We work closely with our clients to obtain an agreement on what they want to achieve and to understand the characteristics of the novel imaging data. We create a statistical analysis plan with mock tables and figures to ensure that we deliver analyses that meet the needs of the client. Throughout the process, we are in constant contact with the client.
Due to the novelty and complexity of the data, we need to make sure that we received data necessary to perform the analysis, including contextual information and meta-data. We design dataset standards with the help of the client to enable seemless transfer of the data when it was available.
The analysis plan describes numerous approaches to characterise the new endpoints including association between new and traditional measures, estimations of treatment effects in treatment groups, assessment of potential lags between the imaging data and traditional measures over time. We also perform further exploratory analysis on the potential use of imaging data as tool for selecting patients most likely to respond to treatment.
Rather than produce large numbers of tables, we utilise visualisation tools and graphics to enable the interpretation of large amounts of information. This includes the generation of multi-panel graphs, clustering approaches as well as time profile plots of individual patient data. In one example, our approach resulted in the identification of a new endpoint that showed significant treatment effect over 12 weeks compared to 52 weeks for the traditional endpoint.
We work hard to understand the needs of customers so that we can efficiently deliver the best results. We have set up document templates that allow us to capture the scope of work, to ensure that we are aligned with the customer and to confirm that we are agreed on the project requirements. We also work across multiple functions including statistics, clinical, bioinformatics, computational biology, genetics. With all interactions, we adopt a collaborative approach whereby we know that the best solutions are those that have been developed together.
We have worked with a broad range of different data platforms for the evaluation of surrogate markers covering protein arrays, gene expression arrays, imaging modalities, metabolomics platforms. Our experience with high-dimensional data allows us to implement practical solutions with novel data types.
We recognise that some studies require alternative approaches to analysis and reporting of results. Where there is a large amount of information to be presented, the use of graphics allows that information to be put in proper context. We have broad experience in use of multi-dimensional data displays and the development of bespoke graphical displays.
We have seen many times the problems caused by poor decision-making as a result of over-interpretation of exploratory analysis leading to wasted time and money. We understand where exploratory analyses can fit into the discovery and development process. We work with clients to place exploratory analysis into proper context and to make them aware of the potential risks. Our focus on helping our clients to conduct the best science and enabling studies to generate and test hypotheses.