Applying uncertainty analysis to assess the variation of operator performance when manually gating flow cytometry data [Poster presentation]

The use of flow cytometry (FC) for cell identification and counting is prevalent within the biomanufacture of Advanced Therapy Medicinal Products (ATMPs) both in clinical centres and biomanufacturing facilities. Operators involved in the use of FC and the analysis of the resultant datasets have the option of manually gating cell clusters or using a range of automated software platforms (supplied by FC instrument manufacturers, or third party). It is well known that many operators prefer to manually gate on the basis of speed and simplicity, however with increasing data complexity (especially when dealing with multi-colour FC instruments) there is significant potential for variation in final cell count to increase as a function of operator subjectivity. Understanding this relationship between increasing data dimensionality and operator subjectivity during manual gating techniques is important because it allows process facilities to comprehend and quantify the potential for process error as a function of incorrect measurement data. This research has specifically considered how uncertainty analysis can be used to better define operator variation over and beyond traditional measures such as the Coefficient of Variation (CV), where CV is typically only representative of the final gating process and does not provide resolution of understanding at each gating step [1, 2]. Three Stages of cell model complexity were designed and trialled, measuring variation in manual gating sequence steps to obtain a target cell population, with operators selected from academic and industrial research, and national measurement laboratory process team environments. Increasing operator variation was found as a function of increasing FC data complexity. When considering 3 step, 5 step and 8 step FC analysis, operator CV was found to range from 6% to 57% (3 step to 8 step data respectively) [3]. At the same time uncertainty analysis was applied to each gating analysis step providing clear definition of variation at each step allowing the generation of inter- and intra-operator variation statistics. Resultant combined uncertainty statements expressed at a one sigma coverage factor (67%) showed variation from 12% (3 step data) through to 34% (8 step data). Process improvement for manually gated FC analysis can therefore take advantage of this more detailed variation analysis framework and provides a benchmark for comparison with FC automated software platforms. In addition, this investigation will be pertinent when defining the human factors requirement for manual gating and evaluation of cell metrics, both for biomanufacturing and clinical scenarios, as well as helping to inform regulatory science.