Reducing the Cost of Clinical Trials with Integrated Visualization

By Jason Neiss, Product Manager, Viz, General Dynamics C4 Systems

How a fresh perspective and new visualization tools can improve clinical trial management.

The innovation wave driving clinical trials toward electronic data capture and centralized data management has been convincing, yet clinical trials are still routinely over budget, suffer from poor communication, lack team awareness, and are unable to forecast and quickly resolve issues. Clinical trials are becoming less of a scientific exercise and more of a business process with three strongly coupled cost centers: design and planning, production and execution, and data analysis. Existing software tools and incumbent vendors have not yet addressed the deep-seated problems that cause cost overruns. A recent study suggests that improvements in a few key areas can result in a dramatic 60-90% reduction in total trial costs, yet beyond EDC, no class of tools were proposed as enablers [1]. This whitepaper offers examples of how an innovative collaborative visualization tool can help achieve these cost reductions.

The low-hanging fruit has been eaten: why gains from incumbent tools are diminishing
EDC, CTMS and other electronic capture tools are necessary but not sufficient for successful trials.
Eisenstein's finding that EDC reduces trial cost by about 10% is both understandable and a bit troubling. The substitution of electrons for ink takes advantage of the powers of modern computing: distributed data storage, instant access, and essentially infinite capacity. The human factors case for EDC is easy to comprehend, and the process change in principle is one of typing instead of writing – both relatively easy propositions. Oddly though, EDC and CTMS usage compliance is poor, mostly because the users of the systems extract little value from these tools [2]. Instead, the promised value is intended to flow to the managers and sponsors. These incentives can be better aligned through tools that provide all study sites tangible value beyond improved efficiencies from data entry. Furthermore, whatever the actual EDC cost savings, it is not reasonable to expect such gains in the other cost areas of study execution and data analysis, because these remaining troubled processes do not lend themselves to purely technology solutions.

The remaining problems require more brilliance, enabled by technology
The issue is not how to get good data - it is how distributed users make the most of it.
Improving the study execution process is a much harder challenge, as the number of people, processes, and inherent variability involved is much larger. From inventory tracking to enrollment profiles, adverse event analysis, financial tracking, and personnel management, clinical trials are a complex process similar in size to a large systems engineering effort. Keys to success include transparency at all levels, smooth coordination of related efforts, and frequent early exposure of problems. The most disruptive situations occur when activity is diverted from the plan. For instance, enrollment at a few sites might be lagging, and it is unlikely that they will be able to catch up. Another variance from plan would be an unexpectedly high number of adverse events in a specific population. In each case, the range of alternative solutions is wide and requires inputs from across functional boundaries, often a difficult proposition. If data, geographic, or procedural barriers prevent fluid communication and information flow among the broader trial team, issues linger for much longer than desired.

Likewise, a different set of experts are charged with data monitoring and analysis. Required to adhere to regulations and study protocols, any variances have both cost and data quality implications. Late or missed visits don't just mean schedule delays; the resulting data may be tainted and have regulatory impacts. So carefully monitoring the incoming data for impending problems is a key capability. Analytics can help here, but just detecting the presence of anomalies is inadequate. Determining the reason for anomalies requires human brilliance at piecing together information fragments and clues across multiple disciplines. Note the prevalence of detective stories in popular culture; forensic and information technology tools provide facts, and people ultimately make the leaps of brilliance. Tools that allow broad inspection and intuitive exploration empower analysts beyond statistical or reporting packages.

What exactly do these new tools need to do?
Encourage shared understanding and communication.

Align the incentives of sites with that of sponsors and regulators.
The variety of clinical sites and their respective management styles and operating procedures is staggering. It's no wonder the market for EDC and CTMS is so vibrant and fragmented. No two offices or medical centers have the same set of incentives, constraints, or experience participating in clinical trials. However, some common aspects tie sites together at some level: the desire to get paid for participating in existing studies, the desire to build a reputation as a good performer, and in some cases the ability to lead adjacent studies or publish results.

Software tools that ignore these incentives are destined to be met with skepticism at sites, or worse jeopardize the study due to poor compliance. Instead of evoking images of Big Brother, tools used for remote site monitoring can involve sites collaboratively and enable them to derive benefits from the data they enter. Empowering sites to actively track their own performance against key metrics allows them to compete against other sites; strong performance can become a strong track record, which can lead to participation in future trials. Tools that encourage site investigator involvement and data perusal can also help identify and resolve issues more quickly than if these problems linger until the scheduled monitor site visits or team meetings.

Level the information playing field with better systems for site monitoring
The largest area for improvement, according to Eisenstein, is in the area of remote site monitoring, estimated at 21% of trial costs. Monitors have the difficult job of ensuring protocol compliance and data quality, so regular review of trial data is critical. Often their desired set of information can be captured in periodic standardized reports or data views that then encourage dialogue. Avoidable costs are incurred with prolonged travel and resolution of data inadequacies. Hastening issue resolution and providing monitors intuitive ways to navigate the data are important cost reduction methods. Information leveling must carefully adhere to data access privileges, as inappropriate access can "unblind" trial data and invalidate the study.

Streamline and Enhance Operations
But maintain reasonable expectations
Although Eisenstein recommends a leaner trial design, this approach is not broadly applicable, at least in the short term. Simpler protocols would certainly focus resources to only the most sensitive and specific tests needed, reducing the need for the large amount of ultimately unused data [3]. However, a leaner design assumes that the set of successful tests are known in advance, which is clearly not the case. The advent of adaptive trials seeks to hasten the identification of successful tests, but operationally adaptive trials are more complex, not more simple. So assuming the existing level of trial design complexity, at least in the short term, tools that allow more vigilance of protocol compliance and visibility into early data can help manage the complexity.

Incumbent vendors are poorly positioned and incentivized to deliver such solutions
Enterprise software lags consumer-directed software most severely in usability. That's tough to change, especially when vendors are scrambling for market share.
Clinical trial IT vendors have been entering the marketplace in hopes of establishing early dominance. Their largely web-based data acquisition software and either hosted or enterprise database management tools are not much different from applications seen in other industries. That's not necessarily a compliment. Clunky interfaces and unintuitive designs are unfortunately hallmarks of most modern software products [4]. When vendor focus is on gaining market share, a long list of features and functions is a great sales tool. Focusing on the interface utility and simplicity isn't as helpful when selling enterprise software. Half-hearted nods to "ease of use" and "interface redesigns" remarkably seem to satisfy. The documented result, though, is poor compliance and usage [2].

The reflex is to partner with successful industry giants known more commonly as enterprise IT providers. Simply boasting integration with these industry giants oftentimes makes things even more complicated for users. Instead of becoming experts in one tool, users must now learn several tools, most of which still suffer from their own legacy usability issues -- but at least these issues are now for someone else's industry to fix. Again, this strategy will ultimately fail; two inadequate tools do not become one good one.

So where can we to look for solutions?
Often where you least expect it
A major initiative among software firms is the rise of human-centered design principles in software design. Firms historically dominated by engineers and programmers are now recognizing that a different set of skills are required to create simple, beautiful, and intuitive interfaces. Engineers excel at problem solving and technical innovation, but visual and interaction designers bring right-brain balance to software development.

As mobile devices and websites have raised the standards for software capabilities and design, enterprise developers must follow suit. As we can see from the consumer marketplace, users desire direct manipulation of data and functions, not layers of menus. When given the chance, users are eager to compose their own spaces and widgets and collaborate almost constantly. A generation of workers is comfortable with basic analytics and visual thinking. Harnessing these urges toward productive endeavors requires designing interfaces and data models that allow creative exploration and personalization. A collaborative workspace based on information visualization is a natural extension of consumer behavior and allows digestion of massive volumes of information that currently overwhelm or confuse people.

Applying collaborative visualization to clinical trials
Empower the teams that need to know
What if it were possible to integrate all relevant data into a single environment where no queries were needed to select and view data? It didn't matter if you had a data warehouse, several databases, or a set of files that you held dear. Your data would always be up-to-date, and charts would change based on new or altered data. Instead of querying with scripts or searches, you could browse at different levels of granularity and directly select and move data among charts, tables, and reports. And what if you could share your live visualizations with others, who could annotate and manipulate the content? There would be premade visualizations to track your most important metrics and information concepts, but you were empowered to compose new charts yourself, without IT involvement. Users could create custom charts and post them to specific team members or to everyone involved in the study. Data security filters would prevent unauthorized access to either data or visualizations.

The use cases for such a system are many. Managers could track milestone progress, sponsors could monitor portfolio performance, site investigators could follow patient timelines and protocols, and data monitors could review data collected to date. Visualizations created for outlier detection or forecasting would always use current data. Tools and data sets not normally correlated could be viewed along shared dimensions; for example, financial data could be overlaid with enrollment events and visits to visualize pay-for-performance over time. Likewise, patient adverse events, concomitant medications, and visits could be overlaid against lab results. Incorporating trial planning and budgeting could help identify which sites are consuming the most resources but producing the least results. Protocol compliance charts identifying which study events are overdue and at risk of being noncompliant could call managers to action. Many problems facing trial teams are due to a lack of understandable information. When that barrier is removed, performance can and normally does improve.

Isn't this just Business Intelligence or Reporting?
Business intelligence is still too passive. Limited interaction and collaboration, coupled with gratuitous visual elements detracts from appeal and stifles adoption.
The Business intelligence (BI) software market has grown at over 10% per year in recent years [5]. The most common domains are sales tracking, marketing, and financial analysis. Penetration into the clinical trials space has been limited. It is unclear whether the reasons for traditional business intelligence's slow adoption in clinical trials are those outlined by visualization experts such as Stephen Few [6]. Recent industry consolidation has left a few large vendors with entrenched platforms along with many small emerging vendors with niche applications. Endemic though, according to Few and others, are useless charts and distracting visual features that tend to confuse users [6]. Further, chart creation is still only the realm of a privileged (cursed?) few; users still have limited facility to compose visualizations, especially with loosely related data. A system that empowers end-users to compose visualizations themselves would break down information silos and allow much faster and more effective analysis and decision making.

Conclusions
Gains from popular clinical trial software tools are achieving diminishing returns, and industry consolidation is shrinking the toolmakers' innovation pipelines. Connecting all clinical trials stakeholders through shared understanding of integrated data will help align their incentives and encourage collaboration and prompt issue resolution. Both cost avoidance and value creation can be derived from such a collaborative visualization system. Decreased travel costs, shorter reporting time, and lower query and issue resolution time are common outcomes. Users are empowered to create understandable, usable, and appealing visualizations to capture and share their ideas. Employing a collaborative visualization system leverages existing IT projects and assets to create an integrated data environment capable of growing with your portfolio while increasing your effectiveness and efficiency at managing clinical trials.

Learn more at www.gdviz.com or contact jason.neiss@gdc4s.com with comments or for more information.

References:
[1] Eric L Eisenstein et. al. "Sensible approaches for reducing clinical trial costs." Clinical Trials, February 2008; 5: 75 - 84.
[2] Robert Musterer. "CTMS Promise vs. Reality" www.clinpage.com/article/ctms_promise_vs_reality
[3] Kenneth Getz et. al. "Assessing the Impact of Protocol Design Changes on Clinical Trial Performance." American Journal of Therapeutics. September/October 2008, Volume 15, Issue 5
[4] Alan Cooper. The Inmates are Running the Asylum. 1999
[5] IDC. "Worldwide Business Intelligence Tools 2005 Vendor Shares." http://download.microsoft.com/download/0/7/6/0760A23B-FBAD-4446-AE12-2D03352BD3B8/Worldwide_Business_Intelligence_Tools2005VendorShares.pdf
[6] Stephen Few. "Visual Business Intelligence." www.perceptualedge.com/blog

Contact details:
Jason Neiss, Product Manager, Life Sciences
T: 412-432-2457, E: jason.neiss@gdc4s.com, W: www.gdviz.com