病例對照研究中以勝算為基礎之因果機轉交互作用檢定

Abstract

背景： 交互作用是流行病學中一個重要的課題。其中一種特別的交互作用類型為「因果機轉交互作用」，近期受到許多關注。評估因果機轉交互作用仰賴的關鍵指標為「歸因於交互作用的相對超額風險」。在稀有疾病下的病例對照研究中，該指標以勝算比為基礎。 方法： 在本研究中，作者以勝算為基礎評估因果機轉交互作用，而非以勝算比為基礎。 作者進行蒙地卡羅模擬，比較以勝算為基礎與以勝算比為基礎的檢定表現。 結果： 以勝算為基礎進行檢定的型一誤差率在控制之中，或稍顯保守。在小樣本時，以勝算比為基礎進行檢定的型一誤差率膨脹嚴重。以勝算為基礎進行檢定的檢定力高於以勝算比為基礎進行檢定的檢定力。 結論： 根據這些令人滿意的特性，我們建議在稀有疾病下的病例對照研究中，使用本研究所提出的以勝算為基礎進行因果機轉交互作用檢定。

Background. Interaction analysis is an important topic in epidemiology. A special type of interactions, the ‘causal mechanistic interaction’, has received much attention recently. Assessment of causal mechanistic interactions relies on a key index,the ‘relative excess risk due to interaction’. For case-control study of rare diseases,the index is based on odds ratios. Methods. In this study, the authors assess causal mechanistic interactions based on odds instead of odds ratios. The authors perform Monte-Carlo simulations to compare the performances of the odds-based and the odds-ratios-based tests. Results. The odds-based tests have type I error rates under control or are slightly conservative. The type I error rates of the odds-ratios-based tests can be severely inflated for small sample sizes. The odds-based tests have higher powers than the odds-ratios-based tests. Conclusions. In light of their desirable properties, we recommend using the odds-based tests proposed in this paper to assess causal mechanistic interactions in case-control studies of rare diseases.