Morikawa Kaoru
Division of Safety Information on Drug, Food and Chemicals, National Institute of Health Sciences, 1-18-1 Kamiyoga, Setagaya-ku, Tokyo 158-8501, Japan.
Kokuritsu Iyakuhin Shokuhin Eisei Kenkyusho Hokoku. 2011(129):1-26.
The worldwide situations of drug safety have changed dramatically. Drugs are used based on the evaluation of safety data collected in clinical practice worldwide. US Food Drug Administration collects spontaneous reports and requires manufacturers to report adverse drug reactions (ADRs) of US marketed drugs occurring worldwide. These worldwide data are available through the Adverse Event Reporting System (AERS) (about 4.1 million reports on about 3,073,340 patients, for 13 years: 1997.4th qr-2010.4th qr.). The current issues are how to analyze and utilize such large-scale safety data. Potential biases should always be kept in mind, because AERS is based on spontaneous reports. However, its huge volumes and exhaustiveness allow for sufficient scientific evaluation with the aid of current IT technology. Therefore, analysis of large-scale ADR database becomes a new research area not only from the medical science but also from the statistical viewpoint. In this report, I introduce some case studies in which we analyzed the AERS data on psychotropics including antipsychotics, antiepileptics, and antidepressants. Antipsychotics caused ADRs specific to each drug, and, in combination therapy, increased the incidences of diabetes mellitus, pancreatitis, and neuroleptic malignant syndrome; antiepileptics caused AEs (adverse events) including serious skin reactions such as Stevens-Johnson syndrome (SJS), congenital anomaly, and closed-angle glaucoma; and antidepressants caused AEs including serotonin syndrome, suicidal events, and congenital anomaly, and AEs occurring at a higher incidence for other indications, drugs often used in the elderly and AEs in combination therapy. We have analyzed ADRs associated with concomitant drug therapies using Bayesian approach. In the analysis we faced difficulties of overdispersion and we have to estimate a number of parameters, given a large number of target drugs as well as ADRs. In addition, ADR reports are not collected from uniform populations, we also have to consider the variations in the target populations. So, we use Bayesian statistics. Bayesian analysis has become feasible with advances in computer technologies and the Markov chain Monte Carlo (MCMC) methods. It allows us to analyze ADRs associated with concomitant drug therapies and estimate the ADR signals for each drug. Therefore, the analysis and evaluation of large-scale ADR database can provide important safety information in clinical practice and the studies on ADR database are the most important issues in ensuring the postmark safety of pharmaceutical products.
全球药品安全形势已发生巨大变化。药品的使用是基于对全球临床实践中收集的安全数据的评估。美国食品药品监督管理局收集自发报告，并要求制造商报告在美国上市药品在全球范围内发生的药品不良反应（ADR）。这些全球数据可通过不良事件报告系统（AERS）获取（1997年第4季度至2010年第4季度的13年间，约有410万份报告涉及约3073340名患者）。当前的问题是如何分析和利用如此大规模的安全数据。由于AERS基于自发报告，所以应始终牢记潜在的偏差。然而，其庞大的数量和详尽性使得借助当前的信息技术进行充分的科学评估成为可能。因此，从医学和统计学角度来看，大规模ADR数据库的分析都成为了一个新的研究领域。在本报告中，我介绍一些案例研究，在这些研究中我们分析了AERS中关于精神药物的数据，包括抗精神病药、抗癫痫药和抗抑郁药。抗精神病药会引发每种药物特有的ADR，并且在联合治疗中，会增加糖尿病、胰腺炎和神经精神性恶性综合征的发生率；抗癫痫药会引发不良事件（AE），包括严重的皮肤反应，如史蒂文斯 - 约翰逊综合征（SJS）、先天性异常和闭角型青光眼；抗抑郁药会引发AE，包括5-羟色胺综合征、自杀事件和先天性异常，以及在其他适应症、常用于老年人的药物和联合治疗中发生率较高的AE。我们使用贝叶斯方法分析了与联合药物治疗相关的ADR。在分析过程中，我们面临着过度离散的困难，并且由于有大量的目标药物以及ADR，我们必须估计许多参数。此外，ADR报告并非从统一的人群中收集，我们还必须考虑目标人群的差异。所以，我们使用贝叶斯统计。随着计算机技术和马尔可夫链蒙特卡罗（MCMC）方法的进步，贝叶斯分析变得可行。它使我们能够分析与联合药物治疗相关的ADR，并估计每种药物的ADR信号。因此，大规模ADR数据库的分析和评估可以在临床实践中提供重要的安全信息，而对ADR数据库的研究是确保药品上市后安全性的最重要问题。
