For decades, the U.S. Food and Drug Administration and its European counterparts had allowed defective medical devices to spill onto global markets, like contaminated water from a broken pipe.
Our Dutch partners, AVROTROS, for instance, had doggedly reported for years about vaginal mesh that had terribly injured thousands.
On Sunday, November 25, 2018, the International Consortium of Investigative Journalists, the Associated Press, NBC News Investigative Unit, AVROTROS, and more than 50 media partners around the world published a yearlong investigation that shows regulators bowing to industry pressure to rush approvals, lower safety standards and cloak critical information, and the consequences: a string of grisly accidents that have left hundreds of thousands disfigured, disabled or dead.
The project required an unprecedented global effort to gather millions of medical-device records from public data and more than 1,500 freedom of information requests. We deployed machine learning and other data analysis to place stories of individual suffering in a global, systemic context.
We found artificial hips that corrode flesh and poison blood, spinal-cord stimulators that shock and permanently injure and surgical mesh that cuts into tissue, causing chronic pain and bleeding — all cleared through a U.S. approval pathway known as 510(k) that okays new devices based, not on human testing, but on their similarity to older products.
A defective metal IUD had remained on sale in the U.S. long after it was pulled in elsewhere, we found. The FDA allowed an exemption to export products — including a carbon shoulder joint that broke inside patients — under lowered standards.
We found that the FDA had allowed device markers to bury hundreds of thousands of breast implant injury reports.
And we found that industry giant Medtronic had repeatedly violated a 2008 pledge to U.S. authorities that it would obey the law, and did so on a global scale.
We found that efforts to overhaul safety regulations of Europe’s industry-friendly system of device regulation had been stymied by an army of lobbyists, not once, but twice, since 2008.
Reporting teams found that In India, as well as the Philippines, Singapore and other countries, devices are waved through, or subjected to less scrutiny, if they have already been certified in Europe and the U.S.
Our analysis of data from India’s Central Drugs Standard Control Organisation and the FDA also revealed that the Indian government failed to publish recalls for dozens of medical devices that were sold in the country.
Our Korean partners, Newstapa, reported that often patients there were left in the dark about the potential risks associated with implants before undergoing surgery and often not notified when a product was recalled.
And our reporting found that around the world, legal redress for device industry harm has been as uneven as regulation, and overseas plaintiffs face daunting hurdles when suing U.S.-based companies.
The project sparked immediate reforms (see below).
What makes this project innovative?
ICIJ’s collaborative model itself is a major innovation. The Implant Files showed the concept could work beyond offshore banking and without benefit of a leak. We assembled the largest healthcare investigative team ever: more than 250 journalists in 36 countries, allowing us to interview hundreds of patients worldwide and sound out thousands more via online surveys. A pre-publication ICIJ survey for breast implant patients, distributed via social media groups, helped to propel our blockbuster exposé. Our model enabled the submission of 1,500 public records requests and the assembly of more than 8 million records from 36 countries. Even public device data required high levels of technical skill to link deaths to individual devices and keen attention to unique identifiers to avoid double counting. The 5.6 million FDA “adverse event” reports collected over a decade contained endless variations of device names (more than 130 for a single hip device) and thousands of corporate subsidiaries (not to mention misspellings). We gathered financial data on more than 40,000 device maker subsidiaries to connect incidents to parent companies. ICIJ used both open source software and a bespoke tool to help 250 reporters from dozens of news outlets navigate records without need of coding skills. ICIJ was ultimately able to identify more than 83,000 deaths and 1.7 million injuries linked to individual devices over the past decade. Using our own machine-learning algorithms, our data team found that 2,100 of the deaths had been misclassified as something less serious. We also found 500,000 reports that described an “explant” — the dangerous removal of a broken implant. We gathered more than 90,000 records from 18 countries (and counting), standardized device makes and models to build the groundbreaking International Medical Device Database (IMDD), allowing the public for the first time to learn whether their device has been cited for official safety concern anywhere.
What was the impact of your project? How did you measure it?
While ICIJ and its partners are proud of the reforms we have helped to generate, particularly in the field of offshore finance, change usually takes years. This time, the reaction began even before ICIJ and its media partners published our first stories. On November 14, 2018, after we asked dozens of questions, the FDA issued an alert about the risks of defective pain pumps that ICIJ had linked to 14,000 explant surgeries. On November 20, the FDA announced action to strengthen oversight of surgical mesh. The same day, it declared an “ambitious new goal” to be “first among the world’s regulatory agencies” to identify harmful devices. ICIJ’s reporting had found it was last in the world to force Essure, a contraceptive that can shred reproductive organs, from the market. The day after we published, FDA Commissioner Scott Gottlieb said that the agency would reform its approval process, following our reporting that showed the FDA had allowed manufacturers to sweep thousands of injury reports under the rug. Then in the spring, the global policy reforms really kicked in. In a major advance for medical device transparency, the FDA in March 2019 pledged to disclose information kept buried for nearly two decades under a program known as “alternative summary reporting” — another issue highlighted in our series. Millions of previously undisclosed cases of patient harm and product malfunction could now soon be released. The decision follows ICIJ reporting that revealed hundreds of thousands of incidents related to breast implants had been kept out of public sight under the summary reporting program. Also in March 2019, the FDA issued a statement acknowledging that a broad array of implanted devices may make some people sick, saying “a growing body of evidence suggests that a small number of patients may have biological responses to certain types of materials in implantable or insertable devices.” The agency said it was gathering information to fill information gaps in the science. The agency held a two-day meeting about breast implants, an issue highlighted in our series In April 2019, French health authorities imposed a sweeping and historic ban on two categories of breast implants associated with an elevated risk of a rare form of cancer. Advocates had pushed for such a measure for years. An activist told us: “I am ecstatic. I think the evidence is just too overwhelming now for them to ignore.” Canadian authorities followed with a plan to suspend the sale of so-called textured breast implants, while plastic surgeons in The Netherlands did so voluntarily after a request from the government there. The UK, Netherlands, Germany and Italy announced they are separately considering mandatory registers of implants, a huge step in tracking post-approval problems. Australian authorities announced a plan to track medical implants like plates, stents and pelvic mesh, described as the biggest shake-up in patient safety in 30 years. A manufacturing giant abruptly ended all production and marketing of vaginal mesh implants, a longtime target of patient groups and another focus of our reporting. Health authorities in Denmark, Spain, Canada, Italy, India and elsewhere have announced they are overhauling aspects of medical device oversight. Such impact rebounds to the U.S., which has loosened rules to compete with more lenient jurisdictions, especially the European Union. The Implant Files’ impact transcended action by official bodies: It sparked a global response from patients and the public. Less than 24 hours after publication of the first stories, LaSexta, our partners in Spain, received more than 5,000 messages and 600 calls. Six weeks after publication, ICIJ’s International Medical Devices Database (IMDD) already had nearly 600,000 visits and won praise from academics, patients and industry insiders. Within days of publication, thousands of patients, healthcare workers and medical device professionals responded to a global ICIJ survey translated into nine languages and shared on mainstream media publications in more than 20 countries. The callout was so popular it broke the website hosting the survey (Screendoor). To date, more than 3,400 people responded, many sharing their experiences in intimat, often heartbreaking, detail. “I’ve been left abandoned in a prison of debilitating pain,” wrote Amanda Tallent, of Spartanburg, South Carolina, in mid-December 2018. She has suffered for nine years from implanted mesh. “Speaking my truth is as close to justice as I will have,” she said. ICIJ partners in several countries began using information gleaned from huge reader engagement to produce follow-up stories. Partners in Tunisia and France worked together to reveal a stream of patients traveling from one country to another in search of cheaper — not necessarily safer — implants. Each week, more people come forward, and the survey continues to propel ICIJ reporting. Some of the most compelling patient stories were captured in video pieces. One of our best-performing videos on Facebook, “What inspired our investigation?”, generated more than 10,000 unique viewers (5,265 for more than 10 seconds), most of them women aged 45 to 54 — a new audience beyond our predominantly 25-to-34-year-old male Facebook following. The video received 996 interactions, including 578 total reactions, 103 total comments and 318 shares. We further engaged this new audience on our growing Instagram account: our largest group of followers, the 25-to-34 age group, is 57% women. The post with our all-time best reach was on defective breast implants and reached 46% non-followers. Much of our breast implant-related content has been reposted by patient advocacy accounts, including @breast_implant_illness, allowing findings to be shared among audience members with their own digital communities.
Source and methodology
Our reporting found that recall and safety alert data were scattered around the world, in different languages, in different formats, without standardized make and model names, leaving patients in the dark about whether their device had been recalled in another jurisdiction. To help fix the problem, we built the International Medical Devices Database (IMDD), which allows for the first time the public to search records — more than 90,000 from 18 countries and counting — to find whether a device has been flagged for official safety concern anywhere in the world. Within six weeks of publication, the database had drawn nearly 600,000 visits and drawn praise from academics, patients and industry professionals. We consider the IMDD, available through https://medicaldevices.icij.org/, to be a reform unto itself. Our FOI campaign, more than 1,500 public records requests from 36 countries, included requests for adverse-event reports, recall notices, safety warnings, lobbying records and documents on communications between authorities. It drew stiff resistance at nearly every turn. Fewer than half of our records requests were even answered. Many of those that were included substantial redactions. In Mexico, reporters submitted 964 requests to federal agencies and received only 44 responses. Authorities found creative ways to deny or delay response. U.S. officials cited a backlog of requests. In Belgium, it was a lack of staff; in Europe, the U.K., Japan and elsewhere: “trade secrets.” In France, Spain and the Netherlands, journalists are still fighting to obtain injury and death information. In Canada, ICIJ partners fought for two years before winning the first-ever release of adverse event reports. In many countries journalists followed up with new requests. ICIJ and its partners were able to identify devices problems even for countries that didn’t supply any records on their own by using information about device distribution and other records found in U.S. recall data. We also reviewed foreign adverse-event reports filed with the FDA available through the agency’s Manufacturer and User Facility Device Experience (MAUDE) database. The effort ultimately helped gather more than 8 million health records that anchored the Implant Files investigation. ICIJ didn’t pay for data, but for a fee paid to the Investigative Reporters and Editors’ data library for its historical record of the FDA’s MAUDE database. The IRE archive contained data from earlier years not publicly available in FDA records. MAUDE contains more than 5.4 million reports sent to the FDA over the last decade on suspected device-associated deaths, serious injuries and malfunctions. Most are from the U.S., but the FDA also receives reports from overseas. ICIJ was able to use MAUDE data to connect death and injury reports to specific medical devices and manufacturers. We used machine-learning techniques to explore device removal procedures known as explants, as well as to identify thousands of cases in which patient deaths were misclassified as “injuries” or “malfunctions” (more at this link: https://www.icij.org/investigations/implant-files/algorithms-analysis-and-adverse-events-how-icij-used-machine-learning-to-help-find-medical-device-issues/). The method involved identifying sentences and nouns, verbs and adverbs used as euphemisms for “death” from thousands of real death reports in the data and refining it based on a verification process vetted by journalists. Once the computer learned to identify deaths based on variations of language, ICIJ ran the analysis through the millions of records from the MAUDE database and checked how the deaths were originally classified. A second algorithm was then used to review whether it was possible to know if a device contributed to the death. A team of journalists then checked the results, one at a time, by reading the event description and comparing it to the classifications obtained through the machine learning process. We gathered similar data on suspected device-related “adverse events” in Australia, Japan, Finland, Canada and Norway. We collected recall, safety alert and field safety notice data from more than 40 countries. In some instances the data was public; in others it was obtained through freedom of information requests. Since devices were listed under myriad make and model names, we used lot numbers, serial numbers and other codes to identify like products. We found dozens of instances in which a device had been flagged for concern or even pulled from the shelves in one country but remained in wide circulation, without warnings, in another. Other datasets obtained and used for analysis included: medical device approval data from the U.S. (source: FDA); similar data from local health authorities or so called notified bodies, the mostly private device-approval companies used in Europe; lobbying data (source: Opensecrets), and payments to physicians (Centers for Medicare and Medicaid Services). Linking injuries and other adverse events to specific manufacturers posed significant challenges. In addition to differences in spelling (or simple misspellings), devices were often associated with subsidiaries — and the biggest manufacturers have hundreds. To make the proper connections, ICIJ collected information on more than 40,000 subsidiaries around the globe. ICIJ used FDA’s product categorization system to cluster similar devices by their general purpose. To avoid double counting events, we paired each report with its unique key number. Our goal of tracking recalls and other safety warnings across borders also posed challenges. To standardize and facilitate our analysis, ICIJ automated translations of each non-English database into English and then checked with translations with local partners. The datasets we gathered had different structures, and the absence of any universal numbering system made it difficult to track specific medical device warnings across the world. We reviewed customs records to check whether devices recalled in one country had entered other countries after a recall had been published. We found that recalls aren’t consistently published at the same time — sometimes months and even years elapse between the first country and last country to act. ICIJ used U.S. Securities and Exchange Commission filings and Dow Jones Factiva reports as primary sources to link parent companies with their manufacturing subsidiaries. ICIJ also used open source tools or in-house tools throughout the project. The team used business analysis tool Saiku to facilitate the analysis of the data, without coding skills. It also used an in-house developed tool (with Ruby on Rails) to facilitate the exploration of unstructured portions of the data as well as its connection with all the other fields. Other technologies that were used in previous projects like the Global I-Hub to facilitate communication and sharing between project members, and the Knowledge Center to facilitate the exploration of documents that were obtained through FOI requests, were also used throughout the project.
Talend Real-time BigData Platform, Microsoft SQL Server 2017, R, Python, Ruby on Rails, Microsoft Excel, Google Sheets, Open Refine, Saiku.
Implant Files Reporting & Data Teams