Clinical calculator summary
Original 2002 Culine CUP Prognostic Model
Clinical calculator summary
Original 2002 Culine CUP Prognostic Model
A final validated two-factor classification using baseline performance status and serum LDH relative to the same-assay ULN.
Evidence-based context for fast calculator use
- Purpose:
- Reproduce the original Good/Poor risk grouping without inventing a point score, intermediate group, or treatment recommendation.
- Population:
- Adults with confirmed carcinoma of unknown primary, no favorable site-specific subset, and initial pretreatment assessment.
- Factors:
- ECOG/WHO performance status, Raw serum LDH, Same-assay LDH ULN, CUP diagnosis and subset, Pretreatment timing
- Reference:
- Culine et al. J Clin Oncol. 2002. DOI 10.1200/JCO.2002.04.019.
Pin your most used calculators here by clicking the star in the dropdown.
Culine CUP Prognostic Model (2002)
Clinical Context & Background
This calculator implements the final validated two-factor model reported by Culine and colleagues in 2002 for adults with carcinoma of unknown primary (CUP). Poor risk is assigned when baseline ECOG/WHO performance status is 2-3 or raw serum LDH is above the upper limit of normal for the same laboratory assay. Patients with performance status 0-1 and LDH not above the ULN are classified Good risk.
The selected final model has two groups and no point total. Liver metastases appeared in an earlier clinical-only analysis but were no longer independently significant after LDH was added. This page therefore does not implement the separate Hess ten-leaf CART model, a weighted eight-variable score, or the Chinese University Prognostic Index for hepatocellular carcinoma.
The source population comprised adults with confirmed CUP assessed before treatment. Performance status 4, pediatric disease, known or unconfirmed primary context, post-treatment or recurrent-only assessment, and favorable site-specific CUP subsets are outside this implementation. Exactly equal LDH and ULN is normal; only raw LDH above the same-assay ULN is elevated.
Historical median survival and one-year survival values are cohort observations from older trials. They are not individualized predictions and do not select chemotherapy intensity or another treatment.
Formula Logic
Original 2002 Culine final model: Poor risk when baseline ECOG/WHO performance status is 2-3 or raw serum LDH is above the same-assay laboratory upper limit of normal; otherwise Good risk.Reference Data
| Group | Source criteria | Historical development / validation outcomes |
|---|---|---|
| Good risk | PS 0-1 and LDH <= same-assay ULN | Median OS 11.7 / 12 months; 1-year OS 45% / 53% |
| Poor risk | PS 2-3 or LDH > same-assay ULN | Median OS 3.9 / 7 months; 1-year OS 11% / 23% |
Clinical Workflow
Use, Interpret, And Continue The Patient Pathway
Expand for workflow guidance, limitations, examples, and related next steps.
Clinical Workflow
Use, Interpret, And Continue The Patient Pathway
Expand for workflow guidance, limitations, examples, and related next steps.
When To Use
- Use at initial pretreatment assessment for an adult with confirmed carcinoma of unknown primary.
- First identify favorable site-specific CUP subsets and use raw LDH with its same-assay laboratory ULN.
How To Interpret
- Good risk requires PS 0-1 and LDH not above the ULN.
- Poor risk is assigned by PS 2-3 or LDH above the ULN.
- Historical survival values are cohort observations, not individualized forecasts.
What To Do Next
- Complete contemporary CUP diagnostic work-up and favorable-subset classification.
- Integrate pathology, molecular findings, disease distribution, current guidance, patient goals, and multidisciplinary assessment.
Limitations
- Do not apply to PS 4, pediatric disease, post-treatment assessment, known primary cancer, noncarcinoma diagnoses, or a favorable CUP subset.
- Do not substitute liver metastases when LDH is unavailable.
- Do not use the group alone to select chemotherapy or treatment intensity.
Validated Population
Adults with confirmed CUP in the original 150-patient development cohort and 116-patient external validation cohort.
Raw LDH boundary example
PS 1 with LDH exactly equal to the same-assay ULN is Good risk; any raw value above the ULN is Poor risk.
Calculation verification history
Verified
Calculation verification history
Verified: Production browser revalidation
- Three source-backed cases passed through visible production controls across Good and Poor Culine groups.
- The former literal undefined trace defect was absent on every result.
- No unexpected N/A, literal undefined value, stale control, unsupported gauge, or browser-console error was observed.
Case-level evidence is retained in validation/browser-revalidation/cancer-unknown-primary-index-2026-08-03.json.
: Local browser remediation verified; production retest pending
- Ten source-backed cases passed through visible controls across 13 submissions, including two exact-wording reproductions for the intentionally not-applicable states.
- Good risk, Poor risk, raw LDH equality, PS 4, unconfirmed scope, hidden dependencies, and reset/repeat all displayed defined trace values with neutral published-rule decisions.
- The original production UI and hydration findings remain open until deployment and public revalidation.
Case-level local remediation evidence is retained in validation/browser-production/cancer-unknown-primary-index.local-remediation.json.
: Production browser validation
- Ten source-backed cases were exercised on the public production route through visible browser controls; 0/10 case records passed their expected-versus-observed assertions across 11 visible submissions.
- Overall browser status: failed-ui-and-console. high: Every calculated result displayed literal `undefined` beside every visible calculation-trace row, including Good risk, Poor risk, PS-4 Not applicable, scope-not-confirmed, and the repeated result after reset. | medium: Production emitted minified React error #418 during route loads/reloads used by this run.
- Case-level inputs, expected and observed results, reset/repeat behavior, failure reproductions, and console evidence are retained in the production-browser evidence record.
Case-level production evidence is retained in validation/browser-production/cancer-unknown-primary-index.json.
: Browser input adapter corrected and reverified
- Accepted finite number values already normalized by the shared calculator form while retaining strict malformed-input rejection.
- No equation, cutoff, unit, classification, or applicability rule changed; calculator-specific verification and the calculator-wide browser-input audit passed.
: Full validation test
- 256/256 source-rule, raw-boundary, applicability, malformed-input, UI, trace, historical-outcome, report, registry, and isolation checks passed.
- No registered master panel calculates the Culine model; no panel receives inherited verification.
: Calculation and input-contract correction
- Replaced the unrelated eight-variable weighted logic with the final two-factor rule: Poor risk for PS 2-3 or raw LDH above the same-assay ULN; otherwise Good risk.
- Added seven explicit strict inputs and CUP applicability gates, corrected historical outcomes, removed the ambiguous CUPI display, and removed unsupported treatment-intensity guidance.
: Public form simplified
- Replaced four repeated population, diagnosis, timing, and favorable-subset questions with one early applicability confirmation.
- Retained only the three result-determining public inputs: baseline performance status, raw LDH, and the same-assay ULN.
- The strict seven-key engine and LDH equality boundary are unchanged; 256/256 engine checks and the shared 29/29 compact-adapter batch audit passed.
Verification confirms the calculator implementation against the cited model; it is not independent clinical validation, regulatory approval, or medical advice.
Frequently Asked Questions
Is this a numerical CUPI score?
No. The final Culine model has two groups and no point total. The page avoids the ambiguous CUPI label, which is also used for a different hepatocellular carcinoma index.
What happens when LDH exactly equals the ULN?
It is not elevated. Only a raw LDH value above the same-assay laboratory ULN is an adverse factor.
Does Poor risk determine treatment?
No. The model describes historical prognosis in a selected CUP population and does not select chemotherapy intensity or another treatment.
Evidence-based oncology decision support. Verify with clinical guidelines.