Drug development

How Do Scientists Evaluate Drug Safety?

Drug safety is assessed in stages, from animal toxicology to clinical trials to monitoring after approval. Because trials are limited in size, rare side effects are often found only once a drug is widely used.

Key points

  • Safety evaluation starts with preclinical toxicology, continues through every clinical trial phase, and carries on for as long as a drug is on the market.
  • International guidelines expect about 1,500 people to be exposed to a new drug for chronic use before approval, including 300 to 600 for six months and about 100 for a year.
  • A trial of a given size can easily miss rare side effects. If no event appears in n patients, the true rate could still be as high as about 3/n.
  • After approval, the FDA monitors safety through spontaneous reports (FAERS) and active surveillance (Sentinel). Individual reports cannot prove a drug caused an event.
  • Among 222 novel drugs approved by the FDA from 2001 to 2010, 32% had a postmarket safety event such as a boxed warning or withdrawal.

No drug is completely safe. Every medicine that affects the body in a useful way can also affect it in harmful ways. The goal of safety evaluation is not to prove a drug is harmless. It is to understand its risks well enough to judge whether its benefits outweigh them, and to keep watching for risks that only appear later.

Stage 1: Before human testing

Safety evaluation begins in the laboratory. Before a drug can be given to people, the sponsor conducts preclinical toxicology studies, usually in more than one animal species. These identify:

  • which organs the drug can damage,
  • the doses at which harm occurs, and the highest dose with no observed harm,
  • whether effects are reversible, and
  • specific risks such as damage to genes, effects on fertility, or harm to a developing fetus.

The results determine whether human testing can begin and at what starting dose. See How Are New Drugs Discovered?

Stage 2: During clinical trials

Phase by phase

  • Phase 1 looks for common side effects and defines a tolerable dose range, usually in a small number of participants.
  • Phase 2 gathers safety data in patients with the disease, at the doses being considered.
  • Phase 3 provides the largest pre-approval safety database, and compares the drug's side effects directly with placebo or an existing treatment.

Recording adverse events

Every unfavorable medical occurrence in a trial participant is recorded as an adverse event, whether or not it seems related to the drug. Investigators then assess whether each event is plausibly a reaction to the drug. Under international definitions (ICH E2A), an event is serious if it:

  • results in death,
  • is life-threatening,
  • requires or prolongs hospitalization,
  • results in persistent or significant disability, or
  • causes a congenital anomaly or birth defect,

or if it is otherwise medically important.

In the United States, sponsors must report serious, unexpected suspected reactions to the FDA quickly. Under 21 CFR 312.32, the deadline is generally 15 calendar days, or 7 days for unexpected fatal or life-threatening reactions.

Independent monitoring

Many larger trials have an independent data monitoring committee that reviews unblinded safety data while the trial is running and can recommend stopping it if participants are being harmed.

How much exposure is enough?

International guidelines (ICH E1) set expectations for drugs intended for long-term treatment of non-life-threatening conditions:

Exposure Typical expectation before approval
Total people exposed to the drug About 1,500
Exposed for 6 months at relevant doses 300 to 600
Exposed for at least 1 year About 100

Source: ICH E1. Requirements can be higher or lower depending on the drug, the disease, and known risks.

The rare side-effect problem

Even a well-run program cannot detect every risk before approval. A useful rule of thumb from biostatistics, known as the rule of three, makes this concrete. If a side effect is not seen at all among n people, the true rate could still be as high as about 3 in n (with 95% confidence). This was described in a 1983 JAMA paper by Hanley and Lippman-Hand.

So if no cases of a particular liver injury appear among 1,500 trial participants, the data are still consistent with a true rate of up to about 1 in 500 patients. A side effect affecting 1 in 10,000 patients will very likely be missed entirely before approval.

Stage 3: After approval

Once a drug is used by many thousands or millions of people, rare and delayed risks can emerge. The FDA uses several tools to find them:

  • FAERS (FDA Adverse Event Reporting System). A database of reports submitted by patients, healthcare professionals, and manufacturers.
  • Sentinel. An active surveillance system that analyzes electronic health records and insurance claims data to study safety questions in large populations.
  • Postmarketing requirements. Studies the FDA can require as a condition of approval, sometimes called Phase 4 studies.
  • REMS (Risk Evaluation and Mitigation Strategies). Programs required for some drugs with serious risks, for example restricting who can prescribe them or requiring patient monitoring.

A caution about adverse event reports

The FDA states plainly that a report in FAERS does not mean the drug caused the event. Reports are voluntary, incomplete, and influenced by publicity. Because nobody knows how many people took the drug without reporting a problem, FAERS cannot be used to calculate how often an event occurs. Reports are signals that may warrant investigation, not measurements of risk. Spectral Genomics always labels adverse event report data this way.

How often new risks are found

A 2017 study in JAMA examined the 222 novel therapeutics approved by the FDA from 2001 through 2010. Over a median follow-up of 11.7 years:

  • 71 (32.0%) had at least one new postmarket safety event: a withdrawal, a new boxed warning, or a safety communication.
  • The median time from approval to the first event was 4.2 years.
  • Events were more frequent among biologics, psychiatric drugs, drugs given accelerated approval, and drugs approved close to their regulatory deadline.

Weighing benefit against risk

Safety is always judged relative to the disease and the alternatives. A side effect that would be unacceptable in a drug for mild allergies might be acceptable in a drug that extends survival in advanced cancer. Regulators approve a drug for a specific use when they judge its benefits outweigh its known risks for the people who will take it, and they keep revisiting that judgment as new evidence arrives.

Sources

  1. International Council for Harmonisation. E1: The Extent of Population Exposure to Assess Clinical Safety for Drugs Intended for Long-Term Treatment of Non-Life-Threatening Conditions.Regulator
  2. International Council for Harmonisation. E2A: Clinical Safety Data Management: Definitions and Standards for Expedited Reporting.Regulator
  3. Hanley JA, Lippman-Hand A. If nothing goes wrong, is everything all right? Interpreting zero numerators. JAMA. 1983;249(13):1743–1745.Peer-reviewed
  4. Downing NS, et al. Postmarket safety events among novel therapeutics approved by the US Food and Drug Administration between 2001 and 2010. JAMA. 2017;317(18):1854–1863.Peer-reviewed
  5. U.S. Food and Drug Administration. Questions and Answers on FDA's Adverse Event Reporting System (FAERS).Regulator
  6. U.S. Food and Drug Administration. FDA's Sentinel Initiative.Regulator
  7. U.S. Food and Drug Administration. Risk Evaluation and Mitigation Strategies (REMS).Regulator
  8. 21 CFR 312.32: IND safety reporting.Statute / regulation

This article is for education. It is not medical advice and not investment advice. Figures are drawn from the sources above and describe historical averages, not the prospects of any particular drug. See our editorial standards and disclaimer.