Performance of continuous controlled attenuation parameter and liver stiffness measurement by the novel SmartExam in metabolic dysfunction‐associated steatotic liver disease

FibroScan® Expert 630 and FibroScan® Mini+430 are novel vibration‐controlled transient elastography devices equipped with the same SmartExam software, which allows continuous measurement of controlled attenuation parameter (CAP) during the entire examination. This study aims to compare the CAP variabilities and the quantification for liver fibrosis and steatosis between the conventional FibroScan and the SmartExam‐equipped machines in patients with metabolic dysfunction‐associated steatotic liver disease (MASLD).

With the global prevalence of metabolic dysfunction-associated steatotic liver disease (MASLD) increasing to over 30%, there is a parallel increase in hepatic adverse outcomes. 1It is therefore of wide interest to diagnose and assess fatty liver disease with rapid, non-invasive and reproducible methods. 2 Vibration-controlled transient elastography (VCTE) is one of the technologies widely applied to various liver diseases in the past two decades.4][5] However, a recent meta-analysis reported the less optimal performance of CAP when extending to a certain population, 6 that is, patients with MASLD with a higher BMI due to the inherent technical limitations of this ultrasound-based steatosis quantification approach.Nevertheless, CAP still represents the first highly standardized and evaluated approach using ultrasound signal modulation for steatosis quantification. 6TE has M and XL probes with different measurement depths for patients with various degrees of adiposity.Two latest models of VCTE, FibroScan® Expert 630 and FibroScan® Mini+430 (Echosens, Paris, France), introduced a new software called the SmartExam.It is featured by the capability of continuous measurement of CAP 7 during the entire examination and automatic measurement depth adaptation to patients' body build.While the original FibroScan 502 machine typically obtains 10 CAP measurements along with LSM, over 200 CAP measurements can be achieved with continuous measurements once the SmartExam software detects that the probe contacts the skin.Although the recent technology should theoretically improve diagnostic accuracy, data remains scarce. 8In a recent single-centre retrospective study, Nogami et al. evaluated the performance of continuous CAP using magnetic resonance imaging proton density fat fraction as the reference standard.To determine whether the performance of CAP in patients with MASLD can be further improved by the introduction of cCAP presents great significance given the suboptimal accuracy of CAP from Petroff et al.'s study. 6In this study, we only included the patients who had undergone liver biopsy.

| Clinical assessment
Body weight, body height and waist circumference were measured for all patients, and body mass index (BMI) was calculated as body weight (kg) divided by body height (m) squared.Clinical parameters involving alanine aminotransferase (ALT), fasting glucose, glycated haemoglobin, lipid profile and history of hypertension and diabetes mellitus were recorded, and all measurements were performed within 48 h before the liver biopsy except for two patients in Hong Kong who had their clinical assessment within 2 months of the liver biopsy.

| Histological assessment
Liver histology was the reference standard for comparing diagnostic performances of measurements based on two VCTE devices in this study.Percutaneous liver biopsy was either performed with the 16G or 18G Temno Tru-Cut needle, and all specimens were ≥15 mm in length and considered acceptable.Liver histology for each specimen was assessed by two experienced histopathologists who were blinded to the clinical data.Liver specimens were scored using the NASH Clinical Research Network system. 9 of hepatocytes with steatosis: S1, 5%-33%; S2, 34%-66% and S3, >66%.

| Transient elastography
CAP and liver stiffness were measured by VCTE within 3 days before liver biopsy.VCTE measurements were performed using both the FibroScan 502 and 630/430 machines, and M and XL probes for each individual regardless of the automatic probe selection suggested by the device.Patients were required to fast overnight for at least 8 h for blood tests and VCTE.We consider the examination as reliable if 10 valid acquisitions for LSM and CAP were obtained, and the interquartile range (IQR)-to-median ratio of LSM was ≤.3.Operators had been fully trained with more than 50 examinations and were blinded to patients' diagnoses and clinical data.
The VCTE machine has an automated probe selection tool to guide the use of M or XL probe based on the skin-to-liver capsule distance.Additionally, as all patients in this study had both M and XL probe examinations, we also performed sensitivity analyses using either M or XL probe results based on (1) the machine recommendation, and ( 2) BMI (M probe data used when BMI <30 and XL probe data used when BMI ≥30 kg/m 2 based on the EASL-ALEH Clinical Practice Guidelines). 10,11

| Statistical analyses
Continuous variables with normal distribution were reported as mean ± standard deviation (SD), and skewed data were reported as median (the first quartile-the third quartile).The per-protocol analysis was performed for this study.Spearman's correlation co-

| RE SULTS
From April 2021 to May 2023, 141 patients underwent liver biopsy and paired VCTE examinations at the two participating centres.After excluding 9 patients with incomplete data and 14 with other liver diseases, 118 patients (84 from Hong Kong, 34 from Japan) were included in the final analysis.All recruited patients were Asians, with a mean age of 56 ± 11 years (Table 1).The mean BMI was 28.4 kg/m 2 , and 30.5% had BMI ≥ 30 kg/m 2 and 49.2% had F3-F4 fibrosis.The median NAFLD activity score (NAS) was 4 (3-5).81.3% had S1-S2 steatosis.The median LSM was 9.8 kPa and 7.9 kPa, and the median sLSM was 8.8 kPa and 7.9 kPa for M and XL probes.For the comparison of measurements from different devices, successful examinations (10 valid measurements for LSM, sLSM and CAP; and >200 valid measurements for cCAP) were obtained for all patients using both probes with all FibroScan devices.All LSM acquisitions were reliable with their corresponding IQR-to-median ratio ≤.3.No adverse events occur from performing the liver biopsy or the VCTE.

| Variability of CAP and cCAP
Because the variability of CAP and cCAP was recorded as IQR and standard deviation, respectively, the latter was converted to standard deviation by the formula SD = IQR/1.35. 12 Standard deviations of cCAP were significantly lower than those of CAP for both the M and XL probes (p < .0001)(Figure 1).CAP variability decreased by 48.6%, 46.0% and 53.1% for the M probe, XL probe and automatically selected probe, respectively.Considering that two cohorts varied in the IQR of CAP values measured by the XL probe, we performed comparisons of CAP and cCAP variance using the XL probe for two cohorts separately, and both cohorts showed a significant reduction of the variability after switching from CAP to cCAP (Figure S3).The variabilities of cCAP showed no significant difference between the two centres (p = .97),while the variability of CAP among the HK cohort was significantly different from that among the JP cohort (p = .046)based on t-test analyses.The linear regression model was performed for potential covariates associated with cCAP variability, which suggested only the effect of probe selection (using an M or XL probe) was significant (p < .001)( (p < .001) as well as centre as the random effect based on the linear mixed regression model, with the percentage of centre variance out of total being 5.5%.

| Diagnostic accuracy of LSM and sLSM
We evaluated the diagnostic performance of LSM and sLSM for identifying liver fibrosis stage by comparing AUROC and their performance with validated LSM cut-offs by Wong et al. 10 with liver histology as the reference standard (Tables S1 and 3).Table S1 pre At the cut-off of <5 kPa to rule out F2-F4 fibrosis, the sensitivity of LSM and sLSM to detect F2-F4 for M probe were both 96.8%; after switching to the XL probe, both LSM and sLSM had a decreased sensitivity of 86% and 93.6%, respectively.The NPV of sLSM for the XL probe remained at 70%, similar to the NPV of measurements with the M probe, whereas NPV of LSM decreased to 48%.Using the <10 kPa cut-off to rule out F4, the sensitivity of LSM was 92.9% and 84.6% for M and XL probes, 78.6% and 76.9% for sLSM, respectively; while at a cut-off of >15 kPa to rule in advanced fibrosis or F4 alone, LSM and sLSM showed comparably good performance in their specificity for both probes, and all specificities were above 80% M probe.

| Diagnostic accuracy of CAP and cCAP
The diagnostic performance of CAP and cCAP was evaluated using validated cut-offs from two meta-analyses. 6,13Because we only recruited patients with biopsy-proven MASLD, the targets of evaluation were S1 versus S2-S3 steatosis and S1-S2 versus S3 steatosis.
F I G U R E 1 CAP and cCAP variabilities for M, XL or automatically selected probes.CAP, controlled attenuation parameter; cCAP, continuous controlled attenuation parameter.

TA B L E 2
Covariates associated with the variability of CAP and cCAP.The random effect of centre was interpreted as the estimated percentage of the centre variance out of the total variance of interquartile range of CAP based on CAP readings measured with both probes for the HK and JP cohort.b Linear regression model was fitted for cCAP analyses given that the random effect estimate of centre on the variability of cCAP was close to zero and singularity was suggested.
The diagnostic performances of CAP and cCAP were moderate or good with AUROC ranging from .70 to .86, and no significant difference was found between CAP and cCAP for both the M and XL probes (Figure 3, Table S5).
With 268, 317 and 310 dB/m as cut-offs for the M probe, XL probe and automatic probe to rule in S2-S3, CAP showed specificity of 44.2%, 79.6%, and 87.8%, while the specificity of cCAP was 51.2%, 86.4%, and 85.4% respectively.To determine S3 steatosis, cut-offs Correlation between CAP and cCAP using the M probe, XL probe, automatic probe selection by the machine and BMI-based probe selection (M probe when BMI < 30 and XL probe when BMI ≥ 30 kg/m 2 ).CAP, controlled attenuation parameter; cCAP, continuous controlled attenuation parameter.

TA B L E 3
Diagnostic accuracy of LSM and sLSM using cut-offs from literature.

Fibrosis stage Probe
Cut-off a (kPa)  respectively.The specificity of CAP and cCAP was 67.4% and 77.9% for the XL probe and 77.3% and 83.0% for the automatic probe (Table S2).

| DISCUSS ION
This retrospective study was conducted on patients with biopsyproven MASLD, and all patients underwent liver stiffness and steatosis measurements with both probes of the FibroScan 502 and 630/430 machines at two centres.Through comprehensive comparisons between readings of two devices across various probes, we confirmed that FibroScan 630/430 can reduce CAP variability for liver steatosis measurement while it has similar diagnostic accuracy for liver fibrosis staging and steatosis grading compared with the FibroScan 502.
The variability of cCAP was significantly lower than that of CAP for all probes.This finding is consistent with the study by Nogami et al. and can be attributed to the capability of the continuous measurements featured by FibroScan 430. 8We observed no significant difference in the variability of CAP readings by the automatic probe between two cohorts, and only a marginal difference in cCAP variability (Table 1).Given that the device performs measurements using the automatic probe selected in routine clinical practice, this comparison also provides references for the application of cCAP by SmartExam-equipped VCTE in real-world clinical settings.The AUROC of CAP and cCAP measurements with both M and XL probes were comparable in detecting S2-3 and S3 (S2-3: .84vs .84,S3: .70 vs .72;both p > .05),suggesting a similar discriminatory power of cCAP.The suboptimal performance of CAP and cCAP, with AUROC at the level of around .7, coincided with the finding of Petroff et al. 6 To further determine the potential factors interfering with the performance of CAPs, we fitted two regression models to analyse parameters associated with variability of CAP readings given that high variability of CAP readings (>40 dB/m) has been reported to be negatively associated with the accuracy of the serial measurements (Table 2). 2,14near mixed regression model was built for CAP given the acrosscentre variability of CAP but not cCAP was justified through t-test analyses.Additionally, both BMI and probe had significant associations with the variability of CAP as fixed effects.Based on the linear regression analysis fitted for cCAP, however, both centre and BMI were insignificantly associated with the variability of cCAP.Our study showed that both BMI and probe had significant associations with the variability of CAP, which echoes with the previous meta-analysis 6 suggesting that the accuracy of CAP is lower among patients with MASLD and high BMI.On contrary, Although the correlations between CAPs across the various probe use seemed to be weaker than those for LSMs, the difference between CAPs was relatively evenly scattered across the range of the average of two measurements (Figure S4), illustrating that no consistent bias existed between the CAPs.Besides, the ICC analysis also suggested consistency between CAP and cCAP (Table S4).
We determined the rule of five interpretation of LSM as proposed by Baveno VI Faculty for the evaluation of chronic liver disease. 15We adopted the same cut-offs for the comparison between LSM and sLSM regardless of the probe used since lower values generated by the XL probe and the increased liver stiffness values due to obesity can partially offset each other. 10At the cut-off of 5 kPa to rule out F1-F4 stage fibrosis, the sensitivity of sLSM with the XL probe remained as high as 86.1% compared to that with the M probe without much sacrifice in its specificity, 62.5%, superior to the performance of LSM with the XL probe.At the cut-offs of 10 and 15 kPa, both LSM and sLSM performed well in the exclusion and inclusion of significant and advanced fibrosis, respectively.
We followed validated cut-offs of CAP based on two metaanalyses. 6,13CAP and cCAP showed similarly relatively low specificity with around 90% sensitivity at these cut-offs.Using MASLD-derived cut-offs from Petroff et al., 6 CAP and cCAP also showed comparable diagnostic properties to determine steatosis levels.cCAP generated slightly better specificity than CAP with the validated cut-offs in the majority of comparisons across the different probes, along with the commensurate lower sensitivity.
As illustrated by the Bland-Altman plots for CAP and cCAP readings (Figure S4), the median difference was 5 dB/m and 13 dB/m for M and XL probes, respectively, showing the consistency between the two readings.Although the limit of agreement between CAP and cCAP spans a relatively wide range, this dispersion of CAP values can be partially attributed to individual differences; furthermore, the median CAP and cCAP measurements were comparable for each steatosis grade.Readings for S1, S2 and S3 steatosis were 279.5 and 267.5 dB/m, 311 and 315 dB/m and 332 and 323 dB/m, respectively.More importantly, the diagnostic accuracy of liver steatosis showed no significant difference between the two CAPs, thus we may regard cCAP as a potentially non-inferior substitute for CAP and further demonstrate if the diagnostic performances of both can be evaluated with unified cut-offs in the future.Providing evidence for the identified cut-off usage can be promising since it reduces the amount of potential labour required for determining optimal cut-offs for cCAP in related studies.
To our understanding, this is the first head-to-head comparison between the original and SmartExam-based VCTE examination using liver histology as the reference standard.However, our study also has a few limitations.First, the sample size was relatively small.However, we performed four measurements with both probes each for the two VCTE devices, which was not done in previous studies. 4is allowed a thorough comparison between the two examinations.
Besides, we only recruited patients with biopsy-proven MASLD,

CO N FLI C T O F I NTER E S T S TATEM ENT
GL-HW has served as an advisory committee member for Gilead Sciences and Janssen and as a speaker for Abbott, Abbvie, Ascletis, Bristol-Myers Squibb, Echosens, Gilead Sciences, Janssen and Roche.She has also received a research grant from Gilead Sciences.
VW-SW has served as an advisory committee member for AbbVie, Boehringer Ingelheim, Echosens, Gilead Sciences, Intercept, Inventiva, Merck, Novo Nordisk, Pfizer, Sagimet Biosciences and TARGET PharmaSolutions and a speaker for Abbott, AbbVie, Gilead Sciences, Novo Nordisk and Unilab.He has also received a research grant from Gilead Sciences and is a co-founder of Illuminatio Medical Technology Limited.TC-FY has served as an advisory committee member and a speaker for Gilead Sciences.

2. 1 |
Patients For this retrospective two-centre cohort study, we recruited 141 patients with clinically suspected MASLD who underwent liver biopsy in the Prince of Wales Hospital, Hong Kong from April 2021 to May 2023 and in Yokohama City University Hospital, Japan from August 2021 to February 2023.Inclusion criteria were ≥18-year-old patients with biopsy-proven MASLD; available measurements of both conventional and SmartExam-equipped transient elastography performed before liver biopsy.We excluded patients with steatosis grade S0 based on their histology and those with unavailable VCTE measurements of either pattern or measurements were not performed within 3 days before their liver biopsy.This study was approved by the institutional review board of the Chinese University of Hong Kong and Yokohama City University and conducted based on the procedures depicted in Figure S1.All patients provided informed written consent.Eighteen patients in this study were included in a previously published study on SmartExam based on MRI-PDFF.
Fibrosis stage ranges from F0 to F4 with F3-F4 representing advanced fibrosis (F3 = bridging fibrosis, F4 = cirrhosis).Steatosis was scored based on the percentage Key points The SmartExam-based vibration-controlled transient elastography reduces the variability of controlled attenuation parameter (CAP) measurement and achieves a similar accuracy as the FibroScan 502 device for the estimation of both hepatic steatosis and fibrosis.Future studies should determine if continuous CAP is a better tool to monitor changes in steatosis than the original CAP.
efficient and intraclass correlation coefficient (ICC) were used for describing the correlation between LSM and LSM measured by the SmartExam-equipped Fibroscan (sLSM) and CAP and continuous CAP (cCAP), respectively.Bland-Altman plots were used to describe the correlations between LSM and sLSM and CAP and cCAP.The Wilcoxon-signed rank test was used for the comparison of CAP and cCAP variabilities after standard transformation of IQR (interquartile range) to SD for CAP readings.T-test was used to compare the CAP and cCAP variabilities between different centres.Two linear regression models were used to analyse the effects of clinical parameters on the variability of CAP and cCAP.Diagnostic accuracy of LSMs and CAPs to detect significant liver fibrosis, advanced fibrosis and steatosis were evaluated by calculating the area under the receiver-operating characteristic curve (AUROC), respectively, with liver histology as the reference standard.A 95% confidence interval has been calculated for the AUROCs of both CAPs and LSMs.AUROC ranging from .60 to .90, a sample size of 118 patients, with roughly one-third of patients each having grade 1, 2 and 3 steatosis, would determine the AUROC of CAP and cCAP with a standard error of .03-.05.The cut-offs of LSM and CAP for optimal AUROC were based on the Youden index.Delong test was used for AUROC comparisons between LSM and sLSM, CAP and cCAP for the M probe, XL probe, automatically adjusted probe and BMI-based probe for each patient.Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and accuracy of LSM, sLSM, CAP and cCAP were calculated with validated cut-offs from the existing literature.A two-sided p value is considered statistically significant if less than .05.No replacement of missing data has been performed.All analyses were performed using R (version 4.2.1;R Core Team 2022).
Measurements by the FibroScan 502 and SmartExam-equipped machines correlated well with each other.A strong correlation was observed between LSM and sLSM with correlation coefficients being .95,.92,.93 and .94for the M probe, XL probe automatic probe selection and BMI-based probe selection, respectively (FigureS2).For CAP and cCAP, the correlation was lower than LSM but remained significant with correlation coefficients of .68,.78,.77and .72 for M probe, XL probe, automatic probe selection and BMIbased probe selection, respectively (Figure2).
sents the comparison of the diagnostic properties of LSM and sLSM based on AUROC analyses.For the M probe, the AUROC of LSM and sLSM detect various stages of fibrosis was similar (.87-.88 for LSM and .83-.87 for sLSM).For the XL probe, the AUROC of LSM was .83-.88, and that of sLSM was .84-.86.Similar results were shown with probe selection by the machine or based on BMI, with all AUROC above .8.No significant difference in AUROC between LSM and sLSM was observed for any level of fibrosis.
of 280 dB/m, 333 dB/m and 331 dB/m were used for the M probe, XL probe and automatic probe selection.Both CAP and cCAP showed a high sensitivity for M probe with the sensitivity of 95.2% and 90.5%,

F I G U R E 3
cCAP is less susceptible to interventions of some clinical parameters as well as different research settings compared to CAP and may serve as a non-inferior and more reproducible monitoring tool for sequential changes of liver steatosis in future multi-centre studies.Diagnostic accuracy of CAP and cCAP to detect hepatic steatosis.CAP, controlled attenuation parameter; cCAP, continuous controlled attenuation parameter.
which reduced the aetiology-derived bias on the accuracy of CAP indicated in the previous study.Second, despite the two-centre design, all the patients were Asians, and patients with severe obesity were underrepresented.Third, because ours was a cross-sectional study, the role of the SmartExam for monitoring is unclear.More longitudinal, standardized studies on populations of various characteristics would help us to establish a greater degree of accuracy in our conclusions.In conclusion, our two-centre study confirms that the SmartExam-based VCTE reduces the variability of CAP measurement and achieves a similar accuracy as the FibroScan 502 device for the estimation of both hepatic steatosis and fibrosis.Our study also provides evidence for using cCAP as a promising alternative for CAP in future practice.FU N D I N G I N FO R M ATI O NThis study was supported in part by a direct grant from The Chinese University of Hong Kong (project reference 2022.031) and the General Research Fund by the Hong Kong SAR Government (project reference 14106923).

Table 2
Clinical characteristics of patients with metabolic dysfunction-associated steatotic liver disease.Continuous variables are expressed in mean ± standard deviation or median (interquartile range).p values represent comparisons between patients from Hong Kong and Japan.