Medical Device Daily Washington Editor
Computer-assisted detection (CAD) of abnormalities on radiological scans may or many not prove superior to careful scrutiny by an experienced physician, but the technology is there, and FDA has responded with a draft guidance for this class of devices.
The guidance was more than a year and a half in the making and was fed in large part by discussions held last year by the radiological devices advisory committee (Medical Device Daily, March 6 and 7, 2008), where a member of the FDA team noted that the agency may need two full years to draft the guidance. During that meeting, FDA heard from clinicians that a trial should include an enriched data set despite the prospect of a heightened rate of false positives, and industry heard from FDA that a trial of CAD for lung scans may have to control for variations in the size and location of any neoplasms appearing in the lung.
The Oct. 21 guidance states that FDA sees devices used during surgical procedures as a different kind of device and hence not subject to the conditions described. Also exempt are units designed to offer a specific diagnosis rather than to just aid in detection of abnormal tissues. Hence, while a CAD algorithm designed to "identify and prompt microcalcifiation clusters and masses on digital mammograms" would fall under the scope of the draft guidance, a system that detects the same types of masses but which are "intended to provide information beyond identifying, marking, highlighting or in any other manner direction attention to portions of an image" would not. A machine that would eliminate the need for interpretation by a clinician is also not covered.
Perhaps not surprisingly to anyone in the business of making CAD systems, FDA indicates it would like to see a trial that deals with the "physical characteristics" of the markings generated by the system software, including the size and shape of the marking as well as the type of boundary the marking employs. This could include a solid or dashed line as well as a circle and an isocontour. The user's knowledge "of the types of abnormalities" the system is built to detect and highlight is another factor FDA would like to be controlled in a trial.
FDA indicates also that it is amenable to the use of a stress test to evaluate a CAD system's performance. Such a study would draw patient data that is enriched "with patient cases that contain more challenging imaging findings ... than normally seen in routine clinical practice."
FDA also indicates a preference for a multiple reader-multiple case study that employs more than one modality for reviewing the images. A fully crossed study, which would have all readers look at all the cases, "offers the greatest statistical power for a given number of cases," FDA notes, but the draft adds that the agency would not necessarily object to a study that is not fully crossed when "interpretations of the same patient data by multiple clinicians may not be feasible."
As for the control arm, FDA recommends that sponsors employ the same scanning equipment on which the CAD is based when the application is for a PMA, but if the application is for the CAD as a second reader rather than as the primary or sole reader, "another possible control is double reading by two clinicians," FDA says.
NIH technologies available for license
The National Institutes of Health (NIH) recently offered another batch of new technologies for licensing, including a very timely offering of attenuated virus material from H1N1, the swine flu virus. However, a product of potentially greater epidemiological impact is one that would offer a biological pacemaker for the human heart.
One of the offerings deals with a potential cancer therapeutic. According to the Oct. 16 edition of the Federal Register, NIH is offering human monoclonal antibodies that are reactive for B-cell chronic lymphocytic leukemia (B-CLL). NIH states in the FR that this strain of antibodies "was generated from a time point after transplantation at which antibodies to B-CLL cell surface antigens peaked, thus indicating its therapeutic value."
NIH notes that the only available potential cure for B-CLL is allogeneic hematopoietic stem cell transplantation, which entails a search for a compatible donor that fails more than half the time. The current generation of anti-carcinogenics, rituximab and alemtuzumab, are both said to hit normal B-cells as well as their malignant replicas. The antibody can be used to develop therapeutics and can be leveraged for development of "methods to inhibit the growth of malignant B-cells." The agency states that the antibodies are in preclinical development and that the potential market for such a therapy is $30 billion next year "largely driven by technological evolution from chimeric and humanized to fully human antibodies."
The offering related to swine flu is for "a recombinant attenuated vaccinia virus, MVA," which NIH states expresses the haemagglutinin (HA) and nucleoprotein (NP) of H1N1 and has demonstrated the ability to offer protective immunity to the swine flu in mice. The announcement offers little else other than to list references in the literature and to note that the product is "fully developed."
As for the biological cardiac pacemaker, NIH seeks a collaborative partnership to develop an early-stage product consisting of cardiac cells or "cardiac-like cells" derived from embryonic or mesenchymal stem cells that would "naturally integrate into the heart." NIH states that the cells would provide "coupling factors" that would restore normal function to two electrophysiological functions in the human heart, one of which deals with calcium ion electrical potentiation.
Mark McCarty, 703-268-5690