The Danger Within Us

by Jeanne Lenzer

  It’s clear that Braunwald was not unaware of the risks of the device. Even in his original 1967 paper, he’d noted several potential dangers of vagus nerve stimulation, including nerve damage, “cardiac standstill or other dangerous arrhythmias,” and the perils present in any surgical intervention. But in his later discussions of the VNS device, he offered few if any cautions about its potential dangers.

  Braunwald’s discovery was rapidly supplanted by coronary artery bypass surgery for coronary heart disease. So in time, Braunwald abandoned his work with the stimulator. But the papers published by this widely admired researcher about the potential of the VNS device remained in the medical literature, waiting to be rediscovered by some other ambitious scientist.

  The moment came in the 1980s, when Jacob Zabara, PhD, a neurophysiologist at Temple University, speculated that stimulating the vagus nerve might interrupt seizures.117 Zabara says the idea occurred to him when he saw his wife using breathing techniques to control labor pains during a Lamaze class. He wondered about the mechanism of pain control. Until this time, doctors thought the vagus nerve carried signals only away from the brain, not to it. Zabara wondered whether it was possible that the vagus nerve, which stimulates the diaphragm in breathing, might be sending information up to the brain and in some way regulating pain. And if the nerve could control pain, might it also control seizures or nausea if stimulated?

  Although his line of reasoning isn’t exactly clear, Zabara filed for a patent on the vagus nerve stimulator. In his patent application, he cited Eugene Braunwald’s work with the device. In 1987, he teamed up with Reese Terry, an experienced medical device-company executive with training in electrical engineering. With money from venture capitalists and early investors, Zabara and Terry cofounded Cyberonics, with its headquarters in Houston, to produce the device for commercial sale.118 In 1990, with an infusion of money from Pfizer, they started clinical trials.119 Their sole product was the VNS device—and unless they could bring it to market, they would have no revenues, no profits, and no business. Getting approval from the FDA was essential.

  Cyberonics set about earning that approval by providing the FDA with data from a clinical trial, code-named E03, which followed 114 patients with epilepsy for fourteen weeks. Half of them received low-level electrical stimulation of the vagus nerve (designated as the “sham,” or control, arm), and half received high-level stimulation (presumed to be therapeutic).120 Bad news followed: in 1994, after reviewing the data, the FDA denied approval of the VNS device to treat epilepsy, saying that the study provided insufficient proof of benefit. If Cyberonics wanted to win approval, the company would need more test subjects and would need to study them for a longer period of time.

  The FDA decision was devastating. Cyberonics was running out of money and had already cut staff. The FDA’s refusal to approve the device looked like it would spell death for the company. The CEO at the time, Allen Hill, resigned.

  Stepping into the breach was Skip Cummins, a flamboyant venture capitalist who helped fund the start-up of Cyberonics in 1987 and had joined the board in 1988. Like everyone associated with the company, he knew that if they didn’t win approval for the VNS device, the company would be dead in the water—in Cummins’s words, Cyberonics would be “essentially a failed technology and [a failed] company.” He describes the situation at the time of Hill’s resignation:121

  [Hill] lost confidence that we would ever get approval…he resigned…we had no CEO. At the time we were running out of money. Morale was terrible. Nobody believed in the science. Nobody wanted to invest in the company; the stock was basically pink sheets. The board was looking for anyone who would put their hand up [to replace Hill]…I figured what the heck, this thing works!

  So Cummins threw his hat in the ring and was appointed CEO in 1995. He was ready to lead the charge on behalf of the upstart company: no one could bum-rush opponents better—or be more persuasive. An imposing six-foot-two-inch former college linebacker, Cummins, with his shaved head and nonstop intensity, has been described as a bully who uses his appearance and demeanor to intimidate. Even the notoriously abrasive hedge fund manager and television host Jim Cramer called Cummins the “most combative, most antagonistic CEO in America.”122 Antagonistic or not, he proved capable of crushing any opposition to the VNS device, whether scientific or regulatory.

  This kind of industry aggressiveness with FDA staff was becoming the new normal. The rise of the biotech and medical device industries in the late 1970s unleashed a stampede of companies that sold just one or two products. Large pharmaceuticals companies with multiple products can generally afford a “miss” or a withdrawn drug. But the new biotech and device companies are often one-trick ponies that can’t afford a loss: they are lean and hungry, and they market their sole prized products with a do-or-die mentality.

  Cummins was unusually well suited for the role of CEO. He later observed, “When you’re trying to commercialize a pioneering science that nobody believes in, that’s not for a shrinking violet or the faint of heart.”121 Cummins went into discussions with the FDA in full battle gear, ready to take on what he says was an unreasonable demand by the agency. He relates the FDA’s position on the first-ever medical device to treat epilepsy by stimulating the vagus nerve:

  FDA’s biggest issue was: we don’t really have a precedent for how we should approve or not approve a device for the treatment of epilepsy. We only have drugs that do that. We are going to hold you to a drug standard, which [requires] huge double-blind randomized controlled trials.121

  Cummins was incensed. The FDA’s standards for approval of medical devices are generally far lower than they are for drugs.123 For the vast majority of implantable devices, the FDA doesn’t require that clinical outcomes be tested; device manufacturers don’t have to prove that their products improve clinical outcomes or make patients live longer. Instead, as one expert commented, the only thing manufacturers have to prove is that a device does what they say it will do—that a cardiac pacemaker fires electrical impulses at regular intervals, that a cardiac stent will increase blood flow through a previously blocked artery, or that an artificial hip actually bends and rotates as a hip should. Those questions are relevant and important. But they don’t address the questions people care most about: Is this device safe, and will it help me live longer or feel better?

  Furthermore, while the FDA doesn’t require clinical data for most high-risk implanted devices, even the few devices that have undergone testing generally don’t measure patient-oriented, clinically important outcomes, such as overall deaths. Instead, 88 percent tracked surrogate markers, such as laboratory tests and symptoms. But surrogate markers have frequently failed—and failed badly.

  Consider coronary angiograms, the surrogate marker widely used to test the benefit of cardiac stents (stents are small metal tubes placed in coronary arteries to hold open arteries that are damaged or blocked by plaque). Angiograms are videos of the heart taken as dye is shot into the newly opened arteries, which often show blood squirting through previously blocked arteries where there had been little or no flow before. But as dramatic as such surrogate-marker evidence on an angiogram may appear, it isn’t evidence that the patient will actually do any better clinically, that he or she will be any less likely to have a heart attack and die.

  But because it seems perfectly logical that, if a coronary artery is blocked by plaque, and it is opened up with a stent, it would be a good thing, doctors—using this surrogate evidence—began implanting stents in millions of patients during percutaneous coronary angioplasty (PCI). The procedure seemed to work brilliantly.

  It took years, and plenty of scientific argument, before a comprehensive study was performed to examine clinically important outcomes, such as future heart attacks and deaths. The results were disturbing: for patients undergoing PCI during elective surgery (as opposed to emergency surgery during an evolving heart attack), the procedure failed to prevent future heart attacks and deaths any better than simply t
reating patients nonsurgically, with medicines.124 Disturbingly, a survey of cardiologists and patients with angina, or heart pain, found that two-thirds of the cardiologists acknowledged that the procedure would not prevent future heart attacks—yet 88 percent of patients believed it would protect them from heart attacks and death.125, 126

  Although the procedure was reported to at least reduce symptoms, Bernard Lown, a now-retired Harvard cardiologist, said, “Optimizing medicine and life style achieves the same benefit as stenting at a far lower cost.” But, he adds, “Stenting is a lucrative business.” Dr. Rita Redberg, a cardiologist at the University of California at San Francisco and editor of JAMA Internal Medicine, said, “Many stents are placed in patients without symptoms or who have not been tried on medical therapy and thus are unnecessary,” and, worse, “It is not clear that they actually reduce symptoms.”127–130 Citing the major study that compared PCI to treatment with medicines only, Redberg said that the small benefit in terms of chest-pain symptoms was gone after a few years131, 132 and that percutaneous coronary interventions have never been tested in a blinded trial, so it is not possible to know if even the temporary benefit of pain relief is simply attributable to the placebo effect of the procedure. Compared to nonsurgical therapy, Redberg said, percutaneous coronary interventions “expose individuals to the harms of an invasive procedure with radiation, which increase[s] cancer risk, and [a] contrast agent, which poses threats to kidney function.”

  Vikas Saini, a cardiologist and president of the Lown Institute, says that researchers with the American College of Cardiology found that only half the patients who underwent stenting were deemed “appropriate,” and while many of the other half were said to be of “uncertain appropriateness,” financial and cultural incentives “completely align for almost all of them to get stented with no real incentives not to do it.”

  Just as stents were approved based on a surrogate marker of vessel openness on angiography, so Cyberonics sought approval of the VNS device on the basis of a reduction in seizures—also a surrogate marker. FDA reviewer Steven Piantadosi, professor of oncology and biostatistics and a clinical-trial methodologist at Johns Hopkins, commented on the unreliability of using a surrogate marker as a measure of success, especially in light of what he called the “high number of deaths” (seventeen of first one thousand) among VNS patients.133 This was critically important, because Cyberonics emphasized the problem of deaths from epilepsy, with the implicit suggestion that the VNS device might reduce that risk by reducing seizures. In the official transcript of the FDA approval hearings, Piantadosi observes:

  [O]ne of the things that’s concerning me is that the endpoint being measured in all of these studies is, in some sense, a surrogate: counting the number of seizures. I realize that to the patient and to others, it is a very important endpoint, but it may not be as definitive as some other things that we could measure. There are numerous examples in the methodologic literature about the weaknesses of accepting clinical trial data based on surrogate outcomes…I would point to, as a recent and a very dramatic example, the cardiac arrhythmia suppression trial, in which the study was designed and the endpoint was selected on the basis of looking at arrhythmias and suppressing them with a drug.

  And the studies originally seemed to show that the drug was effective in suppressing arrhythmias. The problem was that it was so good in suppressing arrhythmias that it was killing people, and the mechanism was not understood until much later and wasn’t even believed until the results of the randomized trial.

  So, I am very nervous when I see high mortality rates associated with a supposed benefit, even though we don’t have a way biologically right now to connect the two. So, that is why I have harped on this this morning and why I am still very nervous with this high death rate.…

  The comparison of the VNS device in the treatment of seizures with anti-arrhythmic drugs in the treatment of abnormal heart rhythms is apt: both seizures and certain heart-rhythm disturbances are associated with an increased risk of death. Yet the CAST (Cardiac Arrhythmia Suppression Trial) showed that correcting heart-rhythm disturbances with certain drugs actually increased the risk of harm and death.3 Could the same be true for the VNS device? Only a study examining overall deaths could provide the answer. But that was another weak link in the way the VNS studies were designed: the device was never directly tested against a group of patients treated with medicines. Instead the studies measured only two different levels of electrical shocks, and one was likely to do better than the other by sheer chance. If the high-stimulation group did better, then it was a winner. If the low-stimulation group did better, the company could claim that the higher dose was just too high.

  It’s not even clear that individuals who did experience fewer seizures had the VNS device to thank. There are several reasons why patients enrolled in clinical trials tend to improve over their baselines. One is that during a trial they generally receive intensive care and oversight that’s not part of “usual and customary” care. Another reason is “regression to the mean,” which occurs in patients with chronic conditions that wax and wane (recall that Fegan’s seizures varied enormously from month to month). Because patients tend to improve after periods of doing poorly, and because patients tend to enroll in clinical trials only when they are doing poorly (rather than when they are doing well), they are likely to improve simply because of the passage of time—in other words, they “regress to the mean,” or average severity, of their condition. Then there’s the placebo effect, which can also cause a patient to improve over baseline. One simple fact remains: what can’t be determined from the Cyberonics studies is whether patients implanted with a VNS device will do any better (or worse) than they would if they were treated with medicines.

  Although the FDA had set a low bar for Cyberonics by allowing seizure counts to serve as a metric of success without requiring patient-oriented clinical outcomes, CEO Skip Cummins was nonetheless seething.121 He knew every detail of the FDA approval requirements, and he knew that plenty of implantable devices had gotten passes without having to conduct a single randomized trial, much less two. And he was worried that the cash-strapped Cyberonics would be unable to find funders to pay for a second study. The company was at its lowest point.

  Despite seemingly insurmountable obstacles, Cummins charged ahead. He managed to persuade St. Jude Medical to invest $12 million in the company with an option to buy.134 Cummins put the cash infusion from St. Jude to work in a second trial, dubbed E05, and the company included the results of E05 in a new submission packet to the FDA.

  The E05 results, however, were far from impressive.135 In fact, St. Jude Medical, which had made an offer of $72 million for the company, decided to withdraw the offer after the new results were revealed.

  The E05 study divided 199 individuals with epilepsy into two groups and followed them for one year. As in the earlier study, E03, one group received “low stimulation,” and the other group received “high stimulation.” The high-stimulation group had an average 28 percent reduction in seizures compared to the low stimulation group, which experienced a 15 percent reduction. The difference didn’t reach statistical significance, according to Piantadosi.133 Yet somehow Cummins managed to persuade the FDA to approve the device. He returned triumphant from the 1997 meeting with the agency: the FDA had granted Cyberonics approval to market the VNS device for individuals with treatment-resistant epilepsy.

  There was just one small glitch: the approval was given conditionally because of concerns about the high number of deaths, as noted by Piantadosi. Under conditional approval, the FDA can order a device off the market if the manufacturer fails to provide postapproval studies showing it to be safe. In other words, despite concerns that the device could be killing people, the company was able to sell it and attempt to prove its safety after it was on the market.

  But that was a tiny bump in the road for Cyberonics. It won the approval it had been seeking, and it was now free to market the VNS device to doctors a
nd the public at large. And since the FDA didn’t require that the company disclose to doctors or patients that the device was effectively on probation because of concerns about high death rates, the public remained in the dark. Only the company and a handful of people at the FDA would know.

  The 1997 triumph for Cyberonics was in no small part due to the research Dr. Eugene Braunwald had done more than thirty years before, as the first expert to use vagus nerve stimulation to treat a medical condition. As I noted, Braunwald’s pioneering work on the VNS was cited in numerous patent applications filed by Cyberonics. Unfortunately, the two deaths among just four patients that Braunwald had witnessed, along with his initial cautionary remarks, appear to have been forgotten by the time the FDA gave its blessing to Cyberonics.

  Skip Cummins was thrilled. He later told Forbes reporter Robert Langreth that the VNS device is “a gigantic opportunity. We are talking about some of the largest medical markets in the world.” He predicted, “Brain stimulation will be to the next 10 years what cardiac pacemakers were to the last 40.”117

  Cyberonics played up the comparison of the VNS device to a cardiac pacemaker. It was a useful marketing ploy: it evoked the image of a remarkably successful device. But the comparison is misleading. The cells of the heart react to the electrical stimuli of a pacemaker by contracting. It’s a simple cause-and-effect response. The brain, on the other hand, is an exquisitely complex organ with myriad feedback loops. Twelve cranial nerves, of which the vagus nerve is one, transmit impulses to and from the brain. Each nerve interconnects with, responds to, and can affect countless other nerve cells that allow us to see, feel, touch, hear, and think. The crude application of an electrical impulse to a large nerve with multiple important functions—from regulating digestion and heart rate to providing feedback to the brain from virtually every major organ in the body—and expecting to reliably achieve a specified desirable outcome is not quite the same as applying an electrical impulse to heart cells that have a single simple function: to contract.