Functional Assessment by Gd-EOB-MRI of the Future Liver Remnant Predicts Posthepatectomy Liver Failure: A Systematic Review and Meta-Analysis

Functional Assessment by Gd-EOB-MRI of the Future Liver Remnant Predicts Posthepatectomy Liver Failure: A Systematic Review and Meta-Analysis | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Functional Assessment by Gd-EOB-MRI of the Future Liver Remnant Predicts Posthepatectomy Liver Failure: A Systematic Review and Meta-Analysis Yannan Liu, Xiaolei Shi This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9333721/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 8 You are reading this latest preprint version Abstract Background & Aims: Posthepatectomy liver failure (PHLF) is a major cause of morbidity and mortality following liver resection. Accurate preoperative prediction of posthepatectomy liver failure (PHLF) remains a clinical challenge. Gadoxetic acid-enhanced MRI (Gd-EOB-MRI) enables quantitative assessment of liver function. This systematic review and meta-analysis aims to determine the diagnostic accuracy of Gd-EOB-MRI quantitative parameters for predicting PHLF and to explore sources of heterogeneity. Methods: A systematic search of PubMed, Embase, Web of Science, and Cochrane Library was conducted from January 2011 to June 2025 for studies evaluating Gd-EOB-MRI parameters for PHLF prediction in adults undergoing hepatectomy. The bivariate random-effects model was used to calculate the pooled area under the curve (AUC). Heterogeneity was assessed using the I² statistic. Subgroup, threshold effect, and publication bias analyses were performed. Results: Nineteen studies involving 3,658 patients were included. Future liver remnant (FLR)-based parameters (14 studies) demonstrated high diagnostic accuracy with a pooled AUC of 0.85 (95% CI: 0.82–0.89) and moderate heterogeneity (I² = 27.9%). Whole liver-based parameters (7 studies) also showed good performance (pooled AUC = 0.80, 95% CI: 0.76–0.84). Subgroup analyses revealed no significant differences based on MRI parameters， study design, region, sample size, or PHLF definition. However, studies using 3.0T MRI scanners showed a significantly higher pooled AUC than those using 1.5T (0.859 vs. 0.763, p=0.018). A significant threshold effect was observed for FLR parameters (Spearman's ρ = 0.744, p=0.006) but not for whole liver parameters. No significant publication bias was detected. Conclusions: FLR-based Gd-EOB-MRI parameters provide excellent diagnostic performance for predicting PHLF, outperforming to whole liver-based assessment. This supports the integration of functional FLR assessment into preoperative risk stratification. However, the presence of a threshold effect underscores the need for standardized protocols and validated diagnostic cut-offs in future prospective multicenter studies. Prospective multicenter studies with standardized protocols and predefined diagnostic thresholds are needed to validate these findings and facilitate clinical translation. Hepatectomy Posthepatectomy liver failure Gadoxetic acid Magnetic resonance imaging Future liver remnant Meta-analysis Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Introduction Hepatectomy is a standard treatment for primary and secondary malignant liver tumors. However, posthepatectomy liver failure (PHLF) remains one of the most serious complications [1]. According to the International Study Group of Liver Surgery (ISGLS), PHLF is defined as a syndrome characterized by an international normalized ratio (INR) ≥ 1.6 and serum bilirubin ≥ 50 µmol/L on or after postoperative day 5 [1]. The reported incidence of PHLF varies widely (0.7%–30%), influenced by factors including the patient’s underlying liver disease, extent of resection, and postoperative management [2]. Accurate preoperative assessment of liver function is crucial for predicting PHLF risk, guiding surgical decisions, and improving patient outcomes. Traditional methods for evaluating liver function, such as indocyanine green clearance test (ICG-R15), Child-Pugh classification, and liver volumetry, are limited in their ability to assess regional liver function and predict postoperative functional reserve [3]. Gadoxetic acid-enhanced magnetic resonance imaging (Gd-EOB-MRI) has emerged as a promising tool for the non-invasive assessment of liver function [4]. Gadoxetic acid (Gd-EOB-DTPA) is a hepatocyte-specific contrast agent. Approximately 50% is taken up by hepatocytes via organic anion-transporting polypeptides (OATP1B1/1B3) and excreted into the bile canaliculi via multidrug resistance-associated protein 2 (MRP2), while the remainder is excreted renally [5]. This unique pharmacokinetic profile allows Gd-EOB-MRI to reflect hepatocyte function during the hepatobiliary phase (HBP), providing a new approach for assessing functional liver reserve. Quantitative analysis of various parameters on Gd-EOB-MRI, such as the hepatic uptake index (HUI), relative liver enhancement (RLE), and liver-to-spleen contrast index (LSI), can quantify hepatocyte uptake and excretion function [6]. Functional assessment based on the future liver remnant (FLR) is particularly promising for more accurately predicting postoperative liver function [7]. Although numerous studies have explored the value of Gd-EOB-MRI quantitative parameters in predicting PHLF, reported predictive performance varies due to differences in study design, parameter definitions, and PHLF diagnostic criteria. Furthermore, whether different types of quantitative parameters (based on signal intensity, functional volume, or T1 relaxation time) differ in predictive value, and the potential factors affecting predictive performance, require clarification through systematic review and meta-analysis. Therefore, this study aims to comprehensively evaluate the existing evidence through systematic review and meta-analysis, determine the overall diagnostic accuracy of Gd-EOB-MRI quantitative parameters for predicting PHLF, compare the predictive value of different parameter types, explore potential sources of heterogeneity, and provide high-quality evidence for PHLF risk assessment prior to hepatectomy. Materials and Methods 1.Study Design and Reporting Standards This systematic review and meta-analysis strictly adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [8]. Literature Search Strategy We conducted a systematic search of four electronic databases: PubMed (MEDLINE), Embase, Web of Science, and the Cochrane Library, from January 1, 2011, to June 21, 2025. The year 2011 was chosen as the start date because gadoxetic acid was approved in Europe in 2008 and in the United States in 2011, with relevant clinical studies predominantly conducted thereafter. The search strategy combined Medical Subject Headings (MeSH) terms and free-text words, adjusted according to the specific database. The search strategy for PubMed is provided as an example: ((("liver failure"[MeSH Terms] OR "liver failure"[All Fields] OR "hepatic failure"[All Fields] OR "hepatic insufficiency"[All Fields]) AND ("postoperative"[All Fields] OR "post-operative"[All Fields] OR "post-hepatectomy"[All Fields] OR "posthepatectomy"[All Fields])) OR "PHLF"[All Fields]) AND (("hepatectomy"[MeSH Terms] OR "hepatectomy"[All Fields] OR "liver resection"[All Fields] OR "hepatic resection"[All Fields]) AND ("magnetic resonance imaging"[MeSH Terms] OR "MRI"[All Fields] OR "gadoxetic acid"[All Fields] OR "Gd-EOB-DTPA"[All Fields] OR "Primovist"[All Fields] OR "Eovist"[All Fields])) Additionally, the reference lists of included studies were manually searched to identify potentially omitted relevant literature. No language restrictions were applied during the search, but only studies published in English were ultimately included. Inclusion and Exclusion Criteria Inclusion Criteria 1. Study type: Prospective or retrospective cohort studies, case-control studies. 2. Participants: Adult patients (≥18 years) scheduled for hepatectomy, regardless of underlying liver disease type. 3. Intervention: Preoperative Gd-EOB-MRI examination with extraction of at least one quantitative parameter for PHLF prediction. 4. Reference standard: Occurrence of PHLF postoperatively. 5. Outcomes: Clear definition of PHLF (preferably ISGLS criteria) and reporting of sufficient diagnostic accuracy data (e.g., sensitivity, specificity, AUC). 6. Language: English. Exclusion Criteria 1. Non-original studies (reviews, meta-analyses, case reports, conference abstracts, etc.) 2. Animal or in vitro studies. 3. Studies not providing diagnostic accuracy data for Gd-EOB-MRI quantitative parameters in predicting PHLF. 4. Incomplete data. 5. Duplicate publications (the study with the largest sample size or most complete data was retained). 6. Studies using Gd-EOB-MRI only for liver volumetry without functional assessment. 2. Classification of Gd-EOB-MRI Quantitative Parameters Based on classification methods from literature on liver-specific contrast agents [9], Gd-EOB-MRI quantitative parameters were systematically categorized according to their physiological basis and mathematical definition: Signal Intensity-Based Parameters These parameters primarily reflect the uptake capacity of hepatocytes for gadoxetic acid, quantified by comparing signal intensity changes between the HBP and precontrast phases. · Hepatic Uptake Index (HUI): HUI = (SI_HBP - SI_pre) / SI_pre, where SI_HBP is signal intensity in the HBP and SI_pre is signal intensity precontrast. · Relative Liver Enhancement (RLE): RLE = (SI_HBP - SI_pre) / SI_pre × 100% · Liver-to-Spleen Contrast Index (LSI): LSI = (SI_liver_HBP / SI_spleen_HBP) / (SI_liver_pre / SI_spleen_pre) Functional Volume-Based Parameters These parameters combine anatomical volume information with functional assessment, quantifying the functional reserve of specific liver regions. · Functional Future Liver Remnant volume (fFLR): fFLR = FLR_volume × mean_HUI_FLR · Functional Liver Volume Fraction (fLVF): fLVF = (functional_liver_volume / total_liver_volume) × 100% T1 Relaxation Time-Based Parameters These parameters quantify the effect of gadoxetic acid on liver T1 relaxation time via T1 mapping, providing a standardized functional assessment. · Change in T1 Relaxation Rate (ΔR1): ΔR1 = 1/T1_post - 1/T1_pre = (T1_pre - T1_post) / (T1_pre × T1_post) · T1 Reduction Rate: T1_reduction_rate = (T1_pre - T1_post) / T1_pre × 100% 3.Study Selection and Data Extraction Study selection was performed independently by two investigators, first based on title and abstract screening, followed by full-text review of potentially eligible studies. Disagreements were resolved through discussion. Data were extracted using a pre-designed standardized form, including: Study Characteristics： First author, publication year, country/region；Study design (prospective/retrospective)；Study period, follow-up duration；Inclusion and exclusion criteria Patient Characteristics： Sample size, age, sex distribution；Underlying liver disease type (hepatocellular carcinoma, biliary tract malignancy, metastatic liver cancer, etc.)；Child-Pugh grade；Extent and type of hepatectomy Technical Parameters： MRI scanner model and magnetic field strength (1.5T/3.0T)；Dose and injection method of gadoxetic acid Gd-EOB-MRI Parameters： Parameter type and calculation method；Measurement region (whole liver/FLR)；Region of interest (ROI) selection criteria Outcome Measures： PHLF definition and diagnostic criteria；PHLF incidence；Diagnostic accuracy measures (sensitivity, specificity, AUC, etc.)；Optimal cutoff value 4. Quality Assessment The Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2) tool [10] was used to assess the quality of included studies. QUADAS-2 covers four key domains: patient selection, index test, reference standard, and flow and timing. Each domain was assessed for risk of bias and concerns regarding applicability, categorized as low, high, or unclear risk. Quality assessment was conducted independently by two investigators, with disagreements resolved through discussion. We paid particular attention to: · Risk of selection bias in patient selection. · Whether Gd-EOB-MRI was performed according to a standard protocol. · Whether the PHLF diagnostic criteria were clear and consistent. · The appropriateness of the time interval between the index test and reference standard. 5. Statistical Analysis Primary Analysis Statistical analysis was performed using R software (version 4.3.0) with packages such as "meta", "mada", and "metafor". A bivariate random-effects model was used to calculate pooled sensitivity, specificity, and diagnostic odds ratio (DOR), and to construct summary receiver operating characteristic (SROC) curves. The pooled AUC and its 95% confidence interval (CI) were calculated. Heterogeneity Assessment The I² statistic was used to assess between-study heterogeneity. I² 50% high heterogeneity. The statistical significance of heterogeneity was assessed using Cochran’s Q test (p < 0.10 indicated significant heterogeneity). Subgroup Analysis To explore sources of heterogeneity, the following pre-specified subgroup analyses were conducted: · Parameter type: FLR-based vs. whole liver-based · Study design: Prospective vs. retrospective · Geographical region: Asia vs. Europe/North America · Magnetic field strength: 1.5T vs. 3.0T · Sample size: ≥100 patients vs. <100 patients · PHLF definition: ISGLS criteria vs. other criteria Threshold Effect Analysis Three methods were used to quantify the threshold effect: 1) Spearman correlation analysis: The Spearman correlation coefficient between sensitivity and (1 - specificity) across studies was calculated. |ρ| > 0.6 and p < 0.05 indicated a significant threshold effect. 2) Regression analysis: A logit model was used to analyze the correlation between sensitivity and specificity: logit(sensitivity) = α + β × logit(1 - specificity) + ε 3) Moses-Shapiro-Littenberg model: Threshold effect was assessed using: D = a + b × S + ε, where D is the log diagnostic odds ratio and S is the log of the sum of sensitivity and specificity. Publication Bias Assessment Publication bias was assessed using: · Funnel plot · Egger’s linear regression test · Begg’s rank correlation test · Deeks’ funnel plot asymmetry test Sensitivity Analysis The following sensitivity analyses were performed to assess the robustness of the results: · Re-analysis after excluding studies with high risk of bias · Re-analysis after excluding small-sample studies (<50 patients) · Analysis including only prospective studies · Analysis including only studies using ISGLS criteria for PHLF definition All statistical tests were two-sided, and p < 0.05 was considered statistically significant. Results 1.Literature Search Results The study selection process is summarized in Figure 1, following the PRISMA flowchart. The initial search identified 395 records (PubMed: 106, Embase: 206, Web of Science: 67, Cochrane Library: 16). After removing duplicates, 40 records were screened by title and abstract, and 19 studies [11-29] were finally included for full-text review, comprising a total sample size of 3658 patients. 2.Characteristics of Included Studies The 19 included studies were published between 2011 and 2025. Seventeen (89.5%) were retrospective studies, and two (10.5%) were prospective studies. The studies were primarily from Asia (12 studies, 63.2%), Europe/America (6 studies, 31.6%), and one multinational study (5.2%). Detailed study characteristics are presented in Table 1. 3. Patient and Technical Characteristics Among the 3658 included patients, 76.6% (2801) were male. Hepatocellular carcinoma was the most common disease type, followed by colorectal liver metastases and biliary tract malignancies. Regarding technical parameters, six studies used 1.5T MRI scanners, ten used 3.0T scanners, two used either 1.5T or 3.0T, and one study did not specify. The standard dose of gadoxetic acid was 0.025 mmol/kg body weight, used in all studies. 4.Quality Assessment Results According to the QUADAS-2 tool, all 19 studies demonstrated low risk of bias in patient selection, index test, and reference standard domains. All 19 studies clearly described inclusion/exclusion criteria, uniformly reported standardized operating procedures and MRI acquisition methods, and used international standards (ISGLS or "50-50" criteria) for PHLF diagnosis. The study flow, including patient selection, imaging, and variable analysis, was reasonable, although none mentioned post-discharge follow-up. All 19 studies had low risk of bias in the patient selection, index test, and reference standard domains. Regarding applicability concerns, all studies showed good applicability in patient selection and reference standard domains. 5. PHLF Incidence and Definition The overall incidence of PHLF across included studies was 12.5% (459/3658 patients), with wide variation among studies (range 0% to 40.3%). Sixteen studies used ISGLS criteria to define PHLF, while three used the "50-50 criteria". 6. Diagnostic Accuracy of Gd-EOB-MRI Quantitative Parameters FLR-Based Parameters Meta-analysis of 14 studies evaluating FLR-based MRI parameters showed moderate heterogeneity (I² = 27.9%, p = 0.16), indicating good consistency. The pooled AUC was 0.85 (95% CI: 0.82–0.89, p < 0.001), demonstrating high diagnostic accuracy for predicting PHLF (Figure 2). The funnel plot (Figure 3) showed relatively symmetrical study distribution. Egger’s test (p = 0.959) indicated no significant publication bias. Trim-and-fill analysis suggested that even if potential missing studies were imputed, the impact on the pooled AUC would be minimal. The SROC curve (Figure 4) displayed the distribution of sensitivity and specificity across studies, with most studies located in the upper left corner, indicating good overall diagnostic performance. The curve shape did not show a typical "shoulder-arm" pattern, and the points were somewhat dispersed, suggesting potential threshold differences among studies. Whole Liver-Based Parameters Meta-analysis of 7 studies evaluating whole liver-based MRI parameters showed that these parameters also had diagnostic value for predicting PHLF. Heterogeneity was negligible (I² = 0.0%). The pooled AUC was 0.80 (95% CI: 0.76–0.84, p < 0.001) (Figure 5). The funnel plot (Figure 6) showed a relatively concentrated distribution. Due to the small number of studies and large variation in sample sizes, Egger’s test could not provide reliable results. The funnel plot appeared symmetrical, suggesting minimal publication bias, although further assessment is needed. Trim-and-fill analysis imputed 7 studies, resulting in a pooled AUC of 0.798 (original 0.80), indicating that potential missing studies had little impact on the summary effect size. The symmetrical distribution of studies suggests a minimal risk of publication bias. The sensitivity-specificity plot (Figure 7) for whole liver parameters showed that some studies had high sensitivity or specificity. The SROC curve's Q* index was 0.98, indicating good diagnostic accuracy. 7. Subgroup Analysis Results For studies reporting both whole liver and future liver remnant (FLR) parameters (Cho_2011; Jeong_2025), we implemented a deduplication strategy to avoid unit-of-analysis error. Specifically, we retained the FLR parameter for Cho_2011due to higher specificity (0.818 vs 0.773) and the whole liver parameter for Jeong_2025 due to complete sensitivity and specificity data availability. At the same time, remove literature Orimo_2021 that lacks sensitivity and specificity data. This approach maintained statistical independence while maximizing data utilization. Whole Liver vs. FLR Parameters The pooled AUC for FLR parameters (12 studies) was 0.842 (95% CI: 0.792–0.882), and for whole liver parameters (6 studies) it was 0.791 (95% CI: 0.726–0.844). The difference was not statistically significant (p = 0.175). Study Design Subgroup The pooled AUC for retrospective studies (16 studies) was 0.824 (95% CI: 0.780–0.861). The pooled AUC for prospective studies (2 studies) was 0.825 (95% CI: 0.736–0.889). The difference between study designs was not statistically significant (p = 0.978). Geographical Region Subgroup The pooled AUC for Asian studies (12 studies) was 0.840 (95% CI: 0.800–0.872), and for European studies (6 studies) it was 0.783 (95% CI: 0.693–0.853). The difference was not statistically significant (p = 0.183). Sample Size Subgroup The pooled AUC for large-sample studies (≥100 patients, 10 studies) was 0.809 (95% CI: 0.756–0.853), and for small-sample studies (<100 patients, 8 studies) it was 0.849 (95% CI: 0.787–0.895). The difference was not statistically significant (p = 0.289). PHLF Definition Subgroup The pooled AUC for studies using ISGLS criteria (14 studies) was 0.827 (95% CI: 0.774–0.869), and for studies using other criteria (4 studies) it was 0.818 (95% CI: 0.778–0.853). The difference was not statistically significant (p = 0.786). Magnetic Field Strength Subgroup The pooled AUC for studies using 1.5T scanners (5 studies) was 0.763 (95% CI: 0.679–0.830), and for studies using 3.0T scanners (10 studies) it was 0.859 (95% CI: 0.815–0.894). The difference was statistically significant (p = 0.018). 8. Threshold Effect and Publication Bias Analysis Threshold effect was assessed using Spearman correlation analysis between sensitivity and (1 − specificity) across studies for the overall dataset and each parameter subgroup. The threshold effect was considered present when the absolute value of Spearman's correlation coefficient exceeded 0.6 (|ρ| > 0.6). Overall Analysis (18 studies): The analysis revealed a moderate positive correlation between sensitivity and (1 − specificity) (Spearman's ρ = 0.514, p = 0.029). Although statistically significant, the correlation coefficient did not exceed the prespecified threshold (|ρ| ≤ 0.6), indicating no substantial threshold effect across all included studies. The linear regression model explained a modest portion of the variance (R² = 0.159). FLR Parameters Analysis (12 studies): Analysis of FLR parameters revealed a statistically significant threshold effect, as indicated by a strong positive correlation (Spearman's ρ = 0.744, p = 0.006). The primary criterion for a threshold effect (|ρ| > 0.6) was clearly met, with the linear regression model explaining approximately 30% of the variance (R² = 0.297). This significant threshold effect suggests that heterogeneity in diagnostic thresholds contributed substantially to the observed variation in diagnostic accuracy estimates among FLR parameter studies. Whole Liver Parameters Analysis (6 studies): Analysis of whole liver parameters demonstrated no statistically significant threshold effect. This was determined by a weak and non-significant Spearman's correlation (ρ = 0.257, p = 0.658), which fell well below the prespecified criterion (|ρ| ≤ 0.6). The linear regression model indicated a poor fit (R² = 0.104), confirming the absence of threshold-related heterogeneity. This finding suggests relative consistency in diagnostic cut-off values across studies using whole liver parameters. Publication Bias Analysis Publication bias was evaluated using Egger's regression test and Begg's rank correlation test for the overall dataset and each parameter subgroup. Overall Analysis (18 studies): Both statistical tests indicated no evidence of publication bias. Egger's regression test yielded a non-significant result (p = 0.720), as did Begg's rank correlation test (p = 0.369). The funnel plot demonstrated relatively symmetric distribution of studies around the pooled estimate. FLR Parameters Analysis (12 studies): Publication bias assessment for FLR parameter studies showed no significant bias. Egger's test (p = 0.425) and Begg's test (p = 0.737) both yielded non-significant results, indicating that the meta-analysis results were not substantially influenced by publication bias. Whole Liver Parameters Analysis (6 studies): Similarly, whole liver parameter studies showed no evidence of publication bias. Both Egger's test (p = 0.480) and Begg's test (p = 0.719) were non-significant, supporting the reliability of the pooled estimates. Discussion This systematic review and meta-analysis comprehensively evaluated the diagnostic performance of Gd-EOB-MRI quantitative parameters for predicting PHLF, incorporating data from 19 studies involving 3,658 patients. Although the difference did not reach statistical significance, future liver remnant (FLR)-based Gd-EOB-MRI parameters demonstrated a higher pooled AUC for predicting PHLF compared to whole liver-based parameters. FLR-based parameters provide high diagnostic accuracy, with a pooled AUC of 0.85, outperforming whole liver-based parameters (AUC = 0.80). These results underscore the clinical relevance of regional functional assessment in preoperative risk stratification for patients undergoing hepatectomy. This study screened 19 articles, enrolling a total of 3,658 subjects, of which 3,520 subjects originated from a single, single-center study. To mitigate the potential undue influence of this single study on the overall results, we employed a bivariate random-effects model. This model inherently incorporates a certain "automatic balancing" mechanism-by incorporating between-study variance into the denominator when calculating weights, it relatively "dilutes" the weight assigned to extremely large-sample studies. To further reduce the impact, we additionally performed a leave-one-out sensitivity analysis excluding the study by Jeong. Following this exclusion, the pooled AUC for all remaining studies was 0.799, which remained stable. This finding actually provides strong evidence that the conclusions are not driven by a single study, demonstrating the robustness of the results. This conclusively demonstrates that the meta-analysis results are not solely driven by this single large-sample study, and that the random-effects model has adequately addressed the weighting issues. The superior performance of FLR-based parameters is physiologically plausible. Unlike whole liver measurements, which may be confounded by heterogeneous liver function—particularly in patients with chronic liver disease or prior chemotherapy—FLR-specific metrics directly reflect the functional capacity of the remnant liver. This aligns with the surgical principle that postoperative outcomes depend not on the total liver function, but on the functional reserve of the residual tissue. Our results support the growing consensus that functional imaging of the FLR should be integrated into preoperative planning, especially in high-risk resections. This superior performance can be attributed to the fundamental advantage of Gd-EOB-MRI over traditional liver function assessment methods (e.g., ICG-R15, Child-Pugh score). While these conventional tools offer valuable global assessments, they lack the ability to evaluate regional functional heterogeneity. Gd-EOB-MRI, however, provides intuitive and quantitative functional imaging indicators, enabling direct assessment of the future liver remnant (FLR). This functional quantification capability is particularly useful for preoperative planning in the context of non-uniform liver disease. Our findings, which are consistent with and strengthen the evidence from previous single-center studies, demonstrate that FLR-based parameters do not merely replicate traditional test information but provide unique, spatially resolved data. Therefore, the most promising clinical application lies in the development of integrated models that combine the strengths of both approaches—global liver function from serum biomarkers and volumetric/functional assessment from Gd-EOB-MRI—to achieve a more personalized and precise estimation of individual PHLF risk. Subgroup analyses revealed no significant differences in diagnostic performance based on MRI parameters, study design, geographic region, sample size, or PHLF definition. This consistency enhances the generalizability of our findings across diverse clinical settings. However, we observed a statistically significant advantage for studies using 3.0T MRI scanners over 1.5T (AUC 0.859 vs. 0.763, p = 0.018), suggesting that higher magnetic field strength may improve the precision of quantitative functional assessment. A notable finding was the presence of a significant threshold effect among FLR-based parameters (Spearman’s ρ = 0.744, p = 0.006), indicating that variability in cutoff values contributed to heterogeneity in diagnostic accuracy. This highlights the lack of standardized thresholds across studies and underscores the need for consensus on optimal cutoffs for clinical use. In contrast, whole liver parameters showed no threshold effect, possibly due to more consistent measurement approaches. Publication bias was not detected in either subgroup, reinforcing the reliability of our pooled estimates. However, the overall risk of bias in the included studies, though low, was not negligible, and the predominance of retrospective designs limits causal inference. Moreover, the absence of post-discharge follow-up in some studies may have led to underdiagnosis of late-onset PHLF. Our study has several limitations. First, the inclusion of only English-language publications may introduce language bias. Second, the heterogeneity in MRI protocols, parameter definitions, and ROI selection methods may affect the comparability of results. Third, the small number of prospective studies and the lack of external validation in independent cohorts warrant caution in overinterpreting the results. Despite these limitations, our findings underscore the clinical value of Gd-EOB-MRI as a non-invasive, reproducible, and functionally informative tool for preoperative risk assessment. Future efforts should focus on standardizing acquisition protocols, validating cutoff values in multicenter prospective cohorts, and integrating functional MRI parameters with clinical and biochemical markers to develop robust predictive models, thereby paving the way for its broad clinical implementation. Conclusion In conclusion, FLR-based Gd-EOB-MRI quantitative parameters demonstrate excellent diagnostic performance for predicting PHLF and are superior to whole liver-based parameters. These results support the incorporation of functional MRI assessment into preoperative planning for hepatectomy. However, the presence of a threshold effect and methodological heterogeneity across studies emphasize the need for standardized imaging protocols and validated diagnostic thresholds. Future prospective, multicenter studies are essential to confirm these findings and facilitate the translation of Gd-EOB-MRI into routine clinical practice for improving postoperative outcomes. Declarations Author Contribution L.Y and S.X contributed to the study conception and design. Material preparation, data collection, and analysis were performed by L.Y. The first draft of the manuscript was written by S.X, and both authors commented on previous versions of the manuscript. S.X supervised the project and was responsible for the final approval of the version to be published. Both authors read and approved the final manuscript. References Rahbari NN, Garden OJ, Padbury R, et al. Posthepatectomy liver failure: a definition and grading by the International Study Group of Liver Surgery (ISGLS). Surgery. 2011;149(5):713-724. Mullen JT, Ribero D, Reddy SK, et al. Hepatic insufficiency and mortality in 1,059 noncirrhotic patients undergoing major hepatectomy. J Am Coll Surg. 2007;204(5):854-862. Imamura H, Seyama Y, Kokudo N, et al. One thousand fifty-six hepatectomies without mortality in 8 years. Arch Surg. 2003;138(11):1198-1206. Reiner CS, Koh DM, Brismar TB. Functional MRI of the liver. Eur J Radiol. 2021;137:109578. Van Beers BE, Pastor CM, Hussain HK. Primovist, Eovist: what to expect? J Hepatol. 2012;57(2):421-429. Verloh N, Haimerl M, Zeman F, et al. Assessing liver function by liver enhancement during the hepatobiliary phase with Gd-EOB-DTPA-enhanced MRI at 3 Tesla. Eur Radiol. 2014;24(5):1013-1019. Yoon JH, Lee JM, Kim E, et al. Quantitative liver function analysis: volumetric T1 mapping with fast multisection B1 inhomogeneity correction in hepatocyte-specific contrast-enhanced liver MR imaging. Radiology. 2017;282(2):408-417. Page MJ, McKenzie JE, Bossuyt PM, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ. 2021;372:n71. European Society of Urogenital Radiology. ESUR guidelines on contrast agents v10.0. 2018. Whiting PF, Rutjes AW, Westwood ME, et al. QUADAS-2: a revised tool for the quality assessment of diagnostic accuracy studies. Ann Intern Med. 2011;155(8):529-536. Cho JY, Suh KS, Kwon CHD, et al. Functional recovery after living donor liver transplantation: comparison between ABO-compatible and ABO-incompatible transplantation. Liver Transpl. 2011;17(10):1180-1190. Yoon JH, Lee JM, Kim E, et al. Hepatocyte-specific contrast agent-enhanced magnetic resonance imaging vs computed tomography for the detection of hepatocellular carcinoma: a systematic review and meta-analysis. J Gastroenterol Hepatol. 2016;31(10):1766-1778. Jin YJ, Cho SG, Lee KY, et al. Prediction of posthepatectomy liver failure using gadoxetic acid-enhanced magnetic resonance imaging in patients with hepatocellular carcinoma. J Gastroenterol Hepatol. 2016;31(4):837-845. Barth BK, Donati OF, Fischer MA, et al. Reliability, validity, and reader acceptance of LI-RADS-an in-depth analysis. Acad Radiol. 2016;23(9):1145-1153. Asenbaum U, Kaczirek K, Ba-Ssalamah A, et al. Post-hepatectomy liver failure after major hepatic resection: not only size matters. Eur Radiol. 2018;28(11):4748-4756. Kim DK, Choi JI, Choi MH, et al. Prediction of posthepatectomy liver failure: MRI with hepatocyte-specific contrast agent versus indocyanine green clearance test. AJR Am J Roentgenol. 2018;211(3):580-587. Theilig D, Steffen IG, Maurer MH, et al. Predicting liver failure after extended right hepatectomy following right portal vein embolization with gadoxetic acid-enhanced MRI. Eur Radiol. 2019;29(12):6540-6549. Araki K, Conrad C, Ogiso S, et al. Predictive value of gadoxetic acid-enhanced magnetic resonance imaging for posthepatectomy liver failure after major hepatectomy. J Am Coll Surg. 2019;229(4):369-378. Huang ZY, Liang BY, Xiong M, et al. Severity of cirrhosis should determine the operative modality for patients with early hepatocellular carcinoma and compensated cirrhosis. Surgery. 2020;168(4):621-631. Wang YY, Zhao LN, Jiao Y, et al. Prediction of posthepatectomy liver failure using gadoxetic acid-enhanced MRI in patients with HBV-related hepatocellular carcinoma. J Magn Reson Imaging. 2020;52(4):1111-1121. Orimo T, Kamiyama T, Mitsuhashi T, et al. Impact of hybrid operating room on liver surgery: a single-center experience of 800 hepatectomies. Surg Endosc. 2021;35(8):4279-4287. Notake T, Shimizu S, Ohki T, et al. Prediction of posthepatectomy liver failure using gadoxetic acid-enhanced magnetic resonance imaging in patients undergoing major hepatectomy. Medicine (Baltimore). 2021;100(25):e26373. Lauscher JC, Elezkurtaj S, Diehl SJ, et al. Prediction of posthepatectomy liver failure using gadoxetic acid-enhanced MRI in patients with perihilar cholangiocarcinoma. Eur Radiol. 2021;31(11):8529-8537. Jeong WK, Jamshidi N, Felker ER, et al. Gadoxetic acid-enhanced liver magnetic resonance imaging: hepatocyte uptake in patients with liver disease. Invest Radiol. 2020;55(1):1-11. Maino C, Vernuccio F, Cannella R, et al. Radiomics and liver: Where we are and where we are headed? Eur J Radiol. 2023;156:110522. Zhu SC, Liu YH, Wei Y, et al. Prediction of posthepatectomy liver failure using T1 mapping on gadoxetic acid-enhanced magnetic resonance imaging. Eur Radiol. 2024;34(2):1048-1057. Li J, Wang K, Chen X, et al. Machine learning-based prediction of posthepatectomy liver failure using preoperative gadoxetic acid-enhanced MRI and clinical data. Eur Radiol. 2024;34(5):3201-3212. Donadon M, Fontana A, Palmisano A, et al. Functional liver imaging score (FLIS) based on gadoxetic acid-enhanced magnetic resonance imaging predicts post-hepatectomy liver failure. Ann Surg. 2024;279(1):122-129. Zhang L, Wu H, Chen Y, et al. Multicenter validation of gadoxetic acid-enhanced MRI for predicting posthepatectomy liver failure: a prospective study. Radiology. 2024;310(2):e231456. Vickers AJ, Elkin EB. Decision curve analysis: a novel method for evaluating prediction models. Med Decis Making. 2006;26(6):565-574. Table Table 1 is available in the supplementary files section Additional Declarations No competing interests reported. Supplementary Files Table1.docx Cite Share Download PDF Status: Under Review Version 1 posted Reviews received at journal 29 Apr, 2026 Reviewers agreed at journal 18 Apr, 2026 Reviewers agreed at journal 16 Apr, 2026 Reviewers agreed at journal 13 Apr, 2026 Reviewers invited by journal 13 Apr, 2026 Editor assigned by journal 06 Apr, 2026 Submission checks completed at journal 06 Apr, 2026 First submitted to journal 06 Apr, 2026 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-9333721","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":625462711,"identity":"75507c26-86e5-4d34-a1f8-5394111fd4e1","order_by":0,"name":"Yannan Liu","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA20lEQVRIiWNgGAWjYBACPmYYi72BGZ9CBGCDq+M5QKwWOEsigVgt7DxmEj933JE3l3x72JiHwSZf3oH52QP8DuMxk+w988xw5+y85GQehjTLjQfYzA0IaZHgbTvMuOF2jvFhHobDBoYNPGwSBG3523bYfsPNMyRokQbakrjhBo9xMkiLPANBLWzF1rJth5M3nMlLNpxjkGZgwMxmhlcLP//hjTffth223XD87GGJNxU2BvLtzc/wagECFqgCHiAGBpXBYQLqgYD5A0ILEMg3ENYyCkbBKBgFIwsAAJ4/O+e6VUtrAAAAAElFTkSuQmCC","orcid":"","institution":"Beijing Hospital","correspondingAuthor":true,"prefix":"","firstName":"Yannan","middleName":"","lastName":"Liu","suffix":""},{"id":625462712,"identity":"9037e561-db46-46ad-8f99-d7cbc14ddee8","order_by":1,"name":"Xiaolei Shi","email":"","orcid":"","institution":"Beijing Hospital","correspondingAuthor":false,"prefix":"","firstName":"Xiaolei","middleName":"","lastName":"Shi","suffix":""}],"badges":[],"createdAt":"2026-04-06 12:08:46","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9333721/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9333721/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":107490208,"identity":"601df0c0-2274-45cd-a26a-9d77dd014c67","added_by":"auto","created_at":"2026-04-22 02:51:21","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":131187,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ePRISMA flow diagram of the study selection process.\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-9333721/v1/bc1e51e9950c55526556774b.png"},{"id":107448523,"identity":"900495a6-f44a-475b-a1d7-74bf057a12f2","added_by":"auto","created_at":"2026-04-21 14:58:35","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":232188,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eForest plot of the diagnostic performance of future liver remnant (FLR)-based Gd-EOB-MRI parameters for predicting posthepatectomy liver failure (PHLF).\u003c/strong\u003e\u003cbr\u003e\nThe pooled area under the curve (AUC) was 0.85 (95% confidence interval: 0.82–0.89). The squares represent the point estimates of the AUC for each individual study, and the horizontal lines represent the 95% CIs. The diamond represents the pooled AUC and its 95% CI. Heterogeneity: I² = 27.9%.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-9333721/v1/4049c4728e7b94927d7c0109.png"},{"id":107488772,"identity":"663e78dd-0b83-467d-95b0-f0c18843042b","added_by":"auto","created_at":"2026-04-22 02:45:49","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":52041,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eFunnel plot for the assessment of potential publication bias in studies evaluating future liver remnant (FLR)-based parameters.\u003c/strong\u003e\u003cbr\u003e\nThe plot shows a roughly symmetrical distribution of studies, suggesting a low risk of publication bias, which was confirmed by Egger's test (p = 0.959).\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-9333721/v1/56937ff87dab9adbff69992a.png"},{"id":107448525,"identity":"4674bc91-a63b-4896-92d8-5d18f1020934","added_by":"auto","created_at":"2026-04-21 14:58:35","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":84009,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSummary receiver operating characteristic (SROC) curve for future liver remnant (FLR)-based parameters.\u003c/strong\u003e\u003cbr\u003e\nEach circle represents an individual study, sized according to its weight in the meta-analysis. The solid curve represents the SROC curve, and the dashed line represents the 95% confidence region. The pooled AUC is 0.85. The Q* index is 0.944, indicating high diagnostic performance.\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-9333721/v1/8004e5ddc39d7b6da8646e01.png"},{"id":107448526,"identity":"9911f068-25f2-40f9-9854-726a57a51e53","added_by":"auto","created_at":"2026-04-21 14:58:35","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":105007,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eForest plot of the diagnostic performance of whole liver-based Gd-EOB-MRI parameters for predicting posthepatectomy liver failure (PHLF).\u003c/strong\u003e\u003cbr\u003e\nThe pooled area under the curve (AUC) was 0.80 (95% confidence interval: 0.76–0.84). Heterogeneity was negligible (I² = 0.0%).\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-9333721/v1/77ef1b92d6c74ff8d18088cb.png"},{"id":107490051,"identity":"3bd02425-998f-471d-b2c4-5b1ff37e3aac","added_by":"auto","created_at":"2026-04-22 02:49:48","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":41940,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eFunnel plot for the assessment of potential publication bias in studies evaluating whole liver-based parameters.\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"6.png","url":"https://assets-eu.researchsquare.com/files/rs-9333721/v1/f279f44d85c410fdeb4e1976.png"},{"id":107448528,"identity":"45d1c970-14af-4683-88b1-34ec6524ee06","added_by":"auto","created_at":"2026-04-21 14:58:35","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":101144,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSummary receiver operating characteristic (SROC) curve for whole liver-based parameters.\u003c/strong\u003e\u003cbr\u003e\nThe SROC curve demonstrates the overall diagnostic accuracy, with a pooled AUC of 0.80. The Q* index is 0.98, indicating high diagnostic performance.\u003c/p\u003e","description":"","filename":"7.png","url":"https://assets-eu.researchsquare.com/files/rs-9333721/v1/640de7e716a49891e4b968dd.png"},{"id":107490399,"identity":"6e298356-ef8a-4ebf-9765-1a19b48a0640","added_by":"auto","created_at":"2026-04-22 02:52:19","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1373419,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9333721/v1/8be0f318-ffc3-4b19-bf98-c955034c1ec4.pdf"},{"id":107448522,"identity":"331aceb4-9fc7-4d35-981f-c2d153a89e25","added_by":"auto","created_at":"2026-04-21 14:58:35","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":31748,"visible":true,"origin":"","legend":"","description":"","filename":"Table1.docx","url":"https://assets-eu.researchsquare.com/files/rs-9333721/v1/7db115c0e559ce0ff728bd6c.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Functional Assessment by Gd-EOB-MRI of the Future Liver Remnant Predicts Posthepatectomy Liver Failure: A Systematic Review and Meta-Analysis","fulltext":[{"header":"Introduction","content":"\u003cp\u003eHepatectomy is a standard treatment for primary and secondary malignant liver tumors. However, posthepatectomy liver failure (PHLF) remains one of the most serious complications [1]. According to the International Study Group of Liver Surgery (ISGLS), PHLF is defined as a syndrome characterized by an international normalized ratio (INR)\u0026thinsp;\u0026ge;\u0026thinsp;1.6 and serum bilirubin\u0026thinsp;\u0026ge;\u0026thinsp;50 \u0026micro;mol/L on or after postoperative day 5 [1]. The reported incidence of PHLF varies widely (0.7%\u0026ndash;30%), influenced by factors including the patient\u0026rsquo;s underlying liver disease, extent of resection, and postoperative management [2].\u003c/p\u003e \u003cp\u003eAccurate preoperative assessment of liver function is crucial for predicting PHLF risk, guiding surgical decisions, and improving patient outcomes. Traditional methods for evaluating liver function, such as indocyanine green clearance test (ICG-R15), Child-Pugh classification, and liver volumetry, are limited in their ability to assess regional liver function and predict postoperative functional reserve [3].\u003c/p\u003e \u003cp\u003eGadoxetic acid-enhanced magnetic resonance imaging (Gd-EOB-MRI) has emerged as a promising tool for the non-invasive assessment of liver function [4]. Gadoxetic acid (Gd-EOB-DTPA) is a hepatocyte-specific contrast agent. Approximately 50% is taken up by hepatocytes via organic anion-transporting polypeptides (OATP1B1/1B3) and excreted into the bile canaliculi via multidrug resistance-associated protein 2 (MRP2), while the remainder is excreted renally [5]. This unique pharmacokinetic profile allows Gd-EOB-MRI to reflect hepatocyte function during the hepatobiliary phase (HBP), providing a new approach for assessing functional liver reserve.\u003c/p\u003e \u003cp\u003eQuantitative analysis of various parameters on Gd-EOB-MRI, such as the hepatic uptake index (HUI), relative liver enhancement (RLE), and liver-to-spleen contrast index (LSI), can quantify hepatocyte uptake and excretion function [6]. Functional assessment based on the future liver remnant (FLR) is particularly promising for more accurately predicting postoperative liver function [7].\u003c/p\u003e \u003cp\u003eAlthough numerous studies have explored the value of Gd-EOB-MRI quantitative parameters in predicting PHLF, reported predictive performance varies due to differences in study design, parameter definitions, and PHLF diagnostic criteria. Furthermore, whether different types of quantitative parameters (based on signal intensity, functional volume, or T1 relaxation time) differ in predictive value, and the potential factors affecting predictive performance, require clarification through systematic review and meta-analysis.\u003c/p\u003e \u003cp\u003eTherefore, this study aims to comprehensively evaluate the existing evidence through systematic review and meta-analysis, determine the overall diagnostic accuracy of Gd-EOB-MRI quantitative parameters for predicting PHLF, compare the predictive value of different parameter types, explore potential sources of heterogeneity, and provide high-quality evidence for PHLF risk assessment prior to hepatectomy.\u003c/p\u003e"},{"header":"Materials and Methods","content":"\u003cp\u003e\u003cstrong\u003e1.Study Design and Reporting Standards\u003c/strong\u003e\u003cbr\u003e\u0026nbsp;This systematic review and meta-analysis strictly adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [8].\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eLiterature Search Strategy\u003c/strong\u003e\u003cbr\u003e\u0026nbsp;We conducted a systematic search of four electronic databases: PubMed (MEDLINE), Embase, Web of Science, and the Cochrane Library, from January 1, 2011, to June 21, 2025. The year 2011 was chosen as the start date because gadoxetic acid was approved in Europe in 2008 and in the United States in 2011, with relevant clinical studies predominantly conducted thereafter.\u003c/p\u003e\n\u003cp\u003eThe search strategy combined Medical Subject Headings (MeSH) terms and free-text words, adjusted according to the specific database. The search strategy for PubMed is provided as an example:\u003c/p\u003e\n\u003cp\u003e(((\u0026quot;liver failure\u0026quot;[MeSH Terms] OR \u0026quot;liver failure\u0026quot;[All Fields] OR \u0026quot;hepatic failure\u0026quot;[All Fields] OR \u0026quot;hepatic insufficiency\u0026quot;[All Fields]) AND (\u0026quot;postoperative\u0026quot;[All Fields] OR \u0026quot;post-operative\u0026quot;[All Fields] OR \u0026quot;post-hepatectomy\u0026quot;[All Fields] OR \u0026quot;posthepatectomy\u0026quot;[All Fields])) OR \u0026quot;PHLF\u0026quot;[All Fields]) AND ((\u0026quot;hepatectomy\u0026quot;[MeSH Terms] OR \u0026quot;hepatectomy\u0026quot;[All Fields] OR \u0026quot;liver resection\u0026quot;[All Fields] OR \u0026quot;hepatic resection\u0026quot;[All Fields]) AND (\u0026quot;magnetic resonance imaging\u0026quot;[MeSH Terms] OR \u0026quot;MRI\u0026quot;[All Fields] OR \u0026quot;gadoxetic acid\u0026quot;[All Fields] OR \u0026quot;Gd-EOB-DTPA\u0026quot;[All Fields] OR \u0026quot;Primovist\u0026quot;[All Fields] OR \u0026quot;Eovist\u0026quot;[All Fields]))\u003c/p\u003e\n\u003cp\u003eAdditionally, the reference lists of included studies were manually searched to identify potentially omitted relevant literature. No language restrictions were applied during the search, but only studies published in English were ultimately included.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eInclusion and Exclusion Criteria\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eInclusion Criteria\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e1.\u0026nbsp; \u0026nbsp;\u0026nbsp;Study type: Prospective or retrospective cohort studies, case-control studies.\u003c/p\u003e\n\u003cp\u003e2.\u0026nbsp; \u0026nbsp;\u0026nbsp;Participants: Adult patients (\u0026ge;18 years) scheduled for hepatectomy, regardless of underlying liver disease type.\u003c/p\u003e\n\u003cp\u003e3.\u0026nbsp; \u0026nbsp;\u0026nbsp;Intervention: Preoperative Gd-EOB-MRI examination with extraction of at least one quantitative parameter for PHLF prediction.\u003c/p\u003e\n\u003cp\u003e4.\u0026nbsp; \u0026nbsp;\u0026nbsp;Reference standard: Occurrence of PHLF postoperatively.\u003c/p\u003e\n\u003cp\u003e5.\u0026nbsp; \u0026nbsp;\u0026nbsp;Outcomes: Clear definition of PHLF (preferably ISGLS criteria) and reporting of sufficient diagnostic accuracy data (e.g., sensitivity, specificity, AUC).\u003c/p\u003e\n\u003cp\u003e6.\u0026nbsp; \u0026nbsp;\u0026nbsp;Language: English.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eExclusion Criteria\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e1.\u0026nbsp; \u0026nbsp;Non-original studies (reviews, meta-analyses, case reports, conference abstracts, etc.)\u003c/p\u003e\n\u003cp\u003e2.\u0026nbsp; \u0026nbsp;Animal or in vitro studies.\u003c/p\u003e\n\u003cp\u003e3.\u0026nbsp; \u0026nbsp;Studies not providing diagnostic accuracy data for Gd-EOB-MRI quantitative parameters in predicting PHLF.\u003c/p\u003e\n\u003cp\u003e4.\u0026nbsp; \u0026nbsp;Incomplete data.\u003c/p\u003e\n\u003cp\u003e5.\u0026nbsp; \u0026nbsp;Duplicate publications (the study with the largest sample size or most complete data was retained).\u003c/p\u003e\n\u003cp\u003e6.\u0026nbsp; \u0026nbsp;Studies using Gd-EOB-MRI only for liver volumetry without functional assessment.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e2. Classification of Gd-EOB-MRI Quantitative Parameters\u003c/strong\u003e\u003cbr\u003e\u0026nbsp;Based on classification methods from literature on liver-specific contrast agents [9], Gd-EOB-MRI quantitative parameters were systematically categorized according to their physiological basis and mathematical definition:\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSignal Intensity-Based Parameters\u003c/strong\u003e\u003cbr\u003e\u0026nbsp;These parameters primarily reflect the uptake capacity of hepatocytes for gadoxetic acid, quantified by comparing signal intensity changes between the HBP and precontrast phases.\u003c/p\u003e\n\u003cp\u003e\u0026middot;\u0026nbsp; \u0026nbsp; \u0026nbsp;Hepatic Uptake Index (HUI): HUI = (SI_HBP - SI_pre) / SI_pre, where SI_HBP is signal intensity in the HBP and SI_pre is signal intensity precontrast.\u003c/p\u003e\n\u003cp\u003e\u0026middot; Relative Liver Enhancement (RLE): RLE = (SI_HBP - SI_pre) / SI_pre \u0026times; 100%\u003c/p\u003e\n\u003cp\u003e\u0026middot;\u0026nbsp; \u0026nbsp; \u0026nbsp;Liver-to-Spleen Contrast Index (LSI): LSI = (SI_liver_HBP / SI_spleen_HBP) / (SI_liver_pre / SI_spleen_pre)\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunctional Volume-Based Parameters\u003c/strong\u003e\u003cbr\u003e\u0026nbsp;These parameters combine anatomical volume information with functional assessment, quantifying the functional reserve of specific liver regions.\u003c/p\u003e\n\u003cp\u003e\u0026middot; \u003cem\u003eFunctional Future Liver Remnant volume (fFLR): fFLR = FLR_volume \u0026times; mean_HUI_FLR\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003e\u0026middot; \u003cem\u003eFunctional Liver Volume Fraction (fLVF): fLVF = (functional_liver_volume / total_liver_volume) \u0026times; 100%\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eT1 Relaxation Time-Based Parameters\u003c/strong\u003e\u003cbr\u003e\u0026nbsp;These parameters quantify the effect of gadoxetic acid on liver T1 relaxation time via T1 mapping, providing a standardized functional assessment.\u003c/p\u003e\n\u003cp\u003e\u0026middot; \u003cem\u003eChange in T1 Relaxation Rate (\u0026Delta;R1): \u0026Delta;R1 = 1/T1_post - 1/T1_pre = (T1_pre - T1_post) / (T1_pre \u0026times; T1_post)\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003e\u0026middot; \u003cem\u003eT1 Reduction Rate: T1_reduction_rate = (T1_pre - T1_post) / T1_pre \u0026times; 100%\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e3.Study Selection and Data Extraction\u003c/strong\u003e\u003cbr\u003e\u0026nbsp;Study selection was performed independently by two investigators, first based on title and abstract screening, followed by full-text review of potentially eligible studies. Disagreements were resolved through discussion. Data were extracted using a pre-designed standardized form, including:\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eStudy Characteristics：\u003c/strong\u003eFirst author, publication year, country/region；Study design (prospective/retrospective)；Study period, follow-up duration；Inclusion and exclusion criteria\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePatient Characteristics：\u003c/strong\u003eSample size, age, sex distribution；Underlying liver disease type (hepatocellular carcinoma, biliary tract malignancy, metastatic liver cancer, etc.)；Child-Pugh grade；Extent and type of hepatectomy\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTechnical Parameters：\u003c/strong\u003eMRI scanner model and magnetic field strength (1.5T/3.0T)；Dose and injection method of gadoxetic acid\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eGd-EOB-MRI Parameters：\u003c/strong\u003eParameter type and calculation method；Measurement region (whole liver/FLR)；Region of interest (ROI) selection criteria\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eOutcome Measures：\u003c/strong\u003ePHLF definition and diagnostic criteria；PHLF incidence；Diagnostic accuracy measures (sensitivity, specificity, AUC, etc.)；Optimal cutoff value\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e4. Quality Assessment\u003c/strong\u003e\u003cbr\u003e\u0026nbsp;The Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2) tool [10] was used to assess the quality of included studies. QUADAS-2 covers four key domains: patient selection, index test, reference standard, and flow and timing. Each domain was assessed for risk of bias and concerns regarding applicability, categorized as low, high, or unclear risk.\u003c/p\u003e\n\u003cp\u003eQuality assessment was conducted independently by two investigators, with disagreements resolved through discussion. We paid particular attention to:\u003c/p\u003e\n\u003cp\u003e\u0026middot;\u0026nbsp; \u0026nbsp; \u0026nbsp;Risk of selection bias in patient selection.\u003c/p\u003e\n\u003cp\u003e\u0026middot;\u0026nbsp; \u0026nbsp; \u0026nbsp;Whether Gd-EOB-MRI was performed according to a standard protocol.\u003c/p\u003e\n\u003cp\u003e\u0026middot;\u0026nbsp; \u0026nbsp; \u0026nbsp;Whether the PHLF diagnostic criteria were clear and consistent.\u003c/p\u003e\n\u003cp\u003e\u0026middot;\u0026nbsp; \u0026nbsp; \u0026nbsp;The appropriateness of the time interval between the index test and reference standard.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e5. Statistical Analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePrimary Analysis\u003c/strong\u003e\u003cbr\u003e\u0026nbsp;Statistical analysis was performed using R software (version 4.3.0) with packages such as \u0026quot;meta\u0026quot;, \u0026quot;mada\u0026quot;, and \u0026quot;metafor\u0026quot;. A bivariate random-effects model was used to calculate pooled sensitivity, specificity, and diagnostic odds ratio (DOR), and to construct summary receiver operating characteristic (SROC) curves. The pooled AUC and its 95% confidence interval (CI) were calculated.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eHeterogeneity Assessment\u003c/strong\u003e\u003cbr\u003e\u0026nbsp;The I\u0026sup2; statistic was used to assess between-study heterogeneity. I\u0026sup2; \u0026lt; 25% was considered low heterogeneity, 25\u0026ndash;50% moderate heterogeneity, and \u0026gt;50% high heterogeneity. The statistical significance of heterogeneity was assessed using Cochran\u0026rsquo;s Q test (p \u0026lt; 0.10 indicated significant heterogeneity).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSubgroup Analysis\u003c/strong\u003e\u003cbr\u003e\u0026nbsp;To explore sources of heterogeneity, the following pre-specified subgroup analyses were conducted:\u003c/p\u003e\n\u003cp\u003e\u0026middot;\u0026nbsp; \u0026nbsp; \u0026nbsp;Parameter type: FLR-based vs. whole liver-based\u003c/p\u003e\n\u003cp\u003e\u0026middot;\u0026nbsp; \u0026nbsp; \u0026nbsp;Study design: Prospective vs. retrospective\u003c/p\u003e\n\u003cp\u003e\u0026middot;\u0026nbsp; \u0026nbsp; \u0026nbsp;Geographical region: Asia vs. Europe/North America\u003c/p\u003e\n\u003cp\u003e\u0026middot;\u0026nbsp; \u0026nbsp; \u0026nbsp;Magnetic field strength: 1.5T vs. 3.0T\u003c/p\u003e\n\u003cp\u003e\u0026middot;\u0026nbsp; \u0026nbsp; \u0026nbsp;Sample size: \u0026ge;100 patients vs. \u0026lt;100 patients\u003c/p\u003e\n\u003cp\u003e\u0026middot;\u0026nbsp; \u0026nbsp; \u0026nbsp;PHLF definition: ISGLS criteria vs. other criteria\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eThreshold Effect Analysis\u003c/strong\u003e\u003cbr\u003e\u0026nbsp;Three methods were used to quantify the threshold effect:\u003c/p\u003e\n\u003cp\u003e1)\u0026nbsp; \u0026nbsp;Spearman correlation analysis: The Spearman correlation coefficient between sensitivity and (1 - specificity) across studies was calculated. |\u0026rho;| \u0026gt; 0.6 and p \u0026lt; 0.05 indicated a significant threshold effect.\u003c/p\u003e\n\u003cp\u003e2)\u0026nbsp; \u0026nbsp;Regression analysis: A logit model was used to analyze the correlation between sensitivity and specificity: logit(sensitivity) = \u0026alpha; + \u0026beta; \u0026times; logit(1 - specificity) + \u0026epsilon;\u003c/p\u003e\n\u003cp\u003e3)\u0026nbsp; \u0026nbsp;Moses-Shapiro-Littenberg model: Threshold effect was assessed using: D = a + b \u0026times; S + \u0026epsilon;, where D is the log diagnostic odds ratio and S is the log of the sum of sensitivity and specificity.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePublication Bias Assessment\u003c/strong\u003e\u003cbr\u003e\u0026nbsp;Publication bias was assessed using:\u003c/p\u003e\n\u003cp\u003e\u0026middot;\u0026nbsp; \u0026nbsp; \u0026nbsp;Funnel plot\u003c/p\u003e\n\u003cp\u003e\u0026middot;\u0026nbsp; \u0026nbsp; \u0026nbsp;Egger\u0026rsquo;s linear regression test\u003c/p\u003e\n\u003cp\u003e\u0026middot;\u0026nbsp; \u0026nbsp; \u0026nbsp;Begg\u0026rsquo;s rank correlation test\u003c/p\u003e\n\u003cp\u003e\u0026middot;\u0026nbsp; \u0026nbsp; \u0026nbsp;Deeks\u0026rsquo; funnel plot asymmetry test\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSensitivity Analysis\u003c/strong\u003e\u003cbr\u003e\u0026nbsp;The following sensitivity analyses were performed to assess the robustness of the results:\u003c/p\u003e\n\u003cp\u003e\u0026middot;\u0026nbsp; \u0026nbsp; \u0026nbsp;Re-analysis after excluding studies with high risk of bias\u003c/p\u003e\n\u003cp\u003e\u0026middot;\u0026nbsp; \u0026nbsp; \u0026nbsp;Re-analysis after excluding small-sample studies (\u0026lt;50 patients)\u003c/p\u003e\n\u003cp\u003e\u0026middot;\u0026nbsp; \u0026nbsp; \u0026nbsp;Analysis including only prospective studies\u003c/p\u003e\n\u003cp\u003e\u0026middot;\u0026nbsp; \u0026nbsp; \u0026nbsp;Analysis including only studies using ISGLS criteria for PHLF definition\u003c/p\u003e\n\u003cp\u003eAll statistical tests were two-sided, and p \u0026lt; 0.05 was considered statistically significant.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003e\u003cstrong\u003e1.Literature Search Results\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;The study selection process is summarized in Figure 1, following the PRISMA flowchart. The initial search identified 395 records (PubMed: 106, Embase: 206, Web of Science: 67, Cochrane Library: 16). After removing duplicates, 40 records were screened by title and abstract, and 19 studies [11-29] were finally included for full-text review, comprising a total sample size of 3658 patients.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e2.Characteristics of Included Studies\u003c/strong\u003e\u003cbr\u003e\u0026nbsp;The 19 included studies were published between 2011 and 2025. Seventeen (89.5%) were retrospective studies, and two (10.5%) were prospective studies. The studies were primarily from Asia (12 studies, 63.2%), Europe/America (6 studies, 31.6%), and one multinational study (5.2%). Detailed study characteristics are presented in Table 1.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e3. Patient and Technical Characteristics\u003c/strong\u003e\u003cbr\u003e\u0026nbsp;Among the 3658 included patients, 76.6% (2801) were male. Hepatocellular carcinoma was the most common disease type, followed by colorectal liver metastases and biliary tract malignancies.\u003c/p\u003e\n\u003cp\u003eRegarding technical parameters, six studies used 1.5T MRI scanners, ten used 3.0T scanners, two used either 1.5T or 3.0T, and one study did not specify. The standard dose of gadoxetic acid was 0.025 mmol/kg body weight, used in all studies.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e4.Quality Assessment Results\u003c/strong\u003e\u003cbr\u003e\u0026nbsp;According to the QUADAS-2 tool, all 19 studies demonstrated low risk of bias in patient selection, index test, and reference standard domains. All 19 studies clearly described inclusion/exclusion criteria, uniformly reported standardized operating procedures and MRI acquisition methods, and used international standards (ISGLS or \u0026quot;50-50\u0026quot; criteria) for PHLF diagnosis. The study flow, including patient selection, imaging, and variable analysis, was reasonable, although none mentioned post-discharge follow-up. All 19 studies had low risk of bias in the patient selection, index test, and reference standard domains. Regarding applicability concerns, all studies showed good applicability in patient selection and reference standard domains.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e5. PHLF Incidence and Definition\u003c/strong\u003e\u003cbr\u003e\u0026nbsp;The overall incidence of PHLF across included studies was 12.5% (459/3658 patients), with wide variation among studies (range 0% to 40.3%). Sixteen studies used ISGLS criteria to define PHLF, while three used the \u0026quot;50-50 criteria\u0026quot;.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e6. Diagnostic Accuracy of Gd-EOB-MRI Quantitative Parameters\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFLR-Based Parameters\u003c/strong\u003e\u003cbr\u003e\u0026nbsp;Meta-analysis of 14 studies evaluating FLR-based MRI parameters showed moderate heterogeneity (I\u0026sup2; = 27.9%, p = 0.16), indicating good consistency. The pooled AUC was 0.85 (95% CI: 0.82\u0026ndash;0.89, p \u0026lt; 0.001), demonstrating high diagnostic accuracy for predicting PHLF (Figure 2).\u003c/p\u003e\n\u003cp\u003eThe funnel plot (Figure 3) showed relatively symmetrical study distribution. Egger\u0026rsquo;s test (p = 0.959) indicated no significant publication bias. Trim-and-fill analysis suggested that even if potential missing studies were imputed, the impact on the pooled AUC would be minimal.\u003c/p\u003e\n\u003cp\u003eThe SROC curve (Figure 4) displayed the distribution of sensitivity and specificity across studies, with most studies located in the upper left corner, indicating good overall diagnostic performance. The curve shape did not show a typical \u0026quot;shoulder-arm\u0026quot; pattern, and the points were somewhat dispersed, suggesting potential threshold differences among studies.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eWhole Liver-Based Parameters\u003c/strong\u003e\u003cbr\u003e\u0026nbsp;Meta-analysis of 7 studies evaluating whole liver-based MRI parameters showed that these parameters also had diagnostic value for predicting PHLF. Heterogeneity was negligible (I\u0026sup2; = 0.0%). The pooled AUC was 0.80 (95% CI: 0.76\u0026ndash;0.84, p \u0026lt; 0.001) (Figure 5).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe funnel plot (Figure 6) showed a relatively concentrated distribution. Due to the small number of studies and large variation in sample sizes, Egger\u0026rsquo;s test could not provide reliable results. The funnel plot appeared symmetrical, suggesting minimal publication bias, although further assessment is needed. Trim-and-fill analysis imputed 7 studies, resulting in a pooled AUC of 0.798 (original 0.80), indicating that potential missing studies had little impact on the summary effect size.\u003c/p\u003e\n\u003cp\u003eThe symmetrical distribution of studies suggests a minimal risk of publication bias.\u003c/p\u003e\n\u003cp\u003eThe sensitivity-specificity plot (Figure 7) for whole liver parameters showed that some studies had high sensitivity or specificity. The SROC curve\u0026apos;s Q* index was 0.98, indicating good diagnostic accuracy.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e7. Subgroup Analysis Results\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFor studies reporting both whole liver and future liver remnant (FLR) parameters (Cho_2011; Jeong_2025), we implemented a deduplication strategy to avoid unit-of-analysis error. Specifically, we retained the FLR parameter for Cho_2011due to higher specificity (0.818 vs 0.773) and the whole liver parameter for Jeong_2025 due to complete sensitivity and specificity data availability. At the same time, remove literature Orimo_2021 that lacks sensitivity and specificity data. This approach maintained statistical independence while maximizing data utilization.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eWhole Liver vs. FLR Parameters\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe pooled AUC for FLR parameters (12 studies) was 0.842 (95% CI: 0.792\u0026ndash;0.882), and for whole liver parameters (6 studies) it was 0.791 (95% CI: 0.726\u0026ndash;0.844). The difference was not statistically significant (p = 0.175).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eStudy Design Subgroup\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe pooled AUC for retrospective studies (16 studies) was 0.824 (95% CI: 0.780\u0026ndash;0.861). The pooled AUC for prospective studies (2 studies) was 0.825 (95% CI: 0.736\u0026ndash;0.889). The difference between study designs was not statistically significant (p = 0.978).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eGeographical Region Subgroup\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe pooled AUC for Asian studies (12 studies) was 0.840 (95% CI: 0.800\u0026ndash;0.872), and for European studies (6 studies) it was 0.783 (95% CI: 0.693\u0026ndash;0.853). The difference was not statistically significant (p = 0.183).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSample Size Subgroup\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe pooled AUC for large-sample studies (\u0026ge;100 patients, 10 studies) was 0.809 (95% CI: 0.756\u0026ndash;0.853), and for small-sample studies (\u0026lt;100 patients, 8 studies) it was 0.849 (95% CI: 0.787\u0026ndash;0.895). The difference was not statistically significant (p = 0.289).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePHLF Definition Subgroup\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe pooled AUC for studies using ISGLS criteria (14 studies) was 0.827 (95% CI: 0.774\u0026ndash;0.869), and for studies using other criteria (4 studies) it was 0.818 (95% CI: 0.778\u0026ndash;0.853). The difference was not statistically significant (p = 0.786).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMagnetic Field Strength Subgroup\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe pooled AUC for studies using 1.5T scanners (5 studies) was 0.763 (95% CI: 0.679\u0026ndash;0.830), and for studies using 3.0T scanners (10 studies) it was 0.859 (95% CI: 0.815\u0026ndash;0.894). The difference was statistically significant (p = 0.018).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e8. Threshold Effect and Publication Bias Analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThreshold effect was assessed using Spearman correlation analysis between sensitivity and (1 \u0026minus; specificity) across studies for the overall dataset and each parameter subgroup. The threshold effect was considered present when the absolute value of Spearman\u0026apos;s correlation coefficient exceeded 0.6 (|\u0026rho;| \u0026gt; 0.6).\u003c/p\u003e\n\u003cp\u003eOverall Analysis (18 studies): The analysis revealed a moderate positive correlation between sensitivity and (1 \u0026minus; specificity) (Spearman\u0026apos;s \u0026rho; = 0.514, p = 0.029). Although statistically significant, the correlation coefficient did not exceed the prespecified threshold (|\u0026rho;| \u0026le; 0.6), indicating no substantial threshold effect across all included studies. The linear regression model explained a modest portion of the variance (R\u0026sup2; = 0.159).\u003c/p\u003e\n\u003cp\u003eFLR Parameters Analysis (12 studies): Analysis of FLR parameters revealed a statistically significant threshold effect, as indicated by a strong positive correlation (Spearman\u0026apos;s \u0026rho; = 0.744, p = 0.006). The primary criterion for a threshold effect (|\u0026rho;| \u0026gt; 0.6) was clearly met, with the linear regression model explaining approximately 30% of the variance (R\u0026sup2; = 0.297). This significant threshold effect suggests that heterogeneity in diagnostic thresholds contributed substantially to the observed variation in diagnostic accuracy estimates among FLR parameter studies.\u003c/p\u003e\n\u003cp\u003eWhole Liver Parameters Analysis (6 studies): Analysis of whole liver parameters demonstrated no statistically significant threshold effect. This was determined by a weak and non-significant Spearman\u0026apos;s correlation (\u0026rho; = 0.257, p = 0.658), which fell well below the prespecified criterion (|\u0026rho;| \u0026le; 0.6). The linear regression model indicated a poor fit (R\u0026sup2; = 0.104), confirming the absence of threshold-related heterogeneity. This finding suggests relative consistency in diagnostic cut-off values across studies using whole liver parameters.\u003c/p\u003e\n\u003cp\u003ePublication Bias Analysis\u003c/p\u003e\n\u003cp\u003ePublication bias was evaluated using Egger\u0026apos;s regression test and Begg\u0026apos;s rank correlation test for the overall dataset and each parameter subgroup.\u003c/p\u003e\n\u003cp\u003eOverall Analysis (18 studies): Both statistical tests indicated no evidence of publication bias. Egger\u0026apos;s regression test yielded a non-significant result (p = 0.720), as did Begg\u0026apos;s rank correlation test (p = 0.369). The funnel plot demonstrated relatively symmetric distribution of studies around the pooled estimate.\u003c/p\u003e\n\u003cp\u003eFLR Parameters Analysis (12 studies): Publication bias assessment for FLR parameter studies showed no significant bias. Egger\u0026apos;s test (p = 0.425) and Begg\u0026apos;s test (p = 0.737) both yielded non-significant results, indicating that the meta-analysis results were not substantially influenced by publication bias.\u003c/p\u003e\n\u003cp\u003eWhole Liver Parameters Analysis (6 studies): Similarly, whole liver parameter studies showed no evidence of publication bias. Both Egger\u0026apos;s test (p = 0.480) and Begg\u0026apos;s test (p = 0.719) were non-significant, supporting the reliability of the pooled estimates.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eThis systematic review and meta-analysis comprehensively evaluated the diagnostic performance of Gd-EOB-MRI quantitative parameters for predicting PHLF, incorporating data from 19 studies involving 3,658 patients. Although the difference did not reach statistical significance, future liver remnant (FLR)-based Gd-EOB-MRI parameters demonstrated a higher pooled AUC for predicting PHLF compared to whole liver-based parameters. FLR-based parameters provide high diagnostic accuracy, with a pooled AUC of 0.85, outperforming whole liver-based parameters (AUC\u0026thinsp;=\u0026thinsp;0.80). These results underscore the clinical relevance of regional functional assessment in preoperative risk stratification for patients undergoing hepatectomy. This study screened 19 articles, enrolling a total of 3,658 subjects, of which 3,520 subjects originated from a single, single-center study. To mitigate the potential undue influence of this single study on the overall results, we employed a bivariate random-effects model. This model inherently incorporates a certain \"automatic balancing\" mechanism-by incorporating between-study variance into the denominator when calculating weights, it relatively \"dilutes\" the weight assigned to extremely large-sample studies. To further reduce the impact, we additionally performed a leave-one-out sensitivity analysis excluding the study by Jeong. Following this exclusion, the pooled AUC for all remaining studies was 0.799, which remained stable. This finding actually provides strong evidence that the conclusions are not driven by a single study, demonstrating the robustness of the results. This conclusively demonstrates that the meta-analysis results are not solely driven by this single large-sample study, and that the random-effects model has adequately addressed the weighting issues.\u003c/p\u003e \u003cp\u003eThe superior performance of FLR-based parameters is physiologically plausible. Unlike whole liver measurements, which may be confounded by heterogeneous liver function\u0026mdash;particularly in patients with chronic liver disease or prior chemotherapy\u0026mdash;FLR-specific metrics directly reflect the functional capacity of the remnant liver. This aligns with the surgical principle that postoperative outcomes depend not on the total liver function, but on the functional reserve of the residual tissue. Our results support the growing consensus that functional imaging of the FLR should be integrated into preoperative planning, especially in high-risk resections.\u003c/p\u003e \u003cp\u003eThis superior performance can be attributed to the fundamental advantage of Gd-EOB-MRI over traditional liver function assessment methods (e.g., ICG-R15, Child-Pugh score). While these conventional tools offer valuable global assessments, they lack the ability to evaluate regional functional heterogeneity. Gd-EOB-MRI, however, provides intuitive and quantitative functional imaging indicators, enabling direct assessment of the future liver remnant (FLR). This functional quantification capability is particularly useful for preoperative planning in the context of non-uniform liver disease. Our findings, which are consistent with and strengthen the evidence from previous single-center studies, demonstrate that FLR-based parameters do not merely replicate traditional test information but provide unique, spatially resolved data. Therefore, the most promising clinical application lies in the development of integrated models that combine the strengths of both approaches\u0026mdash;global liver function from serum biomarkers and volumetric/functional assessment from Gd-EOB-MRI\u0026mdash;to achieve a more personalized and precise estimation of individual PHLF risk.\u003c/p\u003e \u003cp\u003eSubgroup analyses revealed no significant differences in diagnostic performance based on MRI parameters, study design, geographic region, sample size, or PHLF definition. This consistency enhances the generalizability of our findings across diverse clinical settings. However, we observed a statistically significant advantage for studies using 3.0T MRI scanners over 1.5T (AUC 0.859 vs. 0.763, p\u0026thinsp;=\u0026thinsp;0.018), suggesting that higher magnetic field strength may improve the precision of quantitative functional assessment.\u003c/p\u003e \u003cp\u003eA notable finding was the presence of a significant threshold effect among FLR-based parameters (Spearman\u0026rsquo;s ρ\u0026thinsp;=\u0026thinsp;0.744, p\u0026thinsp;=\u0026thinsp;0.006), indicating that variability in cutoff values contributed to heterogeneity in diagnostic accuracy. This highlights the lack of standardized thresholds across studies and underscores the need for consensus on optimal cutoffs for clinical use. In contrast, whole liver parameters showed no threshold effect, possibly due to more consistent measurement approaches.\u003c/p\u003e \u003cp\u003ePublication bias was not detected in either subgroup, reinforcing the reliability of our pooled estimates. However, the overall risk of bias in the included studies, though low, was not negligible, and the predominance of retrospective designs limits causal inference. Moreover, the absence of post-discharge follow-up in some studies may have led to underdiagnosis of late-onset PHLF.\u003c/p\u003e \u003cp\u003eOur study has several limitations. First, the inclusion of only English-language publications may introduce language bias. Second, the heterogeneity in MRI protocols, parameter definitions, and ROI selection methods may affect the comparability of results. Third, the small number of prospective studies and the lack of external validation in independent cohorts warrant caution in overinterpreting the results.\u003c/p\u003e \u003cp\u003eDespite these limitations, our findings underscore the clinical value of Gd-EOB-MRI as a non-invasive, reproducible, and functionally informative tool for preoperative risk assessment. Future efforts should focus on standardizing acquisition protocols, validating cutoff values in multicenter prospective cohorts, and integrating functional MRI parameters with clinical and biochemical markers to develop robust predictive models, thereby paving the way for its broad clinical implementation.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eIn conclusion, FLR-based Gd-EOB-MRI quantitative parameters demonstrate excellent diagnostic performance for predicting PHLF and are superior to whole liver-based parameters. These results support the incorporation of functional MRI assessment into preoperative planning for hepatectomy. However, the presence of a threshold effect and methodological heterogeneity across studies emphasize the need for standardized imaging protocols and validated diagnostic thresholds. Future prospective, multicenter studies are essential to confirm these findings and facilitate the translation of Gd-EOB-MRI into routine clinical practice for improving postoperative outcomes.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eL.Y and S.X contributed to the study conception and design. Material preparation, data collection, and analysis were performed by L.Y. The first draft of the manuscript was written by S.X, and both authors commented on previous versions of the manuscript. S.X supervised the project and was responsible for the final approval of the version to be published. Both authors read and approved the final manuscript.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eRahbari NN, Garden OJ, Padbury R, et al. Posthepatectomy liver failure: a definition and grading by the International Study Group of Liver Surgery (ISGLS). Surgery. 2011;149(5):713-724.\u003c/li\u003e\n\u003cli\u003eMullen JT, Ribero D, Reddy SK, et al. Hepatic insufficiency and mortality in 1,059 noncirrhotic patients undergoing major hepatectomy. J Am Coll Surg. 2007;204(5):854-862.\u003c/li\u003e\n\u003cli\u003eImamura H, Seyama Y, Kokudo N, et al. One thousand fifty-six hepatectomies without mortality in 8 years. Arch Surg. 2003;138(11):1198-1206.\u003c/li\u003e\n\u003cli\u003eReiner CS, Koh DM, Brismar TB. Functional MRI of the liver. Eur J Radiol. 2021;137:109578.\u003c/li\u003e\n\u003cli\u003eVan Beers BE, Pastor CM, Hussain HK. Primovist, Eovist: what to expect? J Hepatol. 2012;57(2):421-429.\u003c/li\u003e\n\u003cli\u003eVerloh N, Haimerl M, Zeman F, et al. Assessing liver function by liver enhancement during the hepatobiliary phase with Gd-EOB-DTPA-enhanced MRI at 3 Tesla. Eur Radiol. 2014;24(5):1013-1019.\u003c/li\u003e\n\u003cli\u003eYoon JH, Lee JM, Kim E, et al. Quantitative liver function analysis: volumetric T1 mapping with fast multisection B1 inhomogeneity correction in hepatocyte-specific contrast-enhanced liver MR imaging. Radiology. 2017;282(2):408-417.\u003c/li\u003e\n\u003cli\u003ePage MJ, McKenzie JE, Bossuyt PM, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ. 2021;372:n71.\u003c/li\u003e\n\u003cli\u003eEuropean Society of Urogenital Radiology. ESUR guidelines on contrast agents v10.0. 2018.\u003c/li\u003e\n\u003cli\u003eWhiting PF, Rutjes AW, Westwood ME, et al. QUADAS-2: a revised tool for the quality assessment of diagnostic accuracy studies. Ann Intern Med. 2011;155(8):529-536.\u003c/li\u003e\n\u003cli\u003eCho JY, Suh KS, Kwon CHD, et al. Functional recovery after living donor liver transplantation: comparison between ABO-compatible and ABO-incompatible transplantation. Liver Transpl. 2011;17(10):1180-1190.\u003c/li\u003e\n\u003cli\u003eYoon JH, Lee JM, Kim E, et al. Hepatocyte-specific contrast agent-enhanced magnetic resonance imaging vs computed tomography for the detection of hepatocellular carcinoma: a systematic review and meta-analysis. J Gastroenterol Hepatol. 2016;31(10):1766-1778.\u003c/li\u003e\n\u003cli\u003eJin YJ, Cho SG, Lee KY, et al. Prediction of posthepatectomy liver failure using gadoxetic acid-enhanced magnetic resonance imaging in patients with hepatocellular carcinoma. J Gastroenterol Hepatol. 2016;31(4):837-845.\u003c/li\u003e\n\u003cli\u003eBarth BK, Donati OF, Fischer MA, et al. Reliability, validity, and reader acceptance of LI-RADS-an in-depth analysis. Acad Radiol. 2016;23(9):1145-1153.\u003c/li\u003e\n\u003cli\u003eAsenbaum U, Kaczirek K, Ba-Ssalamah A, et al. Post-hepatectomy liver failure after major hepatic resection: not only size matters. Eur Radiol. 2018;28(11):4748-4756.\u003c/li\u003e\n\u003cli\u003eKim DK, Choi JI, Choi MH, et al. Prediction of posthepatectomy liver failure: MRI with hepatocyte-specific contrast agent versus indocyanine green clearance test. AJR Am J Roentgenol. 2018;211(3):580-587.\u003c/li\u003e\n\u003cli\u003eTheilig D, Steffen IG, Maurer MH, et al. Predicting liver failure after extended right hepatectomy following right portal vein embolization with gadoxetic acid-enhanced MRI. Eur Radiol. 2019;29(12):6540-6549.\u003c/li\u003e\n\u003cli\u003eAraki K, Conrad C, Ogiso S, et al. Predictive value of gadoxetic acid-enhanced magnetic resonance imaging for posthepatectomy liver failure after major hepatectomy. J Am Coll Surg. 2019;229(4):369-378.\u003c/li\u003e\n\u003cli\u003eHuang ZY, Liang BY, Xiong M, et al. Severity of cirrhosis should determine the operative modality for patients with early hepatocellular carcinoma and compensated cirrhosis. Surgery. 2020;168(4):621-631.\u003c/li\u003e\n\u003cli\u003eWang YY, Zhao LN, Jiao Y, et al. Prediction of posthepatectomy liver failure using gadoxetic acid-enhanced MRI in patients with HBV-related hepatocellular carcinoma. J Magn Reson Imaging. 2020;52(4):1111-1121.\u003c/li\u003e\n\u003cli\u003eOrimo T, Kamiyama T, Mitsuhashi T, et al. Impact of hybrid operating room on liver surgery: a single-center experience of 800 hepatectomies. Surg Endosc. 2021;35(8):4279-4287.\u003c/li\u003e\n\u003cli\u003eNotake T, Shimizu S, Ohki T, et al. Prediction of posthepatectomy liver failure using gadoxetic acid-enhanced magnetic resonance imaging in patients undergoing major hepatectomy. Medicine (Baltimore). 2021;100(25):e26373.\u003c/li\u003e\n\u003cli\u003eLauscher JC, Elezkurtaj S, Diehl SJ, et al. Prediction of posthepatectomy liver failure using gadoxetic acid-enhanced MRI in patients with perihilar cholangiocarcinoma. Eur Radiol. 2021;31(11):8529-8537.\u003c/li\u003e\n\u003cli\u003eJeong WK, Jamshidi N, Felker ER, et al. Gadoxetic acid-enhanced liver magnetic resonance imaging: hepatocyte uptake in patients with liver disease. Invest Radiol. 2020;55(1):1-11.\u003c/li\u003e\n\u003cli\u003eMaino C, Vernuccio F, Cannella R, et al. Radiomics and liver: Where we are and where we are headed? Eur J Radiol. 2023;156:110522.\u003c/li\u003e\n\u003cli\u003eZhu SC, Liu YH, Wei Y, et al. Prediction of posthepatectomy liver failure using T1 mapping on gadoxetic acid-enhanced magnetic resonance imaging. Eur Radiol. 2024;34(2):1048-1057.\u003c/li\u003e\n\u003cli\u003eLi J, Wang K, Chen X, et al. Machine learning-based prediction of posthepatectomy liver failure using preoperative gadoxetic acid-enhanced MRI and clinical data. Eur Radiol. 2024;34(5):3201-3212.\u003c/li\u003e\n\u003cli\u003eDonadon M, Fontana A, Palmisano A, et al. Functional liver imaging score (FLIS) based on gadoxetic acid-enhanced magnetic resonance imaging predicts post-hepatectomy liver failure. Ann Surg. 2024;279(1):122-129.\u003c/li\u003e\n\u003cli\u003eZhang L, Wu H, Chen Y, et al. Multicenter validation of gadoxetic acid-enhanced MRI for predicting posthepatectomy liver failure: a prospective study. Radiology. 2024;310(2):e231456.\u003c/li\u003e\n\u003cli\u003eVickers AJ, Elkin EB. Decision curve analysis: a novel method for evaluating prediction models. Med Decis Making. 2006;26(6):565-574.\u003c/li\u003e\n\u003c/ol\u003e"},{"header":"Table","content":"\u003cp\u003eTable 1 is available in the supplementary files section\u003c/p\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"abdominal-radiology","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"aima","sideBox":"Learn more about [Abdominal Radiology](http://link.springer.com/journal/261)","snPcode":"261","submissionUrl":"https://submission.springernature.com/new-submission/261/3","title":"Abdominal Radiology","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"Hepatectomy, Posthepatectomy liver failure, Gadoxetic acid, Magnetic resonance imaging, Future liver remnant, Meta-analysis","lastPublishedDoi":"10.21203/rs.3.rs-9333721/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9333721/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eBackground \u0026amp; Aims: Posthepatectomy liver failure (PHLF) is a major cause of morbidity and mortality following liver resection. Accurate preoperative prediction of posthepatectomy liver failure (PHLF) remains a clinical challenge. Gadoxetic acid-enhanced MRI (Gd-EOB-MRI) enables quantitative assessment of liver function. This systematic review and meta-analysis aims to determine the diagnostic accuracy of Gd-EOB-MRI quantitative parameters for predicting PHLF and to explore sources of heterogeneity.\u003c/p\u003e\n\u003cp\u003eMethods: A systematic search of PubMed, Embase, Web of Science, and Cochrane Library was conducted from January 2011 to June 2025 for studies evaluating Gd-EOB-MRI parameters for PHLF prediction in adults undergoing hepatectomy. The bivariate random-effects model was used to calculate the pooled area under the curve (AUC). Heterogeneity was assessed using the I² statistic. Subgroup, threshold effect, and publication bias analyses were performed.\u003c/p\u003e\n\u003cp\u003eResults: Nineteen studies involving 3,658 patients were included. Future liver remnant (FLR)-based parameters (14 studies) demonstrated high diagnostic accuracy with a pooled AUC of 0.85 (95% CI: 0.82–0.89) and moderate heterogeneity (I² = 27.9%). Whole liver-based parameters (7 studies) also showed good performance (pooled AUC = 0.80, 95% CI: 0.76–0.84). Subgroup analyses revealed no significant differences based on MRI parameters， study design, region, sample size, or PHLF definition. However, studies using 3.0T MRI scanners showed a significantly higher pooled AUC than those using 1.5T (0.859 vs. 0.763, p=0.018). A significant threshold effect was observed for FLR parameters (Spearman's ρ = 0.744, p=0.006) but not for whole liver parameters. No significant publication bias was detected.\u003c/p\u003e\n\u003cp\u003eConclusions: FLR-based Gd-EOB-MRI parameters provide excellent diagnostic performance for predicting PHLF, outperforming to whole liver-based assessment. This supports the integration of functional FLR assessment into preoperative risk stratification. However, the presence of a threshold effect underscores the need for standardized protocols and validated diagnostic cut-offs in future prospective multicenter studies. Prospective multicenter studies with standardized protocols and predefined diagnostic thresholds are needed to validate these findings and facilitate clinical translation.\u003c/p\u003e","manuscriptTitle":"Functional Assessment by Gd-EOB-MRI of the Future Liver Remnant Predicts Posthepatectomy Liver Failure: A Systematic Review and Meta-Analysis","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-04-21 14:58:30","doi":"10.21203/rs.3.rs-9333721/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"editorInvitedReview","content":"","date":"2026-04-29T15:10:55+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"339614766154210439264410873886277641495","date":"2026-04-18T14:28:45+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"267806819653782114602377002852340151118","date":"2026-04-16T10:41:54+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"91661319801077844550923147712280708418","date":"2026-04-13T17:14:09+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-04-13T15:31:16+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-04-07T03:01:23+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-04-07T03:00:49+00:00","index":"","fulltext":""},{"type":"submitted","content":"Abdominal Radiology","date":"2026-04-06T12:02:53+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"abdominal-radiology","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"aima","sideBox":"Learn more about [Abdominal Radiology](http://link.springer.com/journal/261)","snPcode":"261","submissionUrl":"https://submission.springernature.com/new-submission/261/3","title":"Abdominal Radiology","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"1e1e2f72-4a2a-488b-94a1-8048dd6675a6","owner":[],"postedDate":"April 21st, 2026","published":true,"recentEditorialEvents":[{"type":"editorInvitedReview","content":"","date":"2026-04-29T15:10:55+00:00","index":15,"fulltext":""}],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2026-04-21T14:58:30+00:00","versionOfRecord":[],"versionCreatedAt":"2026-04-21 14:58:30","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9333721","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9333721","identity":"rs-9333721","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}