{"id":"W4412440980","doi":"10.1016/j.esmoop.2025.105509","title":"The use of machine learning models to predict progression-free survival and overall survival outcomes from waterfall plots in randomized clinical trials (MAP-OUTCOMES)","year":2025,"lang":"en","type":"article","venue":"ESMO Open","topic":"Radiomics and Machine Learning in Medical Imaging","field":"Medicine","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Vector Institute; Princess Margaret Cancer Centre; University Health Network","funders":"","keywords":"Waterfall; Clinical endpoint; Confidence interval; Medicine; Randomized controlled trial; Progression-free survival; Logistic regression; Waterfall model; Artificial intelligence; Machine learning; Internal medicine; Statistics; Oncology; Overall survival; Computer science; Mathematics; Software; Cartography","routes":{"ca_aff":true,"ca_fund":false,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":["metaresearch"],"consensus_categories":[],"category_scores_codex":[0.02509921,0.0003127109,0.003248862,0.0001675596,0.00016985,0.0002536331,0.0007654126,0.0001761548,0.00005821129],"category_scores_gemma":[0.04557074,0.0001662405,0.0003737825,0.0001641896,0.0002835272,0.0002341603,0.001366592,0.001043423,0.000003935067],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00004155093,"about_ca_system_score_gemma":0.0001825285,"about_ca_topic_candidate":true,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.01072368,"about_ca_topic_score_gemma":0.0007260067,"domain_scores_codex":[0.9912975,0.004988809,0.002128348,0.0005714217,0.000610141,0.0004037431],"domain_scores_gemma":[0.974453,0.02396819,0.000448577,0.0007544611,0.0001029612,0.0002727599],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.04169929,0.0003154896,0.90945,0.0000512776,0.001054384,0.00004377096,0.0004905806,0.001314718,0.00003247844,0.002215747,0.001925798,0.04140645],"study_design_scores_gemma":[0.3302863,0.0002878642,0.395456,0.001481551,0.0007220971,0.000002405668,0.0001409771,0.246978,0.00003797744,0.006579508,0.01760336,0.0004239539],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9304025,0.0005440198,0.007505089,0.05201969,0.002105172,0.00551974,0.000057397,0.0001013439,0.001745042],"genre_scores_gemma":[0.9794734,0.0008063742,0.01187301,0.002295414,0.0001570113,0.0002231501,0.0001202275,0.00007328225,0.00497814],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.513994,"threshold_uncertainty_score":0.995864,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1472286405677557,"score_gpt":0.4280689458158332,"score_spread":0.2808403052480775,"validation_status":"score_only:v0-immature-baseline","note":"Baseline scores from an immature model (maturity gate not passed). Scores rank; they never assert a category."}}