Positive Predictive Value Research Articles

Diagnostic accuracy of Agile-4 score for liver cirrhosis in patients with metabolic dysfunction-associated steatotic liver disease. A systematic review and meta-analysis of diagnostic test accuracy studies.

A novel noninvasive score, Agile-4 score, combining liver stiffness measurements, aspartate aminotransferase/alanine aminotransferase, platelet count, diabetes status and sex has been developed for the identification of cirrhosis in patients with metabolic dysfunction-associated steatotic liver disease (MASLD). We assessed the performance of Agile-4 for ruling-in/out liver cirrhosis in MASLD patients. We searched Medline, Cochrane library, Web of science, Scopus and Echosens website up to May 2024. Eligible studies assessed the accuracy of Agile-4 for ruling-in (≥0.565) and ruling-out (<0.251) liver cirrhosis, using biopsy as the reference standard, at predefined thresholds. We calculated pooled sensitivity and specificity estimates for both Agile-4 thresholds alongside 95% confidence intervals following bivariate random-effect models. We assessed the risk of bias using Quality Assessment of Diagnostic Accuracy Studies-2 tool. We included seven studies with 6037 participants. An Agile-4 score ≥0.565 yielded a pooled specificity of 0.93 (95% CI, 0.86-0.97). Similarly, an Agile-4 score <0.251 excluded cirrhosis with a summary sensitivity of 0.90 (0.80-0.95). Assuming a cirrhosis prevalence of 30%, the positive predictive value (PPV) for ruling-in cirrhosis was 80%, while the negative predictive value for ruling-out cirrhosis was 95%. Most studies were at high or unclear risk for bias due to concerns regarding patient selection and the blinding status of Agile-4 score interpretation in relation to biopsy results. Agile-4 score performs well for ruling-in/out liver cirrhosis in MASLD patients. Owing to the relatively low PPV, sequential application of the Agile-4 after fibrosis-4 index (FIB-4) testing might further enhance its performance.

XGBoost-SHAP-based interpretable diagnostic framework for knee osteoarthritis: a population-based retrospective cohort study

ObjectiveTo use routine demographic and clinical data to develop an interpretable individual-level machine learning (ML) model to diagnose knee osteoarthritis (KOA) and to identify highly ranked features.MethodsIn this retrospective, population-based cohort study, anonymized questionnaire data was retrieved from the Wu Chuan KOA Study, Inner Mongolia, China. After feature selections, participants were divided in a 7:3 ratio into training and test sets. Class balancing was applied to the training set for data augmentation. Four ML classifiers were compared by cross-validation within the training set and their performance was further analyzed with an unseen test set. Classifications were evaluated using sensitivity, specificity, positive predictive value, negative predictive value, accuracy, area under the curve(AUC), G-means, and F1 scores. The best model was explained using Shapley values to extract highly ranked features.ResultsA total of 1188 participants were investigated in this study, among whom 26.3% were diagnosed with KOA. Comparatively, XGBoost with Boruta exhibited the highest classification performance among the four models, with an AUC of 0.758, G-means of 0.800, and F1 scores of 0.703. The SHAP method reveals the top 17 features of KOA according to the importance ranking, and the average of the experience of joint pain was recognized as the most important features.ConclusionsOur study highlights the usefulness of machine learning in unveiling important factors that influence the diagnosis of KOA to guide new prevention strategies. Further work is needed to validate this approach.

Reliability and Optimal Cut-Off Points of the Test for Upper Limb Apraxia (TULIA) for Spanish-Speaking Post-Stroke Patients.

To evaluate the reliability, identify the optimal cut-off points, and determine the diagnostic accuracy of the TULIA Apraxia test in a sample of Spanish-speaking post-stroke patients. Cross-sectional. Public primary care. 201 post-stroke patients. Reliability was assessed using Cronbach's alpha. Cut-off points were identified for each TULIA Apraxia sub-test and stratified for age (≤65 years, > 65 years) using ROC curve analysis, area under the curve, and Youden index. Diagnostic accuracy was evaluated using sensitivity, specificity, and predictive values. Cronbach's alpha of the sub-tests varied from 0.716 (95% CI 0.653-0.772) for pantomime intransitive to 0.824 (95% CI 0.784-0.858) for imitation non-symbolic. Cut-off points ranged from ≤25 to ≤35 points. For individuals aged ≤65 years, the cut-offs that best balanced sensitivity and specificity were ≤25 points for detecting alterations in imitation transitive and ≤33 points for pantomime non-symbolic. For individuals over 65 years, the imitation transitive cut-off (≤26 points) presented the most adjusted balance. The minimum positive predictive value was 0.667 (95% CI 0.542-0.777), indicating that the TULIA Apraxia tests correctly classified more than 67% of patients with apraxia. The minimum negative predictive value was 0.763 (95% CI 0.644-0.859), showing that the test accurately identified more than 76% of patients without apraxia. The TULIA Apraxia test is reliable in Spanish-speaking post-stroke patients. Optimal cut-off points, along with their respective sensitivity and specificity values, exhibit adequate test accuracy. Predictive values indicate that the test correctly identifies individuals with and without apraxia.

Obstetric-specific compared to general early warning system for predicting severe postpartum maternal morbidity.

Severe maternal morbidity is a major global health concern, and early identification of at-risk postpartum women is essential to improving outcomes. We aimed to compare the predictive values of the Modified Early Obstetric Warning System (MEOWS) versus the non-obstetric general Early Warning System (EWS) for predicting severe maternal morbidity in postpartum women. We retrospectively reviewed hospital documentation of 723 postpartum women admitted to the obstetric high dependency unit between October 2020 and March 2021. Severe maternal morbidity was defined using the American College of Obstetricians and Gynecologists' criteria. We assessed the sensitivity, specificity, positive and negative predictive values, as well as positive and negative likelihood ratios, of the MEOWS and the EWS for predicting severe postpartum maternal morbidity. Twenty-four (3.3%) women included in the study met the criteria for severe maternal morbidity. Hypertensive complications and obstetric haemorrhage were the most prevalent causes of maternal morbidity. The sensitivity of the MEOWS was 92%, specificity 62%, positive predictive value 8%, and negative predictive value 100%. The positive likelihood ratio was 2.4, while the negative likelihood ratio was 0.1. In comparison, the EWS had a sensitivity of 63%, specificity of 66%, positive predictive value of 6%, and negative predictive value of 98%. The positive likelihood ratio for the EWS was 1.8, and the negative likelihood ratio was 0.6. The obstetric-specific early warning system proved to be superior for the early prediction of severe postpartum maternal morbidity compared to the general non-obstetric warning system.

Diagnostic performance of MRI-guided vacuum-assisted breast biopsy (VABB): an essential but still underused technique.

Magnetic resonance imaging (MRI)-guided vacuum-assisted breast biopsy (VABB) is an increasingly requested procedure, but it implies training and experience both in its execution and in determining radiological-pathological concordance and is therefore performed in dedicated breast centers. The purpose of this study is to evaluate the diagnostic performance of MRI-guided vacuum-assisted biopsy and to determine the upgrade rate after surgery or follow-up. We retrospectively evaluated all consecutive patients with suspicious MRI findings without corresponding mammographic and ultrasonographic findings who underwent MRI-guided vacuum-assisted breast biopsy (VABB) at our Institution from November 2020 to March 2023. We determined the sensitivity (SE), specificity (SP), positive predictive value (PPV), negative predictive value (NPV) and accuracy of the procedure; we also assessed upgrade rate to malignancies using surgery or at least 1-year negative follow-up as reference standard. Fisher's exact test was used to evaluate the correlation between enhancement size and type (mass/non-mass) and histological outcomes. A total of 121 patients with 122 suspicious breast lesions have been included. 29.5% (n = 36) of these lesions were classified as malignant (B5), 23% (n = 28) were lesions with uncertain malignant potential (B3 lesions), and 47.5% (n =58) were benign (B2). Among B5 lesions, 47.22% (n =17) were ductal carcinomas in situ (DCIS) and 52.77% (n =19) were invasive carcinomas. Among patients with already diagnosed breast cancer (n = 36), MRI-guided VABB identified additional foci of disease in 36.1% (n = 13) of the cases, specifically 10 foci on the same breast and 3 in the contralateral breast. Accuracy of MRI-guided VABB was 96.7%, SE was 90%, SP was 100%, PPV was 100%, and NPV was 95.3%. 4 benign lesions (B2 and B3) were upgraded to B5 lesions after surgery or follow-up; the upgrade rate to malignancies was 3.28%. Fisher's exact test showed a significant association between enhancement size and histological outcomes (OR = 2.38, p = 0.046), while enhancement type was not significantly correlated (OR = 0.88, p = 0.841). No major complications have been reported. MRI-guided VABB has proven to be a mini-invasive, safe, and accurate procedure for the diagnostic work-up of suspected breast lesions, which can help in the management of patients aiding in the correct surgical decisional process.

Digital monitoring of medication safety in children: an investigation of ADR signalling techniques in Malaysia

BackgroundDigital solutions can help monitor medication safety in children who are often excluded in clinical trials. The lack of reliable safety data often leads to either under- or over-dose of medications during clinical management which make them either not responding well to treatment or susceptible to adverse drug reactions (ADRs).AimThis study investigated ADR signalling techniques to detect serious ADRs in Malaysian children aged from birth to 12 years old using an electronic ADRs’ database.MethodsFour techniques (Proportional Reporting Ratio (PRR), Reporting Odds Ratio (ROR), Bayesian Confidence Propagation Neural Network (BCPNN) and Multi-item Gamma Poisson Shrinker (MGPS)) were tested on ADR reports submitted to the National Pharmaceutical Regulatory Agency between 2016 and 2020. Sensitivity, Specificity, Positive Predictive Value (PPV) and Negative Predictive Value (NPV) of the techniques were compared.ResultsA total of 31 medicine-Important Medical Event pairs were found and examined among the 3152 paediatric ADR reports. Three techniques (PRR, ROR, MGPS) signalled oculogyric crisis and dystonia for metoclopramide. BCPNN and MGPS signalled angioedema for paracetamol, amoxicillin and ibuprofen. Similar performances were found for PRR, ROR and BCPNN (sensitivity of 12%, specificity of 100%, PPV of 100% and NPV of 21%). MGPS revealed the highest sensitivity (20%) and NPV (23%), as well as similar specificity and PPV (100%).ConclusionsThis study suggests that medication safety signalling techniques could be applied on electronic health records to monitor medication safety issues in children. Clinicians and medication safety specialist could prioritise the signals for further clinical consideration and prompt response.

Evaluation of three rapid assays for detecting Mycoplasma pneumoniae in nasopharyngeal specimens

During the 2023 autumn–winter period in China, Mycoplasma pneumoniae (MP) infections have increased. To address this, rapid and accurate MP DNA detection methods are crucial. Three nucleic acid detection assays (Ustar, Coyote Flash10, Coyote Flash 20) that are widely used in China are currently being evaluated for their effectiveness in detecting MP DNA in nasopharyngeal specimens. Reference standard materials for MP and a total of 35 NPS collected from Peking Union Medical College Hospital were tested using the Ustar, Coyote Flash10 and Coyote Flash 20 assays to assess analytical sensitivity, analytical specificity, diagnostic performance and workflow. The assays showed differing limits of detection (LOD) based on the absolute quantification of reference standards, with LODs of 500 copies/mL for the Ustar assays and 200 copies/mL for both Coyote Flash10 and Coyote Flash 20 assays. Additionally, all three assays displayed excellently analytical specificity in detecting MP DNA.The clinical correlation analysis demonstrated that the Ustar assay exhibited a sensitivity of 90.00%, specificity of 100%, positive predictive value (PPV) of 100%, and negative predictive value (NPV) of 62.50%. In contrast, both the Coyote Flash10 and Coyote Flash 20 assays displayed perfect diagnostic accuracy with 100% sensitivities, specificities, PPVs, and NPVs. Despite variations in detection principles, sample volume, and pre-preparation among the three assays, they all had a turnaround time of less than 30 min with low-throughput processing. Overall, all three rapid nucleic acid detection assays displayed excellent clinical performance in detecting MP DNA, offering a solid foundation for the quick clinical diagnosis of MP infection.

A potential predictive model based on machine learning and CPD parameters in elderly patients with aplastic anemia and myelodysplastic neoplasms

BackgroundAplastic anemia (AA) and myelodysplastic neoplasms (MDS) have similar peripheral blood manifestations and are clinically characterized by reduced hematological triad. It is challenging to distinguish and diagnose these two diseases. Hence, utilizing machine learning methods, we employed and validated an algorithm that used cell population data (CPD) parameters to diagnose AA and MDS.MethodsIn this study, CPD parameters were obtained from the Beckman Coulter DxH800 analyzer for 160 individuals diagnosed with AA or MDS through a comprehensive retrospective analysis. The individuals were unselectively assigned to a training cohort (77%) and a testing cohort (23%). Additionally, an external validation cohort consisting of eighty-six elderly patients with AA and MDS from two additional centers was established. The discriminative parameters were carefully analyzed through univariate analysis, and the most predictive variables were selected using least absolute shrinkage and selection operator (LASSO) regression. Six machine learning algorithms were utilized to compare the performance of forecasting AA and MDS patients. The area under the curves (AUCs), calibration curves, decision curves analysis (DCA), and shapley additive explanations (SHAP) plots were employed to interpret and assess the model’s predictive accuracy, clinical utility, and stability.ResultsAfter the comparative evaluation of various models, the logistic regression model emerged as the most suitable machine learning model for predicting the probability of AA and MDS, which utilized five principal variables (age, MNVLY, SDVLY, MNLALSEGC, and MNCEGC) to accurately estimate the risk of these diseases. The best model delivered an AUC of 0.791 in the testing cohort and had a high specificity (0.850) and positive predictive value (0.818). Furthermore, the calibration curve indicated excellent agreement between actual and predicted probabilities. The DCA curve further supported the clinical utility of our model and offered significant clinical advantages in guiding treatment decisions. Moreover, the model’s performance was consistent in an external validation group, with an AUC of 0.719.ConclusionsWe developed a novel model that effectively distinguished elderly patients with AA and MDS, which had the potential to provide physicians assistance in early diagnosis and the proper treatment for the elderly.

Radical prostatectomy without prostate biopsy based on a noninvasive diagnostic strategy: a prospective single-center study.

Prostate biopsy is the most common approach for diagnosing prostate cancer (PCa); however, it has inherent limitations, such as the invasive procedure, postoperative complications, and false negative results. We aimed to provide a noninvasive diagnostic strategy for patients with highly suspected PCa and to evaluate the feasibility of performing biopsy-spared radical prostatectomy. This prospective study included a total of 57 patients between November 10, 2022, and December 1, 2023. All 57 patients underwent radical prostatectomy without prior prostate biopsy based on a noninvasive diagnostic strategy consisting of a diagnostic prediction model [comprised of the prostate imaging-reporting and data system (PI-RADS) score and prostate-specific antigen density (PSAD)] and the 18F-prostate-specific membrane antigen (PSMA)-1007 positron emission tomography (PET)/computed tomography (CT) examination. The primary endpoint was the positive predictive value (PPV) of clinically significant PCa [the International Society of Urological Pathology (ISUP) grade ≥2, Gleason score ≥3 + 4]. The secondary endpoints were a PPV of any-grade PCa (ISUP grade ≥ 1, Gleason score ≥3 + 3) and high-grade PCa (ISUP grade ≥3, Gleason score ≥4 + 3), and the false positive rate of the diagnostic strategy. Of the 371 screened patients with clinically suspected PCa, 57 patients fulfilled the criteria and consented to participate in this study. The median PSAD level was 0.56 (0.42-0.82) ng/mL2; 13 (22.8%) patients were identified as having a PI-RADS score of 4, and 44 (77.2%) patients with a PI-RADS score of 5. The median SUVmax of 18F-PSMA-1007 PET/CT was 21.6 (15.8-33.0). For the 57 enrolled patients who received radical prostatectomy directly, the PPV of clinically significant PCa was 98.2% [56/57, 95% confidence interval (CI): 90.6-100%]. Only 1.8% (1/57, 95% CI: 0.0-9.4%) of patients were diagnosed with clinically insignificant PCa (ISUP grade = 1, Gleason score = 3 + 3). The PPV of any-grade PCa and high-grade PCa were 100% and 73.7% (42/57, 95% CI: 60.3-84.5%), respectively. No one had a false positive result. We proposed a noninvasive diagnostic strategy consisting sequentially of a diagnostic prediction model and the 18F-PSMA-1007 PET/CT examination for diagnosing PCa. Despite some limitations, our initial findings suggest the potential feasibility of radical prostatectomy without prior prostate biopsy.

Evaluation of a Risk Screening Tool for Retinopathy of Prematurity (ROP) in a German Cohort

ABSTRACT Purpose To assess the efficacy of the DIGIROP-Birth algorithm in identifying infants at risk for developing retinopathy of prematurity (ROP). Methods In a retrospective study, we included preterm infants over 11 years, 2010–2020, meeting the inclusion criteria for the DIGIROP-Birth calculator (24 + 0/7 to 30 + 6/7 weeks of gestational age). We assessed the validity of DIGIROP-Birth using receiver-operating characteristic (ROC) curves and calculated area-under-curve (AUC), sensitivity, specificity, and positive and negative predictive values. Results 897 infants were included in the analysis. The median age of the first ophthalmological examination was 40 days (IQR 32–50), the median gestational age was 198 days (IQR 185–209; corresponding to 28 + 2/7 gestational weeks), median birth weight was 1000 g (IQR 790–1300). Of 897 screened children, 458 (51.1%) were diagnosed with ROP, and 34 of 897 (3.8%) required treatment. Analysis of ROP requiring treatment predicted by DIGIROP showed an AUC of 0.860 [95%-CI 0.795–0.925]. An equilibrium of sensitivity and specificity existed at a probability of 4.12%. The positive predictive value was 10.95%, and the negative predictive value was 99.36%. Independent significant peri- and postnatal risk factors were emergency cesarean section and mass blood transfusions. Conclusions The DIGIROP-Birth calculator showed good predictive power in our studied population, with an incidence of 3.79% for therapy-requiring ROP. Peri- and postnatal risk factors should be included in ROP screening.

Machine learning-driven simplification of the hypomania checklist-32 for adolescent: a feature selection approach

BackgroundThe Hypomania Checklist-32 is widely used to screen for bipolar disorder, but its length can be challenging for adolescents with manic symptoms. This study aimed to develop a shortened version of the HCL-32 tailored for adolescents using machine learning techniques.MethodsData from 2,850 adolescents (mean age 15.50 years, 68.81% female) who completed the HCL-32 were analyzed. Random forest (RF) and gradient boosting machine (GBM) algorithms were employed for feature selection. The area under the curve (AUC) was used to evaluate model performance. Receiver operating characteristic (ROC) analysis was conducted to determine optimal cutoff points for the shortened scale.ResultsAn 8-item version of the HCL-32 was derived, maintaining high predictive accuracy (AUC = 0.97). The selected items captured core symptoms of adolescent mania, including increased energy, risk-taking, and irritability. Two cutoff points were identified: a score of 3 offered high specificity (0.98) and positive predictive value (0.98), while a score of 4 provided balanced sensitivity (0.87) and specificity (0.94) with the highest overall accuracy (0.91).ConclusionsThe machine learning-driven 8-item version of the HCL-32 demonstrates strong diagnostic utility for adolescent bipolar disorder, offering a more efficient screening tool without sacrificing clinical sensitivity. This shortened scale may improve assessment feasibility and accuracy in clinical settings, addressing the unique challenges of diagnosing bipolar disorder in adolescents.

Detection of cardiac amyloidosis using machine learning on routine echocardiographic measurements

BackgroundCardiac amyloidosis (CA) is an underdiagnosed, progressive and lethal disease. Machine learning applied to common measurements derived from routine echocardiogram studies can inform suspicion of CA.ObjectivesOur objectives were to test...

The performance of JAM3/PAX1 methylation in the diagnosis of high-grade squamous intraepithelial lesions for women with high-risk HPV infection

ObjectiveTo assess the clinical value of DNA methylation measurement in exfoliated cervical cells for distinguishing high-grade squamous intraepithelial lesions (HSIL) from other cervical abnormalities.MethodsA total of 276 patients were enrolled, and general clinical information was collected. Exfoliated cervical cells were obtained to assess human papillomavirus (HPV) infection, conduct ThinPrep cytology tests (TCT), and measure methylation levels of JAM3 (△CtJ) and PAX1 (△CtP). Logistic regression was performed to identify factors significantly associated with HSIL diagnosis. A conditional inference tree model and the area under the curve (AUC) were employed to evaluate the efficacy of JAM3 and PAX1 methylation in detecting HSIL.ResultsIndependent risk factors for HSIL diagnosis included △CtJ, △CtP, atypical squamous cells of undetermined significance (ASCUS), and HPV16 infection. The conditional inference tree indicated that 96.4% of patients were non-HSIL when △CtJ > 11.66, and 99.1% were non-HSIL when △CtP > 10.97. The diagnostic performance of △CtJ/△CtP surpassed that of TCT/HPV alone. Among six methods, the combination of △CtP, TCT, and high-risk HPV (hr-HPV) testing achieved the highest sensitivity (91.2%), positive predictive value (50.0%), negative predictive value (98.6%), and AUC (0.932).ConclusionIn women with hr-HPV infection, DNA methylation analysis of cervical cytology outperformed traditional TCT or HPV testing. The combination of △CtP with TCT and HPV may offer the most accurate screening approach for HSIL.

Comparison of fine-scale malaria strata derived from population survey data collected using RDTs, microscopy and qPCR in South-Eastern Tanzania

BackgroundMalaria-endemic countries are increasingly adopting data-driven risk stratification, often at district or higher regional levels, to guide their intervention strategies. The data typically comes from population-level surveys collected by rapid diagnostic tests (RDTs), which unfortunately perform poorly in low transmission settings. Here, a high-resolution survey of Plasmodium falciparum prevalence rate (PfPR) was conducted in two Tanzanian districts using rapid diagnostic tests (RDTs), microscopy, and quantitative polymerase chain reaction (qPCR) assays, enabling the comparison of fine-scale strata derived from these different diagnostic methods.MethodsA cross-sectional survey was conducted in 35 villages in Ulanga and Kilombero districts, south-eastern Tanzania between 2022 and 2023. A total of 7,628 individuals were screened using RDTs (SD-BIOLINE) and microscopy, with two thirds of the samples further analysed by qPCR. The data was used to categorize each district and village as having very low (PfPR < 1%), low (1%≤PfPR < 5%), moderate (5%≤PfPR < 30%), or high (PfPR ≥ 30%) parasite prevalence. A generalized linear mixed model was used to analyse infection risk factors. Other metrics, including positive predictive value (PPV), sensitivity, specificity, parasite densities, and Kappa statistics were computed for RDTs or microscopy and compared to qPCR as reference.ResultsSignificant fine-scale variations in malaria risk were observed within and between the districts, with village prevalence ranging from 0% to > 50%. Prevalence varied by testing method: Kilombero was low risk by RDTs (PfPR = 3%) and microscopy (PfPR = 2%) but moderate by qPCR (PfPR = 9%); Ulanga was high risk by RDTs (PfPR = 39%) and qPCR (PfPR = 54%) but moderate by microscopy (PfPR = 26%). RDTs and microscopy classified majority of the 35 villages as very low to low risk (18–21 villages). In contrast, qPCR classified most villages as moderate to high risk (29 villages). Using qPCR as the reference, PPV for RDTs and microscopy ranged from as low as < 20% in very low transmission villages to > 80% in moderate and high transmission villages. Sensitivity was 62% for RDTs and 41% for microscopy; specificity was 93% and 96%, respectively. Kappa values were 0.7 for RDTs and 0.5 for microscopy. School-age children (5–15 years) had higher malaria prevalence and parasite densities than adults (P < 0.001). High-prevalence villages also had higher parasite densities (Spearman r = 0.77, P < 0.001 for qPCR; r = 0.55, P = 0.003 for microscopy).ConclusionThis study highlights significant fine-scale variability in malaria burden within and between the study districts and emphasizes the variable performance of the testing methods when stratifying risk at local scales. While RDTs and microscopy were effective in high-transmission areas, they performed poorly in low-transmission settings; and classified most villages as very low or low risk. In contrast, qPCR classified most villages as moderate or high risk. The findings emphasize that, where precise mapping and effective targeting of malaria are required in localized settings, tests must be both operationally feasible and highly sensitive. Furthermore, when planning microstratification efforts to guide local control measures, it is crucial to carefully consider both the strengths and limitations of the available data and the testing methods employed.

Description of a New miRNA Signature for the Surgical Management of Thyroid Nodules

Background: The diagnosis of malignant thyroid nodules is mainly based on the fine-needle aspiration biopsy (FNAB). To improve the detection of malignant nodules, different molecular tests have been developed. We present a new molecular signature based on altered miRNA expressions and specific mutations. Methods: This is a prospective non-interventional study, including all Bethesda categories, carried out on an FNAB sampled in suspicious nodule(s) during thyroidectomy. miRNA quantification and mutations detection were performed. The reference diagnosis was the pathological assessment of the surgical specimen. Different classification algorithms were trained with molecular data to correctly classify the samples. Results: A total of 294 samples were recorded and randomly divided in two equal groups. The random forest algorithm showed the highest accuracy and used mostly miRNAs to classify the nodules. The sensitivity and the specificity of our signature were, respectively, 76% and 96%, and the positive and negative predictive values were both 90% (disease prevalence of 30%). Conclusions: We have identified a molecular classifier that combines miRNA expressions with mutations detection. This signature could potentially help clinicians, as complementary to the Bethesda classification, to discriminate indeterminate FNABs.

Artificial intelligence in triage of COVID-19 patients

In 2019, COVID-19 began one of the greatest public health challenges in history, reaching pandemic status the following year. Systems capable of predicting individuals at higher risk of progressing to severe forms of the disease could optimize the allocation and direction of resources. In this work, we evaluated the performance of different Machine Learning algorithms when predicting clinical outcomes of patients hospitalized with COVID-19, using clinical data from hospital admission alone. This data was collected during a prospective, multicenter cohort that followed patients with respiratory syndrome during the pandemic. We aimed to predict which patients would present mild cases of COVID-19 and which would develop severe cases. Severe cases were defined as those requiring access to the Intensive Care Unit, endotracheal intubation, or even progressing to death. The system achieved an accuracy of 80%, with Area Under Receiver Operating Characteristic Curve (AUC) of 91%, Positive Predictive Value of 87% and Negative Predictive Value of 82%. Considering that only data from hospital admission was used, and that this data came from low-cost clinical examination and laboratory testing, the low false positive rate and acceptable accuracy observed shows that it is feasible to implement prediction systems based on artificial intelligence as an effective triage method.

Gasdermin D (GSDM D) as a Potential Diagnostic Biomarker in Bladder Cancer: New Perspectives in Detection

Background: Traditional methods of bladder cancer (BC) diagnosis include clinical examination, imaging, urine tests, cystoscopy, and biopsy. Due to the complexity of detection, diagnostic markers of bladder cancer measured in blood are still being sought. The pathogenesis of BC is complex and depends on many factors. However, the available literature data suggest that gasdermin D (GSDM D) has an influence in the pathogenesis of cancers. This study is the first that assesses the significance and diagnostic utility of serum GSDM D in bladder cancer. Methods: A total of 80 serum samples were obtained and analysed from healthy volunteers (C) and bladder cancer patients. The cancer patients were further classified into two groups, low-grade (LG) and high-grade (HG) bladder cancer, according to the TNM classification. The serum levels of GSDM D, CEA, and CA19-9 were assessed following the manufacturer’s instructions using immunoassay techniques. Results: The concentrations of GSDM D were nearly twice as high in the serum of BC patients compared to controls. Additionally, the median of GSDM D concentration was notably elevated in LG and HG bladder cancer than in C. In the total study group, the GSDM D concentration correlated with CRP and CEA levels. The diagnostic sensitivity of GSDM D was higher than that of CEA, but slightly lower in comparison to CA19-9. Moreover, the negative predictive value of GSDM D was the highest among all tested markers. The highest positive predictive value and diagnostic accuracy were observed for CEA, while the accuracy of GSDM D was comparable. GSDM D had an AUC value of 0.741, which was similar to the AUC value of CEA. Conclusions: GSDM D in serum appears to be a valuable diagnostic biomarker, especially when its measurement is used in conjunction with other markers such as CEA. Its high sensitivity makes it effective for the early detection of bladder cancer.

Evaluation of Neonatal Screening Programs for Tyrosinemia Type 1 Worldwide

In The Netherlands, newborn screening (NBS) for tyrosinemia type 1 (TT1) uses dried blood spot (DBS) succinylacetone (SUAC) as a biomarker. However, high false-positive (FP) rates and a false-negative (FN) case show that the Dutch TT1 NBS protocol is suboptimal. In search of optimization options, we evaluated the protocols used by other NBS programs and their performance. We distributed an online survey to NBS program representatives worldwide (N = 41). Questions focused on the organization and performance of the programs and on changes since implementation. Thirty-three representatives completed the survey. TT1 incidence ranged from 1/13,636 to 1/750,000. Most NBS samples are taken between 36 and 72 h after birth. Most used biomarkers were DBS SUAC (78.9%), DBS Tyrosine (Tyr; 5.3%), or DBS Tyr with second tier SUAC (15.8%). The pooled median cut-off for SUAC was 1.50 µmol/L (range 0.3–7.0 µmol/L). The median cut-off from programs using laboratory-developed tests was significantly higher (2.63 µmol/L) than the medians from programs using commercial kits (range 1.0–1.7 µmol/L). The pooled median cut-off for Tyr was 216 µmol/L (range 120–600 µmol/L). Overall positive predictive values were 27.3% for SUAC, 1.2% for Tyr solely, and 90.1% for Tyr + SUAC. One FN result was reported for TT1 NBS using SUAC, while three FN results were reported for TT1 NBS using Tyr. The NBS programs for TT1 vary worldwide in terms of analytical methods, biochemical markers, and cut-off values. There is room for improvement through method standardization, cut-off adaptation, and integration of new biomarkers. Further enhancement is likely to be achieved by the application of post-analytical tools.

Put Down the ACE: Low Clinical Utility for Angiotensin-Converting Enzyme Levels in Sarcoidosis: A Single-Center Retrospective Cohort Study

Background/Objectives: ACE (angiotensin-converting enzyme) is considered a serological marker of sarcoidosis as elevated levels have been reported in 30–80% of patients. However, elevated ACE levels are also encountered in other medical conditions, and the clinical correlation between ACE levels and disease activity in sarcoidosis is disputable as well. To determine the significance of elevated ACE levels in the diagnosis and follow-up of sarcoidosis patients. Methods: All electronic patient records in which an ACE level was recorded in a large tertiary hospital were identified using a computerized algorithm. Medical diagnoses, ACE numerical values, and clinical data were also automatically extracted. Furthermore, all records with a diagnosis of sarcoidosis were manually reviewed for ascertainment of the diagnosis and searched for additional clinical manifestations and treatment responses. Results: A total of 1416 records with a documented ACE level were found in the database, and 146 of the records had a diagnosis of sarcoidosis in the medical record. However, the diagnosis was excluded in 27 of these cases after a manual review of the records. Elevated ACE levels were most commonly encountered among patients with sarcoidosis, non-Hodgkin’s lymphoma, cirrhosis, and interstitial lung disease. Elevated ACE levels had a positive predictive value of 12.76% and a negative predictive value of 94.6% for the diagnosis of sarcoidosis in our cohort, with a sensitivity of 63.5% and a specificity of 59.5%. Among patients with sarcoidosis, ACE levels around the time of diagnosis were higher than ACE levels in remission. However, a paired analysis did not find a statistically significant difference in ACE levels between the two timepoints. A positive correlation between lack of cardiac involvement and elevated ACE levels was found on multivariate analysis. Conclusions: ACE levels are a non-specific serological marker with low specificity and sensitivity for sarcoidosis and a poor positive predictive value, but with a negative predictive value of 94.6%. Furthermore, elevated ACE levels correlated poorly with disease activity in our cohort.

Comparing How Three Frailty Scales Predict Negative Outcomes in Trauma Patients with Rib Fractures

Frailty is a risk factor for adverse outcomes after injury. Herein, we compared three frailty scales: the Canadian Study of Health and Aging clinical frailty scale, the rib fracture frailty index (RFFI) and the modified frailty index-5, to assess which scale is most applicable in predicting risk for negative outcomes in older patients with rib fractures. Patients ≥65 admitted for rib fractures were retrospectively scored for frailty using the RFFI, Canadian Study of Health and Aging clinical frailty scale, and modified frailty index-5. Outcomes examined were in-hospital mortality, pneumonia, in-hospital intubation, hospital length of stay, and discharge to skilled nursing facilities. Areas under the curve, sensitivity, specificity, negative predictive value, and positive predictive value were determined for each frailty scale with each outcome. Agreement was determined using Fleiss' Kappa. P<0.05 was considered significant. Three hundred forty-one patients were included. All three scales demonstrated similar predictive abilities for the measured outcomes. RFFI predicted mortality and pneumonia 70% of the time. All three scales predicted discharge to skilled nursing facilities 60% of the time. The concordance for all three frailty scales was 241/341 (70.7%). Fleiss Kappa was 0.40 [0.34-0.46] (P<0.001), indicating a fair to moderate agreement. The predictive ability of all three scales was higher in patients 65-74y old than in patients ≥75. Overall, no scale appeared to significantly outperform the others by areas under the curve estimation. Interrater reliability was higher in the 65 to 74-y-old population compared to the 75 and older population.