These slides were presented at h/ps:// ...

Theseslideswerepresentedath/ps://www.pmwcintl.com/cur7s-bagne-2018mich/.

Youwilllearnhowtomakedrugdevelopmentmorescien7fic,precise,ethical,produc7ve,andlesscostly.

Otherpresentersatthismee7ngaredriversofprecisiondrugdevelopment.TheseincludeLeeHoodrepresen7ngsystemsbiologyandP4Medicine,FrancisCollinsrepresen7ngmoderngenomics,andEricTopolrepresen7ngthescienceofindividualityinhisbook,TheCrea(veDestruc(onofMedicine.

1

2

Prevailingrandomizedcontrolledtrialdesigns,da7ngbackto1948,wereanimportantscien7ficadvance.However,RCTdesignscomplianttoCONSORT,FDAguidelines,andPCORIMethodologyStandardsdohavefundamentalproblemsandlimita7ons.Amongtheseareconfoundingtreatmenteffectswitheffectsofindividualdifferences,includinggene7cdifferences.Eachpa7entcouldbeadifferentconfoundedmixofac7vetreatmentandindividualdifferencesresponse.Genomicsaccentuatesindividualdifferences.CurrentRCTdesignsaveragethemout.Homogenizingpersonsisan7the7caltogenomics.

Stra7fica7onhelps.However,therearemorecombina7onsofindividualdifferences,typesoftreatment,anddosesoftreatmentthantherearepersonintheworld.Youwilllearnhowtosolvethisproblemwithmorerandomiza7onandSIMA.

3

PrevailingRCTdesignsalsoconfoundtrueresponderstoac1vetreatmentwithrespondersonac1vetreatmentthatwouldhaverespondedtoplacebo.Eachpa7entcouldbeadifferentconfoundedmixofac7vetreatmentandplaceboresponse.

Inaddi7on,classical-designRCTsthatfocusonefficacyneglectsafety,precludedoseop7miza7onforindividualpa7ents,arenotwellsuitedtoaccountfordelayandpersistenceofresponse,anddonotcapitalizeonmoderndatacollec7onandprocessingcapabili7es.Youwilllearnhowtoaddresssuchproblemsasaset.

4

Confoundingcontributesto“imprecisionmedicine”asquan7fiedbyNikSchorkinNature.Hereareresultsfor10top-grossingdrugs.Thebluepersonsarehelped.Theredpersonsarenot.Imprecisiondrivesupcostsandcloudsiden7fica7onofgene7candotherpredictorsofdifferen7aldrugresponse.

5

DrugdevelopmentwithSIMAcanbesimpler,morescien7fic,andmoreprecisebymeasuringthebenefitsandharmsoftreatment.Measurementofbenefitandharmreducesthedimensionalityoftreatmentevalua7onproblems.

Randomizedcontrolledtrialscanprovidecanprovideaccurateandintegratedevalua7onsofsafetyandeffec7venessforeachperson.

Rejec7onofthenullhypothesistotherightindicatesthatbenefitsexceedharms.Rejec7onofthenullhypothesistotheleaindicatesthatharmsexceedbenefits.

Inaddi7on,SIMAprovidesscoresthatcanbeaggregatedandanalyzedsta7s7callyforpopula7onmedicine.

Thisapproachwouldhelpobviatetheclinicalresearchtoclinicalprac7cetransla7onbo/leneck.

6

SIMAisatooltoacceleratebasicandappliedsciencesofcomplexadap7vesystems.

SIMAmeasuresinterac7onsover7methatdescribeandhelppredicthowCASworkover7me.

SIMAquan7fiesedgesinnetworkgraphswheneachnodeisa7meseries.

SIMAcanbeanAItool.

TodayIfocusonmeasurementofbenefitandharmforresponse.

7

Thisisasetofthreesingle-personRCTsthatusethesametypeofdrug,thesamesetoffourdosesincludingplacebo,andthesamethreeresponsevariables.Thesearemockdatafora16weektrialwith4pairsof2-weekperiods.

Fourdoses,includingplaceboaszero-dose,wererandomizedover7meforeachofthreepa7ents.Within-personrandomiza1onofdoseseliminatesbothtypesofconfoundingshownbefore.CONSORT-compliantRCTsdonotrandomizeenough.Seethatdoseisinves1gatedasa1me-dependentdimensionalvariable,NOTacategoricalvariable.

Thissmall-scaleexamplehasonlythreeresponsevariables.Ideally,useenoughsafetyandeffec7venessresponsevariablestoobtaincomprehensiveevalua7onsofsafetyandeffec7veness.

TheseInterac7on-over-Timescores,computedbySIMA,quan7fytheamountofevidenceforinterac1onsover1me.Posi7veIoTscoresquan7fyhigherdoseswithhigherresponsevariablelevels.Nega7veIoTscoresquan7fyhigherdoseswithlowerresponsevariablelevels.

Userssettowardanduntowarddirec7oninaccordwithclinicalsignificanceandpa7entpreferences.Herehigherbloodpressureisuntoward.

Herearetheninebenefitandharmscoresinbagnez-scoreunits,threeforeachpa7ent.

Weightsalsoaresetintermsofclinicalsignificanceandpa7entpreferences.

OverallBenefitandHarmScoresareweightedaveragesforindividualpersons.

Nowcomessta7s7csaaerSIMA.Thenullhypothesisofnooverallbenefitandharmwasrejectedintheposi7veorbeneficialdirec7onwithatwo-tailedt-testonmeanoverallbenefitandharmscore.

8

Thisillustratestheamountofevidencequan7fiedforabenefitscorewithavalueof8.92bagnez-scoreunits.

9

Thisisfromacomputersimula7onthatprocesseddifferentpor7onsofadatasetcreatedbyaddingwhitenoise–randomnormaldeviates–toagivensignal.Halfoftherepeatedmeasurementswereontreatmentandhalfoff.

SeehowsignificancelevelsincreasewiththenumberofsubjectsasexpectedusingchangescoresandwithoutSIMA.

AlsoseehowsignificancelevelsincreasewithnumberofrepeatedmeasurementsandSIMA.

Usingmorerepeatsisbe/erwhenonewantstoavoidconfoundingtheeffectsofindividualdifferenceswithtreatmenteffects,whenmorerepeatsarelessexpensivethanmoresubjects,andforraredisorders.Moredisordersarebecomingrareasdiagnos7cspecificityincreases.ThecurrentversionofSIMAsoawarecanprocessupto500repeatedmeasurements.

10

Thisshowshowyoucoulddrilldownfromthesta7s7callysignificantdemonstra7onresulttoiden7fytheop7malminimumdoseacrossresponsevariablesforeachperson.

SIMAenablesrandomized7tra7ontoop7maldoseforeachperson.

Seehowtheseop7malminimumdosesare40,80,and20forpersons1,2,and3respec7vely.

Seehowthegroup-averageresultcloudstheperson-specificresults.

11

Thecorrectanswersareinbold.

ThishelpsshowhowSIMAandsta7s7csaretwodis7nctandoaencomplementarymethodsthatdoapplytodifferenttypesofdataanddodifferentthings.

Precisiondrugdevelopmentandmedicineneedbothquan7ta7vemethods.

Deathisarealendpoint.Bloodpressureisnot.CONSORT-compliantRCTsoaenalsoconfoundrealendpointswithar7ficialendpoints.

12

SIMAhelpsenabletrulypa7ent-centeredcompara1vesafetyANDeffec1venessresearch.

Increasingnumbersofdrugsmeanthatmoretreatmentsneedtobecompared.

Followthiswithasingle-samplet-testforeachtypeoftreatmenttoseeifeithertreatmentisbeneficialorharmful.

SIMAcangreatlysimplifysta7s7calanalyses.

13

Ourna7onalmo/oisEpluribusunum–Outofmany,one.Thisslideisaboutestablishingthescienceofindividuality,Eunumpluribusandatwo-waystreetbetweenindividuals(SIMA)andpopula7ons(sta7s7cs).

You’vealreadyseenhowthe3-persondemonstra7onyieldedasta7s7callysignificantresult.Thatrepresentsgeneraliza7onatthetopofthepyramid.

Yousawthegroupaverageoverallbenefitandharmscore,thesecondleveldown.

YoualsosawhowSIMAquan7fiedoverallbenefitandharmasnonlinearfunc7onsofdoseforeachpersonfromtheresponsevariablespecificdose-responserela7onships.Thisrepresentswithin-personsummariza7on.Theseweredifferen7allyweightedandaveragedforthegroupofthreepersons–thesecondleveldown.

Yousawtheresponse-variable-specificbenefitandharmscoresforeachperson.Suchdetailedresultsillustratethescienceofindividuality.

Inaddi7on,SIMAcanquan7fybenefitandharmasnonlinearfunc7onsofresponsevariablelevel,delayandpersistenceofresponse,etc.SIMAcanusede-trendingtodis7nguishtreatmenteffectsfordiseaseprogressionandspontaneousrecovery.SIMAcanuseBooleanindependenteventsfordrug-druginterac7onsanddrugcocktails.SIMAcanuseBooleandependenteventsforsyndromessuchasmetabolicsyndromeanddepression.SIMAcanquan7fymechanismsofdiseaseandtreatmenteffect.SIMAcanhelpquan7fyevidenceforcausalitywithinpersonsorotherindividuals.

Thisapproachhaspoten7altoacceleratebothhighlypa7ent-centricprecisionmedicineandpopula7onmedicine.

Thisapproachcanhelpobviatetheclinicalresearchtoclinicalprac7cetransla7onproblemwithbothdrugsandservices.

14

Eroom’slawstatesthatpharmaceu7calindustryproduc7vityhalvedaboutevery9½yearsininfla7on-adjusted$despitealltheinterveningscien7ficandtechnicaladvances.

Coulditbethattheregulatorysciencegatewayisbo/leneckedbyclinicaltrialsthatusecategoricalindependentvariablesandgroupaveragesatendpoints?

Mightweneedtoknowindividualswellthroughthescienceofindividualitybeforewecanclassifythemwell?

Clinicianstreatindividuals.Mightthescienceofindividuality,enabledbyapplyingSIMAtomul7variate7meseriesdata,betheheretoforemissingfounda1onformuchofevidence-basedprecisiondrugdevelopmentandmedicine?

MightthisapproachhelpreverseEroom’slaw?

15

Yousaw,albeitwithmockdata,howitmightbepossibletoachievesta7s7callysignificantresultsinrandomizedsingle-groupRCTswithsmallnumbersofpersons.

Mightitbepossibletolargelyendclinicaldrugsafetyproblems?

16

TherearemanywaystohelpvalidateSIMAonthewaytoprecisiondrugdevelopment.

17

WithSIMA,drugdevelopmentbecomesmorelikeusingafunnellargeendup.SIMAhelpsprovideinforma7onneededtotargettherightdrugattherightdosetotherightperson.

18

Genomicsismilesaheadofphenomics.Wholegenomesequencingisbecomingfeasibleformany.However,weares7llusingoutmodeddiseaseclassifica7onsandconfoundedcategoriesofresponderandnon-responder.

Thisproblemcanbelikenedtotryingtonailphenomicjellytoawallwithgene7cnails.

19

Thisslideemphasizesmechanisms.Mechanismsandresponsetake7me.SIMAcomputes7me-dependentmechanism-specificphenotypesfrommul7variate7meseriesdata.

Thesethreecategories–diagnos7c,treatmentresponse,andagency–aredis7nguishedbyhow7meseriesinternalorexternaltotheindividualareselectedtooperateasindependentanddependentvariableswhenapplyingSIMA.

20

Measurementofbenefitsandharmsisneededwheneverthereisuncertaintyaboutbenefits,harms,anddoses.

21

DataSpeaksisdedicatedtoaccelera7ngprecisiondrugdevelopmentandmedicineforeveryone.

22