Skip to the calculator
AlmeqdadiMD Calculators
All tools

Calculator · Vienna laboratory model

Portal Hypertension Predictor

Probability of clinically significant and severe portal hypertension from platelets, bilirubin, and INR, with liver stiffness if you have it.

Read me first

Use this as a rule-out tool, not a diagnosis. The Vienna models were built in European cohorts with high portal-hypertension prevalence. In the first US external validation, they systematically overestimated risk — so a high probability usually overstates the true likelihood.

Their strength is exclusion: a below-threshold result — especially for severe PH — is reliable (validation NPV ≈ 0.95). Treat an above-threshold result as a prompt to confirm with HVPG, not as a final answer. This tool never replaces HVPG measurement.

Sources: Reiniš et al., J Hepatol 2023;78:390–400 (model) · Mohammed et al., Dig Dis Sci 2026 (US validation, n = 143).
Clinical read
Enter platelets, bilirubin, and INR to begin.
Active model: 3P (platelets · bilirubin · INR)
Clinically significant PHHVPG ≥ 10 mmHg
Awaiting labs
—probability
0%100%
Predicted probability of clinically significant portal hypertension.
US validation:NPV 0.80PPV 0.53AUROC 0.71
Severe portal hypertensionHVPG ≥ 16 mmHg
Awaiting labs
—probability
0%100%
Predicted probability of severe portal hypertension.
US validation:NPV 0.95PPV 0.31AUROC 0.77
Predicted HVPG (regression estimate) —
▸ Add FIB‑4 + albumin cross‑check (CSPH only)

A second, independent lab model for CSPH from the same validation study. Platelets are reused from above.

Ageyears
Albuming/dL
ASTU/L
ALTU/L
CSPH probability · HVPG ≥10
Fill age, albumin, AST, ALT (and platelets above).
—
—
FIB‑4 = (age × AST) / (platelets × √ALT). Model: logodds = 0.7207 − 0.6729·albumin + 0.4408·FIB‑4. Decision threshold 0.68 (validation Youden; NPV 0.80, PPV 0.63).
▸ Model details, coefficients & references

How the estimate is made

Probabilities come from the published Vienna logistic models. With a liver stiffness value entered, the calculator switches to the LSM‑enhanced coefficient set (a separate model, not the 3P result plus an adjustment).

p = 1 / (1 + e−x)
3P:    x = β₀ + βPLT·PLT + βBILI·bilirubin + βINR·INR
3P+LSM: x = above + βLSM·LSM
Units — PLT: 10⁹/L · bilirubin: mg/dL · INR: unitless · LSM: kPa

Decision thresholds

The gauge marks each model's published Youden‑optimal cut‑off: a reading left of the line favours ruling the condition out; right of the line means it cannot be excluded. In the US cohort the models over‑predicted, so weigh above‑threshold results against that.

Coefficients (Vienna, Table S5)

Term3P ≥103P ≥16+LSM ≥10+LSM ≥16
Intercept−1.3682−2.0152−1.5526−2.1685
Platelets−0.009306−0.010163−0.012479−0.015050
Bilirubin0.3785590.1165900.4845830.227967
INR2.5568661.9462511.3475331.032043
LSM——0.0815470.056162
Threshold0.6630.3320.6480.292

Predicted HVPG uses the Lasso regression terms (3P: 13.0609, −0.030378, 0.667001, 2.627117; +LSM: 10.8949, −0.029723, 0.685411, 0.923888, 0.1433).

References

1. Reiniš J, Petrenko O, Simbrunner B, et al. Assessment of portal hypertension severity using machine learning models in patients with compensated cirrhosis. J Hepatol 2023;78:390–400. Online calculator: liver.at/vlsg/HVPG‑Calculator
2. Mohammed FS, Xing M, Mohnasky M, et al. External validation of laboratory‑based models to assess portal hypertension severity in compensated chronic liver disease. Dig Dis Sci 2026.

Clinical use. This is decision support, not a diagnostic device. It has not been prospectively validated or recalibrated for US/MASLD‑predominant populations. Interpret alongside the full clinical picture and confirm with HVPG where management depends on it.
Copied