Back to the Top
Hello Everyone,
I am working on a small study observing the concentration of a drug in obese patients. There is
only 7 patients and they were divided into two groups, one with 3 patients and the other with 4
patients. The concentrations were taken at various times between the patients. For example, some
patients had their concentrations drawn up at 31 minutes, while others at 38 minutes. I was
wondering what the best approach would be to averaging the patients together since their times are
all different. I am hoping to average the 2 groups so I can compare them graphically.
Thanks,
Joe
Back to the Top
Try to keep the times closer than you have now
Back to the Top
Hi, why you need to take average?
Just for graphical representation? I suggest use data as near as possible to make average. To make
PK calculations, you don't need obtain average because you must create data base with real time,
winnonlin for example can calculate PK parameters using real time.
Best regards
Jose
Back to the Top
Joe,
Why do you want to average the patients? Why not fit your model across all patients simultaneously
and see what fits best? This is very easy with GastroPlus(TM) and probably many other software
programs.
If you want to send the data - times and concentrations plus individual doses and body weights, I
would be happy to fit them for you. We don't need to know anything else about the compound to do a
quick estimate, although of the doses are all oral it would help to know the bioavailability if you
know it.
Best regards,
Walt Woltosz
CEO
Simulations Plus, Inc.
and Cognigen Corp
Back to the Top
Joe, Jose is correct you can simply fit your data as you have recorded it either individually,
simple naïve-pooled or as a population (NLME) model with a very user-friendly graphical model
builder and some powerful algorithms.
If you want to see how this might look you could post your data file (e.g. as CSV or Excel) to the
"modelling & simulation" category on www.certara.com/forums and I'd be happy to take a quick look
at it.
Best regards,
Simon.
Simon.Davis.-a-.certara.com
Senior Scientific Consultant
Back to the Top
I absolutely agree with you all and that is what I did. I used Winonlin to fit the data. These two
groups have significant differences in Vd and Cl. The reason I am trying to average these groups
separately is because I want to compare the two different groups against the MIC range. Instead of
comparing individual patients against the MIC, I want to compare them as a group graphically against
the MIC to avoid multiple plots. Is it advisable to group them in this way?
Thanks,
Joe
[Why don't you simulate the data using the fitted parameter values, maybe as lines - db]
Back to the Top
An approach usually taken in studies intended to be submitted to a regulator is to have the method
of how comparisons will be done documented before analysis in a protocol and/or SOP, and that
protocol or SOP normally also describes how large a deviation between the nominal sample time
(30mins = 0.5hr) and the actual sample time is allowed, often with a 5% limit (38min=0.633 Hrs
26.6% deviation !).
So in your case you could model the individual profiles and predict values at the nominal time, and
then average those predicted values, or accept the large deviation and just average the vlaues that
were around 0.5Hr.
Paul.
--
Paul Hurley BSc MSc MBCS
Global Application Administrator, Global Business Technology (GBT)
Back to the Top
Dear Joe:
Make a population model. You can accommodate all the different times of the samples just fine, and
get the max information from the data. You might go to www.lapk.org and click on Pmetrics.
All the best,
Roger
Back to the Top
Joe,
Use a mechanistic model and understand better what is really going on with
your drug/formulation. Stop messing around with simplistic
empirical/statistical models that hide the underlying chemistry,
physiology, with such foolishness as constant Ka absorption models and
learn more about your formulation and drug behavior before you waste time
and money with cut-and-try methods.
This is why the FDA and other regulatory agencies are pushing for
mechanistic absorption modeling (MAM) and physiologically based
pharmacokinetics (PBPK) to be considered more and more for clinical
pharmacology in addition to earlier preclinical and clinical studies.
Best regards,
Walt
Walt Woltosz
Chairman and CEO
Simulations Plus, Inc. (NASDAQ: SLP)
and Cognigen Corp, a wholly owned subsidiary of Simulations Plus
42505 10th Street West
Lancaster, CA 93534
Back to the Top
Hi Joe,
I think the main question is what the objectives of the study were. If the objective was to compare
the two groups, I'm afraid comparing 4 vs. 3 patients won't add that much to your knowledge about
the pharmacokinetics of your drug. This knowledge gained becomes even less when you just look at the
mean values between the two groups. And on the top of that, you have different time-points of
samplings, which makes the mean comparisons even less meaningful.
A better way, as suggested by others on this thread, would be to pool all the data together and
develop a compartmental model based on the observations. This way, you can estimate important PK
parameters and the variability around them (albeit based on a very limited number of subjects). Such
a model can be used to plan your future studies with a better idea of what to expect (through
simulations). Furthermore, data from future studies can be added to the current data set to enrich
the model and its estimates.
Toufigh
Back to the Top
I appreciate all of you taking the time to provide me with different approaches. I know the
question isn't easy to answer without seeing all of the data and that the sample size is not ideal,
but your input has helped a lot.
Thanks,
Joe
Back to the Top
Joe
I broadly agree with Toufigh but I would just point out that his approach works if the
bioavailability is high, not if it is low. For example, if the median of all 7 subjects is 80%
then the range is most unlikely to be over 100% but if the median is, say 10%, then it would be
quite likely that one of the subject will be 2 or 3 times that of the lowest ... Which may be the
answer you are seeking in the first place.
I once was the medical messenger of a project where the the range of bioavailabilities in a group of
12 subjects was 1000 fold. They wanted to shoot me because when they asked me to conduct a clinical
trial I asked them back "what dose should I use?"
Andy Sutton
Back to the Top
I was once charged with a biocompatibility assay of an experimental drug with a marketed product. I
sent out the bioanalytical results. Everyone was down on me because the drugs were by PK different.
I was examined the method was examined the data was examined. I pointed out finally that the
collection times were very different between the two drug studies. One included a time point at 15
minutes which was the tmax and cmax for one drug. The other did not include this point but did have
one at 5 minutes which was the tmax and cmax for that drug. Pre analytical variables are
everything!!
Back to the Top
Andy,
I'm not sure why modeling of concentration-time data should be OK for
drugs with high bioavailability but not good for a low bioavailable
compound. In fact, I would argue that for compounds with low
bioavailability, one can expect a higher variability in drug
concentrations, which makes averaging and comparing small number of
subject (in this case 3 vs. 4) even less informative. Obviously, a
compartmental model on such a small number of subjects and high
variability is not going to provide accurate estimates of the model
parameters. The quality of the data (including the number of data points
available) affects the quality of the model. Nevertheless, I firmly
believe that a modeling approach is better than averaging and comparing,
regardless of the bioavailability.
Toufigh
Back to the Top
Ed,
When is the highest observed concentration equal to Cmax? NEVER!
It's not possible to sample exactly at Tmax - the odds are infinitesimally
small that any sample will be exactly when the plasma concentration is at
its peak. So it's a safe assumption that Cmax is always higher than the
highest observed concentration Maybe not much, but maybe a lot higher
A well-fitted model can provide the best estimate of Cmax and Tmax.
Statistical methods that "connect the dots" and appear to give a perfect
fit to each data point and make Cmax the highest observation are simply
wrong.
Best regards,
Walt
Walt Woltosz
Chairman and CEO
Simulations Plus, Inc. (NASDAQ: SLP)
and Cognigen Corp, a wholly owned subsidiary of Simulations Plus
42505 10th Street West
Lancaster, CA 93534
Back to the Top
Dear Toufigh:
On the other hand, even if you only have a few subjects, you might wish to get the maximum
information out of the data. Whatever the variability, low or high, you might well wish too get the
most info from the data. To me that suggests nonparametric population modeling, which estimated the
entire model parameter distributions, not just means and covariances, and which then lends itself to
multiple model (MM) dosage design. This develops dosage regimens which are designed to hit a desired
target goal specifically with max precision. Parametric modeling approaches cannot do this, as they
can only act on the single points estimates of an assumed distribution. You might look at
1. Bustad A, Terziivanov D, Leary R, Port R, Schumitzky A, and Jelliffe R: Parametric and
Nonparametric Population Methods: Their Comparative Performance in Analysing a Clinical Data Set and
Two Monte Carlo Simulation Studies. Clin. Pharmacokinet., 45: 365-383, 2006.
2. Jelliffe R, Schumitzky A, Bayard D, Milman M, Van Guilder M, Wang X, Jiang F, Barbaut X, and
Maire P: Model-Based, Goal-Oriented, Individualized Drug Therapy: Linkage of Population Modeling,
New "Multiple Model" Dosage Design, Bayesian Feedback, and Individualized Target Goals. Clin.
Pharmacokinet. 34: 57-77, 1998.
3. Jelliffe R, Bayard D, Milman M, Van Guilder M, and Schumitzky A: Achieving Target Goals most
Precisely using Nonparametric Compartmental Models and "Multiple Model" Design of Dosage Regimens.
Therap. Drug Monit. 22: 346-353, 2000.
4. Bustad A, Terziivanov D, Leary R, Port R, Schumitzky A, and Jelliffe R: Parametric and
Nonparametric Population Methods: Their Comparative Performance in Analysing a Clinical Data Set and
Two Monte Carlo Simulation Studies. Clin. Pharmacokinet., 45: 365-383, 2006.
5. Jelliffe R, Schumitzky A, Bayard D, Leary R, Botnen A, Van Guilder M, Bustad A, and Neely M:
Human Genetic variation, Population Pharmacokinetic – Dynamic Models, Bayesian feedback control, and
Maximally precise Individualized drug dosage regimens. Current Pharmacogenomics and Personalized
Medicine, 7: 249-262, 2009.
6. Neely M, van Guilder M, Yamada W, Schumitzky A, and Jelliffe R: Accurate Detection of Outliers
and Subpopulations with Pmetrics, a Nonparametric and Parametric Pharmacometric Modeling and
Simulation Package for R. Therap. Drug Monit. 34: 467-476, 2012.
Very best regards,
Roger Jelliffe
Roger W. Jelliffe, M.D., F.C.P., F.A.A.C.P.
Professor of Medicine Emeritus,
Founder and Director Emeritus
Laboratory of Applied Pharmacokinetics
USC School of Medicine
Consultant in Infectious Diseases,
Children’s Hospital of Los Angeles
4640 Sunset Blvd, MS 30
Los Angeles CA 90027
Back to the Top
Dear Sir,
The sample collection timing are very important in PK studies particularly working with drugs that
follow multiple compartment models. The frequent sampling at early stage may give very interesting
results that may differ from data based on low frequency of sampling.
Dr. Zafar
|
Want to post a follow-up message on this topic?
If this link does not work with your browser send a follow-up message to PharmPK@lists.ucdenver.edu with "Sample Size" as the subject |
Copyright 1995-2014 David W. A. Bourne (david@boomer.org)