Stepwise Variable Selection Research Articles

Association between Periodontitis and Chronic Rhinosinusitis Involving Maxillary Sinus Measured by Lund Mackay Staging System.

This study aimed to evaluate the association between periodontitis and chronic rhinosinusitis (CRS) via cone-beam-computed tomography (CBCT) using the Lund–Mackay staging system. CBCT images from different departments of the school of dentistry, at Prince Sattam University were evaluated for the presence of rhinosinusitis. All the CBCT scans were exposed for multiple indications, and no patients had a scan exposed solely for this study. The Lund–Mackay staging system was used to measure the CRS in the CBCT. Descriptive statistics for the frequencies and percentages were used to summarize the data. Logistic regression was used to examine the associations between periodontitis and CRS. Each variable was assessed individually by using multivariable analysis. Collinearity issues among the variables were solved to select a limited set of factors using a stepwise variable selection procedure. A total of 399 CBCT images were included in the current research. Logistic regression showed that only gender was significantly associated (p = 0.0001) with the presence of CRS. However, a stepwise variable selection procedure included gender and bone loss as significantly associated with CRS. No significant difference was observed between unilateral vs. bilateral CRS in gender, bone loss, medical status, and periodontitis. However, only gender showed a significant difference in both bilateral vs. no CRS and unilateral vs. no CRS. Periodontitis is not associated with CRS. However, gender has a significant influence on CRS.

Associations of SCORE2, circulating cardiovascular biomarkers and carotid intima-media-thickness in a population-based cohort

Abstract Background The updated SCORE2 model to estimate the 10-year risk of fatal and non-fatal cardiovascular disease (CVD) was recently introduced, which is based solely on traditional cardiovascular risk factors. Purpose We aimed to investigate the associations of SCORE2-predicted risk with four circulating cardiovascular biomarkers (high-sensitivity troponin I [hs-cTnI], N-terminal pro B-type natriuretic peptide [NT-proBNP], cystatin C-derived estimated glomerular filtration rate [eGFR] and high-sensitivity C-reactive protein [hs-CRP]) as well as mean carotid intima-media-thickness (cIMT) in a prospective, population-based German cohort. Methods In the first set of 10,000 participants, who were aged 45–74 years and recruited between 2016 and 2019, the SCORE2 model was applied in a cross-sectional manner. Individuals with prevalent CVD were excluded for this purpose. Eligible individuals were then categorized into five risk groups (<2.5%, 2–<5%, 5–<10%, 10–<15%, and ≥15%) according to SCORE2. To test for the associations of circulating biomarkers and cIMT with SCORE2, we created box plots and computed Pearson's correlation coefficients (R). Considering cIMT as a biological surrogate for incident CVD, we explored the incremental utility of circulating biomarkers to predict cIMT beyond SCORE2 by multivariable logistic regression analysis with stepwise selection of variables, quantified by Beta-coefficients per one standard deviation (SD) increase with respective 95% confidence intervals (CI). Discrimination (C-index) and category-free net reclassification improvement (NRI) for predicting mean cIMT >1mm were calculated for this extended model in comparison to SCORE2 alone as the reference. Results In 8,518 individuals free of CVD, median estimated 10-year risk of CVD based on SCORE2 was 6.1 (interquartile range [IQR] 3.2, 9.9) %. All four investigated biomarkers (hs-cTnI, R=0.41; NT-proBNP, R=0.21; hs-CRP, R=0.22; eGFR, R=−0.44; all P<0.001), and mean cIMT (R=0.42, P<0.001) correlated strongly with the continuous SCORE2 risk and the respective SCORE2 risk categories (Figure 1). In multivariable regression analysis, all four circulating biomarkers remained significant independent predictors of mean cIMT when added to SCORE 2 (Figure 2). When compared to SCORE2 as the reference model (C-index 0.763, 95% CI 0.742–0.784) to predict mean cIMT >1mm, the joint addition of all four stepwise-selected investigated biomarkers led to a small but significant improvement of discrimination (C-index 0.770, 95% CI 0.749–0.791, P<0.001) and reclassification yield (NRI 0.154, 95% CI 0.059–0.250, P=0.002). Conclusions All four investigated circulating biomarkers depicting different pathophysiological pathways and mean cIMT correlate strongly with the cardiovascular risk estimated by SCORE2. Circulating biomarkers may further improve CVD-risk prediction when added to the traditional cardiovascular risk factors currently considered by SCORE2. Funding Acknowledgement Type of funding sources: None.

Two-year outcomes of patients with atrial fibrillation and heart failure: the ETNA-AF-Europe registry

Abstract Background Heart failure (HF) is an established risk factor for stroke and systemic embolic events (SEE) in subjects with atrial fibrillation (AF), but it is debated whether this risk varies according to left ventricular ejection fraction (LVEF). Methods We investigated the impact of HF in the ETNA-AF-Europe registry, a prospective, multi-centre, post-authorisation, observational study enrolling patients treated with edoxaban for AF in 825 sites from 10 European countries. This 2-year follow-up analysis is based on a data snapshot from 26 October 2020. HF was defined as a) history of HF or b) ischaemic cardiomyopathy or c) EF <40% or d) dyspnoea not due to chronic obstructive pulmonary disease together with ≥1 of the following: ischaemic heart disease, valvular heart disease, or hypertension treated with ≥3 drugs. Patients' characteristics are summarised descriptively and clinical outcomes are reported as annualised event rates. The hazard ratio (HR) with 95% confidence intervals (CI) for the association of HF with the outcomes was assessed in Cox regression models with stepwise variable selection. Results Of the 13,133 patients, 1,854 (14.1%) had HF; LVEF was available for 1,489 (80.3%), and was <40% in 671 (43.9%) and ≥40% in 857 (56.1%). Patients with HF were more often men and slightly older than those without (Table 1). As expected, they also had more cardiovascular (CV) comorbidities and higher CHA2DS2-VASC and, to a lesser extent, HAS-BLED scores (Table 1). At the end of the 2-year follow-up, the rates of ischaemic stroke/transient ischaemic attack (TIA)/SEE, major bleeding, intracranial haemorrhage (ICH), CV death, and all-cause death were higher in patients with than without HF (Figure 1). When patients with HF were categorized according to LVEF, ischaemic stroke/TIA/SEE was more frequent in those with LVEF ≥40% vs those with LVEF <40%. By contrast, more patients with LVEF <40% died due to any as well as CV causes. The rates of major bleeding and ICH were comparable between the two subgroups (Figure 1). Univariable Cox regression analysis confirmed the association of HF with major bleeding (HR 2.01, 95% CI [1.49–2.71]) and all-cause death (2.62 [2.28–3.02]), but not with ischaemic stroke/TIA/SEE (1.06 [0.72–1.55]). The results were consistent when LVEF was taken into account: the HRs for LVEF <40% or LVEF ≥40%, respectively, were 1.60 (0.99–2.60) and 1.55 (1.02–2.38) for major bleeding, 2.11 (1.69–2.63) and 1.59 (1.28–1.97) for all-cause death, and 0.66 (0.31–1.41) and 1.19 (0.71–1.98) for ischaemic stroke/TIA/SEE. Conclusions In this real-world, large cohort of patients with AF on edoxaban, those with HF at baseline faced more ischaemic, bleeding, and death events, and having HF increased the risk of major bleeding and death, with no differences according to LVEF. Funding Acknowledgement Type of funding sources: Private company. Main funding source(s): This research was funded by Daiichi Sankyo Europe.

High PSQI score is associated with the development of dyskinesia in Parkinson’s disease

Dyskinesia is one of the most disabling motor complications in Parkinson’s Disease (PD). Sleep is crucial to keep neural circuit homeostasis, and PD patients often suffer from sleep disturbance. However, few prospective studies have been conducted to investigate the association of sleep quality with dyskinesia in PD. The objective of the current study is to investigate the association between sleep quality and dyskinesia and build a prediction model for dyskinesia in PD. We prospectively followed a group of PD patients without dyskinesia at baseline for a maximum of 36 months. Univariable and multivariable Cox regression with stepwise variable selection was used to investigate risk factors for dyskinesia. The performance of the model was assessed by the time-dependent area under the receiver-operating characteristic curve (AUC). At the end of follow-up, 32.8% of patients developed dyskinesia. Patients with bad sleep quality had a significantly higher proportion of dyskinesia compared with those with good sleep quality (48.1% vs. 20.6%, p = 0.023). Multivariable Cox regression selected duration of PD, sleep quality, cognition, mood, and levodopa dose. Notably, high Pittsburgh sleep quality index (PSQI) score was independently associated with an increased risk of dyskinesia (HR = 2.96, 95% CI 1.05–8.35, p = 0.041). The model achieved a good discriminative ability, with the highest AUC being 0.83 at 35 months. Our results indicated that high PSQI score may increase the risk of developing dyskinesia in PD, implying that therapeutic intervention targeting improving sleep quality may be a promising approach to prevent or delay the development of dyskinesia in PD.

Endurance Versus Non-endurance Athletes: A Descriptive Discriminant Analysis Based On Echocardiography

PURPOSE: To discriminate between endurance and non-endurance athletes according to echocardiographic measures. METHODS: Echocardiographic exams carried out in 201 male and female athletes of diverse sports were studied (Age = 21.7 ± 5.4 years; mean ± SD). A set of fifteen measures was considered: Left ventricular diastolic diameter (LVDD), Left ventricular diastolic diameter index (LVDDI), Interventricular septum thickness (IST), Left ventricular posterior wall thickness (LVPWT), Relative wall thickness (RWT), Left ventricular mass (LVM), Left ventricular mass index (LVMI), Left auricular diameter (LAD), Aortic root diameter (ARD), Inferior vena cava diameter (IVCD), E-wave velocity (EV), A-wave velocity (AV), E/A ratio (EAR), E/e′ ratio (Ee′R) and Right ventricular systolic velocity (RVS). The athletes were categorized in two groups based on the characteristics of their sport discipline: “Endurance” (cardiorespiratory endurance) and “Other” (Non-Endurance). Gender-stratified analyses were performed. MANOVA was firstly implemented to evaluate for a significance difference between endurance and non-endurance athletes. Post-hoc analysis was done by means of Descriptive discriminant analysis; in order to obtain a more reduced set of variables, Stepwise variable selection method was previously performed using the Wilk’s Lambda criterion. Statistical significance was fixed at the 0.05 level. RESULTS: MANOVA was statistically significant in men (Wilk's λ = 0.78, F = 1.81, p = 0.04) and in women (Wilk's λ = 0.63, F = 2.82, p = 0.002). The Stepwise procedure selected two variables in the male (IST and LVDDI) and female (LVMI and LVDDI) strata. The discriminant function was highly significant in both cases (males: Wilk's λ = 0.86, F = 8.59, p < 0.001; females: Wilk's λ = 0.77, F = 13.21, p < 0.001). In men, the standardized discriminant coefficient and the structure coefficient were higher for IST (0.96 and 0.81) than for LVDDI (0.60 and 0.36). And in women, the coefficients corresponding to LVMI (0.62 and 0.84) were similar to the ones of LVDDI (0.59 and 0.81). CONCLUSIONS: In the male athletes, IST showed a more important contribution for distinguishing between the endurance and non-endurance groups than LVDDI. And in the female athletes, LVMI and LVDDI had a comparable contribution.

ICF-based prediction of return to work after trauma rehabilitation: Results of the icfPROreha study in patients with severe musculoskeletal injuries.

BackgroundPhysical aspects such as the type and severity of an injury are not the only factors contributing to whether or not a person can return to work (RTW) after a serious injury. A more comprehensive, biopsychosocial approach is needed to understand the complexity of RTW fully. The study aims to identify predictors of RTW 78 weeks after discharge from initial inpatient trauma rehabilitation in patients with severe musculoskeletal injuries using a biopsychosocial perspective.MethodsThis is a prospective multicenter longitudinal study with a follow-up of up to 78 weeks after discharge from trauma rehabilitation. Data on potential predictors were collected at admission to rehabilitation using a comprehensive assessment tool. The status of RTW (yes vs. no) was assessed 78 weeks after discharge from rehabilitation. The data were randomly divided into a training and a validation data set in a ratio of 9:1. On the training data, we performed bivariate and multiple logistic regression analyses on the association of RTW and potential predictors. The final logit model was selected via stepwise variable selection based on the Akaike information criterion. The final model was validated for the training and the validation data.ResultsData from 761 patients (n = 561 male, 73.7%; mean age: 47.5 years, SD 12.3), primarily suffering from severe injuries to large joints and complex fractures of the large tubular bones, could be considered for analyses. At 78 weeks after discharge, 618 patients (81.2%) had returned to work. Eleven predictors remained in the final logit model: general health, current state of health, sensation of pain, limitations and restrictions in activities and participation (disability), professional sector, ongoing legal disputes, financial concerns (assets), personality traits, life satisfaction preaccident, attitude to life, and demand for pension claim. A predicted probability for RTW based on the multiple logistic regression model of 76.3% was revealed as the optimal cut-off score based on the ROC curve.ConclusionA holistic biopsychosocial approach is needed to address RTW and strengthen person-centered treatment and rehabilitation. Patients at risk for no RTW in the long term can already be identified at the onset of rehabilitation.

Association of Multiple Sclerosis Prevalence With Sociodemographic, Health Systems, and Lifestyle Factors on a National and Regional Level.

Multiple sclerosis (MS) prevalence varies widely by country and world region, variation that is often attributed to latitude and its association with vitamin D exposure. Given that increasing latitude is also associated with higher national wealth, this study investigated associations between MS prevalence and other factors driving regional differences, with a focus on sociodemographic, health systems, and lifestyle factors on a national and regional level. Using data from multilateral organizations and scientific literature, an ecological study was conducted to evaluate associations between age- and sex-adjusted MS prevalence and prespecified sociodemographic (gross domestic product [GDP] per capita and gross national income [GNI] per capita), health systems (current health expenditure per capita and by percentage of GDP, universal health coverage [UHC] index, and medical doctors per capita), neurology-specific (MRI unit density and neurologists per capita), and lifestyle (obesity and tobacco use) factors. National, regional, and income-stratified data were aggregated and employed in relevant univariable and multivariable regression models. Stepwise variable selection techniques identified independent predictors of MS prevalence. Univariable regression analyses showed significant associations at the national level for all investigated factors, except obesity prevalence and tobacco use. Latitude was significantly associated with MS prevalence in all world regions (β = 0.16-2.16), whereas the UHC index was significantly associated in 5 of 6 world regions (β = 0.18-3.17). MS prevalence was significantly associated with all factors except lifestyle factors and MRI unit density in high-income countries, but no associations were observed in low-income countries. Latitude was associated with MS prevalence for all income strata except low-income countries (β = 0.55-1.62). In multivariable analyses, current health expenditure per capita (β = 0.083, 95% CI 0.048-0.12, p < 0.01) and latitude (β = 1.05, 95% CI 0.63-1.47, p < 0.01) remained significantly associated with MS prevalence. Health expenditure per capita is strongly associated with national MS prevalence, suggesting that theories that attribute variations in MS prevalence primarily to latitude effects on vitamin D are incomplete. Health care access significantly contributes to the global variations in MS prevalence, especially since national wealth rises with latitude and likely results in significant underestimation of MS prevalence in countries with lower health expenditure.

Impact of trans-stent gradient on outcome after PCI: results from a HAWKEYE substudy

To test whether quantitative flow ratio (QFR)-based trans-stent gradient (TSG) is associated with adverse clinical events at follow-up. A post-hoc analysis of the multi-center HAWKEYE study was performed. Vessels post-PCI were divided into four groups (G) as follows: G1: QFR ≥ 0.90 TSG = 0 (n = 412, 54.8%); G2: QFR ≥ 0.90, TSG > 0 (n = 216, 28.7%); G3: QFR < 0.90, TSG = 0 (n = 37, 4.9%); G4: QFR < 0.90, TSG > 0 (n = 86, 11.4%). Cox proportional hazards regression model was used to analyze the effect of baseline and prognostic variables. The final reduced model was obtained by backward stepwise variable selection. Receiver operating characteristic (ROC) was plotted and area under the curve (AUC) was calculated and reported. Overall, 449 (59.8%) vessels had a TSG = 0 whereas (40.2%) had TSG > 0. Ten (2.2%) vessel-oriented composite endpoint (VOCE) occurred in vessels with TSG = 0, compared with 43 (14%) in vessels with TSG > 0 (p < 0.01). ROC analysis showed an AUC of 0.74 (95% CI: 0.67 to 0.80; p < 0.001). TSG > 0 was an independent predictor of the VOCE (HR 2.95 [95% CI 1.77–4.91]). The combination of higher TSG and lower final QFR (G4) showed the worst long-term outcome while low TSG and high QFR showed the best outcome (G1) while either high TSG or low QFR (G2, G3) showed intermediate and comparable outcomes. Higher trans-stent gradient was an independent predictor of adverse events and identified a subgroup of patients at higher risk for poor outcomes even when vessel QFR was optimal (> 0.90).

Kiwi Plant Canker Diagnosis Using Hyperspectral Signal Processing and Machine Learning: Detecting Symptoms Caused by Pseudomonas syringae pv. actinidiae.

Pseudomonas syringae pv. actinidiae (Psa) has been responsible for numerous epidemics of bacterial canker of kiwi (BCK), resulting in high losses in kiwi production worldwide. Current diagnostic approaches for this disease usually depend on visible signs of the infection (disease symptoms) to be present. Since these symptoms frequently manifest themselves in the middle to late stages of the infection process, the effectiveness of phytosanitary measures can be compromised. Hyperspectral spectroscopy has the potential to be an effective, non-invasive, rapid, cost-effective, high-throughput approach for improving BCK diagnostics. This study aimed to investigate the potential of hyperspectral UV–VIS reflectance for in-situ, non-destructive discrimination of bacterial canker on kiwi leaves. Spectral reflectance (325–1075 nm) of twenty plants were obtained with a handheld spectroradiometer in two commercial kiwi orchards located in Portugal, for 15 weeks, totaling 504 spectral measurements. Several modeling approaches based on continuous hyperspectral data or specific wavelengths, chosen by different feature selection algorithms, were tested to discriminate BCK on leaves. Spectral separability of asymptomatic and symptomatic leaves was observed in all multi-variate and machine learning models, including the FDA, GLM, PLS, and SVM methods. The combination of a stepwise forward variable selection approach using a support vector machine algorithm with a radial kernel and class weights was selected as the final model. Its overall accuracy was 85%, with a 0.70 kappa score and 0.84 F-measure. These results were coherent with leaves classified as asymptomatic or symptomatic by visual inspection. Overall, the findings herein reported support the implementation of spectral point measurements acquired in situ for crop disease diagnosis.

A New Clinical Instrument for Estimating the Ambulatory Status after Irradiation for Malignant Spinal Cord Compression.

Simple SummarySince 2005, upfront surgery has been increasingly used in addition to radiotherapy for patients with malignant spinal cord compression (MSCC). As spinal surgery includes significant risks, careful patient selection is crucial. Individual risks and benefits should be considered when choosing an optimal treatment strategy. Benefits include preserving or regaining a patient’s ambulatory function. To facilitate the decision pro or contra upfront surgery, a new prognostic score was developed to predict ambulatory status after radiotherapy alone. This clinical score was created from data of patients previously treated in prospective trials. It includes three prognostic groups (17–21, 22–31, and 32–37 points) with post-radiotherapy ambulatory rates of 10%, 65%, and 97%, respectively. Patients of the 32–37 points group may not require upfront surgery. The new instrument achieved very high accuracy in predicting post-radiotherapy ambulatory and non-ambulatory status and was more precise than a previous prognostic score in predicting non-ambulatory status.Estimating post-treatment ambulatory status can improve treatment personalization of patients irradiated for malignant spinal cord compression (MSCC). A new clinical score was developed from data of 283 patients treated with radiotherapy alone in prospective trials. Radiotherapy regimen, age, gender, tumor type, interval from tumor diagnosis to MSCC, number of affected vertebrae, other bone metastases, visceral metastases, time developing motor deficits, ambulatory status, performance score, sensory deficits, and sphincter dysfunction were evaluated. For factors with prognostic relevance in the multivariable logistic regression model after backward stepwise variable selection, scoring points were calculated (post-radiotherapy ambulatory rate in % divided by 10) and added for each patient. Four factors (primary tumor type, sensory deficits, sphincter dysfunction, ambulatory status) were used for the instrument that includes three prognostic groups (17–21, 22–31, and 32–37 points). Post-radiotherapy ambulatory rates were 10%, 65%, and 97%, respectively, and 2-year local control rates were 100%, 75%, and 88%, respectively. Positive predictive values to predict ambulatory and non-ambulatory status were 97% and 90% using the new score, and 98% and 79% using the previous instrument. The new score appeared more precise in predicting non-ambulatory status. Since patients with 32–37 points had high post-radiotherapy ambulatory and local control rates, they may not require surgery.

Estimation of euphotic zone depth in shallow inland water using inherent optical properties and multispectral remote sensing imagery

Euphotic zone depth (Zeu) is a key indicator of underwater light environment that regulates different biogeochemical and physical processes in aquatic ecosystem. Remote sensing approaches offer a unique opportunity to frequently monitor Zeu, although its estimation from diffuse attenuation coefficient (Kd) requires depth profiles of the down-welling irradiance (Ed). Measuring Ed values at different water depths (i.e., Ed profiling) in shallow waterbodies is a challenging task with the available profiling equipment. In this study, Kd values were estimated from measured absorption coefficient (a) data at multiple locations in the Chilika lagoon, which is a shallow waterbody located in the eastern coast of India and is a prestigious Ramsar site. Specifically, remote sensing reflectance (Rrs), a values (wavelength: 400–715 nm at 4 nm interval), Secchi disc depth (ZSD), suspended particulate matter (SPM) and chlorophyll-a (Chl-a) contents were measured during Oct. 07–08, 2018 and Feb. 02–08, 2019 in the Chilika lagoon. Atmospherically corrected Sentinel 2A/2B (S2A/S2B) MSI- Rrs data (level 2A) were then used to develop algorithms for mapping Zeu for the entire lagoon. Modeling results showed that Rrs values at band 4 (665 nm) or band 5 (704 nm) extracted from the S2A/S2B MSI data may be used to estimate Kd of photosynthetically active radiation (PAR) (observed range: 1.29 m−1 to 10.74 m−1) with the root-mean-squared error (RMSE) as low as 0.81 m−1. However, uncertainties associated with atmospheric correction of S2A/S2B MSI-single band Rrs values are high when compared with in situ Rrs. Thus, a multivariate modeling approach was adopted with Rrs- band ratio and band difference variables selected by stepwise, forward, and backward variable selection method with p < 0.05. Modeling results showed that the band differences between band 2 and band 4 and band 3 and band 5 may be used to estimate Kd(PAR) with improved mean absolute percentage difference (MAPD) while ensuring low uncertainty associated with atmospheric correction. Spatio-temporal dynamics from S2A/S2B MSI for 85 different dates during Jan. 2017 to June 2021 provided seasonal patterns for Zeu in the Chilika lagoon.

Factors associated with death places among elderly patients receiving home-based care.

The quality of end-of-life (EOL) care for patients receiving home-based care is a critical issue for health care providers. Dying in a preferred place is recognized as a key EOL care quality indicator. We explore the factors associated with death at home or nursing facilities among elderly patients receiving home-based care.This retrospective study was based on a medical chart review between January 2018 and December 2019 of elderly patients. Multivariate analysis was conducted by fitting multiple logistic regression models with the stepwise variable selection procedure to explore the associated factors.The 205 elderly patients receiving home-based care were enrolled for analysis. The mean participant age was 84.2 ± 7.8 years. Multiple logistic regression indicated that significant factors for elderly home-based patients who died at home or nursing facilities were receiving palliative service (odds ratio [OR], 3.21; 95% confidence interval [CI], 1.37–7.51; P = .007), symptoms of nausea or vomiting (OR, 5.38; 95% CI, 1.12–25.84; P = .036), fewer emergency department visits (OR, 0.07; 95% CI, 0.03–0.16; P < .001), and less intravenous third-generation cephalosporin use (OR, 0.15; 95% CI, 0.03–0.75; P = .021) in the last month of life. Patients with dementia had a lower probability of dying at home or nursing facilities than patients with other diagnosis (OR, 0.34, 95% CI, 0.13–0.90; P = .030).Among elderly home-based patients, receiving palliative service, with nausea or vomiting, and fewer emergency department visits in the last month of life favored home or nursing facilities deaths. Practitioners should be aware of the factors with higher probabilities of dying at home and in nursing facilities. We suggested that palliative services need to be further developed and extended to ensure that patients with dementia can receive adequate EOL care at home and in nursing facilities.

A novel risk score to identify the need for triple antithrombotic therapy in patients with atrial fibrillation undergoing percutaneous coronary intervention: a post hoc analysis of the RE-DUAL PCI trial.

Current guidelines recommend treating atrial fibrillation (AF) patients who undergo percutaneous coronary intervention (PCI) with triple antithrombotic therapy (TAT) for up to one month in patients at high thrombotic risk. It is unclear how to select these high-risk patients. The aim of this study was to identify patients at high thrombotic risk who might benefit from TAT over double antithrombotic therapy (DAT). This study was a post hoc subanalysis of the RE-DUAL PCI trial. A Cox proportional hazards model was built by stepwise selection of plausible predictor variables for a composite ischaemic endpoint, defined as cardiovascular death, myocardial infarction (MI), stent thrombosis (ST) or ischaemic stroke. The effect of TAT versus DAT was calculated for those patients with the highest proportion of predicted thrombotic risk. A simplified risk score was constructed based on beta-coefficients. For 209 patients (7.7%) the composite ischaemic endpoint occurred during the first year. The simplified risk score contained six variables. In patients with a score ≥5 (n=154, 5.7%), a significant reduction in the composite of MI and ST was observed with TAT versus DAT (6.3% vs 21.0%, p=0.041), without a penalty in terms of bleeding. In patients at low thrombotic risk, a significant increase in bleeding was observed without a reduction of ischaemic events. Our findings support the use of DAT in the majority of patients. A small subgroup of patients might benefit from TAT and we propose a novel clinical risk score to select these patients.

Predictors of antiplatelet cessation in a real-world patient population undergoing non-cardiac surgery after PCI

BackgroundThe optimal perioperative management of antiplatelet therapy (APT) therapy in patients undergoing noncardiac surgery (NCS) following percutaneous coronary intervention (PCI) is unclear. We sought to identify predictors of APT cessation in a real-world cohort of patients undergoing NCS within 1 year of PCI. MethodsConsecutive patients undergoing PCI at a tertiary center between 2011 and 2018 were prospectively enrolled. Perioperative interruption of APT was defined as cessation of either aspirin or P2Y12 inhibitor between 1 and 14 days prior to NCS. Predictors of APT discontinuation were identified by multivariable Cox regression with stepwise selection of candidate variables. ResultsA total of 1092 surgeries corresponding to 747 patients were identified. Overall, there were 487 (44.6%) preoperative antiplatelet interruptions: discontinuation of either P2Y12 inhibitors only (47.4%), aspirin only (7.9%), or both agents (44.7%). Both patient-specific risk factors (prior stroke, lower BMI, anemia, MI) and procedure specific risk factors (chronic total occlusions, multivessel disease, drug-eluting stent use) affected decisions regarding APT cessation. Likelihood of APT cessation increased in higher-risk surgeries and in patients on more potent P2Y12 inhibitors (ticagrelor/prasugrel vs clopidogrel) whereas those undergoing NCS <90 days post PCI were less likely to have cessation of APT. ConclusionIn this contemporary cohort of post-PCI patients undergoing NCS, patient-, angiographic- and surgery-specific factors all affected decision-making regarding APT cessation. Our findings reflective of real-world practice, highlight the importance of a multidisciplinary team approach to individualize decision making in these patients.

Associations between healthcare utilization and access and diabetic retinopathy complications using All of Us nationwide survey data.

PurposeInadequacies in healthcare access and utilization substantially impact outcomes for diabetic patients. The All of Us database offers extensive survey data pertaining to social determinants that is not routinely available in electronic health records. This study assesses whether social determinants were associated with an increased risk of developing proliferative diabetic retinopathy or related complications (e.g. related diagnoses or procedures).MethodsWe identified 729 adult participants in the National Institutes of Health All of Us Research Program data repository with diabetic retinopathy (DR) who answered survey questions pertaining to healthcare access and utilization. Electronic health record data regarding co-morbidities, laboratory values, and procedures were extracted. Multivariable logistic regression with bi-directional stepwise variable selection was performed from a wide range of predictors. Statistical significance was defined as p<0.05.ResultsThe mean (standard deviation) age of our cohort was 64.9 (11.4) years. 15.2% identified as Hispanic or Latino, 20.4% identified as Black, 60.6% identified as White, 2.74% identified as Asian, and 16.6% identified as Other. 10–20% of patients endorsed several reasons for avoiding or delaying care, including financial concerns and lack of access to transportation. Additional significant social determinants included race and religion discordance between healthcare provider and patient (odds ratio [OR] 1.20, 95% confidence interval [CI] 1.02–1.41, p = 0.03) and caregiver responsibilities toward others (OR 3.14, 95% CI 1.01–9.50, p = 0.04).ConclusionsNationwide data demonstrate substantial barriers to healthcare access among DR patients. In addition to financial and social determinants, race and religion discordance between providers and patients may increase the likelihood of PDR and related complications.

The association between mediators of the receptor for advanced glycation end product (RAGE) axis and immune checkpoint inhibitor (ICI)–induced colitis in patients with melanoma.

9587 Background: Colitis and other gastrointestinal (GI) toxicity are a frequent and occasionally severe form of immune-related adverse events (irAEs) in patients treated with ICIs. To date, no definitive mechanism has been identified, and this area remains an active field of investigation. We hypothesized that activation of the RAGE axis, known to be implicated in inflammatory bowel disease through stimulation of signal transduction targeted by pro-inflammatory RAGE ligands, members of the S100 family and High Mobility Group Box 1 (HMGB1), might be associated with irAE- colitis. Methods: We examined sera from 111 advanced melanoma patients prospectively accrued and followed up at NYULH (treated with anti-PD-L1, n = 44; antiCTL4, n = 23; and combination, n = 44). 24 (22%) developed GI toxicity grade &gt;2. Serum biomarkers of the ligand-RAGE pathways, soluble (s)RAGE, endogenous secretory (es)RAGE, S100B, and HMGB1, were measured in the patients’ sera during ICI treatment. Multivariable ordinal logistic regression analyses with all grades of GI toxicity as the primary outcome for all the recorded covariates (including serum biomarkers, clinical covariates) were performed. We then used ordinal multivariable logistic regression with stepwise variable selection. Similar analyses with GI toxicity as a binary outcome ((≥grade 1 vs no toxicity) were also conducted. Only those variables that jointly contributed to the odds of developing toxicity were included in the final stepwise model. No adjustments for multiplicity were included. As sRAGE and esRAGE are highly correlated (r = 0.86), esRAGE concentrations were not used in the joint models. Results: A significant association between GI toxicity and concentrations of sRAGE and S100B was identified. The final stepwise multivariable logistic model includes only sRAGE and S100B. The odds of having a one level increase in GI toxicity grade increase 1.100 times (95% CI: 1.008, 1.199; p = 0.029) for each unit decrease of sRAGE ( = sRAGE/100). The odds of a one level increase in GI toxicity increase 1.059 times (95% CI: 1.004, 1.116; p = 0.035) for each unit increase of S100B ( = s100B/100). All other analyses yielded comparable results. In contrast, concentrations of HMGB1 and other clinical covariates, including response and treatment category, were not associated with GI toxicity. Conclusions: Mediators of the RAGE axis, specifically sRAGE and S100B, might have a role in GI toxicity in patients receiving ICIs. The ligand-RAGE axis may be a novel target for irAE therapies for patients receiving ICIs to mitigate the severity of GI toxicity.

Prospective identification of prognostic factors for patients with early failure of advanced ovarian cancer who undergo primary cytoreductive surgery followed by chemotherapy: The AGO-OVAR 19/FRAGILE study (NCT02828618).

5556 Background: Standard treatment for advanced ovarian cancer (aOC) includes primary cytoreductive surgery followed by chemotherapy (PDS &gt; CTX). There is strong evidence that this strategy is accompanied by high risk for early failure in some patients (pts) but prospective multicenter data for this specific question are lacking. Methods: 64 AGO-sites prospectively registered pts with suspected aOC and collected following variables prior to start of therapy: age, Charlson Comorbidity Index (CCI), timed up and go test (TUG), ASA and ECOG performance score, weight, height, estimated ascites, albumin, creatinine, hemoglobin, leucocytes, platelets, CA125, patient reported outcome measures according to EORTC QLQ-C30 and OV28, hospital anxiety and depression score (HADS), physician-assessed suspected FIGO IV stage, abdominal pain requiring treatment, abdominal bloating, dyspnea and required palliative paracentesis (ascites, pleural effusions). Treatment followed according to investigator’s decision. Primary objective was to predict the unfavorable event of death or progression within 10 months after primary diagnosis. We applied univariate and multiple logistic regression with stepwise variable selection after multiple imputation of missing data in order to identify relevant predictors. Results: 223 pts with aOC FIGO IIIB to IVB and PDS were analyzed. Median age was 63 years (range 31-86). 52 (23.3%) pts experienced progression or death within 10 months from time of enrolment. In univariate regression, age (odds ratio (OR) 1.612 per 10 years), ASA (III vs I/II: OR 3.217), TUG (OR 1.087), body height (OR 0.541 per 10cm), ECOG (1 vs 0 OR 2.326, 2-3 vs 0 OR 4.102), estimated ascites ( &gt; 500cc vs. none OR 2.811), paracentesis (OR 1.991), platelets (upper limit of normal (ULN) vs norm or &lt; lower limit of normal (LLN) OR 1.998), albumin ( &lt; LLN vs norm or &gt; ULN OR 2.053), creatinine ( &gt; ULN vs norm OR 3.969) and baseline QoL single-item subscales appetite loss (“very much” vs “not at all” OR 2.611), constipation (“quite a bit” vs “not at all” OR 4.903), and multi-item subscales global health status (OR 0.982 per 1 point) and nausea/vomiting (OR 1.013 per 1 point) were significant (p &lt; 0.05). Multiple logistic regression identified age (OR 1.459 per 10 years), ASA (III vs I/II: OR 2.427), constipation (“quite a bit” vs “not at all” OR 3.786) and global health (OR: 0.985 per 1 point) as independent predictors of progression or death within 10 months after primary diagnosis (p &lt; 0.05). Conclusions: A significant proportion of aOC pts undergoing PDS &gt; CTX are at high-risk for early failure. The finding of independent risk factors including self-ratings of QoL at time of diagnosis should be confirmed and tested against other models to facilitate a more tailored treatment of high-risk pts and design of future trials in aOC. Clinical trial information: NCT02828618.

Adjuvant abemaciclib for high-risk early breast cancer (EBC): Factors increasing the rate of treatment discontinuations in monarchE.

527 Background: Adjuvant abemaciclib plus endocrine therapy (ET) demonstrated clinically meaningful improvement in reducing the risk of recurrence in patients (pts) with HR+, HER2– high risk EBC. In monarchE, 25.6% pts discontinued abemaciclib before completing the 2-year treatment period due to reasons other than recurrence. This exploratory analysis evaluates the impact of baseline factors on time to discontinuation for abemaciclib-treated patients with an aim to identify pts at higher risk of discontinuation. Methods: Time to discontinuation(TTD) was defined as time from first dose to the discontinuation of abemaciclib or all treatment due to reasons other than recurrence. These exploratory analyses evaluated the association between TTD and baseline demographics, disease characteristics, and the extent of pre-existing medical comorbidities. Factors with p-value &lt;0.05 in the univariate Cox model were considered as potentially associated with TTD and then fitted into a multivariate Cox model with stepwise variable selection, with entry and retaining p-value threshold of 0.05. Discontinuation rates at selected timepoints (6-month, 12-month, 18-month, 24-month) within each subgroup were estimated using the Kaplan-Meier method. Results: The following variables were significantly associated with higher risk of discontinuation in the multivariate model: Age ≥65 yr, enrolled in North America or EU, ECOG PS=1, post-menopausal, 1-3 positive nodes, or 4 or more pre-existing comorbidities. The majority of discontinuations occurred within the first 6 months of the treatment period; thus, this timepoint was selected to estimate discontinuation rates for factors independently associated with TTD (Table). Complete results on discontinuation rates at other timepoints will be reported in the presentation. Conclusions: This exploratory analysis identified several factors associated with a greater risk of discontinuation. These results illustrate the importance of close monitoring and dose adjustments early on, particularly for patients with an increased risk of discontinuation prior to completing the 2-yr treatment period of abemaciclib. Clinical trial information: NCT03155997. [Table: see text]

An alternative approach for estimating large-area indoor PM2.5 concentration – A case study of schools

BackgroundFor indoor air modelling, difficulties in collecting indoor parameters including life activity patterns and building characteristics are dilemmas when conducting a large-area study. Land-use/land cover information which is easier to obtain could represent as surrogates of emission sources for assessing indoor air quality. Moreover, low-cost sensors and machine learning provide a better way to enhance model accuracy. ObjectivesThis study proposed an alternative estimation approach to assess daily PM2.5 concentration for indoor environments of schools in a large area by integrating low-cost sensors, land-use/land cover predictors, and machine learning-based modelling approaches. MethodsIndoor PM2.5 data was collected from 145 indoor AirBox sensors in Kaohsiung and Pingtung Counties of Taiwan. Geospatial predictors were extracted from the circular buffers surrounding each AirBox sensor. Spearman correlation analysis and stepwise variable selection procedures were performed to select variables for land-use regression (LUR) and integrated with XGBoost, Random Forest (RF), and LGBM machine learning models. ResultsThe results revealed that outdoor PM2.5 and distance to the nearest thermal power plant were the main determinants of indoor estimation variations, when there were no indoor sources. When incorporating machine learning, the R2 increased from 0.59 for LUR to 0.85 for LUR-XGBoost while the RMSE decreased from 8.63 to 5.27 μg/m3, which performed better than both LUR-RF and LUR-LGBM. ConclusionsThis study demonstrates the value of the proposed alternative approach by incorporating data from a low-cost sensor with LUR model and machine learning algorithm in estimating the spatiotemporal variability of indoor PM2.5 for a large area.

Factors Affecting Readmission to a Drug Rehabilitation Center in Brunei Darussalam

The objective of this study is to examine the sociodemographic factors contributing to the readmission to a drug rehabilitation center among drug addicts in Brunei Darussalam. This retrospective study used de-identified data obtained from the Al-Islah Rehabilitation Center for the period 1 January 2010 to 31 December 2017. Univariate and multiple logistic regression analyses with a stepwise variable selection method were used to estimate and determine significant factors associated with readmission to the center. A total of 92 out of 705 subjects (13%) were readmitted to the rehabilitation center within the study period, with ages ranging from 20 to 59 years (M = 36.1, SD = 8.6 years). Multiple logistic regression found that subjects who were in the age group of 40 to 49 (OR = 2.24, 95% CI [1.16, 4.37]) resided in the Belait district (OR = 1.94, 95% CI [1.00, 3.63]), divorced or widowed (OR = 2.14, 95% CI [1.19, 3.82]), had committed other criminal offenses (OR = 1.91, 95% CI [1.16, 3.11]), and voluntarily admitted (OR = 2.33, 95% CI [1.34, 4.01]), were found to be significantly more likely to get readmitted to the rehabilitation center. This study provides evidence of sociodemographic factors in relation to readmission to the rehabilitation center in Brunei Darussalam.