CLINICAL TRIALS: Company, Researchers Battle Over Data Access

Carol Cruzan Morton

Volume 290, Number 5494, Issue of 10 Nov 2000, p. 1063.

URL: http://www.sciencemag.org/cgi/content/full/290/5494/1063

A dispute between university-based researchers and the corporation that
funded their study is threatening to erupt into a $7 million to $10
million legal battle. Last week, the researchers reported that a large
clinical trial of an immune system booster to treat HIV-infected people
found that the drug isn't effective in slowing progression to AIDS or
reducing mortality. The company that developed the drug tried to block
publication of the study unless the researchers included the company's
analysis of a subset of the data that suggests the drug might help some people. The researchers refused.

This nasty dispute has again raised the issue of who controls the data
when corporate and academic interests conflict. "This is probably the
unusual case, where investigators and the journal are standing up to the
supporting companies," says science policy analyst Sheldon Krimsky of
Tufts University in Medford, Massachusetts. "More typically, you find investigators willing
to compromise to avoid legal action or loss of funding for future
projects."

The study is believed to be the largest randomized clinical trial among
HIV-infected persons in the last decade. The 3-year, double-blind study
of 2527 otherwise healthy HIV-positive people at 77 U.S. sites tested a
drug called Remune, developed by Immune Response Corp. of Carlsbad,
California. Immune Response and the University of California, San
Francisco (UCSF), Center for AIDS Research funded the research.

The trial ended in May 1999 when an independent safety monitoring board
decided that the drug showed no clinical benefit and was unlikely to do
so. That's when trouble started brewing, says AIDS researcher James Kahn
of UCSF, the study's national principal investigator. In the news release announcing the trial's early end, Immune Response claimed that
an analysis of a subset of people who underwent more frequent blood tests indicates that Remune
reduced the amount of HIV in their blood--the "viral load." This effect,
the company noted, would be the basis of its application to the U.S.
Food and Drug Administration for marketing approval for the immunogen.
(Another trial, focusing on the drug's effect on viral load, began in
September 1999.) Kahn and the study leadership team conducted a preliminary analysis of a
larger sample, which they presented at a fall meeting in San Francisco.
Remune, they concluded, had no apparent clinical effect and no discernible effect on viral load.

In January, the disagreement about how to summarize the virologic effects led the company to propose that Kahn, biostatistician Stephen
Lagakos of the Harvard School of Public Health in Boston, and two other researchers could have access to the complete data set--which the
company controlled--if they agreed to written company approval of "the content, analysis, results
and discussion" before publication, according to a memo from the company to Kahn and his team. The memo also asked for prior approval of any
further analysis and to limit the researchers' access to the data to 1 year. "We were flabbergasted that they would put new conditions on
getting the data, and we objected strenuously," Kahn says.

The researchers didn't accept the terms. Instead, they used the data
submitted to the safety board, which they say is 95% of the results on
clinical progression. This summer, they sent both a draft and then the
final manuscript to Immune Response for review to ensure that it
contains no proprietary information, as required by their contract. The
researchers incorporated some revisions, but refused to add a figure
illustrating the company's analysis of the subset data.

Ronald Moss, Immune Response's vice president for medical and scientific
affairs, says that "despite the failure to show significant differences
in clinical endpoints, [the company's subset analysis] gave us valuable
insight into the potential effects of Remune on viral load and T cell
help. ... The Remune group is favored at weeks 36, 48, 60, 84, 96, and
120 ... we felt it was extremely important to have a section in the
paper describing and discussing the results." Lagakos counters, however,
that he also analyzed the data for the subset, and "there were no
significant differences using the analytical methods specified in the
protocol." Kahn adds: "Immune Response used a statistical test that was
inappropriate for the data. The company thinks 'data dredging' makes
sense. There are no differences at certain interim time points, and
there are differences at others. One cannot pick and choose data points
to suit one's needs."

When it was clear that they would not reach agreement, Immune Response
invoked a contract clause asking for legally binding arbitration,
seeking damages of $7 million to $10 million. The company claims that
the research agreement gives the researchers access only to data
generated by the UCSF site, and that data from other sites are
confidential. UC has filed a counterclaim asking for the complete data
set and maintaining the right to publish further analyses. "In every one
of three key documents--the research agreement, the protocol, and the
site agreement--there is very clear language that the study team will
publish the results of the study and that they have the right to do
that," says UC counsel Christopher Patti.

The researchers submitted their manuscript in September to the Journal
of the American Medical Association, which quickly published the paper
in the 1 November issue along with a cluster of articles and a
commentary that address academic conflicts of interest with industry
research sponsors.

The controversy has sparked curiosity about the company's version of the
subset analysis. Alexandra Levine of the University of Southern
California in Los Angeles says that the researchers should have included
the company's data. "Why not give the reading audience full access to
the data?" she says. "If the authors are presenting data fairly, then
present all of it."

Moss says the full story will be out soon; other investigators involved
in the clinical trial will publish another analysis, with the disputed
figure, possibly in January.