Analysis of Adverse Events in the Presence of Discontinuations
- PDF / 8,095,967 Bytes
- 9 Pages / 648 x 864 pts Page_size
- 105 Downloads / 221 Views
Analysis of Adverse Events in the Presence of Discontinuations Gerd Rosenkranz, PhD Biosfaiisficsa r i d Siaiisf icul Reportirig. Novariis Plicrr-riiaAG. Busel. Switxr-lurid
Key Words Advrrsr el’rliis; Discorifi r i u c t f ioris:
Deperidenf cerisoririg: Semiparitmeiric rneihods Correspondence Address Gerd Rosriikf’fltiz. PhD. Biosfutisfics arid Sfafisficul Relm’iiiig. Novurtis Phcii-ma AG. CH-4002 Busel. Swifzerlitrid (e-mail: grrd.roseiikraii: @rior,ar.fis.coiii).
This paper- is a r-evisrd vet-sioii of’ it prrseiiiufiori
defiiwed at llie 15111 Inieriiuiioiial Workshop ori
SfafisticulMefhodoloLv iri Cliiiical RtiD. Dubliii. 11-eluiid.A p i l 19-21, 2004.
INTRODUCTION The occurrence of adverse events under drug therapy is commonly summarized in terms of incidences (ie, number or percentage of patients having experienced a specific event). This is the case for reports of clinical studies. for publications of results of such studies, and for the prescription information after a drug has been approved by health authorities. However, to summarize information about adverse events and to compare the frequency of them in this way is incomplete for several reasons. Often, it is not clear in which time frame the adverse events did occur, and the same incidence rates observed over different time periods have different meanings. Even if observation periods are similar, incidences are hardly comparable under different patterns of discontinuations. Ironically, a drug that causes patients to discontinue earlier can look safer because it gives events a smaller chance to occur. In an attempt to address the issue, time to the first occurrence of an adverse event has sometimes been considered. Consequently, survival analysis methods have been applied with the idea in mind that these methods account for discontinuations. Incidences at specific time points can then be derived from a Kaplan-Meier
79
Commonly, the occurrence of adverse events in clinical studies is summarized in terms of incidences. “his approach has several drawbacks, such as the lack of reference to Observation time and absence of information about discontinuations. lf suwival analysis methodology is used instead, it is rarely considered that the time to an adverse event and the time to discontinuation might not follow a random censorship model. In this situation, classical statistical methods like Kaplan-Meier estimates or log-rank tests are no longer appropriate. In this article, 1 investigate the applicability of several methods that account for a potential dependence between event times and discontinuation times on the analysis of adverse events.
estimate (1).Although such an approach is generally preferable over the calculation of proportions, survival time methods are mainly appropriate under random censorship (ie. when the censoring time is independent of the time to an adverse event). Sensitivity analyses are sometimes performed to assess whether this assumption is appropriate. Dependent censoring has been investigated primarily in the context of effi
Data Loading...
