Hi everyone!Back to the Top
In a study I have to do a PK analysis on, we had quite a few deviations
of actual sampling times from planned ones (12 sampling times within 24
hours, deviations of up to 30 minutes from planned sampling times).
Does anyone have THE way to deal with this problem? Do you suggest
using the actual sampling times always (a LOT of work) or does anyone
suggest a "threshold" or do you simply ignore this and accept some
additional inaccuracy in estimates of e.g. tmax?
Thank you all for your thoughts and sharing your experiences!
Regards
Hanns
Dear Hanns,Back to the Top
I always follow the simple rule of using the actual time of an event.
When people suggest to me that one should use the nominal time, I ask
them the following question: why do you then bother writing down the
actual time? We instruct the investigators, demanding that they write
down the exact time of an event, and then don't pay attention to the
noted time and use the nominal time! You have done a lot of work
getting the study done. Compared the amount of work put on the conduct
of the study, efforts put on getting the true time an event seems quite
small.
How much impact use of nominal vs. actual time will have on your PK
analysis depends on the kinetics of your compound. If you have a drug
with an elimination half-life of 2 weeks, it probably won't matter if
you use 48 hrs nominal instead of 48.5 hr actual. I think the question
of how much deviation you can tolerate depends on the rate of change of
the measured variable. However, if you have the actual time of an
event, I don't see any reason why not to use it.
Best regards,
Toufigh Gordi
Dear Hanns and Toufigh:Back to the Top
Of course, it makes a difference. It is always good (better) to
write down (look at the clock, or your watch, and write down the
military time when the event happened. With regard to aminoglycoside
therapy and the ability to achieve target goals precisely, knowing when
the dose was given of the infusion started and stopped was the single
biggest factor in being able to achieve a target goal, precisely. This
was done in a simulation study of a good ward, lab, pharmacy, and
phlebotomy service, in
Jelliffe RW, Schumitzky A, Van Guilder M: Nonpharmacokinetic Clinical
Factors Affecting Aminoglycoside Therapeutic Precision. A Simulation
Study. Drug Invest. 4: 20-29, 1992.
In addition, similar findings have been reported, comparing
untrained nurses versus trained pharmacy residents in their ability to
record data and the resulting therapeutic precision, in
Charpiat B, Breant V, Dumarest C, Maire P, and Jelliffe RW: Prediction
of Future Serum Concentrations with Bayesian Fitted Pharmacokinetic
Models: Results with Data Collected by Nurses versus Trained Pharmacy
Residents. Therapeutic Drug Monitoring, 16: 166-173, 1994.
Maybe this might be useful.
Very best regards,
Roger Jelliffe
Roger W. Jelliffe, M.D. Professor of Medicine,
Division of Geriatric Medicine,
Laboratory of Applied Pharmacokinetics,
USC Keck School of Medicine
2250 Alcazar St, Los Angeles CA 90033, USA
Phone (323)442-1300, fax (323)442-1302, email= jelliffe.at.usc.edu
Our web site= http://www.lapk.org
I suggest that you use that data that you have - actual measurementsBack to the Top
and actual times. Both have intrinsic errors but this is science. You may want to determine why the protocol was not followed in terms of timing, so that you do not fall into this hole again
Dave
In my view, you ALWAYS have to use the exact sampling times. a lot ofBack to the Top
work is no excuse. the only possible exception is a drug with a very
long half-life, where the (in this case known systematic) sampling
error is negleglible compared to other errors.
johan mouton
Johan W Mouton, MD PhD
Dept Medical Microbiology and Infectious Diseases
Canisius Wilhelmina Hospital C-70
Weg door Jonkerbos 100
6532 sz Nijmegen
The Netherlands
tel + 31 24 3657514
fax + 31 24 3657516
email mouton.-a-.cwz.nl
Hi,Back to the Top
In case of next time you may want to design your study using the
idea of sampling windows instead of specifying just the actual
sampling times. Obviously the length of the windows will depend on
the half-life of your drug.
Kayode Ogungbenro
Centre for Applied Pharmacokinetics Research
School of Pharmacy and Pharmaceutical Sciences
University of Manchester
Manchester, M13 9PL
U.K.
Back to the Top
Dear Hanns,
In general we always use the actual time. Only for summary statistics
per
time point we use the planned time.
Some pharmaceutical companies have policies to use only the actual time
in
case the deviation is more than 3 minutes. If that is more conveniant
for
you than you can do that too.
Deviations in sampling times have more impact when the sampling time is
closer to tmax. A deviation of 30 minutes can make a large difference
in AUC
in that case. Therefore, the actual time should be used.
Kind regards,
Berber Snoeijer
Biometric Support
The Netherlands
Dear All:Back to the Top
More about times. It has been popular in the past to use trough
serum samples because then the errors in recording the time at which
the sample was drawn make the least difference in understanding the
significance of the measured concentration.
However, that strategy turns out to be the time at which the
measured concentration actually has the least information about the
behavior of the drug in the patient. It usually has the least
information about the pharmacokinetic parameter values, as the various
rate constants or clearances all are time-related. It is must more
useful to use sampling strategies that are at least designed to contain
the most information about the PK parameter values. There are a number
of optimal sampling strategies, and the D-optimal one has been very
useful for over 30 years.
The general strategy is to get samples at times when a change
in the model parameter value(s) causes the greatest change in the serum
concentrations. These are the times when the partial derivatives of the
measured concentrations with respect to each parameter value are
maximum. A common useful pair, in the setting of intermittent IV
therapy, for example, is at the real peak, out of the opposite arm at
the end of the IV infusion, and at about 1.44 half-times later, when
the serum concentrations have fallen to about 36% of the original peak
value.
So if the expected trough is more than about 36% of the peak,
then it is a good time, but if, as with the aminoglycosides, for
example, the usual trough is much less than that, or maybe even
undetectable, then the time at which the expected concentration is
about 36% of the peak has much more information about the behavior of
the drug in that patient. This often is about 3 hours before the
trough. A good pair is a peak with the very first dose, and then
another sample at about 20 hours into the regimen. For patients with
CCr at least 50, this is a good pair, whatever the dosing frequency,
and if you center the patient's dosing around a time which is about 3
hours after routine morning blood drawing time, it gives a practical
approximation of a D-optimal strategy for those patients.
Many drug companies are using similar approaches to optimal
experimental design to make sure they get maximum information from the
serum concentrations they spend money to obtain.
A useful reference, with a very nice comparison of the advantage of
using a D-optimal strategy versus a conventional one for lidocaine, is
D'Argenio D: Optimal Sampling Times for Pharmacokinetic Experiments.
J. Pharmacokin. Biopharmacol., 9: 739-756, 1981.
Very best regards,
Roger Jelliffe
Roger W. Jelliffe, M.D. Professor of Medicine,
Division of Geriatric Medicine,
Laboratory of Applied Pharmacokinetics,
USC Keck School of Medicine
2250 Alcazar St, Los Angeles CA 90033, USA
Phone (323)442-1300, fax (323)442-1302, email=" jelliffe.-at-.usc.edu
Our web site=" http://www.lapk.org
PharmPK Discussion List Archive Index page
Copyright 1995-2010 David W. A. Bourne (david@boomer.org)