Aucun titre de diapositive

1
TOWARDS A RISK BASED METHODOLOGY TO ASSESS THE
ACCEPTABILITY OF AN ANALYTICAL METHOD TRANSFER
COMPARISON OF DIFFERENT APPROACHES. E. Rozet1,
W. Dewe², B. Govaerts³, B. Boulanger², A.
Ceccato2, P. Chiap4, J. Crommen4 and Ph.
Hubert1 1 Laboratory of Analytical Chemistry and
4 Laboratory of Analytical Pharmaceutical
Chemistry, Bioanalytical Chemistry Research
Unit, ULg, B 36, B-4000 Liège, Belgium. 2 Lilly
Development Centre, rue Granbompre, 11, B-1348
Mont-Saint-Guibert, Belgium. 3 Institute of
Statistics, Université Catholique de Louvain, 20,
voie du Roman Pays, B-1348 Louvain-la-Neuve,
Belgium.
ANALYTICAL METHOD TRANSFER DEFINITION AND
OBJECTIVE

• Compute the probability P of having results
  falling outside the acceptance limits and compare
  it with the maximum tolerated risk 1- ß.
• If P ? 1- ß ? the transfer is accepted
• If P gt 1- ß ? the transfer is rejected

• Definition
• An analytical method transfer consists in
  transferring a previously validated analytical
  method from a sending laboratory (called
  sender) mastering this method, to a receiving
  laboratory (called receiver). Experiments are
  made in both laboratory in order to demonstrate
  the acceptability of the transfer.

EXAMPLES
Example 1 Transfer of a LC-UV method for the
quantitation of Fenofibrate and Fenofibric acid
in Lidoses 2 Results of transfer are shown in
Table 2. The acceptance limits (?) for
Fenofibrate and Fenofibric acid were set at 5 and
10 respectively. The maximum risk (1-ß), was set
at 10 for both analytes. As the maximum risk of
10 is not exceeded, the transfer is accepted.
Indeed, the effective risk of having results out
of specifications during routine use, i.e. Risk
C, at the receiver is of 1.9 for Fenofibrate and
6.6 for Fenofibric acid.

• The Objective of a transfer is therefore
• To provide users the guarantee that each
  measurement (xi) on unknown sample is close
  enough to the true value (mT).
• Enough, means for example less than 15 away
  from sample true unknown value (l acceptance
  limits, e.g. 5 (dosage form), 10 (impurities)
  or 15 (bio-analysis)).
• Guarantee, means that it is very likely that
  whatever the measurements, it will be close
  enough from the unknown true value (1-ß risk),
  e.g. 5, 20 or 33.

WHAT RISKS DURING ANALYTICAL METHOD TRANSFER ?
Table 2. Results of the transfer for example 1.
3 risks to take into account A. Type I error
rejecting an acceptable method transfer B. Type
II error not rejecting an unacceptable method
transfer C. Risk of having measurements out of
specifications during routine use of the
method at the receiving site
Example 2 Transfer of an on-line SPE-LC-ECD
method for the quantitation of 3 catecholamines
in human urine Results of the transfer are shown
in Fig. 3 as risk profiles for norepinpehrine
(NE), epinephrine (E) and dopamine (DA) at each
concentration level evaluated during method
transfer. The acceptance limits ? were set at
15. The transfer is rejected if the observed
risk is above the maximum risk accepted. As can
be seen, the risk is smaller than the maximum
risk of 5 for all the catecholamines studied
irrespective of the concentration levels.
Therefore, the risk of having results out of
specification during routine use of the method
for the receiver is of maximum 2.5, 3.4 and
3.1 for NE, E, DA respectively.
CLASSICAL APPROACHES
The classical approaches assess separately two
criterion trueness (systematic error) and
precision (random error). Table 1 shows the risks
controlled by each approach.
Table 1. The classical approaches used to
accept/reject a method transfer.
As can be seen in Table 1, no risks are managed
by both descriptive approaches. The difference
and equivalence approaches partially manage risks
A or B. However, risk C, that is the risk of
having measurements out of specifications, is not
managed by any of the classical approaches.
Furthermore when combining back those two
criterion there is a non negligible risk to
obtain a high proportion of unacceptable results
at the receiving site as illustrated in Fig. 1.
Fig. 3. Risk profile for example 2.
CONCLUSIONS
The objective of an analytical method transfer is
to provide users guarantee in order to minimize
the risks to have future results out of
specifications. The total error approach achieves
this by computing the risk of having future
individual results outside the acceptance limits.
It gives the guarantee that the receiver masters
the analytical method and furthermore allow to
manage the risks of having results out of
specifications during routine use 3. The
classical approaches should be avoided to
evaluate the acceptability of an analytical
transfer.
Fig. 1. Risk in routine with classical
approaches.
TOTAL ERROR BASED APPROACHES
AKNOWLEDGMENTS
Research grant from the Walloon Region and the
European Social Fund to one of the author (E.
Rozet) is gratefully acknowledged (First Europe
Objective 3 project n215269).

• A novel total error based approach is proposed
  1 the Risk approach
• Compute the ß-expected tolerance interval of the
  receiver results LRUR, i.e. the interval in
  which one can expect that at least a proportion ß
  (e.g. 90) of future individual results will lay.
• Set a priori acceptance limits ?. The ideal
  acceptance region is 1- ?1 ?
• Take into account that the sender gives results
  with more or less uncertainty compute the
  confidence interval of the mean of the sender
  results LSUS. This penalize the acceptance
  region (1- ?)US(1 ?)LS
• Compare the ß-expected tolerance interval of the
  receiver results LRUR to the acceptance
  region (1- ?)US(1 ?)LS

REFERENCES
1. Dewé, W. and al. Chemom. Intell. Lab. Syst.
Accepted for publication. 2. Rozet E. and al.
J. Pharm. Biomed. Anal. In press. 3. FDA,
Process Analytical Technology (PAT) Initiative,
2004.