Exploring synthetic controls in rare diseases with a proof of concept in spinal cord injury
Bibliographic record
Abstract
BACKGROUND: Successfully completing clinical trials for rare and heterogeneous disorders, like spinal cord injuries (SCI), remains challenging, thereby reducing the ability to test and translate promising preclinical findings. We propose synthetic controls, derived from data-driven predictions of recovery in patients undergoing standard treatments, to mitigate these challenges, in particular related to patient recruitment. METHODS: Based on data from the European Multicenter Study about Spinal Cord Injury (EMSCI) and the Sygen trial, we construct synthetic controls from personalized predictions of neurological recovery of sequences of segmental motor scores. A total of six architectures (linear, tree, and deep learning models) are compared. We demonstrate the applicability of synthetic controls through a simulation framework modeling the randomization process in a clinical trial and a case study that re-evaluates the recently completed Nogo Inhibition in SCI (NISCI) trial as a single-arm trial post hoc. RESULTS: The primary dataset included 4196 patients from EMSCI and 587 patients from the Sygen trial for external validation. We identified a convolutional neural network as the best-performing architecture to predict segmental motor score sequences, achieving a median root mean squared error below the neurological level of injury of 0.55. Our trial simulations demonstrate that synthetic controls are a viable alternative to randomization, as the proposed solution reduces intercohort heterogeneity and leads to no significant differences with randomized controls in our case study reassessing a clinical trial. CONCLUSIONS: We provide a comprehensive benchmark of data-driven prediction architectures for neurological recovery after SCI. Apart from offering individual patients a specific recovery prediction, these models constitute the basis for synthetic controls. Using real-world data from a completed trial in SCI, we show that synthetic controls could mitigate the challenges of small cohorts and patient recruitment in rare disorders, offering the opportunity to maximize the number of patients receiving an investigative treatment. GITHUB REPOSITORY: https://gitlab.ethz.ch/BMDSlab/publications/sci/sci-in-silico-trials .
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.002 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".