Diagnostic performance of quantitative ultrasound techniques for the diagnosis and grading of hepatic steatosis in metabolic dysfunction-associated steatotic liver disease: a systematic review and network meta-analysis

Quantitative ultrasound-derived techniques are core non-invasive tools assessing liver steatosis in metabolic dysfunction-associated steatotic liver disease (MASLD), providing direct fat fraction measurements for screening, grading, and monitoring. This study compares the diagnostic performance of varied quantitative ultrasound-derived approaches relative to magnetic resonance imaging-derived proton density fat fraction (MRI-PDFF) via a network meta-analysis (NMA). PubMed, Embase, the Cochrane Library, and Web of Science were systematically searched until December 7, 2025, for English diagnostic accuracy studies comparing quantitative ultrasound-based techniques with MRI-PDFF for MASLD grading. The primary outcomes extracted were sensitivity (SEN) and specificity (SPE). Study quality was assessed via the QUADAS-2 tool. The primary analysis employed a Bayesian arm-based bivariate random-effects network meta-analysis of diagnostic accuracy implemented in RStan, jointly estimating the pooled SEN, pooled SPE, diagnostic odds ratio (DOR), and superiority index (SI) of each technique within the same model. In addition, exploratory univariate network meta-analyses were performed for SEN and SPE separately using gemtc to calculate the relative sensitivity, relative specificity, and corresponding SUCRA values. Seventeen studies from 10 countries were included, comprising 1,904 MASLD patients and 1,041 controls and covering 17 quantitative ultrasound-derived techniques. The overall risk of bias (RoB) was moderately low. Except for the SEN analysis of S1 steatosis, no significant publication bias was detected across the remaining grading analyses, although evidence of small-study effects was observed. The bivariate random-effects model demonstrated that SWA (0.96), UDFF (0.92), and TAI (0.98) achieved the highest SEN for diagnosing S1, S2, and S3 steatosis, respectively (e.g., S1 (mild):≥S1 vs. S0, S2 (moderate):≥S2 vs. S0-S1, S3 (severe):≥S3 vs. S0-S2). For SPE, RFM + ACE + HI showed superior performance for S1 (0.95) and S2 (0.91), whereas UDFF/Deep-USFF demonstrated the highest SPE for S3 steatosis (0.89). The superiority index, which comprehensively evaluates overall diagnostic performance, showed that RFM + ACE+HI performed remarkably well in S1 (SI = 14.25) and S2 (SI = 15.72), whereas Deep-USFF showed excellent performance in S3 (SI = 14.61). The diagnostic odds ratios indicated that SWA exhibited the highest DOR in S1 (373,719.55, 95% CrI: 8.08–853,672.04) and S2 (234.01, 95% CrI: 1.86–1,077.23), while TAI showed the highest DOR in S3 (2,498,255.36, 95% CrI: 34.43–1,309,055.51). Some estimates were accompanied by extremely wide credible intervals (owing to zero-event studies), indicating uncertainty and warranting cautious interpretation. This NMA indicates that quantitative ultrasound-derived techniques exhibit distinct diagnostic strengths for grading hepatic steatosis in MASLD. SWA, UDFF, and TAI demonstrated superior SEN for detecting mild (S1), moderate (S2), and severe (S3) steatosis, respectively, whereas RFM + ACE + HI and UDFF/Deep-USFF showed greater SPE for their corresponding steatosis grades. This study relied primarily on the composite indicators from the bivariate model as the core basis for evaluation, with some indicators exhibiting certain uncertainties due to the presence of zero-event studies. The SUCRA values calculated based on SEN and SPE separately reflect only the relative ranking of a single diagnostic indicator and cannot independently represent the overall diagnostic performance. Future high-quality, multicenter studies are still needed to validate the clinical utility of different quantitative ultrasound techniques in the screening and staging of MASLD.