Non-influential Eta

From: Mahesh Samtani Date: April 16, 2008 technical Source: mail-archive.com
Dear NMusers, I am trying to sequentially model the PD of a drug that has a simple inhibitory Emax effect. There is a clear placebo response from the control group and placebo effect goes away once patients stop taking their placebo pills (Hence the TY code below). Both drug & placebo effect (decrease from baseline) are bigger in patients with higher baseline. I want the etas on placebo and drug parameters (C1 and SMAX) to be correlated with baseline (BS) eta, which is done using the Block Band matrix. I think this problem may have the non-influential eta bug (eta on Smax for placebo subjects???). When I run the code below, the model runs quite well. How do I know if I have been bit by the bug and how can I fix it. Kindly advice...Mahesh $SUBR ADVAN4 TRANS4 $PK CALLFL = -2 ; PK PARAMETERS DRUG SPECIFIC VALUES REMOVED KA = # CL = # Q = # V3 = # V2 = # F1 = # F2 = 1-# D2 = # ALAG1 = # L1 = THETA(1)*EXP(ETA(1)) SC50 = THETA(2)*EXP(ETA(2)) C1 = THETA(3)*EXP(ETA(3)) DURA = THETA(4) BS = THETA(5)*EXP(ETA(4)) SMAX = THETA(6)*EXP(ETA(5)) $ERROR ; BS IS THE BASELINE ; TY IS THE PLACEBO RESPONSE ; HFN IS HILL FUNCTION FOR DRUG EFFECT IF (TIME.LE.DURA) THEN TY=C1/L1*(1-EXP(-L1*TIME)) ELSE TY=C1/L1*(1-EXP(-L1*DURA))*EXP(-L1*(TIME-DURA)) ENDIF HFN=(SMAX*F)/(F+SC50) EFF=BS-TY-HFN IPRE = -5 IF (EFF.GT.0) IPRE=LOG(EFF) Y=IPRE+ERR(1) $OMEGA ; DRUG SPECIFIC VALUES REMOVED # # $OMEGA BLOCK(3) ; DRUG SPECIFIC VALUES REMOVED # # # 0 # #
Apr 16, 2008 Mahesh Samtani Non-influential Eta
Apr 16, 2008 Nick Holford Re: Non-influential Eta