Getting SMART: Drug Review in the Computer Age

The Food and Drug Administration is in the early stages of launching its drug review and approval system into cyberspace. This will transform a process that has traditionally involved laborious hands-on scrutiny of mountains of paper into a computerized system able to accomplish in an instant what now might take hours or days or weeks of tedious checking and rechecking.

And the system will work not just faster, but better. Electronic data handling will let agency reviewers of drugs and biological products spot problems, such as unexpected side effects or subtle variations in drug metabolism, that might be buried in the millions of bits of information that tell whether a drug merits approval or rejection. It will allow rapid comparison of a new investigational drug with other drugs of the same type or other drugs used to treat the same health problem. The result: more complete understanding of how a new drug or biological product works, its benefits, and potential role in health care.

This electronic transformation in the making is called SMART, Submission Management and Review Tracking. Full SMART implementation is several years away, but elements of the plan are already in place and more will be added in coming months. Some of the foundation blocks for SMART were in fact laid down more than a decade ago.

The Paper Chase

It helps to have a general understanding of how FDA's drug review and approval process works to grasp what SMART is and what it promises. (See "From Test Tube to Patient: New Drug Development in the United States," an FDA Consumer Special Report published in January 1995.)

To seek FDA marketing approval for a new drug, the drug's sponsor, usually a pharmaceutical company, submits a new drug application (NDA) consisting of voluminous, detailed reports on the drug's properties, development, manufacture, and testing results. Similar applications are required for biological products, animal drugs, certain medical devices, and food additives.

The drug sponsor provides a summary of this mass of data, but FDA requires that its reviewing scientists have access to the primary information on which the NDA is based. Consequently, an NDA comprises tens of thousands of pages of text, charts, tables, graphs, case reports, and clinical notes--the complete history of a drug from discovery and the first studies in animals to advanced clinical testing, from how it will be manufactured and packaged to how it will be described and promoted to health professionals and the public.

This means that paper NDAs don't arrive at FDA in envelopes; they come in trucks. But over the last decade or so, NDAs have begun to arrive at FDA on computer tapes and disks. With the agency's encouragement, drug sponsors now furnish some or all of the information contained in an NDA in a format that can be read and processed by computer. The companies also often lend FDA the computer hardware that agency scientists need to process these computer-assisted new drug applications (CANDAs). At the end of the review, FDA keeps the computer disks and the sponsor takes back the hardware.

CANDAs, and computer-assisted product licensing applications submitted for vaccines and other biological products, shorten review time by reducing the need to sift through reams of paper to get the answers to reviewers' questions. Ordinarily, if a medical reviewer has a question about a specific patient in a drug trial, the reviewer sends a written request for the patient's record to the central document room where NDA files are kept. It could take a day or more to get the information. Moreover, reviewing scientists often go back to the drug sponsor for clarification or reworking of statistical data, which can delay the process for days or weeks. CANDAs eliminate some of this delay.

Between 1991 and 1994, CANDAs were approved about six months faster than traditional paper applications--in an average 18.4 months compared with about 24.6 months. FDA hopes that by the end of this year all new drug applications will contain data that can be processed by computer.

The Leap into Cyberspace

But CANDAs are a far cry from state-of-the-art information technology. For one thing, each CANDA is a one-of-a-kind production. The sponsor assembles NDA information, organizes it on computer disks, and literally wheels the hardware and software into the FDA reviewer's office. Electronic information handling speeds the review for that specific drug, but not for any other. The next CANDA to arrive may use a totally different data processing system employing different hardware and software.

It's not unusual in fact for a drug reviewer's office to house several computer systems, each devoted exclusively to a separate new drug under review. Until fairly recently, each member of the team reviewing a computerized application--typically as many as four people--would get a stand-alone computer system with identical information and specialized software tailored to work with that specific CANDA. But now local area computer networks in FDA's Center for Drug Evaluation and Research allow CANDAs to be stored centrally and shared by all team members.

Though an exciting step forward, networking--and indeed CANDAs themselves--barely hint at what will be possible when the drug review and approval process is reengineered to take full advantage of advanced information technology.

When SMART is fully up and running, all marketing applications will be reviewed using a single, unified data processing system. Agency reviewers will be able to access the system, retrieve information in response to specific inquiries, perform statistical and other kinds of analyses, confer with other reviewers, communicate with drug sponsors--in short, carry out all the tasks involved in the review process with greater speed and accuracy than is possible without the aid of a fully integrated electronic data processing system.

Drug sponsors won't have to redesign their information systems, some of which are considered valuable trade secrets by the companies that use them. Instead, FDA will receive an electronic NDA submission from the sponsor, ensure its compatibility with SMART, verify that no errors have occurred in the electronic acceptance, and electronically file the NDA, allowing agency scientists to begin the review.

Although details of the system remain to be worked out, it's likely that each reviewer's desktop computer will have a set of "tools"--computer programs that allow the reviewer to carry out data searches, create and manipulate spreadsheets, produce tables and graphs, and even display pictures of clinical effects, such as wound healing. (See accompanying article.)

Eventually, as more and more information about investigational and approved drugs enters the system, reviewers will be able to compare the information in a pending NDA with data on other drugs, an enormously time-consuming task when an NDA consists of scores or hundreds of bound volumes that can only be cross-checked by hand.

FDA estimates that reviewers working on paper NDA submissions currently spend much of their time doing routine clerical work, which lessens the amount of time they can devote to their primary task--assessing drug safety and effectiveness. SMART will completely change this pattern, making much better use of the time and skills of reviewing physicians, pharmacologists, chemists, microbiologists, statisticians, and other highly trained scientists. The time saving is, in fact, one of the primary goals of SMART.

A Commitment to Do Better

In 1992 when Congress passed the Prescription Drug User Fee Act, FDA committed itself to completing action on "standard" drug applications in 12 months and on "priority" applications in six months. Applications for drugs similar to products already on the market are termed standard; priority applications are those for drugs believed to offer significant advances over existing treatments.

User fees paid by drug and biological product companies help cover the cost of reviewing new drug and biologic submissions and inspecting sponsors' testing and manufacturing facilities. Expected to total more than $325 million by September 1997, user fees enable FDA to speed the review process by hiring more reviewing scientists and physicians. But agency managers are also counting on the computerization initiative to further reduce review times

A second force driving the SMART program is the trend toward international harmonization of drug development and regulation. Because almost all major drug companies operate on a global scale, they generally support efforts to adopt regulatory systems compatible among many nations.

SMART supports a basic framework for information systems standards usable by regulatory authorities in Japan, Europe, the Western Hemisphere, and elsewhere. Regulatory requirements will almost certainly differ from one country to the next, but drug industry officials say they welcome progress toward a single package of marketing approval information that would meet the varying needs of multiple regulatory systems.

Technical Challenges

Much of what is needed to bring FDA's product review process fully into the computer age is already in place--both within and outside the agency. For example, FDA's drug reviewers are now using personal computers linked both to local area networks and to the Internet. This allows them to exchange information with colleagues and use databases and other computer systems throughout the world.

At the same time, physicians carrying out clinical trials of investigational drugs are increasingly recording data, case notes, and related information via electronic systems linked electronically with the drug sponsor's computers. Not only does this make data transfer virtually instantaneous, it also sharply reduces the risk of errors in writing down and copying patient information.

SMART is being developed in collaboration with representatives of the pharmaceutical and other industries and with officials from foreign regulatory agencies. High on the list of technical challenges is security of the information. One industry official noted that while CANDAs don't represent a problem in terms of protecting confidential, proprietary information about new drugs, the concept of computerized NDA information moving over telephone lines is cause for concern. There are ways of ensuring security--encryption programs and dedicated phone lines, for example--but precisely how this issue will be handled remains to be decided.

By the same token, any system for moving massive amounts of data electronically has to have built-in, highly reliable means to make sure that the data received is identical to the data sent--that there haven't been any inadvertent electronic "typos" along the way. FDA planners are now exploring numerous error-checking systems to determine which will best meet SMART requirements.

Finally and perhaps most important, any scheme for assembling, transmitting and processing product approval submissions to FDA must ensure patient privacy and anonymity. Individuals who consent to take part in studies of experimental drugs, biologics, and medical devices have a right to know that their identity and health records will be kept in strict confidence. SMART is being designed to do that.

When fully operational, SMART will save substantial amounts of time and effort in the product review and approval process. It also promises to maintain, and very likely elevate, FDA's high standards for gauging safety and efficacy and for monitoring product performance. Thus, the ultimate beneficiaries of SMART are the people who rely on the health-care products FDA regulates.

Ken Flieger is a writer in Washington, D.C.

Think of a Giant Spreadsheet

"First you might have a patient's number. Then you would want to have demographic information about that patient--weight, height, age, sex, etc. That takes up the first ten columns of the spreadsheet.

"What's in the next ten columns? Well it might be things like drug dose, duration of dose, change in dose, cumulative dose, and so forth. Next you might want to see all the clinical laboratory tests for that patient--blood count, urinalysis, biochemical assays, etc." After that, he says, the spreadsheet might show adverse events and other findings and observations about that patient's experience with the drug. "Pretty soon you might have a spreadsheet that has a thousand columns in it."

But, Williams points out, the spreadsheet couldn't have just one row per patient. "A patient may come in for a hundred visits during the clinical investigation. So now you've got a hundred rows each having a thousand columns, a hundred thousand cells of information to describe one patient."

It's not unusual for an investigational drug to be studied in 3,000 patients, which means that a single NDA may have 300 million cells of information on safety data alone. There could be equally big data sets for efficacy, pharmacology, preclinical studies, chemistry, manufacturing--all the elements that make up a typical new drug application.

"That sounds overwhelming, and I think it is," says Williams. "But that's where the wonders of computers come in. They give us a way to get a handle on these massive data sets. The computer is an incredible tool," he says, for the staggering job FDA has to do to decide whether or not a new drug belongs on the American market.

--K.F.