STARD 2015 guidelines for reporting diagnostic accuracy studies: explanation and elaboration

Performance of microscopy and rapid test for the diagnosis of mixed malaria in an endemic area of Colombia

BACKGROUND

Mixed infection by Plasmodium (MIP) occurs when more than one species are infecting one individual at the same time.

METHODS

Frequency of MIP was determined in a cohort of patients with acute febrile syndrome. Diagnostic performance of microscopy (MI) and a rapid diagnostic test (RDT) was evaluated against real-time PCR (qPCR), as reference method.

RESULTS

The frequency of MIP was 3% by qPCR, 1.4% by MI and 0.4% by RDT. The sensitivity of MI and RDT for detecting MIP was 21.43% and 15.25%, respectively, and specificity was over 99% for both techniques.

CONCLUSION

The currently available tests for the diagnosis of Plasmodium in the field, showed poor performance for detecting MIP in this study. It is necessary to improve diagnostic tests sensitivity for the correct detection of these coinfections and to monitor the elimination of parasitemia to detect possible recurrences due to inadequate treatments.

Diagnostic Accuracy of an Add-On, Blood-Based Screening Test for Colorectal Cancer in Two Established Screening Programmes

BACKGROUND

CELTiC is a blood-based test consisting of a panel of four mRNAs (CEACAM6, LGALS4, TSPAN8 and COL1A2) associated with colorectal cancer (CRC). CELTiC has a high sensitivity (90%) for detecting advanced neoplasia (AN) when compared to faecal immunochemical test (FIT)-negative subjects.

AIMS

To evaluate the diagnostic accuracy of CELTiC as an add-on test following a positive FIT in two existing CRC screening programmes.

METHODS

We designed a prospective study in FIT-positive screenees. Analyses were performed in two (partially overlapping) groups: participants with FIT ≥ 47 μg Hb/g faeces (group I) and FIT ≥ 20 μg Hb/g faeces (group II). We estimated CELTIC sensitivity and specificity in detecting AN at a pre-defined and a post hoc threshold (targeted sensitivity: 90%).

RESULTS

We included 809 participants (n = 567 in group I and n = 486 in group II). CELTiC did not reach a sensitivity of 90% at the pre-defined threshold (Group I 41%, group II 27%). At the post hoc thresholds for 90% sensitivity in detecting AN, CELTiC had a specificity of 14% (53/388; 95% CI: 10% to 17%) in group I and 12% (44/354; 95% CI: 9% to 16%) in group II.

CONCLUSIONS

In a FIT-positive population, CELTiC did not reach 90% sensitivity at the pre-defined threshold. At post hoc thresholds corresponding to 90% sensitivity, specificity was low. CELTiC cannot currently be recommended as an add-on test to detect AN in FIT-positive screenees in a CRC screening programme.

TRIAL REGISTRATION

ClinicalTrials.gov identifier: NCT04980443, NCT05205967, NCT04369053.

External Validation of an AI mHealth Tool for Gingivitis Detection among Older Adults at Daycare Centers: A Pilot Study

OBJECTIVES

Periodontal disease is a significant public health concern among older adults due to its relationship with tooth loss and systemic health disease. However, there are numerous barriers that prevent older adults from receiving routine dental care, highlighting the need for innovative screening tools at the community level. This pilot study aimed first, to evaluate the accuracy of GumAI, a new mHealth tool that uses AI and smartphones to detect gingivitis, and the user acceptance of personalized oral hygiene instructions provided through the new tool, among older adults in day-care community centers.

METHODS

Participants were invited from 3 day-care community centers. Intraoral photographs were captured and assessed by both GumAI (test) and a panel consisting of 2 calibrated periodontists and a dentist (benchmark). Mean sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), accuracy, and F1 score were calculated to determine GumAI's diagnostic performance in comparison to the benchmark. User acceptance with this tool was assessed using 2 Rasch Theory-based 5-point Likert-type questions.

RESULTS

44 participants were recruited out of 80 invited older adults. GumAI demonstrated a sensitivity of 0.93 and specificity of 0.50 compared to the panel's assessments, with a PPV of 0.90 and NPV of 0.56. The accuracy and F1 scores were 0.85 and 0.91, respectively. All participants expressed high acceptance of the process.

CONCLUSION

GumAI demonstrates high sensitivity, PPV, accuracy, and F1 score compared to the panel's assessments but falls relatively short in specificity and NPV. Despite this, the tool was highly accepted by older adults, indicating its potential to enhance gingivitis detection and oral hygiene management in community settings. Further refinements are necessary to improve specificity and validate usability measures.

CLINICAL RELEVANCE

This study may pave the way for broader applications of mHealth systems in community settings, enabling greater health coverage and addressing oral health disparities.

Diagnostic performance of procalcitonin for detecting anastomotic leak in older adults with colorectal cancer: A delayed type cross-sectional study

BACKGROUND

The utility of procalcitonin for detecting anastomotic leaks after colorectal surgery in older adults has not been well documented. As the immune system undergoes changes with age, procalcitonin levels may differ at baseline, and its diagnostic performance could vary when an anastomotic leak occurs after colorectal surgery in older adult patients with cancer. The aim of this study is to evaluate the diagnostic performance of procalcitonin on postoperative day 3 for detecting anastomotic leaks in older adults with colorectal cancer.

METHODS

We conducted a diagnostic test study on the basis of a delayed-type cross-sectional design in older adults (≥65 years old) with colorectal cancer. Postoperative day 3 procalcitonin levels were tested. The reference standard was anastomotic leak. We calculated the receiver operating characteristic curve and its area under the curve.

RESULTS

The incidence of anastomotic leak was 7.7%. On postoperative day 3, the receiver operating characteristic demonstrated an area under the curve of 0.68 (95% confidence interval, 0.58-0.78) for the prediction of an anastomotic leak using procalcitonin levels. The cutoff point with the greatest diagnostic performance, according to the Youden index, was 0.61 ng/mL, with a sensitivity of 0.69, specificity of 0.62, a positive likelihood ratio of 1.86, and a negative likelihood ratio of 0.48 for predicting an anastomotic leak. The area under the curve was 0.78 when ileostomies were excluded and 0.81 when evaluating grade C leaks.

CONCLUSION

Our study reveals that procalcitonin levels on postoperative day 3 are a poor diagnostic marker for identifying anastomotic leaks in older adults with colorectal cancer, with an improvement in performance for the grade C anastomotic leak subgroup.

Development and validation of a brief questionnaire for the joint assessment of early maltreatment and early caring experiences

BACKGROUND

Early life experiences significantly impact mental health, yet tools assessing both early maltreatment experiences (EME) and early caring experiences (ECE) remain scarce.

OBJECTIVE

This study aimed to develop and validate the self-reported Early Maltreatment & Caring Experiences (EMCE) questionnaire, a concise tool for dual assessment of these experiences.

PARTICIPANTS AND SETTING

A clinical sample of 272 adults diagnosed with major depressive disorder was recruited from an outpatient clinic in Santiago, Chile.

METHODS

Confirmatory factor analyses (CFA), internal consistency, and test-retest reliability of the EMCE were assessed. Logistic regression with 1000 bootstrap replications evaluated predictive validity for complex and severe depression. Construct validity was assessed against the Childhood Trauma Questionnaire-Short Form (CTQ-SF), and resilience was measured using the Brief Resilient Coping Scale (BRCS).

RESULTS

CFA refined the EMCE into a concise 6-item version (χ2(7) = 9.54, p = .216, RMSEA = 0.037, CFI = 0.995, TLI = 0.989) with strong psychometric properties, including good internal consistency (α = 0.73 for EME, α = 0.80 for ECE), excellent test-retest reliability (ICC = 0.86 and 0.79, respectively). Each EME point increased odds of complex depression (OR = 1.33, 95 % BC CI 1.18-1.52), while higher ECE scores reduced them (OR = 0.83, 95 % BC CI 0.74-0.94). Agreement analyses with the CTQ-SF supported construct validity.

CONCLUSIONS

The EMCE provides a reliable, concise tool for assessing early maltreatment and caregiving experiences, supporting clinicians and researchers in exploring vulnerabilities and protective factors. Its brevity ensures feasibility in clinical and research settings.

Diagnostic Accuracy of Methods Used to Detect Cracked Teeth

OBJECTIVES

Cracked tooth diagnosis is challenging due to the unknown diagnostic accuracy of tools, resulting in misdiagnosis and suboptimal treatment outcomes. The primary objective of this study was to determine the diagnostic accuracy of four commonly used visual tests in diagnosing cracked teeth based on clinical appearance, independent of patient symptoms. The secondary objective was to assess if clinical experience influences the ability to accurately identify the presence of a crack. The tertiary objective was to assess the ability of the index tests to accurately determine the location of the crack.

MATERIAL AND METHODS

The test sample included 30 teeth extracted due to a suspected crack. Index tests included macrophotography, surgical microscope, transillumination, and DIAGNOcam. Microcomputed tomography (micro-CT) served as the gold standard. Four examiners of varying experience assessed images of each tooth paired with each index test. The examiner's findings were compared against micro-CT to determine the diagnostic accuracy of index tests. The relationship between clinical experience and diagnostic accuracy was explored.

RESULTS

Transillumination demonstrated the highest accuracy (65.3%) and sensitivity (68.8%) for diagnosing cracks. Macrophotography and high-magnification microscope had the highest specificity of 92.9%. Positive predictive value (PPV) was greatest with high-magnification microscope (96.7%). The low-magnification microscope demonstrated the lowest accuracy (52.2%). Intra-rater reliability was moderate to substantial, and inter-rater reliability was fair. Experienced dentists were more accurate in detecting cracked teeth.

CONCLUSIONS

Visual diagnostic methods cannot definitively diagnose cracks. Further studies are required to explore the impact of a combination of tools in diagnosing cracked teeth.

Comparison of methods to handle missing values in a continuous index test in a diagnostic accuracy study - a simulation study

BACKGROUND

Most diagnostic accuracy studies have applied a complete case analysis (CCA) or single imputation approach to address missing values in the index test, which may lead to biased results. Therefore, this simulation study aims to compare the performance of different methods in estimating the AUC of a continuous index test with missing values in a single-test diagnostic accuracy study.

METHODS

We simulated data for a reference standard, continuous index test, and three covariates using different sample sizes, prevalences of the target condition, correlations between index test and covariates, and true AUCs. Subsequently, missing values were induced for the continuous index test, assuming varying proportions of missing values and missingness mechanisms. Seven methods (multiple imputation (MI), empirical likelihood, and inverse probability weighting approaches) were compared to a CCA in terms of their performance to estimate the AUC given missing values in the index test.

RESULTS

Under missing completely at random (MCAR) and many missing values, CCA gives good results for a small sample size and all methods perform well for a large sample size. If missing values are missing at random (MAR), all methods are severely biased if the sample size and prevalence are small. An augmented inverse probability weighting method and standard MI methods perform well with higher prevalence and larger sample size, respectively. Most methods give biased results if missing values are missing not at random (MNAR) and the correlation or the sample size and prevalence are low. Methods using the covariates improve with increasing correlation.

CONCLUSIONS

Most methods perform well if the proportion of missing values is small. Given a higher proportion of missing values and MCAR, we would recommend to conduct a CCA and standard MI methods for a small and large sample size, respectively. In the absence of better alternatives we recommend to conduct a CCA and to discuss its limitations, if the sample size is small, and missing values are M(N)AR. Standard MI methods and the augmented inverse probability approach may be a good alternative, if the sample size and/or correlation increases. All methods are biased under MNAR and a low correlation.

miRNA panel from HER2+ and CD24+ plasma extracellular vesicle subpopulations as biomarkers of early-stage breast cancer

BACKGROUND

Mammography screening has improved early breast cancer detection, leading to reduced mortality and lower rates of advanced breast cancer. However, mammography has a high false positive rate that results in over a million invasive breast biopsies of benign lesions in the US each year. Therefore, there is a need for noninvasive, blood-based diagnostics that can accurately assess risk of malignancy for women with indeterminate lesions identified by mammography, such as BI-RADS category 4 breast lesions. The aim of this study is to identify biomarkers from multiplexed extracellular vesicle liquid biopsy that can accurately classify mammographically detected BI-RADS 4 lesions.

METHODS

We analyzed plasma from 113 prospectively enrolled subjects with BI-RADS 4 breast lesions, including 86 women with benign lesions and 27 women with malignant lesions (including 12 with stage I invasive carcinoma and 14 with ductal carcinoma in situ). None of the invasive carcinomas were metastatic. From each plasma sample, we used track etched magnetic nanopore technology to separately isolate HER2 and CD24 expressing extracellular vesicles (EVs) and measured their miRNA cargo using next-generation sequencing. We evaluated the performance of EV-miRNA biomarkers for classifying malignancy and applied LASSO classification to identify a panel of four complementary EV miRNA biomarkers that we validated by qPCR.

RESULTS

We identified 19 differentially enriched miRNA from HER2+ EVs and 11 differentially enriched miRNA from CD24+ EVs of women with malignant lesions compared to benign lesions. We observed individual miRNA with an AUC of up to 0.87 for miR-340-5p from HER2+ EVs and 0.75 for miR-223-3p from CD24+ EVs. LASSO classification selected a panel of four complementary EV miRNA for classifying breast cancer: miR-340-5p (HER2+ EVs), miR-598-3p (CD24+), miR-15b-5p (HER2+), and miR-126-3p (CD24+).

CONCLUSIONS

HER2+ and CD24+ EV subpopulations contain complementary biomarkers suitable for validation in larger studies that can accurately detect early-stage breast cancer among women with BI-RADS category 4 breast lesions.

VISITECT® CD4 advanced disease assay in routine use: Diagnostic accuracy and usability, Ethiopia and Indonesia

We assessed the diagnostic accuracy and usability of the rapid semiquantitative point of care VISITECT® CD4 Advanced Disease Assay, compared to routinely obtained CD4 counts among people living with HIV in Ethiopia and Indonesia. We consecutively enrolled people living with HIV aged ≥10 years who were newly diagnosed, had interrupted antiretroviral treatment for ≥3 months, or were feeling unwell. Venous blood was drawn to obtain CD4 counts on routinely available test instruments and a VISITECT® CD4 Advanced Disease Assay result. The sensitivity of VISITECT® CD4 Advanced Disease Assay was high (97.8%, 95% confidence interval [CI], 94.4-99.4%), but the specificity was low (32.7%, 95%CI, 27.9-37.8%). It significantly increased to 42.5% (95%CI, 33.2-52.1%) after retraining of laboratory staff, but remained low. The low specificity of VISITECT® CD4 Advanced Disease Assay was due to poor capacity in test reading by routine staff. Reading the test strip is error-prone and requires intensive and regular training. Reading aids may improve the usability of this essential point of care test.

Complement levels and their diagnostic utility in neonatal sepsis

BACKGROUND

This study evaluated the diagnostic value of complement levels in neonatal sepsis and their roles in disease progression.

METHODS

This diagnostic accuracy study, conducted in Guangdong Province, China (January 2021-February 2022), involved 41 neonates with sepsis and 41 controls matched by sex and gestational age. Cases included neonates with culture-positive or clinically diagnosed sepsis, while controls consisted of neonates hospitalized for non-septic conditions, confirmed by negative blood cultures. Ten complement components (C1q, C3, C3c, C3b, C4, C5, C5a, H, B, mannose-binding lectin [MBL]) were quantified using residual specimens from routine clinical tests. Descriptive statistics, logistic regression, ROC curve analysis, and correlation assessments were introduced in this study.

RESULTS

The neonatal sepsis group had significantly higher levels of C3c (0.68 vs. 0.49 ng/mL, P = 0.007) and MBL (518.81 vs. 397.06 pg/mL, P < 0.001) compared to controls. In contrast, C5a levels were significantly lower in neonates with sepsis (51.18 vs. 57.25 ng/mL, P = 0.042). C5a demonstrated limited individual performance (AUC = 0.63, 95% CI: 0.51-0.75; sensitivity = 65.9%, specificity = 65.9% at 55.63 ng/mL), while MBL showed moderate accuracy (AUC = 0.75, 95% CI: 0.64-0.85; specificity = 90.2%, sensitivity = 53.7% at 512.86 pg/mL). Notably, their combined C5a + MBL indicator achieved exceptional discrimination (AUC = 0.92, 95% CI: 0.85-0.99) with 85.4% sensitivity and 97.6% specificity, yielding 97.2% positive predictive value (PPV) and 87.0% negative predictive value (NPV).Positive correlations were found between C4 levels and C-reactive protein (CRP), and C3 levels and neutrophil percentage (Neut%), while negative correlations were observed between C5 and MBL levels and total cholesterol (TCH).

CONCLUSIONS

This study highlights the diagnostic significance of combined C5a + MBL indicator in neonatal sepsis and underscores the association between complement levels and disease progression.

Performance of the Large Language Models in African rheumatology: a diagnostic test accuracy study of ChatGPT-4, Gemini, Copilot, and Claude artificial intelligence

BACKGROUND

Artificial intelligence (AI) tools, particularly Large Language Models (LLMs), are revolutionizing medical practice, including rheumatology. However, their diagnostic capabilities remain underexplored in the African context. To assess the diagnostic accuracy of ChatGPT-4, Gemini, Copilot, and Claude AI in rheumatology within an African population.

METHODS

This was a cross-sectional analytical study with retrospective data collection, conducted at the Rheumatology Department of Bogodogo University Hospital Center (Burkina Faso) from January 1 to June 30, 2024. Standardized clinical and paraclinical data from 103 patients were submitted to the four AI models. The diagnoses proposed by the AIs were compared to expert-confirmed diagnoses established by a panel of senior rheumatologists. Diagnostic accuracy, sensitivity, specificity, and predictive values were calculated for each AI model.

RESULTS

Among the patients enrolled in the study period, infectious diseases constituted the most common diagnostic category, representing 47.57% (n = 49). ChatGPT-4 achieved the highest diagnostic accuracy (86.41%), followed by Claude AI (85.44%), Copilot (75.73%), and Gemini (71.84%). The inter-model agreement was moderate, with Cohen's kappa coefficients ranging from 0.43 to 0.59. ChatGPT-4 and Claude AI demonstrated high sensitivity (> 90%) for most conditions but had lower performance for neoplastic diseases (sensitivity < 67%). Patients under 50 years old had a significantly higher probability of receiving a correct diagnosis with Copilot (OR = 3.36; 95% CI [1.16-9.71]; p = 0.025).

CONCLUSION

LLMs, particularly ChatGPT-4 and Claude AI, show high diagnostic capabilities in rheumatology, despite some limitations in specific disease categories.

CLINICAL TRIAL NUMBER

Not applicable.

Validity and reliability of a smartphone-based photographic method for detection of dental caries in adults for use in teledentistry

Introduction

With the arrival of the COVID-19 pandemic, the role of teledentistry increased its need for implementation. In this sense, the study aimed to validate a smartphone-based remote photographic method for diagnosing dental caries in adults attending a polyclinic in Lima, Peru.

Methods

A cross-sectional, descriptive research was conducted; 87 patients were selected, and 2020 teeth were evaluated. Each participant underwent a clinical diagnosis of dental caries by two trained and calibrated dentists, considering the diagnostic criteria of the WHO; during the same visit, after the clinical examination, a family member of the patient was instructed to take a photographic record with a cell phone, through a video. Five photographs of the dental arches were recorded, where the centering of images, resolution, and visualization of all the teeth were evaluated. The photographic evaluation was performed by two independent evaluators blinded to the visual evaluation performed, following the same criteria as the clinical visual evaluation. For the data analysis, Cohen's kappa index was determined for interexaminer reliability; sensitivity, specificity, and positive and negative predictive values were obtained.

Results

Overall, high sensitivity: 90.19% [Interquartile 95% (CI): 88.23-92.16]; and specificity: 95.15% (95% CI: 93.83-96.47). The interexaminer agreement was almost perfect, with a kappa of 0.935 and 0.974 for clinical and photographic evaluation, respectively.

Conclusions

It is concluded that the photographic method using a smartphone has demonstrated a satisfactory level of caries detection in adults.

A Prospective Evaluation of a Three-Gene Host Response Signature to Classify Tuberculosis Severity in Children

BACKGROUND

Children with non-severe TB may benefit from short-course treatment, but point-of-care tools are needed to stratify disease severity. We prospectively evaluated the Cepheid Xpert MTB-Host Response (HR) prototype cartridge for distinguishing TB severity in children with pulmonary TB (PTB) in The Gambia and Uganda.

METHODS

We included children <15 with microbiologically confirmed or clinically diagnosed unconfirmed PTB. Severity was defined using the World Health Organization (WHO) guidelines for a four-month, drug-susceptible regimen. Capillary or venous blood was tested with the HR cartridge for PCR-based detection of 3 mRNA genes and calculation of a TB score from cycle thresholds. We generated receiver operating characteristic curves with the TB score to classify severe TB and assessed if Xpert-HR could achieve the WHO target accuracy for treatment optimization (≥90% sensitivity, ≥70% specificity).

RESULTS

Among 106 children, the median age was 4 years (IQR 1-7), 56.6% were female, and 13.2% were living with HIV. In all children with PTB, Xpert-HR achieved an AUC of 0.67 (95% CI 0.55-0.78), with 89.3% sensitivity (95% CI 71.8-97.7) and 29.5% specificity (95% CI 19.7-40.9, cutoff ≤ -0.60). By confirmation status, Xpert-HR approached the target accuracy in children with Confirmed TB, with 62.5% specificity (95% CI 24.5-91.5) at 91.7% sensitivity (95% CI 61.5-99.8, cut-off ≤ -1.349). Among children with Unconfirmed TB, specificity was lower (24.3%, 95% CI 14.8-36.0) at 93.8% sensitivity (95% CI 69.8-99.8, cutoff ≤ -0.450). Target accuracy was almost achieved in children 5-9 regardless of confirmation status (100% sensitivity [95% CI 71.5-100], 66.7% specificity [95% CI 43.0-85.4], cutoff ≤ -1.35), but specificity (28.2%, 95% CI 18.6-39.5) was lower for children < 5 (92.9% sensitivity, 95% CI 76.5-99.1, cutoff ≤ -0.550).

CONCLUSIONS

Xpert-HR approached the target accuracy to stratify PTB severity in older children and those with Confirmed TB but had lower specificity in children with Unconfirmed TB. Child-specific signatures may be needed to improve performance in younger children with paucibacillary disease.

Prehospital Large-Vessel Occlusion Stroke Detection Scales: A Pooled Individual Patient Data Analysis of 2 Prospective Cohorts

BACKGROUND AND OBJECTIVES

Various prehospital scales have been developed to detect patients with anterior-circulation large-vessel occlusion (aLVO) ischemic stroke to enable direct transportation to a thrombectomy-capable stroke center. To guide implementation, a head-to-head comparison of aLVO stroke detection scales is needed to determine which scale is most useful for prehospital triage in different regional contexts. We aimed to systematically identify and compare these scales.

METHODS

Published prehospital aLVO stroke scales were identified with a systematic literature search. Scales were reconstructed from individual patient data of 2 large prospective observational cohort studies conducted between 2018 and 2019, the Leiden Prehospital Stroke Study and PREhospital triage of patients with suspected STrOke symptoms study. Both studies included consecutive adult patients suspected by paramedics of having a stroke within 6 hours of symptom onset, from 4 Dutch ambulance regions, encompassing 15 stroke centers and serving 3.7 million people. All data used for the reconstruction of scales were acquired by paramedics in the field before hospital arrival. Scales' diagnostic performance to detect aLVO stroke was compared with the area under the receiver operating characteristic curve (AUROC) of the full scale and sensitivity and specificity at the scales' original cut-point. Decision curve analysis was used to evaluate harm-benefit trade-offs between delaying IV thrombolysis and expediting endovascular thrombectomy with direct transportation of patients to a thrombectomy-capable center.

RESULTS

We identified 63 aLVO scales, of which 14 could be reconstructed. Of 2,358 included patients (mean age 70 years; 47% female), 231 (9.8%) had aLVO stroke. The AUROC was highest for Rapid Arterial oCclusion Evaluation (RACE) (0.81, 95% CI 0.78-0.84), Los Angeles Motor Scale (LAMS) (0.80, 95% CI 0.77-0.83), Gaze-Face-Arm-Speech-Time (G-FAST) (0.80, 95% CI 0.77-0.83), and modified Gaze-Face-Arm-Speech-Time (mG-FAST) (0.79, 95% CI 0.76-0.82). The Emergency Medical Stroke Assessment had highest sensitivity (85%, 95% CI 80%-90%) but lowest specificity (58%, 95% CI 56%-61%) while Cincinnati Prehospital Stroke Scale with an adjusted cut-point of 3 + gaze had highest specificity (94%, 95% CI 93%-95%) but lowest sensitivity (35%, 95% CI 29%-41%). In decision curve analysis, RACE had the highest benefit across a clinically reasonable range of harm-benefit trade-offs.

DISCUSSION

RACE, LAMS, G-FAST, and mG-FAST are the best-performing scales, with RACE being preferred in most triage settings. Our findings may support policymakers with implementing a scale suitable for their region.

Validating the falls decision rule: optimizing head CT use in older adults with ground-level falls

OBJECTIVE

Falls are a leading cause of traumatic brain injury in older adults, with ground-level falls being the most common mechanism. Despite the increasing use of head computed tomography (CT) in older adults with ground-level falls, there is an ongoing debate regarding the necessity of routine neuroimaging in all cases. The falls decision rule was developed to safely exclude clinically important intracranial bleeding without head CT in older adults. This study aims to validate the falls decision rule externally and assess its accuracy in identifying low-risk patients while reducing unnecessary imaging.

METHODS

This prospective cohort study at a Level-1 trauma center enrolled consecutive patients aged ≥ 65 years presenting within 48 h of a ground-level fall. Patient management, including the decision to perform head CT, was determined independently by the treating emergency physician. Patients were followed up for 42 days to identify clinically important intracranial bleeding cases. The rule's diagnostic performance was evaluated using sensitivity, specificity, and predictive values using 95% confidence intervals (CI).

RESULTS

A total of 800 patients were included, with a median age of 78 years (IQR 72-85), and 59.9% were female. Clinically important intracranial bleeding was identified in 6.1% (n = 49) of patients. Head CT was performed in 67.6% of cases, identifying 43 initial hemorrhages, with six additional cases detected during follow-ups. The falls decision rule demonstrated 97.9% sensitivity (95% CI 89.1-99.9), 31.9% specificity (95% CI 28.6-35.4), and 99.5% negative predictive value (95% CI 97.1-99.9), potentially reducing CTs by one-third.

CONCLUSION

This validation confirms the falls decision rule's high sensitivity and negative predictive value for identifying low-risk older adults after ground-level falls, potentially reducing unnecessary CT scans by approximately one-third. This approach could alleviate ED overcrowding and resource strain while ensuring diagnostic safety.

Scoping review of deep learning research illuminates artificial intelligence chasm in otolaryngology-head and neck surgery

Clinical validation studies are important to translate artificial intelligence (AI) technology in healthcare but may be underperformed in Otolaryngology - Head & Neck Surgery (OHNS). This scoping review examined deep learning publications in OHNS between 1996 and 2023. Searches on MEDLINE, EMBASE, and Web of Science databases identified 3236 articles of which 444 met inclusion criteria. Publications increased exponentially from 2012-2022 across 48 countries and were most concentrated in otology and neurotology (28%), most targeted extending health care provider capabilities (56%), and most used image input data (55%) and convolutional neural network models (63%). Strikingly, nearly all studies (99.3%) were in silico, proof of concept early-stage studies. Three (0.7%) studies conducted offline validation and zero (0%) clinical validation, illuminating the "AI chasm" in OHNS. Recommendations to cross this chasm include focusing on low complexity and low risk tasks, adhering to reporting guidelines, and prioritizing clinical translation studies.

Biomarker profiling for infection diagnosis in emergency departments: A diagnostic study evaluating C-reactive protein, procalcitonin, Club Cell Protein 16, interleukin-6, chitinase-like protein, and soluble urokinase-type plasminogen activator receptor

INTRODUCTION

Diagnosing infections in emergency departments (EDs) is vital for prompt and effective treatment. This study evaluates the diagnostic accuracy of six inflammatory biomarkers-CC16, CRP, IL-6, PCT, suPAR, and YKL-40-in ED patients with suspected infections.

MATERIALS AND METHODS

This was a multicenter diagnostic accuracy study. Adult patients suspected of infections at four Danish EDs were included. We measured biomarker levels at admission and analyzed their diagnostic performance.

RESULTS

The study included 966 patients, with 789 (81.7%) confirmed infections. CRP, IL-6, and PCT demonstrated the highest positive predictive values, recorded at 93% (95% CI, 91-95), 92% (95% CI, 90-94), and 91% (95% CI, 88-93), respectively. However, the negative predictive values of these markers were low, at 35.5 (95% CI, 30.8-40.5), 38.3 (95% CI, 32.5-44.3) and 34.2 (95% CI, 28.9-39.9) respectively at Youden Index. CRP, IL-6, and PCT were the most predictive, with CRP demonstrating an area under the curve (AUC) of 0.79. IL-6 and PCT showed similar levels of accuracy. CC16, suPAR, and YKL-40 displayed AUCs in the range of 0.53-0.64. CRP levels peaked three days after symptoms appeared, whereas PCT levels were highest upon symptom onset at admission. The influence of fever enhanced the sensitivity of IL-6 and PCT.

CONCLUSION

CRP, IL-6, and PCT showed the best diagnostic accuracy among the biomarkers tested but are influenced by the timing of symptom onset and fever presence. CC16, suPAR, and YKL-40 had poor diagnostic accuracy and were considered poor discriminators. Overall, no single biomarker was sufficient to conclusively rule in or rule out infection. These findings support the need for nuanced interpretation when using biomarkers to diagnose infections in ED settings.

Machine learning of blood haemoglobin and haematocrit levels via smartphone conjunctiva photography in Kenyan pregnant women: a clinical study protocol

INTRODUCTION

Anaemia during pregnancy is a widespread health burden globally, especially in low- and middle-income countries, posing a serious risk to both maternal and neonatal health. The primary challenge is that anaemia is frequently undetected or is detected too late, worsening pregnancy complications. The gold standard for diagnosing anaemia is a clinical laboratory blood haemoglobin (Hgb) or haematocrit (Hct) test involving a venous blood draw. However, this approach presents several challenges in resource-limited settings regarding accessibility and feasibility. Although non-invasive blood Hgb testing technologies are gaining attention, they remain limited in availability, affordability and practicality. This study aims to develop and validate a mobile health (mHealth) machine learning model to reliably predict blood Hgb and Hct levels in Black African pregnant women using smartphone photos of the conjunctiva.

METHODS AND ANALYSIS

This is a single-centre, cross-sectional and observational study, leveraging existing antenatal care services for pregnant women aged 15 to 49 years in Kenya. The study involves collecting smartphone photos of the conjunctiva alongside conventional blood Hgb tests. Relevant clinical data related to each participant's anaemia status will also be collected. The photo acquisition protocol will incorporate diverse scenarios to reflect real-world variability. A clinical training dataset will be used to refine a machine learning model designed to predict blood Hgb and Hct levels from smartphone images of the conjunctiva. Using a separate testing dataset, comprehensive analyses will assess its performance by comparing predicted blood Hgb and Hct levels with clinical laboratory and/or finger-prick readings.

ETHICS AND DISSEMINATION

This study is approved by the Moi University Institutional Research and Ethics Committee (Reference: IREC/585/2023 and Approval Number: 004514), Kenya's National Commission for Science, Technology, and Innovation (NACOSTI Reference: 491921) and Purdue University's Institutional Review Board (Protocol Number: IRB-2023-1235). Participants will include emancipated or mature minors. In Kenya, pregnant women aged 15 to 18 years are recognised as emancipated or mature minors, allowing them to provide informed consent independently. The study poses minimal risk to participants. Findings and results will be disseminated through submissions to peer-reviewed journals and presentations at the participating institutions, including Moi Teaching and Referral Hospital and Kenya's Ministry of Health. On completion of data collection and modelling, this study will demonstrate how machine learning-driven mHealth technologies can reduce reliance on clinical laboratories and complex equipment, offering accessible and scalable solutions for resource-limited and at-home settings.

Results of the Prospective Randomized UroFollow Trial Comparing Marker-guided Versus Cystoscopy-based Surveillance in Patients with Low/Intermediate-risk Bladder Cancer

BACKGROUND AND OBJECTIVE

A growing body of evidence suggests that the intensity of current follow-up in non-muscle-invasive bladder cancer (NMIBC) patients greatly exceeds clinical necessities. The UroFollow trial investigated the diagnostic accuracy of marker-based follow-up in patients with low/intermediate-risk NMIBC against the standard of care (SOC) for noninferiority (margin: <20%).

METHODS

Patients with Ta low- and high-grade (G1-2) NMIBC were randomized to the SOC or 6-monthly marker-based follow-up (algorithm comprising urine markers and ultrasound; marker-based surveillance regimen [MA]). After a negative 3-mo cystoscopy (white light cystoscopy [WLC]), only patients with a positive algorithm underwent WLC in the MA. End-of-study WLC was recommended at 3 yr to recurrence-free patients. Simultaneously, several innovative urine markers were examined.

KEY FINDINGS AND LIMITATIONS

In total, 214 patients were randomized to the SOC (n = 109) and MA (n = 105). The median follow-up was 2.4 yr; 30 and 29 cases of tumor recurrence were diagnosed in the SOC and MA arms, respectively. Sensitivity was 96.5% versus 81.5% (p = 0.1), with one and five Ta low-grade tumors being overlooked in the SOC and MA patients, respectively. No tumor progressing in stage or grade was missed. A total of 589 WLC procedures were performed in the SOC and 148 in the MA arm (p < 0.001). Among five other markers (ADX-Bladder, CellDetect, Bladder EpiCheck, UBC rapid, and Xpert bladder cancer monitor [BC-M]), Bladder EpiCheck and the Xpert BC-M showed similar performance to the algorithm.

CONCLUSIONS AND CLINICAL IMPLICATIONS

UroFollow is the first urine marker-based randomized trial in low/intermediate-risk NMIBC patients. We conclude that 6-monthly marker-based follow-up after negative 3-mo WLC is safe in this cohort. Results of contemporary urine markers suggest that their potential for use in marker-based surveillance, however, requires prospective confirmation.

Variation in sensitivity and specificity of diverse diagnostic tests across health-care settings: a meta-epidemiological study

OBJECTIVES

Diagnostic test accuracy (DTA) may vary among health-care settings, which among other reasons may be due to referral from primary to secondary care. The true magnitude and direction of any difference is not certain. We analyzed the results of meta-analyses of DTA to compare sensitivity and specificity between patients in nonreferred and referred care settings.

STUDY DESIGN AND SETTING

We systematically searched EBSCOhost MEDLINE for systematic reviews that included at least ten original studies of the same diagnostic test, with at least three studies each performed in nonreferred and referred care. Random-effects models, with setting as a binary covariate, were used to calculate pooled sensitivity and specificity estimates per test. Sensitivity analyses were conducted limiting the analyses to studies from countries with gatekeeping systems only.

RESULTS

In total, nine systematic reviews evaluating thirteen diagnostic tests were included. For signs and symptoms (seven tests), the differences in sensitivity and specificity ranged from +0.03 to +0.30 and from -0.12 to +0.03, respectively; for biomarkers (four tests) differences in sensitivity ranged from -0.11 to +0.21 and specificity from -0.01 to -0.19. Differences in sensitivity and specificity for one questionnaire test were +0.1 and -0.07 respectively and for one imaging test were -0.22 and -0.07. Sensitivity analyses limited to countries with gatekeeping health care systems produced similar results.

CONCLUSION

Sensitivity and specificity vary in both direction and magnitude between nonreferred and referred settings, depending on the test and target condition, with no universal patterns governing performance differences.

PLAIN LANGUAGE SUMMARY

Doctors use diagnostic tests to help assess the likelihood if a patient has a certain condition. However, the accuracy of these tests may vary depending on where they are used-such as in primary care (where patients first seek help) or in specialist care (after being referred by a doctor). We wanted to find out how much test accuracy changes between these settings. To do this, we analyzed previous studies that reviewed the accuracy of different diagnostic tests. We compared how well these tests worked in patients who had not yet been referred to a specialist vs those who had. Our analysis included results from thirteen different diagnostic tests, covering symptoms, biomarkers (such as blood tests), a questionnaire, and an imaging test. We found that test accuracy varied depending on the type of test and the condition being diagnosed. Some tests had higher sensitivity (correctly identifying patients with the disease) or specificity (correctly identifying healthy individuals) in primary care, while in specialist care, the same test could perform better, worse, or similarly. There was no clear pattern that applied to all tests. This suggests that researchers should consider how test accuracy may differ across health-care settings when conducting and interpreting diagnostic test accuracy studies.

Calprotectin as a sepsis diagnostic marker in critical care: a retrospective observational study

Diagnosing sepsis in critical care remains a challenge due to the lack of gold-standard diagnostics. Calprotectin (S100A8/A9) has been proposed as a diagnostic marker to identify sepsis in critically ill patients. This study evaluated the diagnostic performance of calprotectin and C-reactive protein (CRP) to distinguish between sepsis and non-sepsis on intensive care unit (ICU) admission. Admission biobank blood samples from adult patients admitted to four ICUs (2015-2018) were used to analyse calprotectin and CRP. All adult patients were screened retrospectively for the sepsis-3 criteria at ICU admission. The diagnostic performance of calprotectin and CRP was evaluated using receiver operating characteristic (ROC) curves. We included 4732 patients, of whom 44% had sepsis. Calprotectin levels were higher in sepsis (p < 0.001). The area under the receiver operating curve (AUROC) to diagnose sepsis was 0.61 for calprotectin compared to 0.72 for CRP (p < 0.001). Among microbiological subgroups of sepsis patients, fungal sepsis had the highest level of calprotectin. We conclude that the diagnostic performance of calprotectin in identifying sepsis patients at ICU admission was inferior to that of CRP.

Development and Validation of an Electronic Health Record-Based, Pediatric Acute Respiratory Distress Syndrome Subphenotype Classifier Model

OBJECTIVE

To determine if hyperinflammatory and hypoinflammatory pediatric acute respiratory distress syndrome (PARDS) subphenotypes defined using serum biomarkers can be determined solely from electronic health record (EHR) data using machine learning.

DESIGN

Retrospective, exploratory analysis using data from 2014 to 2022.

SETTING

Single-center quaternary care PICU.

PATIENTS

Two temporally distinct cohorts of PARDS patients, 2014-2019 and 2019-2022.

INTERVENTIONS

None.

MEASUREMENTS AND MAIN RESULTS

Patients in the derivation cohort ( n = 333) were assigned to hyperinflammatory or hypoinflammatory subphenotypes using biomarkers and latent class analysis. A machine learning model was trained on 165 EHR-derived variables to identify subphenotypes. The most important variables were selected for inclusion in a parsimonious model. The model was validated in a separate cohort ( n = 114). The EHR-based classifier achieved an area under the receiver operating characteristic curve (AUC) of 0.93 (95% CI, 0.87-0.98), with a sensitivity of 88% and specificity of 83% for determining hyperinflammatory PARDS. The parsimonious model, using only five laboratory values, achieved an AUC of 0.92 (95% CI, 0.86-0.98) with a sensitivity of 76% and specificity of 87% in the validation cohort.

CONCLUSIONS

This proof-of-concept study demonstrates that biomarker-based PARDS subphenotypes can be identified using EHR data at 24 hours of PARDS diagnosis. Further validation in larger, multicenter cohorts is needed to confirm the clinical utility of this approach.

The Lever Sign Test Demonstrates Limited Clinical Utility for Diagnosing Full-Thickness Anterior Cruciate Ligament Tears After a Traumatic Knee Injury

Background

Current systematic reviews with meta-analyses have identified the lever sign test as the best clinical examination for ruling out an anterior cruciate ligament (ACL) tear, but the included studies have methodological limitations that could bias the test outcome, potentially overestimating its clinical utility.

Purpose

To investigate the interrater reliability and concurrent validity of the lever sign test after a traumatic knee injury and to investigate the association between test variables (surface used, fist position, effusion grade, force applied, pain reported) and test outcomes.

Study Design

Cohort study (Diagnosis); Level of evidence, 2.

Methods

The lever sign test was performed on hard and soft surfaces in 101 participants after a traumatic knee injury. Magnetic resonance imaging was used as the reference standard, with index testing performed after magnetic resonance imaging was conducted (>3 weeks after injury). Agreement between observers based on the surface used and fist position was evaluated with the Cohen kappa coefficient (κ). Concurrent validity was assessed through sensitivity, specificity, and likelihood ratios. Logistic regression was used to determine whether effusion grade, force applied, and pain reported were significantly associated with test outcomes.

Results

Interrater reliability was superior on the soft surface but demonstrated only moderate agreement (κ = 0.529 [95% CI, 0.368-0.691]). Sensitivity was higher on the soft surface, and specificity was higher on the hard surface, for both assessors. At best, positive and negative likelihood ratios were 3.02 (95% CI, 1.60-5.69) and 0.45 (95% CI, 0.28-0.73), respectively. Test outcomes were affected by the surface used and fist position, but effusion grade, force applied, and pain reported were not significantly associated with correct/incorrect test results.

Conclusion

In participants assessed from 3 weeks after a traumatic knee injury, the lever sign test demonstrated limited clinical utility for diagnosing full-thickness ACL tears. Test outcomes were affected by the surface used and fist position of the assessor.

Registration

NCT05416632 (ClinicalTrials.gov).

Prenatal diagnosis of tracheo-oesophageal fistula/oesophageal atresia: is MRI helpful?

BACKGROUND

Oesophageal atresia (OA) with or without tracheo-oesophageal fistula (TOF) affects 2.75 per 10,000 births within the UK. It is most frequently suspected on antenatal imaging when the stomach is absent or appears small. Studies have shown fetal magnetic resonance imaging (MRI) has greater diagnostic accuracy than ultrasound; however, there remains uncertainty over what size constitutes a small stomach and how frequently this correlates with a diagnosis of TOF/OA.

METHODS

A retrospective study of patients referred for fetal MRI due to suspicions of TOF/OA on antenatal ultrasound from 2011 to 2022. We also included patients with a fetal MRI suspecting TOF/OA who had been referred for other reasons. The indication, MRI findings and postnatal outcome were compared to assess diagnostic accuracy. For each case, the size of the stomach bubble was measured on MRI, and stomach volumes in a control group were measured for comparison.

RESULTS

The positive predictive value for USS was 45.5% and 51.7% for fetal MRI. Fetal MRI had a negative predictive value and sensitivity of 100% (p = 0.027). The control group showed a strong positive correlation between stomach size and increasing gestational age (R2 = 0.69, p < 0.001), but this correlation was less positive in the TOF/OA group (R2 = 0.26, p = 0.03), and the stomach volumes in TOF/OA were consistently lower than the control group. The receiver operating characteristic curve illustrates that an absent stomach or unmeasurably small stomach is more diagnostic of TOF/OA as volumes ≤0.06 ml had 90% sensitivity.

CONCLUSION

Fetal MRI can accurately exclude TOF/OA but only has marginally improved positive predictive value over ultrasound. Research with larger numbers is required to further aid the development of a cut-off value for what can be considered a pathologically small stomach.

IMPACT

There are several features on imaging that raise the suspicion of TOF/OA. Fetal MRI has some improved diagnostic accuracy compared with antenatal ultrasound alone; however, it is only marginally better. Absence of stomach bubble and presence of oesophageal dilatation combined on fetal MRI are more diagnostic of TOF/OA.

Risk of Bias in Liver Imaging Reporting and Data System Studies Using QUADAS-2

Purpose: Use a tailored version of the Quality Assessment of Diagnostic Accuracy Studies tool to evaluate risk of bias and applicability across LIRADS related publications. Method: A tailored QUADAS-2 tool was created through consensus approach to assess risk of bias and applicability across 37 LI-RADS related publications. Studies were selected from 2017 to 2022 using the assistance of experienced hospital librarians to search for studies evaluating the diagnostic accuracy of CT, MRI, or contrast-enhanced ultrasound for HCC using LI-RADS through multiple different databases. QUADAS-2 assessments were performed in duplicate and independently by 2 authors with experience using the QUADAS-2 tool. Disagreements were resolved with a third expert reviewer. Consensus QUADAS-2 assessments were tabulated for each domain. Results: Using the tailored QUADAS-2 tool, 31 of the 37 included LI-RADS studies were assessed as high risk of bias, and 9 out of 37 studies demonstrated concerns for applicability. Patient selection (21 out of 37 studies) and flow/timing (24 out of 37 studies) domains demonstrated the highest risk of bias. 6 out of 37 studies in the index domain demonstrated high risk of bias. 2 out of 37 studies showed high risk of bias in the reference standard domain. Conclusion: A significant proportion of LI-RADS research is at risk of bias with concerns for applicability. Identifying risk of bias in such research is essential to recognize limitations of a study that may affect the validity of the results. Areas for improvement in LI-RADS research include reducing selection bias, avoiding inappropriate exclusions, and decreasing verification bias.

Cognitive assessment tools for Arabic-speaking older adults: A systematic review

This systematic review aims to identify available cognitive assessments for Arabic-speaking older adults and to assess their validity and performance. A comprehensive search was conducted using Medline, Embase, and APA PsycInfo up to November 2023, encompassing studies validating or using cognitive tools in Arabic for individuals aged ≥ 50. We identified 29 validation studies for 20 cognitive tools and 125 studies using cognitive tools. Three tools were validated in more than one study/setting. Cut-offs for dementia were validated for 16 tools (including two domain-specific tools) and for cognitive impairment for three tools. The Mini-Mental State Examination and Montreal Cognitive Assessment were the most frequently validated and used tools. The results highlight a large need for improved psychometric data for cognitive assessments for Arabic-speaking older adults and identify important gaps in knowledge regarding domain-specific tools, the detection of cognitive changes, and the suitability of assessments across different settings and subgroups. HIGHLIGHTS: We reviewed the availability and properties of cognitive assessments in Arabic. Psychometric data on cognitive tools for older Arabic-speaking adults are scarce. Only three tools are validated in more than one study/setting. Data are largely lacking for domain-specific tools and early cognitive changes. The review identifies important methodology, reporting, and reproducibility issues.

Diagnostic Accuracy of On-Premise Automated Coronary CT Angiography Analysis Based on Coronary Artery Disease Reporting and Data System 2.0

Background Chest pain is a leading cause of outpatient and emergency department visits; advancements in artificial intelligence (AI) could improve coronary CT angiography (CCTA) workflows for these patients. Purpose To evaluate the performance of an on-premise AI-based coronary artery calcium scoring (CACS) and CCTA analysis software against expert interpretation based on Coronary Artery Disease Reporting and Data System (CAD-RADS) 2.0. Materials and Methods This retrospective study included consecutive patients undergoing CCTA for coronary analysis at a tertiary academic center between January 2017 and October 2021 across four scanners from three vendors. Patients with stents, bypass grafts, anomalies, or nondiagnostic studies were excluded. On-premise AI output included CACS, CAD-RADS category, and segment involvement score (SIS) within less than 5 minutes. Original CCTA reports were used as the reference, and discrepancies between AI and reports were further adjudicated by two blinded level-III readers with 8 and 5 years of CCTA experience. Agreement among CACS risk categories, CAD-RADS categories, and plaque burden scores was measured with the weighted κ. The area under the receiver operating characteristic curve, positive predictive value, and negative predictive value were used to evaluate diagnostic performance. Bootstrapping was used to estimate 95% CIs. Results A total of 1032 patients (median age, 62 [IQR, 54-69] years; 581 female) with 1041 CCTA images were included: 361 of the 1041 images (35%) were classified as CAD-RADS 0, 274 (26%) as CAD-RADS 1, 186 (18%) as CAD-RADS 2, 101 (10%) as CAD-RADS 3, 95 (9%) as CAD-RADS 4A, 11 (1%) as CAD-RADS 4B, and 13 (1%) as CAD-RADS 5. There was substantial agreement between AI and expert CAD-RADS stenosis severity categories (weighted κ = 0.73). AI demonstrated high performance (per-scan area under the receiver operating characteristic curve, 0.90; 95% CI: 0.87, 0.92) for CAD-RADS greater than or equal to 3 or greater than or equal to 4A and high negative predictive value (98%; 95% CI: 97, 99) but low positive predictive value (39%; 95% CI: 32, 45) for CAD-RADS greater than or equal to 4A. AI-based plaque burden scores derived from CACS reached near-perfect agreement with experts (weighted κ = 0.97), whereas those derived from SIS showed substantial agreement (weighted κ = 0.79). Conclusion On-premise AI accurately ruled out obstructive coronary artery disease at CCTA and achieved substantial to near-perfect agreement with human experts for CAD-RADS 2.0 stenosis severity and plaque burden. © RSNA, 2025 Supplemental material is available for this article. See also the editorial by van Assen and De Cecco in this issue.

Comparing Screening Tools for Predicting Phoenix Criteria Sepsis and Septic Shock Among Children

BACKGROUND AND OBJECTIVES

The Phoenix criteria for pediatric sepsis and septic shock have recently been proposed for worldwide application. The Phoenix sepsis criteria are based on organ dysfunction scoring. Although many screening tools exist, their performance in predicting Phoenix outcomes is not known. We hypothesized that the quick Pediatric Septic Shock Screening Score (qPS4) would demonstrate greater sensitivity compared with the Liverpool quick Sequential Organ Failure Assessment (LqSOFA) and a commonly used 2-stage screening tool created at Children's Hospital of Philadelphia (CHOP).

METHODS

We performed a secondary analysis of the qPS4 validation set data from a retrospective cohort study of pediatric emergency department patients with suspected infection. The exposure was a positive screen prior to outcome occurring. We calculated the predictive characteristics of qPS4, LqSOFA, and CHOP for Phoenix sepsis and septic shock within 24 hours of arrival.

RESULTS

We analyzed 47 176 encounters. Within 24 hours of arrival to the ED, 628 (1.3%) met criteria for sepsis and 228 (0.5%) met criteria for septic shock. The qPS4 predicted sepsis with 67.8% sensitivity and 89.6% specificity compared with LqSOFA (sensitivity 47.0%, specificity 95.7%) and the CHOP screen (sensitivity 49.7%, specificity 92.1%) (P < .05 for all compared to qPS4). The qPS4 predicted septic shock with 85.5% sensitivity and 89.0% specificity compared with LqSOFA (sensitivity 59.2%, specificity 95.2%) and the 2-stage CHOP screen (sensitivity 64.9%, specificity 91.5%) (P < .05 for all compared to qPS4).

CONCLUSIONS

The qPS4 predicted Phoenix sepsis and septic shock with greater sensitivity and clinically similar specificity compared with widely used bedside tools.

Harmonizing neuropathic pain research: outcomes of the London consensus meeting on peripheral tissue studies

ABSTRACT

Neuropathic pain remains difficult to treat, with drug development hampered by an incomplete understanding of the pathogenesis of the condition, as well as a lack of biomarkers. The problem is compounded by the scarcity of relevant human peripheral tissues, including skin, nerves, and dorsal root ganglia. Efforts to obtain such samples are accelerating, increasing the need for standardisation across laboratories. In this white paper, we report on a consensus meeting attended by neuropathic pain experts, designed to accelerate protocol alignment and harmonization of studies involving relevant peripheral tissues. The meeting was held in London in March 2024 and attended by 28 networking partners, including industry and patient representatives. We achieved consensus on minimal recommended phenotyping, harmonised wet laboratory protocols, statistical design, reporting, and data sharing. Here, we also share a variety of relevant standard operating procedures as supplementary protocols. We envision that our recommendations will help unify human tissue research in the field and accelerate our understanding of how abnormal interactions between sensory neurons and their local peripheral environment contribute towards neuropathic pain.

Feasibility, efficacy, and safety of core needle biopsy as a first-line method for cervical lymphadenopathy

OBJECTIVES

This study aimed to determine the feasibility, diagnostic efficacy, and safety of ultrasound-guided core needle biopsy (CNB) as a first-line biopsy method for cervical lymphadenopathy of non-thyroid origin.

MATERIALS AND METHODS

This retrospective cohort study included consecutive patients with cervical lymphadenopathy in whom US-guided CNB was used as the first-line biopsy method for cervical lymph nodes (LNs) of presumed non-thyroid origin. The coaxial CNB technique was routinely used, while the tilting and hydrodissection CNB techniques were selectively employed for small high-risk LNs. The primary endpoint of this study was the diagnostic efficacy of CNB, evaluated by the rate of inconclusive results (nondiagnostic and indeterminate) and diagnostic accuracy (criterion 1: malignant results; criterion 2: malignant or indeterminate result). The secondary outcomes included the feasibility and safety of CNB, assessed based on the technical success rate and complication rate, respectively.

RESULTS

The rates of nondiagnostic, indeterminate, and inconclusive results were 0.7%, 3.4%, and 4.1%, respectively. The sensitivity, specificity, and accuracy of CNB for malignant LNs were 96.2%, 100%, and 97.8%, respectively, with criterion 1, and these values were all 99.8% with criterion 2. The technical success rate of CNB was 99.3%. There were no major complications and 7 cases (0.6%) of minor complications (asymptomatic hematomas).

CONCLUSION

CNB was technically feasible, effective, and safe as a first-line biopsy method for cervical lymphadenopathy of non-thyroid origin with high diagnostic accuracy for malignant nodal disease.

KEY POINTS

Question The role of US-guided CNB as a first-line biopsy method for cervical LNs has not yet been verified and established. Findings US-guided CNB, as a first-line method, demonstrated a high technical success rate and diagnostic accuracy for malignant nodes, with few minor complications. Clinical relevance US-guided CNB can be used as an effective first-line biopsy method for cervical lymphadenopathy and will enable accurate diagnosis of malignant LNs.

Evaluating stratified T-SPOT.TB results for diagnostic accuracy in tuberculosis disease: a retrospective cohort study with sensitivities, specificities, and predictive values

OBJECTIVES

The study aimed to investigate the association between quantitative T-SPOT.TB values and the risk of incident and prevalent tuberculosis disease (TBD), identify risk factors, and evaluate test accuracy.

METHODS

This retrospective cohort study followed patients tested consecutively with T-SPOT.TB at Aarhus University Hospital from 2010 to 2017, with follow-up for incident TBD through 2022. Data on demographics, comorbidities, medications, and TB status were collected from patient records and national registries. Cox proportional hazard models with restricted cubic splines assessed the risk of incident TBD (occurring ≥3 months post-test) by quantitative spot counts. Cox and log-binomial regressions identified risk factors for incident and prevalent TBD (occurring between 3 months before and after the test). Sensitivity, specificity, and predictive values assessed test accuracy. T-SPOT.TB was the index test, and microbiologically and/or clinically confirmed TBD was the reference standard.

RESULTS

Among 8542 individuals with complete follow-up, 59 developed incident TBD over 67 456 person-years. Among 9014 individuals tested once, 162 had prevalent TBD at the time of testing. The risk of incident TBD increased with higher spot counts, plateauing for tests with more than ten spots. The strongest risk factors for both incident and prevalent TBD were categorical T-SPOT.TB results: compared with negative tests (≤4 spots), adjusted hazard ratios for incident TBD were 5.0 (95% CI: 1.9-13.1) for borderline (5-7 spots) and 8.0 (95% CI: 4.0-15.7) for positive tests (≥8 spots). Adjusted risk ratios for prevalent TBD were 14.9 (95% CI: 7.7-28.9) for borderline and 35.6 (95% CI: 21.4-59.2) for positive tests. Sensitivities for incident and prevalent TBD were 54.0% (95% CI: 39.3-68.2%) and 78.4% (95% CI: 71.3-84.5%), respectively. Specificities were 84.8 (84.0-85.4) and 83.7 (82.9-84.4), respectively.

DISCUSSION

Incident TBD risk increases with T-SPOT.TB values but plateaus beyond 10 spots. Borderline and positive T-SPOT.TB results are strongly linked to TBD risk.

Evaluating a novel MR imaging biomarker for retinal hemorrhage

BACKGROUND

Dilated fundus examination is the clinical reference standard for retinal hemorrhage diagnosis. Obtaining a timely examination may be challenging in emergent and inpatient settings, and retinal hemorrhage may go undetected. An imaging biomarker is therefore desirable.

OBJECTIVE

To evaluate the frequency of ringlets of hyperintensity in orbital fat on axial fat-suppressed T2-weighted fast spin echo MR images in children with ophthalmologically-confirmed retinal hemorrhage and in a comparison group of children with benign macrocrania.

MATERIALS AND METHODS

This retrospective single-center study included cases aged > 6 weeks- < 3 years with ophthalmologist-confirmed retinal hemorrhage from 07/2011-12/2019 who underwent full brain MR imaging and reference patients aged > 6 weeks- < 3 years with benign macrocrania who underwent brain MR imaging during the same period. Images were reviewed by two pediatric neuroradiologists blinded to patient group; a subset was randomly selected for duplicate reads. Descriptive data were tabulated for each reviewer and adjudicated results.

RESULTS

After adjudication, ringlets were observed in at least one eye in 67/70 retinal hemorrhage cases (96%) and 3/74 benign macrocrania patients (4%; P < 0.0001). Similar results were observed upon stratification by presence of visible retinal hemorrhage on imaging, and by magnet strength. Intrarater reliability and interrater reliability were high (all κ ≥ 0.86), with concordance rates ≥ 93% across duplicate reads.

CONCLUSION

Ringlets of hyperintensity are present at high frequency in children with confirmed retinal hemorrhage and are neither an MR imaging artifact, nor a normal developmental finding in infants and young children. As such, they may aid in retinal hemorrhage detection following abusive head trauma.

Validation study of the SER/SEPAR screening criteria for interstitial lung disease in early rheumatoid arthritis patients

OBJECTIVE

In 2023, the Spanish Society of Rheumatology (SER) and the Spanish Society of Pulmonology and Thoracic Surgery (SEPAR) proposed screening criteria for interstitial lung disease (ILD) in rheumatoid arthritis (RA) based on expert opinion, requiring validation. This study aimed to evaluate their sensitivity and specificity in a cohort of early RA patients.

METHODS

This cross-sectional study retrospectively assessed the SER/SEPAR criteria in 146 early RA patients screened for ILD at diagnosis. Screening included medical history, respiratory auscultation, chest X-ray (CXR), and complete pulmonary function tests (PFTs). Thoracic high-resolution computed tomography (HRCT) was performed only in the presence of symptoms, velcro crackles, or abnormalities on CXR or PFTs.

RESULTS

Among the 146 patients included, 28 (19.2 %) developed ILD, all confirmed by HRCT. Of these, 12 (43 %) had clinically evident ILD preceding or coinciding with joint symptoms, while the remaining 16 (57 %) were identified after applying the screening protocol. At diagnosis, 90 patients (61.6 %) met the screening criteria, with ILD confirmed in 26 cases (28.9 %). Conversely, among 56 patients (38.4 %) not meeting the criteria, ILD was ultimately identified in 2 cases (3.5 %). The SER/SEPAR screening criteria demonstrated a sensitivity of 92.9 % (95 % CI: 76.5-99.1), specificity of 45.8 % (95 % CI: 36.6-55.2), LR+ of 1.71 (95 % CI: 1.41-2.08), LR- of 0.16 (95 % CI: 0.04-0.60), PPV of 28.9 % (95 % CI: 25.1-33.1), NPV of 96.4 % (95 % CI: 87.5-99.1), and diagnostic accuracy of 54.8 % (95 % CI: 46.4-63.1).

CONCLUSION

The SER/SEPAR criteria for ILD screening demonstrated high sensitivity in recent-onset RA patients, supporting their utility as a tool for early detection in this scenario.

The Inaccuracy of the Mood Disorder Questionnaire for Bipolar Disorder in a Community Sample: From the "DYMERS" Construct Toward a New Instrument for Detecting Vulnerable Conditions

Background/Objectives: The Mood Disorder Questionnaire (MDQ) is a widely used tool for the early detection of Bipolar Disorder (BD), yet its diagnostic accuracy remains debated. In particular, the MDQ often yields false positives in individuals with anxiety, stress-related, or personality disorders, raising questions about its clinical utility. This study aimed primarily to evaluate the sensitivity, specificity, and predictive values of the MDQ in identifying BD within a large, community-based sample using structured clinical interviews. Additionally, we explored the construct of DYMERS (Dysregulation of Mood, Energy, and Social Rhythms Syndrome), a proposed condition characterized by mood instability, hyperactivation traits, and rhythm dysregulation among MDQ-positive individuals without a formal psychiatric diagnosis. Methods: A total of 4999 adults were surveyed across six Italian regions using a stratified random sampling method. Psychiatric diagnoses were established using DSM-IV-TR criteria via the Advanced Neuropsychiatric Tools and Assessment Schedule (ANTAS). The MDQ was administered face to face in its validated Italian version, with a positivity cut-off of ≥7. The MDQ exhibited low sensitivity and high specificity (0.962; 95% CI: 0.961-0.963). Results: Among 2337 analyzable cases, the MDQ showed high specificity (96.2%) but low sensitivity (42.9%) for BD, indicating limited effectiveness as a screening tool. In clinical terms, this implies that while MDQ-positive individuals are unlikely to be false positives, a substantial proportion of true BD cases are not identified. Notably, a significant subgroup of MDQ-positive individuals without psychiatric diagnoses displayed features consistent with DYMERS. Conclusions: Our findings confirm the limited screening value of the MDQ for BD in community samples. However, MDQ positivity may help identify a broader spectrum of mood and rhythm dysregulation not captured by current diagnostic systems. Future research should focus on validating DYMERS as a clinical entity and on developing targeted diagnostic instruments capable of capturing this emerging dimension of psychopathology.

Researcher's Guide for the Preparation of Tables

Almost all scientific articles include tables, but there is little guidance on how to prepare tables. Authors should enhance the focus on the tables' data-ink, the essential information, and minimizing the no-data-ink, nonessential elements. The six principles of Gestalt on human perception can also be applied to the table to increase readability. Through this review, we provide a practical guide and an overview of how to prepare readable, informative tables. The five steps for the preparation of tables include (1) tables with a clear purpose; (2) using a universal layout; (3) selecting relevant data for Table 1 versus other tables; (4) simplifying variables by categorizing, standardizing, and reducing; and (5) enhancing the readability of numbers and decimals. This results in informative tables that contain data, serving a specific purpose for the reader and increasing readability.

Development of a simplified smell test to identify Parkinson's disease using multiple cohorts, machine learning and item response theory

To develop a simplified smell test for identifying patients with Parkinson's disease (PD), we reevaluated the Sniffin'-Sticks-Identification-Test (SST-ID) and University-of-Pennsylvania-Smell-Identification-Test (UPSIT), using three case-control studies. These included 301 patients with PD or dementia with Lewy bodies (DLB), 68 subjects with multiple-system atrophy (MSA) or progressive supranuclear palsy (PSP), and 281 healthy controls (HC). Scents were ranked by area-under-the-curve values for group classification and results leveraged by 8 published studies with 5853 individuals. PD/DLB patients showed markedly worse olfaction than controls, whereas scores for MSA/PSP subjects were intermediate. We identified and validated a subset of 7 shared odorants that performed similarly to the traditional 16-scent SST-ID and 40-scent UPSIT tests in distinguishing PD/DLB from HC. There, the identification of 4 or fewer scents out of 7 served as an effective cut-off between the two groups. We also identified a critical role for distractors (from correct answers) and age on olfaction performance.

Coronary Calcium Scoring Using True and Virtual Non-Contrast Reconstructions on Photon-Counting CT with Differing Slice Increment: Impact on Calcium Severity Classifications

Background: Cardiovascular risk assessment relies heavily on coronary calcium scoring. With an emphasis on varying slice increments, this study investigates the effectiveness of true and virtual non-contrast reconstructions on photon-counting CT. Reconstruction methods' effects on calcium severity classifications are critical to the improvement in imaging techniques. Methods: This study comprised 77 participants (mean age: 63 ± 10 years, 43% female), of whom 0 had a coronary artery calcium score (CACS) of zero. In contrast to true non-contrast (TNC) 3 × 3 mm, the reconstructions included TNC 3 × 1.5 mm, virtual non-contrast (VNC) 3 × 3 mm, and VNC 3 × 1.5 mm. Agatston units served as the basis for classifications into standard clinical diagnostic categories. Results: High concordance between acquisition types was revealed by interclass correlation values (0.97-0.99). Comparing TNC 3 × 1.5 mm reconstructions to their VNC counterparts, misclassifications were less common (Cohen Kappa = 0.94). (K = 0.83-0.85). Significant differences in the average calcium scores and rates of misclassification highlighted the impact of reconstruction methods on precise evaluations. Conclusions: VNC methods demonstrated high agreement; however, with a small rate of misclassifications as compared to the gold standard method. VNC CACS may help optimize workflows but may need differing cutoffs as compared to traditional methods.

Genomic and clinical characterization of HER2 exon 20 mutations in non-small cell lung cancer: insights from a multicenter study in South China

BACKGROUND

The objective of this study was to investigate the clinical and genetic characteristics and clinical relevance of HER2 exon 20 oncogenic variants in non-small cell lung cancer (NSCLC) patients.

METHODS

This prospective study analyzed 51 NSCLC patients with HER2 mutations, identified via next-generation sequencing (NGS) of tissue, blood, cerebrospinal fluid, or pleural effusion samples. Patients were grouped based on the presence of exon 20 mutations (exon 20 vs. non-exon 20) and further divided based on whether they had received prior anti-tumor treatments (baseline vs. non-baseline). Clinical and genetic data, treatment responses were analyzed. Progression-free survival (PFS) and overall survival (OS) were evaluated using Kaplan-Meier methods and compared with log-rank tests. Gene ontology (GO) analysis was performed to uncover the biological significance of the mutated genes.

RESULTS

In a cohort of 651 NSCLC patients, 51 (7.83%) harbored HER2 alterations, including 20 (3.08%) with exon 20 mutations. The median age of the HER2-altered subgroup was 58.5 years. Adenocarcinoma was the most prevalent subtype (96.1%), and most patients presented at stage IV (72.5%). The most common metastatic sites were the lungs (68.6%), lymph nodes (52.9%), and brain (43.1%). Among the HER2 mutated patients, 20 (39.3%) had exon 20 mutations. Exon 20 mutations were more prevalent in the non-baseline group (55.0% vs. 29.0%, P = 0.049) and males (75.0%, P = 0.025). These mutations were associated with a higher rate of metastasis to the lungs, lymph nodes (P < 0.001). Patients with exon 20 mutations demonstrated poorer overall survival (OS) outcomes (P = 0.048). No significant differences were observed in age, smoking history, histological subtype, or TNM stage at diagnosis between groups. The majority of exon 20 mutations were in-frame indel mutations (92.0%), with the most common specific mutation being p.Y772_A775dup (70%). Gene Ontology (GO) analysis linked exon 20 mutations to unregulated protein kinase activity and anoikis.

CONCLUSIONS

Our study found that NSCLC patients with HER2 exon 20 oncogenic variants have a higher risk of metastasis and drug resistance, leading to worse outcomes than non-exon 20 mutations. This highlights the urgent need for targeted therapies aimed at exon 20 insertions to improve survival and treatment outcomes in this subgroup.

Diagnostic accuracy of Truenat MTB Ultima on sputum for pulmonary tuberculosis diagnosis in an HIV-endemic setting

OBJECTIVES

Assess Truenat MTB Ultima (Ultima) diagnostic accuracy, alongside standard-of-care Truenat MTB Plus (MTB Plus) and Xpert MTB/RIF Ultra (Ultra) tests, for pulmonary tuberculosis (TB) diagnosis in a high-burden setting. We secondarily evaluated Truenat MTB-RIF Dx (MTB-RIF Dx) on Ultima- or MTB Plus-positive samples for rifampicin susceptibility diagnosis.

METHODS

Adults (≥18 years; n = 498) with presumptive TB self-presenting to primary care clinics in Cape Town, South Africa (19/02/2016-22/02/2023) provided sputa. The microbiological reference standard was a single culture for TB and MTBDRplus on an isolate for rifampicin susceptibility.

RESULTS

In total, 54% (n = 269) of the participants had HIV, and 42% (n = 210) had previous TB. The proportion of Ultima and MTB Plus unsuccessful results was 14% (95% CI 11, 16) and 20% (17, 23), respectively, with at least half resolving upon retesting the same eluate. In a three-way analysis, Ultima, MTB Plus, and Ultra had TB sensitivities of 90% (85, 93), 84% (78, 88), and 92% (87, 95) and specificities of 85% (80, 88), 95% (92, 97), and 95% (92, 97). Ultima specificity did not improve with Ultra in the reference standard. MTB-RIF Dx had high unsuccessful result rates that varied if done on the day of DNA extraction or on Ultima- [18% (10, 26) vs. 44% (35, 51) if after day of extraction] or MTB Plus-positive eluates [9% (3, 16) vs. 27% (18, 35)]. Same day rifampicin susceptibility testing was often unsuccessful in samples with the "very low" semiquantitation category reported by Ultima [75% (65, 86)] or MTB Plus [73% (58, 89)] but had 100% (40, 100) sensitivity and 99% (96, 100) specificity (on both MTB Plus- or Ultima-positive DNA). Reagent lot variation in unsuccessful and false positive results was observed.

DISCUSSION

Ultima met the minimum sensitivity recommended by the WHO for TB detection, but specificity, reagent lot variation, and unsuccessful results were suboptimal.

Artificial Intelligence Models for Pediatric Lung Sound Analysis: Systematic Review and Meta-Analysis

BACKGROUND

Pediatric respiratory diseases, including asthma and pneumonia, are major causes of morbidity and mortality in children. Auscultation of lung sounds is a key diagnostic tool but is prone to subjective variability. The integration of artificial intelligence (AI) and machine learning (ML) with electronic stethoscopes offers a promising approach for automated and objective lung sound.

OBJECTIVE

This systematic review and meta-analysis assess the performance of ML models in pediatric lung sound analysis. The study evaluates the methodologies, model performance, and database characteristics while identifying limitations and future directions for clinical implementation.

METHODS

A systematic search was conducted in Medline via PubMed, Embase, Web of Science, OVID, and IEEE Xplore for studies published between January 1, 1990, and December 16, 2024. Inclusion criteria are as follows: studies developing ML models for pediatric lung sound classification with a defined database, physician-labeled reference standard, and reported performance metrics. Exclusion criteria are as follows: studies focusing on adults, cardiac auscultation, validation of existing models, or lacking performance metrics. Risk of bias was assessed using a modified Quality Assessment of Diagnostic Accuracy Studies (version 2) framework. Data were extracted on study design, dataset, ML methods, feature extraction, and classification tasks. Bivariate meta-analysis was performed for binary classification tasks, including wheezing and abnormal lung sound detection.

RESULTS

A total of 41 studies met the inclusion criteria. The most common classification task was binary detection of abnormal lung sounds, particularly wheezing. Pooled sensitivity and specificity for wheeze detection were 0.902 (95% CI 0.726-0.970) and 0.955 (95% CI 0.762-0.993), respectively. For abnormal lung sound detection, pooled sensitivity was 0.907 (95% CI 0.816-0.956) and specificity 0.877 (95% CI 0.813-0.921). The most frequently used feature extraction methods were Mel-spectrogram, Mel-frequency cepstral coefficients, and short-time Fourier transform. Convolutional neural networks were the predominant ML model, often combined with recurrent neural networks or residual network architectures. However, high heterogeneity in dataset size, annotation methods, and evaluation criteria were observed. Most studies relied on small, single-center datasets, limiting generalizability.

CONCLUSIONS

ML models show high accuracy in pediatric lung sound analysis, but face limitations due to dataset heterogeneity, lack of standard guidelines, and limited external validation. Future research should focus on standardized protocols and the development of large-scale, multicenter datasets to improve model robustness and clinical implementation.

ECG Criteria and Diagnostic Score for Right Atrial Disease in Children: The CHILDHEART Study

To develop a novel diagnostic electrocardiogram (ECG) score to improve the early detection of right atrial disease (RAD) in pediatric patients with congenital heart disease (CHD). We conducted a retrospective case-control study involving 219 pediatric patients aged 0 to 14 years with confirmed CHD and 1,418 controls without congenital disease. The ECG data were analyzed to identify specific criteria predictive of RAD. Receiver operating characteristic (ROC) curves and multivariate analyses were used to refine and develop a diagnostic score. Key variables assessed included right bundle branch block (RBBB), a pure R wave in V1, an initial q wave in V1, an S wave greater than the R wave in V6, and RV1 + SV6 ≥ 18 mm. The diagnostic score demonstrated robust performance. In children under 3 years, the score achieved an area under the curve (AUC) of 0.826 (95% CI: 0.779-0.879). A score of ≥ 2 had a sensitivity of 63.2% and specificity of 89.7%. For children aged 3 years and older, the score had an AUC of 0.817 (95% CI: 0.764-0.871), with a sensitivity of 48.67% and specificity of 98.1%. The CHILDHEART diagnostic score provides a practical and reliable tool for identifying RAD in pediatric patients using a standard 12-lead ECG. This score offers significant potential for early diagnosis and improved management of RAD in children, especially in clinical settings where advanced imaging modalities may not be readily available.

Rigorous validation of machine learning in laboratory medicine: guidance toward quality improvement

The application of artificial intelligence (AI) in laboratory medicine will revolutionize predictive modeling using clinical laboratory information. Machine learning (ML), a sub-discipline of AI, involves fitting algorithms to datasets and is broadly used for data-driven predictive modeling in various disciplines. The majority of ML studies reported in systematic reviews lack key aspects of quality assurance. In clinical laboratory medicine, it is important to consider how differences in analytical methodologies, assay calibration, harmonization, pre-analytical errors, interferences, and physiological factors affecting measured analyte concentrations may also affect the downstream robustness and reliability of ML models. In this article, we address the need for quality improvement and proper validation of ML classification models, with the goal of bringing attention to key concepts pertinent to researchers, manuscript reviewers, and journal editors within the field of pathology and laboratory medicine. Several existing predictive modeling guidelines and recommendations can be readily adapted to the development of ML models in laboratory medicine. We summarize a basic overview of ML and key points from current guidelines including advantages and pitfalls of applied ML. In addition, we draw a parallel between validation of clinical assays and ML models in the context of current regulatory frameworks. The importance of classification performance metrics, model explainability, and data quality along with recommendations for strengthening journal submission requirements are also discussed. Although the focus of this article is on the application of ML in laboratory medicine, many of these concepts extend into other areas of medicine and biomedical science as well.

Performance of Xpert MTB/RIF ultra for the diagnosis of tuberculous meningitis in children using cerebrospinal fluid

Rapid and reliable diagnostic tests for tuberculous meningitis (TBM) in children remain limited. This study evaluated the performance of Xpert MTB/RIF Ultra (Ultra), an improved version of Xpert MTB/RIF (Xpert), for detecting TBM in children using cerebrospinal fluid (CSF). Children (0-14 years) presumptive of TBM were enrolled prospectively from three tertiary level hospitals in Dhaka, Bangladesh, from December 2019 to January 2022. CSF was tested by Ultra, Xpert, Lowenstein-Jensen (L-J) culture, and acid-fast bacilli (AFB) microscopy. Diagnostic performance was assessed against clinical case definitions, composite microbiological reference standard (CMRS), and L-J culture. Of 187 children, Ultra detected TBM in 23.4% of cases, significantly higher than the 9.1% detected by Xpert (p < 0.001). Compared to the clinical case definition of probable and definite TBM, Ultra exhibited a sensitivity of 88% (44/50), markedly outperforming Xpert (34%, 17/50) and L-J culture (30%, 15/50) (p < 0.001). Against the CMRS, Ultra's sensitivity was 100% (44/44), compared to Xpert (38.6%, 17/44) and L-J culture (34.1%, 15/44) (p < 0.001). AFB microscopy showed very low sensitivity (2.3%, 1/44). Negative predictive values for both clinical case definition and CMRS were higher for Ultra (95.8-100%) than Xpert (80.6-84.1%) and L-J culture (79.7-83.1%) (p < 0.001). Compared to gold standard L-J culture, Ultra's sensitivity (100%, 15/15) was also higher than Xpert (93.3%, 14/15) (p = 0.316), with specificities of 83.1% (143/172) and 98.3% (169/172), respectively. Ultra detected 26 TBM cases missed by other tests, with most (84.6%) in the 'trace detected' category. Due to the high performance, Ultra can be used widely at health care settings of high TB burden countries like Bangladesh for rapid and accurate diagnosis of paediatric TBM using CSF samples.

Criteria for identifying acute respiratory events based on FEV1 decline in home spirometry for lung transplant patients

BACKGROUND

Lung transplantation is a critical treatment for end-stage lung diseases, but long-term survival is challenged by graft rejection and infection. The detection of adverse respiratory events depends on home spirometry, which can exhibit greater fluctuations than laboratory tests and may not provide timely alerts. Our LT-FollowUp system offers an internet-based platform for daily FEV1 monitoring. This paper explores whether a new algorithm using LT-FollowUp data can detect the clinically significant FEV1 declines that predict adverse respiratory events.

METHODS

A retrospective cohort study of lung transplant patients from the University of Tokyo Hospital was conducted using LT-FollowUp. The accuracy of the algorithm was evaluated using a nested case-crossover study comparing FEV1 declines before acute respiratory events with control periods, and a nested case-time-control study comparing cases with matched controls to adjust for time trends and bias.

RESULTS

Of the 95 patients included in this study, 21 experienced acute respiratory events. The odds ratios derived from conditional logistic regression in the nested case-crossover study and the conditional logistic regression in the nested case-time-control study are 5.42 × 105 and 1, respectively. There is a clear association between abnormal FEV1 decline and acute respiratory events. No clear time trend is observed.

CONCLUSION

The proposed algorithm using LT-FollowUp data shows promise for the real-time detection of respiratory events in lung transplant patients, potentially facilitating early interventions that may prevent chronic lung allograft dysfunction. Further validation in larger, multi-centre studies is needed to confirm these findings and enhance clinical utility.

Use of Portable 24-Hour Polysomnography as Alternative Diagnostic Tool for Narcolepsy Type 1 in Adults and Children

BACKGROUND AND OBJECTIVES

The diagnosis of narcolepsy type 1 (NT1) currently requires the multiple sleep latency test (MSLT), or a nocturnal sleep-onset REM period (SOREMP) combined with typical cataplexy, or alternatively the determination of CSF hypocretin-1 (CSF-hcrt-1) deficiency. We evaluated the 24-hour polysomnography (PSG) recordings in adult and pediatric patients as an alternative diagnostic tool.

METHODS

Patients of any age, referred to the narcolepsy center of a university hospital for suspected central disorder of hypersomnolence (CDH), were consecutively recruited between 2013 and 2022. Participants underwent 2 days (day1-night1-day2-night2) of continuous dynamic PSG followed by MSLT. When consent was given, CSF-hcrt-1 was measured. The accuracy of 24-hour PSG variables from night1 and day2 (index test) was assessed with receiver operating characteristic (ROC) curve analysis in identifying NT1 based on current criteria (applied to night2-PSG, MSLT, and CSF-hcrt1). The markers with area under the curve (AUC) ≥0.75 were then tested in adults and children, separately, and to diagnose NT1 and narcolepsy type 2 (NT2) in different scenarios.

RESULTS

Eight hundred seven patients (30.1% pediatric, 52.4% male) were included, and 709 had CSF-hcrt-1 measured. According to the standard criteria, 322 were diagnosed with NT1 (mean age 26.7 ± 17.1 years, 40.4% pediatric, 54.0% male) and 484 with non-NT1 (mean age 32.7 ± 16.5 years, 23.3% pediatric, 51.3% male), encompassing 31 with NT2, 163 with idiopathic hypersomnia, and 281 with other diagnoses. Detecting SOREMP ≥1 during daytime resulted in AUC = 0.84 (95% CI 0.82-0.87), with 84.4% sensitivity and 84.5% specificity for NT1. Performance was superior to all nighttime-PSG measures (p < 0.001) including nighttime-SOREMP (AUC = 0.77, 95% CI 0.74-0.80; sensitivity = 62.1%, specificity = 91.7%) and did not differ from 24-hour SOREMP ≥1 (AUC = 0.85, 95% CI 0.82-0.87; sensitivity = 89.7%, specificity = 80.2%). The combination of daytime-SOREMP ≥1 with cataplexy showed AUC = 0.89 (95% CI 0.86-0.91) for NT1, superior to the combination of nighttime-SOREMP with cataplexy (AUC = 0.78, 95% CI 0.76-0.81, p < 0.001) and similar to MSLT criteria for narcolepsy (AUC = 0.90, 95% CI 0.88-0.92, p = 0.36). Performances were similar in adults and children. Daytime-SOREMP ≥1 identified NT1 and NT2 combined within all CDH with a sensitivity of 80.8% and specificity of 88.0%.

CONCLUSIONS

The detection of daytime-SOREMP during dynamic 24-hour PSG is more accurate than nighttime-SOREMP for diagnosing narcolepsy and, combined with cataplexy, is comparable with MSLT criteria for the identification of NT1. These results offer the prospect of 24-hour PSG diagnostics for NT1 in the home setting.

CLASSIFICATION OF EVIDENCE

This study provides Class II evidence that daytime SOREMP during a 24-hour PSG accurately distinguishes NT1 in patients with a clinical history of possible cataplexy from those who do not have NT1.

Predicting categories of coronary artery calcium scores from chest X-ray images using deep learning

BACKGROUND

The coronary artery calcium (CAC) score (CACS) is recommended in clinical guidelines for coronary artery disease evaluation. However, it is being replaced by coronary computed tomography angiography as the primary diagnostic tool for patients with stable chest pain. This study aimed to develop and validate a deep learning model for predicting the CACS categories from chest X-ray radiographs (CXRs).

METHODS

We included 10,230 patients with available CXRs and CACSs obtained within six months. Three models were trained based on the CACS thresholds (0, 100, and 400) to distinguish zero from non-zero CACSs, CACSs of <100 and ​≥ ​100, and CACS of <400 and ​≥ ​400. The final CXR integration models incorporating clinical factors, including age, sex, and body mass index, were also trained. All models were evaluated using 10-fold cross-validation. External validation was also performed. We experimentally demonstrated the prognostic value of the predicted CACS for major adverse cardiovascular events, comparing it to the actual CACS classification.

RESULTS

The CACS classification performance of the deep learning model was promising, with areas under the curve (AUCs) of 0.74 (zero vs non-zero), 0.75 (<100 vs. ≥100), and 0.79 (<400 vs. ≥400). The accuracy of the model further improved upon the integration of clinical factors; the AUCs reached 0.77, 0.79, and 0.82, respectively, for the same CACS categories. The external validation results were consistent (AUCs of 0.78, 0.79, and 0.81, respectively).

CONCLUSIONS

The deep learning model effectively classified the CACS from CXRs, especially for cases of severe calcification. This approach can cost-effectively improve coronary artery disease risk assessment and support clinical decision-making while minimizing radiation exposure.

Screening for Streptococcus agalactiae: Development of an Automated qPCR-Based Laboratory-Developed Test Using Panther Fusion® Open AccessTM

Laboratory-developed tests (LDTs) are optimal molecular diagnostic modalities in circumstances such as public health emergencies, rare disease diagnosis, limited budget, or where existing commercial alternatives are unavailable, limited in supply, or withdrawn, either temporarily or permanently. These tests reduce access barriers and enhance equitable clinical practice and healthcare delivery. Despite recommendations for the development of nucleic acid amplification tests, procedural details are often insufficient, inconsistent, and arbitrary. This protocol elucidates the methodology used in the development of a fully automated real-time polymerase chain reaction (qPCR)-based test, using the Panther Fusion® Open AccessTM functionality, for the detection of Streptococcus agalactiae in pregnant women, using selectively enriched rectovaginal swabs. In addition, guidelines are provided for oligonucleotide design (primers and TaqMan probes), in silico and in vitro evaluation of design effectiveness, optimization of the physicochemical conditions of the amplification reaction, and result analysis based on experimental designs and acceptance criteria. Furthermore, recommendations are provided for the analytical and clinical validation of the intended use. Our approach is cost-effective, particularly during the design and optimization phases. We primarily used open-source bioinformatics software and tools for in silico evaluations for the test design. Subsequently, the process was manually optimized using a CFX96 Dx analyzer, whose technical specifications and performance are homologous to that of the final platform (Panther Fusion®). Unlike Panther Fusion®, the CFX96 Dx does not require excess volumes of reagents, samples, and evaluation materials (dead volume) to accommodate potential robotic handling-associated imprecisions. The utilization of the CFX96 Dx analyzer represents a strategic approach to enhancing the efficiency of resources and the optimization of time during LDT optimization. Key features • Efficient and robust use of bioinformatics tools for designing primers and TaqMan probes for qPCR-based detection of Streptococcus agalactiae. • Optimization of physicochemical conditions and evaluation of the design effectiveness of a qPCR-based laboratory-developed test (LDT). • Analytical and clinical validation of the qPCR-based LDT developed on an open-access functionality (Open AccessTM) of an automated in vitro diagnostic platform, Panther Fusion® (Hologic).