Disease-area signal mix
Safety and efficacy failure signals change by disease area
Hematology (non-onc) has the most safety-heavy biological-signal mix in the comparison. Dermatology sits at the efficacy-heavy end. The reason mix changes materially by disease area.
- Hematology (non-onc) has the highest safety share among biological signals: 63.3% (19 of 30).
- Dermatology has the highest efficacy/futility share among biological signals: 87.5% (28 of 32).
- Neurology contains 92 efficacy/futility signals and 27 safety signals.
- Oncology contributes 305 efficacy/futility and 277 safety signals, the largest absolute biological-signal count.
- The comparison includes disease areas with at least 200 stopped records and does not estimate overall clinical trial failure rates.
The same failure label can hide different problems
It is tempting to combine efficacy, futility, and safety into one biological-failure number. That is useful for a first filter, but it hides an important difference. A study that stops because benefit is insufficient is not the same analytical event as a study that stops because toxicity or tolerability changes the benefit-risk balance.
The current database shows that the mix between these signals changes across disease areas. Among areas with at least 200 stopped records, some have more safety than efficacy/futility signals. Others show the reverse by a wide margin.
Hematology (non-onc) leans toward safety
Hematology (non-onc) has the clearest safety-heavy biological-signal mix in the comparison. Safety accounts for 63.3% (19 of 30), compared with 11 efficacy/futility signals.
This does not establish that hematology (non-onc) trials are generally less safe. The denominator contains stopped records only, and the biological-signal subset contains 30 records. It shows what kind of source explanation appears more often when a stopped record in this slice carries a biological classification.
Neurology points much more strongly toward efficacy
Neurology sits toward the efficacy-heavy side of the comparison. Its stopped records contain 92 efficacy/futility signals and 27 safety signals. Efficacy therefore represents 77.3% (92 of 119) of its biological-signal subset.
That can direct the research workflow toward endpoints, treatment effect, futility analyses, patient selection, and whether a program produced enough measurable benefit. The pattern remains descriptive rather than predictive: it does not estimate the chance that a new neurology trial will fail.
Oncology is large and relatively balanced
Oncology contributes by far the largest number of biological signals in absolute terms: 305 efficacy/futility and 277 safety records. Safety represents 47.6% (277 of 582) of that subset, making the mix much more balanced than in neurology or dermatology.
That scale makes oncology useful for subgroup analysis, but raw counts should not be confused with a higher underlying risk. Phase, intervention, condition, and source wording can all change the interpretation substantially.
Why percentages need a minimum denominator
Very small categories can produce dramatic percentages from only a few records. To reduce that distortion, this comparison includes disease areas with at least 200 stopped trials. Even then, I would read both the percentage and the underlying counts.
A safety share based on dozens of biological signals is less stable than one based on hundreds. The table therefore reports efficacy and safety counts together instead of presenting a percentage without its denominator.
How to use the comparison responsibly
I would use this analysis to choose the first question, not the final answer. In a safety-heavy area, start with dose, adverse events, monitoring, exposure, and benefit-risk. In an efficacy-heavy area, start with endpoints, effect size, futility rules, population selection, and comparator performance.
Every important conclusion should still return to the individual NCT record and supporting evidence. These classifications organize public source language into a searchable research signal. They do not replace clinical, statistical, or regulatory review.
Disease areas ranked by safety share
| Disease area | Safety / biological signals |
|---|---|
| Hematology (non-onc) | 63.3% (19 / 30) |
| Musculoskeletal | 60.0% (6 / 10) |
| Endocrine & Metabolic | 52.5% (31 / 59) |
| Oncology | 47.6% (277 / 582) |
| Infectious Disease | 40.0% (58 / 145) |
| Cardiovascular | 39.5% (34 / 86) |
| Respiratory | 39.3% (24 / 61) |
| Immunology & Autoimmune | 38.8% (26 / 67) |
| Gastroenterology & Hepatology | 37.4% (52 / 139) |
| Ophthalmology | 37.1% (13 / 35) |
Disease areas ranked by efficacy share
| Disease area | Efficacy / biological signals |
|---|---|
| Dermatology | 87.5% (28 / 32) |
| Neurology | 77.3% (92 / 119) |
| Renal & Urology | 70.6% (24 / 34) |
| Psychiatry & Mental Health | 67.6% (23 / 34) |
| Other | 66.8% (259 / 388) |
| Ophthalmology | 62.9% (22 / 35) |
| Gastroenterology & Hepatology | 62.6% (87 / 139) |
| Immunology & Autoimmune | 61.2% (41 / 67) |
| Respiratory | 60.7% (37 / 61) |
| Cardiovascular | 60.5% (52 / 86) |
Absolute biological-signal counts
| Disease area | Efficacy / safety |
|---|---|
| Oncology | 305 / 277 |
| Other | 259 / 129 |
| Infectious Disease | 87 / 58 |
| Gastroenterology & Hepatology | 87 / 52 |
| Neurology | 92 / 27 |
| Cardiovascular | 52 / 34 |
| Immunology & Autoimmune | 41 / 26 |
| Respiratory | 37 / 24 |
| Endocrine & Metabolic | 28 / 31 |
| Ophthalmology | 22 / 13 |
Continue from here
FAQ
Which disease area has the most safety-heavy biological signal mix?
Among disease areas with at least 200 stopped records, Hematology (non-onc) has the highest safety share in the current dataset: 63.3% (19 of 30).
Does this show which therapeutic area has the highest trial failure rate?
No. The database contains stopped trials rather than all initiated trials, so it cannot estimate an overall failure rate by therapeutic area.
Why compare counts as well as percentages?
Percentages can look unstable when the underlying biological-signal count is small. Counts show the denominator and make differences between disease areas easier to interpret responsibly.
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.