Evaluation of the performance characteristics of a new chemical sensor, device, or method

• Published in: Sensors and actuators. B, Chemical Chemical, 2024-10, Vol.417, p.136087, Article 136087
• Main Authors: Lindner, Ernő, Pendley, Bradford, Zhang, Hongmei
• Format: Article
• Language: English
• Subjects: Method comparison studies; New graphical representation of agreement between methods; New simple formula for assessing the association between methods; Sources of errors
• ISSN: 0925-4005
• DOI: 10.1016/j.snb.2024.136087

For those of us involved in the design of novel sensors and methods, understanding the analytical requirements based upon their intended use is essential. In the case of a medical sensor, we must know how the results of a test, performed with the novel sensor, will increase the probability of a good diagnostic or disease management decisions by a physician. So, to assess whether the new sensor meets the expectations, one must design and execute experiments and analyze the results by appropriate statistical methods. As a last step of the evaluation of a novel sensor, one can measure the agreement between the results provided by an established method and the new sensor in a method comparison study. However, the correct statistical approach in method comparison studies is not obvious. In this paper a new evaluation method for determining the unbiased parameters of the association between the compared methods is discussed. It is based on the relationship between the slope and intercept of the regression line in the scatter plot and the parameters of the regression line in the Bland and Altman plot of the differences, in which the differences between the values measured by the compared methods in the same samples are plotted as a function of the average of the measured values. Moreover, we suggest a graphical representation of the data for assessing both the association and the agreement between the compared methods and show the limits of the proposed evaluation method theoretically and using computer-generated data and real-world examples. •The difficulties in the statistical analysis of method comparison studies are discussed.•A new evaluation method equivalent to the Deming regression is proposed.•The utility of the new method is validated, and its limits are analyzed theoretically.•A residual plot is used to assess the association and the agreement simultaneously.•The implementation of the new method is demonstrated in the Supporting Information.