610 likes | 1.22k Views
Erythropoietin Modeling and Simulation. Immanuel Freedman, Ph.D., SMIEEE. O 2 sensor. A. HIF-1. . enhancer. P. EPO. apoptosis. neocytolysis. EPO. C. D. B. Self renewal. Default suicide pathway. selective hemolysis of early erythroblasts. Bone-marrow sinusoidal endothelium.
E N D
Erythropoietin Modeling and Simulation Immanuel Freedman, Ph.D., SMIEEE
O2 sensor A HIF-1 enhancer P EPO apoptosis neocytolysis EPO C D B Self renewal Default suicide pathway selective hemolysis of early erythroblasts Bone-marrow sinusoidal endothelium Hematopoietic stem cells erythroid precursor cells BFU-e/CFU-e reticulocytes erythrocytes
SC bolus Serum concentration Ka• F Ke Cc Hemoglobin RBC Production RBC Lifespan
RBC lifespan eEPO PK Response cytokines Bone Marrow Status Cancer Type Chemotherapy CL Dose efficiency Ceff Ferrokinetics? Prior Radio/Chemo Therapy
Clinical Trial Simulation Dose withholding If Hgb<1 g/dL @ Week 6 Dose = 5/3*Dose If [Hgb]<14 | 15 g/dL If [Hgb]14 | 15 g/dL Dose = 0.75xDose Dose increase TEST ARM Dose = 200 g Q2W If [Hgb]<8.0 Transfusion Treatment Treat & Monitor Censor for 4 weeks Patient Dropout CONTROL ARM Dose = 3.0 g/kg Q2W Baseline Characteristics Random pt censoring ~3.1%/week cf study [Hgbo]=9.8 ± 0.6 (8-11) g/dL 74.7 ± 18.5 (27-156) kg Male 32%, Female 68% n=254/cohort x 1000
Simulator Customer Groups • Corporate • Clinical • Marketing • Research
Simulator Goals The simulator must be: • Accurate • Responsive • Portable • Easy to Use
Inclusion Criteria (1 of 2) Subjects must be: • at least 18 years of age, • receiving cyclic chemotherapy, • diagnosed with non-myeloid malignancies, • diagnosed with Anemia of Cancer or Chemotherapy Induced Anemia, • anemic (Hb ≥9.0 g/dL and Hb ≤ 11.0 g/dL), except in Amgen Study 990146 (Hb ≤ 13.0 g/dL), • capable of self-care (ECOG 0 to 2), and • diagnosed with adequate renal and hepatic function.
Inclusion Criteria (2 of 2) Subjects must have: • no history of seizures, cardiac or hematologic disorders that could cause anemia, • no rHuEPO treatment before study begins, • less than 2 RBC transfusions within 4 weeks before study drug, and • no RBC transfusions during current chemotherapy cycle before randomization.
Population PK/PD Model Features • eight compartments, • an indirect Emax link model, • non-Gaussian residuals, • censored transfusion data, • allometric parameter scaling, • step-down covariate analysis, • validation on data not used in estimation, and • estimation with NONMEM V and MATLAB software.
Population PK/PD Model Discussion • fitted Emax scales with body weight according to (BWT/70.9)-0.9±0.3
Population PK/PD Simulation Features • a dose adjustment model, • a transfusion censoring model, • a patient dropout model, and • a multilognormal cohort.
NESP PK/PD Model (1 of 7) $PROB TEMPLATE FOR POP PKPD MODEL FOR DARBEPOETIN ALFA ; Run 1011 based on Run 501 and Run 701 for simulation ; Run 501: PD fit Hb data from 290, 162, and 291 studies. No transfusion points. ; Run 701: PK fit Aranesp 162 SC, 146 SC and IV. $INPUT C ID TIME AMT DV HB0 CMT CHEM TYPE ROUT STUD BWT $DATA 20010102PD.csv IGNORE=C $SUBROUTINE ADVAN6 TRANS1 TOL=3 $MODEL COM=(SC) COM=(CONC) COMP=(PERI) COMP=(PR1) COMP=(LS1) COMP=(LS2) COMP=(LS3) COMP=(LS4)
NESP PK/PD Model (2 of 7) $PK CL=THETA(1)*EXP(ETA(1)); Clearance from central compartment (mL/day) V2=THETA(2)*EXP(ETA(2)); Volume of distribution (mL) V3=THETA(3)*EXP(ETA(3)); Q=THETA(4)*EXP(ETA(4)); KA=THETA(5)*EXP(ETA(5)); Absorption rate constant (/day) LT1=THETA(6)*EXP(ETA(6)); F1=LT1/(1+LT1); Bioavailability of SC dose K=CL/V2; Elimination rate constant (/day) K23=Q/V2 K32=Q/V3; S2=V2
NESP PK/PD Model (3 of 7) ;PD MODEL PARAMETERS RBCPT=THETA(7)*EXP(ETA(7)); Maturation time (day) RBCLS=THETA(8)*EXP(ETA(8)); Transit time (day) EMAX=THETA(9)*EXP(ETA(9)); Maximum stimulation effect EC50=THETA(10)*EXP(ETA(10)); Concentration at half maximal effect (ng/mL) KPT=1/RBCPT; Production rate constant (/day) KLS=4/RBCLS; Loss rate constant (/day) KCP=Q/V2; KPC=Q/V3; ; ALLOMETRIC SCALING CL=CL*(BWT/70.9)**0.75; V2=V2*(BWT/70.9); V3=V3*(BWT/70.9); Q=Q*(BWT/70.9); EMAX=EMAX*(BWT/70.9)**THETA(11);
NESP PK/PD Model (4 of 7) $DES ; PK C2=A(2)/V2; E=EMAX*C2/(EC50+C2); ; DADT(1)= -KA*A(1); SC injection site compartment DADT(2)=KA*A(1)-(K+KCP)*A(2); Central compartment DADT(3)=KCP*A(2)-KPC*A(3); Peripheral compartment ; ; PD DADT(4)=KPT*(1+E)-KPT*A(4); Progenitor stimulation DADT(5)=KLS*(A(4)-A(5)); Erythrocyte maturation DADT(6)=KLS*(A(5)-A(6)); DADT(7)=KLS*(A(6)-A(7)); DADT(8)=KLS*(A(7)-A(8)); ;
NESP PK/PD Model (5 of 7) $ERROR EFF=(A(5)+A(6)+A(7)+A(8))*HB0/4.0; W=EFF IPRED=EFF IRES=DV-IPRED IF(W.GT.0) THEN IWRES=IRES/W ELSE IWRES=0 ENDIF Y=EFF+ERR(1)
NESP PK/PD Model (6 of 7) $THETA (2010 FIX); CL (3390 FIX); V2 (251 FIX); V3 (2900 FIX); Q (0.318 FIX); KA (0.795 FIX); LT1 (F1=0.443) (4.68 FIX) ; RBCPT (0, 120); RBCLS (0, 10); EMAX (0, 10); EC50 (0, 5); THETA(11) ; $OMEGA 0.296 FIX 3.22 FIX 1.29 FIX 0.483 FIX 0.004 0.216 FIX 20 0.004 0.004 0.004 ; $SIGMA 10.0
NESP Covariate PK/PD Model (7 of 7) function y=resampleResiduals(residual, numberOfSamples, numberOfPatients) sizeVar=size(residual); maxIndex=sizeVar(1)-1; resplus=residual(2:end); resminus=residual(1:end-1); correlationMatrix=corrcoef(resplus, resminus); correlation=correlationMatrix(1,2); %off diagonal innovation = resplus-correlation*resminus; for subject=1:numberOfPatients y(1, subject)=0; %initial for sample=2:numberOfSamples index=1+round(abs(maxIndex*rand)); while(index==0 | index > maxIndex) index=1+round(abs(maxIndex*rand)); end %if y(sample, subject)=correlation*y(sample-1, subject)+ innovation(index); end %for sample end %for subject return
rHuEPO Baseline PK/PD Model Parameters *EC50 scaled from NESP EC50 using peptide mass