Stop-reason taxonomy

Clinical trial stop reasons

See how sponsor-provided stop language is organized into efficacy/futility, safety, operational, regulatory, and other/unknown research signals.

Stopped records23,617

Terminated, withdrawn, and suspended records.

Likely biological signals1,821

Efficacy/futility and safety classifications.

Comparison groups5

Distinct rows in the comparison below.

Dataset updated2026-08-14

ClinicalTrials.gov API v2

Current dataset

Stop-reason categories in the current dataset

The categories summarize registry language for screening. They do not replace the sponsor's full statement, the protocol, results, publications, or regulatory evidence.

Stop-reason categoryTotal stoppedLikely biologicalEfficacy / futilitySafetyOperational
EFFICACY/FUTILITYWeak benefit, endpoint, or futility language1,1001,100100% of stopped1,10000
SAFETYSafety, toxicity, or risk-benefit language721721100% of stopped07210
OPERATIONALEnrollment, logistics, funding, or execution signals12,11900% of stopped0012,119
REGULATORYRegulator or authorization-related language9100% of stopped000
Other / unknownNo reliable specific signal in the compact record9,58600% of stopped000
EFFICACY/FUTILITYWeak benefit, endpoint, or futility language
1,100 stopped
Biological
1,100 (100%)
Efficacy
1,100
Safety
0
Operational
0
SAFETYSafety, toxicity, or risk-benefit language
721 stopped
Biological
721 (100%)
Efficacy
0
Safety
721
Operational
0
OPERATIONALEnrollment, logistics, funding, or execution signals
12,119 stopped
Biological
0 (0%)
Efficacy
0
Safety
0
Operational
12,119
REGULATORYRegulator or authorization-related language
91 stopped
Biological
0 (0%)
Efficacy
0
Safety
0
Operational
0
Other / unknownNo reliable specific signal in the compact record
9,586 stopped
Biological
0 (0%)
Efficacy
0
Safety
0
Operational
0

How to use the data

What this comparison does and does not show

Efficacy and futility are related, not identical

Lack of efficacy can describe observed insufficient benefit, while futility may follow an interim assessment that continuing is unlikely to achieve the planned objective. The database groups both because they are commonly searched together as biological stop signals.

Safety language needs careful reading

A safety stop may involve toxicity, adverse events, tolerability, risk-benefit review, or a precautionary action. The classification identifies the signal; it does not determine causality or quantify clinical risk.

Operational is not a failed drug

Enrollment, logistics, funding, supply, site execution, and portfolio decisions can stop a study without demonstrating that the intervention lacks efficacy or has an unacceptable safety profile.

Unknown is an honest category

Registry explanations can be missing, generic, or too ambiguous to support a specific classification. Keeping those records in other/unknown is more reliable than forcing every stop into a dramatic narrative.

Method and limitations

Read the classification as a screening signal

The classifier uses structured fields and sponsor-provided stop text from the current dataset. Reason labels are analytical screening categories and can simplify more complex real-world decisions.

Likely biological failure combines efficacy/futility and safety records. Regulatory and operational records are kept separate, while ambiguous language remains other/unknown.

Every high-value conclusion should be checked against the linked NCT record and, where relevant, sponsor disclosures, publications, and regulatory documents.

How classification works →

Source-linked examples

Records behind the comparison

Examples are selected from the current ingest and link to the corresponding NCT evidence page.

EFFICACY/FUTILITYNCT06013371

PDE4 Inhibition in Seborrheic Dermatitis and Papulopustular Rosacea

Icahn School of Medicine at Mount Sinai

The study ended early after interim analysis.

Open source-linked record →
OTHER/UNKNOWNNCT05691478

A Study to Test the Addition of the Drug Cabozantinib to Chemotherapy in Patients With Newly Diagnosed Osteosarcoma

National Cancer Institute (NCI)

Other - Extend the observation period for current patients

Open source-linked record →
OPERATIONALNCT06694454

Neoadjuvant Inhaled Azacytidine With Platinum-Based Chemotherapy and Durvalumab (MEDI4736) - a Combined Epigenetic-Immunotherapy (AZA-AEGEAN) Regimen for Operable Early-Stage Non-Small Cell Lung Cancer (NSCLC)

National Cancer Institute (NCI)

Placed on voluntary hold to new recruitment pending modification to combination monoclonal antibody treatment.

Open source-linked record →
REGULATORYNCT05652504

Safety Evaluation of PfSPZ Vaccine in Pregnant Women in Mali (MalVIP1)

National Institute of Allergy and Infectious Diseases (NIAID)

The study remains on hold due to a funding issue after it's been lifted by the FDA.

Open source-linked record →
SAFETYNCT05401058

Low-dose Droperidol for Prevention of Postoperative Delirium in Elderly Patients After Non-cardiac Surgery

RenJi Hospital

The trial was terminated on the independent DSMB's recommendation after an interim analysis showed no evidence of benefit and a numerically higher incidence of postoperative delirium in the droperidol group,...

Open source-linked record →

FAQ

Questions about this comparison

What are the main reasons clinical trials stop?

The dataset groups stop language into efficacy/futility, safety, operational, regulatory, and other/unknown categories. Operational and unclear explanations are the largest groups.

Is futility the same as a failed endpoint?

Not always. Futility can be based on an interim probability of success, while a failed endpoint usually refers to an observed result. Registry wording and study context determine the interpretation.

Can the classifier determine why a drug failed?

It can surface likely signals from registry language, but it cannot replace clinical, statistical, regulatory, or sponsor-level review.