- Home
- Research & development
- Marathons magazine
- Issue 3
- Make mine espresso and the blood-pressure readings: A new way in
data management
Make mine espresso and the blood-pressure readings: A new way in data management
“Come to Starbucks,” Ted Chin said. It was two years ago. “All I’m asking is that you show up, ask questions, and pay for the coffee.”
Four executives in GlaxoSmithKline R&D got that invitation, one at a time. Arriving at a Starbucks cafe in suburban Philadelphia, a cafe with wireless internet access, each found Chin and two of his colleagues next to a table on which they had arrayed several laptop computers. Soon Chin would be showing the executive clinical data just arrived in databases from study sites around the world—data transmitted to GSK electronically rather than on paper. So did the coffee buyer begin to see the future as it was fast becoming the present.
“In the paper world, we put a lot of slack in the system,” Chin says. “In the e-world, the slack is gone. We can work with data as we get it in real time wherever we are, even Starbucks.”
This conversion from paper to what is called electronic data capture points up a crucial if often overlooked part of pharmaceutical R&D. Yes, laboratory robotics are wondrous, and genetic findings are revelatory. But no medicine gets to patients unless some people in an R&D organization devote themselves to managing its myriad interdependent parts. Call it project management. Call it continuous improvement. Call it sweating the details.
“R&D runs around the world and around the clock,” observes Amber Salzman, a mathematician who heads development operations at GSK. “It needs constant and close attention, attention not only to what we are doing today but to how we can do it better tomorrow.”
Consider clinical activities alone. In any given year, GSK will start hundreds of clinical studies enrolling tens of thousands of people, all this activity coming in addition to studies continuing from prior years.* Thousands of physicians will participate from their respective hospitals and clinics in scores of countries. From this work there flows a torrent of data ultimately serving to support a product-licensing application to regulatory authorities.
Time was when such an application, submitted on paper in multiple copies, would run to millions of pages. It moved by truck. Today GSK generally makes such submissions to regulators electronically, doubtless preserving a forest somewhere. Even so, a single submission remains formidable, swelling to up to eight gigabytes and embedding some 50,000 hyperlinks that cross-reference texts, charts, graphs, and other sources of information. Increasingly, submissions are global: Last year one for a GSK cancer drug was first translated into 24 languages.
The switch to electronic data capture (EDC), then, serves as one example of a ceaseless effort to better manage complexity. It accelerates data management between the conduct of a clinical study and the analysis leading to final preparation of the regulatory submission.
A little background: Physicians and nurses have typically recorded medical information about every patient on a paper chart, whether or not the patient is enrolled in a clinical study. They generally still do, though charting by computer is making some headway. What is changing is the next step. In the past, the investigators would transcribe data of specific relevance to the trial from the medical chart onto still other paper documentation, called a case-report form.** GSK representatives would then fan out to the trial sites to collect these forms and bring them back to GSK sites or have them sent back by overnight mail. Ultimately, GSK staff at GSK sites would review the forms and then essentially replicate them by entering the same information into databases for subsequent analysis.
Traffic lights for data flows: The InForm™ software available to clinical-trial investigators indicates the status of data that must be entered into a database for analysis of trial results. The numbers in the columns here identify individual trial sites and patients. White lights indicate that none of the required data relating to a visit have yet been entered; yellow, that some of the data have been entered, but not all; green, that all the data have been entered, and correctly—and red, that some data may have been entered incorrectly. Red lights serve as signals to investigators working with GlaxoSmithKline that they must review medical charts to confirm that data have been transcribed into the database correctly. Although patients can be identified by number, their personal identity is not reported to the company.
By contrast, investigators making use of EDC can enter trial-relevant data into those databases directly. There need be no manual collection by GSK representatives, no mailing, no time lost to redundant data entry at GSK sites. What once took weeks takes minutes.
Truth to tell, EDC had been a promising technology for some time before anyone really committed to it. When the technology first became available, about a decade ago, few investigative sites in hospitals or clinics had the computers and software required. It would cost them. Pharmaceutical companies also proceeded only cautiously. Force of habit and the sheer magnitude of the clinical-trial enterprise combined to favor the security of the status quo. At GSK as elsewhere, the new way was slow to take hold at first, notwithstanding several pilot studies. Ultimately there came a mandate from the most senior levels in R&D. The organization would embrace EDC—in stages, to be sure, but each stage would lead to the next, and the exceptions would be rare.
Chin explains: “Management had this vision that we limit the number of exceptions and say, ‘We’re doing this, and damn the exceptions. You want exceptions? The hurdles will be very high.’ So the organization looked at that and said, ‘My God, they’re really committed. We’re not doing pilots anymore.”
Chin holds one degree in pharmacology but another in business administration, and he has found his joy as a business-process fix-it man. To kick-start the change, he returned to the US from stints in Japan and Singapore.
“I lured him back,” remembers Frank Rockhold, a statistician who had worked with Chin previously and who now held executive sponsorship of the EDC project. “I told him this was his chance to be closer to his kids in college in the US.”
Chin recoiled at suggestions that would lead to needless bureaucracy. Should he assemble a large group to manage directly? No, he would have no direct reports. Might there be a formal committee, perhaps 20 or so people? No, not that, but a select group—seven people, as it turned out—who represented departments already working on clinical studies and who would bring to the project the enthusiasm of converts. “Keep it small, keep it agile,” he says. “You can tell a high-performing team just by walking in the room. You can sense the emotion, the passion.”
Training people within GSK was the smaller part of the task. They were far outnumbered by people to be trained at clinical-trial sites, some 15,000 of them. Often this training could be done on-line, as so-called e-learning. Additional help came during site visits by GSK’s own clinical-trial monitors and by representatives of Phase Forward, the Boston-area company that had developed the EDC software.
“GSK foresaw the benefits of EDC for its business and for investigators and patients,” says Steve Powell, a senior executive at Phase Forward, “and it carried out the fastest, large-scale adoption of EDC we have seen across the pharmaceutical industry.”
The work began in earnest in mid-2004. Within a year, post-training validation of readiness to adopt EDC was completed at 8,700 sites in 51 countries. “We had insanely aggressive targets,” observes Chin, beaming. If a trial site lacked the computer essentials, GSK provided them at GSK’s cost.
Aside from being a means of data transmission, the Phase Forward software, called InForm™, serves as both a to-do list and a flag to errors. With a click on the computer keyboard, investigators can call up the list of tests to be done and observations to be made during a patient’s visit: heights and weights, blood-cell counts, respiratory volumes, quality-of-life scores, and so on as needed for any particular study.
Then, the visit ended, the software displays traffic-light-like icons, the “light” of each icon indicating the status of data entry. A red light signals what might be a mistake. The light turns red if, for instance, an investigator has entered a blood-pressure reading outside the expected range for the study. The investigator then knows to double-check the entry against the original medical record and correct it if it is wrong; if it isn’t, the software provides a way to so indicate, and the light turns to green.
Such small enhancements have big ramifications. Recall how staff at GSK inputted data in the past: They were working from the paper case-report forms that had been delivered to them, not the original medical charts, which must remain with the physicians and nurses rendering care. Even previously these GSK staff, as they inputted the data, were themselves using software that flagged possible errors; however, to look into any questionable detail, they had to re-establish contact with the clinical-trial site. That meant catching busy investigators and asking them to take the time to check apparent errors. Hence follow-up phone calls, or faxes, or visits by GSK trial monitors, and possibly still more paperwork to create an audit trial justifying any change in documentation. The chase could go on for weeks.
“What EDC helps us all to do is to get it right the first time,” says Laura Zupko, a biomedical engineer who now concentrates on the uses of information technology in data management. “We are minimizing delays, delays in getting new medicines to patients.” GSK has, in fact, reduced by nearly two months the median period for “clinical-study close-out,” that being the time to receive and interrogate data after the final visit of any patient to a study site.
By now 350 GSK studies have reported data through InForm, and nearly all sizeable studies will do so going forward. The roll-out is truly global: An investigator in, say, Sydney, may enter data that is accessed from a server near San Francisco by a technical writer in Mumbai who provides a report to a statistician on the outskirts of London. The main point is speed—speed with no compromise of accuracy—not direct savings, but there are those as well, £2 million annually in printing, scanning, and other data-entry costs.
The job is not done. Still to be integrated into EDC are data originated outside routine trial sites, genetic data, for example, or PET scans. Nor have the potential advantages been exhausted. The sooner clinical data on a drug under trial arrive for analysis, the sooner signals of unexpected side effects or surprising benefits might be spotted.*** This acceleration could then enable an early halt or course change in the study. Indeed, EDC is gaining favor just as pharmaceutical companies think more about “adaptive trial design,” that is, design that allows the objective of a still-running trial to adapt to what is learned from the data as they stream in.
Ted Chin, for his part, has moved on to other tasks, all different yet all the same in that they fine-tune the dynamic forces in R&D. “Things in our industry are moving so fast,” he says, “that the landscape is always changing.”
###
*All these figures concern therapeutics only. A large clinical program is also run by GSK Biologicals, which devotes most of its studies to vaccines. One example: A single study enrolled 60,225 infants to evaluate a vaccine designed to prevent rotavirus gastroenteritis, which is estimated to cause 600,000 deaths annually. The vaccine is now available in more than 100 countries.
**The identity of the patient, however, is not transcribed and is kept confidential. Only a number is associated with any patient in records transmitted to GSK. If any finding in the data concerning specific patients needs to be communicated to them for their own welfare, GSK notifies the treating physicians, who then deal with the patients directly.
***Any early looks at data would be taken within the trial protocol. Typically in large studies an independent monitoring board is established with authority to review data at specific points during the progress of a study in the interest of patient safety. Any early review of data by GSK is prohibited or carefully constrained to avoid bias during the conduct of the study.
