Failure signals by phase
Safety dominates Phase I signals; efficacy dominates Phase II and III
Safety accounts for 60.0% of Phase I biological signals. In Phase III, efficacy/futility accounts for 71.2%. The composition changes as development advances.
- Phase I has 257 safety signals and 171 efficacy/futility signals among stopped records.
- 60.0% of classified Phase I biological signals are safety-related.
- Phase II has 599 efficacy/futility signals and 374 safety signals.
- Phase III has 354 efficacy/futility signals and 143 safety signals.
- These are signal counts inside stopped records, not overall failure probabilities for trials entering each phase.
The pattern across development
The biological failure signal is not the same at every stage of clinical development. In Phase I stopped records, safety signals outnumber efficacy/futility 257 to 171. The relationship reverses in Phase II (599 efficacy/futility versus 374 safety) and becomes wider in Phase III (354 versus 143).
This is one of the clearest phase-level patterns in the database. It also matches the different questions the phases are designed to answer: early development emphasizes tolerability, exposure, and dose, while later development increasingly tests whether benefit is strong and reliable enough.
Why Phase I looks different
Phase I trials are commonly built around safety, tolerability, pharmacokinetics, dose escalation, and dose selection. A safety-led stop is therefore closer to the central purpose of the phase than it would be in many later-stage programs.
That does not mean every Phase I safety signal invalidates the mechanism. Toxicity can depend on dose, schedule, formulation, combination partner, route of administration, or patient population. The trial-level context still determines how broadly the result should be interpreted.
Why efficacy becomes more visible in Phase II
Phase II is often the point where a program must demonstrate enough activity to justify larger and more expensive studies. Futility analyses, weak treatment effects, insufficient responses, and endpoint problems therefore become more visible in the stop language.
The shift from safety to efficacy signals should not be read as proof that safety no longer matters. It shows that efficacy and futility become the more common classified biological reason within the stopped Phase II records.
The gap widens in Phase III
By Phase III, efficacy and futility signals substantially outnumber safety signals in the stopped-trial slice. Confirmatory designs are intended to establish clinically and statistically meaningful benefit, so an efficacy-led stop can become a decisive program signal.
Safety remains material, particularly when larger populations or longer exposure reveal risks that were not clear earlier. But in this database, the balance of classified biological stops moves increasingly toward efficacy as development advances.
Why absolute counts need denominators
There are more records in some phases than others, and trials can carry more than one phase label. Raw counts therefore need both a phase denominator and a clear definition of the signal being counted.
The tables below show counts and signal shares within stopped records. They do not represent the failure probability of all trials entering each phase, and they should not be compared directly with industry success-rate studies that use different cohorts and denominators.
How I would use the phase pattern
For early-stage research, I would inspect safety language, dose-limiting toxicity, exposure, and whether the issue appears molecule-specific or mechanism-related. For Phase II and III, I would begin with endpoint design, futility rules, patient selection, comparator performance, and whether the effect size was clinically meaningful.
The phase pattern is a useful prior for where to look, not a substitute for source review. The decisive evidence remains in the trial record, protocol, results, sponsor communication, publication, and regulatory history.
Safety versus efficacy signals by phase
| Development phase | Efficacy / safety |
|---|---|
| Phase I | 171 / 257 |
| Phase II | 599 / 374 |
| Phase III | 354 / 143 |
Biological signals within stopped records
| Development phase | Signals / stopped records |
|---|---|
| Phase I | 6.5% (428 / 6,629) |
| Phase II | 9.1% (973 / 10,743) |
| Phase III | 12.5% (497 / 3,975) |
Continue from here
FAQ
Which failure signal is more common in stopped Phase I trials?
Safety signals are more common than efficacy or futility signals in the current stopped Phase I records.
When do efficacy signals become more common than safety signals?
In this dataset, efficacy and futility signals outnumber safety signals in Phase II and Phase III stopped records.
Are these overall clinical trial failure rates by phase?
No. They describe the composition of stopped records only and do not include every successful, completed, or ongoing trial entering each phase.
Source note: counts are generated from the current ClinicalTrials.gov-derived stopped-trial dataset used by ClinicalTrialFailures.com. These labels are analytical screening signals, not medical advice.