Statistical Methods and Software for Multivariate Meta-analysis

 描述(由申请人提供)：比较有效性研究(CER)旨在为卫生保健决策提供有关不同诊断和治疗方案的好处和风险的信息。针对特定情况的评估工具和治疗选择的数量不断增加，其成本也迅速上升，这使得在临床实践中通过多变量荟萃分析对多种诊断测试和多种治疗进行科学严格的比较的需求越来越大。虽然多变量Meta分析方法非常有用，因为它们可以提供具有更好统计特性的估计，但与传统的单变量Meta分析相比，还有许多额外的挑战。 针对PA-13-303，这项建议的总体目标是开发尖端和健壮的多元荟萃分析方法，以增强一致性、适用性和普适性，以及一个完全开源、跨平台、公开可用和易于使用的R软件包。此外，我们将把待开发的R包集成到公开可用的软件Open Meta-Analyst中，以进行高级荟萃分析。在这项建议中，我们将：(1)开发多变量网络荟萃分析框架和软件，以同时比较多个治疗与多变量结果，以及多个诊断测试(由于在患病率、敏感性和特异性方面的共同利益，本质上涉及多变量结果)；以及(2)开发多变量可视化工具和非参数/参数方法，以检测和调整发表偏倚。 我们将通过大量的真实数据分析和精心设计的模拟研究来评估这些建议方法与现有荟萃分析方法的优缺点。所提出的统计方法将广泛适用于多变量Meta分析。这两个目标的实现将使CER方案直接受益，因为它提供了使用JAGS和R统计语言的用户友好的软件中实施的最先进的方法，这些方法将免费向公众提供。它将通过促进各种癌症、心血管疾病、传染病和其他疾病的诊断和治疗来改善公共健康。

 DESCRIPTION (provided by applicant): Comparative effectiveness research (CER) is aiming at informing health care decisions concerning the benefits and risks of different diagnosis and treatment options. The growing number of assessment instruments and treatment options for a given condition, as well as the rapid escalation in their costs, has generated the increasing need for scientifically rigorous comparisons of multiple diagnostic tests and multiple treatments in clinical practice via multivariate meta-analysis. While multivariate meta-analysis methods are certainly very useful as they can provide estimates with better statistical properties, there are many additional challenges compared to traditional univariate meta-analysis. In response to PA-13-303, the overall goal of this proposal is to develop cutting-edge and robust multivariate meta-analysis methods to enhance the consistency, applicability, and generalizability, and a completely open-source, cross-platform, publicly available and easy-to-use R software package. In addition, we will integrate the to-be-developed R package into the publicly available software Open Meta-Analyst for advanced meta-analysis. In this proposal, we will: (1) develop multivariate network meta-analysis frameworks and software to simultaneously compare multiple treatments with multivariate outcomes, and multiple diagnostic tests (which intrinsically involve multivariate outcomes due to the joint interests in prevalence, sensitivity ad specificity); and (2) develop a multivariate visualization tool and nonparametric/parametric methods to detect and adjust for publication bias. We will evaluate the strengths and weaknesses of these proposed methods versus existing meta-analysis methods through many real data analyses and carefully designed simulation studies. The proposed statistical methodology will be broadly applicable to multivariate meta-analysis. Completion of these two aims will directly benefit the CER program by providing state-of-the art methods implemented in user-friendly software using the JAGS and R statistical languages that will be made freely available to the public. It will improve public health by facilitating the diagnosis and treatment of various cancers, cardiovascular, infectious and other diseases.