Abstract
SPLEEN STIFFNESS MEASUREMENT BY A DEDICATED 100Hz VIBRATION-CONTROLLED TRANSIENT ELASTOGRAPHY PROBE IMPROVES THE NON-INVASIVE DIAGNOSIS OF CLINICALLY SIGNIFICANT PORTAL HYPERTENSION IN A PROSPECTIVE MULTICENTER STUDY
Background:
Non-invasive diagnostic tools for clinically significant portal hypertension (CSPH) have been developed in the ANTICIPATE study/subsequent work (liver stiffness measurement (LSM) and platelet count (PLT) ± BMI) and endorsed by the Baveno VII consensus. A dedicated 100Hz probe for spleen stiffness measurement (SSM) substantially increased its clinical applicability, but limited data exists on the diagnostic performance of SSM (100Hz) for CSPH as well as its ability to improve the ANTICIPATE±NASH models.
Methods:
Twelve specialized, high-volume European centers contributed data from prospectively characterized compensated advanced chronic liver disease (cACLD, defined by LSM ≥10kPa or F3/4 fibrosis) patients who underwent paired assessment of the hepatic venous pressure gradient (HVPG), LSM, and SSM from 2021-2023. The goal was to refine the non-invasive diagnosis of CSPH by adding SSM to LSM/PLT±BMI, and to compare the algorithm’s diagnostic performance to the ANTICIPATE±NASH models. In line with previous work, the analysis was restricted to patients with Child-Turcotte-Pugh A stage disease. All area under the receiver operating characteristics curve (AUC) analyses were derived from bootstrapping (n=2000).
Results:
Overall, 244 cACLD patients were included. Most had NAFLD (40.2%) or ALD (36.1%), followed by viral (14.8%) and other (8.9%) etiologies. The median BMI was 28.9 (IQR: 24.8; 33.6) kg/m2, and 43.0% were obese (BMI ≥30). The median HVPG was 11 (8; 15)mmHg (CSPH prevalence: 64.8%), and the median LSM, SSM and PLT were 22.5 (14.8; 33.2)kPa, 45.6 (33.0; 66.5)kPa, and 129 (92; 183)G/L, respectively.
SSM and LSM yielded a comparable AUC for CSPH (SSM: 0.778, LSM: 0.755, DeLong-test: P=0.546). In logistic regression analysis, SSM, LSM, PLT and obesity (BMI ≥30) were independently associated with CSPH. The combined model had an AUC for CSPH diagnosis of 0.864. Notably, its diagnostic performance was superior to the ANTICIPATE±NASH model (AUC: 0.814, DeLong test: P=0.008). A nomogram was developed to facilitate point-of-care risk stratification (Figure). Results were confirmed by cross-validation after randomly splitting the data into a 1:1 train/test cohort matched by CSPH prevalence (data not shown)
Conclusion:
SSM by the novel, dedicated 100Hz module improves the non-invasive diagnosis of CSPH vs. established tools, which tend to be slightly less accurate in the context of contemporary (predominantly NAFLD/ALD) cohorts.