Thrombospondin-2 is a performant biomarker of at-risk MASH and advanced MASH fibrosis in a large multicentre European cohort

Background Non-invasive biomarkers with biological rationale are needed for identifying patients with progressive metabolic dysfunction-associated steatohepatitis (MASH). Objective To test the diagnostic performance of thrombospondin-2 (TSP2), insulin-like growth factor binding protein-7 (IGFBP7), growth differentiation factor (GDF15) and CD163, which were identified by liver transcriptomics as associated with the severity of MASLD. Design Retrospective study from three European centres in patients with liver biopsy for suspected MASLD. TSP2, IGFBP7, GDF15 and CD163 were measured by ELISAs. Liver stiffness (vibration controlled transient elastography, VCTE), FIB4, Agile3 and FAST scores served as comparators. Outcomes were at-risk MASH (ARM, steatohepatitis NAS>4 and fibrosis stages 2–4) and advanced fibrosis (AF, stages 3–4). Combinatorial scores included fibrosis-biomarkers and clinical/biological variables using multivariate logistic regression controlling for centres, age, body mass index (BMI) and gender. Training and validation sets were defined using a machine learning algorithm with a random split of the total cohort, for 100 iterations, and averaging of area under the receiver operating characteristic curve (AUROC) predictions. Results 469 patients, 61% males, mean age 51.4 years, BMI 30.6 kg/m², 44% with type 2 diabetes, 54% with hypertension, 31% with AF and 46% with ARM were included. The TSP2 AUROC for AF was 0.812 (95% CI: 0.766 to 0.859) higher than for the other biomarkers (ranging from 0.680 to 0.769, p<0.001) and for the comparators (VCTE: 0.739, p<0.04; FIB4: 0.726, p<0.01 and Agile3: 0.755, p=0.20). The TSP2 AUROC for ARM was 0.812 (95% CI: 0.758 to 0.858) higher than for other biomarkers (ranging from 0.722 to 0.748, p<0.001) and comparators (FAST: 0.761, p=0.17). Combining TSP2 with clinical/biological variables (AST, age, platelets, albumin, GGT) increased AUROCs to 0.846 (95% CI: 0.807 to 0.885) for AF and 0.845 (95% CI: 0.806 to 0.880) for ARM and was significantly more performant than all other biomarkers. Conclusions Serum TSP2, alone or with clinical/biological variables, potently discriminates AF and ARM in patients with MASLD.