[Feasibility of near-infrared fluorescence imaging in assisting with the determination of the resection range of radiation intestinal injury]

Objective: To investigate the feasibility of near-infrared fluorescence imaging (NIRFI) to assist in determining the resection range of radiation intestinal injury (RII). Methods: A descriptive cohort study was conducted. Clinical data of 10 RII patients who presented intestinal obstruction and received operation with more than 100 cm of small intestine had been resected atGeneral Department of Jinling Hospital from October 2014 to January 2015 were retrospectively analyzed. The Novadaq SPY Intra-operative Imaging System was used in capturing and viewing fluorescent images. Firstly, the dense adhesion was mobilized and the obstructive intestine was fully freed under laparoscopy, then entering into abdomen from the corresponding incision. The surgeon determined the resection range according to the color of the intestinal serous layer of the diseased intestinal wall, the thickness of the intestinal wall, and the degree of swelling of the mesentery. Afterwards, intra-operative NIRFI was performed by intravenous injection of 2 ml indocyanine green (ICG) and the imaging results of the diseased intestinal arteriovenous phase were observed and recorded. The evaluation criteria for the final resection range were mainly based on the changes in mesenteric arterial phase imaging. In RII lesions, mesenteric vessels in mesenteric artery phase were disordered, and the comb-like distribution of normal mesenteric vessels completely disappeared. Only the clouded appearance in the intestinal wall was observed. Imaging results of the diseased intestinal tissue during the development phase and mesenteric vein phase were not significantly different from normal intestinal tissue. Intraoperative and postoperative conditions under NIRFI-assisted positioning, including the resection range, anastomosis site, operation-related complications, hospitalization time and cost were recorded. Data of abdominal contrast-enhanced CT and gastrointestinal angiography during 5 years of follow-up were collected to evaluate whether there was anastomotic stenosis or insufficient resection of diseased bowel. Results: Based on the imaging of mesenteric arterial phase of NIRFI, the median resection length of the small intestine was 185 (120-260) cm. After NIRFI imaging, only local lesion of ileum was excised in 6 patients, and jejunum-ileum anastomosis was performed to preserve ileocecal flap. No serious complications such as anastomotic leakage and anastomotic hemorrhage, or chronic intestinal failure such as short bowel syndrome occurred. The median hospitalization time was 32 (22-51) days, and the median hospitalization cost was 142 000 (90 000-175 000) RMB. The hospitalization time and cost were mainly used for the enteral and parenteral nutrition support treatment during the perioperative period. All the patients had normal oral diet and/or oral enteral nutrient. After 5 years of follow-up, no recurrence was found. Abdominal contrast-enhanced CT and gastrointestinal angiography showed no thickening of the intestinal wall or stenosis of the lumen. Conclusion: Mesenteric arterial phase imagingof NIRFI can help surgeons to determine the site and range of resection of RII lesions.

A Machine Learning Approach for Predicting Early Phase Postoperative Hypertension in Patients Undergoing Carotid Endarterectomy

BACKGROUND

This study aimed to establish and validate a machine learning-based model for the prediction of early phase postoperative hypertension (EPOH) requiring the administration of intravenous vasodilators after carotid endarterectomy (CEA).

METHODS

Perioperative data from consecutive CEA procedures performed from January 2013 to August 2019 were retrospectively collected. EPOH was defined in post-CEA patients as hypertension involving a systolic blood pressure above 160 mm Hg and requiring the administration of any intravenous vasodilator medications in the first 24 hr after a return to the vascular ward. Gradient boosted regression trees were used to construct the predictive model, and the featured importance scores were generated by using each feature's contribution to each tree in the model. To evaluate the model performance, the area under the receiver operating characteristic curve was used as the main metric. Four-fold stratified cross-validation was performed on the data set, and the average performance of the 4 folds was reported as the final model performance.

RESULTS

A total of 406 CEA operations were performed under general anesthesia. Fifty-three patients (13.1%) met the definition of EPOH. There was no significant difference in the percentage of postoperative stroke/death between patients with and without EPOH during the hospital stay. Patients with EPOH exhibited a higher incidence of postoperative cerebral hyperperfusion syndrome (7.5% vs. 0, P < 0.001), as well as a higher incidence of cerebral hemorrhage (3.8% vs. 0, P < 0.001). The gradient boosted regression trees prediction model achieved an average AUC of 0.77 (95% CI 0.62 to 0.92). When the sensitivity was fixed near 0.90, the model achieved an average specificity of 0.52 (95% CI 0.28 to 0.75).

CONCLUSIONS

We have built the first-ever machine learning-based prediction model for EPOH after CEA. The validation result from our single-center database was very promising. This novel prediction model has the potential to help vascular surgeons identify high-risk patients and reduce related complications more efficiently.

Characterization of Urinary N-acetyltaurine as a Biomarker of Serum Acetate in Experimental Animal Models of Hyperacetatemia

Objectives

Acetate is an intermediate metabolite originated from multiple important metabolic pathways. Even though blood acetate level has been associated with many health events, it is not commonly monitored in clinical practice, partially due to the needs for invasive blood collection and the challenges in acetate analysis. N-acetyltaurine (NAT) was earlier identified as a novel urinary metabolite of ethanol from the reaction between taurine and ethanol-derived acetate. As a direct metabolite of acetate, NAT has the potential to function as a urinary biomarker that reflects the acetate level inside the body. To test this hypothesis, this study examined the correlations between serum acetate level and urinary NAT level in three experimental animal models of hyperacetatemia.

Methods

Glycerol-triacetate (GTA) dosing, ethanol dosing, and streptozotocin (STZ)-induced Type 1 diabetes were used to achieve hyperacetatemia in mice. In GTA model, serum samples were collected at 2 h and urine samples were collected for 24 h after dosing the mice with 5.8 g/kg GTA. In ethanol dosing, serum were collected at 2 h after intraperitoneal injection of 4 g/kg ethanol, while 24 h urine samples were collected before and after 14-day feeding of modified semi-solid diet containing 2.2%–6.7% (v/v) ethanol. In the Type I diabetes model, urine and serum samples were collected before and 5 days after the intraperitoneal injection of 180 mg/kg STZ. The concentrations of NAT and creatinine in urine, as well as acetate in serum, were measured using their respective liquid chromatography-mass spectrometry (LC-MS) methods.

Results

The occurrence of hyperacetatemia in three animal models was confirmed by the clear elevation of serum acetate concentrations. The concentrations of urinary NAT were also dramatically increased in three animal models, suggesting the correlations between serum acetate and urinary NAT.

Conclusions

Urinary NAT is an effective metabolic marker of hyperacetatemia in three experimental models. The results warrant further investigation on its application in other pathophysiological conditions and in humans.

Funding Sources

This research was partially supported by the Agricultural Experiment Station project MIN-18–125 (C. C.) from the United States Department of Agriculture (USDA).

Investigating ultrasound-induced acoustic softening in aluminum and its alloys

Ultrasonic vibration has been observed to lower the flow stress necessary to initiate plastic deformation, a phenomenon known as "acoustic softening". This unique effect of ultrasound has been extensively applied in welding, machining, forming of metals, and ultrasonic additive manufacturing to lower the yield stress necessary to initiate plastic deformation, it nevertheless lacks fundamental investigation. Some prior studies showed experimental errors due to the design of experimental setups and the associated testing methods that have been introduced, leading to questions about their observations and conclusions. Therefore, an experimental setup described in this paper is designed to minimize the constraints identified from the setups in prior studies. Three types of aluminum are studied: Al 1100-O a commercially pure aluminum, Al 6061-O an aluminum alloy without precipitate strengthening, and Al 6061-T6 a precipitate-strengthened aluminum alloy. The acoustic softening and residual effect are compared based on the similarities and differences in microstructures of the three types of aluminum. In both acoustic softening and residual effect, linear relations are obtained between stress change and ultrasound intensities. The slope defined by the linear relations, i.e. the acoustic softening factor, depends on the microstructure of the specific material. The underlying mechanism of acoustic softening is associated with the activation of dislocations by ultrasonic energy and subsequently their interactions with other dislocations and precipitates, whereas the residual effects are attributed to the permanent changes in dislocation density due to dislocation annihilation, dynamic annealing, and dislocation-precipitate interaction.

Metabolomic Profiling of Staphylococcus aureus

Metabolomics is becoming increasingly important in bioscience research as it provides a comprehensive analytical platform for a better understanding of the metabolic functions of cells and organisms. Recently, microbial metabolomics has been utilized in diverse research areas, including detection and diagnosis of pathogens, metabolic engineering, and drug discovery. An efficient and reproducible method to measure the intracellular metabolites of a specific microbial organism is a key prerequisite for utilizing metabolome analysis in microbiological research. In this chapter, we describe a workflow focusing on the extraction and quantification of intracellular metabolites of Staphylococcus aureus. Fast quenching with chilled methanol is applied to minimize metabolite leakage, while solvent extraction is used to obtain both polar and nonpolar fractions, which are then analyzed by respective liquid chromatography-mass spectrometry (LC-MS) methods for characterizing and quantifying the intracellular metabolites of S. aureus. This protocol is demonstrated to be an efficient method for analyzing polar and nonpolar intracellular metabolites of S. aureus.

Identification of two novel anti-HCV E2 412-423 epitope antibodies by screening a Chinese-specific phage library

The hepatitis C virus (HCV) E2 412-423 linear epitope has been found to be highly conserved across multiple HCV genotypes. The antibodies against this epitope have broadly neutralizing activity. Considering the poor immunogenicity of the epitope in humans and significant diversity in the global distribution of HCV genotypes, the aim of this study was to construct an anti-HCV phage library by using a series of optimal strategies to screen novel broadly neutralizing antibodies from Chinese donors. mRNA was isolated from peripheral blood samples of 39 patients who were anti-HCV positive. A phage library was constructed by inserting a single-chain variable fragment (scFv) gene repertoire into the T7Select10-3b vector. A synthetic peptide representing the HCV E2 N-terminal 412-423 region was used as "bait" for bio-panning. The binding affinities of phage clones to the synthetic peptide were evaluated through peptide-ELISA. Two scFv clones (R3-19 and R4-85) showing the strongest binding affinities were selected. The complementarity-determining regions (CDRs) of these clones were aligned with those of other previously reported broadly neutralizing anti-HCV antibodies, and multiple conserved amino acid sites were found. The optimized procedures ensured that two novel scFv antibodies were isolated from a constructed phage library and showed specific binding to the poorly immunogenic HCV E2 412-423 linear epitope. Keywords: phage antibody library; hepatitis C virus; broadly neutralizing antibody; synthetic peptide.

Evaluation of Cronobacter sakazakii inactivation and physicochemical property changes of non-fat dry milk powder by cold atmospheric plasma

Cronobacter sakazakii can cause life-threatening infections in neonates. Exposure to contaminated powdered food, especially milk powder, is a major route for C. sakazakii infection. Cold atmospheric plasma (CAP) is well known as a non-thermal method for inactivating microbial pathogens. This study evaluates the effectiveness of CAP on C. sakazakii in non-fat dry milk (NFDM) powder using a fluidized reaction system. The CAP treatments for 20-120 s led to 1.17-3.27 log₁₀ reductions of C. sakazakii. C. sakazakii inactivation increased with increasing flow rate from 8 to 20 L/min. In terms of quality attributes of NFDM after the CAP treatments, no noticeable color changes (ΔE < 1.5) were observed. Moreover, no significant changes in crystallinity, amino acid composition, or phenolic content occurred following a 120s-CAP treatment. These results indicate that this fluidized reaction system combined with CAP can provide an effective antimicrobial activity with minimal effects on some physicochemical properties of NFDM powder.

[Clinical effect and safety of pegylated interferon-α-2b injection (Y shape, 40 kD) in treatment of HBeAg-positive chronic hepatitis B patients]

Objective: To investigate the clinical effect and safety of long-acting pegylated interferon-α-2b (Peg-IFN-α-2b) (Y shape, 40 kD) injection (180 μg/week) in the treatment of HBeAg-positive chronic hepatitis B (CHB) patients, with standard-dose Peg-IFN-α-2a as positive control. Methods: This study was a multicenter, randomized, open-label, and positive-controlled phase III clinical trial. Eligible HBeAg-positive CHB patients were screened out and randomized to Peg-IFN-α-2b (Y shape, 40 kD) trial group and Peg-IFN-α-2a control group at a ratio of 2:1. The course of treatment was 48 weeks and the patients were followed up for 24 weeks after drug withdrawal. Plasma samples were collected at screening, baseline, and 12, 24, 36, 48, 60, and 72 weeks for centralized detection. COBAS® Ampliprep/COBAS® TaqMan® HBV Test was used to measure HBV DNA level by quantitative real-time PCR. Electrochemiluminescence immunoassay with Elecsys kit was used to measure HBV markers (HBsAg, anti-HBs, HBeAg, anti-HBe). Adverse events were recorded in detail. The primary outcome measure was HBeAg seroconversion rate after the 24-week follow-up, and non-inferiority was also tested. The difference in HBeAg seroconversion rate after treatment between the trial group and the control group and two-sided confidence interval (CI) were calculated, and non-inferiority was demonstrated if the lower limit of 95% CI was > -10%. The t-test, chi-square test, or rank sum test was used according to the types and features of data. Results: A total of 855 HBeAg-positive CHB patients were enrolled and 820 of them received treatment (538 in the trial group and 282 in the control group). The data of the full analysis set showed that HBeAg seroconversion rate at week 72 was 27.32% in the trial group and 22.70% in the control group with a rate difference of 4.63% (95% CI -1.54% to 10.80%, P = 0.1493). The data of the per-protocol set showed that HBeAg seroconversion rate at week 72 was 30.75% in the trial group and 27.14% in the control group with a rate difference of 3.61% (95% CI -3.87% to 11.09%, P = 0.3436). 95% CI met the non-inferiority criteria, and the trial group was non-inferior to the control group. The two groups had similar incidence rates of adverse events, serious adverse events, and common adverse events. Conclusion: In Peg-IFN-α regimen for HBeAg-positive CHB patients, the new drug Peg-IFN-α-2b (Y shape, 40 kD) has comparable effect and safety to the control drug Peg-IFN-α-2a.

[Outcomes of splenectomy in relapsed/refractory autoimmune hemolytic anemia]

Objective: To evaluate the outcomes of splenectomy in the treatment of relapsed/refractory autoimmune hemolytic anemia (AIHA). Methods: Retrospective analysis was performed in 30 cases with relapsed/refractory AIHA who were treated with splenectomy in our hospital. The pre- and post-operative blood routine indexes and responses were followed up. Results: Among the 30 relapsed/refractory AIHA patients, 20 were pure AIHA (including 13 patients with warm antibody AIHA, 2 with warm-cold double antibody AIHA and 5 with Coombs negative AIHA) and 10 were Evans syndrome. The short-term response was evaluated 10-14 days after operation, and the overall response rate (ORR) of short-term response was 90% [12 cases in complete response (CR), 6 cases in partial response (PR)] in 20 therapeutic evaluable cases. Among 13 patients with long-term follow-up data, except 3 patients with Evans syndrome died (2 cases were refractory to splenectomy, 1 case relapsed after surgery), the ORR of 10 patients with relapsed/refractory pure AIHA at 6 months and 12 months were 90% (9/10) and 70% (7/10), respectively, with a median follow-up of 14 (4-156) months. At the end of follow-up, 3 cases had maintained CR for more than 3 years. Conclusion: The short-term response of splenectomy as a second-line treatment for relapsed/refractory AIHA is satisfactory, and long-term outcome of splenectomy is up to 70% at 1 year. Approximately one-third of patients could maintain sustained remission.

A comprehensive study of risk factors for post-operative pneumonia following resection of meningioma

BACKGROUND

Post-operative pneumonia (Pop) following meningioma surgery is the dominant systemic complication which could cause serious threats to patients. It is unclear whether hematological biochemical markers are independently associated with the Pop. This study attempted to perform a more comprehensive study of taking both clinical factors and hematological biomarkers into account to promote the management of patients after meningioma surgery.

METHODS

We collected clinical and hematological parameters of 1156 patients undergoing meningioma resection from January 2009 to January 2013. According to whether the symptoms of pneumonia had manifested,patients were divided into the Pop group and the Non-Pop group. We analyzed the distinctions of clinical factors between the two groups. We successively performed univariate and multivariate regression analysis to identify risk factors independently associated with the Pop.

RESULTS

4.4% patients infected with the Pop (51 of 1156). The median age at diagnosis of the Pop patients was significantly older than the Non-Pop group (p = 0.002). There were strike distinctions of post-operative hospital stays between two groups, with 21 days and 7 days each (p < 0.001). On multivariate analysis, tumor relapse (p < 0.001), skull base lesions (p = 0.001), intra-operative blood transfusion (p = 0.018) and cardiovascular diseases (p = 0.001) were linked with increased risk of the Pop following meningioma resection. For hematological biochemical markers, it was the factor of Red blood cell distribution width-standard deviation (RDW-SD) (OR 5.267, 95%CI 1.316, 21.078; p = 0.019) and Neutrophils lymphocytes ratio (NLR) (OR 2.081, 95%CI 1.063, 4.067; p = 0.033) that could appreciably predict the Pop.

CONCLUSIONS

Apart from tumor recurrence, localizations, intra-operative blood transfusion and cardiovascular diseases are independent risk factors for the Pop. We initially found hematological RDW-SD and NLR are also important predictors.

Evaluation of a partially de-oiled microalgae product in nursery pig diets

Although microalgae can be used as a source of energy and macronutrients in pig diets, there is limited information on the use of partially de-oiled microalgae coproducts in swine feeding programs. The objectives of this study were to evaluate the effects of a partially de-oiled microalgae extract (MAE) in nursery pig diets on growth performance and health status. A total of 300 pigs (initial BW = 6.3 ± 2.1 kg) were used in a 42-d experiment. Treatments included a standard corn-soybean meal control diet, and diets containing 1, 5, 10, or 20% MAE replacing primarily corn. The ME content of MAE was calculated from the chemical composition, and diets were formulated to meet or exceed nutrient requirements for nursery pigs. Pigs were stratified by weaning BW into 12 blocks in a randomized complete block design, with sex distributed evenly among blocks. Pens of pigs (5 pigs/pen) were assigned randomly within block to one of five dietary treatments. Pig BW and feed disappearance were recorded weekly. On day 42, 30 pigs were harvested and sections of the jejunum and ileum were collected for gut morphology analysis, and a liver sample was collected for metabolomic analysis using liquid chromatography-mass spectroscopy. Data were analyzed by ANOVA with diet as treatment effect, and contrasts were used to test linear or quadratic effects of dietary MAE inclusion level. Overall, pigs fed 1% and 5% MAE had the greatest (quadratic P < 0.05) ADG, resulting from greater (quadratic P < 0.05) ADFI. There was a tendency for a greater number of pigs requiring injectable treatments (P = 0.16) and a greater mortality (P = 0.14) in pigs fed the control diet than pigs in any of the diets with the MAE. Final BW increased (P < 0.05) for pigs fed 1% and 5% MAE diets. The improvements in ADG were not explained by differences in mucosa height or goblet cell count among dietary treatments. Pigs fed diets containing 1% or 5% MAE had relatively less concentration (P < 0.05) of ammonia in the liver and had changes in metabolites associated with the urea cycle. In conclusion, feeding MAE resulted in increased growth responses and may have beneficial health effects when fed to nursery pigs.

Multicentre validation of a sepsis prediction algorithm using only vital sign data in the emergency department, general ward and ICU

OBJECTIVES

We validate a machine learning-based sepsis-prediction algorithm (InSight) for the detection and prediction of three sepsis-related gold standards, using only six vital signs. We evaluate robustness to missing data, customisation to site-specific data using transfer learning and generalisability to new settings.

DESIGN

A machine-learning algorithm with gradient tree boosting. Features for prediction were created from combinations of six vital sign measurements and their changes over time.

SETTING

A mixed-ward retrospective dataset from the University of California, San Francisco (UCSF) Medical Center (San Francisco, California, USA) as the primary source, an intensive care unit dataset from the Beth Israel Deaconess Medical Center (Boston, Massachusetts, USA) as a transfer-learning source and four additional institutions' datasets to evaluate generalisability.

PARTICIPANTS

684 443 total encounters, with 90 353 encounters from June 2011 to March 2016 at UCSF.

INTERVENTIONS

None.

PRIMARY AND SECONDARY OUTCOME MEASURES

Area under the receiver operating characteristic (AUROC) curve for detection and prediction of sepsis, severe sepsis and septic shock.

RESULTS

For detection of sepsis and severe sepsis, InSight achieves an AUROC curve of 0.92 (95% CI 0.90 to 0.93) and 0.87 (95% CI 0.86 to 0.88), respectively. Four hours before onset, InSight predicts septic shock with an AUROC of 0.96 (95% CI 0.94 to 0.98) and severe sepsis with an AUROC of 0.85 (95% CI 0.79 to 0.91).

CONCLUSIONS

InSight outperforms existing sepsis scoring systems in identifying and predicting sepsis, severe sepsis and septic shock. This is the first sepsis screening system to exceed an AUROC of 0.90 using only vital sign inputs. InSight is robust to missing data, can be customised to novel hospital data using a small fraction of site data and retains strong discrimination across all institutions.

[Efficacy and safety of pegylated interferon α-2b injection (Y shape, 40 kD) in treatment of patients with genotype 1/6 chronic hepatitis C]

Objective: To investigate the efficacy and safety of the new investigational drug pegylated interferon α-2b (Peg-IFN-α-2b) (Y shape, 40 kD) injection (180 µg/week) combined with ribavirin in the treatment of patients with genotype 1/6 chronic hepatitis C (CHC), with standard-dose Peg-IFN-α-2a combined with ribavirin as a positive control. Methods: A multicenter, randomized, open-label, and positive-controlled phase III clinical trial was performed. Eligible patients with genotype 1/6 CHC were screened out and randomly divided into Peg-IFN-α-2b(Y shape, 40kD) group and Peg-IFN-α-2a group at a ratio of 2:1. The patients in both groups were given oral ribavirin for 48 weeks in addition and then followed up for 24 weeks after drug withdrawal. Abbott Real Time HCV Genotype II was used to determine HCV genotype, and Cobas TaqMan quantitative real-time PCR was used to measure HCV RNA level at 0, 4, 12, 24, 48, and 72 weeks. Adverse events were recorded in detail. The primary efficacy endpoint was sustained virological response (SVR), and a non-inferiority test was also performed. Results: A total of 561 patients with genotype 1/6 CHC were enrolled, among whom 529 received treatment; 90.9% of these patients had genotype 1 CHC. The data of the full analysis set showed that SVR rate was 69.80% (95% CI 65.00%-74.60%) in the trial group and 74.16% (95% CI 67.73%-80.59%) in the control group (P = 0.297 0). The data of the per protocol set (PPS) showed that SVR rate was 80.63% (95% CI 76.04%-85.23%) in the trial group and 81.33% (95% CI 75.10%-87.57%) in the control group (P = 0.849 8), and the 95% CI of rate difference conformed to the non-inferiority standard. The analysis of the PPS population showed that of all subjects, 47.9% achieved rapid virologic response, with a positive predictive value of 93.8%. The incidence rate of adverse events was 96.30% in the trial group and 94.94% in the control group, and the incidence rate of serious adverse events was 5.13% in the trail group and 5.06% in the control group. Conclusion: In the regimen of Peg-IFN-α combined with ribavirin for the treatment of genotype 1/6 CHC, the new investigational drug Peg-IFN-α-2b(Y shape, 40 kD) has comparable clinical effect and safety to the control drug Peg-IFN-α-2a.

Entecavir maleate versus entecavir in Chinese chronic hepatitis B predominantly genotype B or C: Results at week 144

Reports on the efficacy and safety of long-term entecavir treatment in chronic hepatitis B (CHB) predominantly genotype B or C are insufficient. This study presents the efficacy and safety of entecavir maleate in Chinese CHB patients. Patients were randomly assigned to receive 48-week treatment with either 0.5 mg/day entecavir (group A) or 0.5 mg/day entecavir maleate (group B), and then all patients received treatment with 0.5 mg/day entecavir maleate from week 49. Two hundred and seventy-five patients with CHB (HBeAg-positive: 218) were analysed, predominantly (98.5%) with genotype B or C. Baseline characteristics were balanced. For the HBeAg-positive CHB patients, the mean HBV DNA level decreased similarly (A: by 6.36 log₁₀ IU/mL vs B: by 6.31 log₁₀ IU/mL) between groups at week 144. The percentages of patients who achieved undetectable HBV DNA were similar (A: 70.59% vs B: 66.67%) between groups. Similar HBeAg loss rates (A: 43.53% vs B: 40.23%; P>.05) and HBeAg seroconversion rates (A: 21.52% vs B: 21.18%) were achieved. For the HBeAg-negative CHB patients, similar reductions in HBV DNA levels from baseline (A: by 6.13 log₁₀ IU/mL vs B: by 5.65 log₁₀ IU/mL) and percentages of patients who achieved undetectable HBV DNA (A: 100% vs B: 100%) were achieved. The overall incidence of adverse events was comparable between groups. In conclusions, 48-week administration of entecavir maleate and entecavir showed similar efficacy and safety in Chinese patients with CHB. Long-term entecavir maleate treatment was effective and safe in CHB patients.

Using Transfer Learning for Improved Mortality Prediction in a Data-Scarce Hospital Setting

Algorithm–based clinical decision support (CDS) systems associate patient-derived health data with outcomes of interest, such as in-hospital mortality. However, the quality of such associations often depends on the availability of site-specific training data. Without sufficient quantities of data, the underlying statistical apparatus cannot differentiate useful patterns from noise and, as a result, may underperform. This initial training data burden limits the widespread, out-of-the-box, use of machine learning–based risk scoring systems. In this study, we implement a statistical transfer learning technique, which uses a large “source” data set to drastically reduce the amount of data needed to perform well on a “target” site for which training data are scarce. We test this transfer technique with AutoTriage, a mortality prediction algorithm, on patient charts from the Beth Israel Deaconess Medical Center (the source) and a population of 48 249 adult inpatients from University of California San Francisco Medical Center (the target institution). We find that the amount of training data required to surpass 0.80 area under the receiver operating characteristic (AUROC) on the target set decreases from more than 4000 patients to fewer than 220. This performance is superior to the Modified Early Warning Score (AUROC: 0.76) and corresponds to a decrease in clinical data collection time from approximately 6 months to less than 10 days. Our results highlight the usefulness of transfer learning in the specialization of CDS systems to new hospital sites, without requiring expensive and time-consuming data collection efforts.