Artificial neural networks for assessing the risk factors for urinary calcium stones according to gender and family history of stone

Objective. In a previous work, we evidenced that artificial neural networks (ANNs) were more informative than classical statistical analyses for assessing the risk of idiopathic calcium nephrolithiasis in male stone-formers. Material and methods. We compared risk factors for idiopathic calcium nephrolithiasis (age, body mass index, calcemia, calcium oxalate supersaturation, and 24-h calciuria, oxaluria, uricosuria, citraturia, urea, and sodium) in four populations: men and women with and without a family history of stone (FHS). A total of 119 males (58 with an FHS, 61 without) and 59 females (30 with an FHS, 29 without) were compared to healthy controls. For each variable, receiver operating characteristic (ROC) curve indices were calculated by means of ANNs. Results. In men without an FHS, the most discriminant variables were 24-h urea (ROC curve index 0.76), supersaturation (ROC curve index 0.72), 24-h calciuria (ROC curve index 0.68), 24-h uricosuria (ROC curve index 0.64), 24-h oxaluria (ROC curve index 0.63), 24-h sodium (ROC curve index 0.62), and calcemia (ROC curve index 0.60). In men with an FHS, only supersaturation (ROC curve index 0.67) was discriminant. In women without FHS, calcemia (ROC curve index 0.67), 24-h calciuria (ROC curve index 0.64), and 24-h uricosuria (ROC curve index 0.62) were discriminant. In women with an FHS, supersaturation (ROC curve index 0.70), 24-h uricosuria (ROC curve index 0.69), 24-h urea (ROC curve index 0.68), and 24-h calciuria (ROC curve index 0.67) were discriminant. Conclusions. Risk factors for idiopathic calcium nephrolithiasis were roughly the same in men with or without an FHS, and in women with an FHS. In these patients, calcium oxalate supersaturation and 24-h urea were the most discriminant factors. Conversely, in women without an FHS, calcium abnormalities (calcemia, 24-h calciuria) were discriminant and should prompt a search for infraclinical primary hyperparathyroidism or sarcoidosis.