Evaluation of Precision and Bias Specifications Required to Achieve the 2018 FDA Guidance Criteria for Glucose Meter Performance Using Simulation Models

摘要

Background: The objective of this study was to estimate the combinations of total bias and total imprecision required for devices to meet the Food and Drug Administration (FDA) specifications using Monte Carlo simulation rather than collection and analysis of experimental data. Methods: A model Gaussian distribution of true-glucose values was altered by adding bias and imprecision to create measured-glucose values affected by analytic error. The fraction of measured-glucose values that met the 2018 FDA criteria for blood glucose monitoring system (BGMS) or self-monitoring blood glucose (SMBG) devices was determined as a function of bias and imprecision. Results: The BGMS model determined that a maximum total imprecision of 6% was required with no bias, and with a total bias of +10?mg/dL the total imprecision allowed was reduced to 5% to achieve the 95% FDA performance expectation: 95% of results ≥75?mg/dL within ±12% and 95% of results <75?mg/dL within ±12?mg/dL. The SMBG model determined that a maximum total imprecision of 6% was required at no bias, and with a total bias of +10?mg/dL the total imprecision allowed was reduced to 4% to achieve the 98% FDA expectation: 98% of results ±75?mg/dL within ±15% and 98% of results <75?mg/dL within ±15?mg/dL. Conclusions: The 2018 FDA guidance criteria require strict conditions for glucose meter clinical trials to achieve <10?mg/dL total bias and total imprecision of <5%. Total imprecision and bias values assessed in models in this study represent the cumulative imprecision and bias errors for the glucose meters, the reference method, and preanalytic processes.