基于贝叶斯理论构建中医药单病例随机对照试验模型及其临床验证研究

Randomized controlled trials in individual patient (N-of-1 trials) are multicycle, double-blinded controlled two-stage cross-over trials based on individuals. N-of-1 trials have been widely used in traditional Chinese medicine (TCM). However, Chinese herbal medicinal ingredients were complex, and their half-life periods were longer than western medicine, which resulted in residual effects in the later stages. There were no appropriate models to measure residual effects. Based on the previous study on N-of-1 trial with no residual effects, we established N-of-1 trials with no washout period (with residual effects) according to TCM, and their corresponding mathematical models. The simulation researches based on Bayesian theory are carried out. We consider the different residual effects of TCM, simulate lots of actual N-of-1 trials, generate data using computer methods, compare paired t test, t test, mixed effect model and meta-analysis, and explore the best analysis models for the different residual effects. We also carry out N-of-1 trials about the patients with ulcerative colitis, using the examples to verify these models. The mixed effects model is expected to be more suitable to analyze the data of N-of-1 trials in TCM. At last, we use the mathematical models to guide N-of-1 trials in TCM, remove the washout period, shorten the time of N-of-1 trials, and accurately estimate the residual effect of TCM.

单病例随机对照试验（N-of-1试验）是一种对单个病人进行随机、多周期二阶段交叉设计的研究方案，已经广泛用于中医药研究领域。然而，中医药N-of-1试验的中药成分复杂，半衰期较长，容易引起残留效应，可是缺乏度量残留效应的模型。因此，本研究在无残留效应N-of-1研究的基础上，构建符合中医药的无洗脱期（有残留效应）N-of-1，及对应的数学模型；采用贝叶斯理论开展模拟研究，重点考虑中医药存在的各种残留效应，模拟实际临床研究可能出现的各种情况，使用计算机方法生成数据，比较配对t检验、t检验、meta分析和混合效应模型的优劣，探索不同残留效应的最佳分析模型；并开展中医药治疗溃疡性结肠炎患者的N-of-1试验，使用实例验证构建的模型。本研究预计混合效应模型更加符合中医药N-of-1试验。最终使用数学模型指导中医药N-of-1试验的临床研究，去除洗脱期，缩短临床试验时间，精确估计中医药的残留效应。

背景：单病例随机对照试验(N-of-1试验)的个体化治疗理念与中医辨证论治有相通之处，因而在中医药临床研究中的应用日益广泛。然而，学界在N-of-1数据分析的方法以及洗脱期的设置方面仍未达成共识。.研究内容：探索适合中医药N-of-1试验的理论模型，并建立对应的混合效应模型。通过对服从正态分布、有残留效应的系列N-of-1试验定量数据进行模拟研究，比较配对t检验、混合效应模型和贝叶斯混合效应模型三种方法的统计性能。最后，通过健脾清热化湿方联合美沙拉秦治疗脾虚湿热型缓解期溃疡性结肠炎的N-of-1试验进行验证。.结果：(1)建立了无洗脱期的中医药N-of-1试验的理论模型，并建立了对应的混合效应模型：Yij = αi +ui*group + τj + λa + λb。其中Yij为第i个受试者（i = 1，2，3，…n）第j个阶段（j = 1，2，3，…6）的效应值，μ表示第i个受试者两种干预措施之间（A-B）的平均得分，服从正态分布。group为指示变量，表示干预方法：A干预（group = 1）和B干预（group = 0）。τj表示残留效应。(2)模拟研究证实：当两组处理存在残留效应时，相比配对t检验，混合效应模型和贝叶斯混合效应模型能很好控制Ⅰ类错误（0.05），估计值接近真实值（效应差 值估计值的偏差约为0），而且两种方法的检验功效大致相等。(3)健脾清热化湿方联合美沙拉秦治疗脾虚湿热型缓解期溃疡性结肠炎的N-of-1试验证实无洗脱期的中医药N-of-1试验模型及其对应的混合效应模型的可行性，实用性，值得推广。.科学意义：无洗脱期中医药临床试验数学模型及其混合效应模型，有助于校正个体自身因素和阶段效应的影响，不仅缩短试验周期，而且精确估计中医药的残留效应，弥补以往中医药N-of-1试验在方法学方面的不足，可以为中医药和循证医学的结合提供一个较好的桥梁。