Prediction of acute kidney injury during the prerace stage of a 48-hour ultramarathon

Acute kidney injury (AKI) is commonly seen in ultrarunners, and we hypothesized that an AKI prediction model for a 48-hour ultramarathon runner could be constructed with the runner's prerace blood, urine, and body composition data. Fifteen male and three female ultrarunners were recruited from a 48-hour Ultramarathon Festival. AKI prediction models were built based on the support vector machine algorithm. The models’ performance was evaluated by the accuracy of cross-validation tests. Moreover, we used the Friedman test to determine physiological changes from prerace to post-race in blood, urine, and body composition data. The best AKI prediction model reached an accuracy of 85% with the sensitivity and specificity being 78% and 93%, respectively. The major components of the best model were potassium, triglyceride, troponin, cholesterol, low-density lipoproteins, and creatine kinase MB of the blood; blood urea nitrogen of the urine; and muscle and creatinine clearance rate of the body composition. Furthermore, the biochemical and physiological responses of ultrarunners showed consistencies with related studies in traditional marathons and ultramarathons. In conclusion, a promising AKI prediction model was proposed, and ultrarunners are suggested to maintain healthy kidneys, heart, muscle mass, and decrease fat mass to reduce the risk of acquiring AKI.