validation_report.md

HC-ONC-005 — Pancreatic Cancer (PDAC)

Validation Report

• Generated: 2026-05-26T21:18:53.369921+00:00
• N patients: 500 (primary) + ~4,400 CA 19-9 panel rows + ~100 surgical rows + 500 molecular rows + ~860 treatment-history rows
• Seed: 42
• Weighted Score: 10.0/10
• Grade: A+

Scorecard

Metric	Value	Target	Score	Status	Source
age_mean	70.704	[66.0, 75.0]	10.0	PASS	SEER PDAC median age 70-71
female_pct	45.8	[38.0, 52.0]	10.0	PASS	SEER PDAC female ~45-48%
diabetes_pct	41.2	[28.0, 48.0]	10.0	PASS	PDAC enriched in diabetes ~30-40% (Aggarwal 2013, Mizrahi 2020)
stage_iv_pct	47.0	[38.0, 58.0]	10.0	PASS	SEER PDAC Stage IV at-dx ~50-55%
stage_i_pct	5.2	[2.0, 9.0]	10.0	PASS	SEER PDAC Stage I at-dx ~5% (most present advanced)
head_tumor_pct	71.4	[60.0, 80.0]	10.0	PASS	PDAC: pancreatic head ~65-75% (Hidalgo 2010)
jaundice_in_head_pct	64.986	[60.0, 80.0]	10.0	PASS	Head PDAC presents with obstructive jaundice ~70-80%
kras_mutation_pct	93.6	≥90.0	10.0	PASS	TCGA PAAD KRAS mutation ~93%, FLOOR (PDAC hallmark)
kras_g12d_pct	30.4	[28.0, 42.0]	10.0	PASS	TCGA PAAD KRAS G12D ~36% (most common variant)
kras_g12v_pct	24.6	[15.0, 28.0]	10.0	PASS	TCGA PAAD KRAS G12V ~22%
kras_variants_exclusive_pct	100.0	≥100.0	10.0	PASS	Structural: KRAS variants categorical (exactly one per patient), FLOOR
tp53_mutation_pct	73.2	[65.0, 82.0]	10.0	PASS	TCGA PAAD TP53 mutation ~70-75%
smad4_lost_pct	48.2	[38.0, 55.0]	10.0	PASS	TCGA PAAD SMAD4 loss ~50% (Iacobuzio-Donahue 2009)
cdkn2a_lost_pct	63.2	[52.0, 68.0]	10.0	PASS	TCGA PAAD CDKN2A loss ~60%
brca2_germline_pct	2.8	[1.5, 8.0]	10.0	PASS	POLO PDAC trial: BRCA2 germline 5% (3-7% real-world)
brca1_germline_pct	2.2	[0.5, 4.0]	10.0	PASS	PDAC BRCA1 germline ~1-2%
parp_eligible_pct	9.2	[6.0, 18.0]	10.0	PASS	PDAC HRR/DDR composite (BRCA1/2, PALB2, ATM) ~10-15%
msi_h_pct	0.8	[0.1, 4.0]	10.0	PASS	PDAC MSI-H rare ~1% (Hu 2018; KEYNOTE-158 enrolled <1%)
surgery_rate_pct	20.6	[12.0, 30.0]	10.0	PASS	PDAC overall surgical resection rate ~15-20% (cohort-driven by Stage IV)
whipple_in_head_surgical_pct	88.571	≥80.0	10.0	PASS	Head PDAC undergoing surgery → Whipple/PPPD ~85-95%, FLOOR
r0_resection_pct	72.816	[60.0, 85.0]	10.0	PASS	Modern PDAC R0 ~65-75% (Strobel 2017)
fistula_bc_pct	25.243	[15.0, 32.0]	10.0	PASS	ISGPF Grade B/C fistula ~20-25%
dge_bc_pct	20.388	[12.0, 32.0]	10.0	PASS	ISGPS DGE Grade B/C ~15-25%
folfirinox_in_palliative_pct	38.631	[25.0, 50.0]	10.0	PASS	PRODIGE-4 era FOLFIRINOX uptake in fit patients ~30-40%
gem_nabp_in_palliative_pct	43.046	[35.0, 55.0]	10.0	PASS	MPACT era Gem+NabP uptake ~40-50%
pembrolizumab_in_msi_h_pct	75.0	≥40.0	10.0	PASS	KEYNOTE-158: Pembro for MSI-H PDAC, FLOOR (high variance at small n)
olaparib_in_parp_eligible_pct	28.261	[15.0, 50.0]	10.0	PASS	POLO era olaparib maintenance in HRR+ PDAC ~25-40% real-world
os_median_overall_mo	5.75	[4.5, 9.0]	10.0	PASS	PDAC overall median OS ~6-11mo; cohort heavily Stage IV-enriched
os_median_stage4_mo	5.25	[3.5, 8.0]	10.0	PASS	PDAC Stage IV median OS ~6-8mo (PRODIGE-4, MPACT)
pfs_median_mo	2.385	[1.5, 4.0]	10.0	PASS	PDAC mPDAC PFS ~3-6mo (regimen-dependent)
orr_pct	22.6	[15.0, 35.0]	10.0	PASS	PDAC palliative ORR ~20-30% (NAPOLI-3 ~42%, MPACT ~23%)
ctdna_in_stage4_pct	89.362	[70.0, 95.0]	10.0	PASS	PDAC ctDNA detection in advanced ~80-90% (Cohen 2018, Bachet 2020)
liver_mets_in_stage4_pct	48.936	[40.0, 65.0]	10.0	PASS	PDAC Stage IV liver mets ~50% (Yachida 2010)
stage_resectability_monotonic_pct	100.0	≥100.0	10.0	PASS	Structural: Stage I→Resectable, Stage IV→Metastatic ordering preserved, FLOOR
ca19_9_panel_median_visits_per_pt	6.0	[4.0, 12.0]	10.0	PASS	CA 19-9 longitudinal: variable visits per patient (depends on OS); median 6-7

Notes

• Floor metrics (kras_mutation_pct, whipple_in_head_surgical_pct, kras_variants_exclusive_pct, pembrolizumab_in_msi_h_pct, stage_resectability_monotonic_pct) are one-sided ≥ threshold checks.
• Five-table output: primary cohort + CA 19-9 longitudinal panel + surgical-cohort subset + molecular panel + treatment history. All tables join on patient_id.
• Patched generator: the original generator had a fatal rng.integers(0, 2**128) overflow at line 60 (numpy max int64). The wrapper applies a minimal in-place patch composing a 128-bit UUID from four 32-bit draws. See README.md for full disclosure.