By Mark D. Uehling
Senior Science Editor
August 13, 2002 | Six locked rooms in New York City. Each is filled with paper. The documents contain vital information about a single drug — all of the scientific and clinical data submitted to the FDA for its approval. Incredible as it seems, this hoard of perishable paper is maintained by a pharmaceutical company that is one of the world's largest and most sophisticated research organizations — except when it comes to managing its critical clinical trial data.
Now the kicker: This company isn't the only one that does this. They all do.
In Gaithersburg, Md., the FDA has an entire warehouse filled to the rafters with similar
|Clinical Trials: The Ultimate IT Challenge
|Every new drug approved by the FDA may take 15 years and $403 million to bring to market when capital costs are excluded. Seventy percent of that sum, or $282 million per drug, is spent on clinical trials — complex scientific experiments that digital and electronic technologies have only begun to streamline.
documents, all from pharmaceutical companies, all testifying to the industry's continuing reliance on paper. Routinely, semitrailer trucks pull up to unload a new drug application. For all the technology that is brought to bear on other stages of the drug research process, marshaling IT to help manage the analysis of drugs in human patients is a task so big, so hairy, that it has resisted the best efforts of software and hardware to tame it.
So it goes in the pharmaceutical and biotech industries, which have never managed information as effectively as they have chemical reagents and cell cultures. The hope, of course, is that change is ahead.
In the industry's defense, its main challenge — finding safe, effective medicines — is so monumental that data management has been an afterthought. The software for clinical trials may be the most neglected component of the life science IT portfolio simply because clinical trials are the final stage of the scientific process, following years of preliminary research on cells and animals.
Not surprisingly, the companies involved are not always comfortable discussing the topic. Novartis Pharmaceutical's Iso Moshfegh, global head of development for IT, is polite but circumspect. "We have a lot of homegrown software we use in this area," he says. He knows the problem well. Based in Switzerland, he's spent 22 years in IT. Some 25 IT professionals at his company are dedicated to clinical trials.
Moshfegh will allow that Novartis is making slow progress, moving information to a data warehouse. And he does say the company is beginning the process of deciding whether to adopt XML for clinical trials and other processes. But the pace is sluggish. "We have started the evaluation of how we should proceed, mainly in the data entry area," he says. "We haven't made up our mind whether it's XML or something else."
This slowness to adapt is crippling Big Pharma, suggests Bernard P. Wess Jr., president of Needham Heights, Mass.-based Perseid Software Ltd., who has survived designing large medical databases for Ford and Procter & Gamble.
"The orientation of senior management [at pharmaceutical companies] needs to be about the acceleration of drug discovery and clinical trial processes," says Wess. But "that has not been their orientation. The orientation toward the information systems part of it is rather backward compared to other industries."
Something has to change, says Wess, and the sooner the better. "We need to collect all of the clinical trial variables in a database so that a multidisciplinary, multisite trial is completely electronic," he says. "We need the data to be available for analysis, annotated and connected to the proteomic and genomic databases in real time with relational database connections."
No Simple Solution
The drug industry is not likely to get there soon. That's a result, at least in part, of the ungainly clinical trials themselves. They go on for years. They proliferate. According to the Tufts Center for the Study of Drug Development, a single drug requires an average of 37 separate clinical studies prior to approval. Just managing the data gathered from instruments is a headache. Most companies keep 20-year-old equipment in mothballs. Why? To fire it up if needed to read or regenerate an old graph to satisfy the FDA.
Clinical data comes from vast hospital labs, from nurses scrawling on checklists, and from patients themselves. Then there is input from physicists, chemists, biologists, and physicians. By default, paper simply emerged as the only viable technology to standardize around. Complicating matters, some of the
"The management of clinical data is generally not the No. 1 focus of the boardrooms and other people thinking about how to improve the whole process. It should be. There are tremendous advantages. Not only in timeliness, but in quality and cost and overall access to data."
—Paul Bleicher, Phase Forward
people involved in recording clinical trial data often have more pressing responsibilities. On any given day, other tasks in hospitals and doctors' offices — keeping patients alive, say — loom larger than filling out forms for a drug company.
So pharmaceutical companies have a limited ability to dictate broad technological change. Even if one site accepts new hardware or software, a key physician-investigator may balk at using it. Software developed in-house at the drug company may have its own proponents who resist change. The bottom line: There is no magic IT pill that can make all of the aspects of clinical trials work smoothly.
With so many obstacles to overcome, it's a little surprising that a few companies offering solutions have not only survived but prospered as the industry wakes up to the problem. Phase Forward Inc., of Waltham, Mass., is one example: It racked up $106 million in sales last year. Phase Forward's InForm and Clintrial software applications are two important tools for managing medical studies.
Phase Forward founder and CEO Paul Bleicher says that other pressures affecting the pharmaceutical industry — from Wall Street, consumers, and politicians — have distracted management. "The management of clinical data, as important as it is, is generally not the No. 1 focus of the boardrooms and other people thinking about how to improve the whole process," he says. "It should be. There are tremendous advantages. Not only in timeliness, but in quality and cost and overall access to data."
Phase Forward's software can run trials the traditional way, on paper, with data entry improvements—or entirely over the Internet. Paperless Internet-based clinical trials will probably prove as elusive as the paperless office. But unlike paper, the software can interrupt a nurse if, contrary to the rules of a clinical trial, he or she has just signed up a patient who is overweight, for example.
Sorting out the eligibility of one obese patient may seem mundane, but it is exactly the sort of chore that makes clinical trials take so long, with dozens, hundreds, or even thousands of patients in a study. Bleicher has built his company on removing some of the innumerable delays in collecting snippets of highly technical information over years. "There is a lag. I call it the data visibility gap," says Bleicher, a dermatologist and former medical professor at Harvard. "Is the data clean? Is it being collected properly? Do I need to stop the clinical trial because of a safety issue? You're making decisions on a process that has, on average, a three-month lag in it."
The Paperless Chase
Some companies are pushing harder to get rid of paper. Morrisville, N.C.-based etrials Inc., whose slogan
|Clinical Trial Software Providers
|A listing of clinical trial software providers.
is "Helping you leave paper behind," reports a 177 percent jump in 2002 second quarter revenues over the same quarter in the previous year. "Direct data entry at the point of patient care totally eliminates paper in the clinical trial process and probably in the doctor's office," promises company President John Cline. "Paper is not going away immediately. But over time, we will significantly decrease the amount of [it]."
Cline notes a strange disparity between drug companies' high-tech sophistication in some areas and lack thereof in others. "A lot of the industry is familiar with technology on the drug discovery side. Once those compounds are discovered and have gone through the preclinical phase, the whole system reverts back to a fairly archaic paper process."
Admittedly biased, Cline insists that savings from electronic trials are real. One recent job would have cost $451,000 to do in paper, he says, but etrials' bill was $261,000. "The savings are materializing, and [companies are] getting the data faster," he says.
Big IT companies are entering the clinical trials market, too. The best known example is database giant Oracle Corp., which offers Oracle Clinical, an application that grew out of the company's consulting work. Oracle Clinical components include the main application, which allows investigators to set up forms and generally manage the project, and modules that keep track of tricky technical vocabulary (a thesaurus/dictionary) and adverse events (serious complications or death). The cost: $15,000 per named user.
Business is good, says Keith Howells, Oracle's vice president for pharmaceutical applications. A few big pharmas (such as a Wyeth or a Pfizer) could have hundreds of Oracle Clinical users.
Howells explains that basic respect for the pharmaceutical industry's intricate workflow has caused IT vendors to incorporate more flexibility in pharmaceutical applications than they would in software for a Toyota factory, for instance. So Oracle Clinical software allows for both electronically and manually captured data.
"Ours is the only one where there is one system and you set it up identically for paper and electronic capture," says Howells, citing the conventional wisdom that 95 percent of all clinical trials are done on paper. "Everybody accesses the same information. That is critical from a productivity and control/validation point of view."
|The Least Automated Industry in the World
|Bernard Wess discusses how drug companies got into their predicament — and how to fix it
Howells disputes the idea that installing Oracle Clinical can be a long process. "If a client said, 'Put this protocol together as fast as you can, train our people, and set up the labs and batch feeds,' we could get it done in under a couple of months. The primary limiting factor is not our software," he says, "it is the size and culture of the company we're dealing with."
More typically, he concedes, Oracle Clinical projects take longer than a few months. Blame the workflow, he suggests. Vendors and consultants who try to work with pharmaceutical companies often wrestle with multiple layers of authority, he says. "You say, 'We've got this demographic thing we're trying to model, this field called 'Sex.' What are its valid values? Is it M and F? Zero and 1?' Whatever you pick, you are advised to stick with it until the end of time for all studies for all therapeutic areas worldwide. The big companies go into shock — 'We've got to organize a committee to decide this thing.'" Some life sciences companies need six months to choose such variables, he reports.
While it can be painful to set up Oracle Clinical the first time, the bright side is that once it's up and running, the next studies can begin almost instantly. "You go back a year later, and they say, 'I put a protocol together in a day because I got my standards set — standard demographics, all my edit checks, standard batch loads,'" Howells says.
He says that the disappearance of small vendors in the clinical trials arena has brought customers knocking. "There is a general feeling of menace in the air," he says, citing large FDA fines (which are part of the public record), and rumored shutdowns and future or ongoing FDA inspections (which are not). "We are seeing a conservative attitude. If a single trial is invalidated, let alone a whole project, let alone all of them, the cost would be staggering. Everyone is in extreme defensive mode at the moment."
The New Rules
As Howells suggests, it is hard to underestimate the looming burden of the FDA regulation known as 21 CFR Part 11 (see right), which officially pertains only to electronic signatures. Part 11, as the rule
|Decoding The FDA's Title 21 CFR Part 11
|Understanding and obeying 21CFR Part 11 is becoming a cottage industry, one that poses difficulties, costs, and risks that are both organizational and technological.
is called, is another Y2K-type problem for the pharmaceutical industry. It touches nearly every piece of software and hardware connected to clinical trials.
Under Part 11, when a drug company submits an eighteen-wheeler full of paper to the FDA, it must be able to prove that when an apparently 400-pound patient is allowed to stay in a trial, it was because his actual weight was 290 pounds, but wrongly recorded initially. More importantly, it must prove that the health-care, IT and statistical professionals who discovered, investigated, and corrected the error did so properly.
Failure to document such changes to the database, however trivial they might seem, can result in six-figure fines or the shutting down of a pharmaceutical production line. To its credit, the FDA wants to know that the patient really belonged in the study, and that the medicine really had the effect the company reported. Tracking such minutiae fastidiously is the only way the government can ensure that drugs work as advertised. So all the paperwork must be in order.
|Pushing EDC into Overdrive
|Electronic data capture forces patients to record data in a timely manner — not in the parking lot before their next appointment.
Thus, much like the Y2K phenomenon in the late 1990s, Part 11 is becoming an industry unto itself and having a pervasive effect on the relevant IT tools. NetRegulus Inc. is just one of many software and services companies specializing in the field. Lisa Ensign, director of biostatistic and clinical studies for the Oakbrook Terrace, Ill.-based firm, notes that there can be sharply different appraisals of Part 11 solutions by those who pay for them (the drug companies or universities) and those who live with them (staff in hospitals and doctors' offices).
Ensign says the vaunted productivity gains from electronic approaches to clinical trials may not be materializing, at least not so far. "The people in the trenches have not seen a benefit. It hasn't eased their workloads yet," Ensign says. "Fifty to 60 percent of companies say they intend to move into [electronic trials]. It's just not happening as quickly as people expected."
Some clients come to NetRegulus after brushes with the FDA. "We had a recent situation [in which] a client had an FDA audit, and they were asked whether they could change a record. Yes, they could. They were asked to show how that worked, and they did. And the auditor said, 'Can you show the audit trail that allowed that change?' But they couldn't show the fundamental audit trail of who made the change and when they made it. The audit trail needs to be automatic, and it needs to be unavailable to the person making the change."
In some cases, Ensign says, there can be hard-to-quantify returns from an investment in better IT. These may emerge in Phase IV of a clinical trial — the period after which a drug is available to any member of the public by prescription, not just small numbers of carefully chosen human guinea pigs during stages I through III.
Consider, says Ensign, the hypothetical instance of a company that makes pacemakers and glucose monitors. The company might have a hard time noticing problems in small numbers of patients if the information is being tracked on paper. "If you see a problem that occurs once or twice, it may or may not raise a red flag on its own. But if you can see a problem across all products, that can be very valuable knowledge, [leading you to realize,] 'Ah, what we really have is a battery problem.'"
IT veterans report that some of the pressure driving adoption of better IT for the clinical trials process is coming not from the FDA but from Big Pharma's fear of litigation. The cost of data that merely appear to be mishandled can be big enough to dent the bottom lines of large companies.
|Decoding The FDA's Title 21 CFR Part 11
|Understanding and obeying 21CFR Part 11 is becoming a cottage industry, one that poses difficulties, costs, and risks that are both organizational and technological.
A case in point is the withdrawal of the diet-drug duo Fen-Phen, which is believed to have killed 31 patients among the millions it helped. The first clues to the problem came not from the companies that made the drugs but from the Mayo Clinic. A lag in acting on the Fen-Phen information helped patients' lawyers collect a $3.75 billion class-action settlement from American Home Products (now known as Wyeth). Could better management of clinical trial data have helped detect the problem sooner?
Bayer, to take another case, pulled Baycol off the market in August 2001. At least 100 people are now estimated to have died from the cholesterol-lowering drug. Bayer logged $586 million in Baycol sales in 2000, and losing those revenues prompted the company to cut 4,000 jobs. Hundreds of Baycol lawsuits are now pending.
In both the Baycol and Fen-Phen cases, the drug companies were not simply criticized for making drugs with infrequent but grave side effects. Instead, the legal questions were, what did the drug companies know, when did they know it, and what did they do about it? In contrast to Phases I through III of clinical trials, data from Phase IV are more public. They can flow to the FDA and to the companies themselves, and be used against drug companies in court.
Thus, software from NetRegulus, Phase Forward, Oracle and other companies that tracks adverse events poses a corporate dilemma. Use the software, and overcautious managers might pull lucrative drugs from the market prematurely. Don't use such software, and risk horrendous publicity and lawsuits from not identifying and preventing problems sooner.
Ironically, there may be one constituency in clinical trials more ominous in the minds of pharma executives than lawyers. They are statisticians. If the math wizards rule that a new blood pressure medicine is not in fact lowering anyone's blood pressure, millions of dollars in research are written off. If they say the numbers add up, on the other hand, those figures become ammunition for marketing campaigns.
IT is affecting even this segment of clinical trials. Statistician Elizabeth Yetisir, a Canadian consultant to pharmas, argues that it can be tricky to figure out what happened in a clinic months or years ago based on information in a PDA or electronic data capture device.
Was the patient's hemoglobin really 128? Or 48? Was that a normal value or out of range? Yetisir concedes new technologies can be programmed to ask users to confirm their entries, but says, "it's just a little easier to make a mistake" with the PDAs. "Most people can put what they mean on paper."
She says the inherent caution bred into pharmaceutical companies is part of the reason the transition
|ImClone, Erbitux, and Data Management
|Fiasco illustrates that eliciting conclusive proof from clinical trials is difficult
from paper may be occurring slowly. "Every industry has its own level of inertia. It's hard to move from one technology to another. You don't want to jump into something and then find out it's really not giving you the security you were after."
Security, an important consideration, is built into most software in the clinical trials space. But life science companies unwilling to tear out their existing applications may elect to protect documents by inserting encryption technology between the user and the application.
Alchemedia Technologies Inc., for example, has a server-based encryption program, Mirage, that is Part 11-compliant. Its ability to create and decrypt PDF documents on the fly is unique, says Buddy Wilson, senior vice president for the Grapevine, Texas-based firm. An upgrade last year gave Mirage the ability to prevent copying and printing even while the information is displayed on a screen.
Some of the new tools for managing clinical trial data are adaptations of existing products. For example, eRoom, from eRoom Technology Inc., is a collaboration tool for any industry. Michael Artinger, director of the pharmaceutical practice at the Cambridge, Mass.-based company, says Pfizer and other pharmas are using eRoom to expand collaboration and inform top execs. "Senior management needs to see what's happening, where the bottlenecks are, what the status issues are, so they can base decisions on current information."
One customer, Aventis, agrees. Peter Loupos, the company's vice president of drug innovation and approval, praises eRoom's ease of use and expects the software to promote cooperation between scientists inside and outside the company.
"They begin to collaborate more quickly than with phone, fax, or e-mail," Loupos says. "It is a driver for cultural change." Aventis, he notes, is hoping to have the first truly networked drug discovery effort, and to move beyond the balkanized specialties of traditional R&D. Loupos says eRoom doesn't try to do too much. "Upfront, we define what it is and what it is not."
Aventis may have been fortunate — it was able identify one discrete IT problem and select the appropriate tool. Other pharmaceutical companies may well face a more wrenching predicament. When can their existing IT assets be tweaked to make clinical trials run more efficiently? If minor adjustments don't yield results, larger changes loom. The challenge then will be to reconsider the entire drug development cycle, and to design a foundation of IT that supports a modern and less paper-dependent process.
PHOTOGRAPH BY STEPHEN WEBSTER