Accounting for the time at which an adverse event occurs
In a previous post, I described how JMP Clinical allows you to specify time windows within an incidence analysis. Specifying time windows can provide a more informative analysis since it is possible to view how the risk of adverse events (AEs) changes over the course of a clinical trial. However, the influence of time, which is related to the exposure to study medication, tends to be ignored in most presentations of AEs. Review the safety tables in any medical journal or drug label. More often than not, AEs are presented for the entire duration of treatment.
Why is this? There may be numerous reasons: The rarity of most adverse events could make breaking the study into mutually exclusive time periods rather pointless; there is rarely sufficient space within a journal to report all findings of interest; the sheer scope may make it difficult to summarize the results meaningfully. All of these reasons can be addressed using a graphical presentation for AE incidence.
However, there is an important statistical concern – multiplicity. Simply put, the more statistical tests you perform, the more likely you are to see significant results based on chance alone. These are referred to as type I errors or false-positive findings. Breaking the study into time windows creates even more tests, and the potential for more type I errors. The JMP Clinical Adverse Event Incidence Screen includes a number of methods to minimize the risk of false positive findings, even novel methods that consider the medical grouping of AEs.
How many time windows should be specified? In general, this is a difficult question to answer. You must balance the need to specify a sufficient number of windows to see how the risk changes over time, without specifying too many windows, which would result in a loss of power for the treatment comparisons. The particular therapeutic area may suggest appropriate windows, and you should consult with your clinical staff.
JMP Clinical has other methods that account for the timing of AEs. When the AE rates appear similar between treatments, there can still be important differences in terms of the timing that the AEs occur. Sometimes, it’s only a matter of time before a subject has an adverse event, so the best thing we can hope for is the treatment to maximize the time before an event occurs.
For each type of AE within the study database, the AE Time-to-Event analytical process compares treatments using the time to the first event for each subject. Figure 1 includes a summary of all AEs for the Nicardipine example using two popular statistical tests for Nicardipine, Log Rank and Wilcoxon. Similar to the incidence screen, you can quickly identify differences between the treatments (adjusted for multiplicity) and drill down to view individual Kaplan-Meier (KM) Curves for events of interest. For example, Figure 2 displays the KM curves for Isosthenuria. Not only does the Nicardipine treatment arm have more subjects with events, but these events also tend to occur earlier than for subjects on Placebo.
This type of analysis can further enhance your understanding of the safety profile of your new drug or intervention.For more information, feel free to download the following white paper or view these short videos.