Paolo is correct in his understanding. A bug was introduced into PREDPP when the Initial Steady State feature was introduced, in NONMEM VI 2.0. SS=2 doses do not work correctly. This same bug is present in NONMEM 7. More details will be sent when available.
On Mon, 05 Oct 2009 09:37:05 +0200, "Paolo Denti" <[email protected]> said: > Dear all, > I am creating a dataset with the SS=2 option and I noticed an unexpected > prediction profile that indicated that I am misinterpreting the way SS > works in NONMEM... I would appreciate any feedback. > > My understanding of the SS option in NONMEM is that the dose is assumed > as previously given at regular intervals (specified by II), and the > level of drug concentration resulting from this process is calculated. > There are different options (1, 2 and 3), which I list below to clarify > the way I understand them to work, so that it might be clear where I am > mistaken. > > According to the NONMEM guide, SS=1 should zero-out all system > compartments and reset them to the value resulting only from the > steady-state dose, thus virtually forgetting the past dosing history > completely, with the exception of the steady-state dose and its previous > infinite homologous doses. SS=3 - besides all the Greek in the > explanation - seems to be doing the same, but using different values for > the initial estimates (?!?). In any case, I tried and it zeroes-out the > compartments as SS=1 does. > > SS=2, instead, is supposed to calculate the steady-state level of drug > resulting from the implied series of doses, and add that amount ON TOP > of whatever other amount would be in the compartments resulting from > extra sources, i.e., I would add, other amounts present in the > compartments due to events PRIOR to the SS=2 record. > > I want to build an integrated model to analyze multiple drugs at the > same time, so I would avoid the SS=1 option, as the doses for drug x > would cancel out the ones for drug y... So I used SS=2, but it behaves > in a way I did not expect. > > I have dosing history for three days before the experiment, but not for > all patients. The drug is given once daily, normally in the morning, but > the exact time generally changes from day to day, so I decided to model > the first dose I have record of as steady-state, and then follow the > dosing history. Some subjects, however, always declared to have taken > the dose at 8 am. I pasted a fragment of the dataset below. > > #ID TIME DV AMT SS II MDV > 1 0 . 450 2 24 1 > 1 24 . 450 0 . 1 > 1 48 . 450 0 . 1 > 1 72 . 450 0 . 1 > 1 72.2 0.05 . . . 0 > 1 73.3 3.83 . . . 0 > 1 74.4 3.27 . . . 0 > 1 75 2.66 . . . 0 > 1 96 . . . . 1 > > I would expect to observe the very same predictions at 24 hours > distance, as 24 hours is the inter-dose interval. > However, after a minimization successful and that's what I get as an > output: > ID TIME Y DV PRED RES WRES IPRE IRES IWRE > 1 0 */0.11356 /* 0 0.16226 0 0 */0.11356/* > -0.11356 -0.090191 > 1 24 */0.11356/* 0 0.16226 0 0 */0.11356/* > -0.11356 -0.090191 > 1 48 */0.056889/* 0 0.081288 0 0 */0.056889/* > -0.056889 -0.045196 > 1 72 */0.056782/* 0 0.081132 0 0 */0.056782 /* > -0.056782 -0.045111 > 1 72.2 0.59151 0.05 0.81647 -0.76647 -0.4746 > 0.59151 -0.54151 -0.42489 > 1 73.3 2.557 3.83 3.5268 0.30316 0.49445 2.557 > 1.273 0.83272 > 1 74.4 3.3771 3.27 4.6667 -1.3967 -0.27596 > 3.3771 -0.10714 -0.062947 > 1 75 3.5245 2.66 4.8757 -2.2157 -0.6256 3.5245 > -0.86455 -0.49795 > 1 96 0.056782 0 0.081132 0 0 0.056782 > -0.056782 -0.045111 > > The first two records (0 and 24) have a value which is about twice as > much as the following (48, 72 and 96), which seem to be going towards a > lower steady-state level. > > With SS=1 it does work as expected, but when I will add the compartments > for the other drugs, SS=1 will interfere with those. > > I am sure I am misinterpreting something here, but in my understanding, > it should not matter whether I choose SS=1 or SS=2, as long as the dose > is the first one in the dataset... SS=1 forgets the past dosing history, > while SS=2 remembers it, but what difference does it make if there was > no other dose in the past? > > Thank you in advance to anyone who can help - or even just had the > patience to read all this.. ;) > Paolo > > -- > ------------------------------------------------ > Paolo Denti, PhD > Post-Doctoral Fellow > Division of Clinical Pharmacology > Department of Medicine > University of Cape Town > > K45 Old Main Building > Groote Schuur Hospital > Observatory, Cape Town > 7925 South Africa > phone: +27 21 404 7719 > fax: +27 21 448 1989 > email: [email protected] > ------------------------------------------------ > -- Alison Boeckmann [email protected]
