Li Y J, Kan H, He Y N, Li Y X, Mu Y T, Dai J H, Zheng Y J
Department of Epidemiology, School of Public Health, Fudan University, Shanghai 200032, China; Key Laboratory for Health Technology Assessment, National Commission of Health and Family Planning, Fudan University, Shanghai 200032, China; Key Laboratory of Public Health Safety, Ministry of Education, School of Public Health, Fudan University, Shanghai 200032, China.
Department of Epidemiology and Biostatistics, School of Public Health, Xinjiang Medical University, Urumqi 830011, China.
Zhonghua Liu Xing Bing Xue Za Zhi. 2020 Apr 10;41(4):589-593. doi: 10.3760/cma.j.cn112338-20191030-00770.
Due to the flaws inherited in synchronicity, statistical association and survivor bias on variables under measurement, a common 'consensus' has been reached on "cross-sectiional studies (CSS) can lead to failure on causal inference". In this paper, under both causal thinking and diagram, the real and measured cross-sections are clearly defined that these two concepts only exist theoretically. In real CSS research, the temporal orders of measured variables are all non-synchronic, equivalent to the assumption that measurement variables are independent to each other, or there is no differentiated classification bias. Similar to cumulative case-control or historical cohort studies, both exposure and outcome that exist or occur before their measurements in cross-sectional studies, are actions of historical reconstruction or doing 'Archaeology'. One of the common preconditions for causal inference in such studies is that: there must be a causal relation between the measured variables and their historical counterparts. The measured variables are all agents of their corresponding real counterparts, and the temporal orders are not that important in causal inference. It is necessary to better understand the analytic role of the CSS.
由于同步性、统计关联性以及测量变量的幸存者偏差中存在的固有缺陷，人们就“横断面研究（CSS）可能导致因果推断失败”达成了一种常见的“共识”。在本文中，在因果思维和因果图之下，明确界定了真实横截面和测量横截面，这两个概念仅在理论上存在。在实际的CSS研究中，测量变量的时间顺序都是非同步的，这等同于测量变量相互独立的假设，或者不存在差异化分类偏差。类似于累积病例对照研究或历史性队列研究，横断面研究中在测量之前就已存在或发生的暴露和结局，都是历史重建或进行“考古”的行为。此类研究中因果推断的一个常见前提是：测量变量与其历史对应物之间必须存在因果关系。测量变量都是其相应真实对应物的代理，并且时间顺序在因果推断中并非那么重要。有必要更好地理解CSS的分析作用。
