Improved conversion of creatine kinase MB (CK-MB) activity into mass data by generalized additive modeling

Creatine kinase isoenzyme CK-MB is used to monitor myocardial damage. We investigated the feasibility of an interconversion of CK-MB mass and activity data to simplify multicenter trials. We modeled CK-MB data with ordinary least squares regression and with generalized additive models (GAMs) which permit the inclusion of covariates. An optimized GAM predicted CK-MB masses from activities, sex, and sampling time with an Rsquare of 0.981. Interconversion of CK-MB masses and activities by GAMs created from representative patient cohorts may help to include study centers with incompatible data into multicenter trials.

肌酸激酶同工酶CK-MB（Creatine kinase isoenzyme CK-MB）常用于监测心肌损伤。本研究探讨了将CK-MB质量与活性数据进行相互转换的可行性，以简化多中心试验的开展流程。我们分别采用普通最小二乘回归与允许纳入协变量的广义相加模型（generalized additive models, GAMs）对CK-MB数据进行建模。经优化的广义相加模型可基于活性值、受试者性别及采样时间预测CK-MB质量，其决定系数（Rsquare）达0.981。通过由代表性患者队列构建的广义相加模型实现CK-MB质量与活性数据的相互转换，可助力将数据采集标准不统一的研究中心纳入多中心试验范畴。