Clinical Trials & Translational Oncology
Note: This page is educational and reflects the state of oncology trial design in 2025. It does not replace medical or regulatory advice.
TL;DR
A clinical trial is the formal mechanism by which a drug, device, or strategy moves from biology to approved use. Modern oncology trials are no longer purely sequential Phase I → II → III machines — they include basket, umbrella, platform, and adaptive designs, often anchored on biomarkers instead of organ of origin. Roughly 3–5 % of cancer drugs that enter Phase I reach approval, so understanding why trials succeed or fail is part of the science. Brazilian readers should also know CONEP/Plataforma Brasil, ANVISA, and the Lei 14.874/2024 (new clinical research law).
1. Why trials exist
Three failures motivate the trial system:
- Pre-clinical promise often does not translate to humans (mouse ≠ human, cell line ≠ tumor).
- "It worked in my hands" is not a substitute for prospective, controlled, pre-registered evidence.
- Approval requires acceptable risk/benefit for a defined population — and that has to be measured, not assumed.
Trials answer two questions, in this order: does it work? and does it work better than what we already have?
2. Phases — what each one actually asks
| Phase | Question | Typical n | Outcome |
|---|---|---|---|
| Phase 0 / first-in-human microdosing | Does the drug do anything in humans (PK/PD)? | 5–15 | Pharmacodynamics, target engagement |
| Phase I | What dose is tolerable? | 15–50 | DLT, MTD/RP2D, toxicity profile |
| Phase II | Does it have biological/clinical activity? | 30–300 | ORR, PFS, signal of efficacy |
| Phase III | Is it better than standard of care? | 300–3 000+ | OS, PFS, head-to-head superiority/non-inferiority |
| Phase IV | Real-world safety post-approval | thousands | Rare AE, long-term outcomes |
In oncology, the line between phases blurs — Phase I expansion cohorts can produce registration-quality efficacy data; Phase II baskets can lead directly to approval (e.g., pembrolizumab for MSI-H tumors). Sources: [1]
3. Modern trial designs
Adaptive designs
Allow pre-specified modifications during the trial — sample size re-estimation, dropping ineffective arms, or switching to a more promising dose — without breaking statistical validity. The I-SPY 2 trial in breast cancer is the classic example. Sources: [1]
Basket trials
Test one drug across multiple tumor types that share a biomarker (e.g., NTRK fusions, BRAF V600E). Backed pembrolizumab's tissue-agnostic approval in MSI-H/dMMR tumors and larotrectinib in NTRK fusions.
Umbrella trials
Test multiple drugs in a single tumor type, each arm matched to a different biomarker. Lung-MAP and NCI-MATCH are landmark examples.
Platform trials
Long-running infrastructure where new drugs enter and leave as evidence accumulates. The control arm is shared across cohorts — efficient and patient-friendly. RECOVERY (COVID-19) showed how powerful this is; oncology adopted it for AML (Beat AML), brain tumors (GBM AGILE), and breast cancer (I-SPY). Sources: [1]
N-of-1 / personalized trials
Increasingly relevant for ultra-rare driver mutations and individualized neoantigen vaccines.
4. Endpoints — what we actually measure
| Endpoint | What it is | Strengths | Limits |
|---|---|---|---|
| OS (overall survival) | Time from randomization to death | Hardest, most patient-centric | Slow, confounded by post-progression therapy |
| PFS (progression-free survival) | Time to progression or death | Faster than OS | Surrogate; investigator-assessed bias |
| ORR (objective response rate) | % of patients with tumor shrinkage by RECIST | Quick read | Doesn't capture stable disease or duration |
| DOR (duration of response) | How long responses last | Captures durability | Only in responders |
| MRD negativity (e.g., AML, multiple myeloma) | Absence of detectable disease | Earlier readout | Not yet validated for all tumors |
| Patient-reported outcomes (PROs) | Symptoms, QoL | Patient-centric | Subjective, missing data |
Surrogate endpoints (PFS, ORR) speed up approval but their correlation with OS varies by tumor type — a recurring controversy. Sources: [1]
5. Biomarkers and companion diagnostics
A modern oncology drug rarely launches without a companion diagnostic (CDx) — an FDA/EMA-cleared test that identifies the patients most likely to benefit (e.g., HER2 IHC for trastuzumab; PD-L1 IHC for many checkpoint inhibitors; EGFR PCR/NGS for osimertinib).
Key distinctions:
- Predictive biomarker — predicts response to a specific therapy (PD-L1, HER2, EGFR mutation).
- Prognostic biomarker — predicts disease behavior regardless of treatment (Oncotype DX recurrence score, Ki-67).
- Pharmacodynamic biomarker — confirms the drug hit its target (target engagement assays).
For a deeper dive, see Biomarkers & companion diagnostics.
6. The drug development pipeline
Roughly:
Target discovery → Lead identification → Preclinical (cell + animal) →
IND filing → Phase I → Phase II → Phase III → Marketing application
(BLA/NDA in US, MAA in EU, registration with ANVISA in Brazil) →
Post-approval (Phase IV)Median time from first patient dosed in Phase I to FDA approval ≈ 6–8 years for oncology agents in the modern era. Cost estimates vary wildly ($500 M–$2.6 B per approved drug, depending on accounting).
Failure modes (in rough order of frequency):
- Insufficient efficacy (most common in Phase II → III).
- Unacceptable toxicity.
- Trial design / endpoint failures.
- Manufacturing or commercial decisions.
7. Brazilian regulatory context
Anyone running or analyzing trials with Brazilian sites should know:
- ANVISA — federal regulator; approves clinical trial applications and drug registrations.
- CONEP / CEPs / Plataforma Brasil — research ethics review system; all human research must be registered and approved.
- Lei 14.874/2024 — the new federal law on clinical research with humans, modernizing the framework (timeline obligations on regulator/CEPs, protections for participants, post-trial access). Replaces older patchwork.
- Resolução CNS 466/2012 and 510/2016 — foundational ethics regulations (still relevant where the new law refers to them).
- REBEC — Brazilian Clinical Trials Registry; trials with Brazilian sites should be registered (in addition to ClinicalTrials.gov).
- SUS-conducted trials require additional considerations on equity and post-trial access.
For ethics deep dive, see Regulatory & ethics.
8. Where technologists fit
Clinical trials produce some of the densest, best-curated data in medicine — and they need:
- Trial matching algorithms — patient → eligible trial (HemOnc, ClinicalTrials.gov API, NLP on EHRs).
- Real-world data + synthetic controls — augment small trials, especially in rare cancers.
- Digital twins — simulated patient cohorts for sensitivity analysis.
- Smart endpoint capture — wearables, ePROs, image-based response.
- Federated analysis — letting institutions train on each other's data without moving it.
- AI-driven recruitment — identify under-represented populations earlier.
Caveats — see ML pitfalls in oncology. Trial data have lots of footguns: censoring, immortal-time bias, informative dropout, and the gap between intent-to-treat and per-protocol populations.
9. Reading a trial paper — a checklist
When you see a press-released "breakthrough":
- Phase and design — Phase I expansion is not Phase III randomization.
- Population — eligibility criteria can make the trial population unlike real-world patients.
- Comparator — placebo? Standard of care? Investigator's choice? In what country?
- Primary endpoint — was it OS, PFS, or ORR? Pre-specified?
- Effect size and confidence interval — not just the p-value.
- Adverse events — Grade 3+ rate, treatment discontinuations.
- Subgroup analyses — exploratory or pre-specified? Beware fishing.
- Generalizability — would your patient have been eligible?
See also
- Clinical trials 101 — deeper dive into terminology and statistics.
- Biomarkers & companion diagnostics
- Regulatory & ethics
- ML pitfalls in oncology
- Emerging therapies
References
- Sayour EJ, Boczkowski D, Mitchell DA, Nair SK. Cancer mRNA vaccines: clinical advances and future opportunities. Nat Rev Clin Oncol 2024;21:489-500. PMID 38760500. https://doi.org/10.1038/s41571-024-00902-1 (modern review covering trial design innovation in vaccine programs)
- del Carmen MG, Joffe S. Informed consent for medical treatment and research: a review. Oncologist 2005;10:636-641. PMID 16177288. https://doi.org/10.1634/theoncologist.10-8-636
- U.S. National Cancer Institute. Clinical trials information. https://www.cancer.gov/about-cancer/treatment/clinical-trials
- ClinicalTrials.gov. https://clinicaltrials.gov
- ANVISA — Agência Nacional de Vigilância Sanitária. Pesquisa clínica. https://www.gov.br/anvisa/pt-br
- CONEP / Plataforma Brasil. https://plataformabrasil.saude.gov.br
- REBEC — Registro Brasileiro de Ensaios Clínicos. https://ensaiosclinicos.gov.br
- Lei nº 14.874, de 28 de maio de 2024 — pesquisa clínica com seres humanos. https://www.planalto.gov.br/ccivil_03/_ato2023-2026/2024/lei/L14874.htm
- A.C. Camargo Cancer Center. https://accamargo.org.br
- Fundação do Câncer (Brasil). https://www.cancer.org.br/
- Ministério da Saúde / BVS. ABC do câncer. https://bvsms.saude.gov.br/bvs/publicacoes/abc_do_cancer.pdf
- American Cancer Society. Cancer A-Z. https://www.cancer.org/cancer.html
- Cleveland Clinic. Cancer (overview). https://my.clevelandclinic.org/health/diseases/12194-cancer