Data Analysis and Presentation

DATA ANALYSIS AND PRESENTATION

Here we are concerned with the analysis–including quantifying–and presentation of adverse event (and other medical) data that has been coded using MedDRA, as distinct from searching the database and finding the relevant cases. The principal issues are the large number of terms in MedDRA; the most appro-priate levels and groupings for the required purpose; multiaxiality; and version changes.

The large number of terms in MedDRA (in particu-lar, PTs) may complicate analyses involving counts of events. For example, if adverse events from a parallel group clinical trial with 100 patients in each treatment arm had been coded using a legacy dictionary such as COSTART, we might see an event such as paraes-thesia occurring in 10% of patients receiving Treat-ment A and in 2% of those receiving Treatment B. However, if there were actually more specific symp-toms of paraesthesia described by trial subjects and recorded by the investigators, the differences between treatment groups might be less apparent when using MedDRA PTs, as shown in Table 13.9.

By contrast, if analysis had been carried out using the respective HLT, Paraesthesias and dysesthesias, the 10% versus 2% difference would have been main-tained. The use of group terms in analyses might also prove problematic, however. Thus, some HLTs – especially in the Investigations SOC – include PTs that represent opposing concepts. For example, the HLT Platelet analyses includes PTs Platelet count decreased, Platelet count increased, Platelet count abnormal and Platelet count normal. Such a group-ing would not be helpful if comparing effects on platelets between two treatments. In other instances, MedDRA groupings may include terms representing concepts that, whilst not in opposition, are signifi-cantly different medically. As an example, ten reports of an adverse event represented by the HLT Ventric-ular arrhythmias and cardiac arrest might relate to ten cases of Torsade de Pointes (a particularly seri-ous type of arrhythmia) or ten cases of Ventricular extrasystole (a generally benign and mild form of rhythm disorder).

A review of some of the adverse events commonly seen in clinical trials (Brown, 2004) showed that the use of MedDRA PTs might increase the number of available terms (and hence ‘dilute’ differences between treatment arms) dramatically. However, in practice, the ratio of MedDRA to WHO-ART PTs in clinical trials has been reported as around 2:1 (Kubler et al., 2005).

Presentation of adverse event data using MedDRA is also the subject of ‘Points to Consider’ Guide-lines (MedDRA  Data Retrieval and Presentation, 2004). Many of the points are identical to those concerning data retrieval, or have been described above. However, the Guidelines describe a number of approaches to presenting data for various purposes and discuss the advantages and disadvantages of these.

Thus, an overall presentation of the safety profile highlights the distribution of ADR/AEs across SOCs and helps identify areas where more in-depth analysis should be conducted. The Guidelines recommend that data should be presented in a way that allows ready recognition of patterns of terms potentially associated with relevant medical conditions. They point out that the conventional display of data as PTs under the respective SOCs alone may not optimally represent the frequency of events and can be misleading. For example, if a number of reports describe a similar medical condition, they could be represented under various specific PTs, so that it may not be apparent that they are associated.

As a first look, one should display all data as an overview according to primary SOC, including HLGTs, HLTs and PTs (as in Table 13.6). This applies to standard tables for clinical trial and post-marketing adverse reaction data and post-marketing cumulative summaries. It assures that all events will be seen and the overview might be useful in identifying clusters, perhaps in an HLT or HLGT. For a small data set, this might be all that is required. This overview can be used as the basis for planning more in-depth analyses.

Line listings (both clinical and post-marketing data) can also be displayed by primary SOC and PT. While it might be sufficient to use these for small data sets, the Guidelines indicate that it might be preferable to display data by HLGTs and HLTs as well as show-ing SOC and PTs for more complex data. Graphical display such as histograms, bar charts and pie charts showing event terms might facilitate understanding by the viewer.

In some situations, the Guidelines suggest that a more focused presentation of data may be required, in addition to the Overview by Primary SOC. For exam-ple, when reviewing in more depth any clusters seen in Primary SOC output, or for looking at previously identified safety concerns, monitoring events of special interest or responding to regulatory and other queries, it may be appropriate to expand the presentation by showing additionally terms in their secondary (mulitaxial) SOC locations. The Guidelines recommend display of the SOC or the HLGT/HLT relevant to the search, showing all the primary and secondary terms. An example of such a presentation is shown in Table 13.10. It must be remembered that, using this method, displaying more than one SOC may lead to double (or a higher multiple) counting of terms. The Guidelines also suggest, if appropriate, the linking of relevant PTs from the three non-multiaxial SOCs (i.e., SOC Investigations, SOC Surgical and medical procedures and SOC Social circumstances). However, for medical conditions that are likely to involve terms in more than one SOC, it is proposed that users should consider using an SMQ.