Conceptualizing Agonistic Interaction in a Marginal Sufficient Component Cause Model: An Alternative Interpretation for Subadditive Interaction

視覺化同效型交互作用之充分組成病因模型：對次加成作用的另類詮釋

戴安順(An-Shun Tai) ； 黃彦棕(Yen-Tsung Huang) ； 李文宗(Wen-Chung Lee) ； 林聖軒(Sheng-Hsuan Lin)

《中國統計學報》 58卷3期 (2020/09) Pp. 168-198

摘要

The sufficient-component cause (SCC) framework, one of the most sophisticated techniques for the methodological development of causal inference, has the advantage of visualizing the interaction effect of synergism or antagonism. However, statistical interaction occurs even without synergism or antagonism, and vice versa. In this study, we propose a marginal SCC (mSCC) model and incorporate it into the counterfactual framework. The mSCC model can visualize agonism, which is a crucial subtype of interaction apart from synergism and antagonism. Causal pie weight (CPW) and population attributable fraction are also illustrated in the mSCC framework. A hypothetical example is used to demonstrate the use of the mSCC model to identify the agonist and connect mechanistic and causal interaction. Our method is applied to Taiwanese cohort data. We perform simulation studies to evaluate the performance of our proposed CPW estimator in separate scenarios of no synergistic interaction and no agonistic interaction. The small biases indicate that the estimated CPWs are close to the true values, and the coverage rates are approximately 0.95. Among all cases of hepatocellular carcinoma (HCC), independent effects due to hepatitis C virus (HCV), hepatitis B virus (HBV), and the lower bound of agonistic interaction account for 20.3%, 30.8%, and 12.6% cases of HCC, respectively. Under the assumption of no synergistic interaction, the proportion of agonistic interaction is exactly 12.6%. Our finding regarding agonism.

關鍵字

Interaction ； Agonism ； Synergism ； Sufficient Component Cause Model ； Causal inference ； Mechanism Investigation

並列摘要

充分組成病因模型架構為目前因果推論中極重要的模型之一，特別是能夠定義以及視覺化機制型交互作用（包含協同型交互作用與拮抗型交互作用）。然而，機制型交互作用與統計上的交互作用往往並不一致。因此，本篇研究在反事實模型的框架下，提出邊際充分組成病因模型；並使用該模型識別出一個不同於「協同」與「拮抗」的機制型交互作用—同效型交互作用，以及重新詮釋因果圓派權重與族群可歸因比率。除此之外，邊際充分組成病因模型能使統計上的交互作用都能有特定的機制型交互作用的解釋。在數值分析的部分，本研究採用模擬分析以評估該模型之效能。模擬結果顯示該模型估計式具有不偏的性質且其覆蓋率能有效地被控制在95%。同時，此方法被套用至台灣肝癌資料庫以探討C型肝炎病毒與B型肝炎病毒對於肝癌的致病機制。分析結果顯示C肝病毒獨立作用、B肝病毒獨立作用以及兩病毒的同效作用在全體肝癌患者中分別佔有20.3%、30.8%以及12.6%。我們對同效作用的發現成功地解釋了統計上的交互作用與機制型交互作用之間的不一致，這將有助於更深入地暸解因果機制。

並列關鍵字

交互作用 ； 同效作用 ； 協同作用 ； 充分組成病因模型 ； 因果推論 ； 機制型交互作用

參考文獻

Darroch, J (1997). Biologic synergism and parallelism. American journal of epidemiology, 145(7), pages 661-668. https://doi.org/10.1093/oxfordjournals.aje.a009164

Flanders, WD (2006). On the relationship of sufficient component cause models with potential outcome (counterfactual) models. European Journal of Epidemiology 21, pages 847-853. https://doi.org/10.1007/s10654-006-9048-3.

Lee, M-H, Yang, H-I, Lu, S-N, Jen, C-L, Yeh, S-H, Liu, C-J, Chen, P-J, You, S-L, Wang, L-Y, Chen, WJ and Chen, C-J (2010). Hepatitis C virus seromarkers and subsequent risk of hepatocellular carcinoma: long-term predictors from a community-based cohort study. J Clin Oncol, 28(30), pages 4587-93. https://ascopubs.org/doi/10.1200/JCO.2010.29.1500

Lee, W-C (2013b). Assessing causal mechanistic interactions: A peril ratio index of synergy based on multiplicativity. PLOS ONE 8(6): e67424. https://doi.org/10.1371/journal.pone.0067424

Lin, J-H and Lee, W-C (2015) Testing for mechanistic interactions in longterm follow-up studies. PLoS One 10(3): e0121638. https://doi.org/10.1371/journal.pone.0121638.