Proof-of-concept: AI-assisted, natural-language-guided survival analysis achieves concordance with human-conducted results in glioblastoma
Bibliographic record
Abstract
Background The use of large language models with natural-language prompting may simplify survival analysis but requires validation against standard analyses carried out by trained analysts. We compared a human-conducted SPSS survival analysis to the Replit cloud development platform for conversational statistical analysis using a fully observed cohort of 1265 glioblastoma patients. Patients and methods Two statistical procedures—Kaplan–Meier estimation and Cox proportional hazards regression—were implemented. The reference analysis was carried out by an experienced clinical researcher using SPSS; the Replit-based analysis by an oncologist with no graduate level statistical or programming training employed the lifelines package through Claude language model integration via conversational chatbox interface. Statistical results were reviewed by a data science professor. Artificial intelligence- (AI) generated code was reviewed by a machine learning engineer. All patients had died at last follow-up, so no censoring occurred. Concordance was defined as exact matches in median survival times and hazard ratios (HR), reflecting the scientific principle that identical datasets processed with identical statistical methods must produce identical results. Results Initial comparison of median survival across 12 molecular/age subgroups found exact concordance in 7 of 12 subgroups (58.3%). The remaining discrepancies represented preprocessing differences rather than acceptable analytical variation. Natural-language troubleshooting through Replit's conversational interface identified three sources of discrepancy: patient inclusion differences, age-group boundary definitions, and inconsistent molecular encoding. After harmonizing these factors, median survival times and HRs were identical across both analyses, achieving 100% concordance. The Replit-based approach required 1 h 40 min total time compared with 8.5 h for traditional analysis, representing a 80% time reduction while maintaining statistical rigor. Conclusions This proof-of-concept demonstrates that the Replit platform can achieve exact replication of standard Kaplan–Meier and Cox analyses carried out by an experienced analyst when subgroup definitions and data preprocessing are aligned. Although our findings are limited to a single dataset and workflow, they suggest conversational AI interfaces could reduce barriers to statistical analysis for clinical researchers. Broader validation across varied analytical scenarios is essential before widespread clinical implementation. Statistical expertise remains essential for data quality assessment and model validation.
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 | 0.013 | 0.040 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.000 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.002 | 0.001 |
| Open science | 0.002 | 0.002 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.014 | 0.005 |
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".