应用R软件crossnma程序包实现跨研究设计与跨数据格式的网状Meta分析

Performing network meta-analysis using cross-design evidence and cross-format data in crossnma package of R software

网状Meta分析(NMA)是一种通过综合多个临床研究数据且比较多种干预措施的疗效和安全性的统计技术,可为证据网络中所有干预方案提供优劣排序结果,并为临床决策提供直接证据支撑。当前,NMA通常基于同一类型数据集的汇总,而对于实现跨研究设计与跨数据格式的数据集合并仍存在方法学与软件操作上难点。R软件crossnma程序包基于贝叶斯框架与马尔可夫链蒙特卡罗算法,将三级分层模型扩展到标准NMA数据模型中以实现不同数据类型的差异化合并。crossnma程序包通过引入模型变量充分考虑不同类型数据间合并所带来的偏倚风险对结果的影响。此外,该程序包还提供结果输出和简易图形绘制等功能,这为实现跨研究设计与跨数据格式证据的NMA提供可能。本研究将通过4个个体参与者数据集与2个聚合数据集的实例对crossnma程序包模型方法和软件操作进行实例演示和讲解。

Network meta-analysis(NMA) is a statistical technique that integrates data from multiple clinical studies and compares the efficacy and safety of multiple interventions, which can provide pro and con ranking results for all intervention options in the evidence network and provide direct evidence support for clinical decision-making. At present, NMA is usually based on the aggregation of the same type of data set, and there are still methodological and software difficulties in achieving cross-study design and cross-data format data set merging. The crossnma package of R programming language is based on Bayesian framework and Markov chain Monte Carlo algorithm, extending the threelevel hierarchical model to the standard NMA data model to achieve differential merging of varied data types. The crossnma package fully considers the impact of risk bias caused by the combination of different types of data on the results by introducing model variables. In addition, the package provides functions such as result output and easy graphing,which makes it possible to combine NMA across study designs and evidence across data formats. In this study, the model based on crossnma package method and software operation will be demonstrated and explained through the examples of four individual participant datasets and two aggregate datasets.