LightGBM outperforms other machine learning techniques in predicting graft failure after liver transplantation: Creation of a predictive model through large-scale analysis.

BACKGROUND: The incidence of graft failure following liver transplantation (LTx) is consistent. While traditional risk scores for LTx have limited accuracy, the potential of machine learning (ML) in this area remains uncertain, despite its promise in other transplant domains. This study aims to determine ML's predictive limitations in LTx by replicating methods used in previous heart transplant research. METHODS: This study utilized the UNOS STAR database, selecting 64,384 adult patients who underwent LTx between 2010 and 2020. Gradient boosting models (XGBoost and LightGBM) were used to predict 14, 30, and 90-day graft failure compared to conventional logistic regression model. Models were evaluated using both shuffled and rolling cross-validation (CV) methodologies. Model performance was assessed using the AUC across validation iterations. RESULTS: In a study comparing predictive models for 14-day, 30-day and 90-day graft survival, LightGBM consistently outperformed other models, achieving the highest AUC of.740,.722, and.700 in shuffled CV methods. However, in rolling CV the accuracy of the model declined across every ML algorithm. The analysis revealed influential factors for graft survival prediction across all models, including total bilirubin, medical condition, recipient age, and donor AST, among others. Several features like donor age and recipient diabetes history were important in two out of three models. CONCLUSIONS: LightGBM enhances short-term graft survival predictions post-LTx. However, due to changing medical practices and selection criteria, continuous model evaluation is essential. Future studies should focus on temporal variations, clinical implications, and ensure model transparency for broader medical utility.

Can the Timed Up & Go Test and Montreal Cognitive Assessment predict outcomes in patients waitlisted for renal transplant?

PURPOSE: Frail patients who undergo renal transplantation (RT) have more complications; however, little is known if these patients can sustain the wait to RT. We used the Timed Up and Go Test (TUGT) and Montreal Cognitive Assessment (MoCA) to determine outcomes of RT candidates. METHODS: In this retrospective study, 526 RT candidates underwent TUGT and MoCA (2015-2019) and were divided into "favorable" (transplanted or remained on the list) or "unfavorable" (not listed, removed from list, or died) outcome. Demographics, education, language, comorbidities, dialysis type, use of a walking device, TUGT, and MoCA were compared by outcome. RESULTS: Overall, 230 patients (43.7%) passed TUG, 268 (51%) passed MoCA, 133 (25.3%) passed both, and 161 (30.6%) failed both tests. Multivariate analysis demonstrated age ≥ 65 (OR 1.58, CI 1.03-2.43), cardiac disease (OR 3.09, CI 2.02-4.72), ≥36 months on dialysis (OR 1.80, CI 1.24-2.69), EPTS < 20% at time of MoCA (OR 0.26, CI 0.07-0.98), and failing TUGT (OR 2.14, CI 1.43-3.19) were associated with unfavorable outcome. Failing MoCA was not associated with outcome. CONCLUSIONS: MoCA test results were not associated with RT waitlist outcomes; however, passing the TUGT was associated with receiving RT or remaining on the list. Additional studies are needed to validate this and determine outcome after RT.

Breaking distance barriers in liver transplantation: Risk factors and outcomes of long-distance liver grafts.

BACKGROUND: Long-distance-traveling liver grafts in liver transplantation present challenges due to prolonged cold ischemic time and increased risk of ischemia-reperfusion injury. We identified long-distance-traveling liver graft donor and recipient characteristics and risk factors associated with long-distance-traveling liver graft use. METHODS: We conducted a retrospective analysis of data from donor liver transplantation patients registered from 2014 to 2020 in the United Network for Organ Sharing registry database. Donor, recipient, and transplant factors of graft survival were compared between short-travel grafts and long-distance-traveling liver grafts (traveled >500 miles). RESULTS: During the study period, 28,265 patients received a donation after brainstem death liver transplantation and 3,250 a donation after circulatory death liver transplantation. The long-distance-traveling liver graft rate was 6.2% in donation after brainstem death liver transplantation and 7.1% in donation after circulatory death liver transplantation. The 90-day graft survival rates were significantly worse for long-distance-traveling liver grafts (donation after brainstem death: 95.7% vs 94.5%, donation after circulatory death: 94.5% vs 93.9%). The 3-year graft survival rates were similar for long-distance-traveling liver grafts (donation after brainstem death: 85.5% vs 85.1%, donation after circulatory death: 81.0% vs 80.4%). Cubic spline regression analyses revealed that travel distance did not linearly worsen the prognosis of 3-year graft survival. On the other hand, younger donor age, lower donor body mass index, and shorter cold ischemic time mitigated the negative impact of 90-day graft survival in long-distance-traveling liver grafts. CONCLUSION: The use of long-distance-traveling liver grafts negatively impacts 90-day graft survival but not 3-year graft survival. Moreover, long-distance-traveling liver grafts are more feasible with appropriate donor and recipient factors offsetting the extended cold ischemic time. Mechanical perfusion can improve long-distance-traveling liver graft use. Enhanced collaboration between organ procurement organizations and transplant centers and optimized transportation systems are essential for increasing long-distance-traveling liver graft use, ultimately expanding the donor pool.

Coatings of polyethylene glycol for suppressing adhesion between solid microspheres and flat surfaces.

This article describes the development and the examination of surface coatings that suppress the adhesion between glass surfaces and polymer microspheres. Superparamagnetic doping allowed for exerting magnetic forces on the microbeads. The carboxyl functionalization of the polymer provided the means for coating the beads with polyethylene glycol (PEG) with different molecular weight. Under gravitational force, the microbeads settled on glass surfaces with similar polymer coatings. We examined the efficacy of removing the beads from the glass surfaces by applying a pulling force of ~1.2 pN. The percent beads remaining on the surface after applying the pulling force for approximately 5 s served as an indication of the adhesion propensity. Coating of PEG with molecular weight ranging between 3 and 10 kDa was essential for suppressing the adhesion. For the particular substrates, surface chemistry and aqueous media we used, coatings of 5 kDa manifested optimal suppression of adhesion: that is, only 3% of the microbeads remained on the surface after applying the pulling magnetic force. When either the glass or the beads were not PEGylated, the adhesion between them was substantial. Addition of a noncharged surfactant, TWEEN, above its critical micelle concentrations (CMCs) suppressed the adhesion between noncoated substrates. The extent of this surfactant-induced improvement of the adhesion suppression, however, did not exceed the quality of preventing the adhesion that we attained by PEGylating both substrates. In addition, the use of surfactants did not significantly improve the suppression of bead-surface adhesion when both substrates were PEGylated. These findings suggest that such surfactant additives tend to be redundant and that covalently grafted coatings of PEGs with selected chain lengths provide sufficient suppression of nonspecific interfacial interactions.

Prognostic ability of inflammation-based markers in radioembolization for hepatocellular carcinoma.

AIM: Inflammation-based markers, such as the neutrophil-to-lymphocyte ratio (NLR) and platelet-to-lymphocyte ratio (PLR), have recently been used as prognostic indicators in hepatocellular carcinoma (HCC). We aimed to determine whether NLR and PLR may predict response to yttrium-90 transarterial radioembolization (TARE) as primary treatment for HCC. METHODS: We performed a retrospective review of a prospectively collected database of HCC cases (1994-2019) and selected patients who received TARE as primary treatment (n = 42). Laboratory studies were used to calculate NLR and PLR. Response to TARE was determined using the modified response evaluation criteria in solid tumors (mRECIST). Patients were classified as non-responders (stable or progressive disease) or responders (partial or complete response) to treatment based on mRECIST. RESULTS: Receiver operating characteristic curves identified a pre-treatment NLR cutoff of ≥ 2.83 and a pre-treatment PLR cutoff of ≥ 83 for predicting non-response to treatment. Pre-treatment NLR ≥ 2.83 was the only significant predictor of non-response to TARE in multivariate logistic regression analysis (odds ratio 7.83, P = 0.036). On time to progression analysis, both pre-treatment NLR ≥ 2.83 and pre-treatment PLR ≥ 83 were associated with a higher proportion of tumor progression at 6 months post-treatment (43.6% vs. 10.0%, P = 0.014, log-rank) and (38.6% vs. 0%, P = 0.010, log-rank), respectively. CONCLUSION: NLR confers prognostic value and may be superior to PLR in determining response to TARE as primary treatment for HCC. Future studies are necessary to validate these findings in a larger cohort.

Underlying liver disease and advanced stage liver cancer are associated with elevated neutrophil-lymphocyte ratio.

BACKGROUND/AIMS: Inflammation-based scores, such as the neutrophil-to-lymphocyte ratio (NLR), have been associated with prognosis in hepatocellular carcinoma (HCC); but variable cut-off values and potential lack of specificity have limited the utility of NLR. This study evaluates NLR in a large cohort of HCC patients. METHODS: We retrospectively reviewed 789 HCC cases (1993-2017) for demographics, tumor characteristics, treatment, and survival. NLR was stratified into NLR ≥1.5 and NLR ≥3 and analyzed for correlation with American Joint Committee on Cancer (AJCC) and Barcelona Clinic Liver Cancer (BCLC) stages. In 235 patients who underwent liver resection, survival and recurrence were evaluated by NLR. RESULTS: In 789 HCC cases, mean NLR was increased with advanced AJCC and BCLC stages. Hepatitis C patients were less likely to have NLR ≥1.5 and ≥3. Non-alcoholic steatohepatitis patients were more likely to have NLR ≥3. Patients with tumor size >5 cm, rupture, or macrovascular invasion were more likely to have NLR ≥3. In patients treated with resection, NLR ≥3 predicted early recurrence (odds ratio [OR] 4.14, P<0.01) and overall recurrence (OR 4.05, P<0.01). Mean NLR was 4.30 in those with recurrence and 2.75 in those without recurrence. Patients with NLR ≥3 showed significantly worse survival compared to those with NLR <3 (P<0.01 by log-rank test). CONCLUSION: Elevated NLR is associated with advanced cancer stage and aggressive tumor characteristics, such as large size, rupture, and invasion. NLR ≥3 was associated with early and overall recurrence after resection but varied with etiology. NLR may be a useful biomarker in predicting recurrence for HCC patients undergoing curative resection, but further studies are required to elucidate the effect of disease etiology.