One year ago we reviewed the status of transcranial magnetic stimulation for depression, and concluded with the following TCPR Verdict: “TMS for depression: Approval is highly unlikely.”
We were wrong. On October 10, 2008, the FDA approved the Neuronetics Neurostar TMS machine for the treatment of patients with major depression who have failed one prior antidepressant trial.
Did they make the right decision? Or is this a replay of the embarrassing approval of vagus nerve stimulation, which is still not covered by insurance a full 3 ½ years after its approval in July 2005? And if they did make the right decision, how feasible will the device be for office-based psychiatrists?
Efficacy The Neuronetics FDA application was based on a placebo-controlled clinical trial of 301 patients with various degrees of treatment-resistant depression. These patients were randomly assigned to either fast frequency TMS delivered to the left dorsolateral prefrontal cortex (see this issue’s interview with Dr. John O’Reardon for a good explanation of all the TMS jargon) or to sham TMS. As we covered last year, the results were not impressive, with a four week response rate of 18.1% on TMS vs. 11% for sham. This difference was statistically significant, as were most of their other results using various outcome scales, but the clinical significance of an 18.1%response rate is debatable. In fact, two years ago, when an FDA advisory panel looked at this data, the experts were so unimpressed that they recommended that TMS not be approved.
So what happened? Why did the FDA change its mind? Because Neuronetics reanalyzed their data, focusing specifically on the 164 patients who had failed a single prior antidepressant trial (the larger data set included patients who had failed up to four antidepressant trials). Looking specifically at these less treatment-resistant patients, the six week response rate for active TMS was 27.3% vs. 10.5% in the placebo group (see poster with this data at http://www.neuronetics.com/Clinical_Significance_Thase.pdf). Among patients with more than one prior antidepressant trial, there was no significant effect of TMS (Lisanby SH et al., Neuropsychopharmacology 2008 August 13, advance online publication).
Thus, when the FDA saw that patients who had failed only one prior trial responded at rates approaching three times the rate of placebo, it relented, and awarded the approval only for these less treatment-resistant patients. While we can appreciate why the FDA was more impressed with these new numbers, it’s important to note that this reanalysis was done post-hoc, that is, it was conceived after all the results were in. These post-hoc analyses are exploratory by nature, and never conclusive. The appropriate scientific next step would be to conduct another trial and hypothesize a priori that those who have failed one previous trial will do the best. What are the chances of the company doing such a study, now that they have FDA approval? Can you spell “infinitesimal”?
Nonetheless, we view the treatment more positively than in the past largely because newer studies of TMS for depression show more benefit than did those on which we based our earlier negative impression. In one meta-analysis, for example, the results of older TMS studies (those published up to January 2002, including 13 studies and 324 patients) were compared with the results of newer studies (December 2005 to November 2006, five studies, 274 patients). While the older studies showed a moderate effect size of 0.35, the newer studies yielded a combined effect size of 0.76, which is considered large (Gross M et al., Acta Psychiatr Scand 2007;116:165 173). The authors suggest that later studies used more TMS sessions and larger sample sizes.
Recently, a particularly informative metaanalysis was published in a British journal (Schutter DJLG, Psychological Medicine 2008 Apr 30 early online), and reviewed 30 well designed sham controlled trials of TMS. All were trials of patients with non-psychotic depression, and all randomly assigned patients to either high-frequency TMS of the left dorsolateral prefrontal cortex (the LDPFC) or sham TMS.
A total of 1164 patients with major depression were enrolled in these 30 studies, of which 606 received real TMS and 558 received sham. The author reported that the overall weighted mean effect size for TMS treatment was 0.39, a moderate effect size. Such an effect size is comparable to many placebo controlled antidepressant trials, so if you can believe these studies, the treatment looks fairly impressive.
However, an alternative viewpoint is that this moderate difference between active and sham does not represent a true treatment effect, but rather a placebo effect. While all the studies tried to control for a placebo effect by randomly assigning patients to a sham condition, it is not clear whether patients were always fooled. This is because real TMS produces a clicking sound and causes scalp sensations, whereas sham TMS generally does not. If patients in the real TMS condition were able to guess their assignment, they might have felt more hopeful than those in the sham group, and any difference in depression scores might have been due to expectancy effects alone. Compounding this problem is the potential for investigator bias. In most studies, the researchers actually administering the treatment knew when they were delivering real or fake TMS, and they might have unconsciously communicated their enthusiasm for the new treatment (and lack of enthusiasm for the sham treatment) to patients.
The Schutter meta-analysis reported that of the 30 trials analyzed, six took measures to check the adequacy of the blind, generally by asking patients to guess which treatment they received. In five of those six trials, patients were unsuccessful in guessing their treatment condition, which is reassuring.
What about real life use of TMS? How feasible is it? Unfortunately, the economics of the device remain murky. I called Neuronetics to ask how much the Neurostar machine would cost me if I wanted to set up shop in my office, and I was told that the company “does not have a policy of issuing price quotes publicly.” They are planning a “slow roll out,” primarily to psychiatrists who already have experience using the device, for example, by having participated in clinical trials.
Reading between the lines, there are some delicate politics involved here. In order for the Neurostar to become successful, insurance companies are going to have to decide to pay for it. They will only pay if they are convinced that it actually works. Presumably, the Neuronetics strategy is to provide the device to psychiatrists who are more likely to achieve good results, and these would be those who have used the device before.
How much are insurance companies likely to reimburse psychiatrists for a typical 40 minute treatment session? The company is hoping for close to $400 per treatment session, which is the rate at which their researchers claim that TMS is as cost-effective as standard anti-depressant treatment (see http://tiny.cc/ENlee for their analysis). If insurance companies believe this analysis, they might reimburse at close to this rate. This may sound like a lot for 40 minutes of work, but you have to factor in various hidden costs. For example, you are going to have to gradually pay off the upfront cost of buying the Neurostar, which may be anywhere from $50,000 to $100,000. You’ll need a room for the device, and you’ll probably want to hire a nurse or a medical assistant to actually administer the treatments. Otherwise, you’ll be spending your days placing magnetic coils on heads and pushing buttons – probably not the work you were hoping for when you went into psychiatry. In addition, you’ll have to factor in another expense: about $100/session for a gadget called the “SenStar Treatment Link.” These are replaceable sensors that the company says are necessary to ensure good contact with the head and to mitigate scalp discomfort. That may be true, but it’s also a lot of extra income for the company – and less money for you.
But the really crucial question, and the one upon which the ultimate success of the Neuro Star device will hinge, is whether psychiatrists will refer patients who have failed only one antidepressant trial for TMS. According to the Star-D trial, the response rates of patients switching from a failed SSRI trial to another antidepressant were on the order of 30%, similar to the TMS 27.3% response (see TCPR May 2006). Therefore, it would make sense to try at least two, if not more, antidepressants before trying TMS, which is an inconvenient treatment that requires office visits five days a week for three weeks, and then on a tapering schedule for three additional weeks. If patients have not responded to several trials, then it might be reasonable to go to the next level of treatment. But therein lies the rub. Once patients have been through several meds and are ready to accept a trial of TMS, they fall into the category of highly treatment resistant patients who are unlikely to respond.
TMS: Probably as effective as meds, but feasibility is highly questionable.