Comparing FIB-4, VCTE, pSWE, 2D-SWE, and MRE Thresholds and Diagnostic Accuracies for Detecting Hepatic Fibrosis in Patients with MASLD: A Systematic Review and Meta-Analysis

Abstract

Objectives: To compare thresholds and accuracies of FIB-4, vibration-controlled transient elastography (VCTE), point shear wave elastography (pSWE), 2D shear wave elastography (2D-SWE), and MR elastography (MRE) for detecting hepatic fibrosis in patients with MASLD. Materials and Methods: Systematic searching of MEDLINE, EMBASE, Cochrane Library, Scopus, and the gray literature from inception to March 2024 was performed. Studies evaluating accuracies of FIB-4, VCTE, 2D-SWE, pSWE, and/or MRE for detecting significant (≥F2) and/or advanced (≥F3) hepatic fibrosis in MASLD patients compared to histology were identified. Full-text review and data extraction were performed independently by two reviewers. Multivariate meta-analysis and subgroup analyses were performed using index test and fibrosis grading. Risk of bias was assessed using QUADAS-2. Results: 207 studies with over 80,000 patient investigations were included. FIB-4 1.3 threshold sensitivity was 71% (95% CI 66–75%) for detecting advanced hepatic fibrosis, which improved to 88% (85–91%) using a <0.75 threshold. FIB-4 specificity using a 2.67 threshold was 96% (94–97%). Sensitivities of 88–91% were achieved using thresholds of 3.2 kPa for pSWE, 4.92 kPa for 2D-SWE, 7.18 kPa for VCTE, and 2.32 kPa for MRE. No significant differences were identified for sensitivities in subgroup analysis with thresholds between 7 and 9 kPa. Most imaging-based studies were high risk of bias for the index test. Conclusions: A FIB-4 threshold of <0.75 and modality-dependent thresholds (VCTE < 7 kPa; pSWE <3 kPa; 2D-SWE <5 kPa; and MRE <2.5 kPa) would achieve sensitivities of around 90% when defining low-risk MASLD in population screening. A modified two-tier algorithm aligning with existing Society of Radiologists in Ultrasound guidelines would improve risk stratification accuracies compared to existing guidelines by European and American liver societies.

1. Introduction

Metabolic dysfunction-associated steatotic liver disease (MASLD), formerly non-alcoholic fatty liver disease (NAFLD), has become the most common cause of chronic liver disease (CLD) in the Western world with a rising incidence [1]. MASLD consists of a spectrum of steatotic liver diseases ranging from isolated liver steatosis (metabolic dysfunction-associated steatotic liver [MASL]), hepatocellular ballooning and lobular inflammation (metabolic dysfunction-associated steatohepatitis [MASH]), and hepatic fibrosis ultimately to cirrhosis. Hepatic fibrosis and cirrhosis are associated with liver failure, portal hypertension and other complications including hepatocellular carcinoma, ultimately resulting in approximately 2 million deaths annually worldwide [2]. Early detection and risk stratification of patients on the MASLD spectrum can help identify patients who may benefit from lifestyle and medical intervention to prevent and even reverse early stages of CLD. The traditional gold standard, liver biopsy, is subject to several limitations restricting widespread use at the population level. As such, several alternative non-invasive approaches to detect significant and advanced liver fibrosis have been developed, including various combinations of clinical and laboratory investigations (the most popular calculation now used is termed Fibrosis-4 [FIB-4]), as well as several imaging-based tools such as vibration-controlled transient elastography (VCTE), point shear wave elastography (pSWE), 2D shear wave elastography (2D-SWE), and MR elastography (MRE). These elastography technologies utilize different physical properties including variations of vibration-controlled technologies (TE and MRE) and acoustic radiation force impulses (pSWE and 2D-SWE).

Several guidelines have recently been published offering recommendations for population-level risk stratification using non-invasive techniques [1,2,3,4]. These guidelines have included a range of recommendations from a single vendor agnostic image-based multi-threshold risk stratification tool independent of underlying pathophysiology [4] to variations of a two-step blood-based followed by imaging-based vendor and modality agnostic approach, to risk stratification thresholds [1,3] and disease severity grading [2]. In guidelines using multi-modality image-based risk stratification tools, only one differentiated thresholds for MRE from other liver stiffness measurement (LSM) techniques [2] but did not differentiate between VCTE, pSWE, nor 2D-SWE. Given that each technique differs fundamentally in underlying technology, a nuanced evaluation of accuracy is needed, comparing the accuracy and threshold grading of these modalities to inform current and future revisions of these population-level risk stratification guidelines. This meta-analysis aims to evaluate the individual diagnostic accuracies of FIB-4, VCTE, pSWE, 2D-SWE, and MRE for detecting significant (METAVIR ≥ F2) and advanced (METAVIR ≥ F3) hepatic fibrosis in patients with MASLD at differing thresholds. Subgroup analyses are performed to determine the accuracies of frequently recommended modality-specific thresholds.

2. Methods

This systematic review and meta-analysis were reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analysis–Diagnostic Test Accuracy (PRISMA-DTA) guidelines (Supplementary Table S1) [5]. Prior to initiation, a study proposal was submitted to the PROSPERO database (CRD42024528716). Institutional ethics approval and patient consent were not required as this review included pooled analysis of previously published studies.

2.1. Literature Search

A systematic search of multiple databases was performed to identify studies evaluating the diagnostic accuracy of one or more of (1) FIB-4, (2) VCTE, (3) pSWE, (4) 2D-SWE, or (5) MRE for detecting significant and/or advanced hepatic fibrosis in patients with MASLD. Individualized search protocols for MEDLINE, EMBASE, Cochrane Library (systematic reviews and registry of controlled trials), and Scopus from the date of inception up to March 25, 2024 were developed by a reviewer with 10 years of imaging experience and expertise in the performance of diagnostic test accuracy systematic reviews (M.P.W.). Several combinations of title/abstract/keywords and medical subject headings pertaining to information related to patient population (NAFLD/MASLD), index test(s) (FIB-4, VCTE, pSWE, 2D-SWE, and MRE) and diagnostic accuracy were customized by database. Individualized database search criteria are shown in Appendix A. No language restriction was applied. The electronic search was conducted according to best practices [6]. Studies from individual databases were then collated, and duplicates were removed. A title and abstract review was performed by a separate reviewer with 10 years of imaging experience and experience in diagnostic test accuracy studies (R.S.). A full-text review was then conducted separately in a blinded fashion by this reviewer and a medical student with prior experience in scoping reviews (R.S., S.M.), with a gray literature search performed in tandem by S.M. evaluating the most recent three years of meetings by the Radiological Society of North America (RSNA), the American Roentgen Ray Society (ARRS), and the American College of Gastroenterology (ACG). References for relevant articles were manually evaluated by reviewers, and forward searching of relevant included articles was also performed on Google Scholar.

2.2. Selection Criteria

Studies were identified and ultimately selected for inclusion when evaluating one or more index test diagnostic accuracies using the following criteria: (1) MASLD patient population; (2) one or more of FIB-4, VCTE, pSWE, 2D-SWE, and/or MRE used as the index test; (3) histopathology used as the reference standard; (4) studies evaluating a minimum of 5 patients; and (5) sufficient information to construct a 2 × 2 contingency table (true positive [TP], false positive [FP], false negative [FN], and true negative [TN]). Histopathology reference standards were referenced to METAVIR 0–1 versus 2–4 (“significant hepatic fibrosis”) and/or METAVIR 0–2 versus 3–4 (“advanced hepatic fibrosis”). When studies used alternative histopathology staging systems for hepatic fibrosis (such as Ishak score or Batts–Ludwig system), studies were correlated to METAVIR grading using previously proposed standardized correspondence between systems [7]. Studies were excluded from review if (1) MASLD was not the selected patient population and/or the performance of index tests in only MASLD patients could not be extracted from pooled data with other patient populations; (2) an index test other than the above was used; (3) patients without a histopathological reference standard were included; and (4) they were non-original articles, including review articles, guidelines, consensus statements, letters, and editorials.

2.3. Data Extraction

Two reviewers (R.S., S.M.) independently extracted data. Study, patient, and index test characteristics were recorded, including author, year, country of institution, study design (prospective or retrospective), number of centers involved in patient recruitment (single- or multicenter), reference standard, total number of patients, mean age and range, percentage of male sex, total number of cases of METAVIR ≥ 2 and/or ≥ 3, index test(s) used, thresholds by index test and fibrosis severity, reader agreement, and pertinent notes by study. For studies evaluating one or more of pSWE, 2D-SWE, and/or VCTE, details related to the index test including vendor used, probe frequency (when applicable), number of acquisitions of elastography, and technician (person performing the examination) experience were extracted. For studies evaluating MRE as an index test, details including vendor used, reader experience, MRI field strength (1.5 T or 3 T), pulse sequence used (gradient echo [GRE] versus spin echo echoplanar imaging [SE-EPI]), passive driver amplitude, and segmentation technique used (manual/freehand or automated) were extracted. Contingency tables were developed for each individual index test and fibrosis severity, and corresponding thresholds were recorded. Thresholds were reported as Young’s modulus (kPa) for imaging-based modalities. For studies reporting elasticity with shear wave velocities (m/s), a conversion calculation of E = 3pc² was used, where E = Young’s modulus, p = density [1000 kg/m³], and c = shear wave velocity [8]. When true positive, false positive, false negative, and true negative cases were not reported directly, a contingency table was constructed using the total number of patients included in that index test, the total number of positive cases (based on fibrosis severity), and sensitivity and specificity. If contingency tables were provided by multiple readers within a single index test and fibrosis severity, results were averaged as a single result based on the a priori design [9]. After independent extraction, the data were combined into a single file where discrepancies were resolved by re-review and consensus between the two data extractors and a third author (R.S., S.M., and M.P.W.).

2.4. Risk of Bias Assessment

Risk of bias and applicability concerns were evaluated on a per study basis using the Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2) tool [10]. The patient selection, reference standard, and flow and timing were assessed using standardized signaling questions outlined in QUADAS-2. Any single signaling question marked as high risk would result in a high-risk grading for a given domain.

2.5. Data Analysis

A bivariate mixed-effects regression model was initially planned in an a priori protocol with stratification by index test, fibrosis severity, and threshold used. However, after extraction of the data and recognition of inconsistent threshold performance variably reported across a continuous spectrum by study and index test, a multilevel random effects model was instead performed with stratification by index test and fibrosis severity (METAVIR 0–1 versus 2–4 and/or 0–2 versus 3–4) for the primary analysis. This model links the range of thresholds and retrospective pairs of sensitivity and specificity to identify thresholds at which the test is likely to perform optimally. The model assumes a logistic distribution and estimates the distribution parameters in patients without significant and/or advanced hepatic fibrosis, applying a linear mixed-effects model to the transformed data. The model accounts for between-study variability and dependence of sensitivity and specificity. Three separate sensitivity weightings were applied (sensitivity to specificity weights of 0.3:0.7, 0.5:0.5, and 0.7:0.2). These weightings were selected to recognize likely “rule in”, “balanced”, and “rule out” thresholds, respectively. Continuity correction using a value of one was applied to any cells in a 2 × 2 contingency table where a zero value was encountered. Summary sensitivities and specificities with a 95% confidence interval (CI) were estimated for each threshold. A separate multilevel random effects model was applied for FIB-4 studies evaluating thresholds of 1.3 and 2.67 for detecting both significant and advanced hepatic fibrosis based on current guideline recommendations. Finally, a vendor-specific multilevel random effects model was applied for pSWE, 2D-SWE, and VCTE studies in the same fashion.

Statistical measures of study variability and publication bias were not assessed as these are proven to be unreliable and no longer recommended in the PRISMA-DTA checklist [5]. Variability across studies was assumed, and exploration for potential causes of variability between image-based investigations using a priori subgroup analyses was planned. These included a number of study characteristics (year of publication, location, study design, reference standard, and duration between index test and reference standard), patient characteristics (percent male sex, mean age, total number of included patients, and total number of patients stratified by METAVIR grading), and imaging characteristics (VCTE/pSWE/2D-SWE: vendor used, technician experience, number of acquisitions, and threshold used; MRE: vendor used, MRI field strength, passive driver amplitude, segmentation method, and threshold used). Based on available data, completed subgroup analyses included location (east Asian country versus other countries), study design (prospective versus retrospective), number of centers (single-center versus multicenter), risk of bias (low risk of bias versus single domain with unclear or high risk of bias), vendor used, number of measurements for pSWE (<10 versus ≥ 10), MRI field strength (1.5 T versus 3 T), and MRI pulse sequence used (GRE versus SE-EPI). Analysis was performed using STATA 16 software (StataCorp, 2019, Stata Statistical Software: Release 15, College Station, TX, USA: StataCorp LP) and R software (R Core Team (2021, R: A language and environment for statistical computing; R Foundation for Statistical Computing, Vienna, Austria).

3. Results

The literature search PRISMA flow diagram is shown in Figure 1. Of 3806 articles reviewed, a total of 207 were included, with 121 articles evaluating FIB-4 (57,100 patients) [11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100,101,102,103,104,105,106,107,108,109,110,111,112,113,114,115,116,117,118,119,120,121,122,123,124,125,126,127,128,129,130,131], 13 articles evaluating pSWE (1033 patients) [14,132,133,134,135,136,137,138,139,140,141,142,143], 24 articles evaluating 2D-SWE (2592 patients) [11,13,116,140,144,145,146,147,148,149,150,151,152,153,154,155,156,157,158,159,160,161,162,163], 93 articles evaluating VCTE (24,628 patients) [12,15,19,21,22,23,24,26,27,28,29,30,31,32,33,34,35,36,37,39,40,41,42,43,44,45,46,47,48,49,67,116,133,134,136,137,138,140,142,146,150,151,152,154,159,160,162,163,164,165,166,167,168,169,170,171,172,173,174,175,176,177,178,179,180,181,182,183,184,185,186,187,188,189,190,191,192,193,194,195,196,197,198,199,200,201,202,203,204,205,206,207], and 24 articles evaluating MRE (3420 patients) [15,16,17,18,19,20,116,139,145,146,152,172,174,175,198,208,209,210,211,212,213,214,215,216]. Some articles included the evaluation of more than one modality.

3.1. Study, Patient, and Imaging Characteristics

Study and patient details for FIB-4 are shown in Appendix B. Included studies were performed across numerous countries using a mix of prospective and retrospective study designs. All studies used liver biopsy as the reference standard.

Study, patient, and imaging characteristics for studies evaluating pSWE, 2D-SWE, and VCTE are shown in Appendix C. Studies evaluating 2D-SWE used a mix of vendors including Canon, General Electric (GE), Siemens, SuperSonic Imagine, Toshiba, and Ultrasign. Studies evaluating pSWE used a mix of Siemens, Phillips, and Samsung. All VCTE studies used Echosens (FibroScan). Studies for each modality were performed internationally with a mix of prospective and retrospective study designs. Readers demonstrated varying degrees of experience with the technologies (reported as number of exams and/or years of experience). All studies used liver biopsy as the reference standard.

Study, patient, and imaging characteristics for studies evaluating MRE are shown in Appendix D. Most studies were performed in North America, with east Asian countries the next most common. A mix of magnet strengths (1.5 Tesla or 3 Tesla) and pulse sequences (2D gradient echo [2D-GRE] or spin echo echoplanar [SE-EPI] sequences) were used. Most studies used a 60 Hz passive driver and a manual region of interest. All studies used liver biopsy as the reference standard.

3.2. Diagnostic Accuracy by Modality

Pooled and weighted sensitivity and specificity values reported by threshold for each modality for detecting significant (METAVIR 0–1 versus 2–4) and advanced (METAVIR 0–2 versus 3–4) hepatic fibrosis are shown in

Table 1. Pooled and weighted sensitivity and specificity values reported by threshold for FIB-4, pSWE, 2D-SWE, VCTE, and MRE for detecting significant and advanced hepatic fibrosis in MASLD patients. Weighted sensitivities to specificities included 0.3:0.7, 0.5:0.5, and 0.7:0.3.

		Sensitivity Weight 0.3			Sensitivity Weight 0.5			Sensitivity Weight 0.7
Modality	Fibrosis Severity	Threshold	Sensitivity [%, 95% CI]	Specificity [%, 95% CI]	Threshold	Sensitivity [%, 95% CI]	Specificity [%, 95% CI]	Threshold	Sensitivity [%, 95% CI]	Specificity [%, 95% CI]
FIB-4	Significant	2.34	36 [27–47]	89 [84–93]	1.06	68 [61–75]	68 [61–75]	0.48	89 [84–93]	36 [27–47]
	Advanced	2.39	42 [35–50]	88 [85–91]	1.33	70 [63–76]	70 [65–75]	0.74	88 [85–91]	42 [35–50]
pSWE	Significant	35.90	52 [10–92]	88 [42–99]	5.37	74 [60–84]	74 [59–84]	0.80	88 [42–99]	52 [10–92]
	Advanced	12.97	61 [33–83]	88 [74–95]	6.43	78 [60–89]	78 [65–87]	3.20	88 [74–95]	61 [33–83]
2D-SWE	Significant	13.59	54 [35–72]	88 [78–94]	8.31	74 [60–85]	74 [62–84]	5.10	88 [78–94]	54 [35–72]
	Advanced	18.63	60 [38–79]	88 [79–94]	9.58	77 [61–88]	77 [67–85]	4.92	88 [79–94]	60 [38–79]
VCTE	Significant	11.88	48 [37–60]	88 [82–92]	8.48	72 [62–80]	72 [63–79]	6.10	88 [82–92]	48 [37–60]
	Advanced	13.21	56 [48–64]	88 [85–91]	9.74	75 [69–81]	75 [70–80]	7.18	88 [85–91]	56 [48–64]
MRE	Significant	3.70	69 [50–83]	90 [80–95]	2.93	82 [67–90]	82 [68–90]	2.32	90 [80–95]	69 [50–83]
	Advanced	5.98	75 [61–85]	91 [84–95]	2.73	85 [75–91]	85 [76–90]	1.24	91 [84–95]	75 [61–85]

Significant hepatic fibrosis = METAVIR 0–1 versus METAVIR 2–4; advanced hepatic fibrosis = METAVIR 0–2 versus METAVIR 3–4; FIB-4 = Fibrosis 4; pSWE = point shear wave elastography; 2D-SWE = 2D shear wave elastography; VCTE = vibration-controlled elastography; and MRE = magnetic resonance elastography.

. Specific evaluation of studies assessing FIB-4 thresholds of 1.3 and 2.67 are shown in

Table 2. Pooled sensitivity and specificity values for FIB-4 thresholds of 1.3 and 2.67 for detecting significant and advanced hepatic fibrosis in MASLD patients.

Hepatic Fibrosis Severity	FIB–4 Threshold	No. Studies	No. Patients	Sensitivity [% (95% CI)]	Specificity [% (95% CI)]
Significant	1.3	13	7975	65 (55–73)	74 (66–81)
Advanced	1.3	44	21,409	72 (68–76)	71 (66–75)
Significant	2.67	7	3741	25 (18–33)	97 (94–99)
Advanced	2.67	50	29,953	32 (27–38)	96 (94–97)

Significant hepatic fibrosis = METAVIR 0–1 versus METAVIR 2–4; advanced hepatic fibrosis = METAVIR 0–2 versus METAVIR 3–4.

. Sensitivities are 65% (95% CI 55–73%) and 72% (68–76%) for detecting significant and advanced hepatic fibrosis at a FIB-4 threshold of 1.3. Sensitivities are increased to 89% (84–93%) and 88% (85–91%) for excluding significant and advanced hepatic fibrosis at FIB-4 thresholds of 0.48 and 0.74, respectively. Specificities are 97% (94–99%) and 96% (94–97%) for detecting significant and advanced hepatic fibrosis at a FIB-4 threshold of 2.67 with sensitivities of 25% (18–33%) and 32% (27–38%), respectively. A lowered specificity of 88% (85–91%) for detecting advanced hepatic fibrosis is achieved with a threshold of ≥2.4 with a corresponding increased sensitivity of approximately 10% (42% [35–50%]).

Studies evaluating pSWE showed lower threshold values for achieving similar accuracies compared to 2D-SWE and VCTE. Thresholds with sensitivities nearing 90% for detecting significant and advanced hepatic fibrosis are 0.80 (88% [42–99%]) and 3.20 (88% [74–95%]), respectively, for pSWE, 5.10 (88% [78–94%]) and 4.92 (88% [70–94%]) for 2D-SWE, and 6.10 (88% [92–92%]) and 7.18 (88% [85–91%]) for VCTE. Sensitivities of 90% (80–95%) and 91% (84–95%) for detecting significant and advanced hepatic fibrosis are achieved with MRE thresholds of 2.32 and 1.24, respectively.

Evaluations of vendor-specific thresholds and accuracies are shown in

Table 3. Vendor-specific pooled and weighted sensitivity and specificity values reported by threshold for pSWE, 2D-SWE, and MRE for detecting significant and advanced hepatic fibrosis in MASLD patients. Weighted sensitivities to specificities included 0.3:0.7, 0.5:0.5, and 0.7:0.3.

			No. Studies	No. Contigency Tables	Sensitivity Weight 0.3			Sensitivity Weight 0.5			Sensitivity Weight 0.7
Modality	Fibrosis Severity	Vendor			Threshold	Sensitivity [%, 95% CI]	Specificity [%, 95% CI]	Threshold	Sensitivity [%, 95% CI]	Specificity [%, 95% CI]	Threshold	Sensitivity [%, 95% CI]	Specificity [%, 95% CI]
2D-SWE	Significant	Canon	4	6	NA	NA	NA	NA	NA	NA	NA	NA	NA
		GE	7	11	14.5	49 (12–87)	91 (0–100)	9	73 (47–89)	73 (51–87)	5.6	88 (72–95)	49 (30–68)
		Siemens	2	4	NA	NA	NA	NA	NA	NA	NA	NA	NA
		Supersonic Imagine	8	10	13.6	52 (23–80)	88 (80–96)	8.2	74 (55–87)	74 (57–86)	5	88 (63–97)	52 (23–80)
	Advanced	Canon	4	5	NA	NA	NA	NA	NA	NA	NA	NA	NA
		GE	7	10	12.3	59 (28–84)	88 (70–96)	8.7	77 (54–90)	77 (61–87)	6.1	88 (71–96)	59 (37–77)
		Siemens	2	4	NA	NA	NA	NA	NA	NA	NA	NA	NA
		Supersonic Imagine	9	12	35	58 (10–95)	88 (36–99)	10.8	76 (58–88)	76 (64–85)	3.3	88 (39–99)	58 (12–93)
pSWE	Significant	Siemens	5	8	15.1	54 (61–96)	88 (29–99)	4.4	75 (62–84)	75 (61–84)	1.3	88 (37–99)	54 (8–94)
		Phillips	5	5	NA	NA	NA	NA	NA	NA	NA	NA	NA
		Samsung	1	3	NA	NA	NA	NA	NA	NA	NA	NA	NA
	Advanced	Siemens	7	10	8.8	67 (41–86)	89 (77–96)	5.9	81 (64–91)	81 (69–89)	4	89 (78–95)	67 (52–79)
		Phillips	5	5	NA	NA	NA	NA	NA	NA	NA	NA	NA
		Samsung	1	3	NA	NA	NA	NA	NA	NA	NA	NA	NA
MRE	Significant	GE	13	19	3.9	70 (52–83)	80 (81–95)	3	82 (67–91)	82 (69–90)	2.4	90 (72–97)	70 (41–88)
		Siemens	2	2	NA	NA	NA	NA	NA	NA	NA	NA	NA
	Advanced	GE	10	14	5.9	75 (57–88)	91 (83–96)	2.6	85 (73–92)	85 (76–91)	1.1	91 (72–98)	75 (45–92)
		Siemens	1	1	NA	NA	NA	NA	NA	NA	NA	NA	NA
		Phillips	1	1	NA	NA	NA	NA	NA	NA	NA	NA	NA

Significant hepatic fibrosis = METAVIR 0–1 versus METAVIR 2–4; advanced hepatic fibrosis = METAVIR 0–2 versus METAVIR 3–4; No. studies = number of studies; No. contingency tables = number of contingency tables (which may include >1 table per study if multiple thresholds are reported); NA = not analyzable.

. Accuracies of most vendor-specific thresholds for pSWE, 2D-SWE, and MRE could not be performed due to the limited number of individual studies and/or contingency tables available by fibrosis severity. More than one vendor comparison was only available for 2D-SWE (GE and Supersonic Imagine), with GE demonstrating higher thresholds for similar accuracies. For example, sensitivities of 88% (81–96) and 88% (39–99%) for detecting advanced hepatic fibrosis were achieved with thresholds of 6.1 for GE and 3.3 for Supersonic Imagine, respectively.

3.3. Subgroup Analysis

Subgroup analyses evaluating reported thresholds of 7–9 kPa (8 ± 1 kPa) for pSWE, 2D-SWE, VCTE, and MRE are shown in

Table 4. Pooled sensitivity and specificity subgroup analyses evaluating reported thresholds of 7–9 kPa (8 ± 1 kPa) for pSWE, 2D-SWE, VCTE, and MRE for detecting significant and advanced hepatic fibrosis in MASLD patients.

Modality	Fibrosis Severity	Subgroup	No. Studies	No. Patients	Sensitivity, % (95% CI)	Specificity, % (95% CI)	Significance
pSWE	Significant	East Asian Countries	0	0	N/A	N/A	N/A
		Other Countries	3	232	78 (70–84)	80 (71–87)
		Vendor: Siemens	1	27	N/A	N/A	N/A
		Vendor: Philips	1	159	N/A	N/A
		Vendor: Samsung	1	46	N/A	N/A
		<10 measurements	2	92	72 (59–83)	68 (52–81)	NS
		≥10 measurements	3	232	77 (69–93)	80 (71–87)
		RoB: Low	3	138	73 (62–81)	70 (57–81)	NS
		RoB: Unclear and/or High	2	186	78 (69–85)	82 (72–89)
	Advanced	East Asian Countries	2	117	75 (53–88)	62 (40–80)	NS
		Other Countries	5	518	72 (62–81)	82 (74–89)
		Vendor: Siemens	2	263	62 (51–71)	86 (67–95)	NS
		Vendor: Philips	4	326	80 (70–87)	77 (61–88)
		Vendor: Samsung	1	46	67 (41–87)	71 (51–87)
		<10 measurements	2	146	71 (52–85)	57 (41–72)	p < 0.05
		≥10 measurements	4	472	73 (60–83)	84 (76–89)
		RoB: Low	3	299	73 (62–81)	70 (57–81)	NS
		RoB: Unclear and/or High	4	336	78 (69–85)	82 (72–89)
2D–SWE	Significant	East Asian Countries	9	1005	84 (72–91)	85 (77–91)	NS
		Other Countries	6	649	83 (74–89)	83 (75–88)
		Prospective Study Design	12	1360	84 (74–90)	82 (73–89)	NS
		Retrospective Study Design	4	450	84 (76–90)	81 (73–87)
		Vendor: Canon	3	339	92 (75–97)	87 (73–94)	NS
		Vendor: GE	5	518	82 (66–91)	81 (63–91)
		Vendor: Siemens	2	301	85 (64–94)	74 (51–88)
		Vendor: Supersonic Imagine	6	652	83 (70–91)	79 (64–89)
		RoB: Low	9	1216	85 (74–92)	75 (63–84)	NS
		RoB: Unclear and/or High	7	594	84 (76–90)	81 (73–87)
	Advanced	East Asian Countries	10	1199	86 (78–92)	81 (75–86)	NS
		Other Countries	5	478	87 (74–94)	83 (74–90)
		Prospective Study Design	12	1260	86 (80–91)	84 (78–89)	NS
		Retrospective Study Design	2	223	86 (68–95)	82 (76–86)
		Vendor: Canon	3	339	98 (84–100)	82 (72–98)	p < 0.05
		Vendor: GE	7	668	81 (73–87)	86 (78–91)
		Vendor: Siemens	1	222	91 (84–96)	76 (68–83)
		Vendor: Supersonic Imagine	2	230	90 (77–96)	62 (45–76)
		RoB: Low	7	860	86 (77–92)	83 (77–90)	NS
		RoB: Unclear and/or High	7	623	87 (76–93)	81 (69–88)
VCTE	Significant	East Asian Countries	12	1774	80 (71–86)	74 (63–82)	NS
		Other Countries	23	5065	80 (74–85)	66 (58–73)
		Prospective Study Design	28	5010	79 (75–83)	72 (66–78)	NS
		Retrospective Study Design	9	2380	84 (79–89)	59 (50–69)
		Single Hospital Design	22	2820	77 (72–82)	70 (64–77)	NS
		Multicenter Design	18	5122	81 (76–85)	69 (61–75)
		RoB: Low	23	2937	77 (72–82)	72 (64–78)	NS
		RoB: Unclear and/or High	17	4773	82 (76–86)	65 (56–73)
	Advanced	East Asian Countries	14	2178	87 (80–92)	71 (59–81)	NS
		Other Countries	25	8691	87 (83–91)	69 (60–77)
		Prospective Study Design	29	7541	88 (84–91)	72 (66–78)	NS
		Retrospective Study Design	12	5328	87 (82–91)	64 (55–71)
		Single Hospital Design	22	3148	84 (79–89)	71 (61–79)	NS
		Multicenter Design	22	9881	89 (86–92)	70 (60–78)
		RoB: Low	22	7065	88 (84–91)	61 (52–69)	p < 0.05
		RoB: Unclear and/or High	19	5804	87 (82–90)	79 (72–84)
MRE	Significant	East Asian Countries	7	1125	88 (82–92)	84 (77–89)	NS
		Other Countries	10	970	77 (67–84)	88 (83–91)
		Prospective Study Design	10	1277	83 (73–90)	87 (81–91)	NS
		Retrospective Study Design	3	394	88 (73–95)	83 (71–91)
		Vendor: GE	9	1276	86 (75–92)	84 (78–89)	NS
		Vendor: Siemens	2	166	85 (59–96)	93 (80–98)
		Magnet Strength: 1.5T	4	389	84 (69–93)	82 (71–90)	NS
		Magnet Strength: 3T	8	1112	83 (71–91)	87 (80–91)
		Pulse Sequence: GRE	7	642	84 (72–91)	82 (72–88)	NS
		Pulse Sequence: SE–EPI	4	458	83 (66–93)	89 (77–95)
		RoB: Low	10	1241	82 (72–89)	87 (81–91)	NS
		RoB: Unclear and/or High	4	498	83 (71–91)	86 (78–92)
	Advanced	East Asian Countries	1	201	82 (73–89)	92 (84–96)	NS
		Other Countries	4	305	86 (75–93)	85 (80–89)
		Prospective Study Design	5	403	86 (76–93)	88 (83–92)	N/A
		Retrospective Study Design	0	0	N/A	N/A
		Vendor: GE	3	292	82 (68–91)	91 (80–96)	N/A
		Vendor: Siemens	0	0	N/A	N/A
		Magnet Strength: 1.5T	1	59	92 (73–99)	89 (73–97)	NS
		Magnet Strength: 3T	3	292	82 (68–91)	90 (81–95)
		Pulse Sequence: GRE	3	253	85 (73–92)	86 (80–91)	NS
		Pulse Sequence: SE–EPI	1	98	87 (60–98)	96 (90–99)
		RoB: Low	3	292	82 (68–91)	91 (82–96)	NS
		RoB: Unclear and/or High	2	111	93 (74–98)	81 (62–92)

Subgroup analysis performed by modality and fibrosis severity for studies reporting thresholds between 7–9 kPa (8 ± 1 kPa). Categories with an insufficient number of studies could not be analyzed and are labeled as “N/A”; No. studies = number of studies; No. patients = number of patients included in subgroup assessment; GE = General Electric; RoB = Risk of Bias stratified by all domains that are deemed low risk versus studies with any single domain deemed unclear or high risk of bias; GRE = gradient echo; SE-EPI = spin echoplanar; N/A = not analyzable; NS = not significant.

. No significant differences were identified between sensitivities across all modalities and fibrosis severity. Studies using more than 10 measurements with pSWE had a higher specificity for detecting advanced hepatic fibrosis (84% vs. 57%, p < 0.05). GE demonstrated a higher specificity than Supersonic Imagine for detecting advanced hepatic fibrosis using 2D-SWE (86% vs. 62%, p < 0.05) but no difference was seen for Canon or Siemens. Studies with a single domain rated as unclear or high risk of bias demonstrated a higher specificity than low-risk studies for detecting advanced hepatic fibrosis using VCTE (79% vs. 61%, p < 0.05). No other differences were seen between specificities across modalities and fibrosis severity.

3.4. Risk of Bias Assessment

Risk of bias and applicability concerns are shown for each study separated by modality in Appendix E. Most imaging-based modalities (pSWE, 2D-SWE, VCTE, and MRE) were deemed high risk of bias for the index test as most studies used a retrospective threshold to evaluate accuracy. Many studies would also report a 90% sensitivity and/or 90% specificity, and these were included for multilevel evaluation when reported. Most studies evaluating imaging-based modalities were deemed low or unclear risk for patient selection, index test, and flow and timing. Most studies evaluating FIB-4 were deemed low or unclear risk of bias across all four domains.

4. Discussion

The European Association for the Study of Liver (EASL), American Association for the Study of Liver Disease (AASLD), and the American Gastroenterological Association (AGA) have each recently published guidelines for population-level screening of patients with suspected or diagnosed MASLD [1,2,3]. These guidelines use a combination of blood-based, then imaging-based recommendations. An initial blood-based investigation is suggested to be a more cost-effective first step with each guideline recommending the use of FIB-4 as the investigation of choice. In these pathways, a FIB-4 score of <1.3 characterizes a patient as low risk for liver-related outcomes, while a FIB-4 score of >2.67 characterizes a patient as high risk. Scores of 1.3–2.67 are deemed to be indeterminate and warrant further evaluation with an imaging-based investigation. Our meta-analysis of over 50,000 international patients receiving FIB-4 testing suggests that a “rule out” threshold of 1.3 results in sensitivities of only 71–74% for significant and advanced hepatic fibrosis. By contrast, a threshold of <0.75 can result in a sensitivity of nearly 90% for excluding advanced hepatic fibrosis. Conversely, a “rule in” threshold warranting referral to hepatology of 2.67 can result in specificities of 96–97% for detecting significant and advanced hepatic steatosis but with a sensitivity of only 32% (95% CI 27–38%). A slightly lower “rule in” threshold of ≥2.4 would result in a specificity nearing 90% and could also increase sensitivity by approximately 10% (42% [35–50%]), but it may result in an increase of too many patients without hepatic fibrosis being referred to hepatology services.

For imaging-based investigations recommended in indeterminate scenarios, existing guidelines recommend a modality- and vendor-agnostic threshold of 8 kPa as the low-risk threshold with two guidelines recommending follow-up with FIB-4 in 1–3 years [1,3] and one guideline recommending either follow-up or liver biopsy in these low-risk patients [2]. However, our meta-analysis indicates that a threshold of <8 kPa is likely to have a sensitivity between 70–80% when using 2D-SWE or VCTE for detecting advanced hepatic fibrosis, and a lower sensitivity when using pSWE. By contrast, individualized sensitivities of nearly 90% can be achieved with independent modality thresholds of 3 kPa for pSWE, 5 kPa for 2D SWE, and 7 kPa for VCTE (“3/5/7 rule”) for detecting advanced hepatic fibrosis. Conversely, a “rule in” threshold of 13 kPa could result in specificities of nearly 90% detecting advanced hepatic fibrosis with pSWE and VCTE and significant hepatic fibrosis with 2D-SWE. These thresholds would more closely align with the comparatively more conservative thresholds recommended by the Society of Radiologists in Ultrasound (SRU) for pSWE and 2D-SWE, which use a “rule out” threshold of 5 kPa and an indeterminate range of 5–13 kPa [4].

Finally, the ASSLD guideline recommends an MRE threshold of ≤3 kPa to risk stratify indeterminate risk patients into a low-risk category [2]. Our meta-analysis suggests that a threshold of 3 kPa would result in a sensitivity of around 80% for detecting significant and advanced hepatic fibrosis, while a threshold closer to 2.5 kPa would increase the sensitivity to around 90% for detecting significant hepatic fibrosis. A lower threshold was identified in our meta-analysis for detecting advanced hepatic fibrosis, which is felt to be a function of limited underlying studies evaluating this disease severity, so preference is placed on the identified threshold for significant hepatic fibrosis.

Based on these results, a revised proposed risk stratification pathway is shown in Figure 2. This proposed pathway lowers the “rule out” FIB-4 threshold from existing guidelines and provides modality-specific thresholds for imaging-based investigations. While this is likely to result in higher frequency use of imaging-based investigations and possibly a higher number of hepatology referrals for “indeterminate risk” patients, the ability to risk stratify MASLD patients into a “low-risk” category would be improved with the sensitivity increasing from below 80% to approximately 90%.

4.1. Limitations

Our meta-analysis is limited by near-universal issues of presumed variability in diagnostic test accuracy systematic reviews [5]. In an effort to identify potential sources of variability within different modalities, an a priori subgroup analysis explored several potential causes for variability separated by modality and fibrosis severity with thresholds between 7 and 9 kPa. No significant differences in sensitivity were identified in these thresholds, and only three differences in specificity were identified (higher specificities identified with ≥10 measurements for pSWE, for GE versus Supersonic Imagine vendors for 2D-SWE, and for studies with unclear and/or high-risk domains for VCTE). This likely under-recognizes other potential sources for variability which were not explored due to lack of available data or non-extraction. One area of particular concern was vendor-specific accuracies for pSWE, 2D-SWE, and MRE as these technologies rely on vendor-specific proprietary software parameters. Unlike VCTE, which uses a single-vendor product, these modalities may be dependent on the vendor-specific acquisition. However, subgroup analysis identified only a statistically lower accuracy for Supersonic Imagine compared to GE for 2D-SWE with thresholds between 7 and 9 kPa for detecting advanced but not significant hepatic fibrosis. A separate vendor-specific analysis of all thresholds was also performed to determine the appropriateness of vendor-specific threshold recommendations, although limited available primary data precluded detailed comparison between these vendors stratified by modality and fibrosis severity. The available data suggest that Supersonic Imagine may benefit from a lower “rule out” threshold recommendation than GE to sustain sensitivities around 90% for 2D-SWE, although this was not statistically significant. Another potential area for variability between modalities was the conversions needed for some studies such as shear wave velocities to Young’s modulus (most often needed in pSWE studies) and conversion to a METAVIR grading system from other histologic grading systems. It is possible that differences between vendor-specific conversion calculations and assumptions (such as density of soft tissue and liver) may have contributed to lower threshold values for pSWE compared to 2D-SWE and VCTE, for example. As such, it is recommended that manual conversions from shear wave speed to Young’s modulus should utilize the same calculation as was used in this study when referencing the modified threshold recommendations outlined here. Additionally, to reduce the risk for potential variability within an individual, surveillance of the same patient would best be performed using the same modality and, ideally, the same machine as was used previously. The SRU suggests a variance of 10% using the same machine to indicate a clinically significant change in individualized surveillance for pSWE and 2D-SWE [4].

4.2. Future Research

Several avenues of future research are needed. First, continued accuracy studies, particularly for image-based investigations including pSWE, 2D-SWE, and MRE will help future data synthesis identify important subgroups to further refine threshold recommendations. In particular, further studies evaluating vendor-specific accuracies and thresholds will help define the potential utility of vendor-specific threshold recommendations. Additionally, as population screening gains traction on a national and international level, large-scale longitudinal evaluation of the change in referral patterns and patient outcomes will be required. Through this process, optimal surveillance recommendations—currently based on expert consensus and dependent on individual guidelines and regional practice—can be further refined. For those receiving imaging-based follow-up, evaluation of individual modality-specific and clinically significant thresholds for change are needed. This includes studies evaluating specific subgroups of MASLD which can inform referral pathways and threshold recommendations. For example, limited literature evaluating patients <18 years of age is available, and recommended guidelines for the pediatric population with MASLD are not clearly established. Alternatively, type 2 diabetes is known to be the strongest predictor for developing MASLD/MASH-related advanced fibrosis and liver related complications and may warrant shorter surveillance periods [1]. Additionally, a cost-analysis study to determine the economic consequences of the proposed changes outlined in this study can help inform economic decisions necessary for these algorithmic changes.

4.3. Conclusions

In summary, this comprehensive systematic review, including more than 200 studies and over 80,000 index tests, offers novel insights into the accuracies of individual blood-based and imaging-based guidelines available to date. Our study suggests that existing guideline thresholds for determining low-risk patients with FIB-4 should be lowered, and image-based investigations for indeterminate patients should be individualized by modality. A revised proposed algorithm has been provided which can be used to inform future guideline development and revisions, and, more importantly, population-based screening and risk stratification practices.