Post-marketing withdrawal of 462 medicinal products because of adverse drug reactionsa systematic review of the world literature

A drug is banned in Germany on a Tuesday. That same week, it sits on pharmacy shelves in three other countries, dispensed without restriction and prescribed without warning. The same compound. The same harm. Different borders, different answers. That asymmetry is not a hypothetical — it is what Onakpoya, Heneghan, and Aronson documented when they built the first systematic map of sixty years of drug withdrawals. What they found should change how we think about the systems that are supposed to protect us. Post-marketing withdrawal means a regulator pulling a drug from the market after it has already been approved because adverse reactions have emerged that weren't visible in clinical trials. The key phrase is "after approval." Trials are powerful but finite. They enroll thousands of patients over months or years, optimized to detect efficacy. The real world enrolls millions over decades, and it catches things trials miss. When those signals appear, the question becomes: what do regulators do with them, how quickly, and how consistently? Before this study, nobody had answered that question systematically. Onakpoya and colleagues searched PubMed, Google Scholar, and the World Health Organization's drug databases, including the websites of national regulatory agencies like the UK Medicines and Healthcare products Regulatory Agency, the US Food and Drug Administration, the European Medicines Agency, Health Canada, and many more. They covered the years from 1950 through December 2014, included only drugs withdrawn because of suspected adverse reactions, and excluded products pulled for commercial reasons, contamination, or efficacy concerns. They also rated the quality of evidence behind each withdrawal using the Oxford Centre for Evidence-Based Medicine hierarchy, which is a standard ladder that runs from systematic reviews at the top down through randomized trials, cohort studies, case series, to single case reports at the bottom. What they counted: four hundred sixty-two medicinal products withdrawn between 1953 and 2013. The single most common reason was hepatotoxicity, which is drug-induced liver injury, accounting for eighty-one of those four hundred sixty-two withdrawals, or eighteen percent. Immune-mediated reactions came next at seventeen percent. Together, those two categories explain more than a third of all withdrawals. Neurotoxicity, cardiotoxicity, and carcinogenicity each accounted for between thirteen and sixteen percent of cases. Deaths were associated with withdrawal decisions in one hundred fourteen instances, which is one in four of all the products in the dataset. Now here is the finding that stops you cold. When Onakpoya and colleagues looked at what evidence regulators actually had in hand when they pulled those drugs, three hundred thirty of the four hundred sixty-two withdrawals, or seventy-one percent, were based primarily on case reports. These included individual patients and anecdotes in the clinical literature. Not randomized trials, not systematic reviews. Only twenty-seven withdrawals, or fewer than six percent, were supported by individual randomized studies. Just six, which is a little over one percent, were backed by a systematic review. The paper is blunt about this: formal studies are often not conducted when adverse reactions are reported anecdotally. Regulators were removing drugs from the market — sometimes drugs that killed people — based on the thinnest thread of formal evidence. The dominance of case reports did decline over time, from eighty-five percent of withdrawal decisions in the nineteen fifties to thirty-five percent for withdrawals since the year 2000. Progress, yes. But even that improvement needs context. Because here is what happened with the time. The median interval between the first published report of an adverse reaction and the year of first withdrawal was six years, with an interquartile range of one to fifteen years. Six years of warnings in the literature before regulatory action. And when you extend the window further, the median time from a drug's launch to its first withdrawal was eighteen years overall. Think about what accumulates in eighteen years of prescriptions. The most damaging detail is this: that six-year gap between the first report and the first withdrawal did not consistently shorten over time. Even as reporting improved, even as pharmacovigilance systems matured, the lag between recognizing harm and acting on it held stubbornly in place. Faster reporting of adverse reactions did not translate into faster regulatory action. The pipeline got more information sooner, but the valve at the end didn't open any quicker. Part of the explanation lies in the geography. And this is where the paper becomes genuinely disturbing. Of the four hundred sixty-two withdrawn drugs, only forty-three, or nine point three percent, were withdrawn worldwide. A drug recognized as dangerous enough to pull in one jurisdiction was pulled everywhere in fewer than one in ten cases. Meanwhile, one hundred seventy-nine products, or thirty-nine percent, were withdrawn in a single country only. The rest were pulled in some countries but not others. The global drug supply is integrated, but the global safety response is not. The continental pattern is stark. Europe recorded three hundred nine withdrawals; North America one hundred thirty-four; Asia one hundred fifty; South America sixty-five; Africa sixty-three. When you standardize those numbers by country, Africa's rate was one point seventeen withdrawals per country. Europe's was six point eighteen. North America's was five point eighty-three. Every other continent's relative rate compared to Africa was statistically significant, with p-values well below zero point zero five. Per million population, Africa had zero point zero six withdrawals per million, against zero point four two in Europe and zero point two five in North America. Onakpoya and colleagues link this directly to pharmacovigilance infrastructure. Drug regulatory systems in most African countries, they write, are not well developed. Data from the World Health Organization showed that only four percent of African nations had moderately developed pharmacovigilance systems, while thirty-nine percent lacked adequate regulatory capacity altogether. As of 2009, more than half of African countries did not have a drug regulatory website. Reporting is suppressed by lack of knowledge of reporting systems, uncertainty about causality, time pressures, and structural conflicts of interest. Licensing in many African countries relies on assessments done elsewhere. The result is that drugs withdrawn in Europe because of documented harm can remain in active use in African clinics — not because of a deliberate choice, but because the infrastructure to recognize and act on the signal simply isn't there. What needs to change? Onakpoya and colleagues identify three levers. The first is international coordination. A withdrawal decision in one country should trigger formal review in others, not sit quietly in a regulatory database while the drug continues to be prescribed across borders. The fragmentation documented here is not incidental — it is structural, and coordination is the structural fix. The second is transparency. The paper calls explicitly for access to clinical study reports as a priority for future drug regulation, alongside more transparency in reporting adverse events observed during clinical trials. The reliance on case reports as the primary basis for withdrawal is in part a symptom of a system where full trial data are not available to regulators or independent researchers who might detect signals earlier. The third lever is building pharmacovigilance capacity where it is weakest. The authors note the World Health Organization's proposal, in collaboration with African Union countries, to establish an African Medicines Agency — and position that kind of structural investment as a direct route to preventing harmful drugs from remaining available longer in regions with fewer resources to catch the signals. They also call for better education and incentives to increase reporting by clinicians and patients, because the chain from harm to regulatory action begins with someone noticing and saying something. This study is, at its core, a sixty-year audit of how the world has handled the discovery that approved medicines cause harm. The answer is: inconsistently, slowly, and with profound geographic inequity. The evidence base for pulling drugs is often thin. The lag from first warning to first action is measured in years, sometimes decades. And whether a patient is protected often depends less on what a drug does to the human body than on which country that patient happens to live in. That is the gap this research maps. Closing it requires coordination, transparency, and investment — not in the science of drug discovery, but in the quieter, less glamorous science of watching what happens after the drug is already out there in the world. This lecture was created by ennepō. Go to https://ennepo.ai to Discover, Create and Follow the latest research in your field. Read when you can. Listen when you want to.