The interpretation of hepatic venous pressure gradient tracings - excellent interobserver agreement unrelated to experience

Background and aims: The hepatic venous pressure gradient (HVPG) plays an important role in the diagnosis, prognosis and therapy of patients with cirrhosis and portal hypertension. One barrier to its widespread use is the potential for a low reproducibility. We aimed to evaluate the interobserver agreement in the interpretation of optimally acquired HVPG tracings from patients with cirrhosis and different degrees of portal hypertension.

Methods: Two hundred and fifteen tracings obtained from 51 patients with cirrhosis in a single centre were interpreted independently by two hepatologists: one experienced observer and one inexperienced observer. Correlation was performed by Pearson linear regression and the intraclass correlation coefficient (ICC). A Bland-Altman plot was constructed to visualize how the differences between observers compared across the range of measurements. Logistic regression was used to identify predictors of ≥10% variation between observers' readings.

Results: There was a significant linear relationship between observers' readings r = 0.98 (P = 0.001). The ICC between observers (interobserver agreement) was also excellent at 0.991 (P = 0.001). Using the Bland-Altman technique, the mean difference between the observers' readings was 0.2 mmHg (95% CI: -1.2 mmHg to 1.6 mmHg). Thirteen per cent of all readings and 9% of readings with an HVPG of ≥10 mmHg differed by ≥10%. As expected, a lower baseline HVPG was a predictor of this variability.

Conclusions: Interobserver reproducibility in the assessment of optimally acquired HVPG tracings is excellent without differences related to experience. The data provide further support that the HVPG can be used accurately in clinical and research settings.