Background: TI is an inhaled mealtime insulin with ultra-rapid onset and short duration of action closely mimicking physiologic insulin. Clinical studies demonstrated a higher initial dose could result in faster titration to overcome a perceived lack of effect. The aim was to develop an in silico model to augment results of a study (NCT04849845) evaluating an increased initial conversion of TI from injected insulin vs. the current label in 20 adults with diabetes that showed significantly improved post-prandial control after 45 minutes. Methods: The in silico model characterizes blood glucose trends and outcomes utilizing published equations and distributions derived from real-world data sources. Inputs with the largest effects on glucose outcomes were combined to match study parameters to generate one million meal boluses. The model was validated using common modeling and simulation recommendations. Results: Model inputs were constrained to match clinical study parameters; in silico results substantiated and complemented in vivo safety and efficacy results with similar predictive values and acceptable hypoglycemia distributions. Conclusions: Clinical data augmented by in silico data support that an increased initial dose is not associated with any new safety concerns. Leveraging in silico data generates high quality evidence faster and with less risk to patients. Disclosure L.Desborough: Consultant; Luna, MannKind Corporation, Roche Diabetes Care, Securecell AG, Tandem Diabetes Care, Inc., Employee; Nudge BG, Inc. K.B.Kaiserman: Advisory Panel; Medtronic, Consultant; Medtronic, Employee; MannKind Corporation, Research Support; Medtronic, Speaker's Bureau; Medtronic, Stock/Shareholder; MannKind Corporation. J.Hanna: Employee; MannKind Corporation, Medtronic. K.Jaffe: None. J.Ulloa: Employee; MannKind Corporation, Medtronic, Stock/Shareholder; MannKind Corporation, Medtronic.

中文翻译：

---

1496-PUB：使用计算机证据增强临床试验数据以评估 Technosphere 胰岛素 (TI) 的初始转换剂量变化

背景：TI 是一种吸入餐时胰岛素，起效超快，作用持续时间短，与生理胰岛素非常相似。临床研究表明，较高的初始剂量可能会导致更快的滴定，以克服感知到的效果不足。目的是开发一种计算机模型，以增强一项研究 (NCT04849845) 的结果，该研究评估了 20 名糖尿病成人的 TI 从注射胰岛素与当前标签的初始转化增加，该标签显示 45 分钟后餐后控制显着改善。方法：计算机模型利用来自真实世界数据源的已发表方程和分布来表征血糖趋势和结果。结合对葡萄糖结果影响最大的输入以匹配研究参数以产生一百万次膳食推注。使用通用建模和仿真建议对该模型进行了验证。结果：模型输入被限制以匹配临床研究参数；计算机模拟结果证实并补充了体内安全性和有效性结果，具有相似的预测值和可接受的低血糖分布。结论：通过计算机数据增强的临床数据支持增加的初始剂量与任何新的安全问题无关。利用计算机数据可以更快地生成高质量的证据，并且对患者的风险更低。披露 L.Desborough：顾问；Luna, MannKind Corporation, Roche Diabetes Care, Securecell AG, Tandem Diabetes Care, Inc.，员工；Nudge BG, Inc. KBKaiserman：咨询小组；美敦力，顾问；美敦力，员工；MannKind 公司，研究支持；美敦力，发言人局；美敦力，股票/股东；曼金公司。J.Hanna：员工；MannKind 公司，美敦力。K.Jaffe：没有。J.Ulloa：员工；MannKind Corporation，美敦力，股票/股东；MannKind 公司，美敦力。