1. Trials and Tribulations of Developing a Method for Quantification of Aerosol Chamber Concentration Richard Snell, Dose Concentration Analysis GSK

2. Overview Inhalation chamber concentration analysis Meth Dev and Validation of methods for aerosol chamber concentrations The problem Child  Method Development and Validation  Study Samples  The Investigation  We Have a Winner

3. Chamber Concentration Analysis - 1 7 or 14 day candidate selection studies supported including 2-3 weeks prelim phase We perform:  Concentration checks using GF/C filters  Particle size distribution analysis using washings of stainless steel foils and back-up GF/A filters Samples quantified against calibration line Quality control samples run to check performance of assay – must be within 10% of nominal concentration UPLC is the preferred system due to faster run times leading to shorter turnaround times.

4. Chamber Concentration Analysis - 2 Particle size distribution (PSD) solutions and GF/A Filters: Cascade impactor used in place of animal tube: contains up to 8 stainless steel foils + back-up GF/A filter Foils trap different size particles - particle size gets progressively smaller towards the rear of the impactor with GF/A filter trapping smallest particles. Foils from stages 3-8 are collected and washed with 3 mL of diluent An aliquot is analysed from each 3 mL washing solution of foils 3-8 GF/C Filters: Aerosol chamber concentration check filters used in place of animal tube Gives representation of the dose animals will be exposed to

5. Method Development and Validation - 1 Solubility check - 3 to 5 mg weighed, analytical diluent added in 100 µL steps (increased if no signs of going into solution) Compound dissolution assessed visually. λ max Determined by running samples prepared in diluent on a UPLC system equipped with a photo diode array (PDA) Chromatography developed Linearity investigated Interference from filters checked

6. Method Validation  Standard line bracketing QCs  QCs at 2 Levels – HLQ and LLQ  One QC solution prepared at each level, 10 mL added to each filter type in replicates of 6  6 replicates of blank filters  10% acceptance criteria on standards and QCs QCs re-injected with freshly prepared calibration line after storage at room temperature to establish stability Method Development and Validation - 3

7. The Problem Child – Method Development and Validation - 1 Solubility in initial diluent (50/50 MeCN/Water) was good Linear range of 0.25 to 30µg/mL QCs at low level failed to meet acceptance criteria. Re-run, as usual reason for failure is spiking error, or difference when spiking vastly different volumes This time marked difference between QCs containing filters and those without, ≈ 15-20%, indicating drug binding to the filters

8. High and low filters not exhibiting same trend, leading to a change in diluent. Consistent with binding. 40/40/20 (MeCN/MeOH/Water) was tried, resulting in the same approximate 20% difference between QCs with and without filters LLQ Raised to 0.5ng/mL Investigate different Solvents as diluent  Addition of 0.1% Formic acid to diluents  Addition of 0.05% Triflouroacetic Acid (TFA)to diluents  Aqueous Tertrahydrofuran (THF) 20%, 30% and 50% The Problem Child – Method Development and Validation - 2

9. The Problem Child – Method Development and Validation - 3 Addition of acid had no effect 50%, and 30% THF both had detrimental effect on chromatography 20% no effect on chromatography and appears to stop the compound binding to filters Validation run and stability runs successful METHOD VALIDATED................. Or so we thought...

10. First submission of chamber characterisation sample GF/C Filters only - run fails standard line fails and QCs outside acceptance criteria Repeated with fresh line and QCs, same issues, on closer examination line exhibits an exponential rather than linear correlation The Problem Child – Study Samples - 1

11. The Problem Child – Study Samples - 2 Consistent with non-specific binding Since study required Low levels, range truncated 0.5 to 8µg/mL FAILURE! No improvement on linearity. Samples sent for analysis by another group to eliminate our system

12. The Problem Child – Study Sample - 3 Analysis shows – Compound sticking to glass as well as filters  Worse at lower concentrations  Worsens over time

13. All glassware eliminated from extraction procedure, and replaced with polypropylene Raised LLQ to 1µg/mL Highlights compound still binding to filters Binding worse at higher levels indicating that in the absence of glass compound has an affinity for itself Options,  Different Filter types  Changes to diluent addition of NH 3 or TEA  Change to the way compound is trapped The Problem Child – The investigation

14. The use of Quartz filters were tried but did not offer any advantage The addition of 1% NH 3 to 40/40/20 (MeCN/MeOH/Water) The Problem Child – We have a Winner QC 1 GF/A QC 1 GF/C QC 30QC 30 GF/AQC 30 GF/C 11.0361.0201.02629.66929.84429.867 21.0281.0231.03729.57929.89629.544 31.0391.0181.04229.68129.75830.422 41.0441.0181.03629.72230.19130.172 51.0351.0241.03330.63129.95030.136 61.0431.0261.02929.56730.18030.283 Mean1.0371.0221.03429.80829.97030.071 S.D.0.0060.0030.0060.4080.1780.317 Precision (%CV) 0.60.30.51.40.61.1 Bias %3.72.23.4-0.6-0.10.2 n666666

15. Acknowledgements Jessica Benton Aida Merchan Bob Gibbon Alan Ferrie Kay Rush