Estimation of Gestational Age-Specific Reference Intervals for Coagulation Assays in a Neonatal Intensive Care Unit Using Real-World Data.

BACKGROUND: Interpretation of coagulation testing in neonates currently relies on reference intervals (RIs) defined from older patient cohorts. Direct RI studies are difficult, but indirect estimation may allow us to infer normative neonatal distributions from routinely collected clinical data. METHODS: We analyzed first-in-life coagulation testing results from all patients admitted to a level IV neonatal intensive care unit between 1/1/2018-1/1/2024. Results obtained after transfusion of any blood product were excluded. Indirect RIs were estimated across gestational age groups using refineR and compared to currently reported intervals for patients less than one year of age. RESULTS: Prothrombin times (PT) and international normalized ratios (INR) were available for 1,128 neonates, while activated partial thromboplastin times (aPTT) were available for 790 neonates. The indirect RI was 10-25s in preterm, 10-22s in term, and 10-24s in all neonates for PT, 0.7-2.1 in preterm, 0.8-1.8 in term, and 0.8-1.9 in all neonates for INR, and 25-68s in preterm, 25-58s in term, and 25-62s in all neonates for aPTT. Compared to our current intervals, the indirect RIs would flag 58% fewer PT, 43% fewer INR, and 17% fewer aPTT results as abnormal. CONCLUSIONS: Indirectly estimated RIs in neonates admitted to intensive care show substantial divergence from current, first-year-of-life RIs, leading to an abundance of abnormal flags. The associations between these flags and provider behavior, transfusion practice, or clinical outcomes is an area of future exploration.

Validating, Implementing, and Monitoring Machine Learning Solutions in the Clinical Laboratory Safely and Effectively.

BACKGROUND: Machine learning solutions offer tremendous promise for improving clinical and laboratory operations in pathology. Proof-of-concept descriptions of these approaches have become commonplace in laboratory medicine literature, but only a scant few of these have been implemented within clinical laboratories, owing to the often substantial barriers in validating, implementing, and monitoring these applications in practice. This mini-review aims to highlight the key considerations in each of these steps. CONTENT: Effective and responsible applications of machine learning in clinical laboratories require robust validation prior to implementation. A comprehensive validation study involves a critical evaluation of study design, data engineering and interoperability, target label definition, metric selection, generalizability and applicability assessment, algorithmic fairness, and explainability. While the main text highlights these concepts in broad strokes, a supplementary code walk-through is also provided to facilitate a more practical understanding of these topics using a real-world classification task example, the detection of saline-contaminated chemistry panels.Following validation, the laboratorian's role is far from over. Implementing machine learning solutions requires an interdisciplinary effort across several roles in an organization. We highlight the key roles, responsibilities, and terminologies for successfully deploying a validated solution into a live production environment. Finally, the implemented solution must be routinely monitored for signs of performance degradation and updated if necessary. SUMMARY: This mini-review aims to bridge the gap between theory and practice by highlighting key concepts in validation, implementation, and monitoring machine learning solutions effectively and responsibly in the clinical laboratory.

Cardiac Troponin to Adjudicate Subclinical Heart Failure in Diabetic Patients and a Murine Model of Metabolic Syndrome.

BACKGROUND: Cardiovascular disease, kidney health, and metabolic disease (CKM) syndrome is associated with significant morbidity and mortality, particularly from congestive heart failure (CHF). Guidelines recommend measurement of cardiac troponin (cTn) to identify subclinical heart failure (HF) in diabetics/CKM. However, appropriate thresholds and the impact from routine screening have not been elucidated. METHODS: cTnI was assessed using the Abbott high sensitivity (hs)-cTnI assay in outpatients with physician-ordered hemoglobin A1c (Hb A1c) and associated with cardiac comorbidities/diagnoses, demographics, and estimated glomerular filtration rate (eGFR). Risk thresholds used in CKM staging guidelines of >10 and >12 ng/L for females and males, respectively, were used. Multivariate logistic regression was applied. hs-cTnI was assessed in a high-fat-diet induced murine model of obesity and diabetes. RESULTS: Of 1304 patients, 8.0% females and 15.7% males had cTnI concentrations above the risk thresholds. Thirty-one (4.2%) females and 23 (4.1%) males had cTnI above the sex-specific 99% upper reference limit. A correlation between hs-cTnI and Hb A1c (R = 0.2) and eGFR (R = -0.5) was observed. hs-cTnI concentrations increased stepwise based on A1C of <5.7% (median = 1.5, IQR:1.3-1.8), 5.7%-6.4% (2.1, 2.0-2.4), 6.5%-8.0% (2.8, 2.5-3.2), and >8% (2.8, 2.2-4.3). Male sex (P < 0.001), eGFR (P < 0.001), and CHF (P = 0.004) predicted elevated hs-cTnI. Obese and diabetic mice had increased hs-cTnI (7.3 ng/L, 4.2-10.4) relative to chow-fed mice (2.6 ng/L, 1.3-3.8). CONCLUSION: A high proportion of outpatients with diabetes meet criteria for subclinical HF using hs-cTnI measurements. Glucose control is independently associated with elevated cTnI, a finding replicated in a murine model of metabolic syndrome.

Worse fibro-inflammatory activity on diagnostic liver biopsy adversely impacts biochemical remission in autoimmune hepatitis.

BACKGROUND: Autoimmune hepatitis (AIH) patients can present with advanced fibrosis at diagnosis or may progress to the same if biochemical remission on treatment is not achieved. METHODS: We conducted a single-center retrospective analysis of 34 pediatrics and 39 adult AIH patients. Three pathologists, blinded to clinical information, reviewed the diagnostic liver biopsy (DLB) slides of AIH patients. We evaluated the impact of clinical, laboratory, and histopathologic parameters on outcomes including biochemical remission (BR). RESULTS: Incidence of advanced (Ludwig stage 3 or 4) fibrosis on DLB was 45.2 %. AIH patients with advanced fibrosis had higher median Ishak score (p < 0.001) and higher IgG level (p = 0.01) at diagnosis. The incidence of BR at 6-month (31.2% vs. 88.6 %, p = 0.001) and 1-year (68.8% vs. 88.6 %, p = 0.04) post-diagnosis was significantly lower in AIH patients with advanced fibrosis. Although not statistically significant, a higher proportion of AIH patients with advanced fibrosis were on high dose of steroids (58% vs. 37.9 %, p = 0.1) at 1 year post diagnosis. Higher serum IgG level at diagnosis was associated with lower odds of achieving BR at 6-month (p = 0.004) and 1-year (p = 0.03) post-diagnosis in multivariate analysis. Pediatric age at diagnosis (p = 0.02) was associated with higher steroid dose at 1-year post-diagnosis in univariate analysis. CONCLUSIONS: Findings of advanced fibrosis on DLB of AIH patients was accompanied by more pronounced necro-inflammatory activity and higher serum IgG level, which translated to lower rates of BR and higher exposure to steroids during the first year after diagnosis.

Process improvement in thyroid fine needle aspiration: Standardizing number of smears for enhanced adequacy and diagnosis.

Fine-needle aspiration (FNA) is a safe, cost-effective diagnostic procedure used in the evaluation of thyroid nodules. The number of thyroid FNAs has dramatically increased over the past few years. In the absence of standardized procedures regarding the number of needle passes needed for diagnosis and the lack of clarity on the use of conventional smears (CS) versus liquid-based preparations (LBP), the demand of thyroid FNAs has led to increased workload on cytology laboratories, which can negatively affect patient safety. We implemented a standardized two needle passes for CS and collection of all needle rinses and additional pass material in CytoRich Red for ThinPrep LBP and compared the non-diagnostic and diagnostic rates before and after this intervention. There were 290 pre-intervention cases and 348 post-intervention cases; of which, there were 17 (5.9%) non-diagnostic cases of the pre-intervention group and 27 (7.8) non-diagnostic cases of the post-intervention group. There was no statistically significant difference in non-diagnostic and diagnostic rates before and after the change (p = 0.347 by two-tailed Z test).

Automating the Detection of IV Fluid Contamination Using Unsupervised Machine Learning.

BACKGROUND: Intravenous (IV) fluid contamination is a common cause of preanalytical error that can delay or misguide treatment decisions, leading to patient harm. Current approaches for detecting contamination rely on delta checks, which require a prior result, or manual technologist intervention, which is inefficient and vulnerable to human error. Supervised machine learning may provide a means to detect contamination, but its implementation is hindered by its reliance on expert-labeled training data. An automated approach that is accurate, reproducible, and practical is needed. METHODS: A total of 25 747 291 basic metabolic panel (BMP) results from 312 721 patients were obtained from the laboratory information system (LIS). A Uniform Manifold Approximation and Projection (UMAP) model was trained and tested using a combination of real patient data and simulated IV fluid contamination. To provide an objective metric for classification, an "enrichment score" was derived and its performance assessed. Our current workflow was compared to UMAP predictions using expert chart review. RESULTS: UMAP embeddings from real patient results demonstrated outliers suspicious for IV fluid contamination when compared with the simulated contamination's embeddings. At a flag rate of 3 per 1000 results, the positive predictive value (PPV) was adjudicated to be 0.78 from 100 consecutive positive predictions. Of these, 58 were previously undetected by our current clinical workflows, with 49 BMPs displaying a total of 56 critical results. CONCLUSIONS: Accurate and automatable detection of IV fluid contamination in BMP results is achievable without curating expertly labeled training data.

GPT-4 Underperforms Experts in Detecting IV Fluid Contamination.

BACKGROUND: Specimens contaminated with intravenous (IV) fluids are common in clinical laboratories. Current methods for detecting contamination rely on insensitive and workflow-disrupting delta checks or manual technologist review. Herein, we assessed the utility of large language models for detecting contamination by IV crystalloids and compared its performance to multiple, but variably trained healthcare personnel (HCP). METHODS: Contamination of basic metabolic panels was simulated using 0.9% normal saline (NS), with (n = 30) and without (n = 30) 5% dextrose (D5NS), at mixture ratios of 0.10 and 0.25. A multimodal language model (GPT-4) and a diverse panel of 8 HCP were asked to adjudicate between real and contaminated results. Classification performance, mixture quantification, and confidence was compared by Wilcoxon rank sum. RESULTS: The 95% CIs for accuracy were 0.57-0.71 vs 0.73-0.80 for GPT-4 and HCP, respectively, on the NS set and 0.57-0.57 vs 0.73-0.80 on the D5NS set. HCP overestimated severity of contamination in the 0.10 mixture group (95% CI of estimate error, 0.05-0.20) for both fluids, while GPT-4 markedly overestimated the D5NS mixture at both ratios (0.16-0.33 for NS, 0.11-0.35 for D5NS). There was no correlation between reported confidence and likelihood of a correct classification. CONCLUSIONS: GPT-4 is less accurate than trained HCP for detecting IV fluid contamination of basic metabolic panel results. However, trained individuals were imperfect at identifying contaminated specimens implying the need for novel, automated tools for its detection.

Generalizability of a Machine Learning Model for Improving Utilization of Parathyroid Hormone-Related Peptide Testing across Multiple Clinical Centers.

BACKGROUND: Measuring parathyroid hormone-related peptide (PTHrP) helps diagnose the humoral hypercalcemia of malignancy, but is often ordered for patients with low pretest probability, resulting in poor test utilization. Manual review of results to identify inappropriate PTHrP orders is a cumbersome process. METHODS: Using a dataset of 1330 patients from a single institute, we developed a machine learning (ML) model to predict abnormal PTHrP results. We then evaluated the performance of the model on two external datasets. Different strategies (model transporting, retraining, rebuilding, and fine-tuning) were investigated to improve model generalizability. Maximum mean discrepancy (MMD) was adopted to quantify the shift of data distributions across different datasets. RESULTS: The model achieved an area under the receiver operating characteristic curve (AUROC) of 0.936, and a specificity of 0.842 at 0.900 sensitivity in the development cohort. Directly transporting this model to two external datasets resulted in a deterioration of AUROC to 0.838 and 0.737, with the latter having a larger MMD corresponding to a greater data shift compared to the original dataset. Model rebuilding using site-specific data improved AUROC to 0.891 and 0.837 on the two sites, respectively. When external data is insufficient for retraining, a fine-tuning strategy also improved model utility. CONCLUSIONS: ML offers promise to improve PTHrP test utilization while relieving the burden of manual review. Transporting a ready-made model to external datasets may lead to performance deterioration due to data distribution shift. Model retraining or rebuilding could improve generalizability when there are enough data, and model fine-tuning may be favorable when site-specific data is limited.

Genetic characterization of primary and metastatic high-grade serous ovarian cancer tumors reveals distinct features associated with survival.

High-grade serous ovarian cancer (HGSC) is the most lethal histotype of ovarian cancer and the majority of cases present with metastasis and late-stage disease. Over the last few decades, the overall survival for patients has not significantly improved, and there are limited targeted treatment options. We aimed to better characterize the distinctions between primary and metastatic tumors based on short- or long-term survival. We characterized 39 matched primary and metastatic tumors by whole exome and RNA sequencing. Of these, 23 were short-term (ST) survivors (overall survival (OS) < 3.5 years) and 16 were long-term (LT) survivors (OS > 5 years). We compared somatic mutations, copy number alterations, mutational burden, differential gene expression, immune cell infiltration, and gene fusion predictions between the primary and metastatic tumors and between ST and LT survivor cohorts. There were few differences in RNA expression between paired primary and metastatic tumors, but significant differences between the transcriptomes of LT and ST survivors in both their primary and metastatic tumors. These findings will improve the understanding of the genetic variation in HGSC that exist between patients with different prognoses and better inform treatments by identifying new targets for drug development.

Data-Driven Anomaly Detection in Laboratory Medicine: Past, Present, and Future.

BACKGROUND: Anomaly detection is an integral component of operating a clinical laboratory. It covers both the recognition of laboratory errors and the rapid reporting of clinically impactful results. Procedures for identifying laboratory errors and highlighting critical results can be improved by applying modern data-driven approaches. CONTENT: This review will prepare the reader to appraise anomaly detection literature, identify common sources of anomalous results in the clinical laboratory, and offer potential solutions for common shortcomings in current laboratory practices. SUMMARY: Laboratories should implement data-driven approaches to detect technical anomalies and keep them from entering the medical record, while also using the full array of clinical metadata available in the laboratory information system for context-dependent, patient-centered result interpretations.

CIViCdb 2022: evolution of an open-access cancer variant interpretation knowledgebase.

CIViC (Clinical Interpretation of Variants in Cancer; civicdb.org) is a crowd-sourced, public domain knowledgebase composed of literature-derived evidence characterizing the clinical utility of cancer variants. As clinical sequencing becomes more prevalent in cancer management, the need for cancer variant interpretation has grown beyond the capability of any single institution. CIViC contains peer-reviewed, published literature curated and expertly-moderated into structured data units (Evidence Items) that can be accessed globally and in real time, reducing barriers to clinical variant knowledge sharing. We have extended CIViC's functionality to support emergent variant interpretation guidelines, increase interoperability with other variant resources, and promote widespread dissemination of structured curated data. To support the full breadth of variant interpretation from basic to translational, including integration of somatic and germline variant knowledge and inference of drug response, we have enabled curation of three new Evidence Types (Predisposing, Oncogenic and Functional). The growing CIViC knowledgebase has over 300 contributors and distributes clinically-relevant cancer variant data currently representing >3200 variants in >470 genes from >3100 publications.

Distinct clonal identities of B-ALLs arising after lenolidomide therapy for multiple myeloma.

Patients with multiple myeloma (MM) who are treated with lenalidomide rarely develop a secondary B-cell acute lymphoblastic leukemia (B-ALL). The clonal and biological relationship between these sequential malignancies is not yet clear. We identified 17 patients with MM treated with lenalidomide, who subsequently developed B-ALL. Patient samples were evaluated through sequencing, cytogenetics/fluorescence in situ hybridization (FISH), immunohistochemical (IHC) staining, and immunoglobulin heavy chain (IgH) clonality assessment. Samples were assessed for shared mutations and recurrently mutated genes. Through whole exome sequencing and cytogenetics/FISH analysis of 7 paired samples (MM vs matched B-ALL), no mutational overlap between samples was observed. Unique dominant IgH clonotypes between the tumors were observed in 5 paired MM/B-ALL samples. Across all 17 B-ALL samples, 14 (83%) had a TP53 variant detected. Three MM samples with sufficient sequencing depth (>500×) revealed rare cells (average of 0.6% variant allele frequency, or 1.2% of cells) with the same TP53 variant identified in the subsequent B-ALL sample. A lack of mutational overlap between MM and B-ALL samples shows that B-ALL developed as a second malignancy arising from a founding population of cells that likely represented unrelated clonal hematopoiesis caused by a TP53 mutation. The recurrent variants in TP53 in the B-ALL samples suggest a common path for malignant transformation that may be similar to that of TP53-mutant, treatment-related acute myeloid leukemia. The presence of rare cells containing TP53 variants in bone marrow at the initiation of lenalidomide treatment suggests that cellular populations containing TP53 variants expand in the presence of lenalidomide to increase the likelihood of B-ALL development.

Neoadjuvant and Adjuvant Pembrolizumab in Resectable Locally Advanced, Human Papillomavirus-Unrelated Head and Neck Cancer: A Multicenter, Phase II Trial.

PURPOSE: Pembrolizumab improved survival in patients with recurrent or metastatic head and neck squamous-cell carcinoma (HNSCC). The aims of this study were to determine if pembrolizumab would be safe, result in pathologic tumor response (pTR), and lower the relapse rate in patients with resectable human papillomavirus (HPV)-unrelated HNSCC. PATIENTS AND METHODS: Neoadjuvant pembrolizumab (200 mg) was administered and followed 2 to 3 weeks later by surgical tumor ablation. Postoperative (chemo)radiation was planned. Patients with high-risk pathology (positive margins and/or extranodal extension) received adjuvant pembrolizumab. pTR was quantified as the proportion of the resection bed with tumor necrosis, keratinous debris, and giant cells/histiocytes: pTR-0 (<10%), pTR-1 (10%-49%), and pTR-2 (≥50%). Coprimary endpoints were pTR-2 among all patients and 1-year relapse rate in patients with high-risk pathology (historical: 35%). Correlations of baseline PD-L1 and T-cell infiltration with pTR were assessed. Tumor clonal dynamics were evaluated (ClinicalTrials.gov NCT02296684). RESULTS: Thirty-six patients enrolled. After neoadjuvant pembrolizumab, serious (grades 3-4) adverse events and unexpected surgical delays/complications did not occur. pTR-2 occurred in eight patients (22%), and pTR-1 in eight other patients (22%). One-year relapse rate among 18 patients with high-risk pathology was 16.7% (95% confidence interval, 3.6%-41.4%). pTR ≥10% correlated with baseline tumor PD-L1, immune infiltrate, and IFNγ activity. Matched samples showed upregulation of inhibitory checkpoints in patients with pTR-0 and confirmed clonal loss in some patients. CONCLUSIONS: Among patients with locally advanced, HPV-unrelated HNSCC, pembrolizumab was safe, and any pathologic response was observed in 44% of patients with 0% pathologic complete responses. The 1-year relapse rate in patients with high-risk pathology was lower than historical.

Standard operating procedure for curation and clinical interpretation of variants in cancer.

Manually curated variant knowledgebases and their associated knowledge models are serving an increasingly important role in distributing and interpreting variants in cancer. These knowledgebases vary in their level of public accessibility, and the complexity of the models used to capture clinical knowledge. CIViC (Clinical Interpretation of Variants in Cancer - www.civicdb.org) is a fully open, free-to-use cancer variant interpretation knowledgebase that incorporates highly detailed curation of evidence obtained from peer-reviewed publications and meeting abstracts, and currently holds over 6300 Evidence Items for over 2300 variants derived from over 400 genes. CIViC has seen increased adoption by, and also undertaken collaboration with, a wide range of users and organizations involved in research. To enhance CIViC's clinical value, regular submission to the ClinVar database and pursuit of other regulatory approvals is necessary. For this reason, a formal peer reviewed curation guideline and discussion of the underlying principles of curation is needed. We present here the CIViC knowledge model, standard operating procedures (SOP) for variant curation, and detailed examples to support community-driven curation of cancer variants.

Open-Sourced CIViC Annotation Pipeline to Identify and Annotate Clinically Relevant Variants Using Single-Molecule Molecular Inversion Probes.

PURPOSE: Clinical targeted sequencing panels are important for identifying actionable variants for patients with cancer; however, existing approaches do not provide transparent and rationally designed clinical panels to accommodate the rapidly growing knowledge within oncology. MATERIALS AND METHODS: We used the Clinical Interpretations of Variants in Cancer (CIViC) database to develop an Open-Sourced CIViC Annotation Pipeline (OpenCAP). OpenCAP provides methods to identify variants within the CIViC database, build probes for variant capture, use probes on prospective samples, and link somatic variants to CIViC clinical relevance statements. OpenCAP was tested using a single-molecule molecular inversion probe (smMIP) capture design on 27 cancer samples from 5 tumor types. In total, 2,027 smMIPs were designed to target 111 eligible CIViC variants (61.5 kb of genomic space). RESULTS: When compared with orthogonal sequencing, CIViC smMIP sequencing demonstrated a 95% sensitivity for variant detection (n = 61 of 64 variants). Variant allele frequencies for variants identified on both sequencing platforms were highly concordant (Pearson's r = 0.885; n = 61 variants). Moreover, for individuals with paired tumor and normal samples (n = 12), 182 clinically relevant variants missed by orthogonal sequencing were discovered by CIViC smMIP sequencing. CONCLUSION: The OpenCAP design paradigm demonstrates the utility of an open-source and open-access database built on attendant community contributions with peer-reviewed interpretations. Use of a public repository for variant identification, probe development, and variant interpretation provides a transparent approach to build dynamic next-generation sequencing-based oncology panels.

Standard operating procedure for somatic variant refinement of sequencing data with paired tumor and normal samples.

PURPOSE: Following automated variant calling, manual review of aligned read sequences is required to identify a high-quality list of somatic variants. Despite widespread use in analyzing sequence data, methods to standardize manual review have not been described, resulting in high inter- and intralab variability. METHODS: This manual review standard operating procedure (SOP) consists of methods to annotate variants with four different calls and 19 tags. The calls indicate a reviewer's confidence in each variant and the tags indicate commonly observed sequencing patterns and artifacts that inform the manual review call. Four individuals were asked to classify variants prior to, and after, reading the SOP and accuracy was assessed by comparing reviewer calls with orthogonal validation sequencing. RESULTS: After reading the SOP, average accuracy in somatic variant identification increased by 16.7% (p value = 0.0298) and average interreviewer agreement increased by 12.7% (p value < 0.001). Manual review conducted after reading the SOP did not significantly increase reviewer time. CONCLUSION: This SOP supports and enhances manual somatic variant detection by improving reviewer accuracy while reducing the interreviewer variability for variant calling and annotation.

Author Correction: The prognostic effects of somatic mutations in ER-positive breast cancer.

The original version of this Article contained errors in the depiction of confidence intervals in the NF1 BCSS data illustrated in Figure 3b. These have now been corrected in both the PDF and HTML versions of the Article. The incorrect version of Figure 3b is presented in the associated Author Correction.