Moving beyond the traditional two-step approach for prognosis prediction: The BayeSarc model.
Bibliographic record
Abstract
11572 Background: Extremity Soft tissue sarcomas (eSTS) are rare and heterogeneous, limiting the collection of large datasets for robust predictive modeling. Sarculator, a Cox model-based tool for overall survival (OS) prediction, was built using the traditional two-step paradigm (1) model building and (2) external validation. However, this method can underperform on external cohorts, often yielding low predictive accuracy and limited generalizability. We introduced a Bayesian Sequential Learning strategy to iteratively refine Sarculator, incorporating new data while preserving prior properties. Methods: The initial model was built on the Italian Sarculator development cohort, with age, tumor size, tumor grade, and histology as covariates. Sequential updates were then performed with the three original Sarculator external validation cohorts , and a more recent Italian cohort. Each step used the results from the previous update as prior information for the next. Performance was assessed as discriminative ability (C-index) and calibration. Key differences from the original Sarculator were Bayesian Cox modelling, and a piecewise-constant hazard. Results: The two-step approach yields separate performance metrics for each cohort, making generalizability unclear when performance drops (e.g. French cohort, Table). Conversely, the sequential approach progressively increases the total information (number of patients and follow-up), without discarding previous evidence, and readjusts performance metrics at each step. Occasional declines in the C-index reflect cohort-specific divergences but can be reversed in subsequent updates if newer cohorts share similar features. Ultimately, the final BayeSarc outperformed the initial model in discriminative ability, calibration, and reduced uncertainty in predictions. Conclusions: BayeSarc is an accurate, generalizable OS prediction model for eSTS, preserving external validation properties while moving beyond the conventional two-step approach. By building on prior evidence, the model dynamically adapts over time, ultimately relying on 4713 patients, with results independent of cohort order. BayeSarc sets a benchmark for future rare-disease prognostic research, paving the way for incorporating new cohorts and/or prognostic variables (e.g. emerging biomarkers). Cohorts Istituto Nazionale Tumori, Milan, Italy1994-2013 Mount Sinai Hospital, Toronto, Canada 1994-2013 Royal Marsden Hospital,London, UK2006-2013 Institut Gustave Roussy, Villejuif, France 1996-2012 Istituto Nazionale Tumori, Milan, Italy2014-2021 Two-step procedure Dev N=1452 Val 1 N=1436 Val 2N=440 Val 3 N=420 Val 4N=965 C-index 0.767 0.775 0.762 0.698 0.765 Bayesian updating Dev N=1452 Upd 1 N=2888 Upd 2 N=3228 Upd 3 N=3748 Upd 4 N=4713 C-index 0.761 0.775 0.771 0.707 0.796
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 machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.012 | 0.020 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.002 | 0.002 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.002 | 0.002 |
| Open science | 0.003 | 0.002 |
| Research integrity | 0.002 | 0.004 |
| Insufficient payload (model declined to judge) | 0.005 | 0.002 |
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 source (direct Gemma or distilled Codex), 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".