Health Decisions Celebrates 20 Years of Adaptive Clinical Research

By Kevin Davies

May 18, 2009 | In building a company of some 150 employees over the past two decades, Health Decisions founder and CEO Michael Rosenberg has presumably done something right. But he maintains the industry is still doing a lot wrong. “The drug industry’s remarkable success has made it reluctant to change, but some changes are long overdue,” says Rosenberg .

The Durham, N.C.-based CRO espouses the concept of “agile clinical research,” an adaptive system that allows mid-trial improvements to both operations and design that  can help sponsors get drugs and diagnostics to market faster and cheaper. Rosenberg says his firm’s success is predicated on technology as a platform. “You can’t manage effectively without timely information, and you can’t get that information without the kind of technology we  use. Clinical research has not taken advantage of technologies that other industries have been using for years. When you look at the economic difficulties the industry is having, I think that technology as an enabler is way, way overdue,” says Rosenberg.

“The drug industry tries to manage complex, clinical operations without timely information because it doesn’t collect and process information fast enough,” Rosenberg sighs. “The industry hasn’t seen a need to manage trials based on near-real-time information.” But that provides “a great opportunity to improve efficiency,” says Rosenberg. “Efficiency is going to determine who leads the industry into the next era, not megamergers and off shoring.”

‘Agile’ clinical research encompasses design adaptations, operational adaptations, and underlying technology that make continuous adaptation possible—changing in response to changing information. (That philosophy will be detailed in Rosenberg’s forthcoming book—Adaptive Research the Agile Way: Optimizing Efficiency in Clinical Development—to be published by Wiley later this year.)

Rosenberg recalls a recent interview with a big pharma executive who admitted his firm was so poor at early development that it was shifting strategy and buying drugs developed by other companies. “I thought, well, that’s a very striking statement. Outsourcing early drug development doesn’t reduce risks—it delays risks and raises the stakes. It makes more sense to develop your own drugs, but learn how to do it more efficiently. It’s pretty sad when a major company thinks it’s incapable of becoming efficient.”

Rosenberg believes fast, accurate data capture is essential for running clinical studies efficiently. Web-based EDC is not up to the task, Rosenberg says. Health Decisions placed its first remote EDC system more than 15 years ago, but there were problems, particularly around data entry. “If you look today at the problems in web-based EDC, it’s exactly the same problem. Somebody’s got to sit down at a keyboard and enter data. Until they have the time and inclination to do that tiresome chore, data sits around in batches. That delays finding out about problems. Meanwhile, the same problems recur again and again.” Moreover, web-based EDC systems don’t collect performance metrics. “You need more than the patient data—you need timely information on the status of all key operations,” he says.

A major reason the industry is so expensive and error prone is that it recruits about 85% of studies late. Why is that? “The problem with web EDC is, it doesn’t tell you which recruitment strategy is working and which isn’t, or what’s responsible for screen failures, or how different inclusion/exclusion criteria are affecting recruitment. There’s no quick way to find out if I’m spending money well on recruitment strategies. Our system tracks information that lets you manage recruitment efficiently.” 

Waste Not
“Typical clinical studies waste a tremendous amount of time and money,” Rosenberg continues. “About 1/3 study budgets are spent redoing stuff that wasn’t done right the first time.” He says the clearest indicator of waste in the industry is the query rate. When a site submits data, Health Decisions runs the data through a validation engine. Sometimes, the site receives a note saying, we didn’t get a response for this question, could they clarify what it is? Queries are necessary to ensure accuracy, Rosenberg says, but that does not mean study managers should be complacent.

“The query ratio for web EDC systems is usually 5-7 queries per 100 fields. We looked at this and said, we need to make it easier for sites to enter data. But once you’ve submitted the data, we want to identify the problems right away…  You don’t want sites to make the same inadvertent or dumb mistakes while waiting for queries to come back. You don’t want to find out six months later that ambiguous instructions for the CRF were causing the errors.”

Rosenberg sees parallels between the state of clinical studies and the downfall of the U.S. car industry 50 years ago, when assembly line workers counted on fixing mistakes at the end of the line. “Toyota stopped the assembly line and fixed problems immediately, diagnosed the cause, and fixed that, too,” says Rosenberg. “The institutionalized query system with all its delays is like the rework area in an obsolete car factory. You expect queries, and clinical studies focus on processing queries, but there’s not much urgency about it. Queries often get resolved when regularly scheduled site visits roll around, not immediately. Meanwhile, errors and queries proliferate.”

Rosenberg says the most important metric is, what’s the interval between when a data point is generated and when the CRO gains access to it in house? “In typical studies, there’s a very long feedback loop. We know from other industries, if you want people to decrease their frequency of errors, you have to provide quick feedback.” This is where machine-read entry and the digital pen (a Health Decisions hallmark) come in handy.

Digital Penmanship
There are usually 1-2 digital pens per site. The digital pen writes on a piece of special paper that has a very fine grid pattern, which orients the pen as to which form it is and where it is on the form. Once the pen is docked, the data are transmitted to Health Decisions, much like a PDF. Upon arrival, it is machine read into the system, where it can undergo immediate validation. “When I’ve written your name, I dock the pen,” says Rosenberg. “If I’m collecting data in Croatia, and its 4 p.m. there and somebody’s interviewing a patient, I can have this information on your desktop in digested form before the patient walks out the door.” The system is poised to validate information, issue queries, and update operational performance metrics as soon as the data comes in.

If the system works, one expects to see a lower query rate, and that’s exactly what Rosenberg delivers. The query rate is 1 per 100 fields, even lower in bigger studies. According to a commentary last year in Science, each query costs about $350. “That’s pretty reasonable when you consider all the costs,” says Rosenberg. But “take that $350 and multiply by 10,000 queries over the life of a study…. You’re wasting a couple of million dollars.”

He adds: “We can tell where our recruitment dollars are best spent. We know why patients come in… If someone doesn’t show up for an appointment, we jump on that.” By contrast, he says, “Phase Forward and Medidata grew up around the data technology nerds and they’re not focused around the users of the system… I think the technology world that we work in misses 90% of the important stuff that’s helpful to shortening timelines and reducing costs.” A new release of Health Decisions includes web 2.0 capabilities to enhance collaboration between sites.

Just as Apple puts things together by thinking about the user and making their life easy, Rosenberg says his priority is “making the user’s life dead easy. You get fewer data entry errors and delays if you make data entry really easy—as easy as writing with an ordinary pen.” Could he be considered the Steve Jobs of the clinical world, I ask? “That’s probably a stretch,” he laughs.

The Health Decisions central system integrates all the components in a clinical study: data collection, adaptive randomization, site payments, administrative functions, and a reporting suite. “We’ve been using dedicated study websites for reporting and coordinating for 7-10 years,” says Rosenberg. “I assumed everyone was doing this. But almost nobody else does that. From a technology standpoint, this is a no brainer. It’s the quickest, easiest, most efficient way to distribute information.”

David and Goliath
Health Decisions recently finished a sizeable study of 4,000 patients for a start-up company that was battling a big pharma. “This was a David v. Goliath story,” says Rosenberg, who says the CEO of his client said: “You did for $7 million what Roche failed to do for $20 million.” “We catapulted them ahead of Roche,” says Rosenberg. “This [diagnostic] product was just approved by the FDA.”
Health Decisions was able to do everything quickly: recruitment, data entry, and study turnaround. “The FDA audited this study, and normally they’d come out to the CRO. They were comfortable enough with the pen that we could provide everything online. I believe this sets a new benchmark. We did this without a single 483 [FDA notice of a discrepancy].”

At a recent CHI conference, one IT staffer from a pharma company was quoted as saying: “All EDC vendors lie… None can do adaptive trials.” Rosenberg begs to differ. Almost every study Health Decisions does is adaptive. “Some types of EDC can do adaptive studies, but they have to be able to collect, transmit, and validate information quickly, and to generate timely metrics on study progress and operational efficiency… The digital pen and our systems make that routine.”

Rosenberg says his company is “ a little unusual for a CRO. We work with a lot of companies to consult on the design. Most companies are just beginning to dip their toe in this stuff.”