ROCKVILLE, Md.-- The Arthritis Advisory Committee of theFDA on Thursday recommended against approval of GreenwichPharmaceuticals Inc.'s treatment for rheumatoid arthritis witha vote of 5-0 (with two abstentions).

The decision turned on Therafectin's apparent lack of efficacy.The Fort Washington, Pa., company had conducted "fiveoverpowered studies to get two that were significant," saidcommittee member David Felson, professor of medicine atBoston University School of Medicine.

The die seemed cast when committee chair George Ehrlich, aprofessor of medicine at the University of Pennsylvania, askedto know precisely what the FDA's mandate is regardingefficacy.

Events leading up to the committee meeting suggested thatGreenwich might have saved itself the trouble of preparing forthe meeting. Last September, FDA sent the company a "notapprovable" letter, saying that while one clinical trial, RA9,demonstrated efficacy, "similar effects were not replicated inmultiple larger studies conducted under this NDA."

Experimental Design

The studies RA9 through RA11 had four end points: swollenjoints, painful joints, physician assessment of the patient'scondition and the patient's own assessment. A fifth end pointwas improvement in the other four.

In RA9 and RA10, patients were enrolled who showed"adequate worsening" upon ending treatment with non-steroidal anti-inflammatory drugs (NSAIDs). This created abaseline. Two hundred patients were then randomized toreceive placebo or therapy for 12 weeks.

Forty-one percent of 98 patients on Therafectin and 21 percentof 103 patients on placebo improved in all four variables(p=.003).

In RA10, 36 percent of 206 patients on Therafectin and 30percent of 105 patients on placebo improved in all fourvariables. Neither these overall results nor those for theindividual variables were statistically significant.

Furthermore, the biweekly analysis the company had done ofRA9 and RA10 results "are much more likely to say somethingis positive when it isn't," than combining the data, committeemember Peter Lachenbruch, a professor of Biostatistics at UCLASchool of Public Health, told BioWorld.

In RA11, patients who were on Therafectin for nearly twoyears were randomized to placebo or continued Therafectin.

RA12 was a labeling-and-use trial in whch patients onTherafectin or placebo also took NSAIDs. These werewithdrawn after 12 weeks. The primary end point wasprevention of NSAID withdrawal flare -- sudden worsening ofinflammation.

Gary Cook, a statistician consulting for GreenwichPharmaceuticals, claimed that RA10 was supportive of RA9.When the data from both were pooled the overall resultremained positive, he said.

But Hoi Leung of FDA's office of drug evaluation II, spent 45minutes dissecting weaknesses in data from all the studies."RA9 showed statistically significant differences in three of thefour standard efficacy variables at week 12," he said.

"RA10, with identical design and twice as many Therafectinpatients as RA9, did not provide statistical evidence to supportthe efficacy of Therafectin," said Leung. "There was no credibleexplanation for its failure."

The pooling of RA9 and RA10 eliminated statistical significancefrom one of the positive end points, leaving only two significantpositive end points for the combined studies, said Leung.

As for RA11, "statistical significance in the whole population(nominal p=0.035) is questionable since the effect ofTherafectin in adjunctive patients was considered to be ofquestionable value," Leung said. "Statistical adjustment for it'sp-value would seem to be in order." Additionally, Therafectinpatients used an average of .10 to .20 tablet of propoxypheneper day more than placebo patients.

John Decker, also of FDA's office of drug evaluation II, dealt thecompany another blow: No data on RA5, an earlier clinical trial,had been included in the company's packet; "nonetheless, it'spart of the NDA," Decker said.

In fact, RA5 was a bigger study of longer duration withpatients more functionally impaired than RAs 9-12, Deckerpointed out. In RA5, 381 subjects in functional classes two orthree instead of one were followed for a year (higher functionalclass numbers indicate sicker patients).

There was no quarrel with the quality of the study's design orimplementation, but Therafectin demonstrated no statisticalsuperiority over placebo.

The company disputed neither Leung nor Decker. The onlymitigating factor was the apparent safety of the drug; therehad been a total incidence of proteinuria of 2.3 percent but thishad fallen off with time. And there had been some patientconstipation.

The committee's discussion suggested that further studiesmight be futile. "If a drug that is very weak is going to beapproved, I'd like to know that it's absolutely safe," Felson said.

Greenwich said in a press release late Thursday that "thecompany is considering a number of alternatives as a result ofthis decision, both in terms of Therafectin and the othercompounds in the company's clinical development program."

Greenwich's stock (NASDAQ:GRPI) closed at $2.50 a share onThursday, down 56 cents, in heavy trading of more than 2.3million shares.

-- David Holzman Washington Editor

(c) 1997 American Health Consultants. All rights reserved.