Letter Urging Against Approval of Cardiac Device REPEL-CV
December 20, 2007
Daniel G. Schultz, M.D.
9200 Corporate Blvd.
Rockville, MD 20850
Dear Dr. Schultz:
Public Citizen urges the Food and Drug Administration (FDA) to deny approval to SyntheMed’s REPEL-CV, a device to reduce cardiac adhesions (abnormal scar tissue formation in which surfaces normally separate adhere to each other) from surgery, because the device has not demonstrated a “reasonable assurance that the device is safe and effective,” the standard necessary for approval under the law. SyntheMed has said repeatedly that it expects FDA approval for REPEL-CV by the end of this year.
We incorporate by reference our public testimony at the Circulatory Systems Devices Panel meeting of September 19, 2007, in which we argued that the device should not be approved based on the submitted data. Since then, the company has generated no new data, to our knowledge. Transcripts of the meeting show that, of the panelists who voted to approve the device with conditions in an 8-3 vote, only three appeared to agree completely with the conditions for approval. The panelists quoted below all voted for approval.
SyntheMed submitted a single randomized, evaluator-blinded pivotal trial with an intent-to-treat population of 110 neonates (56 treatment, 54 control) who required at least two sternotomy (opening the chest cavity) procedures for the repair of congenital heart malformations. The device was implanted in the first surgery and adhesions were measured in the second. In the treatment group, REPEL-CV was sutured to the margins of the open pericardium (a sac surrounding the heart) underneath the sternum in the first operation, while in the control group the pericardium was left open. Study endpoints were based upon a four-point severity scale that graded adhesions based upon dissection difficulty, with Grade 0 being no adhesions, and Grade 3 being severe adhesions requiring “extensive sharp dissection” to separate the heart surface from the sternum.
The primary endpoint was the percentage of the investigational surgical site (ISS, an area defined by the edges of the heart and diaphragm) with Grade 3 adhesions detected during the second surgery. Three of the four secondary endpoints also drew from this scale by assessing the 1) the percentage of the ISS with Grade 0, 1, and 2 adhesions; 2) the patient’s most severe adhesion grade; and 3) the percentage of patients with Grade 3 adhesions as their worst adhesion. The terms “incidence,” “severity” and “extent” used in the initially proposed indication were not defined in the protocol, although they came to have relatively specific meanings during the panel discussion. The company sought a label claiming that REPEL-CV was effective for all three. “Incidence” appeared to refer to the occurrence of adhesions of any severity (i.e., the percentage of patients without Grade 0 adhesions in secondary endpoints 1 and 2), “severity” the grade of adhesion (e.g., secondary endpoints 1 and 2) and “extent” the patient-specific percentage of the ISS covered in adhesions (i.e., primary endpoint and secondary endpoint 1). Both the adhesion grading system and consequently the chosen endpoints had never before been used, let alone validated, for clinical significance. The final secondary endpoint measured the time required to dissect adhesions at the second sternotomy.
The imprecision of the adhesion endpoints was a recurrent theme in the panel’s deliberations. Even Dr. Eli Pines, the Vice President and Chief Scientific Officer of SyntheMed, acknowledged that “it was critical to have blinded observers making the assessments because we knew the assessment was somewhat quantitative (sic; presumably “qualitative” was intended), if you will.” He admitted that measurements were not reproducible among surgeons, raising serious questions about the appropriateness and clinical relevance of the endpoints. Panel members also questioned the procedure used to assess the endpoints, especially since the ISS is such a small area in neonates. Dr. Valluvan Jeevanandam asked, “if you start taking down adhesions blindly, how can you go back and say what percentage of areas had severe adhesions? ...it was the surgeon’s gestalt of that area.” Dr. Carl Lewis Backer, a SyntheMed consultant and surgeon, answered that the surgeon would “look at that area and way (sic) 70% of this area, 40% of this area, was severe, 20% was ... you didn’t, in terms of actually drawing it, no, but in your mind’s eye you know, you were just there, and you said okay, this percentage of this area would contain this grade of adhesions.” Dr. Richard Hopkins, a consultant to the panel, said, “in the zero to three grading system, the only one I really believe is the no adhesions to minimal adhesions.” Dr. Judah Weinberger agreed, saying, “I’m particularly perplexed that (sic) the weakness of the methodology for gathering this information … this might have just as easily been graded one plus two, plus three, plus four, plus, rather than assigning continuous variable numbers to that. That’s the way I feel about the precision of the data.” This assessment methodology essentially allowed the surgeon to use her judgment of adhesion severity, extent and incidence, an inadequate standard for a clinical trial of this importance.
Regardless of their precision, subjective measures are especially prone to unblinding problems. SyntheMed acknowledged that 26 out of 110 patients (half of them randomized to REPEL-CV) had procedures that were evaluated at reoperation by an unblinded surgeon, a breach of trial protocol. The sponsor claimed to have addressed this problem by doing an analysis only of the primary endpoint that excluded the 26 unblinded patients and still found statistically significant results. But, as we emphasized in our testimony, there was a separate unblinding problem. SyntheMed revealed that “implanted test material or a fibrous capsule, or other abnormal tissue” was present in 30% (17/56) of REPEL-CV patients and 1.9% (1/54) patients in the control group (p<0.0001). Such unblinding could be a significant source of bias, inadvertently alerting the surgeon that the patient had received REPEL-CV. Indeed, Dr. Backer, the SyntheMed surgeon, admitted under questioning that he did observe “tiny remnants of kind of whitish-yellow fibrocollagenous looking material at the very edges of the pericardiums,” and that the Vicryl sutures used to keep REPEL-CV in place were potentially an indication to the surgeon that a device had been placed. This evidence was completely ignored by the panel during its deliberations. It is thus possible that up to 54% (17+13/56) of the REPEL-CV group and 26% (1+13/54) of the control group were unblinded.
Setting aside the issues that the endpoints have not been validated, that the assessments were subjective and that significant unblinding occurred, the study only demonstrated at best modest efficacy for the device. Overall the study showed a decrease in the “severity” of the adhesions, but not in either their “extent” or “incidence.” In other words, adhesions were just as common (“incidence”) and covered as much of the ISS (“extent”) in REPEL-CV patients as in the control group; only the percentage of the ISS with Grade 3 adhesions was reduced (“severity”). Commendably, the advisory committee struck the “extent” and “incidence” indications from the proposed label as conditions of approval.
The fourth secondary endpoint was the time to dissect adhesions and was the only clinical endpoint evaluated in the study. It was also the most objective study outcome. Unfortunately, there was no reduction in operative time for REPEL-CV patients whatsoever. The sponsor claimed that the study was under-powered to detect an up to five-minute decrease in operative time, and thus it was possible that the device did reduce operative time by five minutes, but that the study had failed to detect it. Of course, with greater power, it is equally likely that an increase of five minutes in operative time for the REPEL-CV group would have been observed. But a glance at the raw data indicates how unlikely either scenario is. In a slide that the FDA, for reasons unclear to us, chose not to present (see Figure), it is clear that the distribution of dissection times was nearly identical between the two study groups.
Figure: Histogram of the number of patients with a given dissection time.
These negative clinical data troubled many of the panel members. Dr. Hopkins, supporting his contention that the device was effective, said, “at some level, [reduction in severe adhesions] should be reflected in the overall outcomes…the logic appeals to me as a surgeon.” It is unclear how this “logic” trumped data revealing no clinical significance. Dr. Mark Katz commented, when asked whether he would use the device in his own surgical practice, “I’m not convinced from this that [adhesions were] significantly reduced, from a functional standpoint.” Both surgeons later voted to approve the device.
In addition to these efficacy concerns, there were signs of possible dangers associated with the device. There were trends toward higher risk of death (16.4% vs. 13.0%, or 12 vs. 9 patients), mediastinial infection (5.5% vs. 1.4%, or 4 vs. 1 patient), and adverse events possibly, probably or definitely related to the study (8.2% vs. 1.4%, or 6 vs. 1 patient). The consistent direction of the adverse effects observed, even if non-significant, is concerning. Dr. James Neaton felt that “there is, as a consequence of the sample size for this study, substantial uncertainty about the safety … the absence of a difference in p-value in a study like this is meaningless just because of the power issues.” Dr. Michael Domanski said, “My concern is that this trial was inadequately, may have been inadequately designed to assess safety. And if that’s the case, then you know, you just don’t have a trial.” In sum, all that the study can assure us is that there is not more than a three-fold increase in mortality or more than a 10-fold increase in mediastinitis. How can this be considered reasonable assurance of safety?
It was clear to most of the panelists that the device should not be used in adults because of their increased comorbidities and different immune systems, as well as the lack of data upon which to assess REPEL-CV’s effectiveness in the adult population. But some panelists were reluctant to restrict usage to neonates alone (the only population that had been tested), based on the logic that older patients who were at risk for adhesions might not be implanted with the product. Given that the mean age of patients in the study was 12 days at initial operation, other panelists were very clear that data about device effectiveness only extended to infants with anticipated reoperation. Dr. Jeevanandam said that, as a surgeon, he would only use the device in neonates in whom there was an expectation of reoperation within six months. Dr. Domanski agreed, saying that “the indication should be limited to very early in life. I mean, we can argue about whether it’s neonate or not, but I think pediatric is too broad.” Dr. Weinberger, immediately after voting to approve, said, “In terms of youths and adults, I think that given the magnitude of the benefit and given the relatively limited likelihood of re-operation, I need to have a much more efficacious device to even want to proceed to ask that question.” Despite their reservations, these panelists all voted to approve the device with an FDA-defined pediatric limit of 21 years. Of course, surgeons are free to use the product, if approved, in older patients in whom it has never been adequately tested.
Summary of Votes
In reviewing the committee’s votes, it appears that at most three of the eight panelists who voted to approve the device with conditions did so wholeheartedly. Not one of the panelists made a statement concluding that the device was either very efficacious or very safe. Dr. Katz could not name a single population in which he would use REPEL-CV, yet still, inexplicably, voted to approve the device. Dr. Jeevanandam, despite saying repeatedly that the device was only appropriate in neonates, voted not only to approve the device, but also for the specific motion that REPEL-CV was appropriate for the FDA-defined pediatric population of under 21 years of age. Drs. Weinberger, Domanski, and Neaton all voted against the motion defining the patient population as under of 21 years of age because they felt the data were only narrowly applicable to neonates. Nonetheless, these three members later voted for approval of the device with the under-21 restriction, presumably reasoning that access for a larger age group than they wished was preferable to no access at all. Of the remaining three panelists, Dr. Hopkins felt the device had limited efficacy while Drs. Kenneth Kahka and Eugene Blackstone made few assessments during the meeting. Thus, of the eight panelists voting to support the approval package, two appeared to contradict the panelist’s prior statements and three would have preferred a more restrictive age indication.
It is also noteworthy that six panelists were in fact consultants rather than voting members of the advisory committee. All six consultants voted to approve the device. Thus, this product would not have received a favorable vote without these consultants.
REPEL-CV bears an eerie resemblance to another device, Intergel, an anti-adhesion device intended for the pelvic cavity that also demonstrated reduced adhesions without clinically validated endpoints. That trial was also underpowered for safety and showed a consistent but non-statistically significant increase in infection rates. Initially, the FDA rejected the sponsor’s application, but the company appealed to an external Dispute Resolution Panel that recommended approval. The FDA then reversed itself and approved the device. On April 16, 2003, less than two years after the device was approved, the company removed the product from the market due to dozens of reports of post-operative pain requiring repeat surgery, foreign body reactions and tissue adhesions, including three deaths. This history should give one pause before once again approving an adhesion barrier with only surrogate endpoints and a questionable safety record.
We are left with an unblinded trial that failed to prove the efficacy of the product for two of its three intended indications, failed to generate any evidence of clinical effectiveness and showed worrisome, though not statistically significant, trends toward increases in mediastinitis and mortality. At most, only three panelists logically supported the final approval vote in its entirety. As is often noted, these committee recommendations are indeed advisory; they are reversed by the FDA 28% of the time. Given the weak or contradictory nature of most of the supporting votes we urge the agency to overrule the advisory committee recommendation and save the American public from exposure to this unproven and possibly dangerous device.
Peter Lurie, M.D., M.P.H.
Sidney Wolfe, M.D.
Public Citizen Health Research Group
 21 CFR 860.7(4)(c)(1).
 FDA, Transcript of the Circulatory System Devices Panel. September 19th, 2007, p. 101. Accessed December 20, 2007. http://www.fda.gov/ohrms/dockets/ac/cdrh07.htm#circulatory. Under questioning later in the day, he claimed that the Vicryl sutures were not visible after all, and any possible “tags of tissue” he thought he saw were simply due to his “hyperacuity regarding these patients.” It is not clear why he retracted his statements.
 Of the three panelists who voted against the device, Dr. John Hirschfield did so because he felt the device did not meet the standard for approval; Dr. Richard Page felt the indication should not be expanded beyond neonates, and Dr. John Somberg felt that the post-market study should have been expanded to adults.
 The vote protocol was conducted as follows: first, the panel agreed to approve the device with conditions. Motions were brought by panelists imposing conditions for approval. A motion, once approved, became part of a package of conditions that were then voted on in bulk. Thus, a panelist could object to motions that passed, but still approve the device despite objecting to some of the conditions.
 Sullivan MG. Intergel sales halted pending investigation of deaths, pain. Ob/Gyn News, May 15, 2003.
 Tapley AL. Lurie P, Wolfe SM. Suboptimum use of FDA advisory committees. Lancet. 368: 2210.