bayesian post hoc estimation Alfonso Iorio Abstract Symposium AS 122 Wednesday June 24 2015 1655 Room 701 Shareholder No ne Grant Research Support Funds managed via Institution ID: 785385
Download The PPT/PDF document "A population approach to hemophilia phar..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
A population approach to hemophilia pharmacokinetics. WAPPS: a web-service for bayesian post hoc estimation
Alfonso Iorio
Abstract Symposium AS 122
Wednesday
June
24
2015,
16:55
Room
701
Slide2Shareholder
No
neGrant / Research SupportFunds managed via Institution(Bayer, Baxter, BioGen, NovoNordisk, Octapharma)ConsultantFunds managed via Institution(Pfizer, Bayer, Biogen)EmployeeMcMaster UniversityPaid InstructorNoneSpeaker bureauNoneOtherPI of the WAPPS projectChair of the Data&Demographics Committee WFH,CFGD RG Cochrane Collaboration Editor
Disclosures for A. Iorio
Slide3BacKGround
Slide4Factor levels and bleeding
Clinical
severity of haemophilia A: Does the classification of the 1950s still stand?den Uijl IEM, Mauser Bunschoten EP, Roosendaal G, et al. Haemophilia 2011;17:849–53.
Slide5Factor levels and bleeding
Collins
PW, et al.J Thromb Haemost 2010;8:269–75.
Slide6PHARMACOKINETIC CHARACTERISTICS OF FACTOR VIII AND IX CONCENTRATES –
A
SYSTEMATIC REVIEW 75 articles 2050 patients included in PK analyses.38 on factor VIII concentratesHL(hr) forwild type 7.8 to 19.2,BDD 7.5 to 17.9prolonged HL 11.5 to 23.1 25 on factor IX concentrates.HL(hr) forwild type 12.9 to 36.0 ,prolonged HL 53.5 to 110.4Xi M, Navarro-Ruan T, Mammen S, Blanchette V, Hermans C, Morfini M, Collins P, Fischer K, Neufeld EJ, Young G, Kavakli K, Radossi P, Dunn A, Thabane L, Iorio A for the WAPPS study group - PO262-MON.
Slide7A story of two tails..
Individual case
Estimated terminal t1/2 (hr)Unpublished data, 1 single molecule
Slide8AIM
Slide9implement a population PK engine for all FVIII / FIX concentrates and make it available onlineprovide support to PK estimation in hemophilia based on flexible reduced set of data point (2-3 post infusion samples)
AIM of the project
Slide10Concentration in blood is a biomarker for concentration at site of actionAre we able to measure PK?PK parameters are not directly measured
While you can measure
Cthrough in blood directly, you can’t measure Clearance and VolumePK ESTIMATION
Slide11Population pharmacokinetic
Is it reliable, precise, accurate?
PEAK & TROUGHPRECISION, ACCURACY11 data points classic PK
Slide12Population pharmacokinetic
Slide13Population pharmacokinetic
Slide14Population pharmacokinetic
Slide15Population PK
Two main applications:
Drug oriented (derivation phase)Estimating the PK properties of a drug using sparse data from a population of subjectsPatient oriented (estimation phase)Estimating the PK in one individual using sparse data from the subject and a population model
Slide16Materials and Methods
Slide17Funding support
Slide18Trial registration
Slide19The WAPPS core team
PI:
Iorio, Alfonso and Hermans, CedricAdvisory Committee: Blanchette, Victor; Collins, Peter; Morfini, Massimo; Project coordinator: Navarro, TamaraInformation Technology: Cotoi, Chris; Hobson, Nicholas; McKibbon, Ann;Pharmacokinetics: Edginton, Andrea;Statistics: Foster, Gary; Thabane, Lehana; Consultant: Bauer, Rob (Consultant at ICON)Literature service, data entry: Xi, Mengchen; Mammen, Sunil; Yang, Basil;User testing: Bargash, Islam
Slide20Estimating PK for single individuals on the base of 2-4 samples
Web-application
Single patient reportSingle patient data
Slide21Estimating PK for single individuals on the base of 2-4 samples
Web-application
Single patient reportSingle patient dataOnline PPK engine(NONMEM)
Slide22Estimating PK for single individuals on the base of 2-4 samples
Web-application
Single patient reportSingle patient dataOnline PPK engine(NONMEM)Control files for bayesian individual estimationADVATEKOGENATEBENEFIXALPROLIXELOCTATEOthers..Brand specific Source individual PK dataOffline PPK modelingBrand specific PPK models
Slide23Web-application
Single patient report
Single patient dataOnline PPK engine(NONMEM)Control files for bayesian individual estimationADVATEKOGENATEBENEFIXALPROLIXELOCTATEOthers..Brand specific Source individual PK dataOffline PPK modelingBrand specific PPK modelspatientspatientspatients
Slide24The WAPPS network
Slide25The network
Active centers
US Guy YoungTU Kaan Kavakli US Ellis J. NeufeldCA Shannon JacksonIT Paolo RadossiCA Paula JamesCE Jan BlatnyCA Jerry TeitelVZ Arlette Ruiz-SàezNL Kathleen FischerUS Amy DunnCA Victor BlanchetteD Rainer Kobelt In processCA Alan TinmouthCA MacGregor SteeleUK Savita RangarajanSA Johnny MahlanguIT Alberto TosettoD Cristoph BidlingmaierIT Giancarlo CastamanSL Barbara Faganel KotnikUS Craig Kessler 47 more attended the introductory webinar
Slide26Disclaimer:
This is a research service under development, not yet validated for clinical practice use. Any use of the results of the population pharmacokinetic estimation in the care of individual patients is not recommended and cannot be considered part of the service in this phase. The local investigator is solely responsible for any such use.
Reporting
Slide27Published models
Drug
RefsCompAdvateBjorkmann, Eur J Clin Pharm, 2009;Blood, 2013; JTH 20102AlprolixDiao, Clin Pharmacokinet 20143EloctateNestorov I, Clin Pharm in Drug Dev 20142XynthaKarafoulidou, Eur J Clin Pharmacol 20092
Slide281-cmt
FO
FOCEFOCEILaplacianAdditiveCCVExponentialClVolTypeEstimation MethodIIVModelParametersResidualVariabilityAdditiveCCVAdditive+CCVLog ErrorAssessment1-cmtFOCEExponentialAdditive + CCVClVolModeling:Base Structural Model28 OBJF Diagnostic plots
Slide29Systemic clearance is assumed to be a random effectClearance and volume are assumed to be independent, which reduces the number of parameters in the covariance matrix
Parametrization
Slide30Post-modelling bayesian individual estimation
Sparse data from a subject are modeled iteratively with priors from the population distribution
Credibility intervals around the estimates are generatedIndividual estimation
Slide31RESULTS - MODELLING
Slide32Source data
Subjects
ReplicatesnKineticsBrand A4021161B80584312C3030290D167301197E25--25F129201149Total471159
834
Slide33Age distribution
Slide34Total
477
Min - max4.3 – 71.5Media +/- SD32.2 +/- 13.8Median (25 – 75 PCT)29 (21.4 – 42.6)10th percentile16Derivation cohort:Age distribution
Slide35patient
1
2345dose/kg55595653.150.9dose34054086544847674920one-compV2985.9726294217.3053887.613676.66k0.0010520.0013090.0010890.0012380.000796HL (min)659529
636
559
871
HL (h)
10.98
8.82
10.60
9.32
14.52
C(0)
approx
D/V
1.14
1.55
1.29
1.23
1.34
two-comp
A
0.486937
1.274402
0.476461
0.934315
0.987822
B
0.726386
0.302367
0.8767
0.303577
0.379788
alpha
0.004996
0.001823
0.004597
0.001656
0.001333
beta
0.000532
0.000315
0.000645
0.000528
0.000207
alphaHL
138
380
150
418
519
betaHL
1302
2197
1074
1312
3342
Hlbeta (h)
21.70
36.62
17.90
21.87
55.70
Classical PK estimation
Slide36WAPPS models
Drug
TypeClassGen ctrlPref Mod(Comp)Alt model(Comp)ActiveAdvateF8R-wildN21YAlprolixF9R-longN32YBenefIXF9R-wildN2--YEloctateF8R-longN22YHumate PF8PDY21YKogenateF8R-wildN21YWilateF8PDY21YXynthaF8R-BDDN2--Y
Slide37PPK MODELS
Drug
ClV1V2K10K12K21Advate0.252.870.810.08620.27700.9778Kogenate0.182.900.840.06130.03440.1193Xyntha0.334.520.890.07370.38991.9869Eloctate0.163.400.520.04700.55123.6285Benefix0.518.816.560.0580.13060.1753Alprolix0.227.214.350.03020.34870.5784V3K13K31Alprolix13.790.02570.0134
Slide38Drug
Terminal HL
MRT (Plasma)MRT (body)Advate, hrs(95% CrI)10.5(5 – 16)8(3 – 12)10(5 – 15)Kogenate, hrs(95% CrI)16(6 – 27)11(4 – 18)14.5(5 – 24)Xyntha, hrs(95% CrI)11(8 – 14)9.5(6 – 13)11(8 – 14.5)Eloctate, hrs(95% CrI)17(6 – 26)15(3 – 26)17(6 – 28)FACTOR VIII PPK half-life
Slide39Drug
Terminal HL
MRT (Plasma)MRT (body)Benefix, hrs(95% CrI)23(12 – 34)12(7 – 17)21(12 – 29)Alprolix, hrs(95%, CrI)116(65 – 167)23(12 -34)81(48 – 113)FACTOR IX PPK
Slide40“Time to” critical concentrations
Concentrate
0.0595 % CrI0.0295 % CrI0.0195 % CrIAdvate4421-675828-886833 - 104Kogenate6224 – 1018432 – 13610038 – 163Xyntha4124 – 595635 – 776744 – 91Eloctate7020 – 1209328 – 15811034 – 186Benefix4430 – 577447 – 1009659 – 133Alprolix7542 – 10919193 – 289307179 – 434Benefix 1006643 – 1089659 – 13311971 - 167Levels as IU/mL – time as hours
Slide41Kogenate
50 IU/kg
Slide42BenefIX
50 IU/kg
Slide43RESULTS – INDIVIDUAL ESTIMATIONRICH VS SPARSE DATA SAMPLE
Slide44Bayesian estimate, rich data
Estimate
X95% CrITerminal HL (hr)12(10.5 – 13)Time (hr) to UI/mL0.0546.5(39.75 - 53.25)0.0262.75(53.5 – 72)0.0175(63.5 – 86.25)
Slide45Slide46Slide47Slide48Reduced sampling sets
Samples
HL0.050.020.010:15 481246.5(38.75 - 53.25)62.75(53.5 – 72)75(63.5 – 86.25)0:15, 3, 2813.552.5(36.75 – 67.75)70(49.25 – 90.5)83.25(58.75 – 107.75)0:30, 4811.039(31.15-46.75)53.5(44 – 63)64.5(53.25 – 73.5)0:15, 311.544.25(22.5 – 66)59.75(32 – 87.25)71.5(39.25 – 103.75)
Slide49RESULTS – SENSITIVITY
Slide50Patient with fast clearance
Time
[C]1:300.6912:000.1324:00<0.01HL (t) = 5.5 hours(CrI 5.0 – 7.5)
Slide51Patient after ITI
Time
[C]35:080.03HL (t) = 7.5 hours(CrI 6.5 – 9.5)
Slide52RESULTS – PERFORMANCE, TESTING AND VALIDATION
Slide53Internal Validation: bootstrap
Slide54221 patients257 kinetics
Validation cohort
Emoclot16FANHDI7Aafact17Kogenate62Helixate23Recombinate20Advate41Xyntha30Others41
Slide55CONCLUSION AND FUTURE DEVELOPMENTS
Slide56PPK assisted estimate
It feasible - Is it reliable, precise, accurate?
PEAK & TROUGHPRECISION, ACCURACY11 data points classic PK
Slide57Step I: prediction on sparse data for new patientsReference: full set for the same patient
Index measure:
difference, ratio, predictive valueStep II: prediction on sparse data for new patientsReference: Prospective verification in the same patientIndex measure: agreement difference, ratio, predictive valueStep III: prediction on sparse data for new patientsComparison: “guess” from the treater on the same dataIndex measure: agreement difference, ratio, predictive valueExternal validation
Slide58Release of the multilingual interfaceOptimal sampling time analysis
Development of a regimen simulation module
Future steps
Slide59The WAPPS core teamThe WAPPS network membersBayer, Biogen
,
Kedrion, Octapharma, Pfizer. AKnowledgments
Slide60Thank you !!!
Join the WAPPS network at:
www.wapps-hemo.orgDownload these slides at:Hemophilia.mcmaster.ca