1. A generic platform approach for performing HPLC assays
The Yieldaliser
A generic platform approach for performing HPLC assays
Phil Borman
1st April 2008 1

2. Summary i.) Background ii.) Development of the ‘Yieldaliser’
iii.) Universal external standard approach
iv.) Sample preparation protocols
v.) Software
vi.) Proof of Concept
vii.) Utility

3. Background
Generic* open access# HPLC systems as a tool for chemists to generate impurities data have been in operation within Chemical Development, GSK for over 10 years
The uptake of these systems has been very high
The application of traditional techniques such as TLC are now rarely used
Individually crafted HPLC methods are not used as often
Time, resource and cost savings
*Generic – relating to a wide range of compound applicability
#Open Access – Available for use by all (non experts)

4. Background
In addition to obtaining impurity profile data, the availability of HPLC assay data via a similar platform* system was identified as a valuable tool for process development chemists
Results of a survey undertaken in 2002 indicated that developing a Yieldaliser platform would meet the demands of an increasing number of requests for:
Yield in Solution samples
Assay of isolated solids
Chemists indicated that they require precision of ±3%
*Platform – Standard technical solution

5. Development of ‘The Yieldaliser’
A number of ideas were investigated based around the existing HPLC open access methodology
Standardization Type
Analysis of Standard
Analysis of Compound
External
Uses single λ with broad bandwidth ( nm)
Uses 4-5 standards each suited to different 20nm λ bands
Analyse compound against standard with closest λmax
Internal
Internal Std used in diluent used to prepare compound solution
Analyse standard at its λmax
Analyse compound at its λmax

6. Development of ‘The Yieldaliser’
The Yieldaliser was integrated into the existing open-access HPLC platform (non-GXP)
All materials are assayed against a stable reference material (external standard)
Benzophenone
Enables rapid estimation of yield in solution and assay of isolated solids without the user having to routinely prepare standards
Standards are only prepared on one occasion in order to generate a response factor

7. Standard preparation External Standard:
Benzophenone was chosen
commercially available, cheap and available in high purity
Very soluble in compatible solvents (MeOH, MeCN)
Good long term solution stability
Good chromatographic properties on 8min gradient HPLC method
Very high extinction co-efficient at its max of 254nm
1 stock solution of Benzophenone can be prepared periodically and transferred into HPLC vials for daily use
0.35 mg/mL (typically prepared using 350 mg/L)
MeOH used as the dissolving solvent

8. Standard Solution analysed at start of each day
i.) Automatic daily start up routine (~5:30 a.m.)
Blank Injections
Test mix
ii.) Following this, 6 standards of benzophenone are injected and the %RSD (precision) and average response of the benzophenone* (accuracy) are checked against pre-determined limits.
*Note: each system has a different limit for the response

9. Example of Daily Calibration
Standard vial: 94
Standard reference:
Standard conc (mg/mL):
Standard areas:
Standard mean area:
Standard area RSD:
Standard response:
Calibration completed at 07:00:33 on 18 February 2008
Calibration successful.
Removed
Information ed to Lead User on a daily basis
If system fails calibration then system will disable assay functionality

10. Adding Compounds to the ‘The Yieldaliser’
A Lead User has to perform a couple of one-off tasks for each compound
Calculate the compound's response factor (using its UV response at max or most suitable ) against benzophenone (at max of 254nm)
Compound, response factor and dissolving solvent used is saved within a globally accessible file
The assay wavelength and retention time window are also stored in the file, and this compound-specific information is automatically recalled when a Chemist uses the system.
Full Protocol

11. Protocols for sample preparation
Liquid samples
Original protocol aimed at solution concentrations of between 40 (25 mg/mL) and 2 (500 mg/mL) volumes
20 L of sample added to 20 mL of solvent*
Additional protocols for dilute and concentrated liquid samples were added to the system
1 mL of sample diluted with 20 mL of solvent* for >40 volumes
20 L of sample added to 20 mL of solvent (solution A) and then 1mL of solution A diluted with 20 mL of solvent* for <2 volumes
*Methanol or Acetonitrile

12. Sample preparation for mg/mL solution sample assay protocol
For addition of 20 mL of solvent
25 L EDP pipette
Syringe for ‘problem’ solutions (e.g. DCM)

13. Protocols for sample preparation
Solid samples
Approximately 15 mg of sample (accurately weighed#) dissolved in 1 mL DMSO and then diluted with 40 mL of solvent*
#Using a five placed balance to meet ±3% expectation of chemists
*Methanol or Acetonitrile

14. Balance & disposable containers for % w/w solid assay protocol

15. Sample preparation for % w/w solid assay protocol
For addition of 40 mL of solvent
For addition of 1 mL DMSO

16. Integration into existing Open Access platform
Transition Point – Need to tell the audience that we are moving from sample prep to analysis

17. Sample Log-On
Extra protocols added to system after system was in operation for 1-2 years
Feedback from Chemists highlighted need for dilute and concentrated solutions to be analysed

18. Management of information
Results are ed out to chemist
Assay data are quoted in
mg/mL for solution samples
%w/w for solid samples
The system records metrics
Number of samples analysis
Type of analysis used
Compounds analysed
Users
Facilitates continuous improvement

19. Proof of Concept Results
Note Outliers – reason?

20. Proof of Concept Results
Met Chemists expectations
Note Similar results observed for liquids protocol

21. Improving method variability
Perform replicate measurements (n) and take mean of results
If standard deviation = 1.5% taking
Mean from n=2 gives standard error of σ/2 = 1.06%
Mean from n=3 gives standard error of σ/3 = 0.87%
Improve uncertainty associated with:
Liquid Handling devices
Electronic dispensing pipettes and syringes
Most challenging aspect as wide range of solvents are used
Pressmatic dispensers
Balance

22. Examples of how Yieldaliser is used
In order to design the parameters of a final stage crystallisation, accurate solubility curves (concentration versus temperature) were required.
Compound X is crystallised from acetone by addition of water.

23. Examples of how Yieldaliser is used
Reaction Monitoring
Process Mass Balance Determinations
Determination of Crystallisation Endpoints
Purity Assessment of intermediates

24. Metrics and Continuous Improvement
>50 Open access HPLC systems (with Yieldaliser functionality) exist across 8 sites (R&D and Manufacturing)
16 systems are being used for assays
~200 Compounds have been added to platform to date
On average ~ assays performed per month across GSK sites
>95% of assays performed on solutions
62% of samples use typical solution protocol
35% of samples use dilute solution protocol

25. Other uses of relative response factors based on a single reference material
Implementation of a GXP version of the Yieldaliser with more precise sample preparation protocols
Standardize way majority of HPLC assay methods are operated
Eliminates diversity
Reduce the need for standards for each analysis
Estimation/quantification of related substances/impurities
Evolving Quality by Design approaches for analytical methods* may lead to more flexibility in changing methods
*The Application of Quality by Design to Analytical Methods
Phil Borman et al, Pharm Tech Volume 31, No 10 , Oct 2007 , P

26. Acknowledgements Development of Yieldaliser
John Roberts
Barbara O’Reilly
Ian Barylski
Robin Attrill
Keith Freebairn
Jason Cooke
Steve Wynn (WynnIT)
Current HPLC (Yieldaliser) Platform Owners
Steve Cole
Bill Young
James Roberts
Helen Weston

27. Back-Up Slides

28. Suitability for Yieldaliser
Safety as open access system, typically don't produce response factors for very active (OHC 4 ) compounds on safety grounds (i.e. use of sharps with protocols, general open access safety)
There needs to be a decent set of analytical data backing up the standard before we use to determine response factor. (e.g. ID, imp profile, Water, Solvents) to assign purity. Also compounds must have significant chromophore at >230 nm to be considered. Testing must show good stability (preferably >24 hours), also good reproducibility utilising method (expectation that all 3 weighings come in with a %RSD <1.0%)
3. Implications of compounds with very low (<0.1) or very high (>1.5) response factors wrt to Benzophenone need to be handled. Typically in these cases compounds are judged not suitable for Yieldaliser approach.
Back

29. Preliminary candidate compound assessment:
Identify sample solvent, check solubility under protocol conditions at 0.35mg/mL.
Identify suitable lambda max or min by UV peak apex scan, check generic integration adequate.
Check peak response at chosen wavelength is >30mAU (uncertainty, chemist column loading)
Check for injection carryover
Six replicate injections for reproducibility & short term stability.
Six hour stability in solution
Back

30. Response factor determination, injection sequence
Blank x 2, both wavelengths
Benzophenone Std A x 2, 254nm
Benzophenone Std B x 2, 254nm
Compound solution 1 x 2, chosen wavelength
Compound solution 2 x 2, chosen wavelength
Compound solution 3 x 2, chosen wavelength
Back

31. Response factor determination, calculation & data entry
Automatic response factor calculation for compound.
Check data set uncertainties for benzophenone standards and response factor are within limits.
Enter data required by system, (restricted access)
Inform chemist of sample solvent to be used for Yieldaliser assays.
Back

32. Yieldaliser Calculations
Solid Protocol
Yield (%w/w) = ((CmpArea/(amount/41))/rrf)*(extconc/extmeanarea)*100
Solution Protocol (typical)
Yield (mg/mL) = (CmpArea*(20020/20)*extconc)/(extmeanarea*rrf)
Solution Protocol (Dilute)
Yield (mg/mL) = (CmpArea*(21000/1000)*extconc)/(extmeanarea*rrf)
Solution Protocol (Conc)
Yield (mg/mL) = (CmpArea*(20020/20)*(21000/1000)*extconc)/(extmeanarea*rrf)