23 Food and Drug Administration
Why do we need to know if Dermagraft works?
The legal requirement that a treatment be proven to work originated in a specific tragedy. Today most new drugs and high-risk devices must pass a clinical trial: a prospective study in which participants receive different treatments and their outcomes are measured.
The thalidomide tragedy
The thalidomide tragedy forced the modern standard of evidence into law. Thalidomide was introduced in the late 1950s and sold widely, across Europe and beyond, as a sedative and a remedy for the nausea of early pregnancy (Vargesson 2015). By the standards of its day it looked safe: hard to overdose on, with mild effects on the adults who took it. No study had established whether it was safe during pregnancy, or whether it worked any better than the alternatives, because no law required one.
Thalidomide is a teratogen, a substance that disrupts the development of an embryo. Taken in early pregnancy, it interfered with the formation of the limbs and other organs, producing phocomelia (severely shortened or absent limbs) among other malformations. More than ten thousand children were born affected before the drug was withdrawn in 1961, in what has been called the largest man-made medical disaster on record (Vargesson 2015).
The United States was largely spared, because of one woman: FDA medical reviewer Frances Oldham Kelsey, who reviewed the application to market thalidomide there. She judged the safety data inadequate, declined to approve the drug, and held to that judgment against sustained pressure from the manufacturer (U.S. Food and Drug Administration 2026b). Thalidomide was never marketed in the United States. When the link to birth defects was established abroad, the scale of the disaster the U.S. had avoided became plain (U.S. Food and Drug Administration 2026b).
Thalidomide today is an effective, approved treatment for complications of leprosy and for multiple myeloma, a blood cancer (Vargesson 2015). The failure lay in how it was used: it was given to the wrong patients, for an unproven benefit, with no one obliged to weigh that benefit against its risks.
The Kefauver–Harris Amendments
Before the tragedy, the Federal Food, Drug, and Cosmetic Act of 1938 required a manufacturer to show only that a new drug was safe. The statute said nothing about whether the drug had to work (Van Norman 2016a).
Thalidomide broke a political deadlock over a reform bill stalled in Congress. In October 1962 Congress passed the Kefauver–Harris Drug Amendments to the 1938 Act (Greene and Podolsky 2012; U.S. Food and Drug Administration 2026b). The amendments added a requirement of efficacy to the existing requirement of safety. For the first time, a manufacturer had to prove a drug was effective for its intended use before selling it (Greene and Podolsky 2012; Van Norman 2016a). That proof had to come from adequate and well-controlled studies: trials run against a comparison group, with patients assigned by a defined, documented method rather than by convenience. Such assignment means the result can be attributed to the treatment and not to who happened to enroll. The amendments also placed new controls on testing drugs in human subjects (Van Norman 2016a).
What the FDA regulates: drugs, devices, biologics, and combinations
The FDA sorts the products it regulates into categories, each with its own marketing application and burden of evidence.
Drugs, in the FDA’s usage, achieve their effect chiefly through chemical action, usually as small molecules synthesized in a factory. The FDA’s Center for Drug Evaluation and Research reviews new drugs, and a maker seeks permission to market one through a New Drug Application (NDA). The statute requires “substantial evidence” of both safety and efficacy, which the FDA generally interprets as at least two adequate and well-controlled clinical trials showing the drug to be effective (Van Norman 2016a).
Devices act by physical rather than chemical means: an instrument, an implant, a piece of diagnostic equipment. “Device” spans everything from a tongue depressor to an implanted pacemaker, so devices are sorted by risk into three classes, and the evidence demanded rises with the class (Van Norman 2016b). Class I devices pose low risk, such as surgical gauze, and are subject only to general controls; many need no premarket review at all. Class II devices pose moderate risk, such as a surgical suture, and most reach the market through a 510(k) premarket notification, in which the maker shows the device is substantially equivalent to one already legally marketed (a “predicate”), often without a new clinical trial (Van Norman 2016b). Class III devices sustain life, are of substantial importance in preventing harm, or otherwise present high risk: a pacemaker, or any genuinely new device with no predicate. These require Premarket Approval (PMA), the most stringent device pathway, in which the maker must supply clinical evidence the device is safe and effective (Van Norman 2016b; U.S. Food and Drug Administration 2026d). The Dermagraft trial was run to meet this standard.
Biologics are products derived from living sources rather than chemically synthesized: vaccines, blood products, therapeutic proteins, and cell and gene therapies. The FDA’s Center for Biologics Evaluation and Research reviews them under a Biologics License Application (BLA), which, like an NDA, must show the product to be safe and effective (U.S. Food and Drug Administration 2026a).
Combination products exist because real products cross these boundaries. A drug-eluting stent is a device and a drug. A prefilled autoinjector is a drug and a delivery device. A therapy of living cells grown on a scaffold blurs biologic and device. For such products, the FDA assigns a lead reviewing center by the product’s primary mode of action: the single mode that provides its most important therapeutic effect (U.S. Food and Drug Administration 2026c). Dermagraft, a sheet of living human cells on a dissolvable scaffold, sits close to the boundary between categories rather than squarely inside one.
The drug-development pipeline
A new drug advances through a sequence of phases, each larger than the last and gated by the results of the one before, so the smallest, cheapest studies come first.
Preclinical work comes before any human contact. The candidate is characterized in the laboratory (in vitro) and tested in animals (in vivo) for basic activity and toxicity. Attrition here is severe: on the order of one candidate compound in a thousand survives to be tested in people (Van Norman 2016a). Before the first human dose, the sponsor must file an Investigational New Drug (IND) application, the FDA’s authorization to begin testing in humans (Van Norman 2016a).
Phase I is the first administration to people. It is small, on the order of twenty to a hundred participants, often healthy volunteers, and its question is safety, not efficacy: is the drug tolerated, and what dose can be given without unacceptable harm? (Van Norman 2016a; U.S. Food and Drug Administration 2026e)
Phase II enlarges the study, typically to patients who have the condition the drug is meant to treat. The goal shifts to efficacy: does the drug show a signal of benefit while remaining acceptably safe? These trials are frequently run against a placebo (Van Norman 2016a).
Phase III is typically a large trial, commonly several hundred to a thousand patients (Van Norman 2016a; U.S. Food and Drug Administration 2026e). Its purpose is to establish efficacy against a control or comparator, in enough patients that chance alone is an unlikely explanation for the result.
Once the phases are complete, the accumulated evidence is submitted for review and approval: an NDA for a drug, a BLA for a biologic, a PMA for a high-risk device. The FDA weighs whether it amounts to substantial evidence of safety and efficacy (Van Norman 2016a). Phase IV, or post-market surveillance, follows the treatment into the far larger population of ordinary use, where rare harms and long-term effects that no trial of a few thousand patients could reveal may become visible (Van Norman 2016b).
Roughly nine of every ten drugs that enter human testing fail somewhere along the way. Carrying a single drug from concept to approval typically takes ten to fifteen years and costs on the order of a billion dollars (Van Norman 2016a). Devices, whose pathways are generally shorter, average nearer three to seven years (Van Norman 2016b).