Few healthcare consumers realize what a limited role impartial research plays in shaping the medical treatments they receive. In a poll conducted by the Campaign for Effective Patient Care, a nonprofit advocacy group based in California, 65 percent of the eight hundred California voters surveyed said they thought that most or nearly all the healthcare they receive is based on scientific evidence.157 The reality would probably shock them. A panel of experts convened in 2007 by the prestigious Institute of Medicine estimated that “well below half” of the procedures doctors perform and the decisions they make about devices, surgeries, drugs, and tests have been adequately investigated and shown to be effective. The rest are based on a combination of guesswork, theory, and tradition, along with a strong dose of marketing by drug and device companies. Distinguishing false from true medical claims is a serious challenge; that’s why the way studies are designed and carried out is so important. As FDA historian Suzanne White Junod observes, “The function of the formal controlled clinical trial is to separate the relative handful of discoveries which prove to be true advances in therapy from a legion of false leads and unverifiable clinical impressions.”147 The reality of medical science is that the vast majority of highly touted, seemingly promising cures—whether they take the form of new drugs, new therapeutic regimens, new surgical techniques, or new medical devices—simply don’t work.
This is not a message that the ambitious, profit-oriented corporate leaders of the medical-industrial complex are eager to embrace. And so it’s no wonder they have used their power and influence to resist attempts by medical researchers and some government agencies to challenge their claims or insist on more reliable scientific evidence to ensure that their products are safe and effective before they are approved by the FDA.
The emergence of fast-growing, ambitious companies like Cyberonics in the 1970s, ’80s, and ’90s posed significant challenges for the US medical regulatory system. Agencies like the FDA are supposed to safeguard the health and safety of the public. Yet given the complexity, size, wealth, and power of today’s medical-industrial complex, that’s a nearly impossible task for an underfunded agency subject to control by politicians beholden to industry.
Theoretically, the system for ensuring the safety of drugs and devices after they are approved for sale is supposed to work this way: The FDA receives reports from doctors, manufacturers, and hospitals in its adverse-events database. If an unusually high number of problems is reported—“red flags,” as the agency puts it—the FDA can request a scientific study to determine whether a drug or device is causing a problem. But as Dennis Fegan began to discover by studying the MAUDE database, the FDA’s red-flag system is hamstrung by a series of problems.
First, as the FDA itself acknowledges, its assessment of medical device safety is hampered by manufacturers and hospitals that submit “incomplete, inaccurate, untimely, unverified, or biased data.”158 One problem is that reporting is voluntary. Although manufacturers and facilities (hospitals) are required to report serious adverse events, doctors and other healthcare providers are not required to do so. As Diana Zuckerman, president of the National Center for Health Research, in Washington, DC, points out, even when facilities learn of a device problem, they often fail to report it to the FDA—nor does the FDA enforce the reporting requirement.159 The upshot is that underreporting is commonplace, and only a tiny fraction of adverse events and deaths related to medical devices is ever reported. Zuckerman adds that there are “huge disincentives” for doctors when it comes to reporting adverse events: reporting takes time, and doctors are time-pressured already, with just minutes to see each patient.159
Even more important, Zuckerman said, is that doctors worry about being sued, and “they know that they will be blamed for problems [with devices] as the manufacturers [virtually] always say a bad outcome was due to ‘operator error’ (meaning that a doctor or patient used the device in an unapproved manner),” thereby exculpating the manufacturer. This is dangerous territory not only for patients but also for doctors, who are often lured by manufacturers that encourage doctors to use devices in an off-label manner, because it increases market share. But when something goes wrong, the manufacturer will say it bears no responsibility because the doctor used the product in an unapproved manner. As a result, patients drop product-liability suits against manufacturers and instead charge the healthcare provider with malpractice.
The only incentive to report, said Zuckerman, “is caring about patient outcomes, but that incentive generally isn’t enough to overcome the disincentives.” Companies like Cyberonics, which are financially dependent on a single device, have every reason in the world to want to hide or obfuscate evidence that might suggest that the device is dangerous.
The FDA requires that if a death or serious injury “reasonably suggests” that a drug or device “may have caused or contributed” to death or injury, the manufacturer must report the event to the FDA. But if it’s determined that serious harm was not the fault of a drug or device, the company doesn’t have to report it.
So who decides whether a device “caused or contributed” to a serious complication or death? The FDA leaves that decision up to the manufacturer.100
Mark E. Bruley, vice president of accident and forensic investigation at ECRI, an independent nonprofit patient-safety organization based in Plymouth Meeting, Pennsylvania, says there is a reason to allow companies not to report all deaths: “Virtually every patient who dies in a hospital dies on a stretcher, but we don’t want stretcher manufacturers to report every hospital death.” Such reports, said Bruley, would have no patient safety or regulatory value.160
Although Bruley’s point is well taken, leaving it up to a manufacturer to determine whether its product was the cause of a patient’s death throws open the doors to powerful biases—biases made all the more evident by imagining how different the conclusion might be if the patient’s doctor or family were to make the determination. Manufacturers have powerful incentives to reach conclusions that don’t implicate their drugs or devices, and there is almost always a plausible alternative explanation for individual events.
Allowing manufacturers to be the “deciders” amounts to a virtual get-out-of-jail-free card for device makers—an escape hatch from responsibility that must be tremendously tempting for even well-intentioned company employees. Of course, the FDA can overrule an obviously bad decision by a manufacturer, but out of the hundreds of thousands of reports filed in the MAUDE database, only a relative handful are challenged by the agency. Yet it is a disinterested third party and independently conducted studies that are most needed in these instances.
In Dennis Fegan’s case, Steve Parnis, the Cyberonics staffer present at Fegan’s August 2006 visit with Bahamon, was tasked with ferreting out the “root cause” of Fegan’s brush with death.161 Parnis was no ordinary Cyberonics employee with merely the usual incentive to protect his employer and his job. Unbeknownst to Fegan, between 2003 and 2005, the company had filed eight patents on the VNS device in Parnis’s name, including one co-filed with Jacob Zabara, the original patent holder, in 2003.162–169 This may have given Parnis, a senior manager with Cyberonics, an incentive to protect the reputation of the device and that of the company that manufactured and marketed it. Yet it was Parnis, a radiology technologist whose training consisted of a one-year program to become an invasive cardiovascular technologist,* who examined Fegan’s EKGs and then wrote the official corporate report on the cause of Fegan’s asystole.161
The fact that manufacturers can decide whether their own products do or do not contribute to an adverse event severely limits the value of the FDA’s reporting requirement.
There are still other problems with the way the FDA tracks data about the safety of medical devices. For example, the agency does not collect data on the total number of devices (or drugs) on the market. This lack of a denominator makes statistics about the number of injuries all but useless. If there are fifty deaths among thirteen million users, that’
s very different from fifty deaths among two thousand users.
The FDA’s failure to take full advantage of the potential power of the MAUDE database has become increasingly glaring in today’s era of “big data.” A fully functional, truly comprehensive MAUDE database could be a powerful tool for safeguarding patient health and ensuring the efficacy of drugs and devices. But it would take wide-ranging changes to the system.
Madris Tomes knows more than most experts about the shortcomings of MAUDE. After graduating from American University with an MBA, she worked as a consultant with Booz Allen Hamilton, a management consulting company, and became adept at computer analytics—the craft of examining oceans of data to find patterns that mean something. Tomes went on to spend seventeen years in health information technology, mostly as a consultant to the Centers for Medicare and Medicaid Services and the FDA but also as an employee of the FDA itself.170
While working with the FDA’s MAUDE database, she realized that researchers who relied on the agency’s database would be misled if they didn’t know how to mine the data as she did, since a portion of deaths was misfiled. Although the FDA was in the process of purchasing new software, when Tomes conducted an analysis using the new product, she says she found that it “will not perform the big data analytics that are vital to ensure patient safety.”
Tomes decided to leave the agency to form her own company, Device Events, where she developed a program (at a price that she says was a mere fraction of the rumored $60 million the FDA paid for its program). The agency declined to review Tomes’s software tool.
After she left her work with the FDA, she presented a demonstration of her database tool to Art Sedrakyan, a professor of healthcare policy and research and cardiothoracic surgery at Weill Cornell Medical College. He was having trouble tracking harmful consequences associated with Essure, an implanted medical device intended to provide permanent sterilization for women. The device, manufactured by Bayer, consists of two thin coiled wires about an inch long, which are inserted through the cervix and uterus into the fallopian tubes, where they cause a dense inflammatory reaction and scarring. The scarring in turn blocks eggs from descending into the uterus, where they could be fertilized. But the device can migrate and cause severe pain, bleeding, infections, and even perforation of internal organs, including the uterus and fallopian tubes.171–174 Sedrakyan and his colleagues wanted to know how often these problems arose.
Tomes says Sedrakyan found “only a few hundred” adverse events related to Essure in the FDA’s database. But, she says, “when I plugged [the word Essure] into my database, I got back over a thousand adverse events.” As of July 2016, she said in a telephone call to me, the number stands at “well over twelve thousand adverse events.”
Using Tomes’s database, electronic health records, and claims data, Sedrakyan and his colleagues found that within a single year, one of every twenty-five women implanted with the Essure device underwent surgery to have it removed, most often because of bleeding and pain, and they were ten times as likely to have to undergo reoperation as women who underwent a different sterilization procedure.175
The FDA awarded expedited approval to the device in July of 2002 based on the manufacturer’s claim of a 99.8 percent success rate.171 However, during the approval hearings, it was determined that the claim was based on the first year of use only: at two years of use, only 149 of 507 women in the study (a rate of 29 percent) were even evaluated, and there was no report on how effective it was among those women. At three years of use, only five of 507 women (a rate of 1 percent) had been evaluated.176 The success claimed by Bayer was undermined by low follow-up rates and frequent complaints of miscarriages and unintended pregnancies.176, 177
The twenty thousand or so women complaining of pain, perforations, miscarriages, and the need to undergo surgeries and abortions caught the attention of activist Erin Brockovich. She took up their cause and joined them in demanding that that the FDA withdraw approval of the device. She deplored the FDA’s inaction, saying that if “20,000 penises were falling off, the world would stop.”178
When Tomes dug further into her database for fetal deaths related to Essure, only five deaths popped up in the MAUDE database using the “deaths” filter. But when she used her program, 303 deaths appeared—sixty times as many. She took her concerns to Congressperson Mike Fitzpatrick of Pennsylvania, who queried the FDA about her finding, and as a Fitzpatrick staffer put it, the agency confirmed that her finding was “spot on.” In a letter to Fitzpatrick, the FDA acknowledged that it had 296 deaths on file, saying that the difference between 296 and 303 was simply “methodological.”179 Whether the correct number was 296 or 303, it was a far cry from the five deaths reviewers would see using the FDA’s “deaths” filter.
Fitzpatrick sponsored a bill calling for the FDA to withdraw its approval for Essure, helping trigger an FDA hearing on the device. Although the agency ultimately declined to order the device removed from the market, it did order that a “black box” warning must be placed on the device label and that a checklist of cautions must be provided to patients before they undergo the procedure. However, severe underreporting of device complications continues to plague the FDA and its MAUDE database.
When problems with the FDA’s database surfaced during public protests by women over safety issues related to Essure, a WNBC television reporter asked Tomes which device was the worst in terms of discrepancies in the MAUDE database. It didn’t take long for her to come up with an answer: the vagus nerve stimulator. On August 3, 2016, Tomes conducted a side-by-side comparison between deaths associated with the VNS device found in the MAUDE database and deaths found in the Device Events database. During 2015, there were only 206 deaths in the MAUDE database, while the Device Events database yielded 2,093 deaths. Since 1997, when the VNS device first came on the market, through June 30, 2016, the disparity was even worse: there were 2,006 death reports in the MAUDE database, while the Device Events database yielded 11,223 death reports. Making matters worse, the FDA has deleted from their online database all adverse events dating back ten years or more, while Tomes has twenty years of FDA data.
Of course no one knows how many of the deaths reported in the MAUDE database are actually deaths caused by the device. Many are likely to be unrelated. Nor do the additional deaths found in the Device Events database mean that each death was caused by the device. In many other cases, however, the device may well be the cause of death or serious harm.
The FDA acknowledges significant limitations of the MAUDE database. It acknowledges that it can’t determine causality on the basis of individual reports, nor can it determine the true rate of harmful events caused by a particular device, because it doesn’t track the number of devices in use. What the agency does say is that the database helps it to detect red flags—a sudden uptick in the overall number of adverse events or the occurrence of new or unexpected injuries could cause the agency to ask the manufacturer to conduct a safety study.
But as reports pile up in the MAUDE database, it’s unclear if and when the agency is actually watching for red flags. A number of devices, including the VNS, have large numbers of deaths that failed to trigger action by the FDA.
The FDA responded to Tomes’s findings by stating, “The [MAUDE] database and analysis tools allow the FDA to distinguish mentions of deaths…and those where the device caused or contributed to a death.” Tomes said if that’s true, the FDA is not making that clear by moving the misclassified deaths into the “deaths” category, as the case with Essure demonstrated.
Yet these discrepancies refer only to data already in the FDA’s database. They pale in comparison to a far more significant problem acknowledged by the FDA itself as well as by the Government Accountability Office, which revealed in 1986 that hospitals report fewer than 1 percent of adverse device events to the FDA. According to the report, “the more serious the problem with a device, the less likely it was to be reported.”102 Put another way, 99 percent of device-related adverse
events never make it into the FDA’s database.
Sometime after 2010, the FDA removed that statistic—cited by former FDA commissioner David A. Kessler—from its website, but when questioned, the agency did not offer any newer statistic on underreporting.
Given the multiple uncertainties about the MAUDE database, the FDA’s surveillance of negative outcomes appears to be as effective as using a Ouija board. And as we’ll see, the problems with MAUDE are all too typical of the many shortcomings of the FDA. The sad truth is that guesswork, not science, is the main source for the agency’s assurances of safety.
Chapter Six
No Safety Net
BACK IN 2006, SHORTLY after Fegan’s encounter with Steven Parnis of Cyberonics, his neurologist, Juan Bahamon, formally dropped Fegan from his patient roster. Fegan was unable to find another neurologist until 2008, two full years after his hospitalization. Eventually he found Dr. JoAnne Sullivan.* His family accompanied him to some of his visits, and both Fegan and his family took an enormous liking to her. Sullivan was compassionate. She explained things. She seemed to genuinely care about Fegan and showed concern when he told her about his terrible experience with the VNS device.
Fegan was still eager to obtain justice for his brush with death. But by 2008, he had begun to despair of getting help. He had contacted the FDA, state regulators, politicians, and the media and gotten nowhere. No longer embroiled in day-to-day skirmishes with various regulators and agencies, he settled into daily searches through the medical literature and the MAUDE database, scrutinizing new reports as they came in, hoping to find something, anything, that would turn things around.
His days took on a steady rhythm. He rarely went out, saying, “Something could happen and I wouldn’t know it”—an oblique reference to having a seizure in public. His life as an outgoing, hail-fellow-well-met kind of guy who liked the beach, jogging, and square dancing was over. His world had contracted. On occasion, if a friend with a car came to visit, he might get a ride to the grocery store. But he would breathe a sigh of relief when he got home again and was alone, for it was only then that he felt comfortable.
The Danger Within Us Page 11