Stene Lars C
Department of Chronic Diseases, Norwegian Institute of Public Health, Oslo, Norway.
Oslo Diabetes Research Centre, Norwegian Institute of Public Health, Oslo, Norway.
Diabetologia. 2025 Oct;68(10):2104-2115. doi: 10.1007/s00125-025-06512-5. Epub 2025 Aug 6.
Preventing type 1 diabetes remains a significant challenge but ongoing efforts are bringing us closer to this goal. This article discusses some implications of heterogeneity and chance in relation to prevention of type 1 diabetes, particularly regarding interpretation of evidence and planning of future trials. Using simulations, I illustrate uncertainties in efficacy estimates in prevention trials with time-to-event endpoints, using the TN10 teplizumab trial as an example. I emphasise that risk heterogeneity does not equate to treatment effect heterogeneity. When factors modifying efficacy are taken into account, robust identification of treatment effect heterogeneity may require sample sizes approximately four times larger than those needed to determine overall efficacy in certain scenarios. Efficiency of prevention trials can be increased using 2 × 2 factorial designs investigating two treatment options. I also simulate statistical power in exploratory studies involving multiple testing with different strategies for handling potential type 1 diabetes endotypes. If endotypes are defined as subtypes of type 1 diabetes-related phenotypes with at least partially unique risk factors, it becomes clear that we should aim to discover actionable aetiological factors that are not endotype-specific. Subjective judgements and pragmatism will influence whether and how a prevention trial is planned. Current approaches target high-risk individuals, which reduces the required number of trial participants but increases the cost of identifying trial participants. A prevention trial targeting infants in the general population with a multivalent antiviral vaccine will likely need over 50,000 participants, depending on circumstances and assumptions. While such a trial is conceivable, it would demand robust safety data before initiation.
预防1型糖尿病仍然是一项重大挑战，但持续的努力正使我们离这一目标越来越近。本文讨论了异质性和偶然性在1型糖尿病预防方面的一些影响，特别是在证据解读和未来试验规划方面。通过模拟，我以TN10替普珠单抗试验为例，说明了在具有事件发生时间终点的预防试验中疗效估计的不确定性。我强调风险异质性并不等同于治疗效果异质性。当考虑到影响疗效的因素时，在某些情况下，要可靠地识别治疗效果异质性，所需样本量可能比确定总体疗效所需的样本量大约四倍。使用2×2析因设计来研究两种治疗方案，可以提高预防试验的效率。我还模拟了探索性研究中的统计功效，这些研究涉及采用不同策略处理潜在的1型糖尿病内型进行多重检验。如果将内型定义为具有至少部分独特风险因素的1型糖尿病相关表型的亚型，那么很明显我们应该致力于发现并非内型特异性的可采取行动的病因因素。主观判断和务实态度将影响预防试验是否以及如何规划。目前的方法针对高危个体，这减少了试验参与者的所需数量，但增加了识别试验参与者的成本。根据具体情况和假设，一项针对普通人群中婴儿的多价抗病毒疫苗预防试验可能需要超过50,000名参与者。虽然这样的试验是可以想象的，但在启动之前需要可靠的安全数据。
