The Role of Predictive Models in the Assessment of the Poor Outcomes in Pediatric Acute Liver Failure

Objectives: In children, acute liver failure (ALF) is a severe condition with high mortality. As some patients need liver transplantation (LT), it is essential to predict the fatal evolution and to refer them early for LT if needed. Our study aimed to evaluate the prognostic criteria and scores for assessing the outcome in children with ALF. Methods: Data of 161 children with ALF (54.66% female, mean age 7.66 ± 6.18 years) were analyzed based on final evolution (32.91% with fatal evolution or LT) and etiology. We calculated on the first day of hospitalization the PELD score (109 children), MELD, and MELD-Na score (52 children), and King’s College Criteria (KCC) for all patients. The Nazer prognostic index and Wilson index for predicting mortality were calculated for nine patients with ALF in Wilson’s disease (WD). Results: PELD, MELD, and MELD-Na scores were significantly higher in patients with fatal evolution (21.04 ± 13.28 vs. 13.99 ± 10.07, p = 0.0023; 36.20 ± 19.51 vs. 20.08 ± 8.57, p < 0.0001; and 33.07 ± 8.29 vs. 20.08 ± 8.47, p < 0.0001, respectively). Moreover, age, bilirubin, albumin, INR, and hemoglobin significantly differed in children with fatal evolution. Function to etiology, PELD, MELD, MELD-Na, and KCC accurately predicted fatal evolution in toxic ALF (25.33 vs. 9.90, p = 0.0032; 37.29 vs. 18.79, p < 0.0001; 34.29 vs. 19.24, p = 0.0002, respectively; with positive predicting value 100%, negative predicting value 88.52%, and accuracy 89.23% for King’s College criteria). The Wilson index for predicting mortality had an excellent predictive strength (100% sensibility and specificity), better than the Nazer prognostic index. Conclusions: Prognostic scores may be used to predict the fatal evolution of ALF in children in correlation with other parameters or criteria. Early estimation of the outcome of ALF is essential, mainly in countries where emergency LT is problematic, as the transfer to a specialized center could be delayed, affecting survival chances.

Effectiveness, Explainability and Reliability of Machine Meta-Learning Methods for Predicting Mortality in Patients with COVID-19: Results of the Brazilian COVID-19 Registry

The majority prognostic scores proposed for early assessment of coronavirus disease 19 (COVID-19) patients are bounded by methodological flaws. Our group recently developed a new risk score - ABC2SPH - using traditional statistical methods (least absolute shrinkage and selection operator logistic regression - LASSO). In this article, we provide a thorough comparative study between modern machine learning (ML) methods and state-of-the-art statistical methods, represented by ABC2SPH, in the task of predicting in-hospital mortality in COVID-19 patients using data upon hospital admission. We overcome methodological and technological issues found in previous similar studies, while exploring a large sample (5,032 patients). Additionally, we take advantage of a large and diverse set of methods and investigate the effectiveness of applying meta-learning, more specifically Stacking, in order to combine the methods' strengths and overcome their limitations. In our experiments, our Stacking solutions improved over previous state-of-the-art by more than 26% in predicting death, achieving 87.1% of AUROC and MacroF1 of 73.9%. We also investigated issues related to the interpretability and reliability of the predictions produced by the most effective ML methods. Finally, we discuss the adequacy of AUROC as an evaluation metric for highly imbalanced and skewed datasets commonly found in health-related problems.

Increasing the efficiency of randomized trial estimates via linear adjustment for a prognostic score

Estimating causal effects from randomized experiments is central to clinical research. Reducing the statistical uncertainty in these analyses is an important objective for statisticians. Registries, prior trials, and health records constitute a growing compendium of historical data on patients under standard-of-care that may be exploitable to this end. However, most methods for historical borrowing achieve reductions in variance by sacrificing strict type-I error rate control. Here, we propose a use of historical data that exploits linear covariate adjustment to improve the efficiency of trial analyses without incurring bias. Specifically, we train a prognostic model on the historical data, then estimate the treatment effect using a linear regression while adjusting for the trial subjects’ predicted outcomes (their prognostic scores). We prove that, under certain conditions, this prognostic covariate adjustment procedure attains the minimum variance possible among a large class of estimators. When those conditions are not met, prognostic covariate adjustment is still more efficient than raw covariate adjustment and the gain in efficiency is proportional to a measure of the predictive accuracy of the prognostic model above and beyond the linear relationship with the raw covariates. We demonstrate the approach using simulations and a reanalysis of an Alzheimer’s disease clinical trial and observe meaningful reductions in mean-squared error and the estimated variance. Lastly, we provide a simplified formula for asymptotic variance that enables power calculations that account for these gains. Sample size reductions between 10% and 30% are attainable when using prognostic models that explain a clinically realistic percentage of the outcome variance.

Development and Validation of a Nomogram for Predicting the Risk of Adverse Cardiovascular Events in Patients with Coronary Artery Ectasia

Coronary artery ectasia (CAE) is a rare finding and is associated with poor clinical outcomes. However, prognostic factors are not well studied and no prognostication tool is available. In a derivation set comprising 729 consecutive CAE patients between January 2009 and June 2014, a nomogram was developed using Cox regression. Total of 399 patients from July 2014 to December 2015 formed the validation set. The primary outcome was 5-year major adverse cardiovascular events (MACE), a component of cardiovascular death and nonfatal myocardial infarction. Besides the clinical factors, we used quantitative coronary angiography (QCA) and defined QCA classification of four types, according to max diameter (< or ≥5 mm) and max length ratio (ratio of lesion length to vessel length, < or ≥1/3) of the dilated lesion. A total of 27 cardiovascular deaths and 41 nonfatal myocardial infarctions occurred at 5-year follow-up. The nomogram effectively predicted 5-year MACE risk using predictors including age, prior PCI, high sensitivity C-reactive protein, N-terminal pro-brain natriuretic peptide, and QCA classification (area under curve [AUC] 0.75, 95% CI 0.68–0.82 in the derivation set; AUC 0.71, 95% CI 0.56–0.86 in the validation set). Patients were classified as high-risk if prognostic scores were ≥155 and the Kaplan–Meier curves were well separated (log-rank p < 0.001 in both sets). Calibration curve and Hosmer–Lemeshow test indicated similarity between predicted and actual 5-year MACE survival (p = 0.90 in the derivation and p = 0.47 in the validation set). This study developed and validated a simple-to-use method for assessing 5-year MACE risk in patients with CAE.

Postoperative Radiotherapy for Endometrial Cancer in Elderly (≥80 Years) Patients: Oncologic Outcomes, Toxicity, and Validation of Prognostic Scores

Endometrial cancer is a common malignancy in elderly women that are more likely to suffer from limiting medical comorbidities. Given this narrower therapeutic ratio, we aimed to assess the oncologic outcomes and toxicity in the adjuvant setting. Out of a cohort of 975 women, seventy patients aged ≥ 80 years, treated with curative postoperative radiotherapy (RT) for endometrial cancer between 2005 and 2021, were identified. Outcomes were assessed using Kaplan–Meier-analysis and comorbidities using the Charlson Comorbidity Index and G8 geriatric score. The overall survival at 1-, 2- and 5-years was 94.4%, 82.6%, and 67.6%, respectively, with significant correlation to G8 score. At 1- and 5-years, the local control rates were 89.5% and 89.5% and distant control rates were 86.3% and 66.9%, respectively. Severe (≥grade 3) acute toxicity was rare with gastrointestinal (2.9%), genitourinary (1.4%), and vaginal disorders (1.4%). Univariate analysis significantly revealed inferior overall survival with lower RT dose, G8 score, hemoglobin levels and obesity, while higher grading, lymphangiosis, RT dose decrease and the omission of chemotherapy reduced distant control. Despite older age and additional comorbidities, elderly patients tolerated curative treatment well. The vast majority completed treatment as planned with very low rates of acute severe side-effects. RT offers durable local control; however, late distant failure remains an issue.

Usefulness of Inflammation-Based Prognostic Scores in Patients with Surgically Treated Pancreatic Ductal Adenocarcinoma

In this study, we evaluated the prognostic value of inflammation-based prognostic scores in patients undergoing curative surgery for pancreatic ductal adenocarcinoma (PDAC). A retrospective analysis was conducted for 914 patients undergoing curative surgical resection for PDAC between January 2011 and April 2016. Inflammation-based scores of modified Glasgow Prognostic Score (mGPS), neutrophil-lymphocyte ratio, and platelet-lymphocyte ratio were assessed. mGPS was classified as high (1 or 2) or low (0). Median age was 63 (range, 33–88) years; 538 patients (58.9%) were male. A high mGPS was independently associated with poor overall survival (OS) and disease-free survival (DFS) (median OS: 25.4 months vs. 20.4 months, p = 0.001; median DFS: 11.6 months vs. 9.3 months, p = 0.002), poor OS in patients with TNM stage I PDAC (44 months vs. 24.8 months, p = 0.001), and poor OS and DFS in patients with tumors located at the pancreatic head or uncinate process (OS: 25.4 months vs. 20.4 months; p = 0.007, DFS: 11.4 months vs. 8.87 months; p = 0.005). Preoperative mGPS was a significant prognostic factor for PDAC after curative resection; thus, mGPS can be a useful prognostic predictive factor in patients with TNM stage I PDAC, especially for tumors located at the head and uncinate.

Facility profiling under competing risks using multivariate prognostic scores: Application to kidneytransplant centers

The performance of health care facilities (e.g. hospitals, transplant centers, etc.) is often evaluated through time-to-event outcomes. In this paper, we consider the case where, for each subject, the failure event is due to one of several mutually exclusive causes (competing risks). Since the distribution of patient characteristics may differ greatly by the center, some form of covariate adjustment is generally necessary in order for center-specific outcomes to be accurately compared (to each other or to an overall average). We propose a weighting method for comparing facility-specific cumulative incidence functions to an overall average. The method directly standardizes each facility’s non-parametric cumulative incidence function through a weight function constructed from a multivariate prognostic score. We formally define the center effects and derive large-sample properties of the proposed estimator. We evaluate the finite sample performance of the estimator through simulation. The proposed method is applied to the end-stage renal disease setting to evaluate the center-specific pre-transplant mortality and transplant cumulative incidence functions from the Scientific Registry of Transplant Recipients.

Prognostic scores and early management of septic patients in the emergency department of a secondary hospital: results of a retrospective study

Background Sequential Organ Failure Assessment (SOFA) and other illness prognostic scores predict adverse outcomes in critical patients. Their validation as a decision-making tool in the emergency department (ED) of secondary hospitals is not well established. The aim of this study was to compare SOFA, NEWS2, APACHE II, and SAPS II scores as predictors of adverse outcomes and decision-making tool in ED. Methods Data of 121 patients (age 73 ± 10 years, 58% males, Charlson Comorbidity Index 5.7 ± 2.1) with a confirmed sepsis were included in a retrospective study between January 2017 and February 2020. Scores were computed within the first 24 h after admission. Primary outcome was the occurrence of either in-hospital death or mechanical ventilation within 7 days. Secondary outcome was 30-day all-cause mortality. Results Patients older than 64 years (elderly) represent 82% of sample. Primary and secondary outcomes occurred in 40 and 44%, respectively. Median 30-day survival time of dead patients was 4 days (interquartile range 1–11). The best predictive score based on the area under the receiver operating curve (AUROC) was SAPS II (0.823, 95% confidence interval, CI, 0.744–0.902), followed by APACHE II (0.762, 95% CI 0.673–0.850), NEWS2 (0.708, 95% CI 0.616–0.800), and SOFA (0.650, 95% CI 0.548–0.751). SAPS II cut-off of 49 showed the lowest false-positive rate (12, 95% CI 5–20) and the highest positive predictive value (80, 95% CI 68–92), whereas NEWS2 cut-off of 7 showed the lowest false-negative rate (10, 95% CI 2–19) and the highest negative predictive value (86, 95% CI 74–97). By combining NEWS2 and SAPS II cut-offs, we accurately classified 64% of patients. In survival analysis, SAPS II cut-off showed the highest difference in 30-day mortality (Hazards Ratio, HR, 5.24, 95% CI 2.99–9.21, P < 0.001). Best independent negative predictors of 30-day mortality were body temperature, mean arterial pressure, arterial oxygen saturation, and hematocrit levels. Positive predictors were male sex, heart rate and serum sodium concentration. Conclusions SAPS II is a good prognostic tool for discriminating high-risk patient suitable for sub-intensive/intensive care units, whereas NEWS2 for discriminating low-risk patients for low-intensive units. Our results should be limited to cohorts with a high prevalence of elderly or comorbidities.

Value of ultrasound in grading the severity of sarcopenia in patients with hepatic cirrhosis

Background Ultrasound is emerging as an efficient significant method for measuring muscle mass in patients with liver cirrhosis. It has been applied in numerous studies as an accurate measuring tool for the muscles of the limbs. This study was conducted to assess the severity of sarcopenia in liver cirrhosis patients, through utilizing ultrasound in measuring the cross-sectional area and consequently estimating the muscle mass of both the upper and lower limb muscles, than correlating the results with hand grip strength as representative of functional status. Also, the severity of sarcopenia was correlated with conventional prognostic scores for liver cirrhosis, like Child or MELD scores and detecting its effect on the duration of hospital stay and mortality. Results This study was conducted on 101 liver cirrhosis patients who were admitted to the internal medicine hospital, 30 healthy participants were added as a control group. Using the FNIH (Foundation for the national Institutes of health) cuff off of hand grip (< 26 kg in male and < 16 kg in female, Quadriceps muscle index cutoff was estimated to be(1.67 cm/m2 for male and 1.58 cm/m2 for female). Ultrasound (mid upper arm, mid-thigh and Quadriceps muscle index) showed significant indirect correlation with (Child even in Child A and MELD) scores, as well as with the duration of hospital stay. Also, they showed a direct correlation with HGS. Conclusion Sarcopenia in cirrhotic patients assessed by ultrasonography of (mid upper arm, mid-thigh muscle thickness) and HGS are independent predictors of disease severity and poor outcome, which is assessed by high Child and MELD scores. Also, ultrasound and HGS are straightforward bedside techniques used for assessment of sarcopenia.

Usefulness of arterial spin labeling perfusion as an initial evaluation of status epilepticus

This study aimed to evaluate the sensitivity and prognostic value of arterial spin labeling (ASL) in a large group of status epilepticus (SE) patients and compare them with those of other magnetic resonance (MR) sequences, including dynamic susceptibility contrast (DSC) perfusion imaging. We retrospectively collected data of patients with SE in a tertiary center between September 2016 and March 2020. MR images were visually assessed, and the sensitivity for the detection of SE and prognostication was compared among multi-delay ASL, DSC, fluid-attenuated inversion recovery (FLAIR), and diffusion-weighted imaging (DWI). We included 51 SE patients and 46 patients with self-limiting seizures for comparison. Relevant changes in ASL were observed in 90.2% (46/51) of SE patients, a percentage higher than those for DSC, FLAIR, and DWI. ASL was the most sensitive method for initial differentiation between SE and self-limiting seizures. The sensitivity of ASL for detecting refractory SE (89.5%) or estimating poor outcomes (100%) was higher than those of other MR protocols or electroencephalography and comparable to those of clinical prognostic scores, although the specificity of ASL was very low as 9.4% and 15.6%, respectively. ASL showed a better ability to detect SE and predict the prognosis than other MR sequences, therefore it can be valuable for the initial evaluation of patients with SE.