An approach to meta-analysis of dose-finding studies

The main goal of a Phase I dose-finding study is the estimation of the maximal tolerated dose (MTD) from a set of available dose levels. For cytotoxic clinical trials in oncology, it is not unusual to find several phases I studies carried out on a new molecule or procedure. For instance, the molecule Sorafenib (which inhibits particular tyrosine kinase enzymes in several cancers) was used alone in five published clinical trials. Each clinical trial was conducted separately in different indications and the resulting data were never pooled in any way. No attempt was made to synthesize or combine the information from the different studies. For dose-finding studies, the toxicity itself may not be related to disease. Integrating information across several Phase I trials may lead to improved inference on the dose level, or levels, corresponding to the MTD. Under strong assumptions, we will provide more accurate estimates of the MTD. Under no assumptions a pooled analysis will perform no less well than several separate analyzes and, under intermediary assumptions, there still may be scope for gains. A difficulty is that many methods are sequential in nature so that, in order to group findings under a single heading, it is necessary to retrospectively analyze data obtained according to a dynamic sequential design. We propose a solution to this difficulty. The approach is illustrated via two real examples.