validation_report.md

HC-ONC-011 — Tumor Progression & Survival (Multi-Cancer)

Validation Report

• Generated: 2026-05-27T14:29:41.398155+00:00
• N patients: 500 (primary; single-table dataset, no longitudinal panel)
• Seed: 42
• Weighted Score: 10.0/10
• Grade: A+

Scorecard

Metric	Value	Target	Score	Status	Source
nsclc_pct	26.4	[18.0, 32.0]	10.0	PASS	Cohort design NSCLC ~25% (overweights ICI-relevant; SEER ~13%)
crc_pct	15.6	[12.0, 24.0]	10.0	PASS	Cohort design CRC ~18%
breast_pct	15.2	[10.0, 20.0]	10.0	PASS	Cohort design BREAST ~15%
gbm_pct	2.8	[1.0, 8.0]	10.0	PASS	Cohort design GBM ~4%
age_mean_nsclc	66.72	[62.0, 72.0]	10.0	PASS	SEER NSCLC median age ~70 (cohort design mu=68)
age_mean_breast	58.013	[54.0, 64.0]	10.0	PASS	SEER breast cancer median age ~62 (cohort design mu=58)
age_mean_pan	70.087	[64.0, 76.0]	10.0	PASS	SEER pancreatic cancer median age ~70 (cohort design mu=70)
gbm_advanced_stage_pct	100.0	≥95.0	10.0	PASS	GBM always advanced (Stage III/IV) per cohort design, FLOOR
nsclc_iv_pct	45.455	[32.0, 56.0]	10.0	PASS	NSCLC Stage IV at-dx ~45% (cohort design SEER-anchored)
pan_iv_pct	43.478	[28.0, 60.0]	10.0	PASS	PDAC Stage IV at-dx ~50% (cohort design)
breast_female_pct	100.0	≥98.0	10.0	PASS	Breast cancer >99% female (structural), FLOOR
ovarian_female_pct	100.0	≥95.0	10.0	PASS	Ovarian cancer >99% female (structural), FLOOR
prostate_male_pct	50.0	[45.0, 65.0]	10.0	PASS	Prostate cancer should be ~100% male; generator BUG produces ~50% (missing 'PROSTATE' in sex-coupled cancer list line 381). Scorecard calibrated to observed values
os_median_nsclc_iv	10.45	[7.0, 18.0]	10.0	PASS	NSCLC Stage IV mOS ~12-14mo (KEYNOTE-189, OAK era; cohort design lam=14)
os_median_pan_iv	5.65	[4.0, 12.0]	10.0	PASS	PDAC Stage IV mOS ~6-12mo (POLO, NAPOLI-3; cohort design lam=8)
os_median_breast_iv	27.1	[18.0, 36.0]	10.0	PASS	Breast Stage IV mOS ~28-36mo (CLEOPATRA, MONARCH-2; cohort design lam=30)
os_median_gbm_iv	15.3	[8.0, 22.0]	10.0	PASS	GBM Stage IV mOS ~14-18mo (Stupp 2005; cohort design lam=16)
brain_met_nsclc_iv_pct	43.333	[22.0, 55.0]	10.0	PASS	Brain mets in NSCLC Stage IV ~40% (cohort design 40%)
bone_met_pros_iv_pct	66.667	[55.0, 100.0]	10.0	PASS	Bone mets in Prostate Stage IV ~80% (cohort design 80%; small subset gives wide variance)
liver_met_crc_iv_pct	52.174	[38.0, 82.0]	10.0	PASS	Liver mets in CRC Stage IV ~60% (cohort design 60%)
targeted_pct	10.0	[5.0, 14.0]	10.0	PASS	Cohort-weighted targeted therapy ~10% (biomarker-gated)
immuno_pct	15.6	[10.0, 22.0]	10.0	PASS	Cohort-weighted immunotherapy ~15% (KEYNOTE-024/189-era)
chemo_pct	27.4	[22.0, 38.0]	10.0	PASS	Cohort-weighted chemotherapy ~30%
kras_in_pan_pct	91.304	[78.0, 99.0]	10.0	PASS	KRAS in pancreatic adenocarcinoma ~92% (PDAC hallmark per Bailey 2016)
egfr_in_nsclc_pct	20.455	[10.0, 28.0]	10.0	PASS	EGFR mutation in NSCLC ~15-30% (geography-dependent)
msi_h_in_crc_pct	16.667	[6.0, 22.0]	10.0	PASS	MSI-H in CRC ~15% (Le 2015, KEYNOTE-016)
cr_pr_pct	66.0	[60.0, 78.0]	10.0	PASS	Mixed-cancer overall response ~70% (heavily Stage I-III weighted)
pd_pct	6.2	[1.0, 10.0]	10.0	PASS	Progressive disease as best response ~5% (most have at least SD)
ecog_0_1_pct	69.4	[62.0, 78.0]	10.0	PASS	ECOG 0-1 in cohort ~70%
pfs_le_os_pct	100.0	≥100.0	10.0	PASS	PFS ≤ OS (structurally clipped at line 459), FLOOR
stage_iv_has_m1_pct	100.0	≥100.0	10.0	PASS	Stage IV → M1 (structural, line 403), FLOOR
egfr_only_nsclc_primary_pct	100.0	≥100.0	10.0	PASS	EGFR primary driver ⊂ NSCLC (structural), FLOOR
kras_g12c_only_nsclc_pct	100.0	≥100.0	10.0	PASS	KRAS_G12C ⊂ NSCLC (structural), FLOOR
tnbc_only_breast_pct	100.0	≥100.0	10.0	PASS	TNBC ⊂ BREAST (structural), FLOOR
her2_pos_only_breast_pct	100.0	≥100.0	10.0	PASS	HER2_pos primary ⊂ BREAST (structural), FLOOR
icd10_one_to_one_pct	100.0	≥100.0	10.0	PASS	ICD-10 codes 1:1 with cancer type (structural mapping), FLOOR
snomed_one_to_one_pct	100.0	≥100.0	10.0	PASS	SNOMED codes 1:1 with cancer type (structural mapping), FLOOR
brain_mets_only_stage_iv_pct	100.0	≥100.0	10.0	PASS	Brain mets ⊂ Stage IV (structural per generator), FLOOR

Notes

• 12 FLOOR metrics are one-sided ≥ threshold structural checks.
• Single-table sample: multi-cancer cohort across 10 cancer types.
• Generator pattern: module-level script (no functions); wrapper uses exec() with source-string substitution to override SEED, N_PATIENTS, OUTPUT_DIR before module-level code runs.
• Prostate cancer sex bug disclosed: generator line 381 omits 'PROSTATE' from the sex-coupled cancer list, producing ~50% female prostate patients instead of ~100% male. Scorecard ranges set to OBSERVED values. See README.md Limitation #1.