Application of clinical indices in pathological acute and chronic index prediction in lupus nephritis

To predict renal biopsy with pathological acute and chronic index via comprehensive analysis of clinical indices from lupus nephritis patients.

We collected 107 inpatients' data from the Third Xiangya Hospital of Central South University from January 2011 to September 2018. These inpatients have already had renal biopsy results. These clinical indices were set as variables. Using multiple linear regression, we built the prediction models for renal pathological acute and chronic index. Simultaneously, we evaluated each vital variable's importance in models by standardized coefficient of regression equation, and model prediction accuracy by 5-fold cross validation.

Acute index and chronic index prediction models were built with 19 and 23 clinical variables, respectively. To evaluate the two models' accuracy of prediction, we used 5-fold cross validation and found that the prediction accuracy level were satisfactory with the acute index model (Q²=0.649, R²=0.791) and chronic index model (Q²=0.563, R²=0.744). Standardized coefficient of regression equation showed serum total cholesterol, serum iron, NAG, β2 microglobulin and BUN were important variables to predict acute index while total cholesterol, β2 microglobulin, homocystein and serum low density lipoprotein-cholesterol were important for chronic index prediction.

Clinical indices are able to effectively predict renal biopsy conditions, which are valuable to assess lupus nephritis severity.