762 - I am interested in the safety data for one of the products. Can i use data from a pooled analysis of 10 clinical trials (same indication and product) in our marketing materials. It contains data from pivot clinical trials as well. Can it be considered as a meta-analysis?
-
The short answer to both questions in your described case is, “Almost certainly no”.
In general terms, there are two instances in which PAAB will accept pooled analysis as stand-alone evidence for a therapeutic claim relating to a healthcare product:
i) The pooled analysis is in the TMA
ii) A preplanned pooled analysis. These are quite uncommon and typically involve 2 or 3 studies as opposed to 10. For example, the same trial was conducted in two sites or jurisdictions and were published separately, but the published paper for the pooled analysis states that the plan had always been to combine the studies in a pooled analysis. When studies are designed with pooling in mind, the endpoints are defined & measured the same way, the drugs are dosed the same way, the same inclusion/exclusion criteria were used, and so on. The degree of similarity between pooled studies has a substantial impact on the validity of the pooled analysis. CAVEAT: Even when the decision to pool is made a priori, the PAAB will not accept a pooled analysis if the individual studies do not meet the rigors of the PAAB Code and guidances.
Failing these two scenarios, pooled analysis are only used to corroborate (and immediately follow) acceptable data from the TMA or an individual RCT (which meets all the standards of the PAAB Code and guidances). In this case, the pooled analysis is not the evidentiary support for the claim. It is simply re-affirmation of the same claim which is already supported by an acceptable individual trial (i.e., same endpoint, same comparator, same or shorter duration, similar magnitude of effect/safety, same study population). It is possible, although not required, that one or more of the individual studies included in the pooled analysis satisfies this requirement (particularly for efficacy endpoints which are more likely to have statistically significant findings in an individual study).
You see that, in general, the ability to use pooled analysis in APS is quite narrow. The possibilities get even narrower when dealing with safety data. Adverse events tend not to undergo statistical analysis in individual studies. The potential option to corroborate the findings in an existing individual study is therefore typically non-existent.
In response, to your question about whether this could be considered a meta-analysis, the answer is, “No, but it wouldn’t really matter from an APS perspective”. Meta and pooled analysis are entirely different types of analysis. For example, the studies that are included in a meta-analysis are dictated by a documented and reproducible systematic analysis. This is not the case for a pooled analysis. This is only one of many differences between meta and pooled analysis but it likely marks the simplest way to differentiate between these two forms of analysis. Nonetheless, the instances in which meta-analysis are acceptable for therapeutic claims are pretty much aligned with the approach for pooled analyses. The only difference being that item ii (preplanned) would not occur in a meta-analysis context. The PAAB accepts meta-analysis data excerpted from the TMA. The PAAB also accepts high-quality meta-analysis data to corroborate (and immediately follow) acceptable data from the TMA or an individual RCT (which meets all the standards of the PAAB Code and guidances). Again by “corroborate” we mean that the presented data simply re-affirms the same claim which is already supported by an acceptable individual trial (i.e., same endpoint, same comparator, same or shorter duration, similar magnitude of effect/safety, same study population). The matter of “high quality” in the context of meta-analyses is discussed in response to question 685.
