AOAC SPIFAN Review Team Meeting Book-July 22, 2015

BVIT-02 FOR ERP/WORKING GROUP USE ONLY DO NOT DISTRIBUTE

Abbott Nutrition Division Simultaneous Determination of Total Vitamin B 6, B 2, B 3 in Infant Formula Products by LC-MS/MS Using Enzymatic Digestion and B 1

5. UPLC ANALYSIS a) Upon verifying equilibration of the UPLC system, inject the working standards (WS1 to WS6) followed by a mobile phase blank, control sample and sample extracts. b) Sample extracts are injected in single. c) Samples should be bracketed between two sets of working standards. Vitamin content in product samples is then determined by comparison of the response ratio for the sample to the calibration curve. d) For larger sample sets, working standards should be injected every 6 hours or approximately every 50 sample extracts. e) Calibration curves must have a correlation coefficient, r 2 , of >0.990 as defined in Section H (Quality Control). f) Calibration curve residuals (relative error) must be < 15% for WS1 and <10% for WS2-WS6 as defined in Section H (Quality Control). g) The Method utilizes calibration curves for each nutrient that cover approximately 2 orders of magnitude in analyte concentration. h) The calibration curve for a compound may be shortened to as few as 4 points, i.e. analyte concentrations. This can occur due to an error in working standard preparation or compromised instrument performance especially for lower standards (WS1, WS2 or WS3). Care must be taken when doing so to ensure that the response factor for each sample extract in the sample set is within the response ratio range of the remaining working standards. For example, if WS1 for Pyridoxal is outside the acceptance criteria, the WS1 level (all injections) may be dropped for Pyridoxal and the samples calculated using a calibration curve from WS2 to WS6. i) If individual standards are to be dropped due to atypical instrument response or injection error, etc, care must be taken to not exclude too many working standard calibration points. As a general rule, no more than 25% of the total number of working standard data points should be excluded from a calibration curve while ensuring that the response factor for each product in the sample set is within the response factor range of the remaining working.

30