Scholarlink

<p>Disproportionality analysis is a method of safety signal detection based on quantitative analysis of spontaneous reports of adverse events. Disproportionality findings are often presented in medical publications as real-world evidence on drug safety. In this paper, we review theoretical properties of disproportionality analysis in the framework of causal inference theory. We show that measures of disproportionality can approximate the causal rate ratio for a specific drug–event combination when the study drug and the set of comparator drugs satisfy all of the following conditions: (1) there is no uncontrolled confounding for the drug–event association of interest, (2) under-reporting for the event of interest is either absent or has the same relative magnitude for the study drug and for the comparator drugs, and (3) reporting rates for all adverse events combined are the same for the study drug and for the comparator drug set. Because these conditions are typically not even approximately satisfied in practice, the overwhelming majority of disproportionality hits represent statistical noise rather than causal associations. Researchers choosing to report disproportionality findings in publications should explicitly acknowledge all key assumptions and the exploratory nature of this data-mining technique.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Disproportionality Analysis and Causal Inference in Drug Safety

  • 期刊论文

  • 发表日期：

    2025年3月4日

查看全文
  • Emil Scosyrev,
  • Sigrid Behr,
  • Devendra Jain,
  • Arun Ponnuru,
  • Christiane Michel

摘要

Disproportionality analysis is a method of safety signal detection based on quantitative analysis of spontaneous reports of adverse events. Disproportionality findings are often presented in medical publications as real-world evidence on drug safety. In this paper, we review theoretical properties of disproportionality analysis in the framework of causal inference theory. We show that measures of disproportionality can approximate the causal rate ratio for a specific drug–event combination when the study drug and the set of comparator drugs satisfy all of the following conditions: (1) there is no uncontrolled confounding for the drug–event association of interest, (2) under-reporting for the event of interest is either absent or has the same relative magnitude for the study drug and for the comparator drugs, and (3) reporting rates for all adverse events combined are the same for the study drug and for the comparator drug set. Because these conditions are typically not even approximately satisfied in practice, the overwhelming majority of disproportionality hits represent statistical noise rather than causal associations. Researchers choosing to report disproportionality findings in publications should explicitly acknowledge all key assumptions and the exploratory nature of this data-mining technique.