自发报告系统中发病时间分析的理论框架与关键考虑
该综述总结了自发报告数据库中发病时间(TTO)分析的定义、计算、特征与局限,并概述了纳入TTO信息的统计信号检测方法。非参数分布比较方法(如Kolmogorov-Smirnov和Anderson-Darling检验)适用于暴露人群未知的自发报告数据,可检测频率分析无法识别的发病时间结构异常。TTO分析与比例失衡分析互补,联合使用可能提高信号检测的敏感性和可解释性,但此类方法应定位为探索性工具而非临床或监管决策的直接依据。
要点
- TTO分析可补充比例失衡分析,捕捉不良事件发生的时间模式。
- 非参数分布比较方法适用于暴露人群未知的自发报告数据。
- TTO分析应作为探索性工具,不可用于估计人群风险或推断因果。
核心结论
- 自发报告数据库的TTO分析应定位为探索性工具,用于刻画发病模式、产生假设并指导后续流行病学研究设计,而非临床或监管决策的直接依据。
研究设计与样本
- 本文为综述,总结了自发报告数据库中TTO分析的定义、计算、特征与局限,并概述了纳入TTO信息的统计信号检测方法。
主要发现
- 非参数分布比较方法(如Kolmogorov-Smirnov和Anderson-Darling检验)适用于暴露人群和暴露规模未知的自发报告数据。
- 这些方法可检测频率分析单独无法识别的发病时间结构异常。
- 比例失衡分析与基于TTO的方法并非竞争关系,而是捕捉安全信号不同维度(报告频率和时间模式)的互补策略,联合使用可能提高信号检测的敏感性和可解释性。
- 综述还讨论了基于生存分析的方法和Weibull建模用于TTO数据的理论背景与应用,并强调其在自发报告系统中的固有局限。
临床怎么用
- 在自发报告数据库的信号检测中,可将TTO分析与比例失衡分析联合使用,以补充时间维度信息。
- TTO分析结果可用于产生假设和指导后续流行病学或安全性研究的设计,不应直接用于临床或监管决策。
局限性
- 自发报告数据存在报告偏倚、时间信息不完整以及缺乏非事件病例的问题。
- 基于TTO的生存分析和Weibull建模等方法不应被用于估计人群水平风险或推断因果关系。
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
原文
Spontaneous reporting databases play a central role in pharmacovigilance for monitoring the safety of drugs and vaccines. Conventional statistical signal detection has relied primarily on disproportionality analyses based on reporting frequencies, whereas information on the timing of adverse event onset has not been fully exploited. Time to onset (TTO), defined as the interval between the initiation of drug administration and the occurrence of an adverse event, provides complementary information that captures temporal patterns of event manifestation beyond simple occurrence counts. This review summarizes the definition, calculation, characteristics, and limitations of TTO analyses in spontaneous reporting databases and provides an overview of statistical signal detection methods incorporating TTO information. In particular, nonparametric distribution-comparison approaches, such as the Kolmogorov-Smirnov and Anderson-Darling tests, are well suited to spontaneous reporting data, in which the underlying population and exposure size are unknown. These methods enable the detection of abnormalities in the temporal structure of adverse event onset that may not be identifiable through frequency-based analyses alone. Furthermore, disproportionality analysis and TTO-based approaches are not competing methods but complementary strategies that capture different dimensions of safety signals-reporting frequency and temporal patterns-and their combined use may improve both sensitivity and interpretability of signal detection. The review also discusses survival analysis-based methods and Weibull modeling for TTO data, outlining their theoretical background and applications while emphasizing their inherent limitations when applied to spontaneous reporting systems. Because of reporting bias, incomplete time information, and the absence of non-event cases, such methods should not be used to estimate population-level risks or to infer causality. In conclusion, TTO analyses using spontaneous reporting databases should be positioned as exploratory tools for characterizing onset patterns, generating hypotheses, and informing the design of subsequent epidemiological and safety studies, rather than as a direct basis for clinical or regulatory decision making.
引用
- Noguchi Y, Ino Y, Yokoyama S, et al. Theoretical Framework and Key Considerations for Time-to-Onset Analysis in Spontaneous Reporting Systems. Drug Saf. 2026. PMID: 42118500. doi:10.1007/s40264-026-01677-3
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。