Use of Hoteling's T2 multivariate control chart for effective monitoring of a laboratory test with a 3-level quality control scheme.

Abstract

Introduction: A control chart based on Hotelling's T² multivariate statistics was used to monitor the quality of an immunoenzymatic assay for plasma levetiracetam. The chart incorporated a multi-level quality control (MLQC) system with three concentration levels of the analyte and included the analytical performance specification (APS) for therapeutic drug monitoring.

Materials and methods: Data were collected from March 1 to August 14, 2024, comprising 84 consecutive triplets of values for the three MLQC levels. The initial 59 triplets were used to estimate the variance-covariance matrix and vector of means (phase I). These estimates were then applied to calculate Hotelling's T² for the remaining 25 triplets (phase II). The pharmacokinetic model of Fraser was employed to derive the APS for levetiracetam, based on a twice-daily dosing scheme and a median half-life of 8 hours.

Results: The three MLQC levels showed significant correlations (r > 0.6) in both control phases. The Hotelling's T² control chart detected no out-of-specifications states (OC), compared to 12 OC signals from individual Levey-Jennings charts monitoring the MLQC levels separately. The integration of the APS into the Hotelling's T² chart provided additional insights into the process quality, and in two instances, it aligned with the OC signal from at least one of the Levey-Jennings charts.

Conclusions: Hotelling's T² multivariate chart is effective for internal quality control of laboratory tests. As MLQC data offer correlated information, this approach is advantageous over multiple individual univariate charts as it ensures the correct level of false positive and false negative alarms.