Medical Device Daily Washington Editor

FDA's use of advisory committees seems to be expanding in terms of both frequency and purpose, and Tuesday's meeting of the orthopedic and rehabilitation devices advisory committee extended that trend, dealing with a PMA supplement.

Genzyme (Cambridge, Massachusetts) filed the supplement with FDA to allow a single injection of 6 mL of Synvisc, a viscosupplement, into the knee rather than the currently approved treatment regimes of three injections of two mL each, but the agency's reviewers evidently found the application confounding, perhaps in part because of the firm's use of data from outside the U.S.

Still, the expression "no-brainer" was heard during the course of deliberations, and the panel voted unanimously in favor of allowing the new dosage regime on the market.

Michael Halpin, Genzyme's VP for regulatory affairs, told the panel, "Synvisc has been commercially available for 16 years and was approved in 1997" by FDA. More than 4.6 million patients have had 13 million injections of the product, he said.

Richard Polisson, MD, senior VP for clinical research at Genzyme, said knee osteoarthritis (OA) affects almost 17% of adults aged 45 or older and that current analgesic regimes carry a lot of baggage. He said that non-steroidal anti-inflammatories "have lots of side effects that we know of and lots of medication interactions, including with Coumadin."

To report patient pain scores, Genzyme used the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC), which Pollison said has been "validated and widely used in clinical research."

He said that the WOMAC A1 measure "deserves special comment" in that the question of pain reported while walking on a flat surface is "more relevant measure of mild to moderate" cases of OA, for which the product would be indicated. He also noted that FDA approved the WOMAC A1 measure in other IDE trials for viscosupplements.

Lena Holmdahl, PhD, the company's VP for clinical research, said the pilot study for Synvisc One entailed four treatment regimes, including the Synvisc standard of three injections of two milliliters (mL) each and one injection of six mL. "All treatments were safe, all were efficacious, but there seemed to be some difference in terms of efficacy," she said, with the single injection of six milliliters outperforming the other approaches.

For this study, enrollees had mild to report moderate disease pain of walking score of 2 to 3 on a Likert scale (which rates on a scale of 1 to 5), and the primary efficacy objective was to establish statistically significant better scores for pain than placebo – which was an injection of saline – at 26 weeks. The trial, which FDA did not review beforehand, stipulated that the intent to treat population was the subject of the statistical analysis.

Holmdahl said that 93% of the subjects getting Synvisc One completed the study, as did 91% in the control arm. "We therefore believe the study is of a high quality." She noted that the level of statistical significance, a P value of .047, "is in the same range as other products used to treat OA pain." She said that the data indicate that "Synvisc was superior to the control throughout the duration of the trial."

One of the difficulties with any such trial, Holmdahl cautioned, is that "Synvisc is not a systemic treatment" and pain location can be difficult to establish at times for patients with multiple sites of OA. However, she said the numbers were decisive for patients when "the treated joint is the only joint that is affected" by OA.

Holmdahl also noted that "the overall safety profile was similar between the two groups." She said, "Most adverse events were mild to moderate in intensity" and none were deemed related to the device.

Clare Elkins, director of biostatistics at Genzyme, addressed the way that the company handled the statistical analysis compared to the approach employed by FDA. For the analysis of the primary endpoint, Elkins said Genzyme treated investigator and site "as a fixed effect and analyzed the change from baseline," whereas FDA saw investigator and site as a random effect and "analyzed absolute value."

Elkins said FDA analyzed the data on six different occasions, and found problems only with the last two analyses. All the same, she said, "all FDA re-analyses confirm the statistical significance" finding by the company. Furthermore, she said, the fact that the study examined secondary endpoints does not necessarily impose a need for any additional statistical analyses.

Because the company is not asking for an additional indication, "there is no need to adjust for multiplicity" due to the inclusion of secondary endpoints in the trial. "Five [of the six] FDA re-analyses confirmed the significance of the primary efficacy endpoint," she added.

Kyung Lee, MD, the lead reviewer for FDA on this application, informed the panel, "The study was not conducted in the U.S. under an IDE. Consequently, FDA did not review the protocol prior to its conduct." He stated that the difference in the WOMAC scores "was .15 on a five-point scale" and said that an analysis of covariance (ANCOVA) of secondary endpoints undertaken by the agency suggested that the results "were variable."

Chang Lao, PhD, of the agency's division of biostatistics, said that the agency's ANCOVA analysis indicated "no difference in overall least square means" between the study article and the saline sham. He said this analysis was undertaken in response to a need to "find a better variance-covariance structure among repeated measures of visits."

While most of his presentation consisted of information that could be understood only by statisticians, he indicated that the ANCOVA analysis over the 26 weeks of follow-up showed a "smaller variance of treatment difference ... than that of mean change from baseline" scores.

Lao asserted that the difference in the measurement of least square means in the control vs. the study subjects is "a more powerful approach" to discerning efficacy. He stated that for the WOMAC A1 score, the P score was .032 (with a least-square mean of 1.4 for Synvisc One and 1.55 for the placebo). At the end of his presentation, Lao acknowledged this score "is statistically significant," but proceeded to question whether it is clinically significant.

One of the more curious assertions in the FDA presentation was offered by Cunlin Wang, PhD, a biostatistician at the agency. Despite the fact that the original Synvisc has been on the U.S. market for 12 years, Wang asserted that hyluronic acid, the compound of which Synvisc is made, is "associated with increased risk of severe acute inflammatory reaction, the ... long-term consequences of which remain unclear."

On the question of whether the statistical improvement seen in the study amounted to clinical significance, Stuart Goodman, MD, of the Stanford University School of Medicine (Stanford, California), said the difference in pain scores was "really negligible," but pointed out "that is the same level of other interventions." Another panelist said the fact that Synvisc One means only one trip to the doctor vs. three for Synvisc is meaningful.

After further discussion, panel chair Jay Mabrey, MD, of the Baylor University Medical Center (Dallas), said in summation that the panel "generally believes that the statistics are appropriate" and that the difference of .15 "is relevant if small."

On the question of the multiplicity adjustments for secondary endpoints, panelist Brent Blumenstein, PhD, of Trial Architecture Consulting (Washington), said, "I can't look at those secondary endpoints without making an adjustment in my own mind," but he nonetheless remarked that "the direction [in which the numbers trend] is the most important thing."

Further discussion essentially echoed Blumenstein's comments, and Mabrey concluded that the panel "generally believes that the secondary endpoint analysis is appropriate," but that "even if some adjustments had to be made, they would be small."

The panel concluded with no dissent that the trial demonstrated efficacy, and Mabrey remarked in conclusion, "the device is modesty effective, at least non-zero, but [panelists] have some concerns about the population the device will be applied to." Regarding safety, he noted, "It's unanimous that the device is safe," and the panel also opted not to recommend a post-approval study.

When it came time for a vote for approvability, there was no motion for conditions and the panel quickly voted unanimously to recommend that FDA approve the product.