Development of a Machine Learning Model to Predict the Use of Surgery in Patients With Rheumatoid Arthritis

OBJECTIVE

One in five patients with rheumatoid arthritis (RA) rely on surgery to restore joint function. However, variable response to disease-modifying antirheumatic drugs (DMARDs) complicates surgical planning, and it is difficult to predict which patients may ultimately require surgery. We used machine learning to develop predictive models for the likelihood of undergoing an operation related to RA and which type of operation patients who require surgery undergo.

METHODS

We used electronic health record data to train two extreme gradient boosting machine learning models. The first model predicted patients' probabilities of undergoing surgery ≥5 years after their initial clinic visit. The second model predicted whether patients who underwent surgery would undergo a major joint replacement versus a less intensive procedure. Predictors included demographics, comorbidities, and medication data. The primary outcome was model discrimination, measured by area under the receiver operating characteristic curve (AUC).

RESULTS

We identified 5,481 patients, of whom 278 (5.1%) underwent surgery. There was no significant difference in the frequency of DMARD or steroid prescriptions between patients who did and did not have surgery, though nonsteroidal anti-inflammatory drug prescriptions were more common among patients who did have surgery (P = 0.03). The model predicting use of surgery had an AUC of 0.90 ± 0.02. The model predicting type of surgery had an AUC of 0.58 ± 0.10.

CONCLUSIONS

Predictive models using clinical data have the potential to facilitate identification of patients who may undergo rheumatoid-related surgery, but not what type of procedure they will need. Integrating similar models into practice has the potential to improve surgical planning.

Predicting osteoporosis from kidney-ureter-bladder radiographs utilizing deep convolutional neural networks

Osteoporosis is a common condition that can lead to fractures, mobility issues, and death. Although dual-energy X-ray absorptiometry (DXA) is the gold standard for osteoporosis, it is expensive and not widely available. In contrast, kidney-ureter-bladder (KUB) radiographs are inexpensive and frequently ordered in clinical practice. Thus, it is a potential screening tool for osteoporosis. In this study, we explored the possibility of predicting the bone mineral density (BMD) and classifying high-risk patient groups using KUB radiographs. We proposed DeepDXA-KUB, a deep learning model that predicts the BMD values of the left hip and lumbar vertebrae from an input KUB image. The datasets were obtained from Taiwanese medical centers between 2006 and 2019, using 8913 pairs of KUB radiographs and DXA examinations performed within 6 months. The images were randomly divided into training and validation sets in a 4:1 ratio. To evaluate the model's performance, we computed a confusion matrix and evaluated the sensitivity, specificity, accuracy, precision, positive predictive value, negative predictive value, F1 score, and area under the receiver operating curve (AUROC). Moderate correlations were observed between the predicted and DXA-measured BMD values, with a correlation coefficient of 0.858 for the lumbar vertebrae and 0.87 for the left hip. The model demonstrated an osteoporosis detection accuracy, sensitivity, and specificity of 84.7 %, 81.6 %, and 86.6 % for the lumbar vertebrae and 84.2 %, 91.2 %, and 81 % for the left hip, respectively. The AUROC was 0.939 for the lumbar vertebrae and 0.947 for the left hip, indicating a satisfactory performance in osteoporosis screening. The present study is the first to develop a deep learning model based on KUB radiographs to predict lumbar spine and femoral BMD. Our model demonstrated a promising correlation between the predicted and DXA-measured BMD in both the lumbar vertebrae and hip, showing great potential for the opportunistic screening of osteoporosis.

Comparing Artificial Intelligence-Enabled Electrocardiogram Models in Identifying Left Atrium Enlargement and Long-term Cardiovascular Risk

BACKGROUND

The role of P-wave in identifying left atrial enlargement (LAE) with the use of artificial intelligence (AI)-enabled electrocardiography (ECG) models is unclear. It is also unknown if AI-enabled single-lead ECG could be used as a diagnostic tool for LAE surveillance. We aimed to build AI-enabled P-wave and single-lead ECG models to identify LAE using sinus rhythm (SR) and non-SR ECGs, and compare the prognostic ability of severe LAE, defined as left atrial diameter ≥ 50 mm, assessed by AI-enabled ECG models vs echocardiography.

METHODS

This retrospective study used data from 382,594 consecutive adults with paired 12-lead ECG and echocardiography performed within 2 weeks of each other at Chang Gung Memorial Hospital. UNet++ was used for P-wave segmentation. ResNet-18 was used to develop deep convolutional neural network-enabled ECG models for discriminating LAE. External validation was performed with the use of data from 11,753 patients from another hospital.

RESULTS

The AI-enabled 12-lead ECG model outperformed other ECG models for classifying LAE, but the single-lead ECG models also showed excellent performance at a left atrial diameter cutoff of 50 mm. AI-enabled ECG models had excellent and fair discrimination on LAE using the SR and the non-SR data set, respectively. Severe LAE identified by AI-enabled ECG models was more predictive of future cardiovascular disease than echocardiography; however, the cumulative incidence of new-onset atrial fibrillation and heart failure was higher in patients with echocardiography-severe LAE than with AI-enabled ECG-severe LAE.

CONCLUSIONS

P-Wave plays a crucial role in discriminating LAE in AI-enabled ECG models. AI-enabled ECG models outperform echocardiography in predicting new-onset cardiovascular diseases associated with severe LAE.

A colonial serrated polyp classification model using white-light ordinary endoscopy images with an artificial intelligence model and TensorFlow chart

In this study, we implemented a combination of data augmentation and artificial intelligence (AI) model-Convolutional Neural Network (CNN)-to help physicians classify colonic polyps into traditional adenoma (TA), sessile serrated adenoma (SSA), and hyperplastic polyp (HP). We collected ordinary endoscopy images under both white and NBI lights. Under white light, we collected 257 images of HP, 423 images of SSA, and 60 images of TA. Under NBI light, were collected 238 images of HP, 284 images of SSA, and 71 images of TA. We implemented the CNN-based artificial intelligence model, Inception V4, to build a classification model for the types of colon polyps. Our final AI classification model with data augmentation process is constructed only with white light images. Our classification prediction accuracy of colon polyp type is 94%, and the discriminability of the model (area under the curve) was 98%. Thus, we can conclude that our model can help physicians distinguish between TA, SSA, and HPs and correctly identify precancerous lesions such as TA and SSA.

Genetic and Environmental Contributions of Primary Angle-Closure Glaucoma and Primary Open-Angle Glaucoma: A Nationwide Study in Taiwan

PURPOSE

To estimate the familial risks of primary angle-closure glaucoma (PACG) and primary open-angle glaucoma (POAG) and assess the relative contributions of environmental and genetic factors to these risks.

DESIGN

Retrospective, population-based cohort study.

METHODS

We used the 2000-2017 Taiwan National Health Insurance Program database to construct 4,144,508 families for the 2017 population (N = 23,373,209). We used the polygenic liability model to estimate glaucoma's heritability and familial transmission. The degree of familial aggregation of glaucoma was obtained from the adjusted relative risk for individuals whose first-degree relatives had glaucoma using Cox's model.

RESULTS

PACG and POAG prevalence rates for individuals whose first-degree relatives had PACG or POAG were 0.95% and 2.40%, higher than those of the general population (0.61% and 0.40%, respectively). The relative risk of PACG in individuals whose first-degree relatives had PACG was 2.44 (95% CI = 2.31-2.58). The relative risk of POAG in individuals whose first-degree relatives had POAG was 6.66 (95% CI = 6.38-6.94). The estimated contributions to PACG and POAG phenotypic variances were 19.4% and 59.6% for additive genetic variance, 19.1% and 23.2% for common environmental factors shared by family members, and 61.5% and 17.2% for nonshared environmental factors, respectively.

CONCLUSIONS

These data highlight the relative importance of genetic contribution to POAG and environmental contribution to PACG. Therefore, future work may need to focus on finding more novel environmental determinants of PACG.

Machine learning-based prediction of acute mortality in emergency department patients using twelve-lead electrocardiogram

Background

The risk of mortality is relatively high among patients who visit the emergency department (ED), and stratifying patients at high risk can help improve medical care. This study aimed to create a machine-learning model that utilizes the standard 12-lead ECG to forecast acute mortality risk in ED patients.

Methods

The database included patients who visited the EDs and underwent standard 12-lead ECG between October 2007 and December 2017. A convolutional neural network (CNN) ECG model was developed to classify survival and mortality using 12-lead ECG tracings acquired from 345,593 ED patients. For machine learning model development, the patients were randomly divided into training, validation and testing datasets. The performance of the mortality risk prediction in this model was evaluated for various causes of death.

Results

Patients who visited the ED and underwent one or more ECG examinations experienced a high incidence of 30-day mortality [18,734 (5.42%)]. The developed CNN model demonstrated high accuracy in predicting acute mortality (hazard ratio 8.50, 95% confidence interval 8.20-8.80) with areas under the receiver operating characteristic (ROC) curve of 0.84 for the 30-day mortality risk prediction models. This CNN model also demonstrated good performance in predicting one-year mortality (hazard ratio 3.34, 95% confidence interval 3.30-3.39). This model exhibited good predictive performance for 30-day mortality not only for cardiovascular diseases but also across various diseases.

Conclusions

The machine learning-based ECG model utilizing CNN screens the risks for 30-day mortality. This model can complement traditional early warning scoring indexes as a useful screening tool for mortality prediction.

Can a Deep Learning Algorithm Improve Detection of Occult Scaphoid Fractures in Plain Radiographs? A Clinical Validation Study

BACKGROUND

Occult scaphoid fractures on initial radiographs of an injury are a diagnostic challenge to physicians. Although artificial intelligence models based on the principles of deep convolutional neural networks (CNN) offer a potential method of detection, it is unknown how such models perform in the clinical setting.

QUESTIONS/PURPOSES

(1) Does CNN-assisted image interpretation improve interobserver agreement for scaphoid fractures? (2) What is the sensitivity and specificity of image interpretation performed with and without CNN assistance (as stratified by type: normal scaphoid, occult fracture, and apparent fracture)? (3) Does CNN assistance improve time to diagnosis and physician confidence level?

METHODS

This survey-based experiment presented 15 scaphoid radiographs (five normal, five apparent fractures, and five occult fractures) with and without CNN assistance to physicians in a variety of practice settings across the United States and Taiwan. Occult fractures were identified by follow-up CT scans or MRI. Participants met the following criteria: Postgraduate Year 3 or above resident physician in plastic surgery, orthopaedic surgery, or emergency medicine; hand fellows; and attending physicians. Among the 176 invited participants, 120 completed the survey and met the inclusion criteria. Of the participants, 31% (37 of 120) were fellowship-trained hand surgeons, 43% (52 of 120) were plastic surgeons, and 69% (83 of 120) were attending physicians. Most participants (73% [88 of 120]) worked in academic centers, whereas the remainder worked in large, urban private practice hospitals. Recruitment occurred between February 2022 and March 2022. Radiographs with CNN assistance were accompanied by predictions of fracture presence and gradient-weighted class activation mapping of the predicted fracture site. Sensitivity and specificity of the CNN-assisted physician diagnoses were calculated to assess diagnostic performance. We calculated interobserver agreement with the Gwet agreement coefficient (AC1). Physician diagnostic confidence was estimated using a self-assessment Likert scale, and the time to arrive at a diagnosis for each case was measured.

RESULTS

Interobserver agreement among physicians for occult scaphoid radiographs was higher with CNN assistance than without (AC1 0.42 [95% CI 0.17 to 0.68] versus 0.06 [95% CI 0.00 to 0.17], respectively). No clinically relevant differences were observed in time to arrive at a diagnosis (18 ± 12 seconds versus 30 ± 27 seconds, mean difference 12 seconds [95% CI 6 to 17]; p < 0.001) or diagnostic confidence levels (7.2 ± 1.7 seconds versus 6.2 ± 1.6 seconds; mean difference 1 second [95% CI 0.5 to 1.3]; p < 0.001) for occult fractures.

CONCLUSION

CNN assistance improves physician diagnostic sensitivity and specificity as well as interobserver agreement for the diagnosis of occult scaphoid fractures. The differences observed in diagnostic speed and confidence is likely not clinically relevant. Despite these improvements in clinical diagnoses of scaphoid fractures with the CNN, it is unknown whether development and implementation of such models is cost effective.

LEVEL OF EVIDENCE

Level II, diagnostic study.

Kinship analysis of type 2 diabetes mellitus familial aggregation in Taiwan

BACKGROUND

Family disease history plays a vital role in type 2 diabetes mellitus (T2DM) risk. However, the familial aggregation of T2DM among different kinship relatives warrants further investigation.

METHODS

This nationwide kinship relationship study collected 2000-2016 data of two to five generations of the Taiwanese population from the National Health Insurance Research Database. Approximately 4 million family trees were constructed from the records of 20, 890, 264 Taiwanese residents during the study period. T2DM was diagnosed on the basis of ICD-9-CM codes 250.x0 or 250.x2, with three consecutive related prescriptions. The Cox proportional hazard model was used for statistical analysis.

RESULTS

Compared with their counterparts, individuals who had first-degree relatives with T2DM were more likely to develop T2DM during the follow-up period (hazard ratio [HR], 2.37-27.75), followed by individuals who had second-degree relatives with T2DM (HR, 1.29-1.88). T2DM relative risk was higher in those with an affected mother than in those with affected father. The HR for T2DM was 20.32 (95%CI = 15.64-26.42) among male individuals with an affected twin brother, whereas among female individuals with an affected twin sister, it was 60.07 (95%CI = 40.83-88.36). The HRs presented a dose-response relationship with the number of affected family members.

CONCLUSION

The study suggests a significant familial aggregation of T2DM occurrence; these findings could aid in identifying the high-risk group for T2DM and designing early intervention strategies and treatment plans.

Cancer survivorship and risk of pregnancy complications, adverse obstetric outcomes, and maternal morbidities in female adolescents and young adults: a nationwide population-based study from Taiwan

BACKGROUND

Cancer treatment in female adolescent and young adult (AYA) cancer survivors (i.e., those diagnosed between 15 and 39 years of age) may adversely affect multiple bodily functions, including the reproductive system.

METHODS

We initially assembled a retrospective, nationwide population-based cohort study by linking data from two nationwide Taiwanese data sets. We subsequently identified first pregnancies and singleton births to AYA cancer survivors (2004-2018) and select AYA without a previous cancer diagnosis matched to AYA cancer survivors for maternal age and infant birth year.

RESULTS

The study cohort consisted of 5151 and 51,503 births to AYA cancer survivors and matched AYA without a previous cancer diagnosis, respectively. The odds for overall pregnancy complications (odds ratio [OR], 1.09; 95% confidence interval [CI], 1.01-1.18) and overall adverse obstetric outcomes (OR, 1.07; 95% CI, 1.01-1.13) were significantly increased in survivors compared with matched AYA without a previous cancer diagnosis. Specifically, cancer survivorship was associated with an increased risk of preterm labour, labour induction, and threatened abortion or threatened labour requiring hospitalisation.

CONCLUSIONS

AYA cancer survivors are at increased risk for pregnancy complications and adverse obstetric outcomes. Efforts to integrate individualised care into clinical guidelines for preconception and prenatal care should be thoroughly explored.

Automatic and human level Graf's type identification for detecting developmental dysplasia of the hip

BACKGROUND

Developmental dysplasia of the hip (DDH) is a common congenital disorder that may lead to hip dislocation and requires surgical intervention if left untreated. Ultrasonography is the preferred method for DDH screening; however, the lack of experienced operators impedes its application in universal neonatal screening.

METHODS

We developed a deep neural network tool to automatically register the five keypoints that mark important anatomical structures of the hip and provide a reference for measuring alpha and beta angles following Graf's guidelines, which is an ultrasound classification system for DDH in infants. Two-dimensional (2D) ultrasonography images were obtained from 986 neonates aged 0-6 months. A total of 2406 images from 921 patients were labeled with ground truth keypoints by senior orthopedists.

RESULTS

Our model demonstrated precise keypoint localization. The mean absolute error was approximately 1 mm, and the derived alpha angle measurement had a correlation coefficient of R = 0.89 between the model and ground truth. The model achieved an area under the receiver operating characteristic curve of 0.937 and 0.974 for classifying alpha <60° (abnormal hip) and <50° (dysplastic hip), respectively. On average, the experts agreed with 96% of the inferenced images, and the model could generalize its prediction on newly collected images with a correlation coefficient higher than 0.85.

CONCLUSIONS

Precise localization and highly correlated performance metrics suggest that the model can be an efficient tool for assisting DDH diagnosis in clinical settings.

Cardiovascular outcomes in patients with atrial fibrillation concomitantly treated with antiarrhythmic drugs and non-vitamin k antagonist oral anticoagulants

AIMS

Limited data compared antiarrhythmic drugs (AADs) with concomitant non-vitamin K antagonist oral anticoagulants in atrial fibrillation patients, hence the aim of the study.

METHODS AND RESULTS

National health insurance database were retrieved during 2012-17 for study. We excluded patients not taking AADs, bradycardia, heart block, heart failure admission, mitral stenosis, prosthetic valve, incomplete demographic data, and follow-up <3 months. Outcomes were compared in Protocol 1, dronedarone vs. non-dronedarone; Protocol 2, dronedarone vs. amiodarone; and Protocol 3, dronedarone vs. propafenone. Outcomes were acute myocardial infarction (AMI), ischaemic stroke/systemic embolism, intracranial haemorrhage (ICH), major bleeding, cardiovascular death, all-cause mortality, and major adverse cardiovascular event (MACE) (including AMI, ischaemic stroke, and cardiovascular death). In Protocol 1, 2298 dronedarone users and 6984 non-dronedarone users (amiodarone = 4844; propafenone = 1914; flecainide = 75; sotalol = 61) were analysed. Dronedarone was associated with lower ICH (HR = 0.61, 95% CI = 0.38-0.99, P = 0.0436), cardiovascular death (HR = 0.24, 95% CI = 0.16-0.37, P < 0.0001), all-cause mortality (HR = 0.33, 95% CI = 0.27-0.42, P < 0.0001), and MACE (HR = 0.56, 95% CI = 0.45-0.70, P < 0.0001). In Protocol 2, 2231 dronedarone users and 6693 amiodarone users were analysed. Dronedarone was associated with significantly lower ICH (HR = 0.53, 95%=CI 0.33-0.84, P = 0.0078), cardiovascular death (HR = 0.20, 95% CI = 0.13-0.31, P < 0.0001), all-cause mortality (HR 0.27, 95% CI 0.22-0.34, P < 0.0001), and MACE (HR = 0.53, 95% CI = 0.43-0.66, P < 0.0001), compared with amiodarone. In Protocol 3, 812 dronedarone users and 2436 propafenone users were analysed. There were no differences between two drugs for primary and secondary outcomes.

CONCLUSION

The use of dronedarone with NOACs was associated with cardiovascular benefits in an Asian population, compared with non-dronedarone AADs and amiodarone.

Sensitive label-free detection of the biomarker phosphorylated tau-217 protein in Alzheimer's disease using a graphene-based solution-gated field effect transistor

Alzheimer's disease (AD) is generally diagnosed using advanced imaging, but recent research suggests early screening using biomarkers in peripheral blood is feasible; among them, plasma tau proteins phosphorylated at threonine 231, threonine 181, and threonine 217 (p-tau217) are potential targets. A recent study indicates that the p-tau217 protein is the most efficacious biomarker. However, a clinical study found a pg/ml threshold for AD screening beyond standard detection methods. A biosensor with high sensitivity and specificity p-tau217 detection has not yet been reported. In this study, we developed a label-free solution-gated field effect transistor (SGFET)-based biosensor featuring a graphene oxide/graphene (GO/G) layered composite. The top layer of bilayer graphene grown using chemical vapor deposition was functionalized with oxidative groups serving as active sites for forming covalent bonds with the biorecognition element (antibodies); the bottom G could act as a transducer to respond to the attachment of the target analytes onto the top GO conjugated with the biorecognition element via π-π interactions between the GO and G layers. With this unique atomically layered G composite, we obtained a good linear electrical response in the Dirac point shift to p-tau217 protein concentrations in the range of 10 fg/ml to 100 pg/ml. The biosensor exhibited a high sensitivity of 18.6 mV/decade with a high linearity of 0.991 in phosphate-buffered saline (PBS); in human serum albumin, it showed approximately 90% of the sensitivity (16.7 mV/decade) in PBS, demonstrating high specificity. High stability of the biosensor was also displayed in this study.

Health-related quality of life improvement by adalimumab therapy in patients with rheumatoid arthritis in Taiwan: A nationwide prospective study

BACKGROUND

To determine the effects of adalimumab on health-related quality of life (HRQoL) in Taiwanese patients with moderate-to-severe rheumatoid arthritis (RA) (NCT02616380).

METHODS

During a 24-week observational period, 100 biologic-naive patients with RA received 40 mg adalimumab subcutaneously, every 2 weeks. The primary endpoint was a change in Health Assessment Questionnaire-Disability Index (HAQ-DI) score at 24 weeks. The secondary endpoints included change in HAQ-DI at week 12, number and percentage of patients achieving a meaningful improvement in HAQ-DI at weeks 12 and 24, and changes in the 36-Item Short Form Health Survey (SF-36), EuroQol 5-dimension 3-level version (EQ-5D-3L) index, and Work Productivity and Activity Impairment (WPAI) questionnaire scores at weeks 12 and 24.

RESULTS

At weeks 12 and 24, mean changes in HAQ-DI from baseline were -0.34 ± 0.46 and -0.44 ± 0.59 (both p < 0.001), and clinically meaningful improvement in HAQ-DI was achieved by 60.4% and 59.6% of patients, respectively. SF-36, EQ-5D-3L index, and WPAI scores significantly improved ( p < 0.001) from baseline to weeks 12 and 24. Exploratory analyses showed diabetes was significantly associated with changes in HAQ-DI, EQ-5D-3L, and WPAI scores whereas peptic ulcer correlated with changes in the SF-36 physical component summary T-score.

CONCLUSION

HRQoL improved after initiation of adalimumab therapy in Taiwanese patients with moderate-to-severe RA.

Reduced versus maximum tolerated methotrexate dose concomitant with adalimumab in patients with rheumatoid arthritis (MIRACLE): a randomised, open-label, non-inferiority trial

BACKGROUND

Efficacy of combination therapy with methotrexate and biological disease-modifying antirheumatic drugs is well established in the management of patients with rheumatoid arthritis; however, the optimal dose of methotrexate to administer with a tumour necrosis factor inhibitor remains unclear. We aimed to clarify the efficacy and safety of adalimumab combined with reduced methotrexate dose compared with the maximum tolerated methotrexate dose in patients with rheumatoid arthritis and an inadequate response to methotrexate monotherapy.

METHODS

In this open-label, randomised controlled trial, we recruited methotrexate-naive patients with rheumatoid arthritis and a disease duration of less than 2 years across 24 secondary or tertiary care hospitals across Japan, South Korea, and Taiwan. At initiation, methotrexate was given orally and increased to the maximum tolerated dose by week 12. Patients who did not achieve remission on the basis of the Simplified Disease Activity Index (SDAI) at week 24 were randomly assigned (1:1) to receive adalimumab (40 mg biweekly) combined with a continued maximum tolerated dose of methotrexate or adalimumab combined with a reduced dose of methotrexate. The primary endpoint was non-inferiority of adalimumab plus reduced-dose methotrexate to adalimumab plus maximal-dose methotrexate based on SDAI remission at week 48, assessed in the modified full-analysis set with a pre-specified non-inferiority margin of -15%, based on a two-sided 90% CI. Adverse events were assessed in the safety analysis set. This trial is registered with ClinicalTrials.gov, NCT03505008 and has been completed.

FINDINGS

From April 18, 2018, to June 2, 2020, from 323 patients screened, 300 were enrolled, and 291 patients were included in the full analysis set. The mean age was 57·7 years (SD 15·2), 217 (75%) were female, 74 (25%) were male, and all patients were of Asian ethnicity. The mean SDAI at study enrolment was 26·5 (SD 12·4). 52 patients discontinued the study before week 24 or at week 24 before randomisation. At week 24, 105 (36%) of 291 patients achieved remission and continued methotrexate monotherapy through week 48. 134 (46%) did not achieve remission at week 24 and were randomly assigned to receive adalimumab plus the maximum tolerated dose of methotrexate (n=68) or adalimumab plus reduced-dose methotrexate (n=66). Remission at week 48 was achieved in 25 (38%) of 66 and 27 (44%) of 61 patients, respectively, with an adjusted risk difference of 6·4% (90% CI -7·0 to 19·8), which met the non-inferiority margin of -15%. Adverse events after week 24 tended to be more frequent in the maximum tolerated dose group than in the reduced-dose group (24 [35%] vs 13 [20%], p=0·054). Between week 24 and 48, there were 14 serious adverse events (6 in the methotrexate monotherapy group, 5 in the adalimumab plus maximal-dose methotrexate, and 3 in the adalimumab plus reduced-dose methotrexate group), and no deaths.

INTERPRETATION

The MIRACLE study showed that the efficacy of adalimumab combined with reduced methotrexate dose was not inferior to that with the maximum tolerated methotrexate dose, with a tendency to a better safety profile.

FUNDING

Eisai.

Diagnostic Evaluation Using Salivary Gland Ultrasonography in Primary Sjögren's Syndrome

The purpose of this study is to investigate the clinical manifestations in patients with early primary Sjögren's syndrome (pSS) based on the severity score found by salivary gland ultrasonography. Thirty-five newly diagnosed patients with early pSS were enrolled and divided into mild (score 0-1) and severe (score 2-3) groups according to the salivary gland ultrasonography grade (SGUS) scores at baseline. Clinical evaluation, ESSPRI and ESSDAI index values, sicca symptoms of the mouth, salivary capacity, and serum autoantibodies and cytokines were investigated. The mean age of pSS patients at diagnosis was 49.9 ± 11.9 years, and the mean duration of sicca symptoms was 0.58 years. ESSPRI (EULAR Sjögren's syndrome patient report index) and ESSDAI (EULAR Sjögren's syndrome disease index) scores were 15.97 and 4.77, respectively. Clinical manifestations, including the low production of saliva and autoantibody production, such as antinuclear antibodies, rheumatoid factor, and anti-SSA antibody, were found. A higher prevalence of rheumatoid factor (p = 0.0365) and antinuclear antibody (p = 0.0063) and a higher elevation of total IgG (p = 0.0365) were found in the severe group than in the mild group. In addition, the elevated titer of IL-25 was detected in the severe group than in the mild group. This observation indicated that salivary gland ultrasonography grade (SGUS) scans may help physicians diagnose pSS and the elevated titer of IL-25 in patients may be implicated in the pathogenesis of pSS.

Effect of febuxostat on adverse events and mortality in gout in Taiwan: An interrupted time series analysis

OBJECTIVES

To evaluate the influence of febuxostat on adverse events and mortality in gout.

METHODS

We retrospectively enrolled patients with newly diagnosed gout and prescribed urate-lowering therapy between 2006 and 2017 from the Taiwan National Health Insurance Database. These patients were divided into 2 groups: with and without comorbidities (n = 294 847 and 194 539). An interrupted time series analysis with adjustments for demographics, comorbidities, and comedication by propensity score-based stabilized weights was used to compare the trend of adverse events and mortality before vs after febuxostat was introduced in 2012.

RESULTS

The proportion of febuxostat use gradually increased from 0% in 2012 to 30% in those with comorbidities and 10% in those without comorbidities in 2017. Allopurinol use decreased from 30% in 2012 to 10% in 2017. The slope of the 1-year incidence rate of Stevens-Johnson syndrome (SJS) or toxic epidermal necrolysis (TEN) (per 10 000 patients) significantly reduced after 2012 in those with and without comorbidities (-0.375 per quarter, P = .015 and -.253 per quarter, P = .049). The slope of the 3-year incidence rate of acute myocardial infarction (AMI) (per 1000 patients), percutaneous coronary intervention (PCI) (per 1000 patients), and all-cause mortality (per 100 patients) significantly increased after 2012 in those with comorbidities (+0.207 per quarter, P = .013; +.389 per quarter, P = .002; +.103 per quarter, P = .001).

CONCLUSIONS

Febuxostat may reduce SJS and TEN in all gout patients but increase AMI, PCI, and all-cause mortality in gout patients with comorbidities.

Association of Echocardiographic Parameter E/e' With Cardiovascular Events in a Diverse Population of Inpatients and Outpatients With and Without Cardiac Diseases and Risk Factors

BACKGROUND

The echocardiographic parameter E/e' has been associated with cardiovascular (CV) events. However, few studies have analyzed multiple associated CV outcomes using E/e' in a diverse population of both inpatients and outpatients with and without cardiac diseases and risk factors.

METHODS

Medical records of 75,393 patients without atrial fibrillation (AF) with first available E/e' were retrieved from our hospital database. Patients with mitral valve disease were excluded, and the remainder were studied in protocol 1 (70,819 patients). Patients with hypertension, diabetes mellitus, hyperlipidemia, CV diseases, prior CV events, CV surgeries, and left ventricular ejection fraction <50% or missing left ventricular ejection fraction were further excluded, and the remaining patients were studied in protocol 2 (14,665 patients). The study outcomes are major adverse CV events (MACE), which included myocardial infarction (MI), AF, ischemic and hemorrhagic stroke (IHS), hospitalization for heart failure (HHF), and cardiac death. The primary outcomes were MACE and each of the MACE components.

RESULTS

At the end of maximal 5-year follow-up (median 22.18 months with interquartile range 7.20-49.08 months for MACE in protocol 1 and 23.46 months with interquartile range 8.15-49.02 months for MACE in protocol 2), compared with an E/e' value of <8, an intermediate value of E/e' 8 to 15 and a high value of E/e' >15 were significantly associated with MACE, MI, AF, IHS, HHF, and cardiac death in protocol 1 (all P < .0001). In protocol 2, an intermediate E/e' value of 8 to 15 and a high value of E/e' >15 were significantly associated with MACE, MI, AF, IHS, HHF, and CV death (all P < .05), except an intermediate value E/e' 8 to 15 was not associated with AF.

CONCLUSIONS

In a diverse population of inpatients and outpatients with and without cardiac diseases and risk factors, the echocardiographic parameter E/e' was associated with CV events and is a useful marker of risk.

Development and Validation of a Novel Score for Predicting Long-Term Mortality after an Acute Ischemic Stroke

BACKGROUND

Long-term mortality prediction can guide feasible discharge care plans and coordinate appropriate rehabilitation services. We aimed to develop and validate a prediction model to identify patients at risk of mortality after acute ischemic stroke (AIS).

METHODS

The primary outcome was all-cause mortality, and the secondary outcome was cardiovascular death. This study included 21,463 patients with AIS. Three risk prediction models were developed and evaluated: a penalized Cox model, a random survival forest model, and a DeepSurv model. A simplified risk scoring system, called the C-HAND (history of Cancer before admission, Heart rate, Age, eNIHSS, and Dyslipidemia) score, was created based on regression coefficients in the multivariate Cox model for both study outcomes.

RESULTS

All experimental models achieved a concordance index of 0.8, with no significant difference in predicting poststroke long-term mortality. The C-HAND score exhibited reasonable discriminative ability for both study outcomes, with concordance indices of 0.775 and 0.798.

CONCLUSIONS

Reliable prediction models for long-term poststroke mortality were developed using information routinely available to clinicians during hospitalization.

Fetal-neonatal and maternal outcomes in women with Sjögren syndrome: a population-based registry linkage study

OBJECTIVE

To provide better preconceptional and prenatal counseling to patients with Sjögren syndrome (SS).

METHODS

In total, 2,100 143 pregnancies between 2004 and 2014 were identified in the Taiwan National Health Insurance database and birth registry. The maternal history of SS was ascertained, and data were compared between pregnant women with and without SS. We assessed the odds ratios and 95% confidence intervals of fetal-neonatal and maternal outcomes.

RESULTS

There were 449 pregnancies in women with SS and 2,099 694 pregnancies in women without SS. Regarding fetal outcomes, the risks of still birth (odds ratio [OR] = 2.14, 95% confidence interval [CI] = 1.01-4.55), low birth weight (<2500 g, OR = 2.53, 95% CI = 1.92-3.33), small for gestational age (OR = 2.03, 95% CI = 1.57-2.03), and fetal distress (OR = 1.72, 95% CI = 1.2-2.45) as well as risks of pulmonary oedema (OR = 11.64, 95% CI = 1.62-83.48), shock (OR = 6.07, 95% CI = 1.51-24.3), and respiratory distress (OR = 5.61, 95% CI = 1.39-22.6) were higher in the SS group than in the non-SS group.

CONCLUSION

Women with SS have significant risks of adverse fetal-neonatal and maternal outcomes and must undergo prenatal counseling to understand the risks involved before conception.

The effect of paternal psoriasis on neonatal outcomes: a nationwide population-based study

Background

Psoriasis is a chronic autoimmune disease involving both environmental and genetic risk factors. Maternal psoriasis often results in poor pregnancies that influence both mothers and newborns. However, the influence of paternal psoriasis on the newborn remains unknown. The aim of this study was to investigate whether paternal psoriasis is associated with increased risk of adverse neonatal outcomes, within a nationwide population-based data setting.

Methods

Singleton pregnancies were identified in the Taiwan National Health Insurance database and National Birth Registry between 2004-2011 and classified into four study groups according to whether mothers and spouses had psoriasis (paternal(-)/maternal(-), paternal(+)/maternal(-), paternal(-)/maternal(+), and paternal(+)/maternal(+)). Data were analyzed retrospectively. Adjusted odds ratios (aOR) or hazard ratios (aHR) were calculated to evaluate the risk of neonatal outcomes between groups.

Results

A total of 1,498,892 singleton pregnancies were recruited. Newborns of fathers with psoriasis but not of mothers with psoriasis were associated with an aHR (95% CI) of 3.69 (1.65-8.26) for psoriasis, 1.13 (1.06-1.21) for atopic dermatitis and 1.05 (1.01-1.10) for allergic rhinitis. Newborns of mothers with psoriasis but not of fathers with psoriasis were associated with an aOR (95% CI) of 1.26 (1.12-1.43) for low birth weight (<2500 g) and 1.64 (1.10-2.43) for low Apgar scores, and an aHR of 5.70 (2.71-11.99) for psoriasis.

Conclusion

Newborns of fathers with psoriasis are associated with significantly higher risk of developing atopic dermatitis, allergic rhinitis and psoriasis. Caution is advised for adverse neonatal outcomes when either or both parents have psoriasis.

Artificial intelligence-enabled electrocardiographic screening for left ventricular systolic dysfunction and mortality risk prediction

Background

Left ventricular systolic dysfunction (LVSD) characterized by a reduced left ventricular ejection fraction (LVEF) is associated with adverse patient outcomes. We aimed to build a deep neural network (DNN)-based model using standard 12-lead electrocardiogram (ECG) to screen for LVSD and stratify patient prognosis.

Methods

This retrospective chart review study was conducted using data from consecutive adults who underwent ECG examinations at Chang Gung Memorial Hospital in Taiwan between October 2007 and December 2019. DNN models were developed to recognize LVSD, defined as LVEF <40%, using original ECG signals or transformed images from 190,359 patients with paired ECG and echocardiogram within 14 days. The 190,359 patients were divided into a training set of 133,225 and a validation set of 57,134. The accuracy of recognizing LVSD and subsequent mortality predictions were tested using ECGs from 190,316 patients with paired data. Of these 190,316 patients, we further selected 49,564 patients with multiple echocardiographic data to predict LVSD incidence. We additionally used data from 1,194,982 patients who underwent ECG only to assess mortality prognostication. External validation was performed using data of 91,425 patients from Tri-Service General Hospital, Taiwan.

Results

The mean age of patients in the testing dataset was 63.7 ± 16.3 years (46.3% women), and 8,216 patients (4.3%) had LVSD. The median follow-up period was 3.9 years (interquartile range 1.5-7.9 years). The area under the receiver-operating characteristic curve (AUROC), sensitivity, and specificity of the signal-based DNN (DNN-signal) to identify LVSD were 0.95, 0.91, and 0.86, respectively. DNN signal-predicted LVSD was associated with age- and sex-adjusted hazard ratios (HRs) of 2.57 (95% confidence interval [CI], 2.53-2.62) for all-cause mortality and 6.09 (5.83-6.37) for cardiovascular mortality. In patients with multiple echocardiograms, a positive DNN prediction in patients with preserved LVEF was associated with an adjusted HR (95% CI) of 8.33 (7.71 to 9.00) for incident LVSD. Signal- and image-based DNNs performed equally well in the primary and additional datasets.

Conclusion

Using DNNs, ECG becomes a low-cost, clinically feasible tool to screen LVSD and facilitate accurate prognostication.

Lumbar Bone Mineral Density Estimation From Chest X-Ray Images: Anatomy-Aware Attentive Multi-ROI Modeling

Osteoporosis is a common chronic metabolic bone disease often under-diagnosed and under-treated due to the limited access to bone mineral density (BMD) examinations, e.g., via Dual-energy X-ray Absorptiometry (DXA). This paper proposes a method to predict BMD from Chest X-ray (CXR), one of the most commonly accessible and low-cost medical imaging examinations. The proposed method first automatically detects Regions of Interest (ROIs) of local CXR bone structures. Then a multi-ROI deep model with transformer encoder is developed to exploit both local and global information in the chest X-ray image for accurate BMD estimation. The proposed method is evaluated on 13719 CXR patient cases with ground truth BMD measured by the gold standard DXA. The model predicted BMD has a strong correlation with the ground truth (Pearson correlation coefficient 0.894 on lumbar 1). When applied in osteoporosis screening, it achieves a high classification performance (average AUC of 0.968). As the first effort of using CXR scans to predict the BMD, the proposed algorithm holds strong potential to promote early osteoporosis screening and public health.

Deep Learning-Based Nuclear Morphometry Reveals an Independent Prognostic Factor in Mantle Cell Lymphoma

Blastoid/pleomorphic morphology is associated with short survival in mantle cell lymphoma (MCL), but its prognostic value is overridden by Ki-67 in multivariate analysis. Herein, a nuclear segmentation model was developed using deep learning, and nuclei of tumor cells in 103 MCL cases were automatically delineated. Eight nuclear morphometric attributes were extracted from each nucleus. The mean, variance, skewness, and kurtosis of each attribute were calculated for each case, resulting in 32 morphometric parameters. Compared with those in classic MCL, 17 morphometric parameters were significantly different in blastoid/pleomorphic MCL. Using univariate analysis, 16 morphometric parameters (including 14 significantly different between classic and blastoid/pleomorphic MCL) emerged as significant prognostic factors. Using multivariate analysis, Biologic MCL International Prognostic Index (bMIPI) risk group (P = 0.025), low skewness of nuclear irregularity (P = 0.020), and high mean of nuclear irregularity (P = 0.047) emerged as independent adverse prognostic factors. Additionally, a morphometric score calculated from the skewness and mean of nuclear irregularity (P = 0.0038) was an independent prognostic factor in addition to bMIPI risk group (P = 0.025), and a summed morphometric bMIPI score was useful for risk stratification of patients with MCL (P = 0.000001). These results demonstrate, for the first time, that a nuclear morphometric score is an independent prognostic factor in MCL. It is more robust than blastoid/pleomorphic morphology and can be objectively measured.

Factors Associated With Quality Care Among Adults With Rheumatoid Arthritis

IMPORTANCE

Although quality care markers exist for patients with rheumatoid arthritis (RA), the predictors of meeting these markers are unclear.

OBJECTIVE

To explore factors associated with quality care among patients with RA.

DESIGN, SETTING, AND PARTICIPANTS

A retrospective cohort study using insurance claims from 2009 to 2017 was conducted, and 6 sequential logistic regression models were built to evaluate quality care markers. Quality care markers were measured at 1 year post-RA diagnosis for each patient. The MarketScan Research Database, which contains commercial and Medicare Advantage administrative claims data from more than 100 million individuals in the US, was used to identify patients aged 18 to 64 years with a diagnosis claim for RA. Patients with conditions presenting similar to RA and missing demographic characteristics were excluded. Data analysis occurred between February 18 and May 5, 2022.

EXPOSURES

Success or failure to meet selected RA quality care markers within 1 year after RA diagnosis.

MAIN OUTCOMES AND MEASURES

Prevalence of meeting successive quality care markers for RA.

RESULTS

Among 581 770 patients, 430 843 (74.1%) were women and the mean (SD) age was 48.9 (11.3) years. Most patients (236 285 [40.6%]) resided in the South and had an income less than or equal to $45 200 (490 366 [84.3%]). Of the total study population, 399 862 individuals (68.7%) met at least 1 quality care marker and 181 908 (31.3%) met 0 markers. Most commonly, patients met annual laboratory testing (299 323 [51.5%]) and referral to a rheumatologist (256 765 [44.1%]) markers. The least met marker was receiving hepatitis B screening prior to initiation of disease-modifying antirheumatic drug (DMARD) therapy (18 548 [3.2%]). Women were most likely to meet all quality care markers except receiving DMARDs with hepatitis B screening (odds ratio, 1.14; 95% CI, 1.12-1.16). Individuals with lower median household income had lower odds of receiving a rheumatologist referral, an annual physical examination, or annual laboratory testing, but greater odds of receiving the other quality care markers. Patients with Medicare and those with comorbidities were generally less likely to meet quality care markers.

CONCLUSIONS AND RELEVANCE

In this cohort study of patients with RA, findings indicated downstream associations with rheumatologist referral and receiving DMARDs and varied associations between meeting quality care markers and patient characteristics. These findings suggest that prioritizing early care, especially for vulnerable patients, will ensure that quality care continues.

Assessing the causal relationships between gout and hypertension: a bidirectional Mendelian randomisation study with coarsened exposures

Objectives

Observational studies have demonstrated associations between gout and hypertension, but whether they are causal remains unclear. Our work aims to assess the causal relationship between gout and hypertension.

Methods

We obtained genetic information from the Taiwan Biobank, including 88,347 participants and 686,439 single-nucleotide polymorphisms (SNPs). A novel model of Mendelian randomisation (MR) with coarsened exposures was used to examine the causality between the liability of gout on hypertension and vice versa, using 4 SNPs associated with gout and 10 SNPs associated with hypertension after removal of SNPs associated with measured confounders. The binary exposure (gout/hypertension) can be considered a coarsened approximation of a latent continuous trait. The inverse-variance weighted (IVW) and polygenic risk score (PRS) methods were used to estimate effect size. The MR analysis with coarsened exposures was performed with and without adjustments for covariates.

Results

Of the 88,347 participants, 3253 (3.68%) had gout and 11,948 (13.52%) had hypertension (men, 31.9%; mean age 51.1 [SD, 11.1] years). After adjusting to measured confounders, MR analysis with coarsened exposures showed a significant positive causal effect of the liability of gout on hypertension in both the IVW method (relative risk [RR], 1.10; 95% confidence interval [CI], 1.03–1.19; p = 0.0077) and the PRS method (RR, 1.10; 95% CI, 1.02–1.19; p = 0.0092). The result of causality was the same before and after involving measured confounders. However, there was no causal effect of the liability of hypertension on gout.

Conclusions

In this study, we showed that the liability of gout has a causal effect on hypertension, but the liability of hypertension does not have a causal effect on gout. Adequate management of gout may reduce the risk of developing hypertension.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13075-022-02933-4.

Fetal-neonatal and maternal pregnancy outcomes in women with rheumatoid arthritis: a population-based cohort study

OBJECTIVES

Rheumatoid arthritis (RA) may adversely influence pregnancy and lead to adverse birth outcomes. This study estimated the risk of adverse fetal-neonatal and maternal pregnancy outcomes in women with RA.

DESIGN

This was a retrospective cohort study.

SETTING

We used both the National Health Insurance database and the Taiwan Birth Reporting System, between 2004 and 2014.

PARTICIPANTS

We identified 2 100 143 singleton pregnancies with 922 RA pregnancies, either live births or stillbirths, delivered by 1 468 318 women.

OUTCOME MEASURES

ORs with 95% CIs for fetal-neonatal and maternal outcomes were compared between pregnancies involving mothers with and without RA using an adjusted generalised estimating equation model.

RESULTS

Covariates including age, infant sex, Charlson Comorbidity Index, urbanisation, income, occupation, birth year and maternal nationality were adjusted. Compared with pregnancies in women without RA, pregnancies in women with RA showed that the fetuses/neonates had adjusted ORs (95% CI) of 2.03 (1.66 to 2.50) for low birth weight (n=123), 1.99 (1.64 to 2.40) for prematurity (n=141), 1.77 (1.46 to 2.15) for small for gestational age (n=144) and 1.35 (1.03 to 1.78) for fetal distress (n=60). Pregnancies in women with RA had adjusted ORs (95% CI) of 1.24 (1.00 to 1.52) for antepartum haemorrhage (n=106), 1.32 (1.15 to 1.51) for caesarean delivery (n=398), and 3.33 (1.07 to 10.34) for disseminated intravascular coagulation (n=3), compared with women without RA. Fetuses/neonates born to mothers with RA did not have a higher risk of being stillborn or having fetal abnormalities. Pregnant women with RA did not have increased risks of postpartum death, cardiovascular complications, surgical complications or systemic organ dysfunction.

CONCLUSIONS

Pregnancies in women with RA were associated with higher risks of multiple adverse fetal-neonatal and maternal outcomes; however, most pregnancies in these women were successful.

Ultrasound Can Be Usefully Integrated with the Clinical Assessment of Nail and Enthesis Involvement in Psoriasis and Psoriatic Arthritis

Objectives: This study aimed to examine and compare the findings of nail and enthesis ultrasonography in patients with psoriasis and psoriatic arthritis. Methods: We identified 154 patients with psoriatic arthritis and 35 patients with psoriasis who were treated at Chang Gung Memorial Hospital, Taiwan, between September 2018 and January 2019. Results: There were significant differences in the Nail Psoriasis Severity Index scores and Glasgow Ultrasound Enthesitis Scoring System scores between patients with psoriasis and those with psoriatic arthritis. B-mode ultrasonography revealed that onychopathic changes were more common in the psoriasis group. The psoriatic arthritis group showed a higher proportion of lower-limb enthesopathy, with significant differences in distal patellar ligament thickness and Achilles tendon thickness. Conclusion: The findings of nail ultrasonography were more severe in psoriasis cases, and the ultrasonographic findings of enthesopathy of the lower limb were more severe in cases of psoriatic arthritis.

Electrocardiographic changes associated with SGLT2 inhibitors and non-SGLT2 inhibitors: A multi-center retrospective study

Background

Sodium-glucose co-transporter 2 (SGLT2) inhibitors has been shown with cardiovascular benefit in type 2 diabetes mellitus (T2DM) patients. However, its osmotic diuresis still concern physicians who may look for possible electrolyte imbalance. We therefore aimed to investigate electrocardiographic (ECG) changes associated with SGLT2 inhibitors.

Methods

Electronic medical records from Chang Gung Research Database between January 1, 2001 and January 31, 2019 were searched for patients with ECG reports and patients on an oral hypoglycemic agent (OHA). We then separate these T2DM patients with EKG into those taking either SGLT2 inhibitors or non-SGLT2 inhibitors. We excluded patients with OHA use <28 days, age <18 years, baseline ECG QTc > 500 ms, and ECG showing atrial fibrillation or atrial flutter. Propensity score matching (PSM) was performed between groups by age, sex, comorbidities, and medications (including QT prolonging medications). Conditional logistic regression and Firth's logistic regression for rare events were employed to compare the difference between SGLT2 and non-SGLT2 inhibitor patients.

Results

After exclusion criteria and PSM, there remained 1,056 patients with ECG on SGLT2 inhibitors and 2,119 patients with ECG on non-SGLT2 inhibitors in the study. There were no differences in PR intervals, QT prolongations by Bazett's or Fridericia's formulas, new onset ST-T changes, new onset CRBBB or CLBBB, and ventricular arrhythmia between the group of patients on SGLT2 inhibitors and the group of patients on non-SGLT2 inhibitors. There were no differences between the two groups in terms of cardiovascular death and sudden cardiac death. In addition, there were no differences between the two groups in terms of electrolytes.

Conclusions

Compared with T2DM patients on non-SGLT2 inhibitors, there were no differences in PR interval, QT interval, ST-T changes, bundle-branch block, or ventricular arrhythmia in the patients on SGLT2 inhibitors. There were no differences in cardiovascular mortality between these two groups. In addition, there were no electrolyte differences between groups. SGLT2 inhibitors appeared to be well-tolerated in terms of cardiovascular safety.

Breast cancer-related lymphedema correlated with incidence of cellulitis and mortality

BACKGROUND

This study investigated breast cancer-related lymphedema (BCRL) and its correlation with the incidence of cellulitis and mortality in the National Health Insurance (NHI) database in Taiwan.

METHODS

Between 2004 and 2014, the NHI database of patients with breast cancer who underwent surgical procedures, adjuvant therapies, BCRL, cellulitis, and mortality were retrospectively reviewed. Cox regression was used to calculate hazard ratios (HRs) and 95% confidence intervals (CIs) for incidence of BCRL and cellulitis in different treatment groups. The associations of BCRL with the incidence of cellulitis and mortality were further analyzed using the Kaplan-Meier curve.

RESULTS

Among 100 301 patients, 5464 (5.4%) developed BCRL with a median onset of 1.3 years. At a mean follow-up of 4.77 years, the incidence of cellulitis in the BCRL group (12.7%, 694/5464 patients) was significantly higher than in the no-BCRL group (2.73%, 2589/94 837 patients) (HR: 3.74; 95% CI: 3.43-4.08; p < 0.0001). At a mean follow-up of 5.77 years, the mortality rate in the cellulitis group (34.21%, 1123/3283 patients) was significantly greater than in the no-cellulitis group (16.29%, 15 804/97 018 patients) (HR: 1.17; 95% CI: 1.1-1.24; p < 0.0001).

CONCLUSIONS

BCRL had a significantly higher incidence of cellulitis and mortality.

Comparative safety of Janus kinase inhibitors and tumor necrosis factor inhibitors in patients undergoing treatment for rheumatoid arthritis

OBJECTIVES

Since 2010, biological disease-modifying antirheumatic drugs (bDMARDs) have been the dominant mode of treatment for rheumatoid arthritis (RA). However, the safety of DMARDs, such as tumor necrosis factor inhibitors (TNFis) and Janus kinase inhibitors (JAKis), in treating patients with RA is a concern. We compared the safety outcomes of JAKis and TNFis in RA patients in clinical settings.

METHODS

Patients diagnosed with RA between 2015 and 2017 were identified from the Taiwan National Health Insurance Research Database and followed till 2018. Propensity score stabilized weighting was used to balance the baseline characteristics of the JAKis and TNFis groups. The incidences of safety outcomes, namely cardiovascular (CV) events, tuberculosis (TB), total hip replacement (THR), total knee replacement (TKR), and all-cause mortality, were compared between the 2 study groups.

RESULTS

A total of 3179 patients with RA who were administered JAKis (n = 822) and TNFis (n = 2357) were included in this study. The mean follow-up duration was 2.02 years in the JAKis group and 2.10 in the TNFis group. All-cause mortality had the highest incidence rate, followed by TKR, THR, CV events, and TB. A lower incidence rate of the study outcomes was observed in the JAKis group than in the TNFis group but without statistical significance.

CONCLUSION

Comparable safety issues and mortality rates were observed for JAKis and TNFis in RA patients treated in real-world settings.

The Application of Artificial-Intelligence-Assisted Dental Age Assessment in Children with Growth Delay

Background: This study aimed to reveal the efficacy of the artificial intelligence (AI)-assisted dental age (DA) assessment in identifying the characteristics of growth delay (GD) in children. Methods: The panoramic films matching the inclusion criteria were collected for the AI model training to establish the population-based DA standard. Subsequently, the DA of the validation dataset of the healthy children and the images of the GD children were assessed by both the conventional methods and the AI-assisted standards. The efficacy of all the studied modalities was compared by the paired sample t-test. Results: The AI-assisted standards can provide much more accurate chronological age (CA) predictions with mean errors of less than 0.05 years, while the traditional methods presented overestimated results in both genders. For the GD children, the convolutional neural network (CNN) revealed the delayed DA in GD children of both genders, while the machine learning models presented so only in the GD boys. Conclusion: The AI-assisted DA assessments help overcome the long-standing populational limitation observed in traditional methods. The image feature extraction of the CNN models provided the best efficacy to reveal the nature of delayed DA in GD children of both genders.

Urine Soluble CD163 Is a Promising Biomarker for the Diagnosis and Evaluation of Lupus Nephritis

Introduction

Urine-soluble CD163 (usCD163) is released from alternatively activated macrophages involved in the resolution of inflammation in glomeruli and plays an important role in glomerulonephritis. This study explored the role of usCD163 in patients with systemic lupus erythematosus (SLE).

Materials and Methods

usCD163 concentrations were measured cross-sectionally in 261 SLE patients in Taiwan. Clinical and laboratory data were collected, and SLE disease activity scores were calculated to assess the correlation with usCD163.

Results

SLE patients with high usCD163 levels tended to be younger, with a higher hospital admission rate, higher prednisolone dose, lower estimated glomerular filtration rate, higher urine protein creatinine ratio (UPCR), more pyuria and hematuria, higher levels of inflammatory markers, higher rates of anemia, neutropenia, and lymphopenia, lower complement 3 (C3) levels, higher anti-double-stranded DNA antibody (anti-dsDNA Ab) levels, and higher disease activity scores (p < 0.05). usCD163 levels were significantly higher in patients with active lupus nephritis (LN) than in those with extrarenal or inactive SLE and correlated with UPCR, disease activity, and anti-dsDNA Ab levels. SLE patients with high usCD163 levels tended to have a higher chronic kidney disease stage.

Discussion and conclusion

The usCD163 level correlates with the severity of LN and disease activity in renal SLE.

Stroke Risk in Patients with Gout: A Nationwide Retrospective Cohort Study in Taiwan

Objectives: To estimate stroke risk in Taiwanese patients with gout. Methods: We enrolled patients from the Taiwan National Health Insurance Database, with gout diagnosed from 2000 to 2008, and followed them up until December 2018. This cohort was propensity score-matched according to birth year, sex, the date of diagnosis of gout, comorbidities, and co-medications with individuals without gout (controls) (n = 310,820 in each group). Stroke was defined as the primary diagnosis at discharge after the index date. To evaluate ischemic and hemorrhagic stroke risks, we calculated their incidence, hazard ratio (HR), and two-year moving average incidence rate. Results: The incidence (95% CI) and HR of ischemic stroke were lower in the gout group than in the control group in the first 3 years (incidence: 4.74 [4.60–4.88] vs. 5.17 [5.03–5.32] per 1000 person-years; HR: 0.92 [0.88–0.96]), then became significantly higher than in the control group after 3 years (incidence: 4.10 [4.04–4.16] vs. 3.81 [3.75–3.87] per 1000 person-years; HR: 1.08 [1.05–1.10]). Similarly, the incidence (95% CI) and HR of hemorrhagic stroke was lower in the gout group than in the control group in the first 3 years (incidence: 1.51 [1.43–1.59] vs. 1.70 [1.62–1.79] per 1000 person-years; HR: 0.88 [0.82–0.92]), then became significantly higher than in controls after 3 years (incidence: 1.43 [1.39–1.46] vs. 1.26 [1.22–1.29] per 1000 person-years; HR: 1.14 [1.10–1.18]). Conclusions: In Taiwan, patients with gout had higher risks of ischemic and hemorrhagic stroke after 3 years.

Comorbidities of systemic lupus erythematosus prior to and following diagnosis in different age-at-onset groups

OBJECTIVES

Systemic lupus erythematosus (SLE) is a female-dominated autoimmune disease that can occur at any age and has a diverse course. The clinical manifestation of this disease can vary depending on the patient's age at onset. The aim of this study was to characterise the comorbidities at the time of SLE diagnosis and after in different age groups.

METHODS

A total 1042 incident cases of SLE with a Catastrophic Illness Card in 2005 and 10,420 age- and sex-matched controls from the general population registered in the National Health Insurance Research Database in Taiwan were enrolled in the study. The risk of comorbidities before (adjusted odds ratio, [aOR]) and after (adjusted hazard ratio, [aHR]) of SLE was analysed. The burden of these SLE-associated comorbidities was weight by the Charlson comorbidity index (CCI). We used the cumulative incidence to evaluate the impact of comorbidities on different age onset groups.

RESULTS

In this study, musculoskeletal diseases had the highest positive association (aOR, 5.29; 95% confidence interval [CI]: 4.25-6.57) prior to the diagnosis of SLE and they were also the most common developing incident comorbidity after the diagnosis (HR, 13.7; 95% CI: 11.91-15.77). It only took less than 1 year for 50% of the late-onset SLE patients to develop any increase in CCI score. The developing comorbidities attributed to 16.3% all-cause mortality and they had the greatest impact on late-onset SLE patients, with 33.3% cumulative incidence to all-cause mortality. There is no difference in the incidence of infectious diseases across different age groups. The herpes zoster infection had the greatest cumulative incidence among the category of infection diseases in child-onset SLE patients.

CONCLUSION

SLE patients had increased risks of multiple pre-existing comorbidities at diagnosis. The developed comorbidity after diagnosis could contribute to all-cause mortality. The herpes zoster infection is primarily an issue in child-onset SLE patients.

Comorbidities in osteoarthritis (ComOA): a combined cross-sectional, case-control and cohort study using large electronic health records in four European countries

INTRODUCTION

Osteoarthritis (OA) is one of the leading chronic conditions in the older population. People with OA are more likely to have one or more other chronic conditions than those without. However, the temporal associations, clusters of the comorbidities, role of analgesics and the causality and variation between populations are yet to be investigated. This paper describes the protocol of a multinational study in four European countries (UK, Netherlands, Sweden and Spain) exploring comorbidities in people with OA.

METHODS AND ANALYSIS

This multinational study will investigate (1) the temporal associations of 61 identified comorbidities with OA, (2) the clusters and trajectories of comorbidities in people with OA, (3) the role of analgesics on incidence of comorbidities in people with OA, (4) the potential biomarkers and causality between OA and the comorbidities, and (5) variations between countries.A combined case-control and cohort study will be conducted to find the temporal association of OA with the comorbidities using the national or regional health databases. Latent class analysis will be performed to identify the clusters at baseline and joint latent class analysis will be used to examine trajectories during the follow-up. A cohort study will be undertaken to evaluate the role of non-steroidal anti-inflammatory drugs (NSAIDs), opioids and paracetamol on the incidence of comorbidities. Mendelian randomisation will be performed to investigate the potential biomarkers for causality between OA and the comorbidities using the UK Biobank and the Rotterdam Study databases. Finally, a meta-analyses will be used to examine the variations and pool the results from different countries.

ETHICS AND DISSEMINATION

Research ethics was obtained according to each database requirement. Results will be disseminated through the FOREUM website, scientific meetings, publications and in partnership with patient organisations.

Effect of Blue Light-Filtering Intraocular Lenses on Age-Related Macular Degeneration: A Nationwide Cohort Study With 10-Year Follow-up

PURPOSE

To determine the incidence rate of age-related macular degeneration (AMD) after cataract surgery and compare the relative incidence of AMD in pseudophakes with blue light-filtering intraocular lenses (BF-IOLs) and non-BF-IOLs.

DESIGN

A nationwide cohort study conducted using the Taiwan National Health Insurance Research Database.

METHODS

We enrolled 186,591 patients who underwent cataract surgery in both eyes between 2008 and 2013 and monitored them from the index date (the date of first cataract surgery) until AMD, death, loss to follow-up, or December 31, 2017, whichever occurred first. Propensity score matching (PSM) was used to balance the baseline characteristics between the BF-IOL and non-BF-IOL groups.

RESULTS

BF-IOLs were implanted in 21,126 patients (11.3%) and non-BF-IOLs were implanted in 165,465 patients (88.7%). Patients in the BF-IOL group tended to be younger, with fewer men, different cataract surgery years, higher income, more nonmanual workers, more patients from urban and suburban areas, and fewer chronic diseases compared with the non-BF-IOL group. With a mean follow-up period of 6.1 years (range, 1-10 years) after cataract surgery, 12,533 and 1655 patients developed non-exudative AMD and exudative AMD, respectively. The incidence rate of non-exudative AMD and exudative AMD (per 1000 person-years) was 9.95 and 1.22 for the BF-IOL group and 11.13 and 1.44 for the non-BF-IOL group, respectively. After PSM, no statistical difference in the incidence rate of nonexudative AMD (hazards ratio, 0.95; 95% CI, 0.88-1.03) and exudative AMD (hazard ratio, 0.96; 95% CI, 0.77-1.18) was observed between the BF-IOL and non-BF-IOL groups.

CONCLUSIONS

In Taiwan, the incidence rate of AMD after cataract surgery was 11.59 per 1000 person-years. The use of a BF-IOL for up to 10 years had no apparent advantage over a non-BF-IOL in the incidence of AMD.

Safety and Effectiveness of Direct Oral Anticoagulants vs. Warfarin in Patients With Atrial Fibrillation and Endoscopy-Diagnosed Peptic Ulcer

Background: Patients with active peptic ulcer (PU) were excluded from direct oral anticoagulant (DOAC) trials for stroke prevention in patients with atrial fibrillation (AF). This study evaluated the safety and effectiveness of DOACs in AF patients with active, inactive and no peptic ulcer (PU).

Methods: This study accessed electronic medical records from January 1, 2009 to May 31, 2019 at a multi-center healthcare provider in Taiwan and involved 2,955 AF patients who had undergone esophagogastroduodenoscopy ≤ 1 year before anticoagulation. Subjects were classified into 3 groups: active (n = 237), inactive (n = 828) and no-PU (n = 1,890) groups. We compared the risks of major bleeding, gastrointestinal bleeding, and ischemic stroke/systemic embolism (IS/SE) between DOACs and warfarin among the 3 groups.

Results: In the active PU group, there were no significant differences in the risks of major bleeding [hazard ratio (HR) = 0.65, 95% confidence interval (CI) 0.08–4.98, p = 0.676], gastrointestinal bleeding (HR = 0.65, 95% CI 0.08–4.98, p = 0.676) and IS/SE (HR = 2.58; 95% CI 0.53–12.70, p = 0.243) between DOAC and warfarin (as the reference). In the inactive PU group, there were no significant differences in the risks of major bleeding (HR = 0.36, 95% CI 0.09–1.39, p = 0.138), gastrointestinal bleeding (HR = 0.21, 95% CI 0.02–1.80, p = 0.153), and IS/SE (HR = 1.04, 95% CI 0.39–2.82, p = 0.934) between DOAC and warfarin (as the reference). In the no-PU group, DOACs were associated with lower risk of major bleeding (HR = 0.26, 95% CI 0.12–0.53, p < 0.001), gastrointestinal bleeding (HR = 0.25, 95% CI 0.01–0.59, p = 0.002), and similar risk of IS/SE (HR = 0.92, 95% CI 0.55–1.54, p = 0.757) compared to warfarin.

Conclusions: DOACs were as effective as warfarin in preventing IS/SE irrespective of PU status and safer than warfarin in reducing major bleeding in the no-PU group. In patients with active or inactive PUs, DOAC and warfarin were not significantly different in their effects on major bleeding or gastrointestinal bleeding.

Evaluation of Disease Complications Among Adults With Type 1 Diabetes and a Family History of Type 2 Diabetes in Taiwan

IMPORTANCE

Patients with type 1 diabetes (T1D) and a family history of type 2 diabetes (T2D) appear to be at a high risk of diabetes complications and other cardiovascular diseases. However, estimates of individual risks in patients in Taiwan are largely unavailable or unreliable.

OBJECTIVE

To evaluate the risk of diabetes complications and major adverse cardiovascular events (MACEs) in patients with T1D with a family history of T2D.

DESIGN, SETTING, AND PARTICIPANTS

A population-based cohort study used the Taiwan National Health Insurance Research Database. Participants included all individuals registered in that database on December 31, 2017, and followed up since March 1, 1995. The data were analyzed from December 6, 2018, to December 5, 2019.

EXPOSURE

Patients with T1D and a family history of T2D were evaluated.

MAIN OUTCOMES AND MEASURES

The prevalence and hazard ratios (HRs) of diabetes complications and other cardiovascular diseases in patients with T1D were analyzed. The MACEs were identified by diagnostic or procedural codes and heritability was formulated by the registry data of beneficiaries.

RESULTS

Of 27 370 965 individuals included in the database, 11 237 (mean [SD] age, 22.7 [14.4] years; 54% were female) had T1D. The crude prevalence of T1D was 0.04%, with a female to male ratio of 1.22: 1. The adjusted HRs in individuals who had a first-degree relative with T2D were 2.61 (95% CI, 1.32-5.16) for MACEs at an age at diagnosis of less than 20 years. Adjusted HRs were 1.44 (95% CI, 1.27-1.64) for diabetic neuropathy, 1.28 (95% CI, 1.12-1.47) for retinopathy, and 1.24 (95% CI, 1.06-1.47) for neuropathy at all ages of diagnosis.

CONCLUSIONS AND RELEVANCE

In this study of patients in Taiwan with T1D, having relatives with T2D was associated with an increase in the individual risks of developing diabetes complications. Patients with T1D and a family history of T2D might have more complications and require close management.

Fecal microbiota changes in NZB/W F1 mice after induction of lupus disease

The association between the gut microbiota and the development of lupus is unclear. We investigated alterations in the gut microbiota after induction of lupus in a murine model using viral peptide of human cytomegalovirus (HCMV). Three treatment arms for the animals were prepared: intraperitoneal injection of HCMVpp65 peptide, adjuvant alone, and PBS injection. Feces were collected before and after lupus induction biweekly for 16S rRNA sequencing. HCMVpp65 peptide immunization induced lupus-like effects, with higher levels of anti-dsDNA antibodies, creatinine, proteinuria, and glomerular damage, compared with mice treated with nothing or adjuvant only. The Simpson diversity value was higher in mice injected with HCMVpp65 peptide, but there was no difference in ACE or Chao1 among the three groups. Statistical analysis of metagenomic profiles showed a higher abundance of various families (Saccharimonadaceae, Marinifiaceae, and Desulfovibrionaceae) and genera (Candidatus Saccharimonas, Roseburia, Odoribacter, and Desulfovibrio) in HCMVpp65 peptide-treated mice. Significant correlations between increased abundances of related genera (Candidatus Saccharimonas, Roseburia, Odoribacter, and Desulfovibrio) and HCMVpp65 peptide immunization-induced lupus-like effects were observed. This study provides insight into the changes in the gut microbiota after lupus onset in a murine model.

Comparison of Indexes to Measure Comorbidity Burden and Predict All-Cause Mortality in Rheumatoid Arthritis

Objectives: To examine the comorbidity burden in patients with rheumatoid arthritis (RA) patients using a nationwide population-based cohort by assessing the Charlson Comorbidity Index (CCI), Elixhauser Comorbidity Index (ECI), Multimorbidity Index (MMI), and Rheumatic Disease Comorbidity Index (RDCI) scores and to investigate their predictive ability for all-cause mortality. Methods: We identified 24,767 RA patients diagnosed from 1998 to 2008 in Taiwan and followed up until 31 December 2013. The incidence of comorbidities was estimated in three periods (before, during, and after the diagnostic period). The incidence rate ratios were calculated by comparing during vs. before and after vs. before the diagnostic period. One- and 5-year mortality rates were calculated and discriminated by low and high-score groups and modified models for each index. Results: The mean score at diagnosis was 0.8 in CCI, 2.8 in ECI, 0.7 in MMI, and 1.3 in RDCI, and annual percentage changes are 11.0%, 11.3%, 9.7%, and 6.8%, respectively. The incidence of any increase in the comorbidity index was significantly higher in the periods of “during” and “after” the RA diagnosis (incidence rate ratios for different indexes: 1.33–2.77). The mortality rate significantly differed between the high and low-score groups measured by each index (adjusted hazard ratios: 2.5–4.3 for different indexes). CCI was slightly better in the prediction of 1- and 5-year mortality rates. Conclusions: Comorbidities are common before and after RA diagnosis, and the rate of accumulation accelerates after RA diagnosis. All four comorbidity indexes are useful to measure the temporal changes and to predict mortality.

Intravenous CT Contrast Media and Acute Kidney Injury: A Multicenter Emergency Department-based Study

Background Although the historical risk of acute kidney injury (AKI) after intravenous administration of contrast media might be overstated, the risk in patients with impaired kidney function remains a concern. Purpose To investigate whether intravenous contrast media administration during CT is associated with a higher risk of AKI and further hemodialysis compared with the risk in patients undergoing unenhanced CT. Materials and Methods This retrospective study evaluated patients who underwent contrast-enhanced or unenhanced CT in five Taiwanese emergency departments between 2009 and 2016. The outcomes were AKI within 48-72 hours after CT, AKI within 48 hours to 1 week after CT, or further hemodialysis within 1 month after CT. The associations between contrast media exposure and outcome were estimated by using an overlap propensity score weighted generalized regression model. Subgroup analyses were performed according to the estimated glomerular filtration rate (eGFR). Results The study included 68 687 patients (median age, 68 years; interquartile range, 53-74 years; 39 995 men) with (n = 31 103) or without (n = 37 584) exposure to contrast media. After propensity score weighting, contrast media exposure was associated with higher risk of AKI within 48-72 hours after CT (odds ratio [OR], 1.16; 95% CI: 1.04, 1.29; P = .007) but no significant risk at 48 hours to 1 week after CT (OR, 1.00; 95% CI: 0.93, 1.08; P = .90). Among patients with eGFR less than 30 mL/min/1.73 m², exposure to contrast media was associated with a higher AKI risk (48-72 hours after CT: OR, 1.36; 95% CI: 1.09, 1.70; P = .007) (48 hours-1 week after CT: OR, 1.49; 95% CI: 1.27, 1.74; P < .001) and a higher risk of hemodialysis (OR, 1.36; 95% CI: 1.09, 1.70; P = .008). For patients with eGFR greater than 45 mL/min/1.73.m², contrast media exposure was not associated with higher AKI risk (P > .05). Conclusion Contrast-enhanced CT was associated with higher risk of acute kidney injury and further hemodialysis among Taiwanese patients with an estimated glomerular filtration rate (eGFR) of less than 30 mL/min/1.73 m² but not those with an eGFR of more than 45 mL/min/1.73 m². © RSNA, 2021 Online supplemental material is available for this article.

Application of deep learning neural network in predicting bone mineral density from plain X-ray radiography

DeepDXA is a deep learning model designed to infer bone mineral density data from plain pelvis X-ray, and it can achieve good predicted value for clinical use.

PURPOSE

Osteoporosis is defined as a systemic disease of the bone characterized by a decrease in bone strength and deterioration of bone structure at the microscopic level, leading to bone fragility and increased risk of fracture. Bone mineral density (BMD) is the preferred method for the diagnosis of osteoporosis, and dual-energy x-ray absorptiometry (DXA) is the gold standard for diagnosing osteoporosis. Conventional radiography is more suited for the screening of osteoporosis rather than diagnosis, and osteoporosis can be detected on radiographs by experienced physicians only. This study explored the possibility of predicting BMD relative to DXA using patient radiographs.

METHODS

A deep learning algorithm of convolutional neural network (CNN) was used for the purpose. The method includes image segmentation, CNN learning, and a convolution-based regression model (DeepDXA) that links the isolated images of the femur bone to predict BMD value. Data were obtained in a single medical center from 2006 to 2018, with a total amount of 3472 pairs of pelvis X-ray and DXA examination within 1 year.

RESULTS

The proposed workflow successfully predicted BMD values of the femur bone with the correlation coefficient (R) of 0.85 (P < 0.001) and the accuracy of 0.88 for prediction osteoporosis, a finding that could be reliably ready for further clinical use.

CONCLUSION

When suspicious osteoporosis is seen on plain films using the deep learning method we developed, further referral to DXA for the definite diagnosis of osteoporosis is indicated.

Contour Transformer Network for One-Shot Segmentation of Anatomical Structures

Accurate segmentation of anatomical structures is vital for medical image analysis. The state-of-the-art accuracy is typically achieved by supervised learning methods, where gathering the requisite expert-labeled image annotations in a scalable manner remains a main obstacle. Therefore, annotation-efficient methods that permit to produce accurate anatomical structure segmentation are highly desirable. In this work, we present Contour Transformer Network (CTN), a one-shot anatomy segmentation method with a naturally built-in human-in-the-loop mechanism. We formulate anatomy segmentation as a contour evolution process and model the evolution behavior by graph convolutional networks (GCNs). Training the CTN model requires only one labeled image exemplar and leverages additional unlabeled data through newly introduced loss functions that measure the global shape and appearance consistency of contours. On segmentation tasks of four different anatomies, we demonstrate that our one-shot learning method significantly outperforms non-learning-based methods and performs competitively to the state-of-the-art fully supervised deep learning methods. With minimal human-in-the-loop editing feedback, the segmentation performance can be further improved to surpass the fully supervised methods.

Identification of nodal micrometastasis in colorectal cancer using deep learning on annotation-free whole-slide images

Detection of nodal micrometastasis (tumor size: 0.2-2.0 mm) is challenging for pathologists due to the small size of metastatic foci. Since lymph nodes with micrometastasis are counted as positive nodes, detecting micrometastasis is crucial for accurate pathologic staging of colorectal cancer. Previously, deep learning algorithms developed with manually annotated images performed well in identifying micrometastasis of breast cancer in sentinel lymph nodes. However, the process of manual annotation is labor intensive and time consuming. Multiple instance learning was later used to identify metastatic breast cancer without manual annotation, but its performance appears worse in detecting micrometastasis. Here, we developed a deep learning model using whole-slide images of regional lymph nodes of colorectal cancer with only a slide-level label (either a positive or negative slide). The training, validation, and testing sets included 1963, 219, and 1000 slides, respectively. A supercomputer TAIWANIA 2 was used to train a deep learning model to identify metastasis. At slide level, our algorithm performed well in identifying both macrometastasis (tumor size > 2.0 mm) and micrometastasis with an area under the receiver operating characteristics curve (AUC) of 0.9993 and 0.9956, respectively. Since most of our slides had more than one lymph node, we then tested the performance of our algorithm on 538 single-lymph node images randomly cropped from the testing set. At single-lymph node level, our algorithm maintained good performance in identifying macrometastasis and micrometastasis with an AUC of 0.9944 and 0.9476, respectively. Visualization using class activation mapping confirmed that our model identified nodal metastasis based on areas of tumor cells. Our results demonstrate for the first time that micrometastasis could be detected by deep learning on whole-slide images without manual annotation.

Automated bone mineral density prediction and fracture risk assessment using plain radiographs via deep learning

Dual-energy X-ray absorptiometry (DXA) is underutilized to measure bone mineral density (BMD) and evaluate fracture risk. We present an automated tool to identify fractures, predict BMD, and evaluate fracture risk using plain radiographs. The tool performance is evaluated on 5164 and 18175 patients with pelvis/lumbar spine radiographs and Hologic DXA. The model is well calibrated with minimal bias in the hip (slope = 0.982, calibration-in-the-large = -0.003) and the lumbar spine BMD (slope = 0.978, calibration-in-the-large = 0.003). The area under the precision-recall curve and accuracy are 0.89 and 91.7% for hip osteoporosis, 0.89 and 86.2% for spine osteoporosis, 0.83 and 95.0% for high 10-year major fracture risk, and 0.96 and 90.0% for high hip fracture risk. The tool classifies 5206 (84.8%) patients with 95% positive or negative predictive value for osteoporosis, compared to 3008 DXA conducted at the same study period. This automated tool may help identify high-risk patients for osteoporosis.

Glomerular Disease Classification and Lesion Identification by Machine Learning

Background

Classification of glomerular diseases and identification of glomerular lesions require careful morphological examination by experienced nephropathologists, which is labor-intensive, time-consuming, and prone to interobserver variability. In this regard, recent advance in machine learning-based image analysis is promising.

Methods

We combined Mask Region-based Convolutional Neural Networks (Mask R–CNN) with an additional classification step to build a glomerulus detection model using human kidney biopsy samples. A Long Short-Term Memory (LSTM) recurrent neural network was applied for glomerular disease classification, and another two-stage model using ResNeXt-101 was constructed for glomerular lesion identification in cases of lupus nephritis.

Results

The detection model showed state-of-the-art performance on variedly stained slides with F1 scores up to 0.944. The disease classification model showed good accuracies up to 0.940 on recognizing different glomerular diseases based on H&E whole slide images. The lesion identification model demonstrated high discriminating power with area under the receiver operating characteristic curve up to 0.947 for various glomerular lesions. Models showed good generalization on external testing datasets.

Conclusion

This study is the first-of-its-kind showing how each step of kidney biopsy interpretation carried out by nephropathologists can be captured and simulated by machine learning models. The models were integrated into a whole slide image viewing and annotating platform to enable nephropathologists to review, correct, and confirm the inference results. Further improvement on model performances and incorporating inputs from immunofluorescence, electron microscopy, and clinical data might realize actual clinical use.

Upregulation of miR-210-5p impairs dead cell clearance by macrophages through the inhibition of Sp1-and HSCARG-dependent NADPH oxidase pathway

The deficiency of dead cell clearance is a prominent pathogenic factor in systemic lupus erythematosus (SLE). In this study, the overexpression of miR-210-5p resulted in the accumulation of secondary necrotic cells (SNECs) in macrophages through the reduction of protein degradation. The upreguation of miR-210-5p inhibited NADPH oxidase (NOX) activation, reactive oxygen species (ROS) generation, and SNEC clearance. miR-210-5p overexpression suppressed Sp1 and HSCARG expression, and the knockdown of SP1 and HSCARG inhibited NOX expression and superoxide production in macrophages. Furthermore, patients with active SLE expressed a higher level of miR-210-5p and lower expression of SP1 and HSCARG in peripheral blood mononuclear cells. In summary, our findings indicate that the upregulation of miR-210-5p increases the accumulation of SNECs through a decrease in the Sp1-and HSCARG-mediated NOX activity and ROS generation in macrophages. Our results also suggest that targeting miR-210-5p may have therapeutic potential for SLE.

Using Slit-Lamp Images for Deep Learning-Based Identification of Bacterial and Fungal Keratitis: Model Development and Validation with Different Convolutional Neural Networks

In this study, we aimed to develop a deep learning model for identifying bacterial keratitis (BK) and fungal keratitis (FK) by using slit-lamp images. We retrospectively collected slit-lamp images of patients with culture-proven microbial keratitis between 1 January 2010 and 31 December 2019 from two medical centers in Taiwan. We constructed a deep learning algorithm consisting of a segmentation model for cropping cornea images and a classification model that applies different convolutional neural networks (CNNs) to differentiate between FK and BK. The CNNs included DenseNet121, DenseNet161, DenseNet169, DenseNet201, EfficientNetB3, InceptionV3, ResNet101, and ResNet50. The model performance was evaluated and presented as the area under the curve (AUC) of the receiver operating characteristic curves. A gradient-weighted class activation mapping technique was used to plot the heat map of the model. By using 1330 images from 580 patients, the deep learning algorithm achieved the highest average accuracy of 80.0%. Using different CNNs, the diagnostic accuracy for BK ranged from 79.6% to 95.9%, and that for FK ranged from 26.3% to 65.8%. The CNN of DenseNet161 showed the best model performance, with an AUC of 0.85 for both BK and FK. The heat maps revealed that the model was able to identify the corneal infiltrations. The model showed a better diagnostic accuracy than the previously reported diagnostic performance of both general ophthalmologists and corneal specialists.

A Multimodal Imaging-Based Deep Learning Model for Detecting Treatment-Requiring Retinal Vascular Diseases: Model Development and Validation Study

BACKGROUND

Retinal vascular diseases, including diabetic macular edema (DME), neovascular age-related macular degeneration (nAMD), myopic choroidal neovascularization (mCNV), and branch and central retinal vein occlusion (BRVO/CRVO), are considered vision-threatening eye diseases. However, accurate diagnosis depends on multimodal imaging and the expertise of retinal ophthalmologists.

OBJECTIVE

The aim of this study was to develop a deep learning model to detect treatment-requiring retinal vascular diseases using multimodal imaging.

METHODS

This retrospective study enrolled participants with multimodal ophthalmic imaging data from 3 hospitals in Taiwan from 2013 to 2019. Eye-related images were used, including those obtained through retinal fundus photography, optical coherence tomography (OCT), and fluorescein angiography with or without indocyanine green angiography (FA/ICGA). A deep learning model was constructed for detecting DME, nAMD, mCNV, BRVO, and CRVO and identifying treatment-requiring diseases. Model performance was evaluated and is presented as the area under the curve (AUC) for each receiver operating characteristic curve.

RESULTS

A total of 2992 eyes of 2185 patients were studied, with 239, 1209, 1008, 211, 189, and 136 eyes in the control, DME, nAMD, mCNV, BRVO, and CRVO groups, respectively. Among them, 1898 eyes required treatment. The eyes were divided into training, validation, and testing groups in a 5:1:1 ratio. In total, 5117 retinal fundus photos, 9316 OCT images, and 20,922 FA/ICGA images were used. The AUCs for detecting mCNV, DME, nAMD, BRVO, and CRVO were 0.996, 0.995, 0.990, 0.959, and 0.988, respectively. The AUC for detecting treatment-requiring diseases was 0.969. From the heat maps, we observed that the model could identify retinal vascular diseases.

CONCLUSIONS

Our study developed a deep learning model to detect retinal diseases using multimodal ophthalmic imaging. Furthermore, the model demonstrated good performance in detecting treatment-requiring retinal diseases.