Skip to content

Development and Application of a Guideline-Based Artificial Intelligence Large Language Model for Therapy Recommendations in Interdisciplinary Hemato-Oncological Tumor Boards at Heidelberg University Hospital

Development and Application of a Guideline-Based Artificial Intelligence Large Language Model for Therapy Recommendations in Interdisciplinary Hemato-Oncological Tumor Boards at Heidelberg University Hospital

Status
Active, not recruiting
Phases
Unknown
Study type
Observational
Source
DRKS
Registry ID
DRKS00038205
Enrollment
20
Registered
2026-02-23
Start date
2026-05-04
Completion date
Unknown
Last updated
2026-04-27

For informational purposes only — not medical advice. Sourced from public registries and may not reflect the latest updates. Terms

Conditions

The 10 most common tumor entities (according to the WHO).

Interventions

Group 1: LLM evaluation (HemOncBot): retrospective generation of therapy recommendations, based on current Oncopedia and ESMO guidelines, for tumor board cases of the last two years (from 01/01/2023)

Sponsors

Deutsches Krebsforschungszentrum (DKFZ)
Lead Sponsor

Eligibility

Sex/Gender
All

Inclusion criteria

Inclusion criteria: ESMO and Onkopedia guidelines from 01/2020 (and more recent) are included in HemOncBot. The experts for evaluating the cases are licensed specialists in hematology and oncology who are currently practicing clinically in Germany. A university position or professorship is not required. The prerequisite is regular participation in interdisciplinary tumor boards as well as practical experience in guideline-based treatment decisions for oncology patients.

Exclusion criteria

Exclusion criteria: Tumor board cases with incomplete or non-standardized documented data, lack of pseudonymization, or outside the defined period (January 1, 2023 – November 1, 2025) are excluded. Also excluded are cases with rare or undefined tumor entities for which no current ESMO or Onkopedia guidelines exist (from January 2020 onwards), as well as cases with deliberately guideline-deviating or experimental treatment decisions.

Design outcomes

Primary

MeasureTime frame
Adherence to guidelines, accuracy/correctness of therapy recommendations

Countries

Germany

Contacts

Public ContactTitus J. Brinker

Deutsches Krebsforschungszentrum (DKFZ), Abteilung für Digitale Prävention, Diagnostik und Therapiesteuerung

titus.brinker@dkfz.de+496221425301

Outcome results

None listed

Source: DRKS (via WHO ICTRP) · Data processed: May 1, 2026