Sugai Keiji, Sasaki Junko, Wada Yuki, Shimizu Norihiro, Ishikawa Takuya, Yanagi Ketchu, Hashimoto Takeshi, Tanaka Akihiko, Suwanai Hirotsugu, Suzuki Ryo, Odawara Masato
Department of General Internal Medicine, Toda Chuo General Hospital, 1-19-3 Honcho, Toda, Saitama 335-0023 Japan.
Department of Diabetes, Metabolism, and Endocrinology, Tokyo Medical University Hospital, 6-7-1 Nishishinjuku, Shinjuku-Ku, Tokyo, 160-0023 Japan.
Diabetol Int. 2024 Jan 23;15(2):253-261. doi: 10.1007/s13340-023-00687-1. eCollection 2024 Apr.
Numerous studies demonstrated the risk factors for urological complications in patients with diabetes before sodium-glucose co-transporter 2 inhibitor (SGLT2i) became commercially available. This study aimed to comprehensively investigate urological characteristics in patients with type 2 diabetes (T2DM) after SGLT2i became commercially available.
We examined 63 outpatients with T2DM suspected of bacteriuria based on urinary sediment examinations. Urine cultures were performed, and lower urinary tract symptoms (LUTS) were assessed via questionnaires. Patients with bacteriuria were assessed using ultrasonography to measure post-void residual volume (PVR). Utilizing demographic and laboratory data, a random forest algorithm predicted LUTS, bacteriuria, and symptomatic bacteriuria (SB).
Thirty-two patients had LUTS and 31 had bacteriuria. High-density lipoprotein cholesterol level was crucial in predicting LUTS, while age was crucial in predicting bacteriuria. In predicting SB among patients with bacteriuria, creatinine level and estimated glomerular filtration rate were crucial. Our models had high predictive accuracy for LUTS (area under the curve [AUC] = 0.846), followed by bacteriuria (AUC = 0.770) and SB (AUC = 0.938) in receiver operating characteristic curve analysis. These predictors were previously reported as risk factors for urological complications. Although SGLT2i use was not an important predictor in our study, all SGLT2i users with bacteriuria had SB and exhibited higher PVR compared to non-SGLT2i users with bacteriuria.
This study's random forest model highlighted distinct essential predictors for each urological condition. The predictors were consistent before and after SGLT2i became commercially available.
The online version contains supplementary material available at 10.1007/s13340-023-00687-1.
在钠-葡萄糖协同转运蛋白2抑制剂（SGLT2i）上市之前，大量研究已证实糖尿病患者发生泌尿系统并发症的风险因素。本研究旨在全面调查SGLT2i上市后2型糖尿病（T2DM）患者的泌尿系统特征。
我们检查了63例因尿沉渣检查怀疑有菌尿的T2DM门诊患者。进行了尿培养，并通过问卷调查评估下尿路症状（LUTS）。对菌尿患者使用超声测量排尿后残余尿量（PVR）。利用人口统计学和实验室数据，采用随机森林算法预测LUTS、菌尿和有症状菌尿（SB）。
32例患者有LUTS，31例有菌尿。高密度脂蛋白胆固醇水平对预测LUTS至关重要，而年龄对预测菌尿至关重要。在预测菌尿患者的SB时，肌酐水平和估计肾小球滤过率至关重要。在受试者工作特征曲线分析中，我们的模型对LUTS具有较高的预测准确性（曲线下面积[AUC]=0.846），其次是菌尿（AUC=0.770）和SB（AUC=0.938）。这些预测指标此前已被报道为泌尿系统并发症的风险因素。虽然在我们的研究中使用SGLT2i不是一个重要的预测指标，但所有有菌尿的SGLT2i使用者都有SB，并且与有菌尿的非SGLT2i使用者相比，其PVR更高。
本研究的随机森林模型突出了每种泌尿系统疾病不同的重要预测指标。这些预测指标在SGLT2i上市前后是一致的。
在线版本包含可在10.100 / s13340 - 023 - 00687 - 1获取的补充材料。
