Purpose
The objective of the present analysis was to explore the use of stochastic differential equations (SDEs) in population pharmacokinetic/pharmacodynamic (PK/PD) modeling.
Methods
The intra-individual variability in nonlinear mixed-effects models based on SDEs is decomposed into two types of noise: a measurement and a system noise term. The measurement noise represents uncorrelated error due to, for example, assay error while the system noise accounts for structural misspecifications, approximations of the dynamical model, and true random physiological fluctuations. Since the system noise accounts for model misspecifications, the SDEs provide a diagnostic tool for model appropriateness. The focus of the article is on the implementation of the Extended Kalman Filter (EKF) in NONMEM® for parameter estimation in SDE models.
Results
Various applications of SDEs in population PK/PD modeling are illustrated through a systematic model development example using clinical PK data of the gonadotropin releasing hormone (GnRH) antagonist degarelix. The dynamic noise estimates were used to track variations in model parameters and systematically build an absorption model for subcutaneously administered degarelix.
Conclusions
The EKF-based algorithm was successfully implemented in NONMEM for parameter estimation in population PK/PD models described by systems of SDEs. The example indicated that it was possible to pinpoint structural model deficiencies, and that valuable information may be obtained by tracking unexplained variations in parameters.
Similar content being viewed by others
Abbreviations
- EKF:
-
Extended Kalman Filter
- FOCE:
-
first-order conditional estimation
- GnRH:
-
gonadotropin releasing hormone
- LLOQ:
-
lower limit of quantification
- ODE:
-
ordinary differential equation
- PK/PD:
-
pharmacokinetics/pharmacodynamics
- RSE:
-
relative standard error
- SC:
-
subcutaneous
- SDE:
-
stochastic differential equation. Variables:
- e :
-
measurement error
- ε:
-
residual error
- η :
-
inter-individual variability
- K :
-
Kalman gain
- Ω:
-
inter-individual covariance
- P :
-
state covariance
- φ :
-
individual parameter
- R :
-
output prediction covariance
- Σ:
-
measurement error covariance
- σ w :
-
diffusion term
- t :
-
time
- θ :
-
population mean parameter
- w :
-
Wiener process
- x :
-
state variable
- y :
-
measurement
References
R. Jelliffe A. Schumitzky M. Guilder ParticleVan (2000) ArticleTitlePopulation pharmacokinetics/pharmacodynamics modeling: parametric and nonparametric methods Ther. Drug Monit. 22 IssueID17–18 354–365 Occurrence Handle10.1097/00007691-200006000-00019 Occurrence Handle10850405
J. K. Lindsey B. Jones P. Jarvis (2001) ArticleTitleSome statistical issues in modelling pharmacokinetic data Stat. Med. 20 2775–2783 Occurrence Handle10.1002/sim.742 Occurrence Handle11523082
M. O. Karlsson S. L. Beal L. B. Sheiner (1995) ArticleTitleThree new residual error models for population PK/PD analyses J. Pharmacokinet. Biopharm. 23 651–672 Occurrence Handle8733951
C. W. Tornøe J. L. Jacobsen H. Madsen (2004) ArticleTitleGrey-box pharmacokinetic/pharmacodynamic modelling of a euglycaemic clamp study J. Math. Biol. 48 591–604 Occurrence Handle10.1007/s00285-003-0257-z Occurrence Handle15164224
C. W. Tornøe J. L. Jacobsen O. Pedersen T. Hansen H. Madsen (2004) ArticleTitleGrey-box modelling of pharmacokinetic/pharmacodynamic systems J. Pharmacokinet. Pharmacodyn. 31 401–417 Occurrence Handle10.1007/s10928-004-8323-8 Occurrence Handle15669774
N. R. Kristensen H. Madsen S. B. Jørgensen (2004) ArticleTitleParameter estimation in stochastic grey-box models Automatica 40 225–237 Occurrence Handle10.1016/j.automatica.2003.10.001 Occurrence HandleMR2145299
R. V. Overgaard, E. N. Jonsson, C. W. Tornøe, and H. Madsen. Non-linear mixed-effects models with stochastic differential equations. Implementation of an estimation algorithm. J. Pharmacokinet. Pharmacodyn. In press (2005).
N. R. Kristensen H. Madsen S. B. Jørgensen (2004) ArticleTitleA method for systematic improvement of stochastic grey-box models Comput. Chem. Eng. 28 1431–1449 Occurrence Handle10.1016/j.compchemeng.2004.08.004
S. L. Beal and L. B. Sheiner (Eds.). NONMEM® Users Guides. NONMEM Project Group, University of California, San Francisco, 1994.
A. H. Jazwinski (1970) Stochastic Processes and Filtering Theory Academic Press New York
R. E. Kalman (1960) ArticleTitleA new approach to linear filtering and prediction problems Trans. ASME, Ser. D J. Basic Eng. 82 35–45
R. E. Kalman R. S. Bucy (1961) ArticleTitleNew results in linear filtering and prediction theory Trans. ASME, Ser. D J. Basic Eng. 83 95–108
P. S. Maybeck (1979) Stochastic Models, Estimation, and Control Academic Press New York
V. P. Shah K. K. Midha S. Dighe I. J. McGilveray J. P. Skelly A. Yacobi T. Layloff C. T. Viswanathan C. E. Cook R. D. McDowall (1991) ArticleTitleAnalytical methods validation: bioavailability, bioequivalence and pharmacokinetic studies. Conference report Eur. J. Drug Metab. Pharmacokinet. 16 IssueID4 249–255 Occurrence Handle1823867
C. W. Tornøe H. Agersø H. A. Nielsen H. Madsen E. N. Jonsson (2004) ArticleTitlePopulation pharmacokinetic modeling of a subcutaneous depot for GnRH antagonist degarelix Pharm. Res. 21 574–584 Occurrence Handle10.1023/B:PHAM.0000022403.60314.51 Occurrence Handle15139513
H. Agersø W. Koechling M. Knutsson R. Hjortkjær M. O. Karlsson (2003) ArticleTitleThe dosing solution influence on the pharmacokinetics of degarelix, a new GnRH antagonist, after s.c. administration to beagle dogs Eur. J. Pharm. Sci. 20 335–340 Occurrence Handle10.1016/j.ejps.2003.08.001 Occurrence Handle14592699
D. S. Bayard R. W. Jelliffe (2004) ArticleTitleA Bayesian approach to tracking patients having changing pharmacokinetic parameters J. Pharmacokinet. Pharmacodyn. 31 75–107 Occurrence Handle10.1023/B:JOPA.0000029490.76908.0c Occurrence Handle15346853
Acknowledgments
The authors wish to acknowledge the help of Professor Stuart Beal (UCSF), who instructed us on how to implement the EKF algorithm in NONMEM. This work was financially supported by Ferring Pharmaceuticals A/S and Center for Information Technology, Denmark.
Author information
Authors and Affiliations
Corresponding author
Appendix 1
Appendix 1
NONMEM SDE Control Stream (See Table I)
1 | $PROBLEM SDE 1-COMP MODEL WITH 2 ABS COMPONENT | 65 | A7 = A7 |
$INPUT ID HOUR TIME AMT CMT DV EVID MDV | A8 = A8 | ||
$DATA SDE.dta IGNORE = @ | A9 = A9 | ||
$SUBROUTINE ADVANCE TOL 6 DP | ENDIF | ||
5 | $MODEL COMP = (ABS1) COMP = (CENT) | ; EKF state update equations | |
COMP = (P11) COMP = (P12) COMP = (P13) | 70 | IF(EVID.EQ.3) THEN | |
COMP = (P22) COMP = (P23) COMP = (P33) | ;Output cov. R_j∣j−1 = C P_j∣j−1C^T + Sig Sig^T | ||
RVAR = A9/(A3*A3) + SIG**2 | |||
$THETA (0,5.50)(0,100)(0,30.0)(0,900)(0,0.10,1) | ;Kalman Gain K_j = P_j∣j−1 C^T R_j∣j−1^−1 | ||
10 | (0,0.70,1)(0.0.20)(0,1)(0,250)(0,1) | K1 = A6/(A3*RVAR) | |
$OMEGA 0.25 0.25 0.25 0.25 | 75 | K2 = A8/(A3*RVAR) | |
$Sigma 1 FIX | K3 = A9/(A3*RVAR) | ||
;State update eq. x_j∣j = x_j∣−1+K_j (y_j−y_j∣j−1) | |||
$PK | A1UP = A1+K1*(OBS-LOG(A3/V1)) | ||
15 | CL = THETA(1) | A2UP = A2+K2*(OBS-LOG(A3/V1)) | |
V1 = THETA(2) | 80 | A3UP = A3+K3*(OBS-LOG(A3/V1)) | |
HL1 = THETA(3)*EXP(ETA(1)) | ;State cov. update eq.P_j∣j = P_j∣j−1K_j R_j∣j−1 K_j^T | ||
HL2 = THETA(4)*EXP(ETA(2)) + HL1 | P1UP = A4-K1*RVAR*K1 | ||
KA1 = LOG(2)/HL1 | P2UP = A5-K1*RVAR*K2 | ||
20 | KA2 = LOG(2)/HL2 | P3UP = A6-K1*RVAR*K3 | |
TFR = THETA(5) | 85 | P4UP = A7-K2*RVAR*K2 | |
RHO = LOG(TFR/(1-TFR)) | P5UP = A8-K2*RVAR*K3 | ||
FRAC = EXP(RHO+ETA(3))/(1+EXP(RHO+ETA(3))) | P6UP = A9-K3*RVAR*K3 | ||
TBIO = 1 | ENDIF | ||
25 | IF(CONC. EQ. 20)TBIO = THETA(6) | ;Update states | |
BIO = TBIO8EXP(ETA(4)) | 90 | IF (A_0FLG.EQ. 1) THEN | |
SIG = THETA (7) | A_0(1)A1UP | ||
SGW1 = THETA (8) | A_0 (2) = A2UP | ||
SGW2 = THETA (9) | A_0 (3) = A3UP | ||
30 | SGW3 = THETA (10) | A_0 (4) = P1UP | |
F1 = FRAC*BIO | 95 | A_0 (5) = P2UP | |
F2 = (1-FRAC)*BIO | A_0 (6) = P3UP | ||
;Initialize state and state covariance equations | A_0 (7) = P4UP | ||
IF (NEWIND.NE.2) THEN | A_0 (8) = P5UP | ||
35 | AHT1 = 0 | A_0 (9) = P6UP | |
AHT2 = 0 | 100 | ENDIF | |
AHT3 = 0 | |||
PHT1 = 0 | $DES | ||
PHT2 = 0 | ;State predicton eq. dx_t∣j/dt = g(x_t∣j,d,phi) | ||
40 | PHT3 = 0 | DADT(1) = −KA1*A(1) | |
PHT4 = 0 | 105 | DADT(2) = −KA2*A(2) | |
PHT5 = 0 | DADT(3) = −KA1*A(1) + KA2*A(2) - CL/V1*A(3) | ||
PHT6 = 0 | ;State cov. dP_t∣j/dt = A_t P_t∣j+P_t∣j A_t^T+SGW SGW^T SGW^T | ||
ENDIF | DADT (4) = −2*KA1*A(4) + SGW1*SGW1 | ||
45 | ;Store observations for EKF update | DADT (5) = −(KA1+KA2)*A(5) | |
IF (EVID.EQ.0) OBS = DV | 110 | DADT (6) = −(KA1+CL/V1)*A(6) + KA1*A(4) + KA2*A(5) | |
;Store one-step predictions for EKF update | DADT (7) = −2*KA2*A(7) + SGW2*SGW2 | ||
IF (EVID.NE.3) THEN | DADT (8) = −(KA2+CL/V1)*A(8) + KA1*(5) + KA2*A(7) | ||
50 | A1 = A(1) | DADT (9) = 2*KA1*A(6)+2*KA2*A(8)−2*CL/V1*A(9)+SGW3*SGW3 | |
A2 = A(2) | |||
A3 = A(3) | 115 | $ERROR (OBS ONLY) | |
A4 = A(4) | IF (ICALL.EQ.4) THEN | ||
A5 = A(5) | IF (DV.NE.0) Y = LOG(DV) | ||
A6 = A(6) | RETURN | ||
55 | A7 = A(7) | ENDIF | |
A8 = A(8) | 120 | IPRED = LOG(A(3)/V1) | |
A9 = A(9) | W = SQRT(A(9)/(A(3)*A(3))+SIG**2) | ||
ELSE | IRES = DV – IPRED | ||
A1 = A1 | IWRES = IRES/W | ||
60 | A2 = A2 | Y = IPRED + W*EPS(1) | |
A3 = A3 | 125 | $SIM (1) | |
A4 = A4 | $EST METH=1 INTER | ||
A5 = A5 | $COV | ||
A6 = A6 |
Rights and permissions
About this article
Cite this article
Tornøe, C.W., Overgaard, R.V., Agersø, H. et al. Stochastic Differential Equations in NONMEM®: Implementation, Application, and Comparison with Ordinary Differential Equations. Pharm Res 22, 1247–1258 (2005). https://doi.org/10.1007/s11095-005-5269-5
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11095-005-5269-5