I read each study carefully and was surprised to find that not one of the five studies included any data on deaths.
I immediately contacted the FDA, thinking that surely it must have additional data to prove the device was safe. But the FDA sent me the exact same five studies Cyberonics had provided.
I figured I must be missing some piece of the puzzle. So I went back to Cyberonics and asked for any unpublished data it had proving the device was safe. A representative said the company had proved it was safe with the five studies that were accepted by the FDA. When I asked the company to send me the death statistics for each of the five studies, because the results were not included in the published reports, to my surprise, Cyberonics explained that it had no death data to provide because it never collected death data for any of the five studies. When I pressed the company about how it could claim the device was safe if it didn’t even know how many implanted people had died, I was told that it had collected mortality data from the Social Security Death Index—but hadn’t published the results. When I asked for those data on behalf of the medical journal The BMJ, Cyberonics refused to release its findings.100
I was stunned. Twelve years after the device was approved for epilepsy, the company hadn’t collected death data for the five studies it submitted as proof of safety and wouldn’t release the death data it did collect outside of the studies. I informed the FDA, assuming that the company would get slammed by the regulator for failing to collect death data on a device approved only conditionally because of the FDA’s concerns about a “high rate of deaths.” But the FDA continued to insist the device was safe. When I pressed the agency about this, it responded in an e-mail that it hadn’t asked the company to count the number of deaths, instead it only required Cyberonics to “characterize mortality.” I wondered: How does one characterize mortality without knowing whether anyone implanted with the device died?
I was beginning to feel as if I were in a Kafka novel.
Fegan’s finding that nearly a thousand VNS patients had died in just over a decade raised an important question: Could it be that the VNS device reduced a surrogate marker (in this case, the number of seizures) but increased the number of deaths? Could anyone tell the difference between SUDEP attributable to epilepsy and asystole attributable to the VNS device?
No one could answer the question with any certainty. The FDA’s MAUDE database appeared to be intentionally useless: it didn’t collect data on the total number of devices in use, making it hard to know whether serious adverse events were one in a million or one in a hundred. Nor were the ages of affected patients included. If a ninety-year-old patient dies three years after having a hip replacement, that’s one thing; if a twenty-three-year-old dies after a hip replacement, it could be a clue to a serious safety problem. Nor does the database include the reason a device is implanted. The FDA would have no way to know how many patients implanted with a VNS device to treat depression developed asystole, since it simply doesn’t collect those data. In an apparent lapse, Cyberonics listed the cause of death for a patient who was given a VNS device for depression as “SUDEP.” There was no mention that the patient was epileptic. If the device was causing sudden death (called SUDEP by the company) in patients without epilepsy, this would be important news, and would be additional evidence that the device, rather than the underlying condition, was causing fatal asystole.
Making matters worse, the agency ignored its own demand for safety studies, giving Cyberonics a pass when it first approved the device and then another pass when it failed to collect death data for any of the five studies it submitted to the FDA as proof of safety.
The FDA had closed its case on the matter, but I continued to search the medical literature for an answer. Eventually I came across a study claiming that death rates were “similar” between patients implanted with the VNS device and those without it—but once again, the claim was based on an historical comparison group—observations of previous patients.257 Jerry Hoffman, the expert on how medical illusions distort our understanding of research, says the study can’t provide conclusive evidence about the impact of the VNS device. “One basis for skepticism,” he said, is that “even though the authors report that the number of deaths among 1,819 patients with a VNS was ‘similar’ to the average mortality in several older studies of patients with ‘refractory epilepsy,’ there’s no way to know if these ‘control patients’ were sufficiently similar to the VNS cohort to make any fair comparison.” The authors acknowledged that they didn’t know critical details about the patients that would tend to drive mortality up or down independently of the VNS device. They didn’t have information on the patients’ ages, seizure frequency, or causes of their seizures—each of which can have a powerful effect on mortality. With regard to the study’s methodology, Hoffman said, “The devil is in the details, and there are far too few details provided.”
Without a fair comparison, it’s impossible to know if apples were being compared to apples. Many early VNS test subjects were included only if they were deemed to be “healthy”; some of the studies specifically excluded patients with a host of illnesses. The authors said they relied on the study sponsor, Cyberonics, to supply the death reports, writing that they “had no reason to assume” the company didn’t supply all relevant data, an assumption that ignores the long history of selective reporting of results by Cyberonics and by industry in general. But one thing was clear, they weren’t claiming a mortality benefit—only that the device didn’t make things worse in terms of mortality.
Even some researchers funded by Cyberonics have grown disillusioned with the company’s claims. Christian Hoppe, a neuroscientist and epileptologist at the University of Bonn, in Germany, had initially written about the benefits of the device.258 However, as he analyzed his subsequent research findings and that of other researchers, he concluded that “the VNS literature” is replete with “bad science.”259, 260 He and his colleagues conducted a study of “therapy-resistant patients” who were treated with the “best drug therapy” with and without the VNS device implanted and found that patients with a VNS device fared slightly worse than patients treated with only drug therapy.261 Their finding was consistent with two studies conducted by different researchers that also found no seizure-control benefit of the VNS device compared to simple drug therapy.
Another researcher on a Cyberonics study who was disturbed by the company’s questionable scientific methods is Howard Barkan, formerly adjunct professor at the School of Psychology and Interdisciplinary Inquiry at Saybrook University.* Barkan served as the statistician on a Cyberonics cost-benefit analysis study of the VNS device. The company claimed that the study proved the VNS is a money-saving investment that reduces long-term costs associated with epilepsy because it helps patients control their seizures—a necessary claim in light of the $40,000 price tag.
But Barkan says, “There was a systematic bias in what [the] Cyberonics folks did.” When all the data were in, Cyberonics wanted to exclude a particular VNS patient from analysis as an “outlier.” The patient had seized and had a car crash. The hospital bill was in the vicinity of $150,000, which made the cost-benefit analysis “not look so good” when the patient was included in the analysis.
Barkan “wrestled and argued for months” over the exclusion, fighting the company’s decision “tooth and nail.” Finally, seeing no end in sight, he says he agreed to have the patient excluded on the condition that the case would at least be reported in a footnote. The authors submitted the paper, and Barkan says he was told the paper was rejected for publication. He forgot about it until several years later, when he saw that the article had been published with his name as a second author.262 Barkan was infuriated: not only had Cyberonics thrown out the one “outlier” patient, it also, he says, “knocked out a subset of deaths that they didn’t like.”262
Barkan insisted on having his name retroactively withdrawn from the paper.
In the end, however, the researchers on this Cyberonics-
funded study got to decide whom to select in its research, whom to exclude, and when and where the results would be published. Two Cyberonics employees conducted the statistical analyses, and the company acknowledges in the text of the article that three patients considered “outliers” incurred high medical costs—and yet all three were excluded from the final analysis, making the VNS device look cost effective.* As a result, Cyberonics can honestly claim that, according to published reports, the VNS device works, is safe, and is cost-effective. Doctors who read Cyberonics-sponsored research reports would have no idea about the contrary evidence swept under the rug.
* * *
The brick wall that Dennis Fegan and I ran into when we tried to investigate the safety problems with the VNS device is not unique to that device or to its maker, Cyberonics. Real information about the safety of implanted medical devices is difficult to unearth—even for people in a position to know.
Consider the case of Steve Tower, an orthopedic surgeon who was implanted with a medical device himself and suffered terrible consequences. Tower was far better placed than most of us to know about the true benefits and harms of the device, yet even he didn’t have the information needed to understand the truth.
Tower, who specializes in complex artificial hip replacements and repair, knew in 2006, after five years of progressive pain in his right hip, it was time for him to get a hip replacement. Artificial hips have provided significant relief to millions of patients worldwide, freeing them from pain, allowing them to be active, and improving their overall health and independence. The implants have, until recently, been celebrated as “one of the most successful operations of the 20th century.”263
Tower chose what he believed at the time to be the best artificial hip on the market: the ASR XL metal-on-metal hip, introduced by Johnson & Johnson’s DePuy division the previous year. The device had been cleared for sale under the 510(k) pathway, which meant the company claimed its artificial hip was simply a modification of a device already on the market. No clinical trials were required.
Tower had his ASR hip implanted in March of 2006. He did well at first. He was able to resume his hobby of long-distance bike riding. But about a year after surgery, things began to unravel. The pain returned and became progressively worse—as bad as it had been before the implant.
Troubled, Tower began to wonder if the ASR hip, which contained cobalt and chromium, might be leaking the metals, which are known to cause intense local inflammatory reactions that can damage tissue.264 He obtained a cobalt level on his blood: it was two hundred times higher than normal. He contacted the manufacturer about possible complications from cobalt, but he says DePuy’s design surgeons, engineers, and field representatives reassured him, “No worries…cobalt is harmless.” Besides, they said, no other surgeons were reporting similar problems.
Over the following two years, the pain became excruciating. Tower’s hip joint would grate and squeak with every step. He began having problems with his memory, vision, mood, and sleep. He developed ringing in his ears, and he could no longer bicycle any distance, because he became breathless with exertion. An echocardiogram revealed that he was in the early stages of heart failure.
Finally, in November of 2009, Tower couldn’t take it anymore, and he had the DePuy ASR XL chrome-cobalt parts taken out and replaced with a new ceramic-on-plastic socket. When his surgeon opened up Tower’s hip, he discovered what looked like a crankcase of dirty automotive oil. The tissues surrounding the hip were black. His surgeon said some of the ligaments had turned to “mush.” Tower had metallosis, a condition caused by metals that can produce local and/or systemic effects. Local effects, the sort seen by Tower’s surgeon when he removed the ASR hip, include tissue death and the destruction of muscles, bones, and ligaments. Systemic effects include nerve damage, mental changes, thyroid disorder, vision and hearing problems, and heart failure.264–268
One month after Tower had his ASR XL hip removed, his cobalt level plunged, and his mood and memory improved. Within a year, his heart function had returned to normal. Although his hip continues to dislocate at times because of damage to his ligaments, he is now feeling better.
A few months after his own corrective surgery, a patient Tower had fitted with a metal-on-metal hip complained of systemic symptoms suggestive of metallosis. Tower found the patient’s cobalt levels were high. Following removal and replacement of his patient’s artificial hip, his symptoms resolved.
Tower immediately began trying to sound the warning bell about the ASR hips and metallosis—specifically cobaltism, the condition he had experienced. But to his surprise, he ran into a brick wall at every turn. In May of 2010, he published two case reports of metallosis, his own and that of his patient, in the state of Alaska’s epidemiology bulletin.269 But the limited readership of the bulletin meant that only a few Alaskan doctors would learn of the problem. However, the Centers for Disease Control and Prevention (CDC) expressed interest in publishing Tower’s report in its own widely read and prestigious publication, Morbidity and Mortality Weekly Report.
Tower was “elated” by the CDC’s interest, but then, at the eleventh hour, according to e-mails dated late July of 2010, the CDC pulled the article, saying it was told by the FDA that it shouldn’t publish anything because the agency was “studying” the issue.
In July of 2010, DePuy/Johnson & Johnson announced that it was withdrawing the ASR hip from the market because of low sales.270 Later, the company issued a recall of 93,000 ASR hips, saying only that there was a high failure rate—12 to 13 percent—within five years. That compared poorly with other types of artificial hips, which last an average of fifteen years or more. But it was worse than the company was letting on: internal Johnson & Johnson documents showed that a whopping 37 percent of ASR hips failed after 4.6 years, sixteen times the rate for other hip implants.271
Early descriptions of metal ion leakage from implants were published as early as the 1980s and ’90s, yet the reports received little attention.272, 273 When mandatory device registries, which exist in Australia, the United Kingdom, Europe, Canada, and Japan—but not in the US—began detecting serious problems with metallosis, European authorities issued recommendations for routine screening of patients with metal-on-metal hips.274, 275 Still, as recently as 2010, the phenomenon remained relatively unknown in the US when Tower published his report on two patients.269
Tower’s seemed to be a voice in the wilderness.
Had US doctors and the public been alerted to the risk of metallosis and cobaltism earlier, it’s possible that other patients might have been spared injury. In 2010, a fifty-nine-year-old woman was admitted to a hospital with heart failure, puzzling doctors, because she had no known risk factors.276 Extensive testing failed to reveal the cause of her condition, which continued to worsen despite treatment with multiple medicines. She lapsed into end-stage heart failure, requiring two implants—first a device to help her heart pump and then a pacemaker-defibrillator. Ultimately she required a heart transplant.
It was only after her heart transplant that she—and her doctors—would learn about cobaltism associated with the ASR hip replacements. She had had both hips replaced with ASR implants several years earlier. When she was tested for cobalt, the cause of her heart failure was finally clear: her blood cobalt level was extraordinarily high. After recovery from her heart surgery, she had both hips replaced. Her cobalt levels fell, and she began to recover from her other symptoms of cobaltism, including thyroid disorder and fatigue. And the function of her new heart improved.
Even a few case reports in the medical literature are hardly adequate to warn doctors or the public about serious problems with devices. A fifty-five-year-old German man went from doctor to doctor seeking an explanation for the onset of strange symptoms, including near blindness, deafness, thyroid dysfunction, fever, and severe heart failure despite normal coronary arteries. It was only when he landed on the doorstep of Juergen Schaefer, director of the Center for Undiagnosed Diseases in Marburg,
Germany, that he would learn the cause of his problems. Schaefer, a fan of the fictional television series House, which features a cantankerous and unconventional doctor who solves mysterious medical cases, recalled an episode of the show from February of 2011 about a woman who had similar symptoms. It turned out they were the result of cobalt poisoning. Schaefer learned that the German man had had a metal-on-metal hip replacement a year and a half earlier and found that his cobalt level was one thousand times higher than normal. Following treatment, including replacement of his artificial hip with a ceramic hip, his heart function improved dramatically. Unfortunately, he had minimal improvement in his vision and hearing.276–278
When the FDA told the CDC not to publish Tower’s reports because it was “studying” the problem, it was presumably to be “sure” there was a problem before alarming the public. But reports of possible adverse events involving medical devices are crucial to uncovering dangers. Clinical trials conducted for FDA approval often use limited numbers of test subjects and are thus too small to detect serious adverse events, which is why the FDA itself describes the value of its MAUDE database as an opportunity to identify “red flags.” Recognizing a red flag can lead to a call for study to determine whether such reports represent a fluke or a genuine threat.
The question then is: What is to be done during the period after a red flag appears but before definitive studies are completed? It would seem rational to alert the public, adding the caveat that studies are pending so people can decide for themselves what steps to take. It’s unlikely that patients already implanted would needlessly have their artificial hips surgically removed—or that surgeons would do so without evidence of a problem. But for individuals who have symptoms, the information provided could be important and could help their doctors know what to look for—and that might just save lives.
The Danger Within Us Page 19