Analytical Performance Specifications for Input Variables: Investigation of the Model of End-Stage Liver Disease

BACKGROUND

Artificial intelligence models constitute specific uses of analysis results and, therefore, necessitate evaluation of analytical performance specifications (APS) for this context specifically. The Model of End-stage Liver Disease (MELD) is a clinical prediction model based on measurements of bilirubin, creatinine, and the international normalized ratio (INR). This study evaluates the propagation of error through the MELD, to inform choice of APS for the MELD input variables.

METHODS

A total of 6093 consecutive MELD scores and underlying analysis results were retrospectively collected. "Desirable analytical variation" based on biological variation as well as current local analytical variation was simulated onto the data set as well as onto a constructed data set, representing a worst-case scenario. Resulting changes in MELD score and risk classification were calculated.

RESULTS

Biological variation-based APS in the worst-case scenario resulted in 3.26% of scores changing by ≥1 MELD point. In the patient-derived data set, the same variation resulted in 0.92% of samples changing by ≥1 MELD point, and 5.5% of samples changing risk category. Local analytical performance resulted in lower reclassification rates.

CONCLUSIONS

Error propagation through MELD is complex and includes population-dependent mechanisms. Biological variation-derived APS were acceptable for all uses of the MELD score. Other combinations of APS can yield equally acceptable results. This analysis exemplifies how error propagation through artificial intelligence models can become highly complex. This complexity will necessitate that both model suppliers and clinical laboratories address analytical performance specifications for the specific use case, as these may differ from performance specifications for traditional use of the analyses.

Impact of Integrating Insulin-Like Growth Factor 1 Levels into Model for End-Stage Liver Disease Score for Survival Prediction in Hepatocellular Carcinoma Patients

BACKGROUND

Liver reserve affects survival in hepatocellular carcinoma (HCC). Model for End-Stage Liver Disease (MELD) score is used to predict overall survival (OS) and to prioritize HCC patients on the transplantation waiting list, but more accurate models are needed. We hypothesized that integrating insulin-like growth factor 1 (IGF-1) levels into MELD score (MELD-IGF-1) improves OS prediction as compared to MELD.

METHODS

We measured plasma IGF-1 levels in training (n = 310) and validation (n = 155) HCC cohorts and created MELD-IGF-1 score. Cox models were used to determine the association of MELD and MELD-IGF-1 with OS. Harrell's c-index was used to compare the predictive capacity.

RESULTS

IGF-1 was significantly associated with OS in both cohorts. Patients with an IGF-1 level of ≤26 ng/mL in the training cohort and in the validation cohorts had significantly higher hazard ratios than patients with the same MELD but IGF-1 >26 ng/mL. In both cohorts, MELD-IGF-1 scores had higher c-indices (0.60 and 0.66) than MELD scores (0.58 and 0.60) (p < 0.001 in both cohorts). Overall, 26% of training and 52.9% of validation cohort patients were reclassified into different risk groups by MELD-IGF-1 (p < 0.001).

CONCLUSIONS

After independent validation, the MELD-IGF-1 could be used to risk-stratify patients in clinical trials and for priority assignment for patients on liver transplantation waiting list.

Limited comparability of creatinine assays in patients with liver cirrhosis and their impact on the MELD score

BACKGROUND & AIM

Patients with end-stage liver disease require valid estimations of mortality for organ allocation and risk stratification. The model of end-stage liver disease (MELD) score is used for this purpose in most countries and incorporates bilirubin, International Normalized ratio, and creatinine. The aim of this study was to evaluate the comparability of creatinine results from different routine assays in the serum samples of patients with liver cirrhosis.

METHODS

Residual material from 60 serum samples was available from patients in different stages of liver cirrhosis. Four centers participated; each center analyzed the samples with Jaffé-based and enzymatic routine assays in parallel. In addition, an accredited calibration laboratory certified the panel of samples by an internationally accepted reference measurement procedure (RMP) based on isotope dilution mass spectrometry (ID-MS). This method served as the independent reference.

RESULTS

All routine methods displayed a high correlation to the RMP (r ≥0.937, p<0.001). Two enzymatic and two Jaffé-based methods provided results that were all within a ±20% range of the RMP. The other methods showed deviations >20% in up to 27% of the samples. The enzymatic methods were systematically lower, whereas the Jaffé-based methods were systematically higher (p<0.001). The resulting MELD scores differed from 0 to 4 points.

CONCLUSIONS

There are systematic deviations from the RMP. Jaffé-based assays gave higher results, whereas the enzymatic-based assays gave lower results compared to the results of the RMP. The comparability of results is limited and could be disadvantageous to patients listed for liver transplantation.

Cholesterol esterification in plasma as a biomarker for liver function and prediction of mortality

Background

Advanced stages of liver cirrhosis lead to a dramatically increased mortality. For valid identification of these patients suitable biomarkers are essential. The most important biomarkers for liver function are bilirubin and prothrombin time expressed as International Normalized Ratio (INR). However, the influence of several anticoagulants on the prothrombin time limits its diagnostic value.

Aim of this study was the evaluation of cholesterol esterification (CE) fraction (esterified cholesterol vs. total cholesterol) as an alternative biomarker for liver synthesis and mortality prediction. Under physiological conditions the CE fraction in blood is closely regulated by lecithin-cholesterol acyltransferase (LCAT) which is produced in the liver.

Methods

One hundred forty-two patients with liver disease clinically considered for orthotopic liver transplant for different indications were enrolled in the study. One patient was excluded because of the intake of a direct oral factor Xa inhibitor which has a strong impact on prothrombin time.

Results

Results of CE fraction were in good agreement with INR (R² = 0.73; p < 0.001). In patients who died or survived within three months mean CE fraction was 56% vs. 74% (p < 0.001) and mean INR was 2.0 vs. 1.3 (p < 0.001), respectively. The predictive value of CE fraction for three-month mortality risk was higher compared to INR (p = 0.04). Results for one-year mortality were comparable.

Conclusions

The cholesterol esterification fraction is a valid biomarker for liver synthesis and allows reliable prediction of mortality. In contrast to INR, it is independent of anticoagulation and other analytical limitations of coagulation tests.

Electronic supplementary material

The online version of this article (doi:10.1186/s12876-017-0614-9) contains supplementary material, which is available to authorized users.

Impact of different creatinine measurement methods on liver transplant allocation

Introduction

The model for end-stage liver disease (MELD) score is used in many countries to prioritize organ allocation for the majority of patients who require orthotopic liver transplantation. This score is calculated based on the following laboratory parameters: creatinine, bilirubin and the international normalized ratio (INR). Consequently, high measurement accuracy is essential for equitable and fair organ allocation. For serum creatinine measurements, the Jaffé method and enzymatic detection are well-established routine diagnostic tests.

Methods

A total of 1,013 samples from 445 patients on the waiting list or in evaluation for liver transplantation were measured using both creatinine methods from November 2012 to September 2013 at the university hospital Leipzig, Germany. The measurements were performed in parallel according to the manufacturer’s instructions after the samples arrived at the institute of laboratory medicine. Patients who had required renal replacement therapy twice in the previous week were excluded from analyses.

Results

Despite the good correlation between the results of both creatinine quantification methods, relevant differences were observed, which led to different MELD scores. The Jaffé measurement led to greater MELD score in 163/1,013 (16.1%) samples with differences of up to 4 points in one patient, whereas differences of up to 2 points were identified in 15/1,013 (1.5%) samples using the enzymatic assay. Overall, 50/152 (32.9%) patients with MELD scores >20 had higher scores when the Jaffé method was used.

Discussion

Using the Jaffé method to measure creatinine levels in samples from patients who require liver transplantation may lead to a systematic preference in organ allocation. In this study, the differences were particularly pronounced in samples with MELD scores >20, which has clinical relevance in the context of urgency of transplantation. These data suggest that official recommendations are needed to determine which laboratory diagnostic methods should be used when calculating MELD scores.

Prothrombotic factors in histologically proven nonalcoholic fatty liver disease and nonalcoholic steatohepatitis

An independent role of nonalcoholic fatty liver disease (NAFLD) in the development of cardiovascular disease has been suggested, probably mediated through increased levels of prothrombotic factors. Therefore, we examined whether NAFLD is linked to a prothrombotic state, independently of metabolic risk factors in a large single-center cohort of overweight/obese patients. Patients presenting to the obesity clinic underwent a detailed metabolic and liver assessment, including an extensive panel of coagulation factors. If NAFLD was suspected, a liver biopsy was proposed. A series of 273 consecutive patients (65% female) with a liver biopsy were included (age, 44 ± 0.76 years; body mass index: 39.6 ± 0.40 kg/m(2)). Increase in fibrinogen, factor VIII, and von Willebrand factor and decrease in antithrombin III correlated with metabolic features, but not with liver histology. Levels of plasminogen activator inhibitor-1 (PAI-1) increased significantly with increasing severity of steatosis (P < 0.001), lobular inflammation (P < 0.001), ballooning (P = 0.002), and fibrosis (P < 0.001). Patients with nonalcoholic steatohepatitis had significantly higher PAI-1 values than those with normal liver (P < 0.001). In multiple regression, including anthropometric and metabolic parameters, steatosis remained an independent predictor of PAI-1 levels, explaining, together with fasting C-peptide and waist circumference, 21% of the variance in PAI-1. No consistent correlations with histology were found for the other coagulation factors.

CONCLUSION

In obesity, NAFLD severity independently contributes to the increase in PAI-1 levels, whereas other coagulation factors are unaltered. This finding might, in part, explain the increased cardiovascular risk associated with NAFLD.

Model for End-stage Liver Disease

Model for end-stage liver disease (MELD) score, initially developed to predict survival following transjugular intrahepatic portosystemic shunt was subsequently found to be accurate predictor of mortality amongst patents with end-stage liver disease. Since 2002, MELD score using 3 objective variables (serum bilirubin, serum creatinine, and institutional normalized ratio) has been used worldwide for listing and transplanting patients with end-stage liver disease allowing transplanting sicker patients first irrespective of the wait time on the list. MELD score has also been shown to be accurate predictor of survival amongst patients with alcoholic hepatitis, following variceal hemorrhage, infections in cirrhosis, after surgery in patients with cirrhosis including liver resection, trauma, and hepatorenal syndrome (HRS). Although, MELD score is closest to the ideal score, there are some limitations including its inaccuracy in predicting survival in 15-20% cases. Over the last decade, many efforts have been made to further improve and refine MELD score. Until, a better score is developed, liver allocation would continue based on the currently used MELD score.