort5,only30volunteers(out of the 36 planned) received the active treatment for this part of the trial withmultipledoses.Thefoodinteractionstudyinvolved12volunteersatthe40mgdose.Ninetyvolunteers in total thereforewereexposed toBIA10‐2474during this trial, atvery different doses (0.25 to 100mg, therefore a ratio of 400), whether in single ormultipledailydoses.TheSAD(singledose)partstartedon9 July2015andended(cohort8:100mg)on9October2015.TheMADpart(10days'treatment)startedon6October2015.Thepenultimatecohort(cohort5,50mg)beganon6January2016,therefore19daysaftertheendofcohort4(20mg).We therefore note a three‐day overlap between the SAD andMAD programmes. Thiswasnotatallproblematicfromasafetypointofviewduetotheweakness(2.5mg)ofthefirstdosetestedinMAD.Ontheeveningofdayfive(10January2016)andthereforeofthefifthadministrationinthe50mgMADcohort(totaldoseof250mg),oneofthe6volunteershavingreceivedthe active treatment was hospitalised in Rennes University Hospital in a seriouscondition. Biotrial did not initially consider the relationship between the acutesymptoms presented by the subject and themolecule tested to be possible since theother5volunteersreceivedtheirsixthdosethenextmorning,11Januaryat8a.m.(totaldose:300mg),withoutwaitingfortheresultsofongoingtests(especiallyMRIscan)onthehospitalisedvolunteer.The5volunteersreceivingtheactivetreatment,andnotthetwo subjects receiving the placebo, were in turn hospitalised at Rennes UniversityHospitalbetween13and15January,therefore2to4daysafterthelastadministration.The trial appeared to have been effectively suspended on the 11th, since theadministrations,whichweretocontinueuntilthe15th,werediscontinuedonthatdate.Severalremarksonthiscrucialpartdeservetoberaised:

‐ The trial in Rennes was conducted in a specialized centre (Biotrial) of soundreputation. The reservationmade above as to volunteer recruitment aside, thetrial followed currently used protocols for first‐in‐human /Phase 1 trials. Inparticular, and unlike what has been wrongly cited by several informationsources, there was no time overlap between the various cohorts, especially inMAD. On the contrary, an interval was provided for between the end of onecohortandthestartofthenext.

‐ Althoughnotspecifictothistrial,wecanonlybeastonishedthat,especiallyasitisamoleculeactingviathecentralnervoussystem,volunteerselection,inclusion

15

and follow‐up did not include a neuropsychological assessment with clinicalinterview and cognitive tests. Such assessments have been carried out usingsuitablescalesamongothersforsomeotherFAAHinhibitors.ThispointisoneofthereasonsbehindtheTSSC'srecommendations.

‐ Thedoseescalationmethod(arithmetic,geometricandFibonaccisequences,etc.)in Phase 1 trials is based neither on a consensus nor on clearly establishedinternational recommendations. In the case of the trial in Rennes, the doseescalationappearstobeclearlyproblematicastooabruptattheendofescalationwhereascommonsensewouldbeinfavourofthereverse.Forexample,thedosejumpbetweenMADcohorts4and5(20to50mg)isequivalenttoaratioof2.5whereasitisonly2(2.5to5mg)betweencohorts1and2,zoneinwhichtheriskcan be considered to be low due to the minor difference in absolute valuebetweenthe twodoses.Thesequenceselected inRennes isbasicallygeometric(inessence lesscautious thanarithmeticandFibonacci sequences)witha ratio(multiplyingfactor)of2,alsonotrespectedbetween20and50mg.ThistypeofsequencedoesnotappeartohavebeenusedintrialsonotherFAAHinhibitors.Itappeared to be important to the TSSC that the scientific and regulatorycommunity address this aspect, a major element of volunteer safety (seerecommendationsattheendofthereport).

‐ AdifficultpointtodecideoninthecaseofthetrialinRennesisthatofthechoiceofmaximumdosetobetestedinthevolunteers.Itwassetat100mg,whetherinsingleormultipledoses.Wesawthat thischoicewas theoretically logical ifweextrapolatethedatameasuredinanimalstohumans:extrapolationoftheNOAELresultsinaverycloseestimation(96mg)andthatoftheconcentrationinhibitingFAAHleadstoadoseofbetween10and40mg.Asitisaproductnotconsideredto be at risk according to generally established criteria (especially by theEuropeanMedicinesAgency),useoftheNOAELinthemostsensitivespeciesandnottheMABEL(MinimalAnticipatedBiologicalEffectLevel)was legally justified.Theproblemcomes from theapparently verydifferent response inhumansonthis lastpoint. Ineffect,weseethatFAAHinhibition(allegedmechanismof thepharmacologicalactivityofBIA10‐2474)isachievedinhumansat1.25mgandisalmostcompleteat5mg.Intheseconditions,thechoiceof100mgistantamounttotestingadose20to50timeshigherthanthatpresumedtobeeffective,whichseems absolutely excessive, even if the rule, which stipulates we should notexceedadoseequivalenttotheNOAELinfirst‐in‐humantrials,wasobservedinthis case. This major safety issue could not be anticipated when the trial wasapproved or when it started (only data in animals was known). However, itwouldhavebeenlogicalandexpectedthatthedoseescalationplanbereviewedinthelightofthepharmacokineticdatacollectedfromthevolunteers,as itwascollectedforotherFAAHinhibitors.Thiswasnotthecaseinthisstudy.

9.SymptomsobservedinthehospitalisedvolunteersFor obvious reasons relating to protection of privacy and medical secrecy, theinformation in this report only covers elements useful to the TSSC's mission and ispresentedinsuchawaytoprotecttheidentityofthevolunteers'.

16

OneoftherecordsforthesixvolunteersinMADcohort5havingbeenexposedtoBIA10‐2474wasnotselectedasthepatient,duringaroutineexamination,complainedofnospecific symptoms and their MRI scan only revealed an image interpreted as an"incidentaloma". 9.1.ClinicalsymptomsThefirstvolunteerwashospitalisedintheeveningof10January2016,dayofthefifthadministrationof the investigational product. Twoother volunteerswerehospitalisedon11January(dayofthesixthadministration),twootherson12January(dayafterthelastadministration)andthe lastvolunteeron13January, thereforetwodaysafterthelastadministration.Themainclinicalsymptomsobservedwere:

‐ headaches, in all five volunteers, very severe in one but not occurring as athunderclapheadache,

‐ cerebellarsyndromeinthreevolunteers,‐ consciousness disorders (in three volunteers) ranging from sedation to coma

(deceasedvolunteer),‐ memoryimpairmentintwovolunteers.

Other symptoms were only noted once: diplopia, paraesthesia of the thighs, andhemiparesiswith"tremor"ofonesideofthebody,withoutpyramidalsyndrome,spinepainandstiffness.The initial clinicalpictureworsened in threevolunteers.The first subjecthospitalisedprogressedtobraindeaththreedaysafteronsetof thesymptoms.Theclinicalpictureworsenedintheothertwooverthreetofourdaysbeforestabilising(overtwotothreedays), and improving. These two volunteers still had symptoms however (essentiallycerebellarandmnestic)whentheyleftRennesTeachingHospital.Thetwovolunteersinwhomsymptomsweremild,orduetothis,hardtointerpret,didnotseeanyaggravationandtheywereabletoleavethehospitalwithoutanyapparentsequelae.The four volunteers (except the one that died) were treated from 13 January withmethylprednisolone (Solu‐medrol°) at 1g/24h,without it being possible to determinewhetherthisstrongcorticotherapyplayedaroleinimprovementoftheclinicalpicture.The clinical symptoms in the five subjectswere remarkable in that theywere purelyneurological,suggestingthatonlythecentralnervoussystemwasaffected(butwithnoseizures),withoutanyothersymptomstosuggestotherorgandamageandwithouttheleast sign of infection. The bradycardia episodes and hemodynamic instability in thedeceasedsubjectaside,nocardiacorbloodpressureanomalieswerenoted.Theprogressive, isolatedprofileoftheneurologicalsymptomsgoesagainstavascular,tumoural or infectious process, and is hardly compatible with an inflammatory,metabolicortoxicprocess.

17

9.2.Bloodandcerebrospinalfluid(CSF)testsNo metabolic anomalies or abnormal blood test results were observed in the fivevolunteers.Theimmunologicalassessmentinthemostseverelyaffectedvolunteerwasnegative.The CSF was tested in three subjects. It was normal in one of them, seat of isolatedincreasedspinalfluidproteininanother,andhighlybutnon‐specificallyabnormalinthemostseverelyaffectedvolunteer,withincreasedspinalfluidproteinandthepresenceofneutrophils,possiblysuggestingtissueinflammationornecrosis.9.3.MRIresultsIn one of the volunteers, the MRI‐scan performed 24 h after administration of theinvestigational product, only revealed a minor punctiform hypersignal of the righthippocampalbody,whichwasnolongerseenontheMRI‐scanstwoandfourdayslater.For the other four volunteers, the MRI‐scan showed anomalies of highly variableintensity, affecting thehippocampus and thepons (protuberance)predominant in theanteriorpart(extendingattimestothebulbortothemesencephalon),bilaterallyandsymmetrically. In the deceased volunteer, and only in this person, theMRI‐scan twodays after the initial examination, showed involvement of the thalamus and of thecerebralcortex.The anomalies observed in the four volunteerswere of the same typebut of variableseverity,with:

‐ 1.Diffusionhypersignalanddropintheapparentdiffusioncoefficient(ADC)seenonce, and indicating water diffusion restriction (possibly suggesting cytotoxicoedema, regardless of the cause, but also inflammatory cell infiltration forinstance).Inthemostseverelyaffectedvolunteer,thediffusionhypersignalcamewith an increase in ADC in the posterior part of the brain stem, which couldsuggestvasogenicoedema.

‐ 2. Hyposignal in SWI (Susceptibility Weighted Imaging), interpreted in thiscontext as suggesting the presence of blood (haemoglobin), in the form ofmultiple,small,roundedhyposignalsreflectingmicrobleeds.

‐ 3.Non‐specific,FLAIR(FluidAttenuatedInversionRecovery)hypersignal,possiblyrelatedtoanincreaseinwatercontent,demyelination,gliosisornecrosis.

Thesesignalanomaliesareidenticalinthehippocampusandponsinthefourvolunteers,andinthecerebralcortexinthemostseverelyaffectedofthem.Thissuggeststhatthecortical damage was caused by the same mechanism and not by anoxia from thebradycardiaepisodes.Thesignalanomaliesobservedreflectthereforethepresenceofmicrostructuralchangeswithavascularcomponent(microbleeds)whicharenotspecifictoagivenmechanism.However, theirbilateral andsymmetrical topographyand thepresenceofhyposignalsveryearlyoninSWI,theoreticallymeanwecanruleoutaninflammatoryprocess.Thehighlyunusual topographymakesaprimarymicrovascularmechanismunlikelyand ismoresocompatiblewithatoxic/metabolicprocess.

18

9.4.CommentsThecircumstancesofoccurrence,thestereotypicalnatureandprogressiveprofileoftheneurological symptoms strongly suggest that the product tested caused them. TheanomaliesseenontheMRI‐scan,whichareconsistentwiththeclinicalsymptoms,alsostronglysuggestatoxicormetabolicmechanism,giventheirsignalcharacteristicsandtheirbilateralandsymmetricaltopography.Imputabilityoftheinvestigationalproducttotheneurologicaldamagethereforeappearstobeclear.The highly unusual topography (and perhaps thus far unique) of the lesions(hippocampus,pons,thalamus,cortex)isakeyelementinidentifyingthemechanismbywhichthesubstancetestedcouldhavecausedsucheffects.10.DetectionofsignsoftoxicityintheothervolunteersOne of themost striking elements of the BIA 10‐2474 case is the absence of toxicity(adverse event of noteworthy intensity, a fortiori serious), in particular neurological,observed in the trial volunteers other than those in MAD cohort 5. This, despiteadministrationofsingledosesofupto100mgor10multipleadministrationsofupto20 mg/24h, for a total dose of 200 mg (note: the total doses in the hospitalisedvolunteersrangedfrom250to300mg).Amongthe76volunteers(exceptMADcohorts)havingreceivedtheactivetreatment,18adverse events were observed, 11 of which (frequency: 14.5%) were cardiovascular(orthostatic hypotension, reflex tachycardia, PR or QT interval prolongation on theelectrocardiogram,etc.),andtherewerecasesofmilddizzinessorheadaches.The observations were of the same type for the volunteers in the MAD cohorts, noevents of noteworthy seriousness or severity, and cardiovascular symptoms beingpredominant. It should be noted however that two volunteers from the 10 mg MADcohort presented with blurredvision on two occasions. These episodes, which lastedbetween ten and thirty minutes each time, cannot, as is, be qualified as diplopia, asymptomwhich corresponds to a specific definition in neurology. Therefore, the trialinvestigatorandthemonitoringcommitteedidnotconsiderthissymptomtoberelevant,andalsoitwasnotobservedinthevolunteersinthecohortsexposedtohigherdoses.Since suspension of the trial, all volunteers towhom BIA 10‐2474was administered,have been contacted for a full clinical assessment and brain MRI‐scan. To date, noclinical or biological or MRI anomalies likely to be related to administration of themoleculeorthetrialconditionshavebeendetected.Thesafetydatafromfirst‐in‐humanandPhase1trialsofotherFAAHinhibitorsseemstohavethesamequalitativeprofile(sedation,digestivedisorders,orthostatichypotension,dizziness, etc.) with one difference, however, in the frequency of the adverse effects,whichalthoughtheyvarygreatlyfromoneproducttoanother,generallyappearstobelowerthaninthecaseofBIA10‐2474.

19

11.PharmacokineticdataGenerally,pharmacokineticstudiesconductedinanimalsdonotgiverisetoanyspecificremarks, even if as is fairly often the case, pharmacokinetics appear to become non‐proportional with the highest doses, at least in dogs andmonkeys. For example, theratios of the areas under the curve (AUC) and of the doses administered, which aresupposedtoremainconstant,were960at0.1mg/Kgand1,886at1mg/Kg.A studyof (oral and intravenous) administrationof the radiolabelledproduct showedtheverygoodtissuediffusionofBIA10‐2474anditsverylargevolumeofdistribution.ThepharmacokineticstudiesconductedonthevolunteersfromtheSADcohortsshowedthat the elimination half‐life of BIA 10‐2474 is gradually extended when dosesadministeredbecomehigh; the areasunder the curve (AUC), reflectingexposure, alsoincreased more rapidly than the doses increased. This, from a purely theoreticalstandpoint, could be explained by the acceleration in absorption beyond a certainthreshold(ofthebarrierbreach,facilitationofpassage,carrierinductiontype,etc.),or,alotmorelikely,bysaturationofeliminationprocessesforadoseofbetween40and100mg,without it beingpossible tomore accurately identify the thresholddose atwhichnon‐proportionalitybegins.During MAD studies, the same phenomenon was observed, the AUC increasing morerapidlythanthedosesfrom20mg.Weespeciallyseethat:

‐ Dispersion in the pharmacokinetic parameters among the volunteers had astronginfluenceonindividualAUCvalues,whichwerelargerat50mgthanat20mg.

‐ Againfor50mg,andlikewhatwaslessclearlyobservedfor20mg,residualBIA10‐2474plasmaconcentrationscontinuedtoincreaseslightlyuptothefifthandsixthadministration.Theplasmaconcentrationsteadystatedidnotappeartobereached for some subjects in MAD cohort 5. This strongly suggests that theeliminationhalf‐lifeofBIA10‐2474waslongerinthesesubjectsthanestimatesmade from single doses or lower multiple doses (10 mg). For instance, theeliminationhalf‐lifecalculatedinthe100mgSADgroupwasaround10hoursonaverage,avaluetheoreticallyincompatiblewithwhatwasobservedinthe20and50mgMADcohorts.

‐ AsinSAD,non‐proportionalityisthereforelikelyasof50mgmultipledoses.Thefourmetabolitesidentifiedinanimalsareexpectedtobethesameinhumans,twoofthem (BIA10‐2639andBIA10‐2445) reachedmeasurableplasma concentrationsbutremained however very low (<3% of those of the parent product). Without directlyadministering the metabolites themselves, it is difficult to determine their individualcharacteristics. However, it seems that the variability in the pharmacokineticparameterswashigher for these twometabolites than thatobserved inanimals,with,forexample,aneliminationhalf‐lifeestimated(highlyapproximatelygiventheverylowconcentrations)tovaryfrom4to23hours.Variability also affected, but to a lesser extent, the pharmacokinetics of themoleculeitself. This is commonly observed with medicinal products due to interindividualvariations in metabolism, among others; some (qualified as outliers) can have verydifferentparametersfromtheothermembersofthegroup.InthecaseofaPhase1trial,variabilitycanbecomeproblematicifthedosecalculationsarebased,asisthecasehere,on themeans of the key parametersmeasured in a small number of volunteers fromlowerdoseslevels.Bydefinition,suchanapproachdoesnottakeaccountoftheextreme

20

values in subjects expressing a specific response, distribution ofwhich can vary fromone group to another. This is likely to induce fairly significant prediction errors (seerecommendationsattheendofthereport).12.HypothesesstudiedinanattempttoexplaintheaccidentinRennesThe accident that occurred during the trial on BIA 10‐2474 at the Biotrial centre inRennesisunquestionablyastonishingandunprecedentedin:

‐ its seriousness (several volunteers from the same cohort having to behospitalised,oneofthebeingdeceasedinthedaysfollowingadmission),

‐ the fact that apparently, the toxicology studies, although conducted on fouranimalspeciesup toveryhighdoses,didnot revealany lesionsorpicture thatcouldpredictsuchaspecificneurologicaltoxicity,

‐ the highly unusual nature of the clinical and radiological picture of the braindamageobservedinseveralvolunteersinMADcohort5,unlikeanythingseentodate,

‐ the fact thatnopatentneurologicalor radiological signsof this typehavebeenfound in theothervolunteers (somehavingabsorbedup to100mg ina singledoseortotaldoseof200mgover10days),

‐ finally, the fact that the accident occurred with a molecule similar to othercompounds (around ten) the development of several ofwhichwas abandoneddue to insufficient efficacy, without any neurological or other toxicity beingobserved.

Intermsofformallogic,thefactthattoxicityonlyoccurredinoneofthe14cohortsofvolunteershavingreceivedBIA‐2474,canonlybeexplainedby:

‐ anadministrationerrororprocedurespecifictothiscohort,‐ acommonfeatureamongthesixsubjectshavingpresentedwithsignsoftoxicity,‐ aneffectrelatingtothecumulativeBIA10‐2474dosethatthesubjectsreceived.

Explorationof the firsthypothesisdidnot fallwithin thescopeof theTSSC'smissionsbut this hypothesis seems very little likely. Procedures for controlling the doseadministeredareverytightinPhase1trials.Also,theproductcontainedinthecapsulesadministered to all volunteer groups was the same as that used for the toxicologystudiesandwaslatertestedandrevealedtobeofveryhighpurity.TheTSSCthereforemainlyfocussedontheothertwohypotheses.12.1.HypothesisofacommonfeatureamongthevolunteersinthefifthMADcohortSeveralpossibilitieswerediscussed:12.1.1. HypothesisofinfectiouscontaminationThecloselivingconditionsofthevolunteersinthesamecohortandconsumptionofthesame foods several times a day could support this hypothesis, especially as certaininfections,possibleinthiscontext,canbeexpressedbythecentralnervoussystem.Forexample,thereisarhombencephalicformoflisteriosiswithlesionslocatedinthesameareas as those observed in Rennes. This hypothesis is however very little plausible.

21

Firstly,thisclinicalformisveryrareinhumans(themostcommonneurologicalformoflisteriosisbeingmeningoencephalitis).Itwouldhavethenhadtoonlyaffectvolunteersexposedtotheactivetreatment.Thetwoconditionsmakeupahighlyunlikelyscenarioin statistics terms. Finally, and above all, analysis of the symptoms presented by thevolunteersandoftheMRIimagesdoesnotfitwiththishypothesis.12.1.2.HypothesisofaninteractionwithotherproductsMentioned several times in the days following the announcement of the accident inRennes, an interaction with medicinal products, foods (such as chocolate) orrecreationalsubstances(alcohol,narcoticsincludingcannabis,etc.)couldhaveoccurred.The "medicinal products" hypothesis appears to be resolutely unlikely given Phase 1goodpractices.Especiallyasthesixsubjectshospitalisedwouldhavetohavetakenoneorseveralofthesamemedicinalproductseventhoughtheywereofdifferentages(27to49years)andhadverydifferentprofiles.Pharmacodynamic or pharmacokinetic interaction betweenBIA 10‐2474 or one of itsmetabolitesandafoodalsoappearslittleplausible.Analysisofliteraturedidnotrevealany examples of central nervous system toxicity, with symptoms suggesting thatobserved inRennes,causedbydrug/food interactionofany type.Thesameapplies toconsumptionofchocolatebythevolunteers(astheaccidentoccurredinthetwoweeksafter Christmas). Chocolate only contains small quantities of anandamide. Overall,hyperstimulationoftheendocannabinoidsystemisnotknowntogenerateapictureofthetypeobservedinRennes(seefurtheron).Modification of the BIA metabolism by food or beverages (i.e. acting as inhibitor orinducerofthecytochromesP450pathway)cannot,perse,beruledout,butseemslittleplausible.Todate,thereisnoevidencetosupportthehypothesisofconsumptionofnarcoticsbythevolunteers,immediatelybeforeorduringthestayattheBiotrialcentre.Theresultsofongoinginspectionsandinvestigationswillfromthisstandpointbeabletoinvalidateor confirm such a hypothesis. However, besides the serious breach of Phase 1 goodpractices(intermsofvolunteermonitoringespecially),thatsuch,theoreticallycollective,consumptionwouldrepresent(allMADcohort5volunteershospitalisedwouldhavetohavetakenthesamesubstanceandnonefrompreviousdoselevelswouldhavetohavetakenit),thishypothesiscomesupagainstthepreviousargument.Iteffectivelyseemstobe accepted by neuroscientists, that direct or indirect, even massive stimulation ofendocannabinoid receptors, CB1 in particular, would not induce toxicity such as thatseeninRennes.Evenifincertainsubjectstheycaninduceseverepsychiatriceffects(i.e.psychotic episode), neither cannabis, nor its main active substance,tetrahydrocannabinolleadtoacutetoxicbraindamage,evenexperimentallyandatveryhigh doses. Clearly, such consumption could at best be considered a cofactor, butcertainlynotatriggeringfactorandevenlesssothecauseoftheaccident.12.1.3. Hypothesis of a specific genetic or metabolic characteristic or commonpharmacologicalresponseamongthesubjectsinthefifthMADcohortThere are several genetic factors, among others, likely to modulate individualsusceptibilitytoadministrationofanFAAHinhibitor.Forexample,aswehaveseen,thishydrolase exists as two isoforms (FAAH‐1 and FAAH‐2)with different activity. In thesameway,thecytochromeP450systemisfoundatseverallevels,theactivityofwhichcanvarysignificantlyfromoneindividualtoanother.Asappealingasitmayseem,this

22

hypothesis clashes with statistics laws. For the FAAH example, if we consider thatprevalenceofcarriersofbothisoformsis38%inthegeneralpopulation,theprobabilityoffindingitin5outofthe6exposedcohortsubjectswasalessthan3in100chanceand3 in1000 chance in the six subjects exposed. Ifwe recalculate,whichwouldbemorerelevant,notwith38%but5%(prevalenceofcarriersofthelowactivityisoformonly),the results thus completely rule out such a possibility. The same applies to theprobabilityofhavingincludedbyaccidentamajorityofrapidmetabolisersubjectsinaprevious cohort, which could have biased the pharmacokinetic predictions forcalculationofthedosetobeadministeredtovolunteersinMADcohort5.Itisthereforeevidentthatthesymptomspresentedbythevolunteersincohort5canonlyberelatedtothecumulativedoseofBIA10‐2474administeredtotheminmultipledailydoses.12.2.HypothesesofathresholdeffectrelatingtototalBIA10‐2474doseEvenifthissecondsetofhypothesesappearsalotmorelikely,thepotentialmechanismsare especially numerous, some remaining purely hypothetical. They may involve themoleculeitselfand/oramediatorsuchasanandamide.Let's not first of all forget the highly unusual nature of this dose‐dependent toxicity,apparentlywithnoportentsignsreportedinvolunteershavingbeenexposedtolowerdoses. It happened "as if somethinggavewayor swung suddenlyata specificdose orconcentrationthreshold".Expressionof this typemaybe compatiblewith the fact thatBIA10‐2474ischaracterisedbyaverysteepdose‐effectcurve(ranging,withinafairlynarrow concentration range, fromabsence of FAAH inhibition to complete andhighlyextended inhibition) and by the fact that the pharmacokinetics of BIA become non‐proportionalinhumansfromadoseofbetween40and100mg.Let'salsonotforgetthatBIAhasthecharacteristicsofamoleculewiththeabilitytobind(and, therefore, potentially inhibit) cerebral hydrolases other than that which is itspharmacological target. The specificity ofBIA10‐2474 for FAAH is clearly lower thanthatofotherinhibitorsdevelopedpreviously.ItwasalsoadministeredtothevolunteersinMADcohort5atdosesaround10timeshigherthanthatwhichappearstocompletelyinhibitFAAHinhumansforaverylongperiod.It is thereforehighly likely,not to sayalmost certain, that themechanismcausing theaccidentinRennesshouldbelookedforoutsidetheendocannabinoidsystem,especiallyasstimulationofendocannabinoidreceptorsbyanandamidecannottheoreticallycausethistypeoftoxicity.Severalmechanismscanbediscussed:12.2.1.InhibitionofothercerebralhydrolasesbyBIA10‐2474Thisavenueandthenext(see12.2.2)shouldbe favoured inthe initialanalysisduetotheirbiologicalplausibility.Let'snot forget thatBIA10‐2474wasadministered to thevolunteersinMADcohort5atadose(50mg)probablymorethan10timeshigherthanthat fully inhibiting FAAH activity. Increasing tissue concentrations beyond thosealready completely inhibiting the enzyme cannot, in any circumstances, increase thepharmacologicaleffect,buthaseverychanceofpromoting(especiallywithsuchalittle‐specificproduct)bindingtootherserinehydrolases.Thiscould,viaadirectorindirectmechanism(unknowntodate),bethecauseofthebraindamageobservedinRennes.

23

It is interesting to note that such damage was not apparently observed in animals,despitetheuseofveryhighdoses,orinthesixvolunteershavingreceivedasingle100mgdoseofBIA.Itoccurredfromdayfiveofdailyadministrationof50mg.Thisfindingisbynomeansincompatiblewiththehypothesis:

‐ themechanisminvolvedcouldbeattributedto,asisfairlycommonintoxicologyandpharmacovigilance(i.e.cumulativetoxicity),tolateexpression,

‐ the specific pharmacokinetic features of the molecule (non‐proportionalpharmacokinetics, largevolumeofdistribution)describedabove,makegradualaccumulationofthemoleculeinthebraintissueplausible.Concentrationscouldhavereachedatriggerpointondayfiveorsixofadministration.

The difference in animal/human susceptibility seems more difficult to explain. Suchfeatures,onewayor theother,havealreadybeenobservedhowever. It isnotamootpointeithertonotethatthedosesrequiredtoachieveFAAHinhibitionwerearound10timeslowerinhumansthanthosereachedduringthestudiesinanimals.12.2.2.BIA10‐2474toxicityviaanothermechanismThestartingpoint,asinthepreviousscenario,isbelievedtobeanexcessoffreeBIA10‐2474 inbrain tissue related (1) to a toohighdose administered for endocannabinoidFAAHblockade,and,undoubtedly,(2)tothemolecule'spharmacokineticcharacteristics(kinetics becoming non‐proportional, possible accumulation in tissue from multipledoses, etc.). The difference here is that the pathogenic mechanism is not believed tooccurbyinhibitionofotherserinehydrolasesbutbyinsitu toxicityofthemoleculeonthe cell structures or proteins involved, for instance, in exchanges in the blood‐brainbarrier.Acrediblehypothesistargetstheimidazole‐pyridine"leaving"groupofBIA10‐2474whichcouldeitherbindtobrainproteinsorinhibitthecytochromesP450‐epoxidesystembyleadingtovasospasm.Ineffect,inthecaseofBIA10‐2474,andunlikeotherFAAHinhibitors,theimidazolenucleusisinadjacentpositiontotheelectrophiliccarbon,thesiteofbindingwithFAAH.Asbefore,thishypothesiscomesupagainstthefactthatno toxicityof this typewasobserved inanimals,despiteadministrationofhighdoses.Theresponsestothisobjectionarethesameasabove.12.2.3.ToxicityfromaBIA10‐2474metaboliteToxicity fromoneof the fourperipherally‐circulatingmetabolites (plasma) inhumansandanimalscouldalsobeenvisaged.TheirspecifictoxicityhasnotbeentestedbyBial,howeverthesemetabolitesareproducedinverysmallquantities(<3%ofBIA10‐2474circulating concentration) even if pharmacokinetic variability seems to be higher inhumans than inanimals.All of theknownBIAmetaboliteshavea structure similar tothatofthemothermolecule,andthereisnoevidenceinthedossiertosupportspecifictoxicity from any one of them. It is also possible (as is often the case duringmetabolisationprocesses) that thesederivativesaremorehydrophilic thantheparentproduct. This would tend to make crossing of the blood‐brain barrier more difficult,unlessitispresumedthatthereisaspecificcarrierand/oreffluxpumpinhibitionduringthe rise in circulating concentrations from multiple doses. This assertion shouldhoweverbeweighedagainstthefactthatBialsaiditplannedtodevelopseveralofthesemetabolitesasFAAHinhibitors,whichtheoreticallysuggeststhattheycrosstheblood‐brainbarrierinsignificantamounts.

24

The purely theoretical hypothesis of a metabolite with strong tissue tropism, non‐quantified inhumansdue to theveryhigh tissue/plasma ratio andvery low resultingplasmaconcentrationsremainstobeenvisaged.Finally, theeffectofgeneticpolymorphism, leading forexample to theproductionofametaboliteinlargerquantitiesinsomeindividuals,wouldhavetoberuledoutitseemsforstatisticsreasons(ifthisexists,itwouldhavetohavebeenpresentinallhospitalisedvolunteersinMADcohort5).12.2.4.Suspectedanandamide‐relatedtoxiceffectsFAAH activity blockade leads, as we have seen, to an increase in intracerebralanandamideconcentrations,whichmakesitpossibletoenvisageseveralpossibilities:

‐ AnandamidebindingtootherreceptorsAnandamide is a mediator, the ubiquity of which largely exceeds theendocannabinoidsystem. It isable,especiallywhen itsconcentrations increase,tointeractwithseveraltypesofreceptors(atleastTRPV1,PPARandNMDA)andwith theMAP‐kinasepathway, having possible consequences on apoptosis andneuroprotection.

‐ ToxicityfromanandamidedegradationproductsIn the event of complete and prolonged FAAH inhibition, anandamide can bedegraded by the cyclooxygenases pathway, giving rise to various compounds(leukotrienes and prostanoids) some ofwhich have known effects on cerebralvasomotricity,whichmaybecompatiblewithsomeofthelesionsobservedintheMADcohort5volunteers.

Theplausibilityofthelasttwohypothesesishoweverstronglychallengedbyaseriousofcounterarguments:

‐ Administration,includingathighdoses,ofanandamideoritsnon‐metabolisableanalogues(i.e.:methanandamide)toanimals,isnotknowntoinduceneurologictoxicity,atleastofthetypeexaminedinthisreport.

‐ Thistypeoftoxicityhasnotbeenobserved(inhumansorinanimals)withothertheoretically more specific FAAH inhibitors, including with those described asirreversible,evenduringadministrationofhighmultipledoses.

‐ CompleteandlastingFAAHinhibition,andapparentlythereforetheintracerebralanandamideconcentrationplateau,wouldappeartobereachedinhumansfromBIA 10‐2474 doses of around 5 mg, whereas no neurological toxicity wasobserved in the volunteers having receivedmultiple doses (10 days), of up tofourtimesthatdose(20mg).

13.TSSCconclusionsTheaccidentaffectingseveralvolunteersinthetrialconductedbyBiotrialwouldclearlyappeartoberelatedtothemoleculetested.It isvery little likelythatthetoxicityberelatedtostimulationoftheendocannabinoidsystemviaFAAH(pharmacologicaltargetofBIA10‐2474)inhibition.Atoxicmechanismrelatingtotheincreaseinintracerebralanandamideconcentrationscanalsoapparentlyberuledout.

25

In the current state of knowledge, imputability to one of the known BIA 10‐2474metabolitesdoesnotappeartobefavouredeither.WeshouldnotehoweverthetimetoonsetofthetoxiceffectsobservedinMADcohort5whichcouldbecompatiblewiththeproductionofametabolitewhich,duetothelongereliminationhalf‐lifethanthatoftheparent compound, could accumulate in the tissue compartment gradually duringadministrationuntiltheconcentrationreachesatriggerpoint.Notwithstanding, themost likely hypothesis to date is that of toxicity specific to themoleculeviaitsbindingtootherbraincellstructures,facilitatedby:

‐ itslowspecificityforitstargetenzyme,‐ use of multiple doses a lot higher than those leading (at least in humans) to

completeandlastingFAAHinhibition,and‐ its probable gradual accumulation in the brain, undoubtedly related to the

specific pharmacokinetic features ofBIA10‐2474.This is likely to explainwhythe accident inRennes only occurred onday five of administration of a 50mgdose,andnot in thevolunteershavingreceivedasingledose thatwas twiceashigh.

Atthisstage,itisdifficulttofavouronetoxicitymechanismoutofthetwomostlikely:inhibition of other serine hydrolases, or harmful effect from the imidazole‐pyridine"leaving"group.Thefactthatthistypeoftoxicitywasnotobservedinanimalsdespiteadministrationofveryhighdoses,remainsunexplainedsofar.WeshouldnotehoweverthatBIA10‐2474isaround10timesmoreactiveinhumansthaninanimalsintermsofFAAHinhibition.The sudden onset of the toxic symptoms could be related to BIA 10‐2474 beingcharacterised, with hindsight, as "little manageable" due to relative low efficacy(inhibitory concentrations in themicromolar range), lowspecificity andaparticularlysteep concentration‐effect curve. In these conditions, the little comprehensibleaccelerationindoseescalationinthe20and50mgMADcohortsprobablysignificantlycontributedtotheaccident.Withthisinmind,suchanincreaseindosewasallthemorerisky given that due to cohort sequence timing and the time required for tests to becarried out, the latest pharmacokinetic data available was that of the 10 mg cohortsubjects. Such a sequence made it almost impossible to adjust the dose to beadministered in the light of the emerging non‐proportionality. This would have beenevenmoreproblematicforthelastdoseinitiallyplanned,thatof100mg(doseforwhichthe pharmacokinetics are probably clearly non‐proportional) since adjustment forsafetyreasonswouldhavebeenbasedonthe20mgcohortdata.ItwasnotwithintheTSSC'sscope(unlikethetwoongoinginspections)tocommentonthebasis to thetrialauthorisation issuedbytheANSMfollowingtheopinionfromtheBrestethicscommittee.Inscientificterms,theTSSCconsidershoweverthat:

‐ BIA 10‐2474 could not, theoretically, be considered as an at‐risk product,accordingtothetermsofapplicablerecommendationsespeciallytheGuidelineonstrategies to identify and mitigate risks in first in human clinical trials withinvestigationalmedicinalproducts (Committee forMedicinalProducts forHumanuse,CHMP,EMA,2007).

‐ The data provided, especially the Investigator Brochure, did not containinformation,especiallydataontoxicology,suggestingaspecificriskduringfirst‐in‐human use. We should recall however that the brochure contains many

26

mistakes, inaccuracies, figure inversions or incorrect translation of sourcedocuments, making understanding difficult in several aspects. This is highlysurprisinggiventheregulatoryimportanceofthisdocument.

14.TSSC'srecommendationsfortheconductoffirst‐in‐humantrialsThe seriousness of the accident in Rennes calls for international legislation and goodpracticeson first‐in‐humantrials tobeamended inseveralareas. Ineffect,even if theBIA10‐2474dossierandprotocolofthetrialconductedbyBiotrialareinkeepingwithapplicable provisions and recommendations, it's more in the rules than in the mind.Meetingregulatoryrequirementsshouldforegoneitherthebasestopharmacologyandclinicalpractice,northeultimatetherapeuticaimofdrugdevelopment.Thesupremacyofrulesovercommonsenseandscientificlogicsuggestapotentiallydangerouschangeof direction and calls for collective awareness to be raised. The accident in Rennesillustratesthisinatragicmanner.The TSSC therefore puts forward six recommendations that it would like to seeexamined by European and international regulatory bodies, and by other relevantassociationsandbodies.

‐ 1. First of all, a medicinal product is developed in the ultimate aim ofdemonstrating its therapeutic effectiveness and its utility to public health.Therefore,justificationanddemonstrationofpharmacologicalactivitypredictiveofefficacyinhumanscannotbeconsideredtobesecondary.InthecaseofBIAforexample, out of the 63 pages of the Investigator Brochure summarizing thepreclinical data, fewer than two discuss demonstration of pharmacologicalactivityfortheapparentlyplannedindication.Thetestsdidnotmakeitpossibletodetermineaneffectivedoseeitherbefore lengthy, costly andnever risk‐freepreclinicalandclinicaldevelopmenttookplace.ThetestsrevealedBIA10‐2474to be a product with, at best, potential moderate efficacy in the plannedindication, and in any case clearly lower efficacy than the comparator product(datadeletedfromtheInvestigatorBrochure).Aprerequisiteessentialtoanyformofclinicaldevelopment,wouldbetoconductsufficiently comprehensive preclinical pharmacology studies, comparativewheneverpossible,onasufficientlybroaddoserange(toestablishadose‐effectcurve where appropriate) so as to be reasonably predictive of real‐life, futuretherapeutic efficacy. Prior justification should be clearly emphasised by thesponsor and studied as a priority during the course of initial opinion requests(ethicscommittees)andauthorisationapplications.

‐ 2. A neuropsychological assessment with clinical interview and cognitive testsshouldbeacompulsorypartofassessmentduringvolunteerscreening,inclusionandclinicalmonitoringinaPhase1trialfordrugswith"centralnervoussystem"tropism.This(notprovidedforintheRennesprotocol)couldidentifypotentiallyat‐risk subjects and to detect behavioural changes or neuropsychologicaldisordersearlyonduringexposuretotheinvestigationalproduct.

27

‐ 3.All first‐in‐humanandPhase1protocolsshould,unlessunnecessary,providefor the doses to be tested in volunteers to be adjusted according to the datacollected in volunteers already having been exposed during the trial. Thisobviouslyconcerns(as isusually thecase,especially in theprotocol inRennes)doseadjustmentaccordingtothepharmacokineticparametersfromthepreviousdose level (in fact, fromleveln‐2 in thecase inRennes). It shouldalsoconcernthepharmacodynamicdata. InthecaseofBIA10‐2474,if itwasconfirmedthatFAAH inhibitory concentrationwas10 times lower inhumans than in animals,the choice of the 100 mg highest dose (20 times the dose inducing completeinhibition) was no longer justified and could carry a risk. Concerning doseadjustment according to thepharmacokinetic parametersmeasured, variabilityand its extremes, and not only themean of these parameters, should be takenintoaccountinthecalculationsinordertoprovideforaworstcasescenario.

‐ 4. During first‐in‐human and Phase 1 trials, volunteer safety should take

precedence over any practical, economic or regulatory considerations. To thiseffect, pluridisciplinary work at international level is required to redefinemethodology options that provide for both an acceptable study duration andoptimal levelofsafety.Forexample,asforsingle‐dosestudypractices,thedoseadministration sequence could be transferred to MAD so as not to expose allsubjects fromthesamecohortatthesametime. Inthesameway,thetimingofthe various cohorts should make it possible to have the pharmacokineticparametersfromsubjectsfromthedoselevelimmediatelybelow(n‐1);alargerjump may be problematic in the event of non‐proportionality of thepharmacokineticswiththedoseadministered.

‐ 5. Dose escalation strategies in first‐in‐human and Phase 1 trials should takeaccountofconsiderationsbasedoncommonclinicalandpharmacologicalsense.For example, in the text by the European agency cited above, it only says thefollowing: "Dose increase should proceed with caution taking into accountidentifiedrisk factors fromnon‐clinicalstudies". Aswe have seen, the triggeringfactor in theaccident inRennescouldbe thechoiceofdose tobe testedwhichwas too high, given the new data collected in humans and more rapid doseincrease within the potential risk area. With this in mind, the geometricsequencing(especiallyof2ormore)maintaineduntiltheendofdoseescalationdidnotappeartobereasonable.TheTSSCthereforerecommendsthatgeometricsequencingbeavoidedwherepossible,oratleasttheratiobereducedattheendofescalation.

‐ 6. Finally, the TSSC would like, notwithstanding industrial property

considerations, to see a debate opened at European and international level, onaccess todata fromongoingorprevious first‐in‐humanandPhase1 trials.Thiswould unquestionably be an advance in terms of the protection of subjectsparticipatinginbiomedicalresearch.Forexample,comparisonwiththeprotocolsof studiesonproductsdevelopedpreviously,oreasieraccess to toxicologyandclinical safety datawould enable highly useful comparative analysis, especiallywhenanalysingprotocolsinviewofissuingopinionsorauthorisations.

28

VersionapprovedbythemembersoftheTSSCbeforesubmissiontotheANSM'sDirectorGeneral(MrDominiqueMartin)onMonday18April2016.