How clinical trials decide, when most treatments tested in people fail

Key takeaways

• An industry analysis of 9,704 drug programs found that only 7.9% of drugs entering the first phase of human testing went on to win FDA approval.
• In a major NIH trial of glucosamine and chondroitin, 60.1% of people taking a placebo reported at least 20% less knee pain, which is why trials need comparison groups.
• Before you join a study, an ethics board must approve it, and federal rules say you can quit at any time without penalty.
• Dietary supplements don’t need FDA review or a clinical trial before they are sold, so “clinically studied” on a label is not FDA approval.
• A ClinicalTrials.gov listing comes from the study’s sponsor, and being listed does not mean the government has evaluated the study.

At 16 medical centers across the United States, 1,583 people with painful knee arthritis agreed to take capsules for 24 weeks¹². Some got glucosamine. Some got chondroitin. Others got both, the prescription pain drug celecoxib (Celebrex), or a placebo, a look-alike pill with no active ingredient².

The results appeared in the New England Journal of Medicine in February 2006. Among volunteers taking the dummy pill, 60.1% said their knee pain dropped by at least a fifth². The supplements did a bit better, but not by enough to rule out chance².

That study, the Glucosamine/chondroitin Arthritis Intervention Trial (GAIT), is a good window into how clinical trials work. A clinical trial tests a treatment in people, following a plan written in advance. Trials are how a promising idea becomes a proven treatment, or doesn’t.

Most ideas don’t. The Biotechnology Innovation Organization (BIO), the biotech industry’s trade group, and two research firms tracked 9,704 drug programs from 2011 to 2020. Only 7.9% of drugs that entered the first phase of human testing reached approval⁵.

Every trial asks one question: compared with what?

Symptoms rise and fall on their own. Arthritis pain flares and eases. Colds end. Hope and extra attention can make people feel better, too. If you give a treatment and simply watch, you can’t tell which of these caused the change.

That’s why a good trial includes a control group. These people get a placebo, usual care or another treatment instead of the one being tested. The FDA lists the control group among the core choices in designing a trial, along with who can join, how long it lasts, the dose and how the data will be analyzed³.

GAIT shows why this matters. If researchers had given the two supplements together to everyone and counted who improved, about two out of three would have reported less pain². That looks like a win. Only the placebo group revealed that most of that improvement would have happened anyway.

The trial also had a “positive control,” celecoxib, a drug already known to ease arthritis pain¹. It beat the placebo, 70.1% to 60.1%². That showed the trial could detect a real effect when one existed.

Chance and blindness keep everyone honest

Two more features protect the answer: randomization and blinding.

Randomization means people are put into groups by chance, like a coin flip. It keeps researchers from steering healthier or more hopeful people into the treatment group. With enough volunteers, chance spreads age, health and habits evenly. A difference at the end is then more likely to come from the treatment.

Blinding means people don’t know which group they’re in. In a double-blind trial, neither the volunteers nor the researchers judging them know. GAIT was double-blind¹⁵. That matters because expectations shape how people rate their pain and how researchers judge progress.

The finish line is drawn before the race starts

Every trial names a primary endpoint, the main result it is built to measure. GAIT’s was simple. Did a person’s score on a standard arthritis pain questionnaire, called WOMAC, drop by at least 20% after 24 weeks²?

Picking the endpoint in advance matters. A big trial collects dozens of measurements. If researchers could choose the best-looking one afterward, almost any treatment could look good by luck.

Public registration helps keep everyone to the plan. Researchers Robert Kaplan and Veronica Irvin studied 55 large heart trials paid for by NIH’s National Heart, Lung, and Blood Institute¹⁴.

Before 2000, 57% reported a clear benefit on their main outcome. After 2000, every trial in the group had registered its plans in advance on ClinicalTrials.gov, and only 8% did¹⁴. The authors said they couldn’t prove registration caused the drop, but it was the factor most strongly tied to it¹⁴.

GAIT also shows how to treat a result that wasn’t the main question. Among the 22% of volunteers with moderate to severe pain, 79.2% improved on the combination, compared with 54.3% on placebo¹². The researchers called this exploratory and said a new trial designed for that purpose would need to confirm it¹².

Two-year follow-up reports didn’t settle the question in the supplements’ favor. No treatment gave a clinically important improvement over placebo in pain or function⁶. Joint-space narrowing, an X-ray sign of cartilage loss, didn’t differ either⁷. Our guide to study limitations explains why subgroup findings so often fade.

Before the first dose: an FDA filing, an ethics board and your consent

A drug can’t be tested in people until its sponsor files an Investigational New Drug application, or IND, with the FDA. It must include animal and toxicity data, manufacturing details, the trial plans and information about the researchers³. The FDA has 30 days to review it and can put a trial on hold, though the agency says holds are rare³.

An institutional review board must also approve the study. This ethics committee reviews the consent materials, and any changes to them need its approval before they are used¹⁰.

Federal rules list eight basic things you must be told before you agree to join, according to the HHS Office for Human Research Protections (OHRP)⁹. In plain terms, you should learn:

• that the study is research, its purpose, how long it lasts and which procedures are experimental
• the risks and discomforts researchers can foresee
• the benefits you or others might reasonably expect
• other treatments that might help you
• how your records will be kept confidential
• for riskier studies, what care or payment is available if you are hurt
• whom to call with questions about the study or your rights
• that joining is voluntary and you can quit at any time without penalty⁹

Consent is more than a signature. OHRP calls it an ongoing exchange that begins with the first contact, such as a flyer or an ad, and says a signed form alone isn’t an adequate process¹⁰. The form should open with a short summary of the information most likely to help you decide¹¹.

Payment is allowed. OHRP calls paying volunteers a common practice that is generally acceptable, as long as the amount is fair and not big enough to sway people unduly¹⁰. For long studies, it recommends paying in installments. A lump sum held until the finish could make people reluctant to use their right to quit¹⁰.

Phase by phase: a funnel most drugs don’t survive

The FDA describes drug testing in four phases³:

Phase	Who takes part	Typical length	Main question
Phase 1	20 to 100 healthy volunteers or patients	Several months	Is it safe, and at what dose?
Phase 2	Up to several hundred patients	Several months to 2 years	Does it seem to work? What side effects show up?
Phase 3	300 to 3,000 patients	1 to 4 years	Does it beat the comparison? What less common side effects appear?
Phase 4	Several thousand patients	After approval	What happens in wider use?

BIO’s analysis puts numbers on the funnel⁵. From 2011 to 2020, 52% of drugs passed phase 1, 28.9% passed phase 2 and 57.8% passed phase 3. Of applications that reached the FDA, 90.6% were approved⁵.

The FDA’s own patient guide, last updated in 2018, offers rougher estimates: about 70% of drugs pass phase 1, 33% pass phase 2 and 25 to 30% pass phase 3³. The two sources cover different years and count differently, so treat any single figure as approximate.

Phase 2 is the biggest hurdle. The BIO report says it consistently has the lowest success rate of any phase⁵. It is usually the first real test of whether a treatment helps patients, not just whether it is safe.

Odds vary by disease. Cancer drugs had a 5.3% chance of approval from phase 1, while drugs for blood disorders had 23.9%⁵. Drugs tested in patients chosen with a biomarker, a measurable sign such as a gene variant, had about twice the odds: 15.9% versus 7.6%⁵. The trip from first human dose to approval averaged 10.5 years⁵.

The FDA’s patient guide, written in 2018, says a developer needs adequate data from two large, controlled trials before applying³. The sponsor then files a New Drug Application with all of its data.

FDA physicians, statisticians and other reviewers check the results, and inspectors visit trial sites to look for faked or hidden data⁴. The same guide gives reviewers 6 to 10 months to decide once a complete application is filed⁴. Approval means the FDA found the drug safe and effective for its intended use⁴.

Supplements don’t have to stand in this line

Dietary supplements follow different rules. The FDA generally doesn’t review supplements before they go on sale, according to NIH’s National Center for Complementary and Integrative Health (NCCIH)⁸. The main exception is a new ingredient that wasn’t sold in the U.S. before 1994, which requires safety information⁸.

Supplement makers are responsible for their own safety evidence and for honest labels. The FDA can step in against a problem product, but only after it is already for sale. A drug, by contrast, must prove it is safe and effective before it reaches pharmacy shelves⁸. A supplement label also can’t claim to treat or cure a disease⁸.

So no law requires a supplement to pass a trial before it reaches the shelf. GAIT happened because NIH paid for it¹.

GAIT also reveals a quality wrinkle. Its glucosamine and chondroitin were made by an FDA-licensed research pharmacy and tested for purity and strength¹. The authors warned that products made to those drug-grade standards may not match what stores sell².

In some other countries, certain glucosamine and chondroitin products are prescription drugs. NCCIH notes that prescription or pharmaceutical-grade versions did better in some analyses, though bias in older, company-funded studies may explain part of that⁷.

Where does that leave the evidence? Expert groups disagree. Two 2019 arthritis guidelines advise against glucosamine and chondroitin for knee osteoarthritis, while a 2021 orthopedic guideline lists them as possibly helpful for mild to moderate knee arthritis⁷.

NCCIH reports no major safety problems in large studies. Still, glucosamine may raise blood sugar in some people, and both have been linked to bleeding risk in people taking the blood thinner warfarin⁷. Tell your clinician if you take them, especially if you have diabetes or take a blood thinner.

If you value natural remedies, trials like GAIT are allies, not enemies. They are the fairest way to separate a remedy that helps from one riding on hope and the body’s natural ups and downs. When NIH announced the results, Stephen Straus, then director of its complementary medicine center, said: “It is important to study dietary supplements with well-designed research in order to find out what works and what does not”¹.

How to read a ClinicalTrials.gov listing

ClinicalTrials.gov, run by the National Library of Medicine at NIH, calls itself the biggest database of clinical studies in the world, covering both public and private funding¹². GAIT is listed there as NCT00032890¹⁵.

Two facts shape how you should read it. The information comes from the study’s sponsor or lead researcher, and a listing does not mean the federal government has evaluated the study¹². Think of it as a public notice, not a seal of approval.

Each record covers the study’s purpose, whether it is recruiting, who can join, where it runs and whom to contact¹³. Completed studies may also post results¹². Read a record in this order:

1. Status. Is it recruiting, finished or stopped early? A study that stopped early deserves a question about why.
2. Design. Look for “randomized,” the type of blinding and what the comparison group gets.
3. Primary outcome. What counts as success, and when is it measured? Later, compare it with what the published paper or press release highlights.
4. Eligibility. Age, diagnosis, other conditions and medicines. These also tell you whom the results will apply to.
5. Sponsor. Who is paying? Funding doesn’t make a study wrong, but you should know it.
6. Results. If a trial ended years ago and its results appear nowhere, that silence is information, too.

A checklist for trial ads and “clinically tested” claims

Use the first table for a study ad and the second for a product claim.

If a study ad says…	Ask	A reassuring answer
“Join our research study”	What is the NCT number?	It matches a ClinicalTrials.gov record with the same sponsor and purpose
“Try a new treatment”	Which phase is it, and what does the comparison group get?	A clear phase and an honest note that you may get a placebo
“Compensation available”	How much, when, and who pays for visits and tests?	Payment spread over the study, with no fee for you to join
“Sign up today”	May I take the consent form home first?	Yes, with time to talk it over with your own clinician
“Questions? Call us”	Whom do I call about my rights as a volunteer?	A name and number listed in the consent form9

If a product says…	Ask	A strong answer
“Clinically tested”	Tested in people, with a placebo group?	A randomized, blinded human trial
“Proven results”	Did it meet its main, pre-declared endpoint?	Yes, not just in a subgroup
“Studied ingredient”	Was the same form and dose tested?	The product matches what was studied
“Doctor recommended”	Who paid, and has anyone repeated it?	Independent funding or an independent repeat

If something goes wrong during a trial, call the contact listed for questions about your rights, and tell your own clinician⁹. To report a bad reaction to a supplement, use the HHS Safety Reporting Portal⁸.

What the evidence shows

Claim	Evidence rating	Basis
Comparison groups, randomization and blinding make trial results more reliable	Strong evidence	Built into FDA trial design; GAIT’s placebo response shows why23
Most drugs that enter human testing never win approval	Strong evidence	7.9% of drugs reached approval from phase 1, 2011 to 20205
Glucosamine and chondroitin ease knee arthritis pain	Promising but early	Mixed results; guidelines disagree267
Registering outcomes in advance curbs overly rosy trial results	Promising but early	One study of 55 trials linked registration to far fewer positive results14

Questions readers ask

The bottom line

Clinical trials are a filter, and a tough one. Most drugs that enter human testing never make it through. The tools that do the filtering are simple: a comparison group, assignment by chance, blinding, and a finish line drawn in advance. Ethics boards and consent rules exist to protect the people who make that testing possible.

Supplements can skip that filter, which makes independent trials like GAIT especially valuable. When you see a trial ad or a “clinically tested” label, ask the same questions researchers ask: compared with what, measured how, in whom, and paid for by whom.

Related reading

• Drugs that quieted extra heartbeats raised deaths, a lesson in reading a study’s limits
• Most health products aren’t scams, but many just haven’t been proven
• Who paid for the study matters, even when the science is real
• Why supplement labels say ‘supports’ and never say ‘cures’
• Only 1 in 20 treatments tested in animals wins approval for people

Sources

1. National Institutes of Health. “Efficacy of Glucosamine, Chondroitin Sulfate May Depend on Level of Osteoarthritis Pain.” NIH News Release, February 22, 2006.
2. Clegg DO, Reda DJ, Harris CL, et al. “Glucosamine, Chondroitin Sulfate, and the Two in Combination for Painful Knee Osteoarthritis.” New England Journal of Medicine, 2006;354:795-808. Full text accessed via farm.ucl.ac.be
3. U.S. Food and Drug Administration. “Step 3: Clinical Research.” FDA, content current as of 2018. fda.gov
4. U.S. Food and Drug Administration. “Step 4: FDA Drug Review.” FDA, content current as of 2018. fda.gov
5. Biotechnology Innovation Organization, Informa Pharma Intelligence and QLS Advisors. “Clinical Development Success Rates and Contributing Factors 2011–2020.” BIO, 2021. go.bio.org
6. Sawitzke AD, Shi H, Finco MF, et al. “Clinical Efficacy and Safety of Glucosamine, Chondroitin Sulphate, Their Combination, Celecoxib or Placebo Taken to Treat Osteoarthritis of the Knee: 2-Year Results From GAIT.” Annals of the Rheumatic Diseases, 2010;69(8):1459. pubmed.ncbi.nlm.nih.gov
7. National Center for Complementary and Integrative Health. “Glucosamine and Chondroitin for Osteoarthritis: What You Need To Know.” NIH, 2023. nccih.nih.gov
8. National Center for Complementary and Integrative Health. “Using Dietary Supplements Wisely.” NIH. nccih.nih.gov
9. HHS Office for Human Research Protections. “What Are the Basic Elements of Informed Consent?” HHS. hhs.gov
10. HHS Office for Human Research Protections. “Informed Consent FAQs.” HHS. hhs.gov
11. Yale Human Research Protection Program. “IRB Essentials: Basics of Informed Consent.” Yale University, May 2025. research-support.yale.edu
12. National Library of Medicine. “ClinicalTrials.gov.” NIH fact sheet, 2021. nnlm.gov
13. National Library of Medicine. “ClinicalTrials.gov Help for Patients and Families.” NIH.
14. Kaplan RM, Irvin VL. “Likelihood of Null Effects of Large NHLBI Clinical Trials Has Increased over Time.” PLoS ONE, 2015;10(8):e0132382. journals.plos.org
15. QxMD Read. “Glucosamine, Chondroitin Sulfate, and the Two in Combination for Painful Knee Osteoarthritis” (abstract and trial registration NCT00032890). qxmd.com