Liver and HCC Parallel Results Panel

The panel-liver-hcc route is a neutral parallel-results orchestrator. It sends one strict input snapshot to six separately verified engines and preserves each component's own result, applicability, source version, and decision trace.

It does not create a combined liver/HCC score, consensus stage, worst-result category, prognosis, transplant decision, or treatment recommendation. Expected differences between components are model-scope differences, not agreement or discordance findings.

Fixed component order

1. Child-Turcotte-Pugh (standard-bilirubin INR variant in this compact panel)
2. Original ALBI grade
3. OPTN Calculated MELD 3.0, Policy 9.1.D
4. Corrected BCLC 2026 stage
5. Original HKLC-9 (2014) stage
6. Original Milan criteria

Shared inputs without duplicate meanings

The compact form contains 18 clinical inputs plus one scope confirmation, down from the earlier 32-question form. The confirmation covers adult established chronic liver disease, confirmed HCC at initial complete staging, contemporaneous unconfounded liver values, and calculated-MELD context. Clinicians outside that scope should use the standalone calculators.

The panel accepts one tumor count, one largest diameter, one intrahepatic venous-invasion flag, one extrahepatic vascular-invasion flag, and one extrahepatic-spread flag.

• BCLC macrovascular invasion is derived from intrahepatic or extrahepatic vascular invasion.
• HKLC intrahepatic venous invasion uses only the intrahepatic flag.
• HKLC EVM uses extrahepatic vascular invasion or extrahepatic spread.
• Milan evident vascular invasion uses either vascular flag; regional or distant spread uses the spread flag.
• Numeric tumor count maps to Milan as single, 2-3, or 4+ without a second user-entered category.

Partial completion is intentional

Each model keeps its own dependencies. Missing sodium suppresses only MELD. Missing ascites or encephalopathy suppresses Child-Pugh and dependent HKLC, not ALBI, MELD, BCLC, or Milan. Missing tumor-burden data suppresses only the dependent staging results.

Diffuse tumor burden can be interpreted by original HKLC without count or diameter, while BCLC and Milan remain unavailable.

Important model boundaries

BCLC 2026 does not use Child-Pugh as a stage determinant. HKLC-9 does require an interpreted Child-Pugh class. Milan is a source-specific tumor-burden classification and is not an algorithmic transplant-candidacy decision. OPTN Calculated MELD 3.0 stops at the calculated score and does not determine listing, exception points, match priority, or eligibility.

Validation status

The compact public contract and unchanged engine passed 12,292/12,292 deterministic software implementation assertions on 2026-07-30. This is implementation verification, not prospective or independent clinical validation. The complete evidence JSON and concise workbook are linked from the collapsed validation-history card on the calculator route.

Use every component only within its own source population and timing. Clinical interpretation remains the responsibility of the treating team using the full record.