Stanulović Vid, Kadlecová Pavla, Zelkó Romána, Kerpel-Fronius Sándor
Int J Clin Pharmacol Ther. 2015 Jul;53(7):499-503. doi: 10.5414/CP202283.
Previous analysis of US FDA Medwatch safety alerts for monoclonal antibody therapeutics demonstrated that premarketing clinical trials can predict more than half of safety concerns. We expanded this analysis to assess whether the predictable alerts are detected sooner than the unpredictable alerts.
Times to alert were compared using Mann-Whitney test, Kolmogorov-Smirnov test, and using curves displaying cumulative frequencies of alerts over time.
Until December 31, 2013 inclusive, 76 Medwatch alerts for therapeutic monoclonal antibodies were reported: 43 predictable vs. 33 unpredictable. Predictable alerts were reported at a median (IQR) of 41 (19-77) months after approval vs. 53 (23-73) months for the unpredictable alerts. The mean (SE) was 52.07 (6.69) months and 55.91 (7.06) months for the predictable and unpredictable, respectively. Although the difference of 12 months between medians of time to alert was observed, the difference was not demonstrated as significant. Cumulative frequency curves show that predictable alerts were detected sooner until month 73 after approval, when ~ 80% of alerts were detected. Immunological reactions (such as infusion reactions, anaphylaxis, and reactions due to antibodies) were identified early; all 12 such alerts were released before the curves of cumulative frequencies cross at month 73. On the other hand, reactions occurring after the curves cross are predominantly late-occurring cancers and opportunistic infections.
The results imply that focus on predictable reactions defined as potential risks may play a role in early detection of important safety concerns.
先前对美国食品药品监督管理局（US FDA）关于单克隆抗体治疗药物的Medwatch安全警报分析表明，上市前临床试验能够预测超过半数的安全问题。我们扩展了该分析，以评估可预测的警报是否比不可预测的警报更早被发现。
使用Mann-Whitney检验、Kolmogorov-Smirnov检验以及随时间显示警报累积频率的曲线来比较警报时间。
截至2013年12月31日，共报告了76例关于治疗性单克隆抗体的Medwatch警报：43例可预测警报与33例不可预测警报。可预测警报在批准后的中位时间（四分位间距）为41（19 - 77）个月，而不可预测警报为53（23 - 73）个月。可预测和不可预测警报的平均（标准误）时间分别为52.07（6.69）个月和55.91（7.06）个月。尽管观察到警报时间中位数相差12个月，但差异未显示出显著意义。累积频率曲线表明，在批准后的第73个月之前，可预测警报更早被发现，此时约80%的警报已被发现。免疫反应（如输液反应、过敏反应以及由抗体引起的反应）被早期识别；所有12例此类警报均在第73个月累积频率曲线相交之前发布。另一方面，曲线相交后出现的反应主要是晚期癌症和机会性感染。
结果表明，关注定义为潜在风险的可预测反应可能在早期发现重要安全问题方面发挥作用。
