That not only was a departure from scientific norms, but also artificially inflated the apparent benefit of the prostate cancer therapy when the results were revealed to doctors and investors through a 2010 paper in the New England Journal of Medicine, according to experts in biostatistics and clinical trials.

Dendreon said the discrepancy is immaterial and that Provenge is safe and effective. The U.S. Food and Drug Administration also stood by its April 2010 decision to approve the treatment.

But the disclosure could mark another setback for Dendreon, whose market value has tumbled over 70 percent in the past eight months after disappointing sales of Provenge. The company also faces at least five lawsuits accusing management of insider trading and misleading shareholders about the treatment's prospects.

All companies seeking U.S. Food and Drug Administration approval for their drugs or other therapies tell the agency beforehand how they will analyze data collected in a clinical trial. Once the trials are finished, the FDA determines whether the therapy met the pre-specified goals and is safe and effective.

According to a copy of Dendreon's Statistical Analysis Plan (SAP) submitted to the FDA in early 2008 and reviewed by Reuters, the company said its key trial would look first at whether patients who received Provenge survived longer than patients who received a placebo. The company also said it would analyze whether the therapy affected patients differently depending on whether they were younger than 65, or 65 and over.

But when it published the results of the IMPACT trial in July 2010, the cut-point for that age-based analysis was 71 years, not 65.

Deviating from pre-specified statistical analyses raises the suspicion "that they didn't like what they saw and went looking for something better," said Harvard School of Public Health biostatistician Scott Evans, who has no link to the clinical trial or Dendreon. "If so, it's data dredging with the possibility of producing a false positive" to reflect well on the drug.

Dendreon denies that it manipulated the data. The company's chief medical officer, Dr Mark Frohlich, said FDA guidelines require using 65 as a dividing line, but trial researchers later determined that 71 made more sense biologically, since it represented the median age of patients in the trial.

A spokeswoman for the FDA said that it approved Provenge based on the reported effect for all patients, regardless of age, and the agency stood by that decision. The FDA would not address queries by Reuters on whether the age-related discrepancies in the effects of Provenge cast doubt on the overall value of the treatment.

The New England Journal of Medicine published the age-specific analysis from the IMPACT trial after receiving signed statements from the trial researchers, including Frohlich, that the data presented matched the plan submitted to the FDA. Following inquiries by Reuters, the journal issued a correction in its October 11 print issue reflecting that the age-based analysis did not.

DOES AGE MATTER?

Provenge is not a traditional drug, but a therapy in which some 25 billion white blood cells are taken from the patient, treated to make them produce immune cells that purportedly target prostate cancer, and infused back into the body.

Using 65 or 71 for an age-based analysis may seem like nitpicking. But a study published in the Journal of the National Cancer Institute (JNCI) in February argued that it matters, and that reporting a different age analysis than the one Dendreon told the FDA it would perform hid problems with Provenge.

Data from the IMPACT trial, as reported in the NEJM, showed that patients who received Provenge lived 4.1 months longer on average than men who received the placebo treatment. That benefit has been seized upon by doctors considering whether to prescribe the $93,000 therapy and by men deciding whether to undergo it. The IMPACT paper also reported that men older and younger than the age of 71 benefited from Provenge, though the younger men benefited slightly less.

A different picture emerges when the men are divided at age 65, according to the JNCI paper. The authors - former hedge fund analyst Marie Huber, two prostate-cancer experts and an immunologist - used unpublished data from Provenge's clinical trials to compare the therapy's effects on men older or younger than 65, and concluded that the treatment did not extend patients' lives and may actually be harmful.

Their analysis found that the main reason Provenge appeared to extend survival was that older men who received the placebo died surprisingly early, and many months sooner than those who received Provenge. Huber and her co-authors argued that this was because the placebo, which consisted of removing white blood cells, was harmful to the men, rather than because Provenge was beneficial.

The placebo "was particularly dangerous to men over 65," said Huber, who has been the target of attacks by Dendreon and its supporters.

An FDA review of the IMPACT data in 2009 also found that men under 65 who received Provenge had a 41 percent greater chance of dying during the IMPACT trial than did men in the control group. That suggested "a trend in survival in favor of the control group compared to the (Provenge) group," the FDA scientists wrote.

Not reporting the data based on the 65-year benchmark, as the company told FDA it would, hid that and other red flags, says Huber.

"To omit an unfavorable pre-specified analysis represents scientific misconduct," she said in an interview. "To substitute a more favorable analysis and represent it as the pre-specified analysis is falsification."

Dendreon and its supporters accused Huber and her JNCI co-authors of "junk science." By making 65 the cut-off instead of 71, Dendreon told Reuters, Huber had analyzed data in a way that had not been planned before the clinical trial began.

Huber defended her findings, saying Dendreon's own Statistical Analysis Plan for the IMPACT trial showed that the analysis Huber performed was what the company itself had specified ahead of time. Huber obtained the SAP through a Freedom of Information Act request, and shared it with Reuters.

The FDA's "Guidance on General Considerations for Clinical Trials" says that trial results should be analyzed according to the pre-specified SAP plan. FDA spokeswoman Heidi Rebello declined to address the discrepancies found in the age-specific analysis.

Susan Ellenberg, former director of the FDA's Office of Biostatistics and Epidemiology, said Dendreon should explain why it deviated from its SAP plan.

"It's very easy to find a cut-point to show what you want," said Ellenberg, now a biostatistician at the University of Pennsylvania. "They shouldn't publish something different from the SAP without explaining why or presenting both."

Dendreon said it had shared all its data and analyses with the IMPACT authors and the FDA, which approved Provenge before the IMPACT trial was published.

"You have to be really careful about calling into question a finding from a randomized clinical trial," said Dendreon's Frohlich. The less favorable results from changing the cutoff age to 65 from 71 are almost certainly "a chance finding," he said.

(Editing by Michele Gershberg, Edward Tobin and Matthew Lewis)

• A Permanent Economic Emergency