Kerus started out life as a software tool built within the pharmaceutical industry to address some immediate and pressing concerns. Multiple pressures are making modern clinical trials rapidly become more complex. These include personalised medicine, inclusion of real world data and an urgent need to streamline the development pipeline by making each trial answer more questions.
The task set for Dr Beattie was to take this prototype software and develop it into a more generalised product that would meet an unmet need in the pharmaceutical industry.
With an existing prototype available the first task was to review all the associated literature, including feedback from early market testing and the experience of the researcher involved in the early prototype, Dr Aiden Flynn.
This review sought to not only understand what had been done, but also what:
Competitor products were available
Gaps existed in the original conception
Key unmet market needs were
Opportunities were present
Kerus Development Strategy
The main outcomes were:
Commercial products supporting clinical trial design were not capable of meeting the needs that had been identified
The main competitor was in fact expert statisticians within the pharmaceutical teams who could manually programme simulations and subsequent analysis
The original prototype was designed to handle specific types of clinical trial designs for a limited range of data types, a commercial product needed to generalise the solution
This generalisation had to allow flexible trial designs, data types, interactions, analyses and ways of designing how success could be measured
There was a gap for a tool that could allow designers of trials to build in real-world complications and require realistic multilevel outcomes to support decision making
This tool would need to be able to allow expert statisticians and domain knowledge experts to collaborate and produce outputs that could be used to communicate with management.
The tool was essential for many cutting edge advancements in biomedical technology, including personalised medicine, companion diagnostics, ‘omics based research and benefit-risk analysis
The overall need for Kerus was to provide a simple intuitive interface that allowed the user to perform the work of an advanced biomedical statistician on complex real-world scenarios.
It required a visible framework that guides the user through the process of setting up a virtual study.In addition smart design in the background was essential to intelligently handle tasks such as controlling user choices available depending on the choices they make at each stage.
Sampling variability and healthy-fracture contrast
Early testing showed that non-intelligent workflows allowed errors to creep in too easily. Building a error tracking tool would have been more work but ironically given the impression the software was not intelligent.
Another key requirement was that the software perform the calculations precisely in order that the end users could depend on the outputs from the software. This required extensive mathematical testing on top of the usual rigorous software testing that is normal in development.
A shopping list of technical requirements (data types, statistical tests, types of interactions, graphical outputs) was built up and prioritised with respect to key market sectors.
Other requirements arose during testing, including the need to migrate to a web based solution that the user could access through a web browser. Many companies in the biomedical sector have stringent supplier and IT policies that make the process of getting a product installed in those companies a bleak challenge.
It also became clear that for the product to be scaleable the end user needed to be able to augment its functionality themselves or to be able to access 3rd party add-ons.
Research was critical to the success of the Kerus platform for ensuring both relevance and accuracy. This came in a range of different flavours, demanding a wide range of skills from the team.
Scientific literature reviews
Develop novel mathematical algorithms
Develop new development processes
User requirements
Market analysis
Performance of Kerus Algorithms
All algorithms used were rigorously validated in house and some off-the-shelf solutions were found to be inadequate. Trawling through the scientific literature uncovered cutting edge solutions to some of the shortfalls which were incorporated into the software. This includes algorithms used to handle missing information in correlation matrices.
In some keystone areas no solutions were found and for each a programme of research was planned and implemented, creating novel algorithms that allowed the user to achieve the results they needed. This included algorithms to allow melding of correlation information from different sources and to produce the required level of correlation between categorical variables.
The scientific research was where we saw the software in action, using the intermediate versions (the advantage of test lead development is that you are confident in intermediate releases) we undertook a number of research programmes in pharma, epidemiology and diagnostics. These fed back into the development of the software, ensuring a real world relevance.
Alongside development we regularly exposed our software to external end users to maintain ongoing feedback on progress and continual updating of our understanding of the market and its needs. This commitment to relevance ensured that we were able to rapidly adapt to important new challenges.
The continual real-world application of the software to client problems and the monitoring of external end-users meant that the team (now up to 7) had to respond to new challenges and needs continually.
This led to ongoing procedural improvements to make the development process more agile and responsive, including extensive standardisation of code, dramatically cutting the time it took to implement new variations of functions.
The Kerus environment
The first major pivot in the product development pathway came early on. It rapidly became clear the software would be large and complex and the structure of the inherited code was not scaleable. The software had to operate as a series of independent modules to allow the rapid addition of new functionality with minimal bugs or errors caused. Although this initial refactoring of the code was painful and slow, this decision quickly paid dividends as it reduced the time taken to implement new functions by a factor of 4.
The second major pivot was to switch from a local installable product to a web based service as the barriers to distribution of the installable would have been huge due to the highly regulated nature of the target industry.The original platform used to develop the fist version was not compatible with web technologies and so the product had to be reworked from scratch again. This time knew a third pivot was coming so we planned to accommodate the next change. The web platform was released within 18 months.
The third major change to the development process was to develop an ‘app store’. This would allow the user to build their own new functionality and add it onto the core Kerus platform. This is essential to ensure that Kerus remains future proofed and capable of adapting to the continual advances in clinical trials.
Osentia materials including test kit and contents plus typical reports adapted to the customer’s results
Following the successful validation of the software in-house, it was used for a number of client projects. One exciting project was the ADAPT-SMART and GET-REAL programmes, which were looking at how new and smarter clinical trial designs could provide more efficient and cost effective routes to regulatory approval and therefore market. Using Kerus it was possible to design a series of connected trials that could have reduced the time to approval by 18 months, the number of people exposed to risk (including potentially avoiding one death) and reducing the cost of developing the drug.
A number of ancillary programmes were implemented to develop resources to provide help and tutorials for users to guide them through the process of using Kerus to obtain insightful results. A number of real-world case studies were developed to show case the breadth of Kerus’s capabilities, each carefully constructed to ensure all the major technical capabilities of the system were demonstrated.
As project lead I was heavily involved in the development of marketing collateral, including advertisements, flyers,web pages and display stands. My main role was to ensure that the content accurately reflected the nature and capabilities of the Kerus platform.
“We had prototyped a simulator for designing modern complex clinical trials and needed someone to take the project to the next level. Dr Beattie was quick to understand the vision for the product and map out the hurdles for achieving a viable product. He focused his tem on a process of rapid evolution, developing a downloadable solution in 18 months before pivoting to development of a cloud based version in a further 18 months. Problem solvers for complex problems are few and far between; his insight and understanding of real-world data was critical in developing a solution that met the needs of our clients.“
The Kerus platform was launched as a local installable in July 2015 and the web based platform in July 2017. The former is no longer supported due to being superceded by the web platform. The platform has already been used to make a number of key advances, including the ADAPT-SMART programme,…