Department of Radiology, Academic Medical Center, University of Amsterdam, PO Box 22700, 1100 DD Amsterdam, The Netherlands.
Department of Clinical Epidemiology, Biostatistics and Bioinformatics, Academic Medical Center, University of Amsterdam, PO Box 22700, 1100 DD Amsterdam, The Netherlands.
Res Synth Methods. 2010 Apr;1(2):136-48. doi: 10.1002/jrsm.10. Epub 2010 Aug 25.
For many diseases (e.g. rectal cancer and the Crohn disease), more than two stages exist and as treatment mostly depends on disease stages, correctly determining this by a diagnostic test is very important. To determine their role in clinical practice, the value of these tests should be carefully evaluated, and summarizing results in meta-analysis should also be done appropriately. A multinomial model for meta-analyzing data with more than two categories has previously been developed; these data were considered as nominal categories. However, there is an ordinal character within staging data. In this study we extended this multinomial model to three ordinal models (models for the logits of adjacent-categories, for continuation-ratio logits and for proportional odds logits) to summarize the ordinal character of staging data. Both fixed- and random-effects approaches were developed and compared. The principles of the multinomial model as well as three ordinal models are shown by fitting these models using the data on staging of rectal cancer by endoluminal ultrasonography and magnetic resonance imaging. The proportions of patients correctly staged, understaged, and overstaged per stage are obtained by these models. Because of the increased interest in meta-analyses for evidence-based guidelines, these models can be helpful in summarizing staging data. Copyright © 2010 John Wiley & Sons, Ltd.
对于许多疾病（例如直肠癌和克罗恩病），存在两个以上的阶段，并且由于治疗主要取决于疾病阶段，因此通过诊断测试正确确定这一点非常重要。为了确定它们在临床实践中的作用，应该仔细评估这些测试的价值，并且还应该适当地在荟萃分析中总结结果。以前已经开发出了用于对具有两个以上类别的数据进行荟萃分析的多项式模型；这些数据被视为名义类别。但是，分期数据中存在有序特征。在这项研究中，我们将该多项式模型扩展到了三个有序模型（用于相邻类别的对数几率、用于连续比对数几率和用于比例优势对数几率的模型），以总结分期数据的有序特征。开发了固定效应和随机效应方法并进行了比较。通过使用腔内超声和磁共振成像对直肠癌分期的数据拟合这些模型，展示了多项式模型以及三个有序模型的原理。通过这些模型可以获得每个分期中正确分期，分期不足和分期过度的患者的比例。由于对循证指南的荟萃分析的兴趣增加，因此这些模型可以帮助总结分期数据。版权所有©2010 约翰·威利父子有限公司。
