cs.AIOct 30, 2026

Advancing Clinical Trial Outcome Prediction with Domain-Specific Large Language Models

Authors: Burak Bakkar, Hana Moalla, Mohamed Salem Jedidi, Haifa Ben Messaoud, Nicolas Lopez Carranza

Organizations: InstaDeep

Abstract

Clinical trial failures are a major barrier to efficient drug development, incurring substantial costs and delaying patient access to effective therapies. Structured approaches such as the Hierarchical Interaction Network (HINT) predict trial outcomes by explicitly modelling interactions between drugs, diseases and protocols, but make limited use of the free text in trial registrations. We evaluate two domain-specific language models, BioBERT (108M parameters) and BioMistral (7B parameters), for phase-level outcome prediction on the TOP benchmark (12,477 trials), representing each trial as text built from ten ClinicalTrials.gov fields. We compare fine-tuning (full fine-tuning for BioBERT, low-rank adaptation for BioMistral) with probing, in which the pretrained model is frozen and only a lightweight classifier is trained on its embeddings. On the held-out test set of each phase, all four language-model variants achieved higher ROC-AUC (0.613--0.772) and PR-AUC (0.649--0.901) than the published HINT results, corresponding to relative PR-AUC gains of 10--20%. F1 differences were small and mixed, and no model, including HINT, clearly exceeded the F1 of a trivial classifier that predicts success for every trial. Under the configurations tested, probing performed comparably to fine-tuning: fine-tuned models scored highest on ROC-AUC and PR-AUC in Phases I and II, and frozen BioMistral embeddings in Phase III (ROC-AUC 0.772; PR-AUC 0.901), although most differences between variants were below 0.03 and were not tested for significance. These results suggest that pretrained biomedical representations already encode signal relevant to trial outcome, and that probing is a computationally cheaper alternative to fine-tuning for this task.

Figures & tables

Explore similar work

Apr 24, 2026

DeepImagine: Learning Biomedical Reasoning via Successive Counterfactual Imagining

Predicting the outcomes of prospective clinical trials remains a major challenge for large language models. Prior work has shown that both traditional correlational predictors, such as random forests and logistic regression, and strong commercial LLMs achieve limited performance on this task. In this paper, we propose DeepImagine, a framework for teaching LLMs biomedical reasoning through successive counterfactual imagining. The central idea is to approximate hidden causal mechanisms of clinical trials by training models to infer how observed trial results would change under controlled perturbations of experimental conditions, such as dosage, outcome measures, study arms, geography, and other trial attributes. To support this objective, we construct both natural and approximate counterfactual pairs from real clinical trials with reported outcomes. For settings where strict counterfactual supervision is available, such as paired outcome measures or dose-ranging study arms within the same trial, we train models with supervised fine-tuning. For broader settings where only approximate counterfactual pairs can be retrieved, we optimize models with reinforcement learning using verifiable rewards based on downstream benchmark correctness. We further augment training with synthetic reasoning traces that provide causally plausible explanations for local counterfactual transitions. Using this pipeline, we train language models under 10B parameters, including Qwen3.5-9B, and evaluate them on clinical trial outcome prediction. We aim to show that DeepImagine consistently improves over untuned language models and traditional correlational baselines. Finally, we aim to show that the learned reasoning trajectories provide interpretable signals about how models represent trial-level mechanisms, suggesting a practical path toward more mechanistic and scientifically useful biomedical language models.
Jun 11, 2026

Small LLMs for Biomedical Claim Verification: Cost-Effective Fine-Tuning, Structural Dataset Shortcuts, and Cross-Domain Generalization

Large Language Models such as GPT-4o and GPT-5 achieve strong zero-shot performance on biomedical claim verification, but cost and opacity limit scalable use. We fine-tune three small LLMs: Phi-3-mini (3.8B), Qwen2.5-3B, and Mistral-7B, via QLoRA on SciFact and HealthVer, providing the first study of QLoRA models against GPT-4o and fine-tuned BioLinkBERT encoders. Mistral-7B QLoRA surpasses both GPT-4o and GPT-5 (up to 12% F1 gain) at a fractional cost using just 1,008 training examples. We conduct extensive in-domain and cross-domain evaluation: models trained on SciFact tested on HealthVer and vice versa, at matched sizes to isolate dataset structure from data quantity. We identify a previously unreported structural artifact in SciFact that inflates in-domain scores, and show through bidirectional out-of-domain evaluation that training on structurally sound data enables robust cross-domain transfer. We plan to release all code and adapter checkpoints.
Jun 29, 2026

CaresAI at CT-DEB26: Detecting Dosing Errors In Clinical Trials Using Domain-Specific Transformer Embeddings and Classification Models

Medication errors, particularly dosing errors in clinical trials (CT), can lead to patient harm, adverse drug events and worse patient outcomes. Dosing errors are preventable, and early identification can improve trial integrity and mitigate subsequent clinical and financial burden. This study aims to detect dosing errors within CT protocols by evaluating text representations of trial information using transformer-based language models trained on biomedical corpora. CT textual data was encoded using several models, including ClinicalBERT, PubMedBERT, BioBERT, and MedCPT, and integrated with categorical features. These text embeddings were used as input to classical machine learning models and neural network architectures within an experimental framework. Performance was primarily assessed using ROC-AUC with respect to predicting dosage error. Under a logistic regression baseline, BioBERT consistently outperformed alternative encoders, achieving an ROC-AUC of 0.794, a 3.95% improvement over the ClinicalBERT baseline. Combining multiple embeddings did not yield improvements, indicating that domain alignment outweighs representational stacking. Gradient boosting models, support vector classifiers, logistic regression, and residual neural networks achieved the strongest performance for predicting dosage error, achieving ROC-AUCs: 0.821 to 0.853. Overall, the integration of domain-specific transformer embeddings with structured metadata enables discrimination of trials meeting a predefined elevated dosing error risk criterion, advancing safety monitoring and supporting informed regulatory decision-making.