Realizing Value from Technology

eClinical 2011: Remedies for the Clinical Trials Machine
The Consultant Susan J. Ward

By Kevin Davies

February 9, 2011 | Despite spending the better part of two decades in the United States, Susan J. Ward hasn’t lost her Mancunian accent. A former senior executive at Wyeth for many years, followed by spells at Millennium Pharmaceuticals and Infinity Pharmaceuticals, Ward has spent the past eight years translating her detailed appreciation of the R&D process for numerous clients, helping companies translate technology into products. “Ask any CIO in the drug industry what keeps him/her up at night, and it won’t be too many sentences before the ‘value’ issue comes up,” says Ward.

As clinical development is the most expensive and time-consuming component of bringing new drugs to patients—typically thought to average 6.7 years and $600 million or more—improving this process should create significant value. Notwithstanding the distinct ambivalence of some, the past decade has seen significant investment in IT to support clinical development as evidenced by the growth of IT budgets within biopharma, the mushrooming of new companies, and within the past 2-3 years, a substantial ramp-up in investment by technology giants, including Oracle (see p. 29), IBM, and Microsoft in this space.

All of which begs the question: Has value been realized in clinical development through investments in technology?

EDC: a Case Study

“When EDC [electronic data capture] was first pioneered by Phase Forward, the elapsed time from ‘last patient visit’ to ‘database lock’ for clinical trials in Phase 2 or 3 was typically about six months,” says Ward. “The entire process was manual. All data transfer from clinical site to the sponsor was via fax, requiring small armies of operators—usually middle-aged ladies—to enter each batch of faxed data twice!” Hardly the most efficient process, yet EDC did indeed solve the problem it was designed to fix—a more efficient and faster population of the clinical database. Ward says the biggest impact EDC has had is that dramatically fewer queries require resolution after the in-life phase of the trial is complete. Using EDC, last patient visit to database lock is routinely accomplished in 2-3 weeks—a tremendous value.

But it was more than a decade before a majority of clinical trials utilized EDC. While that was partly due to price, Ward says “it’s really the conservative nature of the industry that defines the pace of technology adoption—especially when talking about a technology that directly touches the clinical data.”

Realizing value required major changes in long-establish business processes. “Too often vendors compromise their revenue by grossly under-estimating the extent of change management their customers must make for their technology to succeed.”

Making a difference

Solve one bottleneck, and another one rears its ugly head. Even more problematic than getting data accurately into a quality database, Ward says patient enrollment is still the rate-limiting step in drug development.

“A not-so-secret secret in the industry is that barely a third of clinical trials complete on time and within budget,” says Ward. “In a large study, one is almost certain to end-up having sites with so few enrolled patients that the site cannot be included in assessing clinical efficacy. It’s not easy to discern ahead of time which investigator sites will become instrumental in delaying a study.”

Ward points to DecisionView’s clinical enrolment modeling capability, which uses algorithms developed from analysis of recruitment data from hundreds of clinical trials to analyze a sponsor’s initial patient recruitment, then predicts the operational outcome for the remainder of the study. Simulations make suggestions for keeping a trial on track, perhaps jettisoning some sites or bringing on new ones. Given the costs associated with drug development delays, reliable prediction and scenario analysis is a powerful tool, she says.

“Selecting the right investigator sites is much more challenging today than ten years ago,” says Ward. “There are more trials, more companies sponsoring trials, more drugs in development, and drug-naïve patients are harder to find. Companies and CROs are competing more for patients, and competing especially for the more for skilled and experienced clinical investigators.”

The importance of this challenge has spawned a spate of new start-ups, including Cliniworks, GoBalto, and PatientsLikeMe.

Another key issue is the problem of actually transferring data from one place to the next in a demonstrably compliant manner. “You’d think by now we might have solved that,” jokes Ward. While larger pharmas are addressing this problem through software consulting relationships, younger biotech companies and academic institutions are turning to open-source solutions (e.g. BioClinica; see page 15), and “out of the box” solutions such as the compliance suite from Virtify and the integrated clinical and regulatory applications suites from MaxisIT, a past Bio•IT World Best of Show winner.

The Next Decade

If the past decade was all about getting the data in, understanding what those data mean is the big challenge for the upcoming decade. Ward suggests the bottleneck is swinging from operational processes to the challenge of designing the right trial, finding the right type of analyses to discern meaningful patient-based differences, and selecting the right patients to benefit from a new drug in development.

Trial design is becoming increasingly important. “The biostats guys are getting to be ever more critical,” says Ward. She cites a group called Cognika, which provides a searchable database of the inclusion/exclusion criteria used in prior clinical trials, along with the investigators and outcomes associated with those criteria—taking the work out of searching the literature for pointers when designing a clinical trial.

“The medical literature is currently our best data source, but is not necessarily a great reflection of actual practice,” she says. “This is not terribly surprising; people don’t publish negative findings, so you have a skew in the database.” That’s why she likes what Phase Forward founder Paul Bleicher’s new company, Humedica, is doing by with its focus on real-world data gathered from patient medical records.

Moreover, Humedica’s technology captures the trajectory of clinical values for individuals over time. “Longitudinal analysis of patient data could be very valuable, I think—we know disease is dynamic, yet trial design and biomarker strategies in clinical development tend toward a static, snap-shot view of a patient. The ability to define patient cohorts based on longitudinal data could be revelatory.”

“We’re trying to prove biomarkers as surrogates for clinical end-points. We’re trying to segment patient populations by the molecular substrates underlying each patient’s manifestation of disease. The more you’re trying to leverage the genome, the more complicated it gets.” It can feel like one step forward, two steps back. Everyone is looking to get their feet into personalized medicine, but now you’ve collected these data in Phase II, what do you do with it? I think a lot of clinical folks are struggling with that.”

Technology can help, says Ward. When it comes to personalized medicine, “one needs to look at patients as individuals, even if eventually you’re going to put them into distinct cohorts.” She is impressed by the leadership at Roche. They understand that the relationships between data can be just as important as the clinical data itself—and it’s only when you analyze information on a patient-by-patient basis that one can see the extent to which relationships between data, for example pharmacodynamics or biomarker associations, are heterogeneous.

Ward predicts that simulation will become “a technology of choice in this coming decade.” Modeling and simulation in biopharma is still in its infancy, currently focused predominantly on projecting drug pharmacokinetics, or on optimizing dosages and number of patients in a clinical trial.

A prior client, Gene Network Sciences (see p. 50), uses highly sophisticated Bayesian modeling techniques and supercomputers to discern molecular pathways associated with clinical response (or non-response) from patient data—from which one may have a basis for patient segmentation, alternate biomarkers, or new insights for drug discovery.

Among Ward’s current clients is Biovista, which aims to predict biological
effects from molecular fingerprints, re-purposing new indications for compounds that failed first time around in the clinic. Biovista has a partnership with FDA for predicting adverse events, as well as research collaborations with several major pharma to identify new uses for their proprietary clinical stage compounds. •