1. EUSTAS Round-Robin Testing of Steviol glycosides Prof. Jan M.C. Geuns Lab Functional Biology, KULeuven Kasteelpark Arenberg 31, MB 2436 3001 Heverlee-Leuven BELGIUM Jan.Geuns@bio.kuleuven.be

2. Abbreviations SV: steviol, hence: SV gly and SV glu ST: stevioside Reb A – G: rebaudioside A - G Rub: rubusoside DulA: dulcoside A

3. Problems -Different steviol glycosides (10) -Different molecular weights -Difference of a few % is very important! -Validation: each step in analysis should be validated -Very pure standards are required of (all) steviol glycosides -Often an UV detector is used at 210 nm -Often absorption spectra taken in EtOH

4. Problems Extinction coefficients of all compounds are not known Methods used so far were not validated, and are often based on the assumption that extinction coefficients are the same for all compounds If extinction coefficients are the same, it follows that the slopes of calibration curves should be the same (plotted as mM conc), eg. ST (804), RebA (966), Rub (642)

5. Proof of similar slopes

6. Problems Nearly everybody claims that their method is the best one and is accurate.....

7. Problems Nearly everybody claims that their method is the best one and is accurate..... So do we !

8. Extinction coefficients Extinction coefficients measured in water, EtOH, 80% AcCN and 35% AcCN Measured at 31 and 62 µM λmax = 205 (EtOH) 196 (H 2 O) 199 (80% AcCN) 197 (35%AcCN) Molar Extinction coefficient  = A/l.c

9. Extinction coefficients ε-values in H 2 O and AcCN mixtures much larger than in EtOH (± 7500 > ± 4000) ε-values measured at λmax much larger than at λ210 This is also reflected in slopes of calibration curves

10. Calibration Stevioside 190 and 210 nm

11. Problems Often percentual composition is taken as percentage purity! These are not related at all! Proof: measurement of SVgly in presence/absence of 90% glucose!

12. 1 mg/mL Measurement in water

13. 1 mg/mL plus 10 mg glu/mL

14. Measurement in presence of impurities (claim: >95% purity) Sample505171Ip0906Nv2060301 Loss on drying2.3 %2.8%5% Stevioside Reb A Reb C Dulcoside A Other 31.29% 14.67 4.89 1.63 0.65 52.44% 23.93 7.21 2.95 0.98 56.44 28.22 6.51 3.26 Total = Purity53.13%87.51%97.68%

15. Examples of % composition

17. Percentual composition was given Calculation: peak areas without correction for MW! Different MW were not taken into account!

18. Examples of % composition

19. Examples of % composition; Claim: >96% RebA

20. Same sample analysed

21. Same sample, claim >96% purity

22. Impurities in samples (C18 column) Always inject blanks (below)

23. Round-robin testing 10 labs Each lab: use of own standards and methods 2 unknown samples -sample 1: 96.2% purity -sample 2: 80:20 dilution of sample 1 with NaHCO 3

24. AIM of Round-robin testing - check as many parameters as possible - drying process (NaHCO 3 ) - peak integration - calculations - weighing........ - try to explain the differences in analyses - how to improve analysis?

25. Round-robin testing Most labs: NH 2 columns, a few C18, 1 SILIC; 250 x 4.6 mm; 5µm; UV 205 or 210 nm Solvents: - AcCN:H 2 O or AcCN:diluted NH 4 OAc or phosphoric acid (isocratic or gradient) - C18: 2 x Alltech Alltima with gradient AcCN:0.1 mM phosphoric acid

26. Round-robin testing In theory, both HPLC methods should be equivalent and give similar results after proper calibration. Use of only 1 early eluting standard for calibration combined with an isocratic solvent, might give an underestimation of slowly eluting compounds (due to flattening of peaks, partly disappearing in base line).

27. Sample 1, gradient: 25 min

28. Sample 1, isocratic 33%; 60 min

29. Difference gradient vs. isocratic

30. Later peaks are systematically smaller than early eluting peaks Combination of “disappearance in baseline” and greater signal due to longer time in detector, as demonstrated by a chromatogram at 1 mL/min and 0.8 mL/min

31. Flows of 1 and 0.8 mL/min

32. Conclusion flow rate Variation in flow rate (even a few %) might produce huge errors in quantification!

33. Round-robin testing: Sample 1

34. Sample 1 Only 3 labs analysed 8 or 9 compounds Only 2 labs found purity > 95% After correction for compounds not analysed: only 3 labs purity >95%. However, lab 3 excessive values for RebC and DulA

35. Sample 2: Weight loss 20 % NaHCO 3 100 mg: 80 mg sample 1 + 20 mg NaHCO 3 Heating at 105°C, several hours: 2 NaHCO 3 Na 2 CO 3 + CO 2 + H 2 O loss of 36.9 %! 100 mg sample 2: contains 80 mg “wet” sample 1. Wt loss is 4.8 mg (6 %) 20 mg NaHCO 3 : loss of 7.38 mg Total: 4.8 + 7.38 = 12.18 mg or 12.18 %

36. Sample 2: Weight loss Reported weight loss: between 4.9 and 12.7% Insufficient drying! Drying should be done to constant weight. Weak point in JECFA recommendation (2 h at 105°C, wt loss< 6%). Moisture calibration of standard and sample: possible eg. RebA. What about all other compounds: hygroscopic properties???

37. Purity of sample 2 Amount of SV gly in sample 2: -purity of sample 1= 96.2% or 90.4% before correction for weight loss (= 6 %) -100 mg sample 2 contains 80 % of sample 1 or 90.4 x 80 % = 72.32 mg. -Theoretical purity of sample 2: 72.32/0.8782 = 82.35 mg or %. This is 85.6% of purity of sample 1.

38. Purity of sample 2

39. Only 3 labs analysed 8 or 9 compounds Purity should be 85.6% of that of sample 1. Only 2 labs found about this value (85 and 87.3) and a wt loss of 12.1 and 12.4 %. Even after correction for compounds not analysed, purity reported by several labs was far below 82.35 %

40. Accuracy of measurement Only way to know what is the exact measurement Standard addition method: at least 3 known amounts are added of ultra-pure and dry standards Extrapolation gives intercept with Y-axis. This value should be the same as that of an unspiked sample.

41. Standard addition of Sample 1

42. Intercept gave a value of 1277.5 for RebA unspiked sample: value of 1309 Difference: 2.5 %, acceptable as expt. was done only once

43. Possible errors in (G)LP Purity of standards: – absence of other SV gly – drying to constant weight – always validate quality of standards!!!

44. Purity of standards: ST: claim: 98.6% purity!

45. Purity of standards: % composition: ST only 94.97

46. Purity of standards: on dry wt. basis only 93.54 % purity (wt loss: 3.1%)

47. Possible errors (in %) Calibration with impure standards leads to an overestimation of purity of samples! ST: claimed purity 98.6%, real purity: 93.54% ClaimedFound (wet) Found (dry wt.) “100 %” vs. wet corr 98.690.493.54 (100 %)Δ 9 %Δ 5.4 %Δ 10.6 %

48. Conclusions HPLC method, NH 2 vs. C 18 less important than GLP! Purity of standards: utmost importance –No other compounds present –Moisture content –To be validated by each lab Drying of standards and samples to constant weight Use of gradients might avoid underestimation of small peaks eluting at the end in isocratic HPLC

49. Conclusions Weighing process itself - Weigh enough sample/standard (at least 50 mg) on a sensitive balance - Weighing of solution (eg. mg/g) might decrease errors - if using automatic pipettes: do not trust them and calibrate regularly! and again and again.... check quality of tips

50. Conclusions Dissolution of sample! Check that sample is well dissolved Inject enough. Large peaks should be > 500 mV. This reduces relative standard errors of smaller peaks. Check base-line in peak integrations!

51. Conclusions Check accuracy by standard addition method! More round-robin tests are required to improve and unify the analysis of steviol glycosides. Cooperation between different companies would be good for the “Stevia Industry” as a whole.

52. Acknowledgements Hilde Verlinden, Bert Demarsin for their excellent help Peter Grosser, Medherbs, Wiesbaden, Germany for financial support My wife Christine for being a “Stevia Widow”