Clinical and experimental nephrology · 2025
Machine learning using serial changes in proteinuria during initial steroid therapy to predict treatment response and immunosuppressant use in pediatric idiopathic nephrotic syndrome.
СтатьяМногоцентровое исследованиеНазвание и аннотация: Перевод в очереди
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Iida Takaya, Miura Kenichiro, Okamoto Takayuki, Fujinaga Shuichiro, Akioka Yuko, Takeshima Yasuhiro, Urushihara Maki, Hisano Masataka, Gotoh Yoshimitsu, Ohta Toshiyuki, Takaya Eichi, Miyauchi Carlos Makoto, Sonobe Shinya, Kaname Tadashi, Hattori Motoshi
Аннотация
BACKGROUND: Epidemiological studies on idiopathic nephrotic syndrome (INS) in children have identified no definitive factors predicting steroid-resistant nephrotic syndrome (SRNS) or frequent relapsing nephrotic syndrome. Research using machine learning (ML) has been conducted to predict INS prognosis; however, no studies have evaluated serial changes in proteinuria during initial steroid therapy.
METHODS: INS patient data were collected from 23 medical centers. ML using clinical and laboratory data at first presentation and time-series features generated using serial changes in urine protein to creatinine ratio (UPCR) during initial steroid therapy were performed to predict SRNS and immunosuppressant use in 329 and 190 patients, respectively. ML models were run to calculate the area under the curve (AUC) and to identify variables contributing to predicted outcomes using the backward stepwise method.
RESULTS: In the SRNS prediction model, UPCR at the final analysis point (i.e., the last sequential day of UPCR input included for model analysis) and several preceding days substantially contributed to the prediction, with UPCR at the final analysis point being the most significant contributor. The immunosuppressant prediction model achieved an AUC ranging from 0.715 to 0.759 and showed that age, serum albumin, serum total cholesterol, and time-series features (approximate entropy, mean UPCR value between the 20th to 80th percentiles, and 70th percentile UPCR value) were significant contributors.
CONCLUSIONS: Our ML suggested that UPCR at the final analysis point was an important predictor of SRNS. Age, serum albumin, serum total cholesterol and serial changes in proteinuria contributed to immunosuppressant use.
METHODS: INS patient data were collected from 23 medical centers. ML using clinical and laboratory data at first presentation and time-series features generated using serial changes in urine protein to creatinine ratio (UPCR) during initial steroid therapy were performed to predict SRNS and immunosuppressant use in 329 and 190 patients, respectively. ML models were run to calculate the area under the curve (AUC) and to identify variables contributing to predicted outcomes using the backward stepwise method.
RESULTS: In the SRNS prediction model, UPCR at the final analysis point (i.e., the last sequential day of UPCR input included for model analysis) and several preceding days substantially contributed to the prediction, with UPCR at the final analysis point being the most significant contributor. The immunosuppressant prediction model achieved an AUC ranging from 0.715 to 0.759 and showed that age, serum albumin, serum total cholesterol, and time-series features (approximate entropy, mean UPCR value between the 20th to 80th percentiles, and 70th percentile UPCR value) were significant contributors.
CONCLUSIONS: Our ML suggested that UPCR at the final analysis point was an important predictor of SRNS. Age, serum albumin, serum total cholesterol and serial changes in proteinuria contributed to immunosuppressant use.