Novel clinical trial designs for establishing potential efficacy of complex interventions

用于确定复杂干预措施潜在功效的新颖临床试验设计

• 批准号: MR/N015444/1
• 负责人: Duncan Wilson
• 金额: $ 31.94万
• 依托单位: University of Leeds
• 依托单位国家: 英国
• 项目类别: Fellowship
• 财政年份: 2016
• 资助国家: 英国
• 起止时间: 2016 至 无数据
• 项目状态: 已结题
• 来源: https://gtr.ukri.org/projects?ref=MR%2FN015444%2F1
• 关键词: Novel; clinical; trial; designs; establishing

Obtaining evidence about whether or not a new treatment is better than currently available treatments is difficult, time-consuming and expensive. Ideally, it requires a large experiment including hundreds, or even thousands, of patients. To make sure only treatments which have a good chance of being found to have a beneficial effect undergo these expensive experiments, smaller 'exploratory' experiments are often carried out first, giving an initial indication of whether the treatment is worth studying further.When the treatment is a drug, these exploratory experiments are common, but not all treatments are drugs. One example would be a psychotherapy treatment delivered by a psychologist for patients suffering depression. Another example might be a support group like alcoholics anonymous. Other non-drug treatments (known as 'complex interventions') can include surgery, physiotherapy, occupational therapy, speech therapy or nursing. It is generally more difficult to test these complex interventions than drugs. This is partly because we are often interested in a number of ways in which the treatment affects patients, as opposed to focussing on a single measure. Another complication comes from it being difficult to work out what is having the effect - the treatment itself, or the person delivering the treatment. When designing exploratory experiments to examine complex interventions, a fundamental decision is how many patients and care providers to study. This is important because it will determine how confident researchers can be in the results of the experiment. The larger the experiment, the less likely that we will observe an extreme effect of the treatment by chance alone. However, we also want to keep experiments as small as possible to minimise the cost and participant's time. Statisticians therefore need to work out the smallest number of participants which will give the research community enough confidence in the results to allow research funders to make reliable decisions about which treatments to take on to large, expensive experiments. However, methods for doing this for exploratory experiments of complex interventions have yet to be developed.In this project I will develop these methods, enabling researchers to properly design exploratory experiments testing complex interventions in a reliable and efficient way. I will develop several variants of the methods so that they can be applied in a range of situations. I will start by extending existing methods used within drug development, focusing on keeping the chance of making an incorrect decision (e.g. taking an ineffective treatment on to a large expensive experiment) low. After this, I will consider ways to use existing information about the treatment, such as how much we expect its effect to vary among the population, which will help minimise the number of participants needed. Finally, I will develop methods that explicitly consider the consequences of decisions, leading to experiments which better reflect the priorities of all interested parties.All of these approaches will be computer intensive. To help others use them, I will also spend time writing user-friendly and efficient computer software which allows them to be used quickly and easily and will apply the methods to real examples of trial design problems. This will involve working with clinicians and other statisticians, and will help to make sure that the methods developed are easy to apply and will be used in practice.Developing these methods will improve the process of developing and testing complex interventions. They will ensure that the limited resources available for running large clinical trials (including research funds and participant time) are spent wisely on the treatments which show the most promise. In turn, this will increase the rate at which new interventions are identified and made available, improving the standard of care for patients across the breadth of the NHS.

获得关于新治疗方法是否优于现有治疗方法的证据是困难的，耗时且昂贵。理想情况下，它需要包括数百甚至数千名患者的大型实验。为了确保只有那些有很大机会被发现有有益效果的治疗方法才能进行这些昂贵的实验，通常会先进行较小的“探索性”实验，以初步表明该治疗方法是否值得进一步研究。当治疗方法是药物时，这些探索性实验很常见，但并不是所有的治疗方法都是药物。一个例子是由心理学家为患有抑郁症的患者提供的心理治疗。另一个例子可能是一个支持团体，如酗酒者匿名会。其他非药物治疗（称为“综合干预”）包括手术、物理治疗、职业治疗、语言治疗或护理。测试这些复杂的干预措施通常比测试药物更困难。这部分是因为我们通常对治疗影响患者的多种方式感兴趣，而不是专注于单一措施。另一个并发症来自于很难弄清楚是什么产生了效果-治疗本身，还是提供治疗的人。在设计探索性实验以检查复杂的干预措施时，一个基本的决定是要研究多少患者和护理提供者。这一点很重要，因为它将决定研究人员对实验结果的信心。实验规模越大，我们就越不可能偶然观察到治疗的极端效果。然而，我们也希望保持实验尽可能小，以尽量减少成本和参与者的时间。因此，统计学家需要计算出参与者的最小数量，这将使研究界对结果有足够的信心，使研究资助者能够做出可靠的决定，决定采取哪些治疗方法进行大型，昂贵的实验。但是，对于复杂干预的探索性实验，还没有开发出这样的方法。在本课题中，我将开发这些方法，使研究人员能够以可靠和有效的方式正确设计测试复杂干预的探索性实验。我将开发这些方法的几种变体，以便它们可以应用于各种情况。我将从扩展药物开发中使用的现有方法开始，重点是降低做出错误决定的机会（例如，将无效的治疗方法用于大型昂贵的实验）。在此之后，我将考虑如何使用有关治疗的现有信息，例如我们预计其效果在人群中的差异有多大，这将有助于最大限度地减少所需的参与者数量。最后，我将开发出明确考虑决策后果的方法，从而使实验更好地反映所有相关方的优先级。为了帮助其他人使用它们，我也将花时间编写用户友好和高效的计算机软件，使他们能够快速，方便地使用，并将这些方法应用于试验设计问题的真实的例子。这将涉及与临床医生和其他统计学家合作，并将有助于确保开发的方法易于应用并将在实践中使用，开发这些方法将改善开发和测试复杂干预措施的过程。他们将确保可用于开展大型临床试验的有限资源（包括研究资金和参与者时间）明智地用于最有希望的治疗。反过来，这将提高新干预措施的确定和提供速度，提高整个NHS范围内患者的护理标准。

项目成果

A hypothesis test of feasibility for external pilot trials assessing recruitment, follow-up, and adherence rates.

评估招募、随访和依从率的外部试点试验可行性的假设检验。

• DOI: 10.1002/sim.9091
• 发表时间: 2021
• 期刊: Statistics in medicine
• 影响因子: 2
• 作者: Wilson DT

Bayesian design and analysis of external pilot trials for complex interventions.

• DOI: 10.1002/sim.8941
• 发表时间: 2021-05-30
• 期刊: Statistics in medicine
• 影响因子: 2
• 作者: Wilson DT;Wason JMS;Brown J;Farrin AJ;Walwyn REA
• 通讯作者: Walwyn REA