On this page
Before any medicine reaches a pharmacy, someone has to decide whether the evidence for its safety is good enough. In 1960 that job fell to Frances Kelsey, a new reviewer at the US Food and Drug Administration (FDA). The drug was thalidomide. Her refusal to approve it kept a major tragedy out of the United States and changed how drugs are tested. This post covers what she did, using FDA and US National Library of Medicine records, then works through why rare side effects are so easy to miss.
Who was Frances Kelsey?
The National Library of Medicine's biography gives her dates as 1914 to 2015. She was born Frances Oldham on Vancouver Island, Canada, and earned bachelor's and master's degrees from McGill University in 1934 and 1935. She took a PhD in pharmacology at the University of Chicago in 1938 and an MD there in 1950. The FDA's history page on Kelsey adds that she did editorial work for the American Medical Association, reviewing papers on new treatments, and taught pharmacology at the University of South Dakota from 1954.
In 1960 she joined the FDA as one of just a handful of medical officers. Their job was to review new drug applications, in which manufacturers had to show evidence that a drug was safe before it could be sold.
Why did she hold back thalidomide?
One of her first applications was for thalidomide, a sleeping pill already sold in dozens of countries. The NLM says she was concerned by data suggesting dangerous side effects in patients who took it repeatedly. The FDA says she refused to approve it "because of its inadequate evidence", despite constant pressure from the company. Each time new material arrived, she asked for scientifically reliable evidence, which she felt was still missing.
What happened next?
About a year later, the FDA records, researchers in Germany and Australia linked thalidomide to clusters of rare, severe birth defects, in which hands and feet grew directly from the shoulders and hips. Thousands of babies were eventually affected. A 2011 review in Toxicological Sciences notes that the drug was widely used for nausea in pregnant women and that the tragedy prompted regulators to develop systematic toxicity testing.
Thalidomide was never marketed in the United States. On 7 August 1962 President John F. Kennedy gave Kelsey the President's Award for Distinguished Federal Civilian Service. The citation, quoted by the NLM, says: "Her exceptional judgment in evaluating a new drug for safety for human use has prevented a major tragedy of birth deformities in the United States."
What changed in the law?
The near-disaster helped pass the 1962 Drug Amendments. The NLM summarises what the new rules required: drugs had to be shown to be both safe and effective, patients in clinical trials had to give informed consent, and adverse reactions had to be reported to the FDA. Kelsey went on to lead the FDA's Division of Scientific Investigations, which checked the integrity of clinical trial data, from the late 1960s until the 1990s.
The 2011 review adds a twist: thalidomide later proved useful against leprosy and the blood cancer multiple myeloma, and it warns that where pregnant women with leprosy are treated without close medical surveillance, it has continued to cause malformations.
Why are rare side effects easy to miss? A worked example
The numbers below are hypothetical, for a classroom example only. Suppose a drug harms 1 in 500 people who take it, and each person is affected independently.
The chance that one person is not harmed is 499/500 = 0.998.
Trial of 100 people. Chance nobody is harmed = 0.998¹⁰⁰ ≈ 0.82. So there is about an 82% chance the trial sees no harm at all.
Trial of 2,000 people. Chance nobody is harmed = 0.998²⁰⁰⁰ ≈ 0.018, about 2%. A trial this size would almost certainly see the problem.
A small, clean trial can look reassuring simply because it was too small to catch a rare harm. And if a harm affects only one group, such as unborn babies, a trial must include that group, or the harm stays invisible however large the trial is.
Common mistakes:
Multiplying instead of raising to a power. The chance for 100 independent people is 0.998 × 0.998 × ... a hundred times, not 0.998 × 100.
Reading "no harm seen" as "no harm". Absence of evidence in a small sample is weak evidence.
Forgetting who was studied. Results only cover the kind of people tested.
Our post on independent events in probability explains the multiplication rule used here.
What can a student take from Kelsey's way of working?
Ask what the evidence actually shows. Popularity and pressure are not data.
Hold your ground politely but firmly. She kept asking for better evidence for months, until the danger became clear.
Know your subject deeply. Her training in pharmacology let her spot weak data.
Think about who is missing from a study. Rare harms hide in groups nobody tested.
If powers and probabilities in data questions slow you down, the Fermi Problem Library has a step-by-step solver that explains the reasoning, so you can check each line of a calculation like the one above.
Next step: redo the worked example for a harm that affects 1 in 1,000 people, and find how large a trial needs to be before the chance of seeing no harm falls below 5%.
