Re: question about sampling time

On 28/10/2013 8:28 a.m., Winnie Seto wrote: > Hello NONMEM users and Dr. Holford > > I am a novice when it comes to pop PK simulations- You refer to pop PK simulations but what you describe below sounds more like a plan for a pop PK parameter estimation study rather than a pop PK simulation study. Of course, some simulation would be helpful for planning a parameter estimation study (see below). It would be helpful for know what your objectives are. There must be hundreds of published aminoglycoside studies so why are you planning another? What are you hoping to learn? Do you have a hypothesis you want to confirm? If you don't have a clear idea about what you going to learn or confirm, and a scientific plan to show you can achieve the objectives, then I would consider this kind of study unethical. > Our group is planning on a prospective once daily aminoglycoside pharmacokinetic study in patients with proposed serial time levels post dose with 30min, 2hr, 4hr, 12hr and 24hr post dose drug levels. In typical patients given once daily doses of aminoglycosides such as gentamicin it is most likely that a 24 h concentration would be less than the limit of quantification of you laboratory. IMHO a 24 h sample would be unethical in a typical patient over the age of 1 year because it has some discomfort and probably $$ cost for the patient and is of almost no value. This is where some simulation could be helpful. In patients with low clearance (e.g. due to immaturity in neonates) then a 24 h sample might be reasonable. You can see a simple simulation example here in slide 17: http://holford.fmhs.auckland.ac.nz/docs/target-concentration-intervention.pdf This kind of simulation is easily done in Excel. Just doing it yourself will help you understand the answers to many questions about PK study design. > We will have only a small sample size (less than 50 patients) and some patients may not be able to have samples at all these timepoints from logistical reasons. The sample size depends on what your objectives are. 50 patients would be fine to describe the popPK of an aminoglycoside given 5 conc measurements per subject. If your objective is related to a question about covariates predicting variability (weight, renal function, maturation, etc) then you may need a larger sample to get a clear answer. > I would really appreciate some advice on the following: > > 1.Is there a number of minimum patients that I need for each sampling time point for pop pk analysis? This kind of question about the number of patients, number of samples and timing of samples is best answered by using an optimal design program. See http://www.page-meeting.org/pdf_assets/9481-mentre_page07postPage2.pdf for a review. Most of these programs will have improved since then but some of them are no longer available). The trade off between these design quantities will be determined by what your objectives are and how much money and time you have. > 2.For each time point – is it better to have some variability within the sampling time window e.g. 10-12hr with a few patients with samples distributed at 10, 11 or 12 hour mark? Yes -- as a general rule your design will be more robust if you use sampling windows and in real life samples will not be taken at exactly the same time in every patient. For data analysis it is essential to record the dosing and sampling times as accurately as possible. > 3.I have learnt only the basics with NONMEM; I have used WIN NONLIN in the past - recently I have added their WIN-NLME pop PK software to my subscriptions – should I use both NONMEM and NLME to run results ? would there be differences? I suspected that NLME may be easier for novice and students to catch up during short-student rotations to run simulations then to learn NONMEM on a 4 week rotation. Any opinions? For simple popPK problems I don't think there is really much difference in ease of use between NONMEM and Phoenix NLME. A NM-TRAN control stream is probably simpler to learn how to write for this kind of very simple PK problem rather than dealing with the layers of the Phoenix GUI. Phoenix graphics are better than NONMEM but there are many ways to make graphs from NONMEM using other tools such as Xpose. There won't be any important differences in the results. If you have a NM-TRAN control stream or Phoenix NLME workflow set up then as you add more data and want students on rotation to learn the basics of popPK then running the model again is only a matter of seconds. The real time required is learning the principles of PK models and population analysis and how to interpret the results. Software run times are negligible compared with that. > Thank you in advance for your help. > > Winnie > > Winnie Seto, BScPhm, PharmD, MSc, ACPR, R.Ph. > > Therapeutic Drug Monitoring Coordinator & Critical Care Unit Pharmacist > > Clinical Manager > > Department of Pharmacy > > The Hospital for Sick Children > > 555 University Avenue > > Toronto, Ontario > > Canada M5G 1X8 > > phone: 416-813-6236 > > fax: 416-813-5880 > > Confidentiality Notice: This is a medical communication. This e-mail message, including any attachments, is for the sole use of the intended recipient(s) and may contain privileged and confidential information. > > Any unauthorized review, use, copying, saving, re-transmission, disclosure or distribution is strictly prohibited. If you are not the intended recipient, please contact the author by reply email and destroy all original and copies of this email message. > > ------------------------------------------------------------------------ > >