Last week, the American Medical Association and a consortium of medical journal editors called for a national registry of clinical trials, one that would require drug companies to submit results of all their studies, both favorable and unfavorable.
These groups are thinking, quite rightly, in terms of todays patients and therapies. Why should drug companies be able to advertise successful results from one clinical trial while burying results from ones showing a drug has no effect?
But theres even more at stake. The registry could change the economics of drug development by making clinical trials, the most expensive aspect of drug development, cheaper.
The more patients in a clinical trial, the more it costs. But if you had the right kind of data from the registry, “you should be able to build better trials that require fewer patients,” according to David Handelsman, lead strategist for clinical research and development at SAS Institute, a company that supplies statistical and modeling tools to analyze and design clinical trials.
Some of the most useful, and most secret, information is from trials showing the drug doesnt work, says Alex Bangs, CTO of Entelos. “If we assume that it would be nice to stop reinventing the wheel, then theres no doubt that having negative data would help,” he said.
Entelos builds in silico models of physiological systems and uses them to help drug companies diagnose disease and predict drug response more efficiently. The company, along with Roche Diagnostics, recently won an award for developing a much less invasive form of a common diagnostic test for diabetes. To build these models, Entelos uses information from the peer-reviewed literature.
Much of the data out there comes from animals and cell cultures. These require “a lot of interpretation,” he explained. But clinical data, by definition, comes from humans. “Anytime we can get that kind of data,” he said, “we can use that as a validation of the model.”
Such useful data is hard to come by, said Alan Hopkins, president of PharmaStat, which advises companies on clinical development. “Usually the only source of data is published reports,” he said, echoing Bangs that, while difficult to get, reports of failed studies could be particularly useful.
But Hopkins, former head of biostatistics and clinical information technology at Genentech, is highly skeptical that companies would be willing to share the data. “Persuasion by editorial committees probably wont make this happen,” he said. The FDA is soliciting public comment for technologies to shorten clinical trials, but I am unaware of any position on a mandatory registry.
Opponents of the registry may argue that it could be more misleading than informing. All three of these experts would agree that not all clinical trials are equal. You cant determine the efficacy of the drug by taking a poll of trial results. Some trials are better designed and more telling than others. And these trials should matter less in making models or health care decisions.
This is precisely why the registry needs to be constructed carefully. “Theres got to be a standard for putting this information in a repository,” said Hopkins, one that would specify required elements. And it should support more than text searching, he said, slipping into biostatistician patois, imagining an appropriate XML to search sample sizes, endpoints and more.
If companies supply the data, experts will crunch them. And maybe, just maybe, the cost of clinical development will start to decline.