University of California, San Francisco responds to inaccurate claims by company

The University of California, San Francisco and UCSF AIDS researcher James O.
Kahn, MD, strongly rejected claims made in a press statement issued Tuesday
(October 31) by a company that attempted to block publication of an important
AIDS research study.

Kahn cited numerous inaccuracies in the company statement.

The dispute stems from the attempt by the company to block Kahn and his
colleagues at Harvard School of Public Health, Brown University and Cornell
University Medical Center from submitting for publication the results of a
large, multi-institutional trial of a drug, HIV-1 Immunogen, which it was hoped
would boost the immune system of patients with HIV-1.

The study, established in 1996, was led by Kahn, UCSF associate professor of
medicine in the Positive Health Program at San Francisco General Hospital
Medical Center, and senior author Stephen W. Lagakos, PhD, Professor of
Biostatistics at the Harvard School of Public Health. The study was terminated
five months early, in May 1999, after the second of three scheduled interim
analyses of the data by an independent Data Safety and Monitoring Board (DSMB)
showed that the compound showed no evidence of clinical benefit. The Data
Safety and Monitory Board was established by the company and the study team to
monitor the data being gathered in the clinical trial.

However, in January 2000, the company informed the university researchers that
it would not provide them with the final data set from the study, but would
instead have the data analyzed by a consultant hired by the company. This move
was contrary to the provisions of the study protocol, which provided that the
study team would perform the analysis. When the study team challenged this
unauthorized protocol change, the company said it would provide the final data
set only if the study team agreed to give it veto power over any publications.

Because these conditions were unacceptable to the study team, the researchers
decided to publish the data it had available from the DSMB analysis. (The
research team decided to proceed with publication because they estimated they
had 95 percent of the confirmed clinical progressions that would have been
recorded in the final database. However, they still seek the full data set
because it could provide important information about how patients responded to
the drug and what responses patients developed while receiving different
anti-retroviral therapies, which most received while taking Immunogen.)

When Kahn, director of the study, provided the company in July 2000 with a copy
of the manuscript that the researchers had prepared, the company filed an
arbitration demand in September 2000 seeking to block publication.

The researchers submitted the manuscript to the Journal of the American Medical
Association, where it is published in the November 1 issue.

The primary measure for the drug’s effectiveness was HIV progression or death.
Secondary measures were the quantity of the virus in the blood, CD4+ T-cell
counts, and body weight.

Kahn and the University offered the following responses to specific claims and
suggestions made by the company:

The company press release issued on October 31, 2000, suggests that the
research team failed to include in their paper a critical data sub-set
measuring the product’s effect on levels of HIV in the bloodstream. This is
false.  The results from the sub-set were included in the final manuscript and
can be found on page 2,199, column 3 in the text of the article. The method by
which the sub-set data was analyzed is the one called for in the original study
protocol.

The study team did not include the company’s analysis of the sub-set data
because the company used an inappropriate method of analysis that was not
pre-specified, Kahn said.

“This approach was not specified in the study protocol and gives erroneous
results,” Kahn said.

“It is a fundamental tenet of the clinical trial process that a research team
agree beforehand how it is going to analyze the data, rather than try, posthoc,
to see which method of analysis provides the most favorable results,” says Zach
Hall, PhD, UCSF vice chancellor of research. “This approach is critical to
preserving the integrity of the clinical trial process.”

“For UCSF, as an academic health care institution, there are two core values at
stake here: the first is our ability to freely communicate research results;
the second is our special obligation in human subject research, both to those
who participate in trials and to the larger patient community, to publish the
results of studies fully, accurately and objectively.  What is at risk is the
trust of the public in the clinical trial process.  Any attempt to selectively
choose the data taints the process and jeopardizes this trust,” Hall said.

In the sub-set of 250 patients who had their blood drawn more frequently than
other participants, the researchers found no statistically significant
difference in the patients’ viral load when measured using the analysis method
specified in advance in the protocol. The 250 patients had specialized HIV
immunogenicity studies performed and it was clear the drug was active and that
it elicited an immune response, but the patients in this group showed no
statistically significant difference in viral load, CD4 counts or body weight.
This data is included in the manuscript.

The company, however, wanted the researchers to include the company’s analysis
that patients in the sub-set who received the drug had a steeper drop in viral
load than those receiving the placebo. This was not an accurate interpretation
of the data, according to the study team.  Rather than following the viral load
in the blood over time, the company focused posthoc on selected time points
that had not been agreed upon in advance.

“One might see a small difference doing this type of analysis, but it’s not
appropriate,” Kahn said.  “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.”

The company statement falsely contended that Kahn excluded clinical
investigators and sponsor comments in the publication of the data.  The
researchers included some suggestions by the company in their analysis, and
they included all available data from the study in their paper.

The trial began in 1996 and was intended to continue until all patients had
been followed for two and a half years. However, the study was terminated after
the analysis by the independent DSMB showed that the compound had no clinical
benefit and that it was unlikely any clinical benefit would be shown by
extending the study. The study leadership, which held the responsibility for
making the final interpretation of the results, agreed with the DSMB
recommendation, and the trial was stopped.

The study team attempted to share their findings with, and get comments on the
draft manuscript from, the investigators at the 77 sites participating in the
study. However, despite repeated requests to the company to receive the list of
investigators and their addresses, the company refused to provide them.

“We asked many times for the list of investigators to circulate the manuscript
and to solicit their opinions,” says Kahn. “As the national principal
investigator, I had a responsibility to all sites, and if I were to short
circuit the process by sending our analysis only to colleagues we knew, we’d be
doing a disservice to all of the sites.

“By refusing to give addresses, the company was thwarting academic freedom,”
says Kahn. “We took the higher road by saying we must send the information to
all sites, and the only way to effectively distribute was through a scientific
publication like JAMA.”

The researchers submitted a manuscript based on the data available to the
DSMB.  On July 4, the study team provided the company with a copy of the
manuscript, and the company replied with comments on August 8.  On August 23,
the study team returned a revised manuscript incorporating some of the
company’s comments.

On September 1, the company initiated the arbitration proceedings to block
publication.

According to the company’s October 31 press release, “the public was informed”
about the study results in May 1999.

“The public was not informed,” says Kahn. “They received the company’s press
release. It did not reflect the analysis of the study team. The analysis
presented by the company did not reflect the data drawn from the entire study
population, nor the three virologic analyses put out by the study team—which
is why it is so important to publish the results.”

“The company’s press release in May 1999 stated that there was an anti-viral
effect, which our analyses demonstrated was insignificant,” Kahn said. “The
pre-specified analysis of all patients participating in the study, as well as a
pre-specified analysis of the sub-set, were not statistically significant.”