1 Modern HPTLC methods validation Application of prediction intervals to dextrine profiles of enzymatic digestion of starch and baking products J.M. Roussel, Consultant, Mâcon (France) M.P. Luquet, Lesaffre International, Marcq-en-Baroeul (France) G. Baeyens, Lesaffre International, Marcq-en-Baroeul (France)
2 Introduction: the assay methodAmylases are starch degrading enzymes particularly used for baking. Within the last 2-3 years, several new amylases have been commercialized, Differentiating them by traditional in vitro enzymatic assays is quite difficult, A HPTLC assay method for amylase-produced dextrins, ranging from glucose (DP1) to maltoheptose (DP7), which allow differentiation of the enzymes have been developed(&). (&)G. Baeyens, M.P. Luquet, (2016) CBS 117, 5-7
3 Introduction: the assay methodChromatographic layer: HPTLC plates silica gel 60 F254 (Merck), 20x10 cm Standard solution: Methanolic stock solution of all 7 dextrins (DP1 to DP 7), 1 mg/mL, diluted with methanol to the desired concentration levels Sample preparation (for enzyme characterisation): Aqueous enzyme solution (1%) was stirred for 20 min., starch solution (4%) was added 1:1 and incubated at 25 °C for 40 min. Heating at 100 °C for 5 min. stopped the reaction. Sample application: Bandwise with Automatic TLC Sampler (ATS 4, Camag), band length 6.0 mm, 27 bands per plate. Development: Automatic Development Chamber (ADC 2, Camag) with chamber saturation (filter paper) and conditioning for 10 min. at 47% relative humidity (saturated potassium thiocyanate). Development with acetonitrile-acetone-water 3:3:2 up to 60 mm. Drying for 5 min.
4 Introduction: the assay methodDerivatization: Immersion in aniline-diphenylamine-phosphoric acid reagent with Chromatogram Immersion Device (Camag) and heating at 120 °C for 5 min. with TLC plate Heater (Camag). Documentation: TLC Visualizer (Camag) under white light illumination (transmission) Densitometry: TLC Scanner nm, slit dimension 4.00 x 0.30 mm, scanning speed 20 mm/sec. Statistical evaluation: NeoLiCy® (R2 beta) software for analytical method’s life cycle statistical assessment (Prof. M. Righezza, Dr. J.M. Roussel).
5 Pre-validation: calibration function50 to 500 ng/track: DP6 based on peak height Y=(cX/(b+X))+a Y=a+bX+cX² Quadratic model show important residuals values for DP5, DP6 and DP7 (even values having no solution)
6 Pre-validation: calibration function50 to 500 ng/track: DP6 based on peak area Y=(cX/(b+X))+a Y=a+bX+cX² Peak area is preferred for both calibration functions
7 Validation experimental designCalibration samples: Standard solution, 4 concentration levels (50, 100, 300, 500 ng) → K concentration levels (k varies from 1 to K), 5 HPTLC plates, in intermediate precision conditions → I series (i varies from 1 to I), 3 repetitions per level and plate → J repetitions (j varies from 1 to J). 60 data points for calibration function assessment Accuracy samples: Spiked matrix, 4 concentration levels (50, 100, 300, 500 ng), 5 HPTLC plates, in intermediate precision conditions, 3 repetitions per level and plate. Specificity: From TLC Visualizer recordings. 5 HPTLC plates in total !
8 Method validation Specificity Spiked matrix 50 ng Spiked matrix 100 ngUnspiked matrix Std level 1 Std level 2 Std level 3 Std level 4
9 There are 2 proposals for assay methods validationMethod validation Accuracy is verified by means of total error and prediction tolerance intervals estimation. A prediction interval is the interval in which the next result(s) is(are) expected to fall. Calculations are based on trueness and precision (total error). Prediction intervals are used to characterize method performance. There are 2 proposals for assay methods validation
10 Method validation The b expectation tolerance interval:Is defined as the prediction interval in which the probability that the next result obtained on the validation sample may fall is ≥b. Requires data obtained in intermediate precision conditions Based on papers by R. Mee (1984) Required in NF V standard (French standard for method validation in foodstuff/feed industry) Proposed for Pharma industry since 2003 (SFSTP) and part of USP <1210> proposal (USP PF40(5)-2014)
11 Method validation The b expectation tolerance interval: with2 sided Student t value, with a=1-b and neff degrees of freedom Penalty (ability to reproduce a result)
12 Method validation The b expectation tolerance intervalAllows a direct estimation of uncertainty of measurement(&): Standard measurement uncertainty: (equivalent to b expectation tolerance interval standard deviation) Expanded measurement uncertainty: With kp =2 or kp= 3 or , with p the desired proportion of the population of measurements (last option is recommended by the Guide to the expression of uncertainty in measurement) (&)NF V standard (2010) J.M. Roussel et al., STP Pharma Pratiques (in press)
13 Method validation ISO 16269-8 prediction interval:ISO standard describes in its chapter 8 a prediction interval defined as the interval in which not less than a desired number of future results may fall, with a probability ≥P. ISO gives k values for the 1, 2, …, 10,…., 100,…, 1000,…, next results. USP <1210> proposal describes the calculation of the k value for the next result (USP PF 42(5) – 2016):
14 Method validation ISO 16269-8 prediction interval:Does not require intermediate precision conditions for the variance, Is simpler to obtain (k values are given in a table), Is more “brutal”, less subtle, than the b expectation tolerance interval, Does not give directly access to the estimation of uncertainty. Therefore, this proposal should be preferred for accuracy estimation in unbalanced validation designs, where intermediate precision cannot be obtained simultaneously with trueness
15 Method validation: resultsComparison of b expectation tolerance intervals (b=80%) obtained with quadratic and Michaelis-Menten calibration functions DP1 – quadratic calibration DP1 – MM calibration
16 Method validation: resultsComparison of b expectation tolerance intervals (b=80%) obtained with quadratic and Michaelis-Menten calibration functions DP7 – quadratic calibration DP7 – MM calibration
17 Method validation: resultsComparison of b expectation tolerance intervals and ISO prediction intervals (calculated with the same variance) DP2 – b exp. tol. interval DP2 – ISO pred. interval
18 Method validation: resultsComparison of b expectation tolerance intervals and ISO prediction intervals (calculated with the same variance) DP7 – b exp. tol. interval DP7 – ISO pred. interval
19 Method validation: resultsUncertainty of measurement estimation Relative expanded uncertainty results Level (ng) Spec. max. (%) DP1 DP2 DP3 DP4 DP5 DP6 DP7 50 30.0 27.0 23.0 22.6 38.1 24.9 100 20.0 11.8 14.0 18.4 24.6 14.6 10.6 300 15.6 13.2 25.4 16.8 28.1 18.1 500 9.6 16.0 21.2 26.7 13.0 DP4 and DP6, particularly, show unacceptable uncertainty levels
20 Method validation: resultsUncertainty of measurement estimation Uncertainty profiles Relative uncertainty Trend line Y=aXb Absolute uncertainty DP2 DP3
21 Conclusions Regarding HPTLC methods performance assessment:Method performance may be checked with not more than 5 HPTLC plates ! Powerful method validation tools, using total error, exist Standards and official proposals already use these tools (NF V03-110, USP <1210> (PF42(5)), ISO 21748…) Accuracy profiles and uncertainty profiles are easy to obtain and both give a clear view on the real performance of analytical methods,
22 Merci de votre attention !Conclusions And, at last: Who may still say that HPTLC methods performance cannot be checked and that HPTLC is not a quantitative method ? Merci de votre attention !