Synthetic data and big data for cancer research

Overview #

Synthetic data and big data can support cancer research by enabling large-scale analysis and method development.

Big data refers to datasets whose scale, complexity, diversity, or rate of generation exceed what conventional analytical approaches can readily manage. In oncology, these datasets may include electronic health records, cancer registries, pathology reports, genomic and transcriptomic profiles, medical images, wearable-device measurements etc.

Synthetic data are artificially generated data designed to reproduce selected characteristics, statistical relationships, or structures of real-world data. They may be created using statistical models, simulation, or other machine-learning techniques. Depending on the method, synthetic datasets may preserve some properties of the original data or omitting others. Synthetic data are not automatically anonymous, and synthetic data are not automatically scientifically valid.  

Big data in cancer research #

Cancer research increasingly relies on combining data across biological, clinical, and population levels, such as combining genomic data, imaging, clinical and real-world data (such as electronic health records, cancer registries, data from routine clinical practice).

Big data can help researchers investigate rare cancers, identify molecular subgroups, validate models across institutions. However, these benefits depend on data quality, harmonised definitions, appropriate methods for handling bias, and differences between healthcare systems.

Synthetic data in cancer research #

Synthetic data are generated rather than directly collected as observations of individuals, although real data may be used to train or calibrate the generating model.

Common approaches include statistical data, simulation-based data, machine-learning synthesis. Synthetic data may be used for method development and software testing without routinely exposing real patient records, or for AI model development, or for fostering data sharing, or for exploring scenarios about rare cancers and under-represented groups, or for providing shareable datasets for training. Synthetic data should generally be viewed as a complement to real patient data.  

Ethical, legal and social implications (ELSI) #

Privacy, confidentiality, and re-identification #

Cancer datasets can contain highly distinctive combinations of information, including rare tumour types or genomic variants. Removing names and other direct identifiers does not necessarily prevent identification.

Synthetic data may reduce the need to disclose original patient records, but generative models can memorise or reproduce rare observations. Therefore, the risk of identifying the person who contributed to the training data still exists, as well as the risk to reconstruct sensitive information. Synthetic data without privacy protections may remain vulnerable to attacks.

Moreover, synthetic data are not automatically anonymous, and generative models can memorise and leak training records, especially outliers such as rare cancers (Stadler et al., Synthetic Data – Anonymisation Groundhog Day, 2022). Synthetic datasets that preserve rare combinations or are produced without privacy controls may remain personal data, so a case-by-case anonymisation assessment is required and should be documented and repeated over time.

Scientific validity and data utility #

Synthetic datasets must be validated, which means they must be fit for purpose. Similarity to the source dataset alone is insufficient; the evaluation must establish whether the synthetic data support the intended research task.  

Bias, fairness, and representation #

Cancer datasets may under-represent particular populations, including people from minority ethnic groups, people with rare cancers, etc. Synthetic data generated from biased source data can reproduce or amplify these patterns (Obermeyer et al. 2019). Therefore, a proper evaluation across relevant subgroups and comparison against real-world evidence wherever feasible is needed.

Consent, secondary use, and participant expectations #

Cancer research data are frequently reused for purposes beyond the original study; therefore, researchers should assess whether the proposed processing is compatible with the original consent, the applicable legal basis, the approved research purpose, and any relevant governance conditions. They should always explain the creation and use of big data and synthetic data. Social licence and public trust are central (Vayena and Blasimme 2018).

Moreover, participants who withdraw may expect their data to be removed from models and synthetic data derived from them, which is technically difficult.

Return of results #

Anonymous or synthetic data cannot be returned to individuals. Big data analyses may generate individually relevant findings, but re-contact is impossible once data are anonymised, which should be explained in consent materials (see the Knowledge Base entry Return of results in cancer research).

Transparency #

A synthetic dataset should be accompanied by documentation describing its origin, generation method, intended use, limitations. Research outputs should clearly identify synthetic data and disclose when a model was trained or tested on them.

Data quality #

Big data and synthetic data need high data quality. A low quality can undermine the reliability of the resulting analysis. For synthetic data, a strong curation is needed. Harmonisation, metadata standards, data-quality assessment should precede synthesis.

Intellectual property, commercial use, and benefit sharing #

Synthetic datasets and generative models may be developed by academic institutions, hospitals, technology companies, or pharmaceutical sponsors. Governance should address ownership, licensing, access, commercial exploitation, and responsibility for misuse. Moreover, benefits generated from patient data must be shared fairly to maintain trust.

Environmental and computational costs #

Training large generative models and processing big data can require substantial computing resources, energy, storage, and specialist expertise. These costs can create inequalities between well-resourced institutions and smaller research centres. Environmental impact must be taken into account.

Ethics oversight #

Research ethics committees may treat synthetic data work as exempt, but the generation step uses real personal data and downstream use may carry risk. Ethics review capacity for big data and AI is often limited (Ienca et al. 2018).

Security and misuse #

Linkage attacks and genomic membership inference remain concerns even with synthetic or aggregated outputs (Shringarpure and Bustamante 2015).

Legal and ethical international and EU sources #

International Level: #

European Union legal framework: #

  • European Health Data Space (EHDS) — Regulation (EU) 2025/327: it establishes a framework for access to and reuse of electronic health data for specified secondary purposes. It is important for cancer research infrastructures seeking to reuse and link data across institutions and Member States
  • EU Artificial Intelligence Act — Regulation (EU) 2024/1689: it establishes rules for the development, placing on the market, and use of AI systems, such as AI systems used for medical diagnosis, prognosis, treatment planning. In particular, article 50 transparency duties for synthetic content may apply to some generative outputs (such as images and text)

BBMRI-ERIC Resources #

Publications from BBMRI-ERIC ELSI members:

  • Lekadir K, Frangi AF, Porras AR, Glocker B, Cintas C, Langlotz CP, Weicken E, Asselbergs FW, Prior F, Collins GS, Kaissis G, Tsakou G, Buvat I, Kalpathy-Cramer J, Mongan J, Schnabel JA, Kushibar K, Riklund K, Marias K, Amugongo LM, Fromont LA, Maier-Hein L, Cerdá-Alberich L, Martí-Bonmatí L, Cardoso MJ, Bobowicz M, Shabani M, Tsiknakis M, Zuluaga MA, Fritzsche MC, Camacho M, Linguraru MG, Wenzel M, De Bruijne M, Tolsgaard MG, Goisauf M, Cano Abadía M, Papanikolaou N, Lazrak N, Pujol O, Osuala R, Napel S, Colantonio S, Joshi S, Klein S, Aussó S, Rogers WA, Salahuddin Z, Starmans MPA; FUTURE-AI Consortium.. FUTURE-AI: international consensus guideline for trustworthy and deployable artificial intelligence in healthcare. BMJ. 2025. https://doi.org/10.1136/bmj-2024-081554 

External resources #

Relevant literature #

  • Eckardt JN, Hahn W, Prelaj A, Bornhäuser M, Middeke JM, Kather JN. Artificial intelligence-generated synthetic data for cancer research and clinical trials. Nat Rev Cancer. 2026 May;26(5):351-363. doi: 10.1038/s41568-026-00912-4. Epub 2026 Feb 20. PMID: 41720945.
  • Kaabachi B, Despraz J, Meurers T, Otte K, Halilovic M, Kulynych B, Prasser F, Raisaro JL. A scoping review of privacy and utility metrics in medical synthetic data. NPJ Digit Med. 2025 Jan 27;8(1):60. doi: 10.1038/s41746-024-01359-3. PMID: 39870798; PMCID: PMC11772694.
  • Chen RJ, Lu MY, Chen TY, Williamson DFK, Mahmood F. Synthetic data in machine learning for medicine and healthcare. Nat Biomed Eng. 2021 Jun;5(6):493-497. doi: 10.1038/s41551-021-00751-8. PMID: 34131324; PMCID: PMC9353344.
  • Giuffrè M, Shung DL. Harnessing the power of synthetic data in healthcare: innovation, application, and privacy. NPJ Digit Med. 2023 Oct 9;6(1):186. doi: 10.1038/s41746-023-00927-3. PMID: 37813960; PMCID: PMC10562365.
  • Walonoski J, Kramer M, Nichols J, Quina A, Moesel C, Hall D, Duffett C, Dube K, Gallagher T, McLachlan S. Synthea: An approach, method, and software mechanism for generating synthetic patients and the synthetic electronic health care record. J Am Med Inform Assoc. 2018 Mar 1;25(3):230-238. doi: 10.1093/jamia/ocx079. Erratum in: J Am Med Inform Assoc. 2018 Jul 1;25(7):921. doi: 10.1093/jamia/ocx147. PMID: 29025144; PMCID: PMC7651916.
  • Azizi Z, Zheng C, Mosquera L, Pilote L, El Emam K; GOING-FWD Collaborators. Can synthetic data be a proxy for real clinical trial data? A validation study. BMJ Open. 2021 Apr 16;11(4):e043497. doi: 10.1136/bmjopen-2020-043497. PMID: 33863713; PMCID: PMC8055130.
  • Mittelstadt BD, Floridi L. The Ethics of Big Data: Current and Foreseeable Issues in Biomedical Contexts. Sci Eng Ethics. 2016 Apr;22(2):303-41. doi: 10.1007/s11948-015-9652-2. Epub 2015 May 23. PMID: 26002496.
  • Vayena E, Blasimme A. Health Research with Big Data: Time for Systemic Oversight. J Law Med Ethics. 2018 Mar;46(1):119-129. doi: 10.1177/1073110518766026. Epub 2018 Mar 27. PMID: 30034208; PMCID: PMC6052857.
  • Ienca M, Ferretti A, Hurst S, Puhan M, Lovis C, Vayena E. Considerations for ethics review of big data health research: A scoping review. PLoS One. 2018 Oct 11;13(10):e0204937. doi: 10.1371/journal.pone.0204937. PMID: 30308031; PMCID: PMC6181558.
  • Obermeyer Z, Powers B, Vogeli C, Mullainathan S. Dissecting racial bias in an algorithm used to manage the health of populations. Science. 2019 Oct 25;366(6464):447-453. doi: 10.1126/science.aax2342. PMID: 31649194.
  • Shringarpure SS, Bustamante CD. Privacy Risks from Genomic Data-Sharing Beacons. Am J Hum Genet. 2015 Nov 5;97(5):631-46. doi: 10.1016/j.ajhg.2015.09.010. Epub 2015 Oct 29. PMID: 26522470; PMCID: PMC4667107.
  • Wilkinson, M., Dumontier, M., Aalbersberg, I. et al. The FAIR Guiding Principles for scientific data management and stewardship. Sci Data 3, 160018 (2016). https://doi.org/10.1038/sdata.2016.18

Acknowledgements #

The work done stems from research conducted in the context of two projects:

  • project canSERV – Providing Cutting-Edge Cancer Research Services Across Europe, which has received funding from the European Union’s Horizon 2020 research and innovation programme under Grant agreement ID: 101058620. The project started on 1 September 2022 and ended on 31 August 2026, then extended until 28 February 2027. 
  • PERIFORMANCE project (PUBLIC ENGAGEMENT IN RESEARCH INFRASTRUCTURES FOR MISSION CANCER: MANAGING COMPLEXITY OF EMERGING TECHNOLOGIES), Grant agreement ID: 101216808, started on 1 September 2025 and to be finished by 31 August 2028.

Updated on 16 November, 2026
WordPress Cookie Notice by Real Cookie Banner