Therapeutic medication monitoring of thiopurine erythrocyte levels was not accessible in all sites, and obtaining the findings generally takes a lengthy time. For a study, researchers developed a model for predicting low levels of 6-thioguanine and 6-methylmercaptopurine in juvenile inflammatory bowel disease (IBD) patients and a model for predicting nonadherence in azathioprine patients (AZA). A total of 332 observations were made on 88 pediatric patients with IBD as part of the research. 6-thioguanine levels of (< 125 pmol/8 × 108) erythrocytes and 6-methylmercaptopurine levels of (< 5700 pmol/8 × 108) erythrocytes were used to determine low AZA dose. Nonadherence was defined as 6-thioguanine and 6-methylmercaptopurine levels of (< 240 pmol/8 × 108) erythrocytes undetected. The data was split into 2 sections: training and testing. The modified random forest approach with cross-validation and resampling was used to build the model predicting low 6-thioguanine levels, nonadherence, and 6-thioguanine level.
When applied to the testing half of the dataset, the resulting models predicting low 6-thioguanine levels and nonadherence had areas under the curves of 0.87 and 0.94, the sensitivity of 0.81 and 0.82, specificity of 0.80 and 86, and distance of 0.31 and 0.21, respectively. When the final model for predicting 6-thioguanine levels was applied to the testing dataset, it produced a root-mean-square error of 110.
Researchers built a model with enough accuracy to predict patients with low 6-thioguanine levels and a model to predict AZA therapy nonadherence using easily available laboratory data.
Reference:journals.lww.com/jpgn/Fulltext/2019/10000/Prediction_of_Thiopurine_Metabolite_Levels_Based.18.aspx
