International Metastatic Renal Cell Carcinoma Database Consortium (IMDC) classification and regression tree analysis to characterize objective response rates (ORR) in metastatic renal cell carcinoma (mRCC).
Bibliographic record
Abstract
4532 Background: Therapies for mRCC have evolved significantly, making treatment decisions more complex. We used machine learning (ML) to identify whether this could help identify subgroups of patients who have a high probability of response. Methods: Patients from IMDC were identified and a ML classification and regression tree analysis was conducted, in which we grew a complex tree up to a depth of 30 with a minimum node split size of 2 with no constraints on the cost-complexity parameter. The resulting tree was pruned according to the cost-complexity parameter that minimized the leave one out cross-validated error rate and had a minimum bucket size of 25 patients. Results: 2,549 patients were included, 73.2% male, 13.5% non-clear cell histology, 70.3% nephrectomy, and 19.4%, 54.2%, and 26.4% had favorable, intermediate and poor IMDC risk respectively. 1L treatment regimens consisted of VEGF inhibitors (51.5%), IO-IO combinations (32.3%), and IO-TKI combinations (16.2%). The ORR was 36.0% overall, with 29.6% for VEGF inhibitors, 39.1% for IO-IO, and 50.2% for IO-TKI combinations. ML identified 5 hierarchal variables —therapy type, prior nephrectomy (PN), lung metastasis (LM), other metastases, and age— that divided patients into 7 different categories with different response probabilities (see Table). VEGF therapy showed the poorest response, with no additional variables able to predict response. The best ORR was observed in patients treated with IO-TKI and PN; and in those treated with IO-IO, PN, and only lung metastasis. Factors associated with poorer responses included non-clear cell histology, older age, bone and liver metastases, poor performance status, elevated neutrophils, and poor IMDC risk score. Conclusions: This large-scale ML analysis identified five key clinical variables that predict treatment response in mRCC, with treatment type emerging as the primary determinant. These results suggest that treatment selection for mRCC could potentially be optimized by considering these hierarchical variables, though further validation is needed. ML analysis results: Groups of patients and associated outcomes. Risk Groups N (%) ORR (%) Odd Ratio TTNT 18-month survival 1) VEGF 1313 (51.5) 29.6 Ref. 9.4 (8.6-10.3) 0.62 (0.59-0.65) 2) IO-IO or IO-TKI and no PN 443 (17.4) 35.0 1.28 (1.02-1.60) 10.2 (8.8-11.3) 0.59 (0.55-0.65) 3) IO-IO and PN a) No LM 137 (5.4) 29.2 0.98 (0.66-1.43) 17.2 (10.6-30.1) 0.85 (0.78-0.92) b) LM and other met 267 (10.5) 43.8 1.87 (1.42-2.44) 13.0 (10.1-20.5) 0.78 (0.72-0.83) c) Only LM 85 (3.3) 60.0 3.56 (2.28-5.63) 39.2 (14.4-NA) 0.93 (0.87-0.99) 4)IO-TKI and PN a) Age 70+ 78 (3.2) 43.6 1.84 (1.15-2.91) 35.7 (19.8-NA) 0.80 (0.71-0.91) b) Age < 70 226 (8.9) 58.4 3.34 (2.50-4.47) 24.7 (22.4-36.4) 0.88 (0.84-0.93) Overall 2549 36.0 11.5 (10.7-12.2) 0.68 (0.67-0.70)
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 | 0.007 | 0.015 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.000 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.001 | 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 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".