Ordinary Least-squares Linear Regression Research Articles

Comparison of Machine Learning Models in Simulating Glacier Mass Balance: Insights from Maritime and Continental Glaciers in High Mountain Asia

Accurately simulating glacier mass balance (GMB) data is crucial for assessing the impacts of climate change on glacier dynamics. Since physical models often face challenges in comprehensively accounting for factors influencing glacial melt and uncertainties in inputs, machine learning (ML) offers a viable alternative due to its robust flexibility and nonlinear fitting capability. However, the effectiveness of ML in modeling GMB data across diverse glacier types within High Mountain Asia has not yet been thoroughly explored. This study addresses this research gap by evaluating ML models used for the simulation of annual glacier-wide GMB data, with a specific focus on comparing maritime glaciers in the Niyang River basin and continental glaciers in the Manas River basin. For this purpose, meteorological predictive factors derived from monthly ERA5-Land datasets, and topographical predictive factors obtained from the Randolph Glacier Inventory, along with target GMB data rooted in geodetic mass balance observations, were employed to drive four selective ML models: the random forest model, the gradient boosting decision tree (GBDT) model, the deep neural network model, and the ordinary least-square linear regression model. The results highlighted that ML models generally exhibit superior performance in the simulation of GMB data for continental glaciers compared to maritime ones. Moreover, among the four ML models, the GBDT model was found to consistently exhibit superior performance with coefficient of determination (R2) values of 0.72 and 0.67 and root mean squared error (RMSE) values of 0.21 m w.e. and 0.30 m w.e. for glaciers within Manas and Niyang river basins, respectively. Furthermore, this study reveals that topographical and climatic factors differentially influence GMB simulations in maritime and continental glaciers, providing key insights into glacier dynamics in response to climate change. In summary, ML, particularly the GBDT model, demonstrates significant potential in GMB simulation. Moreover, the application of ML can enhance the accuracy of GMB modeling, providing a promising approach to assess the impacts of climate change on glacier dynamics.

Self-identified barriers to health services among migrants 50 years of age or older: population-based survey study of Russian speakers in Finland

BackgroundThe compounded effect of a migratory background and ageing increases the risk of unequal medical treatment opportunities. The aim of this article is to investigate the social determinants of barriers to health services.MethodsThe study uses population-based survey data of Russian-speaking migrants (50 + years) residing in Finland (n = 1082, 57% of men, mean age 63 years). Multiple correspondence analysis was performed as a dimension reduction procedure on six barriers to health services. Multiple ordinary least-squares linear regression was used for the predicted score of the barriers as an outcome variable.ResultsMost of the sociodemographic characteristics were not associated with barriers to health services, except gender, as women tended to face more disadvantages. Migration-related factors, such as the need for interpreters for health services and experienced discrimination, were associated with an increased likelihood of reporting barriers to health services. Using the internet as a primary source of health information was associated with more access barriers to health services.ConclusionsMigrants 50 years of age or older face multiple barriers to health services. Given that the healthcare needs increase with age, addressing this issue becomes crucial, necessitating improved access to health services for older migrants.

Lack of structural brain alterations associated with insomnia: findings from the ENIGMA-Sleep Working Group.

Existing neuroimaging studies have reported divergent structural alterations in insomnia disorder (ID). In the present study, we performed a large-scale coordinated meta-analysis by pooling structural brain measures from 1085 subjects (mean [SD] age 50.5[13.9]years, 50.2% female, 17.4% with insomnia) across three international Enhancing NeuroImaging Genetics through Meta-Analysis (ENIGMA)-Sleep cohorts. Two sites recruited patients with ID/controls: Freiburg (University of Freiburg Medical Center, Freiburg, Germany) 42/43 and KUMS (Kermanshah University of Medical Sciences, Kermanshah, Iran) 42/49, while the Study of Health in Pomerania (SHIP-Trend,University MedicineGreifswald,Greifswald,Germany) recruited population-based individuals with/without insomnia symptoms 75/662. The influence of insomnia on magnetic resonance imaging-based brain morphometry using an insomnia brain score was then assessed. Within each cohort, we used an ordinary least-squares linear regression to investigate the link between the individual regional cortical and subcortical volumes and the presence of insomnia symptoms. Then, we performed a fixed-effects meta-analysis across cohorts based on the first-level results. For the insomnia brain score, weighted logistic ridge regression was performed on one sample (Freiburg), which separated patients with ID from controls to train a model based on the segmentation measurements. Afterward, the insomnia brain scores were validated using the other two samples. The model was used to predict the log-odds of the subjects with insomnia given individual insomnia-related brain atrophy. After adjusting for multiple comparisons, we did not detect any significant associations between insomnia symptoms and cortical or subcortical volumes, nor could we identify a global insomnia-related brain atrophy pattern. Thus, we observed inconsistent brain morphology differences between individuals with and without insomnia across three independent cohorts. Further large-scale cross-sectional/longitudinal studies using both structural and functional neuroimaging are warranted to decipher the neurobiology of insomnia.

Machine Learning-based Inverse Model for Few-Mode Fiber Designs

The medium for next-generation communication is considered as fiber for fast, secure communication and switching capability. Mode division and space division multiplexing provide an excellent switching capability with high data transmission rate. In this work, the authors have approached an inverse modeling technique using regression-based machine learning to design a weakly coupled few-mode fiber for facilitating mode division multiplexing. The technique is adapted to predict the accurate profile parameters for the proposed few-mode fiber to obtain the maximum number of modes. It is for a three-ring-core few-mode fiber for guiding five, ten, fifteen, and twenty modes. Three types of regression models namely ordinary least-square linear multi-output regression, k-nearest neighbors of multi-output regression, and ID3 algorithm-based decision trees for multi-output regression are used for predicting the multiple profile parameters. It is observed that the ID3-based decision tree for multioutput regression is the robust, highly-accurate machine learning model for fast modeling of FMFs. The proposed fiber claims to be an efficient candidate for the next-generation 5G and 6G backhaul networks using mode division multiplexing.

The number of examinations required for the accurate prediction of the progression of the central 10-degree visual field test in glaucoma

The purpose of the study was to investigate the number of examinations required to precisely predict the future central 10-degree visual field (VF) test and to evaluate the effect of fitting non-linear models, including quadratic regression, exponential regression, logistic regression, and M-estimator robust regression model, for eyes with glaucoma. 180 eyes from 133 open angle glaucoma patients with a minimum of 13 Humphrey Field Analyzer 10-2 SITA standard VF tests were analyzed in this study. Using trend analysis with ordinary least squares linear regression (OLSLR), the first, second, and third future VFs were predicted in a point-wise (PW) manner using a varied number of prior VF sequences, and mean absolute errors (MAE) were calculated. The number of VFs needed to reach the minimum 95% confidence interval (CI) of the MAE of the OLSLR was investigated. We also examined the effect of applying other non-linear models. When predicting the first, second, and third future VFs using OLSLR, the minimum MAE was obtained using VF1–12 (2.15 ± 0.98 dB), VF1–11 (2.33 ± 1.10 dB), and VF1–10 (2.63 ± 1.36 dB), respectively. To reach the 95% CI of these MAEs, 10, 10, and 8 VFs were needed for the first, second and third future VF predictions, respectively. No improvement was observed by applying non-linear regression models. As a conclusion, approximately 8–10 VFs were needed to achieve an accurate prediction of PW VF sensitivity of the 10-degree central VF.

Data-driven numerical models for the prediction of the thermal resistance value of the Enclosed Airspaces (EAs) in building envelopes

ABSTRACT Enclosed airspaces (EAs) are a common component of different energy efficient technologies in building envelopes which have intrinsic adaptive behavior under various climatic conditions. The need for comprehensive accurate numerical models without restrictions for different applications still exists. Thus, for the first time, machine learning and regression-based techniques (Ordinary least squares Linear Regression (OLR), Support Vector Regression (SVR), Kernel Ridge Regression (KRR), k-Nearest Neighbors (kNN), r-Nearest Neighbors (rNN), and Artificial Neural Network (ANN)) were applied to develop precise models derived from the most credible experimental data with stochastic logic of testing and training in several runs. It was found that application of KRR, SVR, and ANN leads to the most desirable outcomes with highest R2 (0.97–0.99), and lowest errors, however OLR does not provide satisfactory results (R2 <0.6). Moreover, major deviations are observed for calculations by OLR, kNN, and rNN in horizontal EAs (thicknesses = 40mm) with downward heat flow direction.

MO323: Effects of Patient Frailty and Comorbidity on Duration of Hospitalisation as an Outcome Among Adults With Acute Kidney Injury

Abstract BACKGROUND AND AIMS Acute kidney injury (AKI) is commonly associated with an adverse outcome in hospitalised patients. Frailty and comorbidity are risk factors for acute kidney injury. The aim of this study was to assess the strength of association between frailty and comorbidity as modifiers for duration of hospitalisation (LOS) as an outcome among inpatients with acute kidney injury in our institution. METHOD A retrospective observational study as part of a service assessment of adult inpatients, in a 1-month period, during the pre-pandemic phase was conducted to evaluate for changes needed to existing care pathways in our institution. AKI was identified with the help of the national algorithm endorsed by NHS England, which is incorporated in hospital reporting systems. Clinical frailty and comorbidity were estimated using the Rockwood Clinical Frailty Scale (CFS) and the Charlson comorbidity index (CCI). Pairwise correlations between LOS, CFS and CCI were estimated and tested for statistical significance using Bonferroni-adjusted significance levels. Ordinary least-squares linear regression was used to assess the prediction of LOS using CFS and CCI in separate models. RESULTS In 148 patients in our cohort, the mean age was 76.4 [95% confidence interval (95% CI) 74.1–78.8] years. Of these, 54% were male, 24.3% were diabetic, 29.7% had at least one criterion for prior vascular disease and 16.2% had a history of cancer. A total of 22.3% had a prior history of chronic kidney disease. In terms of severity at presentation, 67.6% had stage 1 AKI, 18.9% had stage 2, and 13.5% had stage 3 AKI. The mean CCI was 5.4 (95% CI 5.06–5.81) and the mean RFI was 4.6 (95% CI 4.3–5.0). 50% were both at least moderately frail (CFS ≥ 5) and had a CCI of &amp;gt;5. The mean duration of hospitalisation was 8.7 days. Spearman's correlation coefficient constant for CFS with LOS was 0.23 and was 0.11 with CCI, with only the correlation between LOS and CFS being statistically significant at the 5% level after Bonferroni correction. As expected, there was a statistically significant correlation between CCI and CFS (0.49). Beta coefficients for ordinary least-squares linear regression individual models with LOS as the outcome variable were 0.69 (95% CI 0.18–1.19; P &amp;lt; .007) for CFS and 0.27 (95% CI –0.16 to 0.70; P = .22) for CCI. CONCLUSION Compared with patient comorbidity, clinical frailty has a stronger and statistically significant association with the duration of hospitalisation among hospitalised adults with AKI. This study quantifies the magnitude of factors over and above those resulting from patient comorbidity that are contributory to frailty in adult inpatients with AKI. Further studies are required to improve understanding of how clinical frailty affects duration of hospitalisation in patients with AKI, to help inform future clinical care pathways and improve duration of hospitalisation as a clinical outcome.

The Impact of COVID-19 on Mortality in Italy: Retrospective Analysis of Epidemiological Trends.

BackgroundDespite the available evidence on its severity, COVID-19 has often been compared with seasonal flu by some conspirators and even scientists. Various public discussions arose about the noncausal correlation between COVID-19 and the observed deaths during the pandemic period in Italy.ObjectiveThis paper aimed to search for endogenous reasons for the mortality increase recorded in Italy during 2020 to test this controversial hypothesis. Furthermore, we provide a framework for epidemiological analyses of time series.MethodsWe analyzed deaths by age, sex, region, and cause of death in Italy from 2011 to 2019. Ordinary least squares (OLS) linear regression analyses and autoregressive integrated moving average (ARIMA) were used to predict the best value for 2020. A Grubbs 1-sided test was used to assess the significance of the difference between predicted and observed 2020 deaths/mortality. Finally, a 1-sample t test was used to compare the population of regional excess deaths to a null mean. The relationship between mortality and predictive variables was assessed using OLS multiple regression models. Since there is no uniform opinion on multicomparison adjustment and false negatives imply great epidemiological risk, the less-conservative Siegel approach and more-conservative Holm-Bonferroni approach were employed. By doing so, we provided the reader with the means to carry out an independent analysis.ResultsBoth ARIMA and OLS linear regression models predicted the number of deaths in Italy during 2020 to be between 640,000 and 660,000 (range of 95% CIs: 620,000-695,000) against the observed value of above 750,000. We found strong evidence supporting that the death increase in all regions (average excess=12.2%) was not due to chance (t21=7.2; adjusted P<.001). Male and female national mortality excesses were 18.4% (P<.001; adjusted P=.006) and 14.1% (P=.005; adjusted P=.12), respectively. However, we found limited significance when comparing male and female mortality residuals’ using the Mann-Whitney U test (P=.27; adjusted P=.99). Finally, mortality was strongly and positively correlated with latitude (R=0.82; adjusted P<.001). In this regard, the significance of the mortality increases during 2020 varied greatly from region to region. Lombardy recorded the highest mortality increase (38% for men, adjusted P<.001; 31% for women, P<.001; adjusted P=.006).ConclusionsOur findings support the absence of historical endogenous reasons capable of justifying the mortality increase observed in Italy during 2020. Together with the current knowledge on SARS-CoV-2, these results provide decisive evidence on the devastating impact of COVID-19. We suggest that this research be leveraged by government, health, and information authorities to furnish proof against conspiracy hypotheses that minimize COVID-19–related risks. Finally, given the marked concordance between ARIMA and OLS regression, we suggest that these models be exploited for public health surveillance. Specifically, meaningful information can be deduced by comparing predicted and observed epidemiological trends.

Why do patients with chronic hepatitis C drink alcohol? An examination of pain, depression and drinking motives.

Alcohol consumption in the setting of chronic HCV is associated with accelerated progression towards cirrhosis, increased risk of hepatocellular carcinoma and higher mortality. This analysis contextualizes how sociodemographic factors, chronic pain and depression relate to the motivations of individuals with chronic HCV to consume alcohol. We conducted a secondary analysis of baseline data from the Hep ART trial of behavioural interventions on alcohol use among patients with HCV. Alcohol consumption was measured using the Drinking Motives Questionnaire and a novel 6-item measure of pain-related drinking motives. Statistical analyses performed included ANOVA for bivariate analyses and multivariable ordinary least-squares linear regression. At study baseline, 181 participants had an average age of 55years; the majority (66.7%) reported beyond-minor pain; and a third (37%) met criteria for depression; drinking motives were higher for individuals with beyond-minor pain (means 9.9 vs. 4.6, p<.001) and who met criteria for depression (means 10.9 vs. 6.4, p<.001) when using the pain-related drinking motives items. Average pain(coef=1.0410067141<.001) was significantly associated with increased motives to drink to relieve pain in the full baseline model specification controlling for all covariates using ordinary at least squares; depression (coef=7.06; 95% CI 1.32, 12.81; p=.016) was significantly associated with increased non-pain-related motives to drink. From baseline to 3-month follow-up, compared to participants who had mean average pain scores among the sample, motives to drink to relieve pain decreased in participants who had higher average pain scores (coef=-0.30; 95% CI -0.59, -0.01; p=.40). Physical pain and depression are associated with increased motives to consume alcohol. Patients with chronic liver disease should be screened for chronic pain and depression and, if present, referred to pain specialists or co-managed in partnership with pain specialists in hepatology clinics.

Modelling heating and cooling energy demand for building stock using a hybrid approach

The building sector accounts for 30% of final energy consumption and 28% of global energy-related carbon dioxide emissions, with space heating and cooling consuming a large share of total buildings’ energy consumption. Building stock modelling for space heating and cooling energy prediction provides critical insights on the stock energy consumption and aid the building retrofit policy-making process with the evaluation of the energy-saving potential. By combining the physical modelling approach and data-driven approach, a hybrid approach is applicable for modelling the heating and cooling energy consumption of the building stock, including both residential buildings and non-residential buildings. Within this framework, the Urban Modelling Interface (UMI) tool has been used for physical modelling to generate heating and cooling energy use intensity. Then, ten different machine learning models, including Gaussian radial basis function kernel support vector regression, linear kernel support vector regression, polynomial kernel support vector regression, random forests, extreme gradient boosting, ordinary least-squares linear regression, ridge regression, least absolute shrinkage and selection operator, elastic net and artificial neural network, have been applied to predict heating and cooling energy use intensity (EUI). The approach has been demonstrated using a case study in Chongqing, China. The results show that machine learning models can achieve accurate building heating and cooling EUI prediction, with the polynomial kernel support vector regression showing the best accuracy at the level of a single building, and the Gaussian radial basis function kernel support vector regression performing the best at the stock level. Machine learning models generated by proposed hybrid approach not only provide quickly prediction of building space heating and cooling energy consumption at the stock level, but also support building retrofit decision makings by evaluate energy saving potential of various retrofit options.

Evaluation of the external validity of a joint structure\u2013function model for monitoring glaucoma progression

The dynamic structure–function (DSF) model was previously shown to have better prediction accuracy than ordinary least square linear regression (OLSLR) for short series of visits. The current study assessed the external validity of the DSF model by testing its performance in an independent dataset (Ocular Hypertension Treatment Study–Confocal Scanning Laser Ophthalmoscopy [OHTS–CSLO] ancillary study; N = 178 eyes), and also on different test parameters in a sample selected from the Diagnostic Innovations in Glaucoma Study or the African Descent and Glaucoma Evaluation Study (DIGS/ADAGES). Each model was used to predict structure–function paired data at visits 4–7. The resulting prediction errors for both models were compared using the Wilcoxon signed-rank test. In the independent dataset, the DSF model predicted rim area and mean sensitivity paired measurements more accurately than OLSLR by 1.8–5.5% (p ≤ 0.004) from visits 4–6. Using the DIGS/ADAGES dataset, the DSF model predicted retinal nerve fiber layer thickness and mean deviation paired measurements more accurately than OLSLR by 1.2–2.5% (p ≤ 0. 007). These results demonstrate the external validity of the DSF model and provide a strong basis to develop it into a useful clinical tool.

Comparing the Rule of 5 to Trend-based Analysis for Detecting Glaucoma Progression on OCT

The rule of 5 is a simple rule for detecting retinal nerve fiber layer (RNFL) change on spectral-domain OCT (SD-OCT), in which a loss of 5 μm of global RNFL on a follow-up test is considered evidence of significant change when compared with the baseline. The rule is based on short-term test-retest variability of SD-OCT and is often used in clinical practice. The purpose of this study was to compare the rule of 5 with trend-based analysis of global RNFL thickness over time for detecting glaucomatous progression. Prospective cohort. A total of 300 eyes of 210 glaucoma subjects followed for an average of 5.4±1.5 years with a median of 11 (interquartile range, 7-14) visits. Trend-based analysis was performed by ordinary least-squares (OLS) linear regression of global RNFL thickness over time. For estimation of specificity, false-positives were obtained by assessing for progression on series of randomly permutated follow-up visits for each eye, which removes any systematic trend over time. The specificity of trend-based analysis was matched to that of the rule of 5 to allow meaningful comparison of the "hit rate," or the proportion of glaucoma eyes categorized as progressing at each time point, using the original sequence of visits. Comparison between hit rates of trend-analysis versus rule of 5 at matched specificity. After 5 years, the simple rule of 5 identified 37.5% of eyes as progressing at a specificity of 81.1%. At the same specificity, the hit rate for trend-based analysis was significantly greater than that of the rule of 5 (62.9% vs. 37.5%; P < 0.001). If the rule of 5 was required to be repeatable on a consecutive test, specificity improved to 93.4%, but hit rate decreased to 21.0%. At this higher specificity, trend-based analysis still had a significantly greater hit rate than the rule of 5 (47.4% vs. 21.0%, respectively; P < 0.001). Trend-based analysis was superior to the simple rule of 5 for identifying progression in glaucoma eyes and should be preferred as a method for longitudinal assessment of global SD-OCT RNFL change over time.

An ultrasound-guided fascia iliaca catheter technique does not impair ambulatory ability within a clinical pathway for total hip arthroplasty

BACKGROUND Both neuraxial and peripheral regional analgesic techniques offer postoperative analgesia for total hip arthroplasty (THA) patients. While no single technique is preferred, quadriceps muscle weakness from peripheral nerve blocks may impede rehabilitation. We designed this study to compare postoperative ambulation outcome in THA patients who were treated with a new ultrasound-guided fascia iliaca catheter (FIC) technique or intrathecal morphine (ITM). METHODS We reviewed the electronic health records of a sequential series of primary unilateral THA patients who were part of a standardized clinical pathway; apart from differences in regional analgesic technique, all other aspects of the pathway were the same. Our primary outcome was total ambulation distance (meters) combined for postoperative days 1 and 2. Secondary outcomes included daily opioid consumption (morphine milligram equivalents) and analgesic-related side effects. We examined the association between the primary outcome and analgesic technique by performing crude and adjusted ordinary least-squares linear regression. A P value < 0.05 was considered statistically-significant. RESULTS The study analyzed the records of 179 patients (fascia iliaca, n = 106; intrathecal, n = 73). The primary outcome (total ambulation distance) did not differ between the groups (P = 0.08). Body mass index (BMI) was the only factor (β = -1.7 [95% CI -0.5 to -2.9], P < 0.01) associated with ambulation distance. Opioid consumption did not differ, while increased pruritus was seen in the intrathecal group (P < 0.01). CONCLUSIONS BMI affects postoperative ambulation outcome after hip arthroplasty, whereas the type of regional analgesic technique used does not. An ultrasound-guided FIC technique offers similar analgesia with fewer side effects when compared with ITM.

Comparison of matrix-effect corrections for ordinary and uncertainty weighted linear regressions and determination of major element mean concentrations and total uncertainties of sixty-two international geochemical reference materials from wavelength-dispersive X-ray fluorescence spectrometry

In X-ray fluorescence (XRF) spectrometry, matrix-effect corrections for two types of linear regressions (ordinary and uncertainty-based weighted) have been performed for the first time and compared for the determination of major elements in rocks and minerals. The analytical data from different laboratories for 62 international geochemical reference materials (GRMs) were first processed to obtain better estimates of the mean values, along with lower uncertainties (narrower confidence limits of the mean), which were used in the XRF calibrations. The blank intensities were subtracted from the response for each GRM and the respective uncertainties were estimated. The uncertainty-based weighted least-squares linear regression (UWLR) model, being statistically more appropriate than the ordinary least-squares linear regression (OLR) model, provided more reliable regression equations than the OLR. All calculations and matrix-effect corrections were achieved through a newly developed online computer program. Similarly, matrix-effect corrections involving 11 α and 26 α (alphas) are presented and compared. The best UWLR method involving 26 α was successfully applied to the 62 GRMs and 4 similarly complex rock matrices as the calibrators. Both evaluations confirmed the usefulness of the UWLR model. This UWLR model was also compared with the OLR 26 α model for 62 GRMs treated as unknowns and shown to perform better. Thus, the UWLR model, along with the proposed matrix-effect correction method of 26 α, can be recommended as the most appropriate procedure for the calibration of XRF instruments, instead of the commonly used OLR models. This is the first time when 26 α are advantageously in a better estimate of influence coefficients used for major element determinations in rock and mineral samples. Because the UWLR calibration is based on a large number of GRMs of many rock and mineral types, it should be useful for all kinds of geological materials.

Corneal Biomechanics and Visual Field Progression in Eyes with Seemingly Well-Controlled Intraocular Pressure

To investigate the incidence and risk factors for glaucomatous visual field progression in eyes with well-controlled intraocular pressure (IOP). Prospective cohort. A total of 460 eyes of 334 patients with glaucoma under treatment. Study subjects had a mean follow-up of 4.3±0.8 years. Patients were classified as well controlled if all IOP measurements were less than 18 mmHg. Rates of visual field progression were calculated using ordinary least-squares linear regression of standard automated perimetry (SAP) mean deviation (MD) values over time. Progression was defined as a significantly negative MD slope (alpha= 0.05). Rates of SAP MD change; mean and peak IOP, and IOP fluctuation; and corneal biomechanics: corneal hysteresis (CH), central corneal thickness (CCT), and corneal index. Of the 179 eyes with well-controlled IOP, 42 (23.5%) demonstrated visual field progression. There was no significant difference between progressing and stable patients in baseline MD (-6.4±7.1 decibels [dB] vs.-6.0±6.2 dB; P= 0.346), mean IOP (11.7±2.0 mmHg vs. 12.1±2.3 mmHg; P= 0.405), IOP fluctuation (1.6±0.6 mmHg vs. 1.6±0.5 mmHg; P= 0.402), or peak IOP (14.3±1.9 mmHg vs. 14.6±2.1 mmHg; P= 0.926). Progressing eyes had significantly lower CH (8.6±1.3 mmHg vs. 9.4±1.6 mmHg; P= 0.014) and thinner CCT (515.1±33.1 μm vs. 531.1±42.4 μm; P= 0.018, respectively) compared with stable eyes. In the multivariate analysis, a 1 standard deviation lower corneal index, a summation of normalized versions of CH and CCT, resulted in a 68% higher risk of progression (odds ratio, 1.68; 95% confidence interval, 1.08-2.62; P= 0.021). Approximately one-quarter of eyes with well-controlled IOP may show visual field progression over time. Thin cornea and low CH are main risk factors.

Attacking Data Transforming Learners at Training Time

While machine learning systems are known to be vulnerable to data-manipulation attacks at both training and deployment time, little is known about how to adapt attacks when the defender transforms data prior to model estimation. We consider the setting where the defender Bob first transforms the data then learns a model from the result; Alice, the attacker, perturbs Bob’s input data prior to him transforming it. We develop a general-purpose “plug and play” framework for gradient-based attacks based on matrix differentials, focusing on ordinary least-squares linear regression. This allows learning algorithms and data transformations to be paired and composed arbitrarily: attacks can be adapted through the use of the chain rule—analogous to backpropagation on neural network parameters—to compositional learning maps. Bestresponse attacks can be computed through matrix multiplications from a library of attack matrices for transformations and learners. Our treatment of linear regression extends state-ofthe-art attacks at training time, by permitting the attacker to affect both features and targets optimally and simultaneously. We explore several transformations broadly used across machine learning with a driving motivation for our work being autogressive modeling. There, Bob transforms a univariate time series into a matrix of observations and vector of target values which can then be fed into standard learners. Under this learning reduction, a perturbation from Alice to a single value of the time series affects features of several data points along with target values.

Abstract P288: The Effect of a Time Restricted Eating Approach vs. Standard Heart Healthy Dietary Counseling on Cardiometabolic Health Measures in Adults With Abdominal Obesity: A Randomized Pilot Trial

Introduction: Basic and population level research suggest that meal timing and circadian rhythms are linked with cardiometabolic health. However, interventions that aim to optimally align meal timing with circadian rhythms have an unclear clinical impact on cardiometabolic risk; a critical gap since insight into the interplay between the noted factors may have important preventive and therapeutic implications. To begin addressing this gap, we carried out a randomized dietary intervention trial comparing a time restricted eating (TRE) intervention to standard dietary counseling (STD) for cardiometabolic health as recommended by the AHA and the Academy of Nutrition and Dietetics. Hypothesis: A STD dietary intervention will differentially improve measures of clinical cardiometabolic health compared to a TRE dietary intervention. Methods: Design: Randomized parallel arm pilot trial with 1-week run in period and 8-week intervention period. Participants : 30 adults with abdominal obesity, free of major chronic disease, with interest in improving cardiometabolic health status via dietary counseling with a Registered Dietitian (RD). Intervention : All participants received 4 RD sessions, and were blinded to the nature of the other arm. TRE - RD counseling specifically on meal timing with no mention of content. Explicit directions for daily 12 hour TRE regimen representing the theoretical optimal conditions for cardiometabolic health to begin upon waking. STD- RD counseling specifically on content of dietary pattern with no mention of meal timing. Measurements: All clinical measures were collected by standardized protocol. Diet was assessed by random 24 hour recalls during the run-in (n=2), and intervention (n=3). Blinded continuous glucose monitors assessed adherence, and physical activity/sleep data were collected objectively by wearable device. Analysis: Standard intent-to-treat analyses utilizing ordinary least-squares linear regression models, adjusted for baseline measurements to compare participants across assignment groups. Results: 30 participants (26 women) were randomized, mean (SD) age 42 (13) years and 28 completed. The STD intervention reduced the triglyceride (TG):HDL ratio (2.7 to 2.5), LDL (mg/dL) (118 to 107), blood pressure (BP) (118/74 to 114/72), and increased dietary quality (HEI-2015 score, 57 to 63). The TRE intervention also reduced the (TG):HDL ratio (2.8 to 2.6), LDL (109 to 108), BP (123/78 to 119/74), and increased dietary quality (HEI-2015 score, 50 to 54). P &lt; 0.05 comparing STD to TRE for LDL. Conclusions: Both interventions improved cardiometabolic health and diet quality suggesting a TRE approach may be a useful counseling approach for clinical cardiometabolic health. The STD arm appeared to have stronger effects for LDL compared to TRE, but only a larger and longer trial will provide a definitive answer for any measure.

DXA-Rectified Appendicular Lean Mass: Development of Ultrasound Prediction Models in Older Adults.

Dual-energy X-ray absorptiometry (DXA)-derived appendicular lean soft tissue mass (aLM) is used to diagnose sarcopenia. However, DXA-derived aLM includes non-skeletal muscle components, such as fat-free component of adipose tissue fat cell. These components, if not accounted for, could falsely inflate the aLM in individuals with a high amount of adipose tissue mass. B-mode ultrasound accurately measures muscle size in older adults. We sought to develop regression-based prediction equations for estimating DXA-rectified appendicular lean tissue mass (i.e. DXA-derived aLM minus appendicular fat-free adipose tissue (aFFAT); abbreviated as aLM minus aFFAT) using B-mode ultrasound. Cross-sectional study. Three hundred and eighty-nine Japanese older adults (aged 60 to 79 years) volunteered in the study. aLM was measured using a DXA, and muscle thickness (MT) was measured using ultrasound at nine sites. An ordinary least-squares multiple linear regression model was used to predict aLM minus aFFAT from sex, age and varying muscle thicknesses multiplied by height. Based on previous studies, we chose to use 4 MT sites at the upper and lower extremities (4-site MT model) and a single site (1-site MT model) at the upper extremity to develop prediction models. The linear prediction models (4 site MT model; R2 = 0.902, adjusted R2 = 0.899, and 1-site MT model; R2 = 0.868, adjusted R2 = 0.866) were found to be stable and accurate for estimating aLM minus aFFAT. Bootstrapping (n=1000) resulted in optimism values of 0.0062 (4-site MT model) and 0.0036 (1-site MT model). The results indicated that ultrasound MT combined with height, age and sex can be used to accurately estimate aLM minus aFFAT in older Japanese adults. Newly developed ultrasound prediction equations to estimate aLM minus aFFAT may be a valuable tool in population-based studies to assess age-related rectified lean tissue mass loss.

Spring-Mediated Cranioplasty in Sagittal Synostosis: Does Age at Placement Affect Expansion?

The aim of this study is to evaluate the effect of timing of surgery and spring characteristics on correction of scaphocephalic deformity in patients undergoing spring-mediated cranioplasty (SMC) for sagittal craniosynostosis. The authors conducted a review of patients with sagittal craniosynostosis who underwent SMC at a tertiary referral center between July 2011 and March 2017, with a primary outcome measure of head shape, both preoperatively and postoperatively, determined by cephalic index (CI). Patient demographics and operative details including timing of surgery and spring characteristics were collected. Differences in CI preoperation and postoperation were compared using Wilcoxon signed-rank test. Ordinary least-squares linear regression was used to assess the impact of timing, number of springs, maximum single spring force, and total spring force on postoperative change in CI. Thirty-six subjects (12 males and 24 females) were included in the study. Mean age at spring placement was 3.9 months (range: 1.9-9.2) with a mean follow-up of 1.4 years (range: 0.3-5.2). The mean number of springs used was 3 (range: 2-4). The mean maximum single spring force was 9.9 Newtons (N) (range: 6.9-13.0) and the mean total spring force was 24.6 N (range: 12.7-37.0). Mean CI increased from 70 ± 0.9 preoperatively to 77 ± 1.0 postoperatively (P < 0.001). Age at spring placement was significantly associated with change in CI: for every month increase in age, the change in CI decreased by 1.3 (P = 0.03). The number of springs used, greatest single spring force, and total spring force did not correlate with changes in CI (P = 0.85, P = 0.42, and P = 0.84, respectively). In SMC, earlier age at time of surgery appears to correlate with greater improvement in CI, at least in the short-term. While spring characteristics did not appear to affect head shape, it is possible that the authors were underpowered to detect a difference, and spring-related variables likely deserve additional study.

Is there a causal role for homocysteine concentration in blood pressure? A Mendelian randomization study

Background: An understanding of whether homocysteine is a cause or a marker of increased blood pressure is relevant because blood homocysteine can be effectively lowered by safe and inexpensive interventions (e.g., vitamin B-6, B-9, and B-12 supplementation).Objective: The aim was to assess the causal influence of homocysteine on systolic and diastolic blood pressure (SBP and DBP, respectively) in adults with the use of Mendelian randomization (MR).Design: Data from the 1982 Pelotas Birth Cohort (Brazil) were used. A total of 4297 subjects were evaluated in 2004–2005 (mean age: 22.8 y). The association of homocysteine concentration with SBP and DBP was assessed by conventional ordinary least-squares (OLS) linear regression and 2-stage least-squares (2SLS) regression (MR analysis). The single nucleotide polymorphism (SNP) methylenetetrahydrofolate reductase (MTHFR) C677T (rs1801133) was used as proxy for homocysteine concentration. We also applied MR to data from the International Consortium for Blood Pressure (ICBP) genomewide association studies (>69,000 participants) using rs1801133 and additional homocysteine-associated SNPs as instruments.Results: In OLS regression, a 1-SD unit increase in log homocysteine concentration was associated with an increase of 0.9 (95% CI: 0.4, 1.4) mm Hg in SBP and of 1.0 (95% CI: 0.6, 1.4) mm Hg in DBP. In 2SLS regression, for the same increase in homocysteine, the coefficients were −1.8 mm Hg for SBP (95% CI: −3.9, 0.4 mm Hg; P = 0.01) and 0.1 mm Hg for DBP (95% CI: −1.5, 1.7 mm Hg; P = 0.24). In the MR analysis of ICBP data, homocysteine concentration was not associated with SBP (β = 0.6 mm Hg for each 1-SD unit increase in log homocysteine; 95% CI: −0.8, 1.9 mm Hg) but was positively associated with DBP (β = 1.1 mm Hg; 95% CI: 0.2, 1.9 mm Hg). The association of genetically increased homocysteine with DBP was not consistent across different SNPs.Conclusion: Overall, the present findings do not corroborate the hypothesis that homocysteine has a causal role in blood pressure, especially in SBP.