Everything about Health Decisions screams small company with big plans. Fancy new office-park headquarters fit the company like hand-me-downs on a kid brother. A half sheet of paper next to the glass front door lists every employee’s name and extension, and people wander by would-be conference rooms in shorts or blue jeans. There’s even a six-passenger corporate plane.
Yet there is something unmistakably special going on at the Chapel Hill, N.C.-based clinical research organization (CRO), which designs and runs clinical trials for pharmaceutical companies and government groups like the National Institutes of Health. Health Decisions, whose $12 million annual revenue makes it the smallest Enterprise Value Award winner ever, is challenging a conservative industry’s accepted drug trial process with the resolve of the Little Engine That Could, an appropriate comparison given founder and CEO Michael Rosenberg’s favorite metaphor for his company.
“When James Watt invented the steam engine, people said, This is great because we don’t have to use water wheels anymore,” Rosenberg says. “But for a long time people took steam engines and put them by the river.” They didn’t realize they could save themselves the river trip by putting the engine where they worked?using the new technology to improve the process.
Health Decisions won an Enterprise Value Award this year because it is using technology to change the way pharmaceutical testing is done, not just retrofitting an old process with new parts. Health Decisions’ clinical management system can shave years off the drug development cycle and save sponsors millions of dollars. “It is an order-of-magnitude gain,” says Partners HealthCare System Vice President and CIO John Glaser, a judge for this year’s awards. “This is beyond remarkable.”
Pharmaceutical companies spend an average of $802 million and 10 to 15 years of research and testing to bring a new drug to market, according to the Tufts University Center for the Study of Drug Development. During that time, they are developing promising compounds, experimenting on animals, and testing on humans to find any adverse effects, determine dosage and ultimately see whether a drug does what it is intended to do. Most drugs don’t; only five out of every 5,000 compounds are ever tested on humans, and only 20 percent of drugs that make it that far are ever approved by the Food and Drug Administration. In short, it is a long and expensive process that often fails.
CROs handle the human testing, or clinical phase of the process. They collect results from the doctors’ offices that see the test subjects, check for mistakes and inconsistencies, and perform the statistical analysis. Currently, the vast majority of them do it all on paper, recording by hand the information they collect from patients?everything from date of birth to pain levels. Because so much time and money is at stake with a new drug, most pharmaceutical companies aren’t willing to experiment with other processes. Those CROs that do electronic data capture (EDC) through the Internet claim to be innovative, but all they’re doing “is substituting technology for something that is essentially a [manual] process,” says Rosenberg. “They can get the data into the system faster, but they haven’t figured out how to use it faster.”
Rosenberg started Health Decisions on the principle that technology could change the process of data collection and analysis. Traditionally, research coordinators at the clinics send the paper forms to the CRO, where data-entry clerks enter the information into a computer and check for mistakes and discrepancies, called queries, which in turn are sent back to the clinics for resolution. Each process is a discrete step that must take place before the next one begins. Rooms filled to the ceiling with unprocessed forms are common, and sometimes it takes six months from the time a form is submitted to the time a query is answered.
I Think I Can, I Think I Can
Health Decisions, on the other hand, realized that if it could connect previously disjointed steps by sharing data it wouldn’t have to be a slave to the old process. “They started to connect these things in their heads and then built the systems that made it possible,” says Bob Reck, a member of the review board for this year’s awards, who spent more than 20 years as a pharmaceutical industry consultant. In a Health Decisions study, clinicians record patient data on bubble forms, similar to those used for academic testing. They then send the completed forms to Health Decisions, which scans the information into its system using an Optical Mark Reader (OMR). It’s not cutting-edge technology, but it reduces the staff needed to input data. Once the information has been scanned into the clinical management system, it goes through validation, query generation and resolution modules. When taken as a whole, a seven-year study can be completed in four, says Rosenberg. (Because no two studies are alike?each varies in terms of numbers of patients, types and amount of data collected, and number of locations involved?comparing one study to another is an apple-to-oranges comparison.)
Its results have gotten the company noticed. “From the beginning, [Health Decisions] has been able to incorporate technology in a way that has made it more efficient from a time and cost perspective than some other [CROs] that have been around longer,” says David Windley, an analyst and senior vice president for Jefferies & Co. in Nashville, Tenn., who covers the CRO market. (Windley owns stock in Health Decisions competitor PPDI.) “It is the way it ought to be done in my mind,” he says.
With a medical degree and a master’s in public health, Rosenberg’s rŽsumŽ reads like a who’s who of nonprofit medical research; he’s served on advisory boards for the NIH and worked for the Centers for Disease Control and several smaller organizations. During that time, he became increasingly frustrated by the clinical testing process. So in 1989 he started Health Decisions in his basement and soon got a chance to prove himself. At that time, San Rafael, Calif.-based Oclassen Pharmaceuticals was preparing a trial on a genital wart medication, and it needed results within several days. When the CRO that Oclassen hired realized it couldn’t process the data fast enough, Rosenberg convinced Oclassen that rather than change the study, he could run it?in the time allotted?using an OMR machine he had recently purchased for $3,000.
The process he went through then was more manual than it is today. “We sent the information to the patients who were using the drug, and if we didn’t get a response, we called them on the phone,” says Rosenberg, who is part Godfather, part man-Friday, simultaneously overseeing his company and, at least to some degree, doing everything himself?including setting policies, running studies and even flying the plane. But the process bookends were the same: OMR forms to collect the data and an SAS system for data handling and statistical analysis. Health Decisions’ first study was an overwhelming success and led directly to Health Decisions’ second and third customers.
A Day at Health Decisions
At 10 a.m. on a typical day at Health Decisions’ headquarters, the FedEx man knocks on the wooden door and hands over an armful of 8-1/2-by-11-inch packages containing filled-in bubble forms. Minutes later the whir of the OMR is audible in an otherwise quiet office. The forms slip through the reader at a startling rate, less than a second per page. Health Decisions’ data-entry module (which, like all of the company’s systems, is homegrown) imports the OMR data and performs the initial validation. The system is capable of scanning at least 8,000 pages in a single day.
Once the data is scanned, the quality-assurance team runs the validation and query management (VQM) module. Like the data-entry module, it was conceived in 1989 and first modernized in 1993. The current version, built in 1999, accesses an Oracle database. It is a client/server application, which could in the long run impede its ability to scale, but CTO Rick Farris, a dotcom refugee who joined Health Decisions in 2001, is designing a Web-based version.
The VQM checks for missing fields, logic errors and answers that violate the study’s protocol. It then generates queries and sends them via e-mail to the submitting clinic. For example, in a contraceptive study, the system would generate a query if the age field says a subject is 13 years old and another field says she is taking birth control pills. The clinic would have to confirm that this is really the case or make a change if someone filled in the wrong bubble.
Queries are the bane of pharmaceutical research and a financial burden: Industry estimates are that each query adds up to $200 to a study’s cost. However, Health Decisions’ OMR technology reduces errors at the origin point and when data is transferred to the data-entry system. Rosenberg estimates the industry average query rate to be one query per page. His company’s rate is one-twentieth of that. The query rate is so small that Data Manager Jim Box jokes that he can tell when a clinic hires a new researcher because the query rate shoots up.
But it is Health Decisions’ ability to turn queries around within a day of receiving a form that leads to the real benefits. Since the clinicians filling out the forms receive queries while patient visits are fresh in their minds, they can provide answers easily and learn from their mistakes.
The queries go out as e-mails containing a link to an extranet where clinicians can enter corrected data directly into Health Decisions’ Web data query module. The newest module in Health Decisions’ system, a Web-based master subject log, tells the company the total number of patients the clinics have seen and exactly what data has yet to be submitted. Once all the data is collected, analysis begins, using an updated version of the SAS system Rosenberg first purchased in 1989.
The Company’s Big Break
Health Decisions’ break came in 1996, when Protodigm, a wholly owned subsidiary of Roche, a $15 billion pharmaceutical giant in Basel, Switzerland, hired the then seven-employee CRO (it quickly grew to 50 to accommodate Protodigm’s needs) to test an experimental Alzheimer’s drug. The trial contained more than 450,000 pages of data and tested 1,500 patients from 100-plus locations?making the number of patients about double that of the largest study the company had ever handled
While the drug was ultimately a failure?it didn’t clear the FDA’s bar for relevant cognitive-skills improvement?the study was an astounding success for Health Decisions. The study took three and a half years from the time it started to the time the drug was killed. Roche estimates that killing the drug early saved the company $32 million.
“The most important difference [in Health Decisions’ process] is the time from the last patient seen to the closing of the database,” says Herman Ellman, senior vice president of clinical development at Rockaway, N.J.-based pharmaceutical company Warner Chilcott. Since Health Decisions takes less time to clean its data before analysis begins?four days versus the industry average of 122 days, according to Rosenberg?drug companies can interpret results sooner. As was the case with the Alzheimer’s drug, pharmaceutical companies can kill an ineffective drug right away. When a drug is successful, they can spend more time on data assessment while still submitting a new drug application to the FDA in a reasonable amount of time, says Cap Gemini Ernst & Young consultant Gary Green, a 25-year veteran of the FDA.
Yet, amazingly, Health Decisions’ success has been an obstacle to its growth. “One problem we have is that our largest competitor [Quintiles, a $1.6 billion CRO in Durham, N.C., a 10-minute drive from Health Decisions] will [tell its sponsors it has] 10 people in data management just for your project,” says CTO Farris. “We say we have three or four data-management people in the company, and one will handle all your needs. In some proposals, [our competitors] have a pricing structure based on the number of queries. We specifically say we will reduce your queries. We are going to save them millions, but they are afraid because we don’t require [as much staff].”
Rosenberg is philosophical. “This is a very conservative industry, and they don’t want to see something that is very different,” he says. “Something that is a little different they are willing to take a chance on, but to walk in the door and see something very new, most people’s reaction is, Gee, I don’t know about this stuff.”
That’s a typical reaction to a disruptive technology, which is exactly what Rosenberg thinks Health Decisions has. And with such technologies, profound implications are associated with them. Using the information faster not only saves time and money, it saves enough so that new markets may become viable. Today, big pharmaceutical companies won’t even bother researching medicines to treat diseases with less than a billion-dollar-a-year market.
If Health Decisions, or any other CRO that emulates its methods, can cut the cost of developing a drug in half, then smaller markets become economically viable. Pharmaceutical companies would have more incentive to develop so-called orphan drugs, which treat relatively rare disorders like sickle-cell anemia that affect less than 200,000 people in the United States. It would be a change long awaited by many and the culmination of Rosenberg’s public health career.
In the meantime, Health Decisions will continue trying to shake up the pharmaceutical industry. “If you look at the automotive industry, [American companies] were really complacent until the Japanese came and put a lot of pressure on them,” says Rosenberg. “And the Americans had to learn how to do things better. I think the pharmaceutical industry is in its complacent stage, and I think it is going to start to get pressure to change. I think there are a lot of efficiencies that are waiting to be realized in this industry, and what we hope is that we will be the catalyst for these changes.”