Ultra-processed food consumption and risk of diabetes: results from a population-based prospective cohort.

AIMS/HYPOTHESIS: Understanding the impact of the overall construct of ultra-processed foods on diabetes risk can inform dietary approaches to diabetes prevention. In this study, we aimed to evaluate the association between ultra-processed food consumption and risk of diabetes in a community-based cohort of middle-aged adults in the USA. We hypothesised that a higher intake of ultra-processed foods is associated with a higher risk of incident diabetes. METHODS: The study included 13,172 participants without diabetes at baseline (1987-1989) in the Atherosclerosis Risk in Communities (ARIC) study. Dietary intake was assessed with a 66-item semiquantitative food frequency questionnaire, and foods were categorised by processing level using the Nova classification system. Ultra-processed food was analysed categorically (quartiles of energy-adjusted intake) and continuously (per one additional serving/day). We used Cox regression to evaluate the association of ultra-processed food intake with risk of diabetes with adjustment for sociodemographic characteristics, total energy intake, health behaviours and clinical factors. RESULTS: Over a median follow-up of 21 years, there were 4539 cases of incident diabetes. Participants in the highest quartile of ultra-processed food intake (8.4 servings/day on average) had a significantly higher risk of diabetes (HR 1.13; 95% CI 1.03, 1.23) compared with participants in the lowest quartile of intake after adjustment for sociodemographic, lifestyle and clinical factors. Each additional serving of ultra-processed food consumed daily was associated with a 2% higher risk of diabetes (HR 1.02; 95% CI 1.00, 1.04). Highest quartile consumption of certain ultra-processed food groups, including sugar- and artificially sweetened beverages, ultra-processed meats and sugary snacks, was associated with a 29%, 21% and 16% higher risk of diabetes, respectively, compared with the lowest quartile. CONCLUSIONS/INTERPRETATION: We found that a higher intake of ultra-processed food was associated with higher risk of incident diabetes, particularly sugar- and artificially sweetened beverages, ultra-processed meats and sugary snacks. Our findings suggest interventions reducing ultra-processed food consumption and specific food groups may be an effective strategy for diabetes prevention.

Trends in Diabetes Treatment and Control in U.S. Adults, 1999-2018.

BACKGROUND: Documenting current trends in diabetes treatment and risk-factor control may inform public health policy and planning. METHODS: We conducted a cross-sectional analysis of data from adults with diabetes in the United States participating in the National Health and Nutrition Examination Survey (NHANES) to assess national trends in diabetes treatment and risk-factor control from 1999 through 2018. RESULTS: Diabetes control improved from 1999 to the early 2010s among the participants but subsequently stalled and declined. Between the 2007-2010 period and the 2015-2018 period, the percentage of adult NHANES participants with diabetes in whom glycemic control (glycated hemoglobin level, <7%) was achieved declined from 57.4% (95% confidence interval [CI], 52.9 to 61.8) to 50.5% (95% CI, 45.8 to 55.3). After major improvements in lipid control (non-high-density lipoprotein cholesterol level, <130 mg per deciliter) in the early 2000s, minimal improvement was seen from 2007-2010 (52.3%; 95% CI, 49.2 to 55.3) to 2015-2018 (55.7%; 95% CI, 50.8 to 60.5). From 2011-2014 to 2015-2018, the percentage of participants in whom blood-pressure control (<140/90 mm Hg) was achieved decreased from 74.2% (95% CI, 70.7 to 77.4) to 70.4% (95% CI, 66.7 to 73.8). The percentage of participants in whom all three targets were simultaneously achieved plateaued after 2010 and was 22.2% (95% CI, 17.9 to 27.3) in 2015-2018. The percentages of participants who used any glucose-lowering medication or any blood-pressure-lowering medication were unchanged after 2010, and the percentage who used statins plateaued after 2014. After 2010, the use of combination therapy declined in participants with uncontrolled blood pressure and plateaued for those with poor glycemic control. CONCLUSIONS: After more than a decade of progress from 1999 to the early 2010s, glycemic and blood-pressure control declined in adult NHANES participants with diabetes, while lipid control leveled off. (Funded by the National Heart, Lung, and Blood Institute.).

Prevalence and Disparities in Telehealth Use Among US Adults Following the COVID-19 Pandemic: National Cross-Sectional Survey.

BACKGROUND: Telemedicine expanded during the COVID-19 pandemic, though use differed by age, sex, race or ethnicity, educational attainment, income, and location. It is unclear if high telehealth use or inequities persisted late into the pandemic. OBJECTIVE: This study aims to evaluate the prevalence of, inequities in, and primary reasons for telehealth visits a year after telemedicine expansion. METHODS: We used cross-sectional data from the 2022 Health Information National Trends Survey (HINTS 6), the first cycle with data on telemedicine. In total, 4830 English- and Spanish-speaking US adults (aged ≥18 years) were included in this study. The primary outcomes were telehealth visit attendance in the 12 months before March 7, 2022, to November 8, 2022, and the primary reason for the most recent telehealth visit. We evaluated sociodemographic and clinical predictors of telehealth visit attendance and the primary reason for the most recent telehealth visit through Poisson regression. Analyses were weighted according to HINTS 6 standards. RESULTS: We included 4830 participants (mean age 48.3, SD 17.5 years; 50.28% women; 65.21% White). Among US adults, 38.78% reported having a telehealth visit in the previous year. Telehealth visit attendance rates were similar across age, race or ethnicity, income, and urban versus rural location. However, individuals with a telehealth visit were less likely to live in the Midwest (adjusted prevalence ratio [aPR] 0.65, 95% CI 0.54-0.77), and more likely to be women (aPR 1.21, 95% CI 1.06-1.38), college graduates or postgraduates (aPR 1.24, 95% CI 1.05-1.46), covered by health insurance (aPR 1.56, 95% CI 1.08-2.26), and married or cohabitating (aPR 1.17, 95% CI 1.03-1.32), adjusting for sociodemographic characteristics, frequency of health care visits, and comorbidities. Among participants with a telehealth visit in the past year, the primary reasons for their most recent visit were minor or acute illness (32.15%), chronic disease management (21%), mental health or substance abuse (16.94%), and an annual exam (16.22%). Older adults were more likely to report that the primary reason for their most recent telehealth visit was for chronic disease management (aPR 2.08, 95% CI 1.33-3.23), but less likely to report that it was for a mental health or substance abuse issue (aPR 0.19, 95% CI 0.10-0.35), adjusting for sociodemographic characteristics and frequency of health care visits. CONCLUSIONS: Among US adults, telehealth visit attendance was high more than a year after telemedicine expansion and did not differ by age, race or ethnicity, income, or urban versus rural location. Telehealth could continue to be leveraged following COVID-19 to improve access to care and health equity.

High-sensitivity troponins and mortality in the general population.

AIMS: Cardiac troponin T and I can be measured using a number of high-sensitivity (hs) assays. This study aimed to characterize correlations between four such assays and test their comparative associations with mortality. METHODS AND RESULTS: Among adults without cardiovascular disease in the 1999-2004 National Health and Nutrition Examination Survey, hs-troponin T was measured using one assay (Roche) and hs-troponin I using three assays (Abbott, Siemens, and Ortho). Cox regression was used to estimate associations with all-cause and cardiovascular mortality. Pearson's correlation coefficients comparing concentrations from each assay ranged from 0.53 to 0.77. There were 2188 deaths (488 cardiovascular) among 9810 participants. Each hs-troponin assay [log-transformed, per 1 standard deviation (SD)] was independently associated with all-cause mortality: hazard ratio (HR) 1.20 [95% confidence interval (CI) 1.13-1.28] for Abbott hs-troponin I; HR 1.10 (95% CI 1.02-1.18) for Siemens hs-troponin I; HR 1.23 (95% CI 1.14-1.33) for Ortho hs-troponin I; and HR 1.31 (95% CI 1.21-1.42) for Roche hs-troponin T. Each hs-troponin assay was also independently associated with cardiovascular mortality (HR 1.44 to 1.65 per 1 SD). Associations of hs-troponin T and all-cause and cardiovascular mortality remained significant after adjusting for hs-troponin I. Furthermore, associations of hs-troponin I remained significant after mutually adjusting for hs-troponin I from the other individual assays: e.g. cardiovascular mortality HR 1.46 (95% CI 1.19-1.79) for Abbott after adjustment for the Siemens assay and HR 1.29 (95% CI 1.09-1.53) for Abbott after adjustment for the Ortho assay. CONCLUSION: This study demonstrates only modest correlations between hs-troponin T and three hs-troponin I assays and that hs-troponin I assays can provide distinct risk information for mortality in the general population.

Satisfaction with Telehealth Care in the United States: Cross-Sectional Survey.

Background: Telehealth use remains high following the COVID-19 pandemic, but patient satisfaction with telehealth care is unclear. Methods: We used cross-sectional data from the Health Information National Trends Survey (HINTS 6). 2,058 English and Spanish-speaking U.S. adults (≥18 years) with a telehealth visit in the 12 months before March-November 2022 were included in this study. The primary outcomes were telehealth visit modality and satisfaction in the 12 months before HINTS 6. We evaluated sociodemographic predictors of telehealth visit modality and satisfaction via Poisson regression. Analyses were weighted according to HINTS standards. Results: We included 2,058 participants (48.4 ± 16.8 years; 57% women; 66% White), of which 70% had an audio-video and 30% an audio-only telehealth visit. Adults with an audio-video visit were more likely to have health insurance (adjusted prevalence ratio [aPR]: 1.55, 95% confidence interval [CI]: 1.18-2.04) and have an annual household income of ≥$75,000 (aPR: 1.18, 95% CI: 1.00-1.39) and less likely to be ≥65 years (aPR: 0.79, 95% CI: 0.70-0.89), adjusting for sociodemographic characteristics. No further inequities were noted by telehealth modality. Seventy-five percent of participants felt that their telehealth visits were as good as in-person care. No significant differences in telehealth satisfaction were observed across sociodemographic characteristics, telehealth modality, or the participants' primary reason for their most recent telehealth visit in adjusted analysis. Conclusions: Among U.S. adults with a telehealth visit, the majority had an audio-video visit and were satisfied with their care. Telehealth should continue, being offered following COVID-19, as it is uniformly valued by patients.

Prediabetes, intervening diabetes and subsequent risk of dementia: the Atherosclerosis Risk in Communities (ARIC) study.

AIMS/HYPOTHESIS: The aim of this work was to evaluate whether the association of prediabetes with dementia is explained by the intervening onset of diabetes. METHODS: Among participants of the Atherosclerosis Risk in Communities (ARIC) study we defined baseline prediabetes as HbA1c 39-46 mmol/mol (5.7-6.4%) and subsequent incident diabetes as a self-reported physician diagnosis or use of diabetes medication. Incident dementia was ascertained via active surveillance and adjudicated. We quantified the association of prediabetes with dementia risk before and after accounting for the subsequent development of diabetes among ARIC participants without diabetes at baseline (1990-1992; participants aged 46-70 years). We also evaluated whether age at diabetes diagnosis modified the risk of dementia. RESULTS: Among 11,656 participants without diabetes at baseline, 2330 (20.0%) had prediabetes. Before accounting for incident diabetes, prediabetes was significantly associated with the risk of dementia (HR 1.12 [95% CI 1.01, 1.24]). After accounting for incident diabetes, the association was attenuated and non-significant (HR 1.05 [95% CI 0.94, 1.16]). Earlier age of onset of diabetes had the strongest association with dementia: HR 2.92 (95% CI 2.06, 4.14) for onset before 60 years; HR 1.73 (95% CI 1.47, 2.04) for onset at 60-69 years; and HR 1.23 (95% CI 1.08, 1.40) for onset at 70-79 years. CONCLUSIONS/INTERPRETATION: Prediabetes is associated with dementia risk but this risk is explained by the subsequent development of diabetes. Earlier age of onset of diabetes substantially increases dementia risk. Preventing or delaying progression of prediabetes to diabetes will reduce dementia burden.

Performance of the Glucose Management Indicator (GMI) in Type 2 Diabetes.

BACKGROUND: The glucose management indicator (GMI) is an estimated measure of hemoglobin A1c (HbA1c) recommended for the management of persons with diabetes using continuous glucose monitoring (CGM). However, GMI was derived primarily in young adults with type 1 diabetes, and its performance in patients with type 2 diabetes is poorly characterized. METHODS: We conducted a prospective cohort study in 144 adults with obstructive sleep apnea and type 2 diabetes not using insulin (mean age: 59.4 years; 45.1% female). HbA1c was measured at the study screening visit. Participants simultaneously wore 2 CGM sensors (Dexcom G4 and Abbott Libre Pro) for up to 4 weeks (2 weeks at baseline and 2 weeks at the 3-month follow-up visit). GMI was calculated using all available CGM data for each sensor. RESULTS: Median wear time was 27 days (IQR: 23-29) for the Dexcom G4 and 28 days (IQR: 24-29) for the Libre Pro. The mean difference between HbA1c and GMI was small (0.12-0.14 percentage points) (approximately 2 mmol/mol). However, the 2 measures were only moderately correlated (r = 0.68-0.71), and there was substantial variability in GMI at any given value of HbA1c (root mean squared error: 0.66-0.69 percentage points [7 to 8 mmol/mol]). Between 36% and 43% of participants had an absolute difference between HbA1c and GMI ≥0.5 percentage points (≥5 mmol/mol), and 9% to 18% had an absolute difference >1 percentage points (>11 mmol/mol). Discordance was higher in the Libre Pro than the Dexcom G4. CONCLUSIONS: GMI may be an unreliable measure of glycemic control for patients with type 2 diabetes and should be interpreted cautiously in clinical practice.Clinicaltrials.gov Registration Number: NCT02454153.

Within-Person and Between-Sensor Variability in Continuous Glucose Monitoring Metrics.

BACKGROUND: The within-person and between-sensor variability of metrics from different interstitial continuous glucose monitoring (CGM) sensors in adults with type 2 diabetes not taking insulin is unclear. METHODS: Secondary analysis of data from 172 participants from the Hyperglycemic Profiles in Obstructive Sleep Apnea randomized clinical trial. Participants simultaneously wore Dexcom G4 and Abbott Libre Pro CGM sensors for up to 2 weeks at baseline and again at the 3-month follow-up visit. RESULTS: At baseline (up to 2 weeks of CGM), mean glucose for both the Abbott and Dexcom sensors was approximately 150 mg/dL (8.3 mmol/L) and time in range (70180 mg/dL [3.910.0 mmol/L]) was just below 80. When comparing the same sensor at 2 different time points (two 2-week periods, 3 months apart), the within-person coefficient of variation (CVw) in mean glucose was 17.4 (Abbott) and 14.2 (Dexcom). CVw for percent time in range: 20.1 (Abbott) and 18.6 (Dexcom). At baseline, the Pearson correlation of mean glucose from the 2 sensors worn simultaneously was r 0.86, root mean squared error (RMSE), 13 mg/dL (0.7 mmol/L); for time in range, r 0.88, RMSE, 8 percentage points. CONCLUSIONS: Substantial variation was observed within sensors over time and across 2 different sensors worn simultaneously on the same individuals. Clinicians should be aware of this variability when using CGM technology to make clinical decisions.ClinicalTrials.gov Identifier: NCT02454153.

Effectiveness of Coronavirus Disease 2019 (COVID-19) Vaccination Against Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) Infection Among Residents of US Nursing Homes Before and During the Delta Variant Predominance, December 2020-November 2021.

BACKGROUND: Residents of nursing homes experience disproportionate morbidity and mortality related to coronavirus disease 2019 (COVID-19) and were prioritized for vaccine introduction. We evaluated COVID-19 vaccine effectiveness (VE) in preventing severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infections among nursing home residents. METHODS: We used a retrospective cohort of 4315 nursing home residents during 14 December 2020-9 November 2021. A Cox proportional hazards model was used to estimate hazard ratios comparing residents with a completed vaccination series with unvaccinated among those with and without prior SARS-CoV-2 infection, by vaccine product, and by time period. RESULTS: Overall adjusted VE was 58% (95% confidence interval [CI], 44% to 69%) among residents without a history of SARS-CoV-2 infection. During the pre-Delta period, the VE within 150 days of receipt of the second dose of Pfizer-BioNTech (67%; 95% CI, 40% to 82%) and Moderna (75%; 95% CI, 32% to 91%) was similar. During the Delta period, VE measured >150 days after the second dose was 33% (95% CI, -2% to 56%) for Pfizer-BioNTech and 77% (95% CI, 48% to 91%) for Moderna. Rates of infection were 78% lower (95% CI, 67% to 85%) among residents with prior SARS-CoV-2 infection and completed vaccination series compared with unvaccinated residents without a history of SARS-CoV-2 infection. CONCLUSIONS: COVID-19 vaccines were effective in preventing SARS-CoV-2 infections among nursing home residents, and history of prior SARS-CoV-2 infection provided additional protection. Maintaining high coverage of recommended doses of COVID-19 vaccines remains a critical tool for preventing infections in nursing homes.

Glycated Albumin for the Diagnosis of Diabetes in US Adults.

BACKGROUND: There is growing interest in using glycated albumin for the diagnosis of diabetes, especially when standard tests (glucose and hemoglobin A1c [Hb A1c]) are unavailable. However, it is unknown how well glycated albumin identifies diabetes in the general population. METHODS: We measured glycated albumin in stored serum samples from the 1999-2004 National Health and Nutrition Examination Survey. We evaluated the ability of glycated albumin to identify undiagnosed diabetes in US adults aged ≥20 (n = 4785), overall and at thresholds corresponding to clinical cut points for Hb A1c and fasting plasma glucose (FPG). We assessed 4 reference definitions for undiagnosed diabetes: increased FPG (≥126 mg/dL) [≥6.99 mmol/L), increased Hb A1c (≥6.5%), either FPG or Hb A1c increased, or both FPG and Hb A1c increased. RESULTS: Among US adults, glycated albumin had excellent diagnostic accuracy across all 4 definitions of undiagnosed diabetes, with the area under the receiver operating characteristic curve (AUC) ranging from 0.824 to 0.951. Performance was generally consistent across patient demographic and clinical characteristics. Glycated albumin cut points of 16.5% and 17.8% were equivalent to an FPG of 126 mg/dL (6.99 mmol/L; 97th percentile) and Hb A1c of 6.5% (98th percentile) and had low to moderate sensitivity (0.273 to 0.707) but high specificity (0.980 to 0.992) for detecting undiagnosed diabetes. CONCLUSION: The excellent diagnostic performance of glycated albumin to identify diabetes defined by either FPG or Hb A1c suggests that glycated albumin may be useful for identifying adults with undiagnosed diabetes when standard tests are unavailable.

Learning and Inference in Sparse Coding Models With Langevin Dynamics.

We describe a stochastic, dynamical system capable of inference and learning in a probabilistic latent variable model. The most challenging problem in such models-sampling the posterior distribution over latent variables-is proposed to be solved by harnessing natural sources of stochasticity inherent in electronic and neural systems. We demonstrate this idea for a sparse coding model by deriving a continuous-time equation for inferring its latent variables via Langevin dynamics. The model parameters are learned by simultaneously evolving according to another continuous-time equation, thus bypassing the need for digital accumulators or a global clock. Moreover, we show that Langevin dynamics lead to an efficient procedure for sampling from the posterior distribution in the L0 sparse regime, where latent variables are encouraged to be set to zero as opposed to having a small L1 norm. This allows the model to properly incorporate the notion of sparsity rather than having to resort to a relaxed version of sparsity to make optimization tractable. Simulations of the proposed dynamical system on both synthetic and natural image data sets demonstrate that the model is capable of probabilistically correct inference, enabling learning of the dictionary as well as parameters of the prior.

Diabetes and the risk of hospitalisation for infection: the Atherosclerosis Risk in Communities (ARIC) study.

AIMS/HYPOTHESIS: The aim of this work was to assess the association between diabetes and risk for infection-related hospitalisation and mortality. METHODS: We conducted a prospective cohort analysis of the Atherosclerosis Risk in Communities (ARIC) study. Diabetes was defined as a fasting glucose ≥7 mmol/l or non-fasting glucose ≥11.1 mmol/l, self-report of a diagnosis of diabetes by a physician, or current diabetes medication use. Hospitalisation for infection was ascertained from hospital discharge records. Participants were followed from 1987-1989 to 2019. RESULTS: We included 12,379 participants (mean age 54.5 years; 24.7% Black race; 54.3% female sex). During a median follow-up of 23.8 years, there were 4229 new hospitalisations for infection. After adjusting for potential confounders, people with (vs without) diabetes at baseline had a higher risk for hospitalisation for infection (HR 1.67 [95% CI 1.52, 1.83]). Results were generally consistent across infection type but the association was especially pronounced for foot infection (HR 5.99 [95% CI 4.38, 8.19]). Diabetes was more strongly associated with hospitalisation for infection in younger participants and Black people. Overall infection mortality was low (362 deaths due to infection) but the adjusted risk was increased for people with diabetes (HR 1.72 [95% CI 1.28, 2.31]). CONCLUSIONS/INTERPRETATION: Diabetes confers significant risk for infection-related hospitalisation. Enhancing prevention and early treatment of infection in those with diabetes is needed to reduce infection-related morbidity and mortality.

Trends in Diabetes Management Among US Adults: 1999-2016.

BACKGROUND: Updating national trends in diabetes management is important for identifying areas of progress and remaining gaps in diabetes care. OBJECTIVE: Evaluate trends in diabetes management. DESIGN: Three nationally representative, serial cross-sectional surveys (National Health and Nutrition Examination Survey [n = 5800], National Health Interview Survey [n = 48,519], and Behavioral Risk Factor Surveillance System [n = 741,497]) were used to estimate trends in diabetes management from 1999 to 2016. PARTICIPANTS: Non-pregnant US adults (aged ≥ 18 years) diagnosed with diabetes. MAIN MEASURES: American Diabetes Association's general recommendations for glycemic and cardiovascular risk factor control, medication usage, physical activity, preventive practices, and dietary intake. KEY RESULTS: From 1999 to 2016, the proportion of US adults with diabetes who attained glycemic control (HbA1c < 7.0%) followed a quadratic trend (49.6% in 1999-2004 to 58.6% in 2005-2010 to 55.8% in 2011-2016, P < 0.05 for trend). Control of blood pressure (< 140/90 mmHg) and lipids (LDL cholesterol < 100 mg/dl) increased by 6.6 and 18.7 percentage points, respectively (P < 0.05 for trends). The proportion that attained glycemic, blood pressure, and lipid control followed a quadratic trend (13.3% in 1999-2004 to 24.8% in 2005-2010 to 20.2% in 2011-2016, P < 0.05 for trend). Use of antidiabetic, antihypertensive, and statin medication among those who were eligible rose by 8.6, 5.0, and 24.0 percentage points, respectively (P < 0.05 for trends). Aerobic inactivity declined 7.1 percentage points, while adherence to aerobic activity (≥ 150 min/week) and resistance training (≥ 2 times/week) recommendations grew 3.4 and 3.2 percentage points, respectively (P < 0.05 for trends). Engagement in preventive practices (e.g., receipt of vaccinations) consistently increased for 6 out of 8 outcomes. However, the adherence to saturated fat (< 10% of total daily calories) and sodium (< 2300 mg/day) recommendations fell by 6.5 and 5.2 percentage points (P < 0.05 for trends). CONCLUSIONS: Despite notable improvements, declines in glycemic control and adherence to dietary recommendations may be growing challenges in diabetes care.

Assessing and reducing PET radiotracer infiltration rates: a single center experience in injection quality monitoring methods and quality improvement.

BACKGROUND: Successful injection of radiolabeled compounds is critical for positron emission tomography (PET) imaging. A poor quality injection limits the tracer availability in the body and can impact diagnostic results. In this study, we attempt to quantify our infiltration rates, develop an actionable quality improvement plan to reduce potentially compromised injections, and compare injection scoring to PET/CT imaging results. METHODS: A commercially available system that uses external radiation detectors was used to monitor and score injection quality. This system compares the time activity curves of the bolus relative to a control reading in order to provide a score related to the quality of the injection. These injection scores were used to assess infiltration rates at our facility in order to develop and implement a quality improvement plan for our PET imaging center. Injection scores and PET imaging results were reviewed to determine correlations between image-based assessments of infiltration, such as liver SUVs, and injection scoring, as well as to gather infiltration reporting statistics by physicians. RESULTS: A total of 1033 injections were monitored at our center. The phase 1 infiltration rate was 2.1%. In decision tree analysis, patients < 132.5lbs were associated with infiltrations. Additional analyses suggested patients > 127.5 lbs. with non-antecubital injections were associated with lower quality injections. Our phase 2 infiltration rate was 1.9%. Comparison of injection score to SUV showed no significant correlation and indicated that only 63% of suspected infiltrations were visible on PET/CT imaging. CONCLUSIONS: Developing a quality improvement plan and monitoring PET injections can lead to reduced infiltration rates. No significant correlation between reference SUVs and injection score provides evidence that determination of infiltration based on PET images alone may be limited. Results also indicate that the number of infiltrated PET injections is under-reported.

Design of optical neural networks with component imprecisions.

For the benefit of designing scalable, fault resistant optical neural networks (ONNs), we investigate the effects architectural designs have on the ONNs' robustness to imprecise components. We train two ONNs - one with a more tunable design (GridNet) and one with better fault tolerance (FFTNet) - to classify handwritten digits. When simulated without any imperfections, GridNet yields a better accuracy (∼98%) than FFTNet (∼95%). However, under a small amount of error in their photonic components, the more fault tolerant FFTNet overtakes GridNet. We further provide thorough quantitative and qualitative analyses of ONNs' sensitivity to varying levels and types of imprecisions. Our results offer guidelines for the principled design of fault-tolerant ONNs as well as a foundation for further research.

Association Between Adolescent Body Mass Index and Adulthood Smoking.

INTRODUCTION: Studies increasingly find a prospective association between adolescent weight status and adverse physical health outcomes. However, less is known about the long-term consequences of adolescent body weight on the adoption of health-risk behaviors. This study sought to determine whether adolescent body mass index (BMI) was associated with cigarette smoking in adulthood. METHODS: Six thousand six hundred eighty-three nonsmoking adolescents were interviewed at baseline (1996, age 11-20) and at follow-up (2008, age 24-32) as part of the National Longitudinal Survey of Adolescent Health. Logistic and fractional regression models tested the association between adolescent BMI and smoking status and frequency in adulthood. Respondent weight and height were measured at baseline and converted into age and sex-specific BMI z-scores. Being a smoker was defined as smoking at least once in the past 30 days, while smoking frequency was defined as the proportion of days smoked over the past 30 days. Both outcomes were measured at follow-up. RESULTS: Adolescent BMI was positively associated with the transition from nonsmoking to smoking 12 years later for women but not men. Adolescent BMI was also positively associated with smoking frequency among women smokers. Both associations persisted after adjusting for established risk factors and were robust to sensitivity analyses. CONCLUSIONS: Adolescent BMI was strongly associated with increased cigarette smoking behavior in adulthood for women, even after adjusting for important risk factors. IMPLICATIONS: Using a large, nationally representative sample, this study found that adolescent BMI was positively associated with smoking behavior during adulthood for women but not men. These results have potentially important public health implications for future smoking rates in the United States, as the prevalence of overweight and obesity among US adolescent females continues to grow. Thus, it may be important for smoking prevention interventions to prioritize overweight adolescent females moving forward.