Search

Seventy-two-hour unscheduled return visits (URVs) by emergency department patients are a key clinical index for evaluating the quality of care in emergency departments (EDs). This study aimed to develop a machine learning model to predict 72 h URVs for ED patients with abdominal pain. Electronic health records data were collected from the Chang Gung Research Database (CGRD) for 25,151 ED visits by patients with abdominal pain and a total of 617 features were used for analysis. We used supervised machine learning models, namely logistic regression (LR), support vector machine (SVM), random forest (RF), extreme gradient boosting (XGB), and voting classifier (VC), to predict URVs. The VC model achieved more favorable overall performance than other models (AUROC: 0.74; 95% confidence interval (CI), 0.69–0.76; sensitivity, 0.39; specificity, 0.89; F1 score, 0.25). The reduced VC model achieved comparable performance (AUROC: 0.72; 95% CI, 0.69–0.74) to the full models using all clinical features. The VC model exhibited the most favorable performance in predicting 72 h URVs for patients with abdominal pain, both for all-features and reduced-features models. Application of the VC model in the clinical setting after validation may help physicians to make accurate decisions and decrease URVs.

Background For female adolescent and young adult (AYA), cancer with treatments may affect their children’s health. Our aim was to determine reliable risk estimates of adverse birth outcomes in AYA cancer survivors and the differential effects of treatments. Methods The study population of 4547 births in the AYA cancer survivor group and 45,463 in the comparison group were identified from two national databases between 2004 and 2014. Detailed maternal health conditions, such as maternal comorbidities, medication use during pregnancy and lifestyles, were adjusted in the statistical analyses. The outcomes included low birth weight, preterm labour, stillbirth, small or large for gestational age, a 5-min Apgar score Results The AYA cancer survivor group had a 9% higher risk of overall adverse birth outcomes (adjusted odds ratio, 1.09; 95% confidence interval, 1.02–1.16), especially low birth weight and preterm labour than the comparison group. The radiotherapy-only group additionally had a higher risk of foetal distress, and a 5-min Apgar score Conclusion AYA cancer survivors, especially those who have received radiotherapy, still have higher risks of adverse birth outcomes after adjusting for detailed maternal health conditions. Preconception counselling and additional surveillance may be warranted in this population.

Detecting human-object interactions (HOI) is an important step toward a comprehensive visual understanding of machines. While detecting non-temporal HOIs (e.g., sitting on a chair) from static images is feasible, it is unlikely even for humans to guess temporal-related HOIs (e.g., opening/closing a door) from a single video frame, where the neighboring frames play an essential role. However, conventional HOI methods operating on only static images have been used to predict temporal-related interactions, which is essentially guessing without temporal contexts and may lead to sub-optimal performance. In this paper, we bridge this gap by detecting video-based HOIs with explicit temporal information. We first show that a naive temporal-aware variant of a common action detection baseline does not work on video-based HOIs due to a feature-inconsistency issue. We then propose a simple yet effective architecture named Spatial-Temporal HOI Detection (ST-HOI) utilizing temporal information such as human and object trajectories, correctly-localized visual features, and spatial-temporal masking pose features. We construct a new video HOI benchmark dubbed VidHOI where our proposed approach serves as a solid baseline., Accepted at ACM ICMR'21 Workshop on Intelligent Cross-Data Analysis and Retrieval. The dataset and source code are available at https://github.com/coldmanck/VidHOI

Background This study investigates the association between exposure to urate-lowering drugs (ULDs) and progression and recovery from chronic kidney disease (CKD). Methods We identified 5860 incident gout patients at Chang Gung Memorial Hospital from 2012 to 2015. Propensity score (PS)-weighted Cox proportional hazards model was used to estimate hazard ratios (HRs) for CKD progression and improvement. A separate analysis was conducted to assess the HR for CKD progression and CKD recovery among those with worsening CKD. Results The incidence of CKD progression among allopurinol, febuxostat and uricosuric agent users were 1.98, 1.88 and 1.64 per 1000 person-days. Compared with allopurinol users, the PS-weighted HR (95% confidence intervals [CIs]) was 1.77 (0.85–1.76) for febuxostat users and 1.37 (0.71–1.37) for uricosuric agent users for CKD progression and 1.43 (1.26–1.62) for febuxostat users and 1.00 (0.88–1.14) for uricosuric agent users for CKD improvement. Compared to allopurinol users, the HRs for CKD progression were 1.14 (0.80–1.66) for febuxostat users and 0.92 (0.67–1.31) for uricosuric agent users. Among 741 patients who had CKD progression, the incidence of CKD recovery was 1.33, 6.21 and 3.53 per 1000 person-days for allopurinol, febuxostat and uricosuric agent users. The HRs (95% CIs) for recovery in febuxostat and uricosuric agent users were 2.17 (1.40–3.47) and 1.80 (1.20–2.83) compared to allopurinol users. Conclusions CKD progression and recovery are common in gout patients using ULDs. Febuxostat and benzbromarone were associated with a similar risk of CKD progression with allopurinol, which has a poorer recovery compared with other ULDs.

The temporal relationships between inflammatory bowel disease (IBD)-associated cutaneous manifestations and IBD remain uncertain, with existing evidence mostly from separate cross-sectional studies. We sought to determine the risks of IBD-related dermatologic diseases before and after the diagnosis of IBD. We identified 2847 cases of IBD and 14,235 matched controls from the Taiwan National Health Insurance Research Database between 2003 and 2014. The risks of cutaneous manifestations before and after the diagnosis of IBD were estimated with multivariable-adjusted analyses. At diagnosis, IBD was associated with atopic dermatitis (odds ratio (OR) = 1.61, 95% confidence interval (CI), 1.14–2.28), erythema nodosum (OR = 7.44, 95%CI, 3.75–14.77), aphthous stomatitis (OR = 2.01, 95%CI, 1.72–2.35), polyarteritis nodosa (OR = 5.67, 95%CI, 2.69–11.98), rosacea (OR = 1.67, 95%CI = 1.19–2.35), and cutaneous T cell lymphoma (OR = 21.27, 95%CI, 2.37–191.00). IBD was associated with the subsequent development of pyoderma gangrenosum (hazard ratio (HR) = 17.79, 95%CI, 6.35–49.86), erythema nodosum (HR = 6.54, 95%CI, 2.83–15.13), polyarteritis nodosa (HR = 2.69, 95%CI, 1.05–6.90), hidradenitis suppurativa (HR = 2.48, 95%CI, 1.03–5.97), psoriasis (HR = 2.19, 95%CI, 1.27–3.79), rosacea (HR = 1.92, 95%CI, 1.39–2.65), and aphthous stomatitis (HR = 1.45, 95%CI, 1.22–1.72). This study clarified the associations and temporal relationships between cutaneous manifestations and IBD, highlighting the need for interdisciplinary care in the patient with specific dermatologic diseases presenting with abdominal symptoms, or the IBD patients with cutaneous lesions.

Given input images, scene graph generation (SGG) aims to produce comprehensive, graphical representations describing visual relationships among salient objects. Recently, more efforts have been paid to the long tail problem in SGG; however, the imbalance in the fraction of missing labels of different classes, or reporting bias, exacerbating the long tail is rarely considered and cannot be solved by the existing debiasing methods. In this paper we show that, due to the missing labels, SGG can be viewed as a "Learning from Positive and Unlabeled data" (PU learning) problem, where the reporting bias can be removed by recovering the unbiased probabilities from the biased ones by utilizing label frequencies, i.e., the per-class fraction of labeled, positive examples in all the positive examples. To obtain accurate label frequency estimates, we propose Dynamic Label Frequency Estimation (DLFE) to take advantage of training-time data augmentation and average over multiple training iterations to introduce more valid examples. Extensive experiments show that DLFE is more effective in estimating label frequencies than a naive variant of the traditional estimate, and DLFE significantly alleviates the long tail and achieves state-of-the-art debiasing performance on the VG dataset. We also show qualitatively that SGG models with DLFE produce prominently more balanced and unbiased scene graphs., Comment: Accepted by ACMMM 2021. Source code will be available at https://github.com/coldmanck/recovering-unbiased-scene-graphs

Background Metformin use in pregnancy is controversial because metformin crosses the placenta and the safety on the fetus has not been well-established. This retrospective study aimed to compare pregnancy outcomes in women with preexisting type 2 diabetes receiving metformin or standard insulin treatment. Methods The cohort of this population-based study includes women of age 20–44 years with preexisting type 2 diabetes and singleton pregnancies in Taiwan between 2003 and 2014. Subjects were classified into three mutually exclusive groups according to glucose-lowering treatments received before and after becoming pregnant: insulin group, switching group (metformin to insulin), and metformin group. A generalized estimating equation model adjusted for patient age, duration of type 2 diabetes, hypertension, hyperlipidemia, retinopathy, and aspirin use was used to estimate the adjusted odds ratio (aOR) and 95% confidence interval (CI) of adverse pregnancy outcomes. Results A total of 1166 pregnancies were identified in the insulin group (n = 222), the switching group (n = 318) and the metformin group (n = 626). The insulin group and the switching group had similar pregnancy outcomes for both the mother and fetus, including risk of primary cesarean section, pregnancy-related hypertension, preeclampsia, preterm birth ( 4000 g), large for gestational age, and congenital malformations. The metformin group had a lower risk of primary cesarean section (aOR = 0.57; 95% CI, 0.40–0.82) and congenital malformations (aOR, 0.51; 95% CI; 0.27–0.94) and similar risk for the other outcomes as compared with the insulin group. Conclusions Metformin therapy was not associated with increased adverse pregnancy outcomes in women with type 2 diabetes as compared with standard insulin therapy.

Indoor localization is a fundamental problem in location-based applications. Current approaches to this problem typically rely on Radio Frequency technology, which requires not only supporting infrastructures but human efforts to measure and calibrate the signal. Moreover, data collection for all locations is indispensable in existing methods, which in turn hinders their large-scale deployment. In this paper, we propose a novel neural network based architecture Graph Location Networks (GLN) to perform infrastructure-free, multi-view image based indoor localization. GLN makes location predictions based on robust location representations extracted from images through message-passing networks. Furthermore, we introduce a novel zero-shot indoor localization setting and tackle it by extending the proposed GLN to a dedicated zero-shot version, which exploits a novel mechanism Map2Vec to train location-aware embeddings and make predictions on novel unseen locations. Our extensive experiments show that the proposed approach outperforms state-of-the-art methods in the standard setting, and achieves promising accuracy even in the zero-shot setting where data for half of the locations are not available. The source code and datasets are publicly available at https://github.com/coldmanck/zero-shot-indoor-localization-release., Accepted at ACM MM 2020. 10 pages, 7 figures. Code and datasets available at https://github.com/coldmanck/zero-shot-indoor-localization-release

Studies have shown the anticancer effects of metformin in vitro. However, whether metformin can prevent cancer recurrence or prolong survival in patients with gastric cancer (GC) and diabetes mellitus (DM) post-gastrectomy remains unknown. We evaluated the beneficial effects of metformin in patients with GC and DM post-gastrectomy. We recruited 2400 patients with GC (1749 without DM, 651 with DM) who underwent surgery between 1997 and 2010. Patients with DM were stratified into metformin (group 1) and non-metformin (group 2) users. Their clinicopathological data were recorded prospectively, and demographics, recurrence-free survival (RFS), and cancer-specific survival (CSS) were compared. Tumour recurrence risk and cause of death were analysed between groups 1 and 2 among patients with DM stratified by tumour stage. We also compared RFS and overall survival among patients with and without DM. Tumour recurrence occurred in 201 patients with GC: 57 (25%) in group 1 and 144 (37%) in group 2. After adjusting for confounders, metformin significantly prolonged CSS (hazard ratio (HR) = 0.54, 95% confidence interval (CI) = 0.38&ndash, 0.77) in patients with stage I&ndash, III GC and DM. In subgroup analysis, metformin users with stage III GC and DM had significantly prolonged CSS compared to non-metformin users (HR = 0.45, 95% CI = 0.30&ndash, 0.68), with an insignificant difference in patients with stage I&ndash, II GC. Adjusted HRs for RFS and CSS were significantly lower in patients with stage I&ndash, III GC and DM than those in patients without DM (0.67 (95% CI = 0.54&ndash, 0.92) and 0.62 (95% CI = 0.50&ndash, 0.77), respectively), with an insignificant difference in patients with stage I GC. Metformin significantly reduces tumour recurrence risk and improves CSS in patients with stage III GC and DM post-gastrectomy. Further prospective studies may confirm the efficacy of metformin as an adjunctive treatment for advanced GC postoperatively.

BACKGROUND: Paroxysmal supraventricular tachycardia (PSVT) is a common arrhythmia. However, its incidence and time course in pregnant women are unclear. This study was conducted to determine the incidence of PSVT in pregnant women by trimester. METHODS: From 2001 to 2012, all pregnant women in Taiwan were monitored for PSVT events. Women who visited the emergency department or were admitted for symptomatic PSVT were enrolled in this study, and those with congenital heart diseases were excluded. RESULTS: A total of 2,387,588 pregnancies (1,623,596 mothers) were analyzed. For the women with no previous history of a PSVT event, the incidence rates of symptomatic PSVT were 15, 33, and 60 per 100,000 pregnancies during the first, second, and third trimester, respectively. For the women with a previous history of PSVT, the incidence rates were 5625, 9525, and 11526 per 100,000 pregnancies, respectively. Most PSVT events occurred during the third trimester. CONCLUSIONS: In this Taiwanese cohort of pregnant women there was a stepwise increase in the incidence of symptomatic PSVT, which peaked during the third trimester. A past history of PSVT was associated with a higher risk of recurrence during pregnancy. We suggest that clinicians should be aware of this trend. Prompt management of PSVT events may prevent maternal and fetal complications.

Additional file 1: Supplemental Table S1. Diseases and outcomes. Supplemental Table S2. List of medications.

// Wei-Heng Kao 1, * , Chang-Fu Kuo 2, 3, * , I-Jun Chou 4 , Lai-Chu See 2, 5, 6 , Wen-Kuan Huang 7 , Meng-Jiun Chiou 2 , Weiya Zhang 3 , Michael Doherty 3 , Chun-Chieh Wang 1, 9 , Jun-Te Hsu 8 , Hsien-Hsin Chen 1 and Ji-Hong Hong 1, 9, 10 1 Department of Radiation Oncology, Chang Gung Memorial Hospital, Taoyuan, Taiwan 2 Division of Rheumatology, Allergy and Immunology, Department of Internal Medicine, Chang Gung Memorial Hospital, Taoyuan, Taiwan 3 Division of Rheumatology, Orthopaedics and Dermatology, School of Medicine, University of Nottingham, Nottingham, UK 4 Department of Paediatrics, Chang Gung Memorial Hospital, Taoyuan, Taiwan 5 Department of Public Health, College of Medicine, Chang Gung University, Taoyuan, Taiwan 6 Biostatistics Core Laboratory, Molecular Medicine Research Center, Chang Gung University, Taoyuan, Taiwan 7 Division of Hematology-Oncology, Department of Internal Medicine, Chang Gung Memorial Hospital, Taoyuan, Taiwan 8 Department of General Surgery, Chang Gung Memorial Hospital, Taoyuan, Taiwan 9 Department of Medical Imaging and Radiological Science, Chang Gung University, Taoyuan, Taiwan 10 Radiation Biology Research Center, Institute for Radiological Research, Chang Gung University/Chang Gung Memorial Hospital, Taoyuan, Taiwan * These authors have contributed equally to this work Correspondence to: Ji-Hong Hong, email: jihong@cgmh.org.tw , fructose1202@gmail.com Keywords: prostate-selective α antagonists; prostate cancer; androgen deprivation therapy; fracture; population-based study Received: May 13, 2017 Accepted: December 24, 2017 Published: January 02, 2018 ABSTRACT Introductions: Prostate-selective α antagonists are recommended for relief of lower urinary tract symptoms in prostate cancer patients despite uncertainty of fracture risk as an addition to androgen deprivation therapy (ADT). The purpose of this study is to estimate fracture risk associated with these medications in prostate cancer patients who did and did not receive ADT. Methods: The Taiwan National Health Insurance database was used to identify prostate cancer patients. We identified all 90-day person-quarters exposed to and not exposed to prostate-selective α antagonists. A generalized estimating equation model was used to estimated adjusted odd ratios (ORs) and 95% confidence intervals (CIs) for fracture associated with prostate-selective α antagonists with consideration for confounding by indication bias using propensity score. Results: During 1997–2008, 16,601 persons received a diagnosis of prostate cancer, among whom 13,694 received ADT. Among prostate cancer patients receiving ADT, fracture was significantly more common in person-quarters with prostate-selective α antagonist use than in quarters without such treatment (OR, 1.08; 95% CI, 1.00–1.18). Prostate-selective α antagonist use was most strongly associated with femur fracture (OR, 1.22; 95% CI, 1.09–1.38), followed by skull fracture (OR, 1.29; 95% CIs: 0.93–1.80). Among patients who did not receive ADT, fracture was more common in person-quarters with prostate-selective α antagonist use than in those without medication use (OR, 1.19; 95% CI, 0.91–1.55). Conclusions: Prostate-selective α antagonist is associated with an increased fracture risk, particular for fractures in skull and femur. Patients should be well-informed on this potential risk before taking prostate-selective α antagonists.

// Shu-Fu Lin 1, 5 , Chang-Fu Kuo 2, 5 , Meng-Jiun Chiou 3 and Shang-Hung Chang 3, 4, 5 1 Division of Endocrinology and Metabolism, Department of Internal Medicine, Chang Gung Memorial Hospital, Chang Gung University, Taoyuan, Taiwan 2 Division of Rheumatology, Allergy and Immunology, Department of Internal Medicine, Chang Gung Memorial Hospital, Taoyuan, Taiwan 3 Office for Big Data Research, Chang Gung Memorial Hospital, Taoyuan, Taiwan 4 Division of Cardiology, Department of Internal Medicine, Chang Gung Memorial Hospital, Taoyuan, Taiwan 5 Chang Gung University, Taoyuan, Taiwan Correspondence to: Shang-Hung Chang, email: afen.chang@gmail.com Keywords: type 1 diabetes, pregnancy, outcomes Received: June 04, 2017 Accepted: August 26, 2017 Published: September 16, 2017 ABSTRACT Pregnancy in women with type 1 diabetes is associated with poor maternal and neonatal outcomes. However, the risk of these outcomes has never been evaluated in an Asian national population. In this work, we report the maternal and fetal outcomes of pregnant women with type 1 diabetes in Taiwan. A total of 2,350,339 pregnancy records created between 2001 and 2012 were obtained from the National Health Insurance database and analyzed. Here, 630 pregnancy records were identified in women having type 1 diabetes. Compared with pregnant women without type 1 diabetes, pregnant women with the disease showed increased risk of multiple adverse outcomes, including preeclampsia, eclampsia, cesarean delivery, adult respiratory distress syndrome, pulmonary edema, sepsis, chorioamnionitis, pregnancy-related hypertension, puerperal cerebrovascular disorders, acute renal failure, and shock. Fetuses of type 1 diabetic mothers were at increased risk of stillbirth, premature birth, large for gestational age, low birth weight, and low Apgar score. Of the studied endpoints, only preeclampsia showed an improvement in the late period (2011–2012) when compared with the early period (2001–2010). These findings reveal that pregnant women with type 1 diabetes are at significantly increased risk of developing many adverse maternal and fetal outcomes. Therefore, pregnancy outcomes in women with type 1 diabetes should be improved.

// Fu-Chao Liu 1, 2, * , Huan-Tang Lin 1, 2, * , Chang-Fu Kuo 3, 4 , Lai-Chu See 5, 6 , Meng-Jiun Chiou 7 , Huang-Ping Yu 1, 2 1 Department of Anesthesiology, Chang Gung Memorial Hospital, Taoyuan, Taiwan 2 College of Medicine, Chang Gung University, Taoyuan, Taiwan 3 Division of Rheumatology, Allergy and Immunology, Chang Gung Memorial Hospital, Taoyuan, Taiwan 4 Division of Rheumatology, Orthopaedics and Dermatology, University of Nottingham, Nottingham, UK 5 Department of Public Health, College of Medicine, Chang Gung University, Taoyuan, Taiwan 6 Biostatistics Core Laboratory, Molecular Medicine Research Center, Chang Gung University, Taoyuan, Taiwan 7 Office for Big Data Research, Chang Gung Memorial Hospital, Taoyuan, Taiwan * These authors contributed equally to this work Correspondence to: Huang-Ping Yu, email: yuhp2001@adm.cgmh.org.tw Keywords: breast cancer, epidemiology, survival outcome, nationwide population study Received: November 02, 2016 Accepted: January 24, 2017 Published: February 09, 2017 ABSTRACT Breast cancer is among the most prevalent cancers in Taiwan. The National Health Insurance database was used to identify patients with breast cancer and estimate the yearly prevalence and incidence of breast cancer between 1997 and 2013. Joinpoint regression analysis was used for the annual percentage change of incidence, prevalence, and survival outcome. Among 12,181,919 female beneficiaries in 2013, the prevalence was 834.37 per 100,000 persons (95% confidence interval, 829.28–839.45) and the incidence was 93.00 per 100,000 person-year (95% confidence interval, 91.27–94.73). The average annual percentage change of the age-standardized breast cancer incidence was 3.5 per 100,000 person-years (3.1–3.8; P 1 was 39.1% (19.2%–59.1%) in 1997 and 21.1% (15.7%-32.0%) in 2008, with an annual percentage change of –0.8 (–1.3 to 2.9), suggesting that the mortality rate was gradually decreasing in patients with comorbidities. In conclusion, 1 in 120 women in Taiwan has breast cancer and the incidence is rising, while the annual percentage change of breast cancer prevalence is decreasing. The mortality rate of breast cancer was essentially stable, but the 1-year, 2-year, and 5-year mortality rates in people with Charlson comorbidity index > 1 were declined.

Editorial, see p 590 Paroxysmal supraventricular tachycardia (PSVT) is the most common symptomatic arrhythmia during pregnancy.1 Although PSVT is usually considered transient and harmless, its association with maternal and fetal outcomes during pregnancy are unknown.2–4 In this study, we used a national population cohort to measure the associations between PSVT events during pregnancy and maternal or fetal outcomes and to evaluate the associations between prior ablation of PSVT and those outcomes. This study was approved by the Institutional Review Board of Chang Gung Memorial Hospital, Taiwan. We obtained records of all pregnancies in Taiwan between 2001 and 2012 from a national insurance database. Gestational age, birth weight, Apgar score, and fetal outcomes were obtained from Taiwan’s national birth registry to which obstetricians are required by law to report fetal and maternal information. Mothers who were 44 years of age or who had had congenital heart disease, a gap between deliveries that was 20 years, or multiparities were excluded. Informed consent was waived because of anonymous data. Symptomatic PSVT was defined as …

Background Pregnancy involves adaptation to the immune system to prevent rejection of the foetus and this might have consequences on the activity of the rheumatoid arthritis (RA). Equally the aberrant immunity and disease activity of the RA might influence the pregnancy and foetal-neonatal outcomes. Objectives This study was designed to estimate the risk of adverse foetal-neonatal and maternal outcomes in pregnancies in women with RA. Methods We identified 2,350,339 singleton pregnancies using the Taiwan National Health Insurance database and birth registry between 2001 and 2012. Maternal history of RA, SLE, Sjogren’s syndrome, systemic sclerosis, vasculitis and poly/dermatomyositis were ascertained; among them, 845 individuals had RA. Odds ratios (ORs) and 95% confidence intervals (CIs) for pregnancy outcomes were estimated using an adjusted generalised estimating equation model. Results Pregnancies in women with RA were associated with an OR (95% CI) of 1.65 (1.37–1.98) for low birthweight ( Conclusions Pregnancies in women with RA were at a higher risk for multiple adverse foetal-neonatal outcomes, especially low birthweight ( References [1] Michet CJ Jr, McKenna CH, Elveback LR, Kaslow RA, Kurland LT. Epidemiology of systemic lupus erythematosus and other connective tissue diseases in Rochester, Minnesota, 1950 through 1979. Mayo Clin Proc1985;60(2):105–113. [2] Chiu YM, Lai CH. Nationwide population-based epidemiologic study of systemic lupus erythematosus in Taiwan. Lupus2010;19(10):1250–1255. [3] Rees F, Doherty M, Grainge M, Davenport G, Lanyon P, Zhang W. The incidence and prevalence of systemic lupus erythematosus in the UK, 1999–2012. Ann Rheum Dis2016;75(1):136–141. [4] See LC, Kuo CF, Chou IJ, Chiou MJ, Yu KH. Sex- and age-specific incidence of autoimmune rheumatic diseases in the Chinese population: A Taiwan population-based study. Semin Arthritis Rheum2013;43(3):381–386. [5] Cutolo M, Matucci-Cerinic M, Lockshin M, Ostensen M. Introduction: New trends in pregnancy and rheumatic diseases. Rheumatology (Oxford)2008;47(Suppl 3):iii1. Acknowledgements The authors would like to thank Centre for Big Data Analytics and Statistics in Chang Gung Memorial Hospital for statistical consultation. The sponsors of the study had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; and preparation, review, or approval of the manuscript. Disclosure of Interest None declared

Background Febuxostat is a new xanthine oxidase inhibitor which is effective in the treatment of hyperuricemia. Its use in people with different levels of renal function has not been thoroughly assessed. Objectives To assess the risk of renal function deterioration in gout patients using febuxostat and compare it with patients using allopurinol and other uricosuric agents. Methods We included 5882 adult patients with gout diagnosed between 2012 and 2015. We divided patients into three groups by usage of urate-lowering drugs used (allopurinol, febuxostat and uricosuric agents including benzbromarone, probenecid and sulfinpyrazone). The primary outcome was a progress of one chronic kidney disease (CKD) stage. The index date was the first date of urate-lowering drugs used. We follow each patient to the date of event, death, switch to other drug or the end of the study. The Cox proportional hazards model was used to estimate the adjusted hazard ratios (HRs), adjusted for age, sex, baseline eGFR, baseline serum urine acid, comorbidity, co-medication. Results The incident rate of the renal function deterioration was 2.37 (95% CI, 1.75–3.14) per 1000 person-days in allopurinol users, 1.54 (95% CI, 1.30–1.81) per 1000 person-days in febuxostat users and 1.43 (95% CI, 1.32–1.56) per 1000 person-days in uricosuric agent users. The HRs of progession of CKD stage in febuxostat and uricosuric agent users were 0.98 (95% CI, 0.70–1.38) and 0.80 (95% CI, 0.59–1.09) compared with allopurinol. Among febuxostat users, the HRs were 1.80 (95% CI, 0.85–3.98) in patients with an estimated glomerular filtration rate (eGFR) ≥90 ml/min/1.73 m2), 1.06 (95% CI, 0.56–2.03) for patients with an eGFR 60–89 ml/min/1.73 m2, 0.78 (95% CI, 0.45–1.42) for patients with an eGFR 30–59 ml/min/1.73 m2 and 0.73 (95% CI, 0.22–3.52) for patients with an eGFR 15–29 ml/min/1.73 m2. Among uricosutic agent users, the HRs were 0.94 (95% CI, 0.52–1.88) in patients with an eGFR ≥90 ml/min/1.73 m2, 0.88 (95% CI, 0.55–1.50) in patients with an eGFR 60–89 ml/min/1.73 m2, 0.71 (95% CI, 0.41–1.28) in patients with an eGFR 30–59 ml/min/1.73 m2 and 0.71 (95% CI, 0.19–3.62) in patients with an eGFR 15–29 ml/min/1.73 m2. Conclusions Renal function deterioration is not uncommon after initiation of urate-lowering treatment and febuxostat has a similar renal safety profile as allopurinols and uricosuric agents. Disclosure of Interest None declared

Background Preventable hospitalizations for specific conditions are considered preventable because appropriate outpatient care should potentially avoid hospitalisation for these conditions. Although not all of the hospitalizations can be prevented, to identify the rate of preventable hospitalizations among patients with systemic lupus erythematosus (SLE) can help us to understand the excess potential preventable hospitalizations and to provide an action to the SLE patients. Objectives To investigate the risk of preventable hospitalisation in patients with SLE before and after initial diagnosis of the disease. Methods We identified 4483 adult patients with incident SLE between 2005 and 2009 using the Taiwan National Health Insurance Database. Each SLE patients were matched to five controls without SLE during the same study period, by age and sex. The index date was the first date of SLE diagnosis and their matched controls. We estimated the incidence and incidence rate ratios (IRRs) of preventable hospitalisation by conditional Poisson regression, adjusted for age, sex, Elixhauser Comorbidity Index, number of outpatient visits and hospitalizations 1-y prior to index date, residence urbanisation, income levels, occupation and the number of physicians at the patients’ residence. Results The overall incidence of preventable hospitalisation was 1.88 (95% CI, 1.74 to 2.03) per 1000 person-months among SLE patients and 0.53 (95% CI, 0.49 to 0.56) per 1000 person-months among controls, giving an adjusted IRR of 3.57 (95% CI, 3.25 to 3.93). The IRRs especially higher in heart failure, bacterial pneumonia and urinary tract infection and the respective estimates were were 2.85 (95% CI, 2.14 to 3.80),4.67 (95% CI, 4.03 to 5.42) and 3.84 (95% CI,3.35 to 4.41). Conclusions Risk of preventable hospitalisation is higher in patients with SLE. Disclosure of Interest None declared

Objective Primary malignant brain tumors are relatively uncommon, and their incidence and survival rates have seldom been reported. Patients and methods We identified all patients with malignant brain tumors in Taiwan between 1997 and 2012 using the National Health Insurance database. We estimated the stratified incidence of malignant brain tumors by age and sex. We estimated the median 1-, 2-, and 5-year survival, taking comorbidities into account. Trends for incidence and survival were analyzed using Joinpoint regression. The incidence in different geographic areas was also evaluated. Results A total of 7746 men and 5846 women were identified. The incidence of malignant brain tumor was 3.34 (95% CI, 3.09–3.59) per 100,000 person-years in 1997 and 3.82 (95% CI, 3.56–4.08) per 100,000 person-years in 2012. The average annual percentage change (APC) of the standardized incidence over this period was 0.1 (95% CI, −1.9 to 2.2), suggesting a relatively stable incidence. However, the incidence significantly decreased between 1999 and 2012, with an APC of −1.8 [95% CI, −2.5 to −1.0]. One- and 5-year survival was 53.8% (50.0%–57.5%) and 27.5% (24.1%–30.9%) in 1997 and 67.6% (64.3%–70.7%) and 32.8% (29.6%–35.9%) in 2012. The average APC was 1.1 (95% CI, 0.7–1.5) for 1-year survival and 0.2 (95% CI, −1.0–1.4) for 5-year survival. The trend of improvement in the survival rate was seen for short-term but not long-term survival, especially in the group with more comorbidities. Conclusions A slightly decreased trend in incidence of primary malignant brain tumors was observed in Taiwanese general population since 1999. Over the past 15 years, the short-term survival of malignant brain tumors has improved, especially in adults.

// Jun-Te Hsu 1 , Puo-Hsien Le 2 , Chang-Fu Kuo 3 , Meng-Jiun Chiou 4 , Chia-Jung Kuo 2 , Tsung-Hsing Chen 2 , Chun-Jung Lin 2 , Jen-Shi Chen 5 , Huang-Pin Yu 6 , Chun-Nan Yeh 1 , Yi-Yin Jan 1 and Ta-Sen Yeh 1 1 Department of Surgery, Chang Gung Memorial Hospital at Linkou, Chang Gung University College of Medicine, Taoyuan, Taiwan 2 Department of Gastroenterology, Chang Gung Memorial Hospital at Linkou, Chang Gung University College of Medicine, Taoyuan, Taiwan 3 Division of Rheumatology, Allergy, and Immunology, Chang Gung Memorial Hospital at Linkou, Chang Gung University College of Medicine, Taoyuan, Taiwan 4 Office for Big Data Research, Chang Gung Memorial Hospital at Linkou, Taoyuan, Taiwan 5 Department of Hemato-Oncology, Chang Gung Memorial Hospital at Linkou, Chang Gung University College of Medicine, Taoyuan, Taiwan 6 Department of Anesthesiology, Chang Gung Memorial Hospital at Linkou, Chang Gung University College of Medicine, Taoyuan, Taiwan Correspondence to: Jun-Te Hsu, email: hsujt2813@adm.cgmh.org.tw Chun-Nan Yeh, email: yehchunnan@gmail.com Keywords: gastrointestinal stromal tumor, imatinib, sunitinib, dose escalation, survival Received: January 12, 2017 Accepted: March 22, 2017 Published: April 03, 2017 ABSTRACT Background: Although treatment with imatinib in advanced gastrointestinal stromal tumor (GIST) patients has led to significant clinical benefits, the disease will eventually progress due to imatinib resistance. Treatment options after failure of first-line imatinib include imatinib dose escalation or shifting to sunitinib. However, there is no large-scale study to compare the efficacy difference between these two treatment strategies or the role of surgery. Results: This study recruited 521 advanced GIST patients including 246, 125, and 150 placed in groups 1, 2, and 3, respectively. Groups 1 and 2 had significantly longer overall survival (OS) as compared with the group 3 (median 37.5 months versus 16.0 months; p < 0.0001). After adjusting for confounding variables, groups 1 and 2 had longer OS than group 3. A favorable survival trend was seen with surgery, although this benefit disappeared after adjusting for confounding factors. Materials and Methods: We conducted a nationwide population-based cohort study using data from the Taiwan National Health Insurance Research Database from July 2004 to December 2010. Advanced GIST patients who no longer responded to first-line imatinib were stratified into three groups: imatinib dose escalation (group 1); imatinib dose escalation and a shift to sunitinib (group 2); a direct shift to sunitinib (group 3). The therapeutic success of the three treatment regimens and the effect of surgery were evaluated by overall survival. Conclusions: For advanced GIST patients who failed first-line imatinib treatment, imatinib dose escalation confers significantly longer OS compared to a direct switch to sunitinib. Surgery does not provide survival benefits.

// Hsin-I Tsai 1,2,3 , Chao-Wei Lee 2,3,4 , Chang-Fu Kuo 2,5 , Lai-Chu See 5,6,7 , Fu-Chao Liu 1,2 , Meng-Jiun Chiou 8 and Huang-Ping Yu 1,2 1 Department of Anesthesiology, Chang Gung Memorial Hospital, Taoyuan, Taiwan 2 College of Medicine, Chang Gung University, Taoyuan, Taiwan 3 Graduate Institute of Clinical Medical Sciences, Chang Gung University, Taoyuan, Taiwan 4 Department of Surgery, Change Gung Memorial Hospital, Taoyuan, Taiwan 5 Division of Rheumatology, Allergy and Immunology, Chang Gung Memorial Hospital, Taoyuan, Taiwan 6 Department of Public Health, College of Medicine, Chang Gung University, Taoyuan, Taiwan 7 Biostatistics Core Laboratory, Molecular Medicine Research Center, Chang Gung University, Taoyuan, Taiwan 8 Office for Big Data Research, Chang Gung Memorial Hospital, Taoyuan, Taiwan Correspondence to: Huang-Ping Yu, email: // Keywords : de novo malignancy, organ transplantation, nationwide study, population-based study, cohort study Received : September 27, 2016 Accepted : October 28, 2016 Published : November 04, 2016 Abstract Organ transplant recipients appear to have a higher risk of de novo malignancy. The aim of the study was designed to estimate cancer risk in heart, lung, kidney and liver transplant recipients. The cohort study used the Taiwan National Health Insurance Research Database (1996-2011) and followed the outcomes of organ recipients until 2012. De novo cancer and mortality rates after organ transplantation were evaluated using standardized incidence ratios, excess absolute risks of cancer, and standardized mortality ratios in recipients were compared with those in the general population. We identified 40, 231, 2, and 115 patients who developed cancer after heart, kidney, lung, and liver transplantation, which corresponded to a cancer incidence of 878.4, 1101.2, 728.9, and 1361.4 cases per 100,000 person-years, respectively. In heart, kidney, lung, and liver recipients, the overall standardized incidence ratios were 1.65 (1.21-2.24), 3.33 (2.93-3.79), 1.82 (0.45-7.27) and 3.37 (2.81-4.05) and the overall standardized mortality ratios were 5.45 (4.96-5.98), 1.47 (1.34-1.61), 8.92 (7.10-11.20), and 3.83 (3.48-4.20), respectively. These results reveal a three-fold increase in de novo cancer risk in organ transplant patients compared with the general population. This study illustrated the importance of de novo malignancy after organ transplantation.

Background: Information about known risk factors for congenital heart disease is scarce. In this population-based study, we aimed to investigate the relation between maternal chronic disease and congenital heart disease in offspring. Methods: The study cohort consisted of 1 387 650 live births from 2004 to 2010. We identified chronic disease in mothers and mild and severe forms of congenital heart disease in their offspring from Taiwan’s National Health Insurance medical claims. We used multivariable logistic regression analysis to assess the associations of all cases and specific types of congenital heart disease with various maternal chronic diseases. Results: For mothers with the following chronic diseases, the overall prevalence of congenital heart disease in their children was significantly higher than for mothers without these diseases: diabetes mellitus type 1 (adjusted odds ratio [OR] 2.32, 95% confidence interval [CI] 1.66–3.25), diabetes mellitus type 2 (adjusted OR 2.85, 95% CI 2.60–3.12), hypertension (adjusted OR 1.87, 95% CI 1.69–2.07), congenital heart defects (adjusted OR 3.05, 95% CI 2.45–3.80), anemia (adjusted OR 1.31, 95% CI 1.25–1.38), connective tissue disorders (adjusted OR 1.39, 95% CI 1.19–1.62), epilepsy (adjusted OR 1.37, 95% CI 1.08–1.74) and mood disorders (adjusted OR 1.25, 95% CI 1.11–1.41). The same pattern held for mild forms of congenital heart disease. A higher prevalence of severe congenital heart disease was seen only among offspring of mothers with congenital heart defects or type 2 diabetes. Interpretation: The children of women with several kinds of chronic disease appear to be at risk for congenital heart disease. Preconception counselling and optimum treatment of pregnant women with chronic disease would seem prudent.

Purpose To estimate the rate of hospitalized eye trauma in Taiwan and investigate the role between principal and secondary diagnoses of such trauma. Methods Nationwide fixed cohort study of 1,000,000 beneficiaries from the Taiwan Longitudinal Health Database was used and 4819 patients who were hospitalized for eye trauma during 1996-2010 were analyzed. Results During 1996-2010, the incidence rates of hospitalized eye trauma (per 100 000 person-years) were 35.0 (95% confidence interval (CI), 34.0 to 36.0) for total diagnosis, 9.8 (95% CI, 9.3 to 10.3) for a principal diagnosis, and 25.3 (95% CI, 24.4 to 26.1) for a secondary diagnosis. The sex risk ratio was 3.1 for a principal diagnosis and 2.1 for a secondary diagnosis. The main causes of eye trauma were traffic accident, work accident, assault (among males

Objectives The aim of this study was to estimate the risk of hyperuricaemia and gout in people with hypothyroid or hyperthyroid status. Methods This study analyzed data from individuals who participated in health screening programs at Chang Gung Memorial Hospital in northern Taiwan (2000–2010). Participants were categorized as having euthyroid, hypothyroid, or hyperthyroid status according to their thyroid-stimulating hormone (TSH) levels. Multinomial logistic regression models were used to calculate the odds ratios (95% CI) for hyperuricaemia and gout in participants with thyroid dysfunction compared to euthyroid participants. Results A total of 87,813 (euthyroid, 83,502; hypothyroid, 1,460; hyperthyroid, 2,851) participants were included. The prevalence of hyperuricaemia was higher in hyperthyroid subjects (19.4%) than in euthyroid subjects (17.8%) but not in hypothyroid subjects (19.3%). The prevalence of gout was significantly higher in both hypothyroid (6.0%) and hyperthyroid (5.3%) subjects than in euthyroid subjects (4.3%). In men, hypothyroid or hyperthyroid status was not associated with hyperuricaemia. However, hypothyroid or hyperthyroid status was associated with ORs (95% CI) of 1.47 (1.10–1.97) and 1.37 (1.10–1.69), respectively, for gout. In women, hypothyroid status was not associated with hyperuricaemia or gout. However, hyperthyroid status was associated with ORs (95% CI) of 1.42 (1.24–1.62) for hyperuricaemia and 2.13 (1.58–2.87) for gout. Conclusions Both hyperthyroid and hypothyroid status were significantly associated with gout and weakly associated with hyperuricaemia. A thyroid function test for gout patients may by warranted.