Sponsors
ADDIS project update June 2010
This page describes our progress since starting to develop ADDIS (April 2009) until now (June 2010). It is focussed on software development, an overview of our research achievements can be found on the publications page.
Achievements
We set out to develop an automated tool and method for evidence-based drug benefit-risk assessment. The initial version (ADDIS v0.2), released in June 2009 reflected this: we implemented a database of clinical studies that could be used to perform (rudimentary) pooling of results and (primitive) benefit-risk assessment. The insights gained in implementing ADDIS v0.2 and the problems we identified drove our research and software development priorities for the following year.
Development of ADDIS v0.4 was started in September 2009, and from October 2009 we hired two part-time developers to aid in the process. ADDIS v0.4 solidified what was done for v0.2, and implemented a statistical method for meta-analysis of clinical trials. The primitive benefit-risk method was removed to be replaced by a sound model later on. ADDIS v0.4 was released in December 2009.
After ADDIS v0.4, we had identified network meta-analysis as the evidence synthesis method that could support our benefit-risk model, but our knowledge was not sufficient to implement it. Therefore, ADDIS v0.6 (February 2010) focussed on usability improvements, such as a better method of inputting studies and the possibility to import trial design from ClinicalTrials.gov.
ADDIS v0.8 (April 2010) finally introduced network meta-analysis, based on the MTC library developed in parallel by Gert van Valkenhoef. This made ADDIS the first software in the world to support automated model generation for network meta-analysis, and therefore automated network meta-analysis. In addition, the ability to import and export the database in XML format was added. ADDIS v0.8.1 transitioned ADDIS to load/save exclusively in XML, dropping the binary format altogether.
Having developed automated network meta-analysis, we were ready to implement our benefit-risk model using the JSMAA software for multi-criteria decision aiding, developed by Tommi Tervonen. Jing Zhao performed her master's thesis with our research group to pilot and evaluate the network meta-analysis based benefit-risk method. This helped us to implement the method in ADDIS v0.10, released June 2010. In addition to the benefit-risk model itself, we provide a method to (heuristically) convert measurement scales to clinically interpretable measures to aid in weighting the criteria.
Future goals
In September 2010, we plan to release ADDIS v1.0, which will have basically the same functionality as ADDIS v0.10, but will offer a more polished experience and more solid basis for further expansion.
After this, our top priority will be to enable editing of studies that have been entered to correct measurements or input additional outcome measures. Then, we will aim to offer integration with external systems such as SmPC databases, ClinicalTrials.gov and the ICTRP search portal. After that, we plan to implement additional analysis methods, both for benefit-risk analysis and evidence synthesis.
In the long term, we would like to move to a more generally applicable and standards-based data model to replace the current one. This would enhance interoperability and enable things such as stratified (meta-)analyses.
Personnel
Initially, development was performed by Gert van Valkenhoef (PhD student) and Tommi Tervonen (post-doc). From October 2009, they were aided by Tijs Zwinkels and Maarten Jacobs (part-time developers). From January 2010 to May 2010, Hanno Koeslag replaced Maarten Jacobs. Tijs Zwinkels' and Tommi Tervonen's contracts will terminate 1 September 2010, while Gert van Valkenhoef will transition to an advisory role a that time (and relocate to the UMCG). The position of lead developer will be taken by Ahmad Kamal, a software engineering PhD student. Florin Schimbinschi is being hired as a software developer, and we hope to attract another developer in the near future. So far, our agile development methodology has enabled us to cope with frequent personnel changes with limited impact on productivity.