AdvancedTools in HPLC methoddevelopment Remco Stol, Enrico Martina and Jeffrey Vos Analytical Sciences Chemistry, Quality Unit API/BT NL FHI symposium, Houten, 22 april 2010
What does the customer want? Be able to determine compounds X, Y, Z at levels of Control quality of product Y, sourced from multiple suppliers Identity, assay, impurities (0.05% or trace) Monitor stability Trouble shooting, once-a-time analysis Routine test Cheap Without issues Fast 2
Typical HPLC Method Development Goals Achieve separation selectivity Quantitative Sensitivity Linear Accurate Precise Robust, rugged 3
Typical system failures in routine operation and method development Resolution Retention time Precision Tailing factor, plate number Studied in detail Not studied Not studied Somewhat studied, SST setting 4
Optimization parameters Continuous parameters easy to adjust Solvent strength (% modifier) Temperature Gradient (slope, start- end %) Discontinuous parameters Hard-to-adjust Column type ph Organic solvent Salt concentration, type (buffer, ion pairing) Old way : 1) select discontinuous parameters, test, find optimum, next parameter (OVAT) 2) optimize continuous parameters Strange approach: Most relevant selectivity parameters are of discontinuous type Current way : 1) screen discontinuous parameters (e.g. column & ph scouting) 2) optimize continuous parameters 3) automate where possible Better approach: Most relevant selectivity parameters are of discontinuous type 5
Start screening right Relevant samples Stress & stability samples Samples from different suppliers/sources Solvents Typically ACN and MEOH, others are possible Select set of columns based on chemistry pka, functional groups, Log P Detection principle UV, λ max Other, MS Select buffers MS compatible UV transparent Sample and column stability Peak tracking (minor added value in this phase) MS and/or DAD 6
Discontinuous parameter screening 7
Discontinuous parameter screening Flow: 0.5 ml/min Column temperature: 20ºC Injection volume: 5ul Concentration: 1 mg/ml Detection wavelength: 210 nm Buffers: 100 mm potassium hexafluorophosphate, ph 2 Modifier: acetonitrile Columns: - Chiracel OD-RH - Chiracel AD-RH - Chiracel OJ-RH - Chiracel AS-RH R R R Screening performed with racematic mixture of all four forms (SS, RR, RS and SR) R 8
Discontinuous parameter screening mau 12 10 8 DAD1 A, Sig=210,4 Ref=off, TT (D:\CSADATA\INSTR1\SCH900435-K\DATA\00003582 2010-02-25 12-19-22\00003582.D) 19.685 Area: 242.577 23.592 25.809 Area: 17641.6 Area: 17491 OJ-RH, 38% ACN mau 14 12 10 DAD1 A, Sig=210,4 Ref=off, TT (D:\CSADATA\INSTR1\SCH900435-K\DATA\00003623 2010-02-26 20-47-02\00003623.D) 15.748 24.493 Area: 17639.1 Area: 17667.7 OD-RH, 40% ACN 6 8 18.662 Area: 251.734 4 2 6 4 2 28.150 Area: 257.221 0 0 10 20 30 40 50 min 10 20 30 40 50 min mau DAD1 A, Sig=210,4 Ref=off, TT (D:\CSADATA\INSTR1\SCH900435-K\DATA\00003524 2010-02-23 17-02-29\00003524.D) 7.879 11.988 Area: 17465.3 Area: 17404.1 AS-RH, 42% ACN mau DAD1 A, Sig=210,4 Ref=off, TT (D:\CSADATA\INSTR1\SCH900435-K\DATA\00003545 2010-02-24 07-42-07\00003545.D) 7.949 9.061 Area: 479.09 Area: 34318.3 AD-RH, 37% ACN 25 20 20 7.130 Area: 244.221 15 15 10 10 5 0 9.669 Area: 55.1914 15.441 Area: 18.3398 5 10.883 Area: 73.9407 0-5 5 10 15 20 25 30 min 5 10 15 20 25 30 35 40 min 9
Discontinuous parameter screening mau DAD1 A, Sig=210,4 Ref=off, TT (D:\CSADATA\INSTR1\SCH900435-K\DATA\00003623 2010-02-26 20-47-02\00003623.D) DAD1 A, Sig=210,4 Ref=off, TT (D:\CSADATA\INSTR1\SCH900435-K\DATA\00003625 2010-02-26 22-46-21\00003625.D) DAD1 A, Sig=210,4 Ref=off, TT (D:\CSADATA\INSTR1\SCH900435-K\DATA\00003621 2010-02-26 18-47-41\00003621.D) 14 12 10 8 6 4 45% ACN 40% ACN 35% ACN 2 0 10 20 30 40 50 min Flow: 0.5 ml/min Column temperature: 20ºC Injection volume: 5ul Concentration: 1 mg/ml Detection wavelength: 210 nm Buffers: 100 mm potassium hexafluorophosphate, ph 2 Column: Chiracel OD-RH 10
Evaluation of Scouting data Pick a winner strategy: optimal discontinuous parameters based # of peaks detected, (critical) resolution Use modeling and chemometric techniques to evaluate data Model can be based on retention or trend response Resolution vs temperature Resolution compound x-y vs % modifier Peak area main vs %ACN Peak height second largest peak Typical goal is to exclude conditions not to optimize! Peak tracking is typically not essential (efficient) during scouting 11
Fine-tuning using Design of Experiments (DoE) A two level factorial design (including center points) was used to optimize the separation and estimate robustness. Variables were: Flow rate Amount of modifier Column temperature Buffer concentration responses were: Resolutions Tailing factor Retention time 12
Estimating optimum conditions Software (design expert) calculates the optimum conditions in the region tested. Ranges tested were: Flow rate in range: 0.3 0.7 ml/min. Amount of acetonitrile in range 38% - 42% Column temperature in range 20 40 C Buffer concentration in range 50 150 mm Maximize resolutions Minimize tailing factor and retention time 13
Optimum conditions Flow: 0.5 ml/min Column temperature: 25ºC Buffer concentration: 150 mm potassium hexafluorophosphate, ph 2 Amount acetonitrile: 38% Injection volume: 5 ul Detection wavelength: 210 nm Concentration: 1 mg/ml Column: Chiracel OD-RH 14
Estimating robustness DoE software capable of simulating data Design expert 6 (stat-ease, inc.) Inputs are the regression coefficients from the equation derived during optimization. (i.e. Y = A + B 1 X 1 + B 2 X 2 ) Be carefull, extrapolation of regression models can be prone to errors. 15
Estimating robustness Resolution 27.00 Overlay Plot Rs 2,1: 1.978 26.00 B: column temperature 25.00 6 24.00 Rs 3,2: 1.418 23.00 0.45 0.47 0.50 0.53 0.55 A: flow rate 16
Estimating robustness Peak width and resolution 27.00 Overlay Plot Rs 2,1: 1.978 26.00 B: column temperature 25.00 6 peak width peak 1: 2.298 peak width peak 4: 4.668 24.00 Rs 3,2: 1.418 23.00 0.45 0.47 0.50 0.53 0.55 A: flow rate 17
Estimating robustness Retention time, peak width and resolution 27.00 Overlay Plot Rs 2,1: 1.978 B: column temperature 26.00 25.00 Rt peak 5: 23.67 Rt peak 4: 20.62 Rt peak 2: 14.4 Rt peak 1: 12.72 peak width peak 4: 4.668 peak width peak 1: 2.298 6 Rt peak 2: 17.6 Rt peak 1: 15.54 Rt peak 4: 25.2 Rt peak 5: 28.93 24.00 Rs 3,2: 1.418 23.00 0.45 0.47 0.50 0.53 0.55 A: flow rate 18
Conclusions Automation and smart solutions used throughout method development Discontinuous scouting is fully automated Experimentation is software supported Experiments are programmed by the software 24/7 unattended experimentation The experimental design model is build throughout the fine tuning The model allows not just better method development, it allows robustness to be a real goal during method development. Automation helps to get better (rugged) methods into routine operation 19
Next goals Introduce DoE during scouting Incorporate numerical and categorical variables: Algorithm design, others. Use scouting data for model building Replace formal robustness testing from validation PAR in addition to set-point values in (validated) method descriptions Get acceptance from internal customers Get acceptance from external customers 20