Assessment Of Agreement Between Two Quantitative Measurements Under Nontypical Data Behavior

The Bland-Altman method has been widely recognized and applied by the biomedical community for its intuitive,easy analysis and interpretation.analytical and interpretation.However,researchers often ignore their harsh application terms and abuse the standard Bland-Altman method,which makes the correctness of the interchangeability conclusions between two quantitative methods seriously questioned.Although some scholars have already paid attention to the importance of assessment of agreement between two quantitative measurements under nontypical data behavior,they are mostly limited to statistical methodologies.There is still no specific evaluation strategy and practical guidance on how to conduct agreement evaluation in combination with clinical practice.Based on the research on the agreement evaluation methods between two quantitative measurement under typical data behavior,this study focuses on assessment of agreement under the nontypical data behavior,and seeks to build a common strategy for agreement evaluation under the overall framework.Based on the description of the distribution characteristics,bias trend,variance performance and outliers of the two quantitative measurement methods,this study uses data transformation methods to transform nontypical data behavior into typical data behavior.By calculating LoAs and its confidence intervals,and then performing inverse transformation,the estimation of LoAs and its confidence interval bands under the original scale of two quantitative measurement methods is realized.However,traditional data conversion methods,such as ratio/log-transformation and percentage transformation,sometimes have the problem of over-correction or under-correction.In this paper,we use generalized log transformation,logit transformation,and try to propose a new data behavior transformation method,which can better meet the purpose of data conversion.In addition,we propose an index to evaluate the effect of the agreement evaluation method under nontypical data behavior,and use the index for method optimization.Based on the actual requirements of agreement evaluation,we construct a set of general strategies from research design,statistical analysis to result reporting,compile the SAS macro package for the corresponding method implementation and results report,and demonstrate the entire application process of the SAS macro implementation with two clinical measurement examples finally.The new data behavior transformation method proposed in this study which considers both proportional bias and heteroscedasticity,has good versatility and can effectively solve the problem of "under-correction" or "over-correction" of common data transformation methods under nontypical data behavior.This study proposes a agreement evaluation strategy covering research design,statistical analysis and results reporting,and clarifies the considerations and selection paths.The strategy is expected to provide practical guidance and important application value for improving the quality of agreement evaluation studies.The SAS macro program can realize the customization of the analysis content in our study and the standardization of the result output,which helps to improve the efficiency and quality of the statistical analysis process of assessing agreement between the two quantitative measurements.