By Deborah Borfitz

July 10, 2019 | Opinions about the merits of purpose-built applications versus do-it-all platforms have seesawed over the years, but it doesn't have to be an either/or proposition, says Jae Chung, vice president of Oracle Global Health Sciences and founder of the recently acquired startup solution goBalto. Even five-star apps can't exist in silo, especially when it comes to the global, complex clinical trials process. He encourages life science companies to combine the best of both worlds through cross-vendor integration and workflow automation.

"It doesn't have to be all Oracle," Chung adds. "We have to be flexible to accommodate connectors into other technologies because we know that's the reality of the industry." This is a clear departure from the stance of other major vendors in the space, who are either "still caught in the mindset that it has to be their platform" or pitching a unified clinical platform that is not truly integrated.

"The onus should be on vendors to stitch together best-of-breed apps and build out a platform," says Chung. Their focus, up to now, has instead been on trying to be "everything for everybody."

Smarter Study Startup

Oracle Health Sciences Study Startup solutions have retained the goBalto brand name, which practically defines the niche. At last count, goBalto was directly or indirectly supporting over 500 life science companies, Chung says. Study startup emerged as a software category only about three years ago.

Difficulties in finding sites and patients who are well matched to study protocols, the root cause of the growing length and expense of conducting clinical trials, have existed for decades. "Endemic silos" in drug development are costing companies the institutional memory that could help solve this perennial problem, he says.

Technology is a convenient way to capture and curate that knowledge, but it is not the answer, Chung continues. Organizations first and foremost need to establish business processes that software tools can then help enable and support. "I am a big proponent of collaborative workflow automation that guides all stakeholders through the clinical trial process."

Must-have components of a robust workflow engine include clear lines of ownership responsibility for tasks and documents, the intelligence to ensure requirements are met, and the flexibility to support prioritization of activities and review cycles, Chung says. A user-friendly interface in managing those complexities for the study startup process is goBalto's "secret sauce," he adds.

The reality is that most companies continue to rely on spreadsheets and checklists, and the juggling skills of clinical research associates, to keep track of a sea of activities and documents across thousands of sites in dozens of countries—and so they repeat rather than repeal mistakes.

The only way to close the feedback loop, and improve performance, is for organizations to measure what they've done so they know where and how to do better, Chung says.

Oracle's goBalto study startup solutions gather data on how good sites are at finding patients and if they are activating as quickly as possible, which feeds a model that auto-generates insights for improved decision-making on subsequent trials, Chung explains. It's an exercise that bears repeating for every single study—not haphazardly, as is more typically the case.

The study startup database is the largest in the industry, comprising aggregated data on 150,000 sites, reports Chung. "We know exactly how long it takes to get these studies up and running by phase, therapeutic area, country, and organization type... in real time." Clients see an average 30% reduction in cycle times when running goBalto on a study, he notes.

Predictive to Prescriptive Analytics

Currently, most technology vendors offer status reporting with varying degrees of reliability, says Chung. Oracle's goBalto study startup solutions are predictive, giving users a sense of what can happen. But the Holy Grail is prescriptive analytics where the technology tells people what to do, much like GPS software directs people to optimal routes. "For this we need a tremendous amount of statistical data, and because we've been at this the longest, we have amassed the largest database of site operational metrics and are in a unique position to delivery meaningful insight."

Oracle is already applying predictive analytics to the study startup process to provide study managers with the "ideal critical path" for getting every country and site up and running, Chung says. In the "very near future," the company will market-test the solution’s first prescriptive analytic features. Users will be able to input data to get recommendations, including when to take specific steps to achieve startup even faster.

Better Enrollment Planning

Studies might also be streamlined by reducing the level of guesswork that currently goes into enrollment planning, according to Hrishikesh Kulkarni, principal statistician and customer experience manager at Cytel. The company's EnForeSys enrollment simulation software is an alternative to the typical "wing it" approach that is costing sponsors an average of $26,000 per patient—all because they are making a lot of unrealistic assumptions, such as all sites will be ready to start on the same date and that no sites will fail to enroll entirely.

The current client base for the self-service simulation tool is primarily mid- to large-sized pharmaceutical and biotech companies for their enterprise-wide study programs but is equally helpful to smaller companies with just one or two compounds in the pipeline where enrollment planning is even more critical, says Kulkarni. While EnForeSys is powered by computational models, it is designed for use by non-statisticians working in clinical operations and feasibility groups. Accordingly, the graphical user interface includes familiar terms such as "screen failure" and "seasonal deviations."

Users can model different what-if scenarios with EnForeSys to determine the probability of hitting enrollment targets, Kulkarni says, after accounting for variation in enrollment rates, site activations, screen failures, and site failures. The tool might tell them that there's a 70% probability of enrolling 1,000 participants 24 months after study startup, based on their current plan, he offers as an example. If the goal is at least 80%, the team can think of adding more sites in the trial to bump up enrollment speed.

EnForeSys is as much a communication as a planning tool, Kulkarni says. "[It] facilitates discussions within the team, to get agreement on what the inputs should be so that the eventual plan has mutual consensus."

The back story to the 2014 launch of EnForeSys was growing consensus in the published literature that around 31% of clinical trials fail to meet their enrollment goals, and about one-third of all publicly-funded studies have had their timelines extended because initial recruitment goals were missed, says Kulkarni. Enrollment is a longstanding problem with a lot of moving parts and uncertainty. Cytel specializes in employing statistics to quantify those variables—both of its co-founders are distinguished statisticians, he notes.

The Inputs

Enrollment planners have access to a lot of "extremely valuable" data that can be fed into sophisticated statistical models to address the difficult questions of when, and if, a study will hit its enrollment target and with what degree of probability, says Kulkarni. A minimum of five data inputs is required to run realistic simulations—participating countries, number of study sites, site activation time (expressed as a range), planned enrollment rate for individual countries and sites, and per-country enrollment cap—and the accuracy of predictions improves if more granular data is also available.

Inputs can often be found in an enrollment plan, he continues. But if no comparable trials have been conducted by the organization the data can alternatively be pulled from related historical trials in the same therapeutic area.

Updating predictions is also important, to get studies back on track if they start falling behind, Kulkarni adds. Study teams can consider mitigation strategies and simulate various options, such as adding sites in countries that are enrolling subjects at a faster rate.

Currently, EnForeSys has no crossover with the budgeting process but certainly contributes to it, Kulkarni says. But the next version of the software may provide the means to "optimize regional and site-level planning, select sites, and open and close them depending on performance [at interim review]."

Quick Calculations

When the original version 1.0 of EnForeSys was released, the focus was on running thousands of simulations to generate predictions that would get shared with multiple teams within an organization. That remains a "solid approach… [but] takes some time and requires more inputs," says Kulkarni.

With the current version 2.0 released earlier this year, users can do more quick calculations with a minimum of information, such as target number of study participants and the countries and sites envisioned in the top-level plan, Kulkarni says. Predictions made with this new "interactive planner," which uses a closed-form computation formula, might not be as accurate as predictions coming from simulations but get users in the ballpark.

Insights on the enrollment process get presented to study planners in simple plots and tables, Kulkarni says. "Users say this helps them in finalizing the inputs for their actual simulations."