FDA Panel Recommends Against Depression-Treatment Device

The Food and Drug Administration's Neurological Devices Panel recommended in late January against approval of a transcranial magnetic stimulation (rTMS) device for treatment of major depression. Although panel members believed the device to be safe, they questioned whether any benefit from the treatment was clinically significant.

The device, called the NeuroStar TMS Therapy System, is manufactured by Neuronetics Inc.

In a statement the company released the day after the hearing, Neuronetics President and CEO Bruce Shook said, “We are confident in the efficacy of our NeuroStar TMS Therapy system, and we look forward to working with the FDA to resolve any remaining questions [the agency has regarding efficacy].”

Neuronetics had sought FDA approval of the rTMS device through what is called a “510(k) submission.” A 510(k) asks the FDA to approve a device on the basis that it “is substantially equivalent” to a“ predicate device”—that is, a similar device that is already approved for marketing and use in the same or similar clinical applications.

Neuronetics asserted in its 510(k) application that the rTMS device“ is substantially equivalent” to electroconvulsive therapy (ECT) devices for treatment of major depressive disorder, although rTMS and ECT devices have “technological differences that include modifications in design, materials, and in energy source.”

FDA neurophysiologist and group lab leader Victor Krauthamer, Ph.D., reminded panel members that the regulations governing a 510(k) submission require that any technological differences “pose no new types of questions regarding safety and effectiveness, and the package submitted by the sponsor must contain clinical evidence of a comparable risk-to-benefit profile.”

Indeed, the comparison of efficacy between the TMS system and ECT would prove to be Neuronetics' downfall.

The company submitted data from three clinical trials that it believed demonstrated that the rTMS system met the regulatory requirements of a 510(k) submission.

Judy Ways, Ph.D., Neuronetics' vice president for regulatory affairs and quality assurance, told the panel that it is important to remember that a 510(k) submission “does not require equal risk and equal benefit. The new device could be safer but less effective than the predicate device, or the new device could be less safe but more effective. The NeuroStar risk/benefit profile compares quite favorably to ECT.”

“NeuroStar safety is superior to ECT,” Ways told the panel.“ NeuroStar is effective, but less so than ECT. However, NeuroStar benefit is sustained and at least as durable as ECT.”

Mark Demitrack, M.D., the company's chief medical officer, described Neuronetics' Study 101. This was randomized, double-blind, sham-controlled trial lasting six weeks and included 325 patients, 155 of whom were randomly assigned to receive active rTMS treatments. The remaining 146 patients were randomly assigned to go through a “sham TMS” treatment: the device had inactive, nonmagnetic treatment coils.

The change in the total score on the Montgomery-Asberg Depression Rating Scale (MADRS) was defined as the a priori primary outcome measure. Nine secondary outcomes were also measured, including the Hamilton Depression Rating Scale for Depression (HamD).

At both four and six weeks, patients in the active and sham groups showed significant improvement; however, the difference between the two groups just missed meeting the test of statistical significance, where a p-value must be 0.05 or smaller.

At four weeks the difference between the two groups had a p-value of 0.057, and at six weeks the p-value was 0.058. Thus, Study 101 was judged a failed trial—it did not support the hypothesis favoring the efficacy of rTMS.

The majority of the secondary outcomes measured in Study 101 supported the efficacy of rTMS and met that magical threshold of statistical significance with p-values much less than 0.05. Yet many panel members were unable to move beyond the failure of the study's primary outcome to achieve statistical significance, even after an outside expert in biomedical statistics told the panel he agreed with the panel's own statistical expert that “the difference between a p-value of 0.057 and 0.050 is virtually clinically undistinguishable.” Both statistical experts appeared to believe the p-value was close enough to not represent a concern.