How we got here: why a medicine must prove itself before sale

The medicine contained diethylene glycol, a poisonous solvent related to antifreeze. The law of 1906 did not ask for any information before a drug was sold, so nobody had to show it was safe first.^1,4 The 1938 Act, approved on 25 June 1938, changed that. A company now had to send the FDA a new-drug application with reports of its safety tests and a description of how the drug was made. The FDA could refuse it if the tests were inadequate, if they did not show the drug was safe, or if the manufacturing controls could not keep the drug’s identity, strength, quality and purity.² The law asked for “adequate tests by all methods reasonably applicable”, but it did not say which tests.^2,7

The gap was effect. Before 1962, US manufacturers did not have to prove to the FDA that their products worked.¹ Senate hearings in the years before 1962 heard about “the sorry state of science supporting drug effectiveness” and exaggerated claims in labels and advertising, as a senior FDA official later put it.⁴

On paper, the Netherlands already asked about effect. The Wet op de Geneesmiddelenvoorziening of 28 July 1958 set up a board to register packaged medicines. A product could be registered if it could reasonably be assumed to have the claimed effect and not to harm health when used as directed; unregistered packaged medicines could not be sold. The act asked for a reasonable assumption, not for a particular kind of study. Its article 37 left the date it came into force to a later royal decree, possibly at different times for different parts.⁸ We could not confirm when the register started to work.

Thalidomide was made by the German company Chemie Grünenthal and sold as Contergan in Germany, Softenon elsewhere in Europe and Distaval in the UK. It was advertised as free of risk, and doctors handed out free samples for morning sickness.⁶

The damage came from timing. The drug harmed the embryo in a short window early in pregnancy, between about day 20 and day 36 after fertilisation. Limbs, face, eyes, ears, heart and other organs could be affected, and many survivors live with severe disabilities today.⁶

The tests done before sale did not reveal the danger, and exactly what was tested is still debated. Mice, the usual species for screening drugs, are less sensitive to thalidomide than rabbits or primates. After the disaster, drug testing changed to use several species, and regulators developed systematic tests for toxic effects.^6,10 The same lesson, that an animal result may not carry over to people, runs through our page on the translation gap.

In 1960 Frances Kelsey joined the FDA as one of a handful of medical officers. One of the first applications she was given was for thalidomide, already sold in dozens of countries. Despite constant pressure from the company, she refused to approve it because the evidence was inadequate. About a year later the link with birth defects was found, and the drug was never marketed in the US.⁹

The US did not escape entirely. The US sponsor had sent samples to thousands of doctors, who gave them to patients without telling them the drug was experimental. More than a dozen children with thalidomide damage are believed to have been born in the US as a result.⁷ In 1962 President Kennedy gave Kelsey the highest award for a federal civil servant.⁹

People often get better or worse on their own, believing in a treatment can change how they feel, and an observer who knows who was treated can see what they expect. US regulations say a clinical study exists to “distinguish the effect of a drug from other influences” like these.¹¹

So a well-controlled study compares the treated group with a control group, such as a placebo or an established medicine. It assigns people to the groups by randomisation, as a rule, and uses blinding so that neither patients nor observers can tilt the result. Uncontrolled studies “are not acceptable as the sole basis” for a claim that a medicine works.¹¹ Our page on how to read a study explains each of these protections.

The 1962 law required investigators to tell people that they were receiving an experimental drug and to obtain their consent, or that of a representative. It also allowed exceptions where the investigator judged consent “not feasible” or not in the person’s interest.³ Later rules added more: a US policy of 1981, adopted across government in 1991 as the Common Rule, set requirements for documenting consent and for ethics review boards.¹

The new rule also applied backwards. In 1966 the FDA asked the US National Academy of Sciences to evaluate the effectiveness of about 4,000 drugs approved on safety alone between 1938 and 1962.¹ In 1970, during this review, new rules stated that “uncontrolled studies are not acceptable evidence to support claims of effectiveness”.⁷

The directive covered ready-made (“proprietary”) medicines, sold under a special name in a pack. None could be placed on the market in a member state without an authorisation, and the member states had 18 months to put the rule into their own law. The application had to include the results of chemical and microbiological tests, pharmacological and toxicological tests, and clinical trials. The authorisation was refused if the medicine was harmful in normal use, if its therapeutic effect was lacking or “insufficiently substantiated by the applicant”, or if its composition was not as declared. A decision was due within 120 days, and an authorisation lasted five years before renewal.⁵

The directive named its own balance. Its first aim was “to safeguard public health”, but by “means which will not hinder the development of the pharmaceutical industry or trade in medicinal products”.⁵ In 2001 the EU gathered its medicines laws into one code, Directive 2001/83/EC, which repealed Directive 65/65 and kept the same grounds for refusal.¹³

The Wet op de Geneesmiddelenvoorziening of 1958 was replaced almost fifty years later. On 1 July 2007 the Geneesmiddelenwet, the Dutch Medicines Act of 8 February 2007, came into force and withdrew it.^8,12 The core rule today is article 40: no medicine may be placed on the Dutch market without an EU or Dutch marketing authorisation, and a medicine without one may not be sold, supplied or imported, with a few exceptions.¹⁷ Our page on Dutch law explains what this means for peptides.

In 1993 the European Community created what is now the European Medicines Agency (EMA), with an EU-wide authorisation procedure that started on 1 January 1995.^14,15 Today medicines made by biotechnology and medicines with a new active substance for diabetes, among other groups, must go through this central route.¹⁸ That is why semaglutide and tirzepatide were authorised by the European Commission for the whole EU at once.^19,20

US regulations say that a medicine tested in a trial must be “standardized as to identity, strength, quality, purity, and dosage form to give significance to the results of the investigation”.¹¹ The logic is simple: a trial measures what was in the trial. If a different product has a different content, amount or purity, the trial says little about it.

This is why a product from outside the legal supply chain that carries the name of an authorised medicine is not covered by that medicine’s evidence. Its content has not been checked against the authorised product. See product risk for what that means in practice and what you can check yourself.

Researchers have long compared how quickly new medicines reach patients in different countries. A 1992 review of this “drug lag” literature found that most studies put the United States, Sweden and Norway among the slowest countries to introduce new drugs, and the United Kingdom and West Germany among the fastest. The US and Norway had also introduced far fewer new drugs than other industrialised countries.²⁴

The picture later reversed. For new medicines approved from 2001 to 2010, the FDA’s first review took a median of 303 days and the EMA’s 366 days, and of the medicines approved in both, about two in three reached the US first.²⁵ FDA review times fell from more than three years in 1983 to less than one year in 2017, while the total time from permission to start tests in people to approval stayed at about eight years.²¹ Most of the wait now goes into the studies themselves.

Estimates vary by a factor of more than two. A study based on a survey of ten companies put the average cost per approved new medicine at $2,558 million, including the cost of capital and of failed candidates (2013 dollars).²⁶ A study using public data for 63 medicines put the median at $985 million and the mean at $1,336 million (2018 dollars); its data came mainly from smaller companies.²³ Both include the cost of the many candidates that fail.

Trials are paid for by whoever expects to earn the money back, usually through a patent. A 2021 study found that developers are “rarely interested” in research on approved medicines that have lost patent or regulatory protection, so new uses are mostly tested by academic groups that struggle to fund large trials.²⁷ The EMA has said the same about old medicines and started a pilot in 2021 to help non-profit researchers, but only for medicines that are already authorised.²⁸

Our reading, not a finding of these sources: the same economics apply to many peptides sold as research chemicals. The FDA found no human trial of BPC-157 beyond a phase 2 study from 2005, reported only as a meeting abstract, and no approved product containing it in any country.²⁹ When no one pays for trials, the questions a regulator would ask, about dose, effect and long-term safety, stay unanswered. People who use such products carry that uncertainty themselves. The history is on the Croatian lab page.

• Accelerated assessment (EU). A review of up to 150 days instead of 210 for medicines of major interest to public health.³⁴
• PRIME (EU, 2016). Early EMA advice to developers of medicines for unmet needs, so their studies answer the right questions.³⁵
• Conditional marketing authorisation (EU, 2006). Granted on less complete data when having the medicine now outweighs the risk of missing data; valid one year at a time.³⁰
• Authorisation under exceptional circumstances (EU). For medicines whose full benefits cannot be shown, because the disease is rare or full data cannot be collected.^13,36 Afamelanotide (Scenesse), for a rare disease that makes light painful, was authorised this way in 2014 and is reviewed every year.³⁶
• Compassionate use (EU). A medicine still in development for a group of seriously ill patients who cannot join a trial. In the Netherlands only a manufacturer can apply.^31,33
• Named-patient use (in the Netherlands, the artsenverklaring). A doctor requests an unauthorised medicine for one patient, on the doctor’s own responsibility, and a pharmacy or wholesaler asks the health inspectorate for an exemption.^13,32
• In the US. Treatment use of investigational drugs for life-threatening diseases from 1987; accelerated approval on an early sign of benefit from 1992, if the company later confirms a real benefit.⁴

Speed moves uncertainty from before approval to after it. A 2019 review looked at 93 cancer indications given US accelerated approval from 1992 to 2017. Confirmatory trials had shown longer survival for 19, about one in five. Another 39 had confirmed only an early marker, and others were still delayed, pending or running.³⁷ Of 68 cancer indications approved by the EMA from 2009 to 2013, about half had still not shown a gain in survival or quality of life after a median of 5.4 years.³⁸

Requirements have also loosened over time. The share of new US approvals supported by at least two pivotal trials fell from 80.6% in 1995–1997 to 52.8% in 2015–2017, and the authors conclude that the FDA has “increasingly accepted less data and more surrogate measures”.²¹ The EMA, for its part, found in a ten-year review that conditional authorisation can speed up access to new medicines.³⁰

Every route except named-patient use needs a company that is developing the medicine, with trials running or an application filed.^33,30 For most peptides sold as research chemicals, the EMA has no record of an EU authorisation or application, and none has an approved Dutch compassionate-use programme, so there is little for these routes to work with.^39,40 Named-patient use needs a doctor willing to take responsibility and a supplier that can deliver a product lawfully.³² See legal routes and talking to your doctor.