近日，美国约翰斯·霍普金斯大学医学院Nestoras Mathioudakis团队比较了人工智能驱动的生活方式干预与糖尿病预防计划中的人类指导的效果。这一研究成果发表在2025年10月27日出版的《美国医学会杂志》上。

前驱糖尿病很常见，但循证生活方式干预未得到充分利用。

为了确定在糖尿病前期和超重或肥胖的成年人中，是否转诊到基于糖尿病预防计划（DPP）的人工智能（AI）主导的生活方式干预，在达到体重减轻、血红蛋白A1c （HbA1c）降低和每周身体活动的推荐阈值方面，不逊色于转诊到人类主导的DPP，研究组进行了一项3期、平行组、实用、非劣效性随机临床试验，于2021年10月11日至2024年12月16日（最后一次随访）在美国马里兰州巴尔的摩和宾夕法尼亚州雷丁的两个临床地点进行。研究对象为18岁或以上患有前驱糖尿病、超重或肥胖的成年人。

将参与者以1:1的比例随机分配，接受通过移动应用程序和蓝牙数字秤提供的人工智能驱动的DPP生活方式干预的转诊，或接受远程提供的人工教练指导的DPP生活方式干预的转诊。在12个月的时间里，两种干预措施都是独立于研究组进行的。主要结局是在整个研究过程中维持HbA1c低于6.5%，实现至少5%的体重减轻，至少4%的体重减轻加上每周至少150分钟的身体活动（用活动描记术评估），或12个月时HbA1c绝对降低至少0.2个百分点。人工智能引导的DPP与人工引导的DPP的非劣效性是预先设定的，以确定风险差异的单侧95% CI下限不超过15%。

研究组共纳入368名参与者（年龄中位数为58岁[50-65]岁；71%为女性，27%为黑人，6%为西班牙裔，61%为白人；BMI中位数[IQR]为32.3[28.5-37.1]）。人工智能主导的DPP组的183名参与者中有58名（31.7%）达到了主要结局，人类主导的DPP组的185名参与者中有59名（31.9%）达到了主要结局（风险差异为0.2%），符合非劣效性标准。该结果在复合终点的各个组成部分和敏感性分析中是一致的。

研究结果表明，在患有前驱糖尿病和超重或肥胖的成年人中，在实现基于体重减轻、身体活动和HbA1c的复合结局方面，转介到全自动人工智能主导的DPP并不亚于转介到人类主导的DPP。

附：英文原文

Title: An AI-Powered Lifestyle Intervention vs Human Coaching in the Diabetes Prevention Program: A Randomized Clinical Trial

Author: Nestoras Mathioudakis, Benjamin Lalani, Mohammed S. Abusamaan, Mary Alderfer, Defne Alver, Adrian Dobs, Brian Kane, John McGready, Kristin Riekert, Benjamin Ringham, Aliyah Shehadeh, Fatmata Vandi, Amal A. Wanigatunga, Daniel Zade, Nisa M. Maruthur, AI-DPP Study Group, Nestoras N. Mathioudakis, Benjamin Lalani, Mohammed S. Abusamaan, Adrian Dobs, Jeromie Ballreich, Nisa Maruthur, John McGready, Kristin Riekert, Amal Wanigatunga, Brian Kane, Aliyah Shehadeh, Defne Alver, Benjamin Ringham, Fatmata Vandi, Daniel Zade, Pamela Marrero, Amy Schiller, Mary (Ellie) Alderfer, Nae-Yuh Wang, Seth Martin, Sara Benjamin-Neelon

Issue&Volume: 2025-10-27

Abstract:

Importance  Prediabetes is common, yet evidence-based lifestyle interventions are underutilized.

Objective  To determine whether referral to an exclusively artificial intelligence (AI)–led lifestyle intervention based on the Diabetes Prevention Program (DPP) is noninferior to referral to a human-led DPP in achieving recommended thresholds for weight loss, hemoglobin A1c (HbA1c) reduction, and weekly physical activity among adults with prediabetes and overweight or obesity.

Design, Setting, and Participants  This phase 3, parallel-group, pragmatic, noninferiority randomized clinical trial was conducted from October 11, 2021, to December 16, 2024 (last follow-up) at 2 US clinical sites in Baltimore, Maryland, and Reading, Pennsylvania. Adults 18 years or older with prediabetes and overweight or obesity were enrolled.

Interventions  Participants were randomized in a 1:1 ratio to receive either a referral to an AI-powered DPP lifestyle intervention delivered via a mobile app and Bluetooth-enabled digital scale or a referral to a human coach–led DPP lifestyle intervention delivered remotely. Both interventions were delivered independently of the study team over a 12-month period.

Main Outcomes and Measures  The primary outcome was a composite of maintaining an HbA1c less than 6.5% throughout the study and achievement of at least 5% weight loss, at least 4% weight loss plus at least 150 minutes of weekly physical activity (assessed with actigraphy), or an absolute reduction in HbA1c of at least 0.2 percentage points at 12 months. Noninferiority of referral to the AI-led DPP compared with referral to the human-led DPP was prespecified to be determined if the 1-sided 95% CI lower boundary of the risk difference did not cross 15%.

Results  A total of 368 participants were included (median [IQR] age, 58 [50-65] years; 71% were female, 27% were Black, 6% were Hispanic, and 61% were White; median [IQR] BMI, 32.3 [28.5-37.1]). After referral, 171 of 183 participants (93.4%) initiated the AI-led DPP and 153 of 185 (82.7%) initiated the human-led DPP. The primary outcome was achieved by 58 of 183 participants (31.7%) in the AI-led DPP group and 59 of 185 (31.9%) in the human-led DPP group (risk difference, 0.2% [1-sided 95% CI, 8.2%]), meeting the criterion for noninferiority. Findings were consistent across individual components of the composite end point and in sensitivity analyses.

Conclusions and Relevance  Among adults with prediabetes and overweight or obesity, referral to a fully automated AI-led DPP was noninferior to referral to a human-led DPP in achieving a composite outcome based on weight reduction, physical activity, and HbA1c.

DOI: 10.1001/jama.2025.19563

Source: https://jamanetwork.com/journals/jama/fullarticle/2840703