How do you keep a competitive landscape from going stale?
How do I keep a current, subtype-level read on my competitive landscape without commissioning a quarter-long report every time?
Competitive landscapes change continuously, but landscape reports are commissioned in quarters and age before they're read. In a custom build on the PharosBio Clinical Trial Retriever API, we filtered 36,776 breast-cancer trial records to a scoreable 135, ranked each on a composite of efficacy, response and safety, and stood up a live portfolio board that flags what changed and what to do.
Who this is for: Competitive-intelligence, strategy, and search-and-evaluation teams tracking a moving therapeutic landscape.
- trial records filtered to the scoreable set
- 36,776 → 135
- efficacy, response & safety into one composite score
- 3 → 1
- success probability that moves on new evidence
- live board
- API also available as a skill in Hydra
- custom build
trial records filtered to the scoreable set
efficacy, response & safety into one composite score
success probability that moves on new evidence
API also available as a skill in Hydra
Why quarterly competitive reviews go stale
A therapeutic landscape changes continuously, new readouts, new entrants, shifting safety signals, but competitive analyses are commissioned in quarters. By the time a deep landscape report lands, parts of it are already out of date, and re-running it for a new question means starting over from scratch.
Strategy and search-and-evaluation teams don't want a static deck; they want to ask specific questions, who leads in this subtype, where is our pipeline exposed, and get a current, evidenced answer they can defend in front of an executive team.
What teams in this space search for
- How do I keep my competitive landscape up to date?
- Which drug leads in which cancer subtype?
- Where is my pipeline exposed competitively?
How we built it
A custom monitoring build on the PharosBio Clinical Trial Retriever API. It fetches clinical endpoints from trial records, scores each trial into one composite of efficacy, response and safety, and keeps a live portfolio board that flags what changed and what to do as new evidence lands.
How the pipeline works
We started from the clinical-trial record for breast cancer - 36,776 rows - and filtered to what the scoring needs: records measuring progression-free survival (3,908), then only those with non-missing hazard ratio, safety risk ratio and ORR difference (135), then named-drug records only.
Each surviving trial was scored on three benefit dimensions - efficacy (hazard ratio), response (ORR difference) and safety (serious-AE risk ratio) - converted so higher is always better, ranked to a 0-100 percentile, and averaged into one composite score. A retrospective Dato-DXd and analogue PFS analysis across subtypes is the worked example.
On top of the scored landscape sits a live portfolio board: each asset carries a moving success probability, and when new evidence lands - a competitor readout, a new publication, your own data - it flags what changed and recommends an action.
What it found
The board turns a static deck into a living view. Drug A's probability drops 14 points in 48 hours after a competitor's Phase 2 failure and a new hepatotoxicity publication - with a recommended tox study and a clear stop criterion. Drug D's rises on a competitor success and strong week-12 data, with a recommendation to accelerate.
Because the same evidence streams are linked - publications, safety signals, clinical outcomes, PK/PD and IC50 - a redirect that would normally surface only after expensive animal studies can be called earlier.
What we learned
The difference from a commissioned report is repeatability: the analysis is a function you can re-run, not a snapshot that ages, and every score traces back to a trial, so it survives executive scrutiny.
This was a custom build on the PharosBio Clinical Trial Retriever API - the same engine behind the safety-signals work, and available as a skill in Hydra - so the board can be repointed at any indication or competitive set.
Data-filtering funnel
36,776
breast-cancer clinical-trial records (raw)
3,908
records measuring progression-free survival (PFS)
135
with non-missing HR, Safety RR and ORR difference
clean
named-drug records only (undefined drugs removed)
| Metric | Formula | Reading |
|---|---|---|
| Hazard Ratio (HR) | control mean / treatment mean | < 1 = treatment extends survival |
| ORR difference | ORR treatment − ORR control | > 0 = higher tumour response |
| Safety RR | serious-AE treatment % / control % | < 1 = fewer serious AEs |
Each metric becomes a benefit score (higher = better), is ranked to a 0-100 percentile across all trials, and the three percentiles are averaged into one composite score - so a single number captures efficacy, response and safety together.
A monitored portfolio, refreshed as new evidence lands. Each asset carries a moving success probability; click an alerted asset to see exactly what changed and what to do about it.
| Asset | Indication | Phase | Success prob. | Status |
|---|---|---|---|---|
| Drug A | T2 Diabetes | Preclinical | 32% → 18% | ALERT |
| Drug B | NASH | Preclinical | 71% → 73% | Stable |
| Drug C | Alzheimer's | Preclinical | 45% → 38% | Review |
| Drug D | Oncology | Preclinical | 82% → 85% | PRIORITIZE |
| Drug E | Parkinson's | On Hold | 12% | Stopped |
Drug A - Immediate review recommended
Probability dropped 14 points in 48 hours
What changed
- Competitor failure: A similar GLP-1/GIP dual agonist failed Phase 2 (liver-enzyme elevation in 23% of patients).
- New publication: A JAMA study links GPR40 agonism to hepatotoxicity in patients with pre-existing NAFLD.
- Your data: Week-8 preclinical results show 15% ALT elevation (borderline).
Recommended actions
- Run an extended 12-week tox study with liver monitoring (~€50K, 3 months)
- Test in a NAFLD disease model (~€30K, 6 weeks)
- If signals persist: stop before Phase 1 (save ~€45M)
Click Drug A or Drug D to see why the probability moved.
The cost of stopping too late
~90% of drugs fail. The issue is rarely failure itself - it is stopping too late, after the most expensive studies are already paid for.
competitor Phase 2 readouts
class hepatotoxicity, AE overlap
HR, ORR, PFS by subtype
exposure, potency, model relevance
Agent links them
Call the stop - or the acceleration - earlier
Connect a programme’s own PK/PD and IC50 to the literature and clinical record, and a redirect that would have surfaced after animal studies can surface before them.
One API, many workflows
This was a custom build for a partner, powered by the PharosBio Clinical Trial Retriever API - which fetches clinical endpoints and scores drug similarity by target and mechanism of action. The same API can be repointed at:
- Safety scoring for combinations
- Extrapolating clinical outcome from in vivo data
- Portfolio monitoring
- Drug repurposing
It is also available as a skill inside Hydra
What you get
- A live portfolio board with success probabilities that move on new evidence
- 36,776 trial records distilled to a defensible composite score per trial
- Per-asset alerts that say what changed and recommend an action
- Source-traceable scoring that holds up in executive review
Data sources used
- ClinicalTrials.gov (trials, design, enrichment)
- Published trial outcomes (efficacy & safety)
- cBioPortal / TCGA (subtype & biomarker context)
Figures reflect analyses PharosBio ran on public datasets and public benchmarks. Named competitors, collaborators, and logos are withheld at this stage; the methods and results shown are real and repointable to your own target.
Sources & methods
- Landscape & outcomes: ClinicalTrials.gov; published trial outcomes
- Subtype & biomarker context, cBioPortal / TCGA
Frequently asked questions
How is this different from a commissioned landscape report?
A commissioned report is a one-off snapshot that ages immediately. This pipeline re-runs against new readouts on demand and keeps a live board, turning a quarterly deliverable into a continuously refreshable view - with every score traceable to a trial.
How does the composite score work?
Three benefit scores - efficacy (1 − hazard ratio), response (ORR difference) and safety (1 − safety risk ratio) - are each ranked to a 0-100 percentile across all filtered trials, then averaged. A higher composite means a trial performs well across all three dimensions relative to the rest of the dataset.
Was this built with Hydra?
No - it was a custom build for a partner on the PharosBio Clinical Trial Retriever API. The same API is available as a skill in Hydra, so the monitoring board can be reproduced or repointed at a different indication there.
Run this analysis on your question
Hydra plans, executes, and validates, so you reach a defensible answer in hours, not weeks.