Deep-learning approaches to identify critically Ill patients at emergency department triage using limited information.

STUDY OBJECTIVE: Triage quickly identifies critically ill patients, facilitating timely interventions. Many emergency departments (EDs) use emergency severity index (ESI) or abnormal vital sign triggers to guide triage. However, both use fixed thresholds, and false activations are costly. Prior approaches using machinelearning have relied on information that is often unavailable during the triage process. We examined whether deep-learning approaches could identify critically ill patients only using data immediately available at triage. METHODS: We conducted a retrospective, cross-sectional study at an urban tertiary care center, from January 1, 2012-January 1, 2020. De-identified triage information included structured (age, sex, initial vital signs) and textual (chief complaint) data, with critical illness (mortality or ICU admission within 24 hours) as the outcome. Four progressively complex deep-learning models were trained and applied to triage information from all patients. We compared the accuracy of the models against ESI as the standard diagnostic test, using area under the receiver-operator curve (AUC). RESULTS: A total of 445,925 patients were included, with 60,901 (13.7%) critically ill. Vital sign thresholds identified critically ill patients with AUC 0.521 (95% confidence interval [CI] = 0.519-0.522), and ESI <3 demonstrated AUC 0.672 (95% CI = 0.671-0.674), logistic regression classified patients with AUC 0.803 (95% CI = 0.802-0.804), 2-layer neural network with structured data with AUC 0.811 (95% CI = 0.807-0.815), gradient tree boosting with AUC 0.820 (95% CI = 0.818-0.821), and the neural network model with textual data with AUC 0.851 (95% CI = 0.849-0.852). All successive increases in AUC were statistically significant. CONCLUSION: Deep-learning techniques represent a promising method of augmenting triage, even with limited information. Further research is needed to determine if improved predictions yield clinical and operational benefits.

Reference Laboratory Testing for Neurologic Disorders.

Laboratory testing plays a critical role in the diagnosis and monitoring of patients with neurologic disorders. Although common tests are often performed in a central hospital laboratory, an increasing number of essential but esoteric tests are performed at reference laboratories or other outside health care facilities. In this article, we analyze recent trends in neurologic disease testing within the overall context of reference laboratory testing and discuss strategies to facilitate the provision of high-quality, cost-effective laboratory services.

Mutations in TP53, PIK3CA, PTEN and other genes in EGFR mutated lung cancers: Correlation with clinical outcomes.

INTRODUCTION: The degree and duration of response to epidermal growth factor receptor (EGFR) inhibitors in EGFR mutated lung cancer are heterogeneous. We hypothesized that the concurrent genomic landscape of these tumors, which is currently unknown in view of the prevailing single gene assay diagnostic paradigm in clinical practice, could play a role in clinical outcomes and/or mechanisms of resistance. METHODS: We retrospectively probed our institutional lung cancer database for tumors with EGFR kinase domain mutations that were also evaluated by more comprehensive molecular profiling, and evaluated tumor response to EGFR tyrosine kinase inhibitors (TKIs). RESULTS: Out of 171 EGFR mutated tumor-patient cases, 20 were sequenced using at least a limited comprehensive genomic profiling platform. 50% harbored concurrent TP53 mutation, 10% PIK3CA mutation, 5% PTEN mutation, among others. The response rate to EGFR TKIs, the median progression-free survival (PFS) to TKIs, the percentage of EGFR-T790M TKI resistance and survival had higher trends in EGFR mutant/TP53 wild-type cases when compared to EGFR mutant/TP53 mutant tumors (all p >0.05 without statistical significance); with a significantly longer median PFS in EGFR-exon 19 deletion mutant/TP53 wild-type cancers treated with 1st generation EGFR TKIs (p=0.035). CONCLUSIONS: Concurrent mutations, specifically TP53, are common in EGFR mutated lung cancer and may alter clinical outcomes. Additional cohorts will be needed to determine if comprehensive molecular profiling adds clinically relevant information to single gene assay identification in oncogene-driven lung cancers.

Experience with targeted next generation sequencing for the care of lung cancer: insights into promises and limitations of genomic oncology in day-to-day practice.

INTRODUCTION: Tumor genotyping using single gene assays (SGAs) is standard practice in advanced non-small-cell lung cancer (NSCLC). We evaluated how the introduction of next generation sequencing (NGS) into day-to-day clinical practice altered therapeutic decision-making. METHODS: Clinicopathologic data, tumor genotype, and clinical decisions were retrospectively compiled over 6 months following introduction of NGS assay use at our institution in 82 patient-tumor samples (7 by primary NGS, 22 by sequential SGAs followed by NGS, and 53 by SGAs). RESULTS: SGAs identified abnormalities in 34 samples, and all patients with advanced EGFR-mutated or ALK-rearranged tumors received approved tyrosine kinase inhibitors (TKIs) or were consented for clinical trials. NGS was more commonly requested for EGFR, ALK, and KRAS-negative tumors (p<0.0001). NGS was successful in 24/29 (82.7%) tumors. Of 17 adenocarcinomas (ACs), 11 (7 from patients with ≤15 pack-years of smoking) had abnormalities in a known driver oncogene. This led to a change in decision-making in 8 patients, trial consideration in 6, and off-label TKI use in 2. Of 7 squamous cell (SC) carcinomas, 1 had a driver aberration (FGFR1); 6 had other genomic events (all with TP53 mutations). In no cases were clinical decisions altered (p=0.0538 when compared to ACs). CONCLUSIONS: Targeted NGS can identify a significant number of therapeutically-relevant driver events in lung ACs; particularly in never or light smokers. For SC lung cancers, NGS is less likely to alter current practice. Further research into the cost effectiveness and optimal use of NGS and improved provider training in genomic oncology are warranted.

Gene signature distinguishes patients with chronic ulcerative colitis harboring remote neoplastic lesions.

BACKGROUND: Individuals with ulcerative colitis (UC) are at increased risk for colorectal cancer. The standard method of surveillance for neoplasia in UC by colonoscopy is invasive and can miss flat lesions. We sought to identify a gene expression signature in nondysplastic mucosa without active inflammation that could serve as a marker for remote neoplastic lesions. METHODS: Gene expression was analyzed by complementary DNA microarray in 5 normal controls, 4 UC patients without dysplasia, and 11 UC patients harboring remote neoplasia. Common gene ontology pathways of significantly differentially expressed genes were identified. Expression of genes which were progressively and significantly upregulated from controls to UC without neoplasia, to UC with remote neoplasia were evaluated by real-time polymerase chain reaction. Several gene products were also examined by immunohistochemistry. RESULTS: Four hundred and sixty-eight genes were significantly upregulated, and 541 genes were significantly downregulated in UC patients with neoplasia compared with UC patients without neoplasia. Nine genes (ACSL1, BIRC3, CLC, CREM, ELTD1, FGG, S100A9, THBD, and TPD52L1) were progressively and significantly upregulated from controls to nondysplastic UC to UC with neoplasia. Immunostaining of proteins revealed increased expression of S100A9 and REG1α in UC-associated cancer and in nondysplastic tissue from UC patients harboring remote neoplasia compared with UC patients without neoplasia and controls. CONCLUSIONS: Gene expression changes occurring as a field effect in the distal colon of patients with chronic UC identify patients harboring remote neoplastic lesions. These markers may lead to a more accurate and less invasive method of detection of neoplasia in patients with inflammatory bowel disease.

miR-143 and miR-145 are downregulated in ulcerative colitis: putative regulators of inflammation and protooncogenes.

BACKGROUND: miR-143 and miR-145 are believed to function as colon cancer tumor suppressors, as they inhibit colon cancer cell growth and are downregulated in sporadic colonic tumors. We speculated that miR-143 and miR-145 might also be downregulated and contribute to malignant transformation of colonic epithelium in longstanding ulcerative colitis (UC). METHODS: Biopsies were obtained 20 cm proximal to the anus from individuals with quiescent UC and from normal controls. RNA and proteins were extracted and measured. miR-143 and miR-145 were quantified by real-time polymerase chain reaction (PCR) and miR-145 was also assessed by in situ hybridization. Putative targets of these miRNAs, K-RAS, API5, MEK-2 (miR-143), and IRS-1 (miR-145) were determined by western blotting. To assess the effects of miR-143 and miR-145 on these predicted targets, HCT116 and HCA-7 colorectal cancer cells were transfected with miR-143 and miR-145 and expression levels of these proteins were measured. RESULTS: In UC, miR-143 and miR-145 were significantly downregulated 8.3-fold (3.4-20.1) (P < 0.0001) and 4.3-fold (2.3-7.8) (P < 0.0001), respectively, compared to normal colon. In contrast, IRS-1, K-RAS, API5, and MEK-2 were upregulated in UC, consistent with their assignments as targets of these miRNAs. Furthermore, transfected miR-143 and miR-145 significantly downregulated these proteins in HCT116 or HCA-7 cells. CONCLUSIONS: Compared to normal colonic mucosa, in chronic UC miR-143 and miR-145 were significantly downregulated and their predicted targets, IRS-1, K-RAS, API5, and MEK-2 were upregulated. We postulate that loss of these tumor suppressor miRNAs predispose to chronic inflammation and neoplastic progression in IBD.

Multiplexed quantification of nucleic acids with large dynamic range using multivolume digital RT-PCR on a rotational SlipChip tested with HIV and hepatitis C viral load.

In this paper, we are working toward a problem of great importance to global health: determination of viral HIV and hepatitis C (HCV) loads under point-of-care and resource limited settings. While antiretroviral treatments are becoming widely available, viral load must be evaluated at regular intervals to prevent the spread of drug resistance and requires a quantitative measurement of RNA concentration over a wide dynamic range (from 50 up to 10(6) molecules/mL for HIV and up to 10(8) molecules/mL for HCV). "Digital" single molecule measurements are attractive for quantification, but the dynamic range of such systems is typically limited or requires excessive numbers of compartments. Here we designed and tested two microfluidic rotational SlipChips to perform multivolume digital RT-PCR (MV digital RT-PCR) experiments with large and tunable dynamic range. These designs were characterized using synthetic control RNA and validated with HIV viral RNA and HCV control viral RNA. The first design contained 160 wells of each of four volumes (125 nL, 25 nL, 5 nL, and 1 nL) to achieve a dynamic range of 5.2 × 10(2) to 4.0 × 10(6) molecules/mL at 3-fold resolution. The second design tested the flexibility of this approach, and further expanded it to allow for multiplexing while maintaining a large dynamic range by adding additional wells with volumes of 0.2 nL and 625 nL and dividing the SlipChip into five regions to analyze five samples each at a dynamic range of 1.8 × 10(3) to 1.2 × 10(7) molecules/mL at 3-fold resolution. No evidence of cross-contamination was observed. The multiplexed SlipChip can be used to analyze a single sample at a dynamic range of 1.7 × 10(2) to 2.0 × 10(7) molecules/mL at 3-fold resolution with limit of detection of 40 molecules/mL. HIV viral RNA purified from clinical samples were tested on the SlipChip, and viral load results were self-consistent and in good agreement with results determined using the Roche COBAS AmpliPrep/COBAS TaqMan HIV-1 Test. With further validation, this SlipChip should become useful to precisely quantify viral HIV and HCV RNA for high-performance diagnostics in resource-limited settings. These microfluidic designs should also be valuable for other diagnostic and research applications, including detecting rare cells and rare mutations, prenatal diagnostics, monitoring residual disease, and quantifying copy number variation and gene expression patterns. The theory for the design and analysis of multivolume digital PCR experiments is presented in other work by Kreutz et al.

Molecular biology underlying the clinical heterogeneity of prostate cancer: an update.

CONTEXT: Recent studies have uncovered a number of possible mechanisms by which prostate cancers can become resistant to systemic androgen deprivation, most involving androgen-independent reactivation of the androgen receptor. Genome-wide expression analysis with microarrays has identified a wide array of genes that are differentially expressed in metastatic prostate cancers compared to primary nonrecurrent tumors. Recently, recurrent gene fusions between TMPRSS2 and ETS family genes have been identified and extensively studied for their role in prostatic carcinoma. OBJECTIVE: To review the recent developments in the molecular biology of prostate cancer, including those pertaining to the androgen receptor and the newly identified TMPRSS2-related translocations. DATA SOURCES: Literature review and personal experience. CONCLUSIONS: Prostatic adenocarcinoma is a heterogeneous group of neoplasms with a broad spectrum of pathologic and molecular characteristics and clinical behaviors. Numerous mechanisms contribute to the development of resistance to androgen ablation therapy, resulting in ligand-independent reactivation of the androgen receptor, including amplification, mutation, phosphorylation, and activation of coreceptors. Multiple translocations of members of the ETS oncogene family are present in approximately half of clinically localized prostate cancers. TMPRSS2:ERG gene rearrangement appears to be an early event in prostate cancer and is not observed in benign or hyperplastic prostatic epithelium. Duplication of TMPRSS2:ERG appears to predict a worse prognosis. The relationship between TMPRSS2:ERG gene rearrangement and other morphologic and prognostic parameters of prostate cancer is still unclear.

Inducible heat shock protein 70 prevents multifocal flat dysplastic lesions and invasive tumors in an inflammatory model of colon cancer.

BACKGROUND: Heat shock protein 70 (Hsp70) regulates protein biosynthesis and refolding of denatured proteins. Since Hsp70 participates in recovery from stress injury, we examined the effect of Hsp70 genetic deletion in the azoxymethane (AOM)/dextran sulfate sodium (DSS) model of inflammation and colon cancer. METHODS: Hsp70 mutant mice (Hsp70.1(-/-)/70.3(-/-)) and wild-type (WT) littermates received AOM and three cycles of DSS and were killed 24 weeks later. Tumors were graded for histology and immunostained for p53, adenomatous polyposis coli, beta-catenin, cyclooxygenase-2 (Cox-2) and inducible nitric oxide synthase (iNOS) and sequenced for p53 mutations. RESULTS: Elevated adenomas developed in 4/10 WT mice with no dysplasia in adjacent mucosa. In contrast, 7/8 Hsp70 knock out (KO) mice developed chronic mucosal inflammation and multifocal areas of flat dysplasia and 4/8 progressed to invasive carcinomas arising in a background of flat dysplastic mucosa. These differences in the incidence of flat dysplasia and invasive cancers were significant (P < 0.05). Nuclear p53 was stronger in Hsp70 KO tumors compared with WT tumors, and sequencing confirmed p53 mutations in 2/5 tumors from Hsp70(-/-) versus 0/5 in WT mice. In Hsp70 WT tumors, beta-catenin was predominantly nuclear, compared with membranous beta-catenin in Hsp70(-/-) tumors, suggesting that Hsp70 regulates beta-catenin in colonic tumorigenesis. Cox-2 and iNOS levels were increased in tumors from Hsp70(-/-) mice compared with Hsp70 WT tumors. CONCLUSIONS: Hsp70-deleted mice treated with AOM/DSS develop flat invasive colonic tumors that mimic many histological and molecular features of ulcerative colitis colon cancer. This model will be useful to dissect the role of Hsp70 in inflammatory bowel disease colon cancer.