Using Bland-Altman plot-based harmonization algorithm to optimize the harmonization for immunoassays

Huiling Fang; Ruifeng Yang; Jiayue Guo; Xinxin Ren; Xin Chang; Lan Kang; Yuqing Zhu

doi:10.1515/cclm-2024-0187

Using Bland-Altman plot-based harmonization algorithm to optimize the harmonization for immunoassays

Clin Chem Lab Med. 2024 May 15. doi: 10.1515/cclm-2024-0187. Online ahead of print.

Authors

Huiling Fang¹, Ruifeng Yang^{2

3}, Jiayue Guo¹, Xinxin Ren⁴, Xin Chang⁴, Lan Kang⁴, Yuqing Zhu¹

Affiliations

¹ Department of Clinical Chemistry and Immunology, Shanghai Center for Clinical Laboratory, Shanghai, P.R. China.
² Department of Clinical Laboratory, Peking University Shougang Hospital, Beijing, P.R. China.
³ Peking University Hepatology Institute, Peking University People's Hospital, Beijing, P.R. China.
⁴ Standardization & Performance Evaluation Laboratory, Chemclin Diagnostics Co., LTD, Beijing, P.R. China.

PMID: 38742665
DOI: 10.1515/cclm-2024-0187

Abstract

Objectives: Harmonization has been recommended by the International Organization for Standard (ISO) to achieve equivalent results across in vitro diagnostic measurement devices (IVD-MDs). We aim to evaluate the effectiveness of Bland-Altman plot-based harmonization algorithm (BA-BHA) created in this study and compare it with weighted Deming regression-based harmonization algorithm (WD-BHA) proposed in ISO 21151:2020.

Methods: Eighty patient sera were used as the harmonization reference material (HRM) to develop IVD-MD-specific harmonization algorithms. Another panel of 40 patient sera was used to validate the effectiveness of harmonization algorithms. We compared regression slopes, intercepts, Bland-Altman plot layouts, percent differences, limits of agreement (LoAs), between-method coefficients of variation (CV) before and after harmonization.

Results: After harmonization by WD-BHA, acceptable slopes and intercepts between measured values and HRM targets were observed in weighted Deming regression, but not in Passing-Bablok analysis. Mean differences were -5.5 to 5.0 % and differences at specific levels were -33.9 to 23.9 %. LoAs were -64.6 to 74.6 %. Between-method CV was 22.9 % (±12.9 %). However, after harmonization by BA-BHA, both weighted Deming and Passing-Bablok regressions equations presented harmonized results. Mean differences were -0.3 to 0.2 % and differences at specific levels were -1.1 to 1.6 %. LoAs were -23.3 to 23.2 %. Between-method CV was 8.4 % (±4.0 %). The data points were evenly distributed at both sides of the mean in Bland-Altman plots.

Conclusions: The inequivalence of test results between different methods can be improved but unacceptable analytical differences at specific levels may be hidden in terms of an acceptable slope and intercept on WD-BHA. The new protocol BA-BHA may be a viable alternative to optimize the harmonization for immunoassays.

Keywords: Bland-Altman plot; harmonization; harmonization reference material; weighted Deming regression.