Search

Background Temporal dataset shift can cause degradation in model performance as discrepancies between training and deployment data grow over time. The primary objective was to determine whether parsimonious models produced by specific feature selection methods are more robust to temporal dataset shift as measured by out-of-distribution (OOD) performance, while maintaining in-distribution (ID) performance. Methods Our dataset consisted of intensive care unit patients from MIMIC-IV categorized by year groups (2008–2010, 2011–2013, 2014–2016, and 2017–2019). We trained baseline models using L2-regularized logistic regression on 2008–2010 to predict in-hospital mortality, long length of stay (LOS), sepsis, and invasive ventilation in all year groups. We evaluated three feature selection methods: L1-regularized logistic regression (L1), Remove and Retrain (ROAR), and causal feature selection. We assessed whether a feature selection method could maintain ID performance (2008–2010) and improve OOD performance (2017–2019). We also assessed whether parsimonious models retrained on OOD data performed as well as oracle models trained on all features in the OOD year group. Results The baseline model showed significantly worse OOD performance with the long LOS and sepsis tasks when compared with the ID performance. L1 and ROAR retained 3.7 to 12.6% of all features, whereas causal feature selection generally retained fewer features. Models produced by L1 and ROAR exhibited similar ID and OOD performance as the baseline models. The retraining of these models on 2017–2019 data using features selected from training on 2008–2010 data generally reached parity with oracle models trained directly on 2017–2019 data using all available features. Causal feature selection led to heterogeneous results with the superset maintaining ID performance while improving OOD calibration only on the long LOS task. Conclusions While model retraining can mitigate the impact of temporal dataset shift on parsimonious models produced by L1 and ROAR, new methods are required to proactively improve temporal robustness.

Background:Bloodstream infections (BSIs) are a frequent cause of morbidity in patients with acute myeloid leukemia (AML), due in part to the presence of central venous access devices (CVADs) required to deliver therapy.Objective:To determine the differential risk of bacterial BSI during neutropenia by CVAD type in pediatric patients with AML.Methods:We performed a secondary analysis in a cohort of 560 pediatric patients (1,828 chemotherapy courses) receiving frontline AML chemotherapy at 17 US centers. The exposure was CVAD type at course start: tunneled externalized catheter (TEC), peripherally inserted central catheter (PICC), or totally implanted catheter (TIC). The primary outcome was course-specific incident bacterial BSI; secondary outcomes included mucosal barrier injury (MBI)-BSI and non-MBI BSI. Poisson regression was used to compute adjusted rate ratios comparing BSI occurrence during neutropenia by line type, controlling for demographic, clinical, and hospital-level characteristics.Results:The rate of BSI did not differ by CVAD type: 11 BSIs per 1,000 neutropenic days for TECs, 13.7 for PICCs, and 10.7 for TICs. After adjustment, there was no statistically significant association between CVAD type and BSI: PICC incident rate ratio [IRR] = 1.00 (95% confidence interval [CI], 0.75–1.32) and TIC IRR = 0.83 (95% CI, 0.49–1.41) compared to TEC. When MBI and non-MBI were examined separately, results were similar.Conclusions:In this large, multicenter cohort of pediatric AML patients, we found no difference in the rate of BSI during neutropenia by CVAD type. This may be due to a risk-profile for BSI that is unique to AML patients.

Pediatric oncology patients frequently experience episodes of prolonged neutropenia which puts them at high risk for infection with significant morbidity and mortality. Here, we review the data on infection prophylaxis with a focus on both pharmacologic and ancillary interventions. This review does not include patients receiving hematopoietic stem cell transplantation.Patients with hematologic malignancies are at highest risk for infection. Bacterial and fungal prophylaxis decrease the risk of infection in certain high-risk groups. Ancillary measures such as ethanol locks, chlorhexidine gluconate baths, GCSF, IVIG, and mandatory hospitalization do not have enough data to support routine use. There is limited data on risk of infection and role of prophylaxis in patients receiving immunotherapy and patients with solid tumors. Patients with Down syndrome and adolescent and young adult patients may benefit from additional supportive care measures and protocol modifications. Consider utilizing bacterial and fungal prophylaxis in patients with acute myeloid leukemia or relapsed acute lymphoblastic leukemia. More research is needed to evaluate other supportive care measures and the role of prophylaxis in patients receiving immunotherapy.

ImportanceDiagnostic codes are commonly used as inputs for clinical prediction models, to create labels for prediction tasks, and to identify cohorts for multicenter network studies. However, the coverage rates of diagnostic codes and their variability across institutions are underexplored.ObjectivePrimary objective was to describe lab- and diagnosis-based labels for 7 selected outcomes at three institutions. Secondary objectives were to describe agreement, sensitivity, and specificity of diagnosis-based labels against lab-based labels.MethodsThis study included three cohorts: SickKidsPedsfrom The Hospital for Sick Children, and StanfordPedsand StanfordAdultsfrom Stanford Medicine. We included seven clinical outcomes with lab-based definitions: acute kidney injury, hyperkalemia, hypoglycemia, hyponatremia, anemia, neutropenia and thrombocytopenia. For each outcome, we created four lab-based labels (abnormal, mild, moderate and severe) based on test result and one diagnosis-based label. Proportion of admissions with a positive label were presented for each outcome stratified by cohort. Using lab-based labels as the gold standard, agreement using Cohen’s Kappa, sensitivity and specificity were calculated for each lab-based severity level.ResultsThe number of admissions included were: SickKidsPeds(n=59,298), StanfordPeds(n=24,639) and StanfordAdults(n=159,985). The proportion of admissions with a positive diagnosis-based label was significantly higher for StanfordPedscompared to SickKidsPedsacross all outcomes, with odds ratio (99.9% confidence interval) for abnormal diagnosis-based label ranging from 2.2 (1.7-2.7) for neutropenia to 18.4 (10.1-33.4) for hyperkalemia. Lab-based labels were more similar by institution. When using lab-based labels as the gold standard, Cohen’s Kappa and sensitivity were lower at SickKidsPedsfor all severity levels compared to StanfordPeds.ConclusionsAcross multiple outcomes, diagnosis codes were consistently different between the two pediatric institutions. This difference was not explained by differences in test results. These results may have implications for machine learning model development and deployment.

Children with Down syndrome (DS) are at a significantly higher risk of developing acute myeloid leukemia, also termed myeloid leukemia associated with DS (ML-DS). In contrast to the highly favorable prognosis of primary ML-DS, the limited data that are available for children who relapse or who have refractory ML-DS (r/r ML-DS) suggest a dismal prognosis. There are few clinical trials and no standardized treatment approach for this population. We conducted a retrospective analysis of international study groups and pediatric oncology centers and identified 62 patients who received treatment with curative intent for r/r ML-DS between 2000-2021. Median time from diagnosis to relapse was 6.8 (range 1.1 - 45.5) months. Three-year event-free (EFS) and overall survival (OS) were 20.9±5.3% and 22.1±5.4%, respectively. Survival was associated with receipt of hematopoietic stem cell transplantation (HSCT) (HR 0.28), duration of first complete remission (CR1) (HR 0.31 for > 12 months) and attainment of remission after relapse (HR 4.03). Patients who achieved CR prior to HSCT, had an improved OS and EFS of 56.0±11.8% and 50.5±11.9% respectively, compared to those who underwent HSCT without CR (3-year OS and EFS of 10.0±9.5%). Treatment failure after HSCT was predominantly due to disease recurrence (52%) followed by treatment related mortality (10%). The prognosis of r/r ML-DS remains dismal even in the current treatment period and serve as a reference point for current prognostication and future interventional studies. Clinical trials aimed at improving the survival of patients with r/r ML-DS are needed.

Background Molecular diagnostics appear promising for early, non-invasive detection of invasive fungal disease (IFD) in immunocompromised patients. Our clinical lab developed and validated cell free DNA (cfDNA) fungal polymerase chain reaction (PCR) assays (Table 1), which have been in clinical use since November 2020 and were recently included in an institutional pediatric clinical care pathway for prolonged febrile neutropenia. We aimed to evaluate the performance of these plasma cfDNA fungal PCR assays in pediatric oncology and hematopoietic stem cell transplant (HSCT) patients with clinical concern for IFD. Methods We initiated an observational study of inpatient oncology and HSCT patients who had plasma mold panel (+/- Candida panel) cfDNA fungal PCR obtained as part of an IFD evaluation at our freestanding quaternary care children’s hospital. The primary outcome was IFD clinical diagnosis (proven, probable, possible, unlikely) as per EORTC/MSG† definitions within one month of the cfDNA fungal PCR, which was assigned independently by two physicians, with a third physician utilized in cases of discrepancy. Patient demographics, hospital course, imaging results, and clinical laboratory data were abstracted and maintained in an encrypted REDCap© database. Results In a preliminary analysis of October 2021-March 2022 data, there were 21 IFD evaluations for 18 patients (Table 2). Most oncology evaluations were for prolonged febrile neutropenia, while many HSCT were non-neutropenic, but on enhanced immunosuppression with new clinical concerns (e.g., respiratory symptoms, persistent fever). Plasma cfDNA detected a mold or yeast consistent with the clinical presentation in 100% of the five proven/probable cases (Figures 1 & 2). All 14 possible IFD cases had a negative cfDNA fungal PCR. Proven cases are designated with blue icon, while the probable case is orange. All five cases were in oncology patients who did not have history of hematopoietic stem cell transplant. Cell free DNA (cfDNA) fungal polymerase chain reaction (PCR) results include the organism identified, followed by the cycle threshold (Ct) in parenthesis. Maximum Ct for the assay is 45. Abbreviations: AG: Aspergillus galactomannan index (Ref BAL: bronchoalveolar lavage; 1,3-BDG: 1,3-beta-D-glucan (Ref Conclusion Upfront plasma cfDNA fungal PCR successfully detected a relevant mold or yeast with short turnaround time in 100% of cases that were later classified as proven/probable. This appears promising for early, noninvasive diagnosis that enables prompt optimization of antifungal and surgical management, particularly for mucormycosis cases. Additional data may permit consideration of Mucorales agents PCR as a new EORTC/MSG mycologic criteria. Disclosures Lauren E. Kushner, MD, Roman: Client of my husband's business..

Invasive fungal disease is a difficult to diagnose complication of therapy in patients with hematologic malignancy. Antifungal prophylaxis is recommended in high-risk populations, but its use in other populations is less clear. This brief report describes a patient with Trisomy 21 on caspofungin prophylaxis who died of disseminated Trichosporon asahii during induction therapy for new diagnosis low-risk B-cell acute lymphoblastic leukemia, accompanied by a review of similar cases in the literature. Her case highlights the utility of relatively novel diagnostic modalities and reinforces the need for caution in placing patients on antifungal prophylaxis.

Neurological paraneoplastic syndromes are exceedingly rare, and often difficult to recognize clinically. Paraneoplastic achalasia is a condition characterized by new-onset dysphagia that is unrelated to tumor burden, most often due to the development of auto-immune antibodies targeting esophageal tissue. Due to the rarity of this condition, diagnosis is often delayed, leading to increased time to treatment. Here we report a case of a rare paraneoplastic achalasia in a female child with EBV + Hodgkin lymphoma (HL), review literature describing paraneoplastic achalasia, and discuss treatment strategies for improving clinical outcome in these patients.

Background Invasive candidiasis is the most common invasive fungal disease in children and adolescents, but there are limited pediatric-specific antifungal effectiveness data. We compared the effectiveness of echinocandins to triazoles or amphotericin B formulations (triazole/amphotericin B) as initial directed therapy for invasive candidiasis. Methods This multinational observational cohort study enrolled patients aged >120 days and Results Seven-hundred and fifty invasive candidiasis episodes were identified. After exclusions, 541 participants (235 in the echinocandin group and 306 in the triazole/amphotericin B group) remained. Crude failure rates at 14 days for echinocandin and triazole/amphotericin B groups were 9.8% (95% confidence intervals [CI]: 6.0% to 13.6%) and 13.1% (95% CI: 9.3% to 16.8%), respectively. The adjusted 14-day risk difference between echinocandin and triazole/amphotericin B groups was −7.1% points (95% CI: −13.1% to −2.4%), favoring echinocandins. The risk difference was −0.4% (95% CI: −7.5% to 6.7%) at 30 days. Conclusions In children with invasive candidiasis, initial directed therapy with an echinocandin was associated with reduced failure rate at 14 days but not 30 days. These results may support echinocandins as initial directed therapy for invasive candidiasis in children and adolescents. Clinical Trials Registration NCT01869829.

Objective The change in performance of machine learning models over time as a result of temporal dataset shift is a barrier to machine learning-derived models facilitating decision-making in clinical practice. Our aim was to describe technical procedures used to preserve the performance of machine learning models in the presence of temporal dataset shifts. Methods Studies were included if they were fully published articles that used machine learning and implemented a procedure to mitigate the effects of temporal dataset shift in a clinical setting. We described how dataset shift was measured, the procedures used to preserve model performance, and their effects. Results Of 4,457 potentially relevant publications identified, 15 were included. The impact of temporal dataset shift was primarily quantified using changes, usually deterioration, in calibration or discrimination. Calibration deterioration was more common (n = 11) than discrimination deterioration (n = 3). Mitigation strategies were categorized as model level or feature level. Model-level approaches (n = 15) were more common than feature-level approaches (n = 2), with the most common approaches being model refitting (n = 12), probability calibration (n = 7), model updating (n = 6), and model selection (n = 6). In general, all mitigation strategies were successful at preserving calibration but not uniformly successful in preserving discrimination. Conclusion There was limited research in preserving the performance of machine learning models in the presence of temporal dataset shift in clinical medicine. Future research could focus on the impact of dataset shift on clinical decision making, benchmark the mitigation strategies on a wider range of datasets and tasks, and identify optimal strategies for specific settings.

ImportanceTemporal dataset shift associated with changes in healthcare over time is a barrier to deploying machine learning-based clinical decision support systems. Algorithms that learn robust models by estimating invariant properties across time periods for domain generalization (DG) and unsupervised domain adaptation (UDA) might be suitable to proactively mitigate dataset shift.ObjectiveTo characterize the impact of temporal dataset shift on clinical prediction models and benchmark DG and UDA algorithms on improving model robustness.Design, Setting, and ParticipantsIn this cohort study, intensive care unit patients from the MIMIC-IV database were categorized by year groups (2008–2010, 2011–2013, 2014–2016 and 2017–2019). Tasks were predicting mortality, long length of stay, sepsis and invasive ventilation. Feedforward neural networks were used as prediction models. The baseline experiment trained models using empirical risk minimization (ERM) on 2008–2010 (ERM[08-10]) and evaluated them on subsequent year groups. DG experiment trained models using algorithms that estimated invariant properties using 2008–2016 and evaluated them on 2017– 2019. UDA experiment leveraged unlabelled samples from 2017–2019 for unsupervised distribution matching. DG and UDA models were compared to ERM[08-16] models trained using 2008-2016.Main Outcome(s) and Measure(s)Main performance measures were area-under-the-receiver-operating-characteristic curve (AUROC), area-under-the-precision-recall curve and absolute calibration error. Threshold-based metrics including false-positives and false-negatives were used to assess the clinical impact of temporal dataset shift and its mitigation strategies.ResultsIn the baseline experiments, dataset shift was most evident for sepsis prediction (maximum AUROC drop, 0.090; 95% confidence interval (CI), 0.080-0.101). Considering a scenario of 100 consecutively admitted patients showed that ERM[08-10] applied to 2017-2019 was associated with one additional false-negative among 11 patients with sepsis, when compared to the model applied to 2008-2010. When compared with ERM[08-16], DG and UDA experiments failed to produce more robust models (range of AUROC difference, −0.003-0.050).Conclusions and RelevanceDG and UDA failed to produce more robust models compared to ERM in the setting of temporal dataset shift. Alternate approaches are required to preserve model performance over time in clinical medicine.KEY POINTSQuestionCan algorithms that estimate invariant properties across environments for domain generalization and unsupervised domain adaptation improve the robustness of machine learning-derived clinical prediction models to temporal dataset shift?FindingsIn this cohort study using 4 clinical outcomes, domain generalization and unsupervised domain adaptation algorithms did not meaningfully outperform the standard model training algorithm – empirical risk minimization – in learning robust models that generalize over time in the presence of temporal dataset shift.MeaningThese findings highlight the difficulty of improving robustness to dataset shift with purely data-driven techniques that do not leverage prior knowledge of the nature of the shift and the requirement of alternate approaches to preserve model performance over time in clinical medicine.

Simple Summary To investigate whether pediatric and adult chronic myeloid leukemia (CML) have unique molecular characteristics, we studied the transcriptomic signature of pediatric and adult CML cells using high-throughput RNA sequencing. We identified differentially expressed genes and pathways unique to pediatric CML cells compared to adult CML cells. The Rho pathway was significantly dysregulated in pediatric CML cells compared to adult CML cells, suggesting the potential importance in the pathogenesis of pediatric CML. Our study is the first to compare transcriptome profiles of CML across different age groups. A better understanding of the biology of CML across different ages may inform future treatment approaches. Abstract Children with chronic myeloid leukemia (CML) tend to present with higher white blood counts and larger spleens than adults with CML, suggesting that the biology of pediatric and adult CML may differ. To investigate whether pediatric and adult CML have unique molecular characteristics, we studied the transcriptomic signature of pediatric and adult CML CD34+ cells and healthy pediatric and adult CD34+ control cells. Using high-throughput RNA sequencing, we found 567 genes (207 up- and 360 downregulated) differentially expressed in pediatric CML CD34+ cells compared to pediatric healthy CD34+ cells. Directly comparing pediatric and adult CML CD34+ cells, 398 genes (258 up- and 140 downregulated), including many in the Rho pathway, were differentially expressed in pediatric CML CD34+ cells. Using RT-qPCR to verify differentially expressed genes, VAV2 and ARHGAP27 were significantly upregulated in adult CML CD34+ cells compared to pediatric CML CD34+ cells. NCF1, CYBB, and S100A8 were upregulated in adult CML CD34+ cells but not in pediatric CML CD34+ cells, compared to healthy controls. In contrast, DLC1 was significantly upregulated in pediatric CML CD34+ cells but not in adult CML CD34+ cells, compared to healthy controls. These results demonstrate unique molecular characteristics of pediatric CML, such as dysregulation of the Rho pathway, which may contribute to clinical differences between pediatric and adult patients.

Chronic myeloid leukemia (CML) accounts for 2-9% of leukemias in children and adolescents, and occurs with much greater frequency in adults. Compared to adults, children with CML tend to present with higher white blood cell counts and larger spleens, suggesting that the biology of pediatric CML is different from adult CML. We hypothesize that the differences in clinical presentation of pediatric CML are due to unique molecular characteristics that differ from adult CML. To test this hypothesis, we compared the transcriptomic signature of pediatric and adult CML CD34+ cells and healthy age-matched CD34+ cells. CD34+ cells were isolated by FACS from pediatric CML (n=9), adult CML (n=10), pediatric healthy (n=10), and adult healthy (n=10) bone marrow samples. Prepared libraries were sequenced on the Illumina HiSeq 4000 instrument. Raw sequences were trimmed and aligned to the hg38 reference genome with STAR/2.5.1b aligner. Gene level counts were determined with STAR -quantMode option using gene annotations from GENCODE (p5). Differential gene expression and pathway analysis were conducted with R/3.5.3. Counts were normalized with trimmed mean of M-values from the EdgeR/ 3.24.3 package and further transformed with VOOM from the Limma/ 3.38.3 package. A linear model using the empirical Bayes analysis pipeline also from Limma was then used to obtain p-values, adjusted p-values and log-fold changes. Four comparisons were performed: (1) pediatric CML vs pediatric healthy, (2) adult CML vs adult healthy, (3) pediatric CML vs adult CML, and (4) pediatric healthy vs adult healthy. A False Discovery Rate of ≤ .05 and absolute log2 fold-change > 1 was used to define differentially expressed genes (DEGs) in each comparison. To identify potentially unique pathways based on DEG, pathway over-representation was calculated with either goana from the limma package or clueGO. At diagnosis, pediatric patients had higher platelet counts (p=0.001) and larger spleen sizes (p=0.010) than adult patients. Median WBC counts were 273,000 and 143,000 in pediatric and adult patients respectively. A total of 1352 genes were differentially expressed in either adult or pediatric CML CD34+ cells compared to healthy CD34+ cells, 174 of which were expressed similarly in pediatric and adult CML CD34+ cells (54 up- and 120 down-regulated). There were 746 differentially expressed genes (325 up- and 421 down-regulated) in adult CML CD34+ cells compared to adult healthy CD34+ cells, and 432 differentially expressed genes (156 up- and 276 down-regulated) in pediatric CML CD34+ cells compared to pediatric healthy CD34+ cells. In direct comparison of pediatric and adult CML CD34+ cells, 446 genes (270 up and 176 down) were dysregulated in pediatric CML CD34+ cells. Pathway analysis showed that Rho signaling pathway was downregulated in pediatric CML CD34+ cells and several genes in Rho pathway were uniquely dysregulated. ARHGAP27 and VAV2 were significantly upregulated in adult CML CD34+ cells by 3.7-fold (p=0.0453) and 11-fold (p=0.0072), respectively, compared to pediatric CML CD34+ cells. In addition, several genes involved in the NADPH oxidase pathway, one of the best-characterized Rho GTPase-regulated systems, were differently expressed in CML. NCF1, CYBB, and S100A8 were significantly upregulated in adult CML CD34+ cells by 4-fold (p=0.0045), 3.26-fold (p These results demonstrate unique molecular characteristics of pediatric CML that may contribute to the clinical differences at presentation between adult and pediatric disease. A better understanding of the molecular biology of CML across the ages will provide new insights into the pathogenesis of pediatric CML and potentially inform future treatment decisions. Disclosures Davis: Jazz Pharmaceuticals: Research Funding; Novartis Pharmaceuticals: Honoraria. Hijiya: Novartis: Consultancy; Stemline Therapeutics: Consultancy.

BACKGROUND Pediatric oncology patients are at an increased risk of invasive bacterial infection due to immunosuppression. The risk of such infection in the absence of severe neutropenia (absolute neutrophil count ≥ 500/μL) is not well established and a validated prediction model for blood stream infection (BSI) risk offers clinical usefulness. METHODS A 6-site retrospective external validation was conducted using a previously published risk prediction model for BSI in febrile pediatric oncology patients without severe neutropenia: the Esbenshade/Vanderbilt (EsVan) model. A reduced model (EsVan2) excluding 2 less clinically reliable variables also was created using the initial EsVan model derivative cohort, and was validated using all 5 external validation cohorts. One data set was used only in sensitivity analyses due to missing some variables. RESULTS From the 5 primary data sets, there were a total of 1197 febrile episodes and 76 episodes of bacteremia. The overall C statistic for predicting bacteremia was 0.695, with a calibration slope of 0.50 for the original model and a calibration slope of 1.0 when recalibration was applied to the model. The model performed better in predicting high-risk bacteremia (gram-negative or Staphylococcus aureus infection) versus BSI alone, with a C statistic of 0.801 and a calibration slope of 0.65. The EsVan2 model outperformed the EsVan model across data sets with a C statistic of 0.733 for predicting BSI and a C statistic of 0.841 for high-risk BSI. CONCLUSIONS The results of this external validation demonstrated that the EsVan and EsVan2 models are able to predict BSI across multiple performance sites and, once validated and implemented prospectively, could assist in decision making in clinical practice. Cancer 2017. © 2017 American Cancer Society.

Introduction Pediatric chronic myeloid leukemia (CML) accounts for 10-15% of pediatric myeloid leukemias and 2-9% of all pediatric leukemias. There are several unique characteristics of CML diagnosed in children, adolescents, and young adults, compared to adults. They present with higher white blood counts and larger spleens, suggesting that the biology of pediatric CML is different from adult CML. We hypothesize that the differences in clinical presentation of pediatric CML patients are due to unique molecular characteristics that differ from adult CML patients. To test this hypothesis, we studied the transcriptomic signature of pediatric CD34+ CML cells compared to adult CML and normal age-matched bone marrow CD34+ cells. Methods CD34+ cells were isolated by FACS from pediatric CML (n=9), adult CML (n=10), pediatric normal (n=10) and adult normal (n=10) bone marrow samples. Total RNA was isolated from cells, and cDNA libraries were generated. Prepared libraries were sequenced on the Illumina HiSeq 4000 instrument. Raw sequences were trimmed and aligned to the hg38 reference genome with STAR/2.5.1b aligner. Gene level counts were determined with STAR -quantMode option using gene annotations from GENCODE (p5). Differential gene expression and pathway analysis were conducted with R/3.5.3. Counts were normalized with trimmed mean of M-values (TMM) from the EdgeR/ 3.24.3 package and further transformed with VOOM from the Limma/ 3.38.3 package. A linear model using the empirical Bayes analysis pipeline also from Limma was then used to obtain p-values, adjusted p-values and log-fold changes (LogFC). We performed three comparisons: (1) Pediatric CML vs Normal, (2) Adult CML vs Normal, and (3) Pediatric CML vs Adult CML. A False Discovery Rate (FDR) of £ .05 and absolute log2 fold-change > 1 was used to define differentially expressed genes in each comparison. Over-representation analysis was used to identify potentially unique pathways based on differentially expressed genes. Clinical and demographic features at diagnosis were extracted for pediatric and adult CML patients and compared using Fisher's exact test (categorical variables) or Wilcoxon rank sum test (continuous variables). Results Pediatric patients were diagnosed with CML at a median of 11 years (interquartile range (IQR): 10-14) compared to 54 years (IQR: 33-62) for adult patients. At diagnosis, pediatric patients had higher platelet counts (p=0.001) and larger spleen sizes (p=0.010) than adult patients, whereas the white blood cell count and phase at diagnosis did not differ. We found 606 genes (210 up- and 396 down-regulated) differentially expressed in pediatric CML CD34+ cells compared to pediatric normal controls. Interestingly, transcriptional regulators involved in blood cell differentiation including GATA1, TAL1, and KLF1 were differentially enriched in pediatric CML. In comparing adult CML patients to normal adult CD34+ cells, we found 920 genes (379 up- and 541 down-regulated) differentially expressed. Among all dysregulated genes we identified (1352 genes), 174 genes (54 up- and 120-down-regulated) overlapped when comparing pediatric and adult CML patients. Significantly enriched pathways in both adult and pediatric CML cells included PI3K/AKT signaling, MAPK signaling, and Notch/Wnt signaling, which have been previously reported. We found 437 unique genes that were dysregulated only in pediatric CML (270 up- and 167 down-regulated). Notch/Wnt signaling and Rho signaling pathways were significantly enriched. DLC1, a tumor suppressor gene that encodes a RhoGTPase-activating protein, has been known to be downregulated in solid tumors and hematologic malignancies. Interestingly, our data showed that DLC1 is significantly upregulated by 3-fold (p=0.0238) in pediatric CML, but not adult CML CD34+ cells. In addition, we observed that ABR, an inducer of C/EBPa that encodes an activator of RhoGEF and GTPase, was significantly downregulated by 2-fold (p=0.0119) in pediatric but not in adult CML CD34+ cells. Conclusion These results demonstrate unique molecular characteristics of pediatric CML that may contribute to the clinical differences at presentation between adult and pediatric disease. A better understanding of the particular biology of pediatric CML might impact the treatment of those patients in the future. Disclosures Gotlib: Deciphera: Consultancy, Honoraria, Membership on an entity's Board of Directors or advisory committees, Other: co-chair of the Study Steering Committee and Research Funding; Blueprint Medicines Corporation: Consultancy, Honoraria, Membership on an entity's Board of Directors or advisory committees, Other: Chair of the Response Adjudication Committee and Research Funding, Research Funding.

Mucormycosis in pediatric oncology patients is a rare invasive fungal infection associated with significant morbidity and mortality. We describe five patients diagnosed with mucormycosis during induction chemotherapy for acute lymphoblastic leukemia at our institution. All of the patients in our series survived, some in spite of having disseminated disease. Most of the patients' chemotherapy was modified with the aim of controlling their leukemia while minimizing immunosuppression until their fungal infection was under control. Although mucormycosis is frequently fatal, rapid diagnosis and a multidisciplinary approach can lead to excellent outcomes, even in patients undergoing intensive chemotherapy.

10050 Background: The COVID-19 pandemic has affected oncology practice in a variety of ways. We sought to evaluate the impact on pediatric oncology parent and young adult (YA) patient experiences, concerns, and perceived stress. Methods: We conducted a cross-sectional Internet-based survey of parents and YA patients in the pediatric oncology and survivorship clinics at Stanford between June-December 2020. Patients were recruited in person by clinic staff or through the electronic patient health portal. Surveys (available in English and Spanish) included the NIH Perceived Stress Scale (PSS-10) and investigator-developed questions evaluating clinical practice changes, concerns about health and cancer care, and pandemic-related challenges. Bivariate analyses evaluated associations between demographic, clinical, and pandemic-related factors and (a) concern about the pandemic affecting health and cancer care, and (b) perceived stress. Results: Among 81 participants (66 parents, mean age 41.6 ± SD 9.6; 15 YAs, mean age 21.9 ± 8.4 years), 37% self-identified as Hispanic/Latino, 36% non-Hispanic white, and 21% Asian. Twenty-eight percent were on treatment and 47% had completed treatment (79% < 5 years prior). Thirty percent reported cancer-related appointment changes, largely rescheduling (75%) and/or switching to telehealth (42%). Nearly half (45%) of parents and 27% of YAs reported feeling ‘very’ or ‘extremely’ concerned about the pandemic affecting their child’s/their health or cancer care. Race/ethnicity emerged as the only demographic feature that was significantly associated with high concern (p = 0.018), with 57% of Hispanic/Latino and only 21% of non-Hispanic white respondents reporting high levels of concern. Specific concerns included fear of severe infection, immunosuppression, and whether infection would change treatment and compromise effectiveness. Parents and YAs reported ‘a lot’ or ‘a great deal’ of challenges in their personal/family life (61%) and work/professional life (48%). Among these were having less support from friends/family (35%), reduced wages/work hours (31%), and job loss (20%) with 20% reporting ≥ 3 challenges. On the PSS-10, stress in the past month was high for parents (mean 30 ± 4) and YAs (mean 31 ± 5.1) on a scale of 0-40. Risk factors for higher stress included: male gender (p = 0.028), less support from family/friends (p = 0.002), and experiencing ≥ 3 pandemic-related challenges (p = 0.013). Conclusions: Our findings confirm the prevalent worry and stress that pediatric oncology patients and families are experiencing during the COVID-19 pandemic. Better communication about cancer care service changes may help to alleviate some concerns. Supportive care resources may also help patients and families cope with psychosocial stressors, particularly among at-risk groups.

We examined the use of Pediatric Cancer Specialty Centers (PCSCs) over time and the length of stay (LOS) in pediatric oncology patients with a diagnosis of febrile neutropenia. PCSCs were defined as Children's Oncology Group and California Children's Services designated centers. We performed a retrospective analysis on all discharges of pediatric (0 to 18) oncology patients with febrile neutropenia in California (1983 to 2011) using the private Office of Statewide Health Planning and Development database. We examined influence of age, sex, race/ethnicity, payer, income, distance, tumor type, and complications on utilization of PCSCs and LOS (SAS 9.2). Analysis of 24,559 pediatric oncology febrile neutropenia discharges showed hospitalizations in PCSCs increasing from 48% in 1983 to 94% in 2011. The adjusted regression analysis showed decreased PCSC utilization for ages 15 to 18, Hispanic patients, and those living40 miles away. The median PCSC LOS was 9 days compared with 7 days at a non-PCSC (P0.0001). Discharge from a PCSC was associated with a LOS8 days after controlling for complications. Inpatient PCSC care for febrile neutropenia in California has increased since 1983. Receiving care at a PCSC is influenced by age, tumor type, ethnicity, geography, and complications.

Invasive fungal disease (IFD) remains a major cause of morbidity and mortality in pediatric patients after allogeneic hematopoietic stem cell transplant (HSCT). We analyzed the outcome of 152 consecutive pediatric patients who underwent allogeneic HSCT from 2005 to 2012: 126 of these without a history of IFD and 26 with IFD before HSCT. Antifungal prophylaxis agent was determined by the primary transplant attending. The rate of IFD after HSCT among patients with or without prior IFD was similar (7.7% with and 7.1% without a history of fungal disease before transplant). Mortality in these 2 populations did not differ (35% vs. 28%, P=0.48, χ). Patients deemed at higher risk for IFD were generally placed on voriconazole prophylaxis; however, this did not affect rates of posttransplant IFD. All-cause mortality in patients with posttransplant IFD was significantly higher than those without posttransplant IFD (67% vs. 21%, P

INTRODUCTION Pediatric acute myeloid leukemia (AML) requires intensive chemotherapy that results in prolonged periods of neutropenia, placing patients at significant risk for life threatening infection. Current supportive care guidelines strongly recommend hospitalization after chemotherapy completion until neutrophil recovery. While approximately 70% of United States (US) centers adhere to this recommendation, there are little data to support inpatient over outpatient management. Therefore, we compared clinical outcomes (Aim 1), patient experiences (Aim 2), and patient quality of life (QoL) (Aim 3), between the two strategies in a nationally representative sample of newly diagnosed AML patients METHODS Aim 1 employed retrospective, standardized chart abstraction for AML patients treated between 2011 and 2019 at 17 centers in the US. Patients were observed from their first course of chemotherapy through treatment completion, stem cell transplant, site transfer, relapse, or death. The unit of analysis was chemotherapy courses, the primary exposure was inpatient versus outpatient management, and primary outcomes were bacteremia and time to next treatment course. Only courses in which the patient met "discharge eligible" criteria were included (Figure 1). Analyses excluded Induction I given the low rate of outpatient management. Log-binomial regression was used to compare the bacteremia incidence and generalized linear regression was used to compare time to next course. Generalized estimating equations were used in overall analyses to account for correlation between courses from the same patient. For Aim 2, patients/ families at 9 centers underwent semi-structured interviews that were transcribed and analyzed using a conventional content analysis approach (InVivo). For Aim 3, parent proxies at 14 centers completed a PedsQL before the start of chemotherapy and then after neutrophil recovery but prior to the start of the subsequent course. Only one chemotherapy course per patient was considered. Only "discharge eligible" patients were included. Parent-proxy generic core QoL scores at the end of the chemotherapy cycle were compared by exposure group using analysis of covariance accounting for baseline scores. RESULTS 573 patients contributed 1188 treatment courses for Aim 1 (Figure 1). Mean age was 9 ±6 years, 47% were female, and AML was classified as low, intermediate, and high risk, respectively in 65%, 4%, and 31%, respectively. Overall rates of bacteremia were not significantly different in patients receiving outpatient versus inpatient management (23.8% vs 29.0%, RR 0.76, 95% CI: 0.56, 1.03; p=0.07). Days to next course also did not differ between the groups (30.2 vs. 31.6, adjusted difference -1.6, 95% CI: -3.2, 0.5; p=0.17). In Aim 2, 86 interviews in 57 families (39 inpatient, 18 outpatient) were performed. 86% of families receiving inpatient management expressed satisfaction and 85% receiving outpatient management expressed satisfaction. Dissatisfaction with inpatient management was driven by concerns for hospital acquired infections and separation from family. Dissatisfaction with outpatient management stemmed from stress of caring for a neutropenic child at home. Patients/families reporting satisfaction with outpatient management emphasized that the approach would not be appropriate for all families. For Aim 3, mean (±SD) parent proxy PedsQL score did not differ between outpatient (74.3 ±18.3) and inpatient management (70.3 ±19.2), adjusted mean difference 3.95, 95% CI: -5.57, 13.5. CONCLUSION In this nationally representative sample of AML patients, outpatient management had bacteremia rates similar to inpatient management, with no difference in time to the start of next treatment course or patient QoL. Semi-structured interviews revealed strong alignment between patient/family satisfaction and center discharge practice. However, families experiencing outpatient management note high stress in caring for profoundly neutropenic patients at home and that this strategy would not be suitable for all families. These clinical and patient-centered results suggest that outpatient management during neutropenia is a viable approach without excess risk for children with AML. However, implementation studies are needed to identify patient/family characteristics that portend a positive experience with an outpatient strategy. Disclosures Raetz: Pfizer: Research Funding. Rubnitz:AbbVie: Research Funding. Fisher:Pfizer: Research Funding; Astellas: Other: Data Safety Monitoring Board Chair for an antifungal study; Merck: Research Funding.

Introduction Pediatric chronic myeloid leukemia (CML) accounts for 10 to 15% of children with myeloid leukemia and 2 to 9% of all pediatric leukemias. Prior to the discovery of tyrosine kinase inhibitors (TKI) such as imatinib, stem cell transplantation was the only curative treatment for both adults and children with CML. However, due to the small numbers of patients, standardized treatment approaches for pediatric CML have not been established. There are several unique characteristics of CML diagnosed in children and adolescents, and young adults (AYA; 16-29 years), compared to adults. Children and AYA with CML present with a higher white blood count and have larger spleens, higher peripheral blast counts, and lower hemoglobin levels, suggesting that the biology of pediatric CML is different than adult CML. In addition, potential side effects of TKIs unique to pediatric CML patients include impaired bone growth, fertility and immune function, however none have been extensively studied. We hypothesize that the differences in clinical presentation of pediatric CML patients are due to unique molecular characteristics that are absent in adult CML patients. To test this hypothesis, we studied the transcriptomic signature of pediatric CD34+ CML cells compared to adult CML and normal age-matched bone marrow CD34+ cells. Methods CD34+ cells were isolated from pediatric CML (n=7), adult CML (n=8), pediatric normal (n=2) and adult normal (n=3) bone marrow samples. Total RNA was isolated from cells, and then cDNA libraries were generated. Prepared libraries were sequenced on the Illumina HiSeq 4000 instrument. We aligned reads using the HISAT2 alignment software, and mapped to genes with HT-Seq. We removed genes that had zero reads across all the samples, resulting in a set of 4,696 genes that were detected in one or more samples. In case of technical replicates, we used mean of replicates. We performed three differential expression comparisons with edgeR: (1) Pediatric CML vs Adult CML, (2) Adult CML vs Adult Normal, and (3) Pediatric CML vs Pediatric Normal. We used a False Discovery Rate (FDR) of £ 20% and absolute log2 fold-change ³ 1 for selecting differentially expressed genes in each comparison. We used Fisher's exact test to identify significant KEGG pathways for the differentially expressed genes in each comparison. Results Pediatric CML vs Adult CML We found 24 differentially expressed genes (15 over- and 9 under-expressed). Though no pathway was found to be significant at the false discovery rate (FDR) £ 20%, we identified a number of sub-pathways that are relevant. For example, the Chemokine Signaling pathway shows at the top of the list (ordered by raw p-value) because of two genes, XCR1 and HCK, associated with VEGF and MAPK pathways involved in cell proliferation, angiogenesis, DNA repair, and cancer pathogenesis. Adult CML vs Adult Normal We found 60 genes (30 over- and 30 under-expressed) differentially expressed when comparing adult CML patients to normal adults. Ten genes overlapped with 24 genes we identified when comparing pediatric and adult CML patients. We found 11 pathways as significant at FDR £ 10%. Multiple pathways, including Cell adhesion, allograft rejection, Graft versus Host Disease, and Type I diabetes pathways, showed downregulation of MHC, with subsequent downstream reduction in expression of apoptosis-related genes. The IL-17 pathway makes sense, as MAPK, well-known to be associated with various cancers, is down-regulated. Lastly, in the NK pathway the gene DAP12 is up-regulated. This gene is known as a tyrosine kinase binding protein, and although tyrosine kinase inhibitors are the standard treatment for CML, the role of DAP12 in relation to leukemia has not yet been described. Pediatric CML vs Pediatric Normal We found 509 genes (350 over- and 159 under-expressed) differentially expressed in pediatric CML patients compared to normal. Interestingly, transcriptional regulators are differentially enriched in the hematopoietic stem cell differentiation function group including GATA1, GATA2, KLF1 and KLF2. RFC is down-regulated. RFC is a mismatch repair gene known to be involved in colorectal cancer. Many of the significant pathways are involved in glucose and fatty acid metabolism. Our pilot study identified novel molecular features of pediatric CML bone marrow stem cells, providing new insights into the novel biomarkers and pathogenesis of pediatric CML. Disclosures Gotlib: Blueprint Medicines: Consultancy, Honoraria, Research Funding; Promedior: Research Funding; Deciphera: Consultancy, Honoraria, Research Funding; Incyte: Consultancy, Honoraria, Research Funding; Kartos: Consultancy; Celgene: Consultancy, Honoraria, Research Funding; Gilead: Consultancy, Research Funding; Novartis: Consultancy, Honoraria, Research Funding.

Treatment-related myelodysplastic syndrome/acute myeloid leukemia (t-MDS/t-AML) is a devastating complication for pediatric patients with osteosarcoma. Even with intensive chemotherapy, overall survival ranges from 10-26%. These regimens cause significant morbidity, poor quality of life and increased toxicity, at times precluding the ultimate goal of hematopoietic cell transplant (HCT). Chromatin remodeling therapy with decitabine and vorinostat has shown some promise in adults with relapsed/refractory AML and MDS. The synergistic combination of methotrexate and asparaginase, also referred to as 'Capizzi methotrexate', is not generally part of upfront treatment for AML but does have some activity in relapsed/refractory disease. Using this rationale, we report the use of this combination regimen in 2 children with t-AML where chromatin remodeling therapy with subsequent Capizzi methotrexate successfully bridged both patients to HCT. Patients were treated at Lucile Packard Children's Hospital Stanford between 2010 and 2018 for osteosarcoma and subsequently for t-AML. An overview of the treatments used is described in Table 1. An adolescent female was treated for metastatic osteosarcoma. She developed an isolated pulmonary recurrence 5 months after completing therapy. She underwent resection of her pulmonary nodule and then was enrolled on a clinical trial. Four months later, she developed another pulmonary recurrence. At this time, she began treatment with ifosfamide and etoposide. After the sixth cycle, she was noted to have blasts in her blood, and a bone marrow evaluation confirmed t-MDS. Cytogenetics were negative for deletion of 5q, 7q or MLL rearrangement. Repeat bone marrow evaluation two months later revealed progression to t-AML. She initially received 5 days of high dose cytarabine without response and was then transitioned to Treatments A and B. Bone marrow evaluation on day 30 of Treatment B showed morphologic remission with 11% MRD. She underwent Treatment C and then proceeded to a matched sibling HCT. She had no detectable osteosarcoma at the time of her transplant. The patient remained in remission from t-AML but subsequently died from recurrent osteosarcoma more than 2 years after HCT. An adolescent male was treated for non-metastatic osteosarcoma of the left distal femur with methotrexate, adriamycin, cisplatin, ifosfamide and etoposide. Six years after completing treatment, he developed weight loss, epistaxis and pancytopenia. Bone marrow aspiration revealed a hypercellular marrow with 18% blasts, increased marrow fibrosis and cytogenetics with a complex del(7q)/+8 clone, consistent with t-MDS. He received Treatment B, and a bone marrow evaluation on day 25 showed a hemodilute specimen with MRD The patient received an alpha/beta T cell depleted haploidentical transplant. He remains in morphologic remission with no evidence of disease more than 3 months after HCT. Treatment-related MDS/AML remains a devastating complication of osteosarcoma treatment with no standard treatment. The two cases presented here demonstrated good response to a novel combination of therapy with decitabine/vorinostat followed by Capizzi methotrexate. Given the significant toxicity associated with most regimens for t-MDS/t-AML, this regimen should be considered when first line salvage chemotherapy has failed. The precise regimen as well as the optimal number of cycles prior to stem cell transplant should be evaluated in the context of a clinical trial. Disclosures No relevant conflicts of interest to declare.

Efficacy of therapeutic strategies relative to patient- and family-centered outcomes in pediatric oncology must be assessed. We sought to identify outcomes important to children with acute myeloid leukemia and their families related to inpatient versus at-home management of neutropenia. We conducted qualitative interviews with 32 children ≥8 years old and 54 parents. Analysis revealed the impact of neutropenia management strategy on siblings, parent anxiety, and child sleep quality as being outcomes of concern across respondents. These themes were used to inform the design of a questionnaire that is currently being used in a prospective, multiinstitutional comparative effectiveness trial.

Objective To evaluate the practice of empiric antibiotics for febrile, nonneutropenic pediatric oncology patients with a central venous catheter (CVC) in place. Study design Episodes of fever without neutropenia (absolute neutrophil count [ANC] ≥500 cells/mm 3 ) were reviewed retrospectively in pediatric oncology patients with a CVC undergoing chemotherapy. Characteristics and symptoms were compared between patients with bacteremia and patients without bacteremia. Results A total of 392 episodes of nonneutropenic fever in 138 subjects (52 females; 38%) were reviewed. In this cohort, the median age at an episode was 7 years, and the majority of patients had a diagnosis of acute leukemia (54%). Median ANC was 3100 cells/mm 3 (IQR, 1570-5980 cells/mm 3 ). Median temperature was 38.7°C (IQR, 38.3-39.2°C). Twenty-four infectious episodes (6%) occurred in 18 subjects, and 5 CVCs required removal; all patients requiring removal admitted and received antibiotics owing to chills. There were no significant difference in age, sex, or ANC between patients with bacteremia and those without bacteremia; however, mean temperature was higher in the patients with bacteremia (39.4°C vs 38.7°C; P = .003). No deaths due to sepsis occurred, and no CVCs were removed because antibiotics were not administered empirically. Conclusion Our practice of observing pediatric oncology patients undergoing chemotherapy with CVCs who are not neutropenic does not appear to lead to increased serious adverse outcomes and avoids antibiotic exposure for >90% of patients without a bacterial infection.

A 10-year-old boy was referred to our clinic for tonsillectomy and was found to have a large mass within his oropharynx. Intraoperative biopsies confirmed Burkitt lymphoma. Further imaging and biopsy revealed pancreatic involvement. He was treated with multiagent chemotherapy. He remains disease-free 6 years later. Review of the literature demonstrates other cases of non-Hodgkin lymphoma with pancreatic involvement with good outcomes. Pancreatic involvement is a relatively rare occurrence in childhood lymphoma.

Objective: The goal of this retrospective study was to analyze the incidence and outcome of invasive fungal disease in a large cohort of pediatric patients after allogeneic stem cell transplant at Stanford University from 2005-2012 treated with modern antifungal prophylaxis. In comparison with prior studies, this study is unique in including prophylaxis with voriconazole and dosing information on antifungal prophylaxis, as well as the prevalence of transaminitis in this population and relation to antifungal prophylaxis. Patients and Methods: Data was analyzed for 152 pediatric patients between the ages of 0-21 years who received allogeneic stem cell transplantation for any diagnosis at Stanford University between 2005 and 2012. Both related and unrelated donors were used, including bone marrow, cord blood, and peripheral blood stem cell sources. All patients received antifungal prophylaxis, although the specific medication used and dosing was at the discretion of the treating physician. Patients were categorized as having proven, probable, or possible fungal disease based on the EORTC/MSG classification. Rates of invasive fungal disease and level of transaminitis were analyzed by type of antifungal prophylaxis as well as dosing in milligrams per kilogram. Results: The overall mortality in this cohort was 29%. Of those, 57% of the deaths were due to relapse, 20% due to infectious complications (3 fungal, 3 sepsis, 2 viral, 1 toxoplasmosis), 16% due to organ failure, and 7% due to GVHD. Twenty five patients were diagnosed with fungal disease prior to transplant and these patients were excluded from this analysis. In the 127 patients with no prior history of fungal disease, 60% were on fluconazole prophylaxis, 32% were on voriconazole, 7% were on micafungin or caspofungin and 2% were on amphotericin. Nine of these patients subsequently developed proven or probable invasive fungal disease. Patients who developed invasive fungal disease had a significantly higher mortality rate (67%) compared with those who did not (25%). However, these two groups did not differ on type of antifungal prophylaxis, underlying diagnosis, donor source, product used or age. There was no difference in level of transaminitis based on type of antifungal prophylaxis (voriconazole vs. fluconazole) or dose used in milligrams per kilogram. Conclusions: In those patients who develop invasive fungal disease after transplant, mortality has decreased over time but remains substantial at 67%. This data did not support the use of one antifungal over another for prophylaxis of invasive fungal disease in pediatric patients. Voriconazole and fluconazole caused similar levels of transaminitis. 237