Repeatedly measured predictors: a comparison of methods for prediction modeling

Effects and Thresholds of Young to Midlife Vascular Risks on Brain Health

BACKGROUND

Vascular risk factors, particularly hypertension, are important contributors to accelerated brain aging. We sought to quantify vascular risk factor risks over adulthood and assess the empirical evidence for risk thresholds.

METHODS

We used SBP (systolic blood pressure) and diastolic blood pressure, total cholesterol, fasting blood glucose, and body mass index measurements collected from participants in the CARDIA study (Coronary Artery Risk Development in Young Adults) at 2- to 5-year intervals through year 30. The Montreal Cognitive Assessment and domain-specific cognitive tests were performed at year 30. White matter hyperintensity volume was measured by magnetic resonance imaging. We used a 2-step method to fit longitudinal vascular risk factor exposures to optimized spline functions with mixed-effects models, then used the participant-specific random effects that characterized individual exposures over time in cross-sectional models adjusted for sex, race, age, and education to study effects on midlife brain health.

RESULTS

Change in SBP up to 33 years of age was negatively associated with Montreal Cognitive Assessment scores (-0.29 Montreal Cognitive Assessment Z score per mm Hg/y change [95% CI, -0.49 to -0.09]; P=0.005), with similar effects for SBP changes from 33 to 49 years of age (-0.08 [95% CI, -0.16 to 0.01]; P=0.08). We observed comparable, significant associations between SBP exposure during those ages, midlife performance on specific cognitive domains, and volume of white matter hyperintensity (all P<0.05). SBP ≤111 mm Hg was the estimated threshold below which no harmful association with midlife cognitive performance was identified.

CONCLUSIONS

SBP in early adulthood is the vascular risk factor most strongly associated with midlife cognitive performance and white matter hyperintensity burden, with SBP 111 mm Hg suggested as a harm threshold.

Maternal and Infant Dietary Patterns Are Not Related to Food Allergy Risk in Singapore Children: GUSTO Cohort Study

BACKGROUND

We previously reported that delayed allergenic food introduction in infancy did not increase food allergy risk until age 4 y within our prospective cohort. However, it remains unclear whether other aspects of maternal or infant diet play roles in the development of childhood food allergy.

OBJECTIVES

We examined the relationship between maternal pregnancy and infant dietary patterns and the development of food allergies until age 8 y.

METHODS

Among 1152 Singapore Growing Up in Singapore Towards healthy Outcomes study mother-infant dyads, the infant's diet was ascertained using food frequency questionnaires at 18 mo. Maternal dietary patterns during pregnancy were derived from 24-h diet recalls. Food allergy was determined through interviewer-administered questionnaires at regular time points from infancy to age 8 y and defined as a positive history of allergic reactions, alongside skin prick tests at 18 mo, 3, 5, and 8 y.

RESULTS

Food allergy prevalence was 2.5% (22/883) at 12 mo and generally decreased over time by 8 y (1.9%; 14/736). Higher maternal dietary quality was associated with increased risk of food allergy (P ≤ 0.016); however, odds ratios were modest. Offspring food allergy risk ≤8 y showed no associations with measures of infant diet including timing of solids/food introduction (adjusted odds ratio [aOR]: 0.90; 95% confidence interval [CI]: 0.42, 1.92), infant's diet quality (aOR: 0.93; 95% CI: 0.88, 0.99) or diet diversity (aOR: 0.84; 95% CI: 0.6, 1.19). Most infants (89%) were first introduced to cow milk protein within the first month of life, while egg and peanut introduction were delayed (58.3% introduced by mean age 8.8 mo and 59.8% by mean age 18.1 mo, respectively).

CONCLUSIONS

Apart from maternal diet quality showing a modest association, infant's allergenic food introduction, diet quality, and dietary diversity were not associated with food allergy development in this Asian pediatric population. Interventional studies are needed to evaluate the efficacy of these approaches to food allergy prevention across different populations.

Bioenergetic function is decreased in peripheral blood mononuclear cells of veterans with Gulf War Illness

Gulf War Illness (GWI) is a major health problem for approximately 250,000 Gulf War (GW) veterans, but the etiology of GWI is unclear. We hypothesized that mitochondrial dysfunction is an important contributor to GWI, based on the similarity of some GWI symptoms to those occurring in some mitochondrial diseases; the plausibility that certain pollutants to which GW veterans were exposed affect mitochondria; mitochondrial effects observed in studies in laboratory models of GWI; and previous evidence of mitochondrial outcomes in studies in GW veterans. A primary role of mitochondria is generation of energy via oxidative phosphorylation. However, direct assessment of mitochondrial respiration, reflecting oxidative phosphorylation, has not been carried out in veterans with GWI. In this case-control observational study, we tested multiple measures of mitochondrial function and integrity in a cohort of 114 GW veterans, 80 with and 34 without GWI as assessed by the Kansas definition. In circulating white blood cells, we analyzed multiple measures of mitochondrial respiration and extracellular acidification, a proxy for non-aerobic energy generation; mitochondrial DNA (mtDNA) copy number; mtDNA damage; and nuclear DNA damage. We also collected detailed survey data on demographics; deployment; self-reported exposure to pesticides, pyridostigmine bromide, and chemical and biological warfare agents; and current biometrics, health and activity levels. We observed a 9% increase in mtDNA content in blood in veterans with GWI, but did not detect differences in DNA damage. Basal and ATP-linked oxygen consumption were respectively 42% and 47% higher in veterans without GWI, after adjustment for mtDNA amount. We did not find evidence for a compensatory increase in anaerobic energy generation: extracellular acidification was also lower in GWI (12% lower at baseline). A subset of 27 and 26 veterans returned for second and third visits, allowing us to measure stability of mitochondrial parameters over time. mtDNA CN, mtDNA damage, ATP-linked OCR, and spare respiratory capacity were moderately replicable over time, with intraclass correlation coefficients of 0.43, 0.44, 0.50, and 0.57, respectively. Other measures showed higher visit-to-visit variability. Many measurements showed lower replicability over time among veterans with GWI compared to veterans without GWI. Finally, we found a strong association between recalled exposure to pesticides, pyridostigmine bromide, and chemical and biological warfare agents and GWI (p < 0.01, p < 0.01, and p < 0.0001, respectively). Our results demonstrate decreased mitochondrial respiratory function as well as decreased glycolytic activity, both of which are consistent with decreased energy availability, in peripheral blood mononuclear cells in veterans with GWI.

Physical Activity and Cardiovascular Risk Factors in Children from 4 to 9 Years of Age

BACKGROUND

Physical activity guidelines for children encourage moderate-to-vigorous intensity activities (MVPA); however, some studies have found that only vigorous intensity activities (VPA) might promote health benefits in young children. Thus, the aim of this study is to investigate cross-sectional and 5-year longitudinal associations of VPA and MVPA with cardiovascular disease (CVD) risk factors in childhood using compositional data analysis.

RESULTS

This study utilized data from the SPINACH study (n = 411). Physical activity was measured with accelerometers at 4- and 9-years of age. CVD risk factors were measured at 9-years of age, and included blood pressure (BP), lipid metabolism, and glucose metabolism biomarkers, as well as a continuous metabolic syndrome risk score (MetS). Cross-sectional and longitudinal linear regression models were built using compositional data analysis standards. Cross-sectionally, reallocating time to VPA from lower-intensity behaviours at 9-years was associated with lower waist circumference (B = - 3.219, P = 0.002), diastolic BP (B = - 1.836, P = 0.036), triglycerides (B = - 0.214, P < 0.001), glucose (B = - 0.189, P = 0.033), insulin (B = - 2.997, P < 0.001), and HOMA-IR (B = - 0.778, P < 0.001). Similarly, reallocating time to VPA at 4-years was associated with lower MetS (B = - 0.831, P = 0.049), waist circumference (B = - 4.211, P = 0.015), systolic BP (B = - 5.572, P = 0.015), diastolic BP (B = - 2.931, P = 0.044), triglycerides (B = - 0.229, P = 0.034), glucose (B = - 0.325, P = 0.032), insulin (B = - 5.114, P = 0.001), and HOMA-IR (B = - 0.673, P = 0.001) at 9-years. Reallocations of time to MVPA at 4- or 9-years were not associated with CVD risk factors at 9-years.

CONCLUSIONS

VPA was associated with CVD risk factors in children both cross-sectionally (9-years) and longitudinally (at 4- and 9-years). MVPA seemed not to be a stimulus of enough intensity to trigger these potential cardiometabolic benefits in healthy children. Thus, these findings suggest the importance of higher intensity activities, i.e., VPA already in early childhood for cardiometabolic health.

Determinants of physical activity engagement in older adults

In order to increase engagement in physical activity, it is important to determine which factors contribute to physical activity engagement in older adults. The current study examined the relative predictive ability of several potential determinants, in terms of both the concurrent level as well as longitudinal trajectories. Clinically normal adults aged 61-92 completed the Physical Activity Scale for the Elderly (n = 189 for cross-sectional models; n = 214 for longitudinal models). Potential determinants included age, gender, education, physical health, sensory health, mood, cardiovascular health, cognitive status, and biomarkers of Alzheimer disease (AD). We observed a novel finding that both concurrent physical health (p < 0.001) and change in physical health (p < 0.001) were significant predictors above and beyond other determinants. Concurrent mood predicted levels of physical activity (p = 0.035), particularly in females. These findings suggest that poor physical health and low mood might be important to consider as potential barriers to physical activity engagement in older adults.

Physical activity and sedentary time across pregnancy and associations with neonatal weight, adiposity and cord blood parameters: a secondary analysis of the DALI study

BACKGROUND/OBJECTIVES

Obesity during pregnancy is associated with neonatal adiposity, which is a risk factor for childhood obesity. Maternal physical activity (PA) and sedentary behaviours during pregnancy might modify this risk. We therefore studied associations between maternal PA and sedentary time (ST) during pregnancy and neonatal anthropometry and cord blood parameters and investigated whether associations differed by offspring sex.

SUBJECTS/METHODS

Participants of the Vitamin D And Lifestyle Intervention for Gestational Diabetes Mellitus Prevention (DALI) study with a BMI  ≥ 29 kg/m2 were analysed as a cohort. Maternal moderate-to-vigorous PA (MVPA) and ST were measured repeatedly with accelerometers across pregnancy. Associations between mean levels and changes in MVPA and ST and birthweight, neonatal adiposity (fat mass (FM)%) and cord blood parameters, including C-peptide, leptin and lipids, were analysed in 213 mother-child pairs with Bayesian multilevel models. Interactions with offspring sex were considered.

RESULTS

Almost all women decreased MVPA levels and increased ST throughout gestation. Both higher maternal mean MVPA and increasing MVPA were associated with lower offspring FM% in males (-0.520%; 95% CI: -1.011%, -0.031% and -4.649%; -7.876%, -1.432% respectively). In female offspring, mean ST was associated with lower cord blood C-peptide (-0.145 µg/l; -0.279 µg/l, -0.005 µg/l). No associations were found with birthweight or other cord blood parameters.

CONCLUSIONS

Maternal MVPA is associated with neonatal fat mass, but not birthweight, in male offspring. Our findings underline the importance of physical activity throughout pregnancy.

Multivariate longitudinal data for survival analysis of cardiovascular event prediction in young adults: insights from a comparative explainable study

BACKGROUND

Multivariate longitudinal data are under-utilized for survival analysis compared to cross-sectional data (CS - data collected once across cohort). Particularly in cardiovascular risk prediction, despite available methods of longitudinal data analysis, the value of longitudinal information has not been established in terms of improved predictive accuracy and clinical applicability.

METHODS

We investigated the value of longitudinal data over and above the use of cross-sectional data via 6 distinct modeling strategies from statistics, machine learning, and deep learning that incorporate repeated measures for survival analysis of the time-to-cardiovascular event in the Coronary Artery Risk Development in Young Adults (CARDIA) cohort. We then examined and compared the use of model-specific interpretability methods (Random Survival Forest Variable Importance) and model-agnostic methods (SHapley Additive exPlanation (SHAP) and Temporal Importance Model Explanation (TIME)) in cardiovascular risk prediction using the top-performing models.

RESULTS

In a cohort of 3539 participants, longitudinal information from 35 variables that were repeatedly collected in 6 exam visits over 15 years improved subsequent long-term (17 years after) risk prediction by up to 8.3% in C-index compared to using baseline data (0.78 vs. 0.72), and up to approximately 4% compared to using the last observed CS data (0.75). Time-varying AUC was also higher in models using longitudinal data (0.86-0.87 at 5 years, 0.79-0.81 at 10 years) than using baseline or last observed CS data (0.80-0.86 at 5 years, 0.73-0.77 at 10 years). Comparative model interpretability analysis revealed the impact of longitudinal variables on model prediction on both the individual and global scales among different modeling strategies, as well as identifying the best time windows and best timing within that window for event prediction. The best strategy to incorporate longitudinal data for accuracy was time series massive feature extraction, and the easiest interpretable strategy was trajectory clustering.

CONCLUSION

Our analysis demonstrates the added value of longitudinal data in predictive accuracy and epidemiological utility in cardiovascular risk survival analysis in young adults via a unified, scalable framework that compares model performance and explainability. The framework can be extended to a larger number of variables and other longitudinal modeling methods.

TRIAL REGISTRATION

ClinicalTrials.gov Identifier: NCT00005130, Registration Date: 26/05/2000.

Pubertal increment in insulin resistance is negatively related to lumbar bone mineral density in 18-year-old males

Transient insulin resistance seen during puberty is expected to favour body growth, but our results show that increment in insulin resistance even in physiological ranges during puberty might compromise lumbar spine bone mineral density accrual independently of body composition parameters, and therefore adult bone quality might be challenged.

INTRODUCTION

Insulin resistance (IR) might have a compromising effect on growing bone, and therefore adult bone quality might be challenged. The aim of the present study was to identify whether increases in IR during puberty contribute to bone mineral characteristics in males independently of body composition parameters.

METHODS

This is a retrospective cohort-based longitudinal observational study. Data from 85 subjects were included. Boys were studied annually during their pubertal years (12 years at baseline) and at follow-up at the age of 18 years. Anthropometry, bone age, fasting blood samples, body composition, total body, and lumbar spine bone mineral characteristics were measured. Insulin resistance was determined by homeostatic model assessment of IR (HOMA-IR). Multiple regression analysis was performed to determine the effect of changes in HOMA-IR during pubertal years as a longitudinal predictor to fixed bone mineral outcome variables at the age of 18 years. All models were adjusted to potential clinically justified confounding variables.

RESULTS

After adjustment to baseline bone indices and body composition-related predictors, the pubertal increment in the HOMA-IR was a negative independent predictor of lumbar spine bone mineral areal density (β =  - 0.202, p = 0.005) and lumbar spine bone mineral apparent density (β =  - 0.235, p = 0.005) in 18-year-old males.

CONCLUSIONS

Pubertal increment in IR has a potential diminishing effect on lumbar spine bone mineral density accrual independently of body composition parameters. Further studies are needed to clarify whether monitoring HOMA-IR during puberty may identify subjects at increased risk of low peak bone mass and possible osteoporosis in the future.

Green and blue spaces and lung function in the Generation XXI cohort: a life-course approach

BACKGROUND

Exposure to natural environments may affect respiratory health. This study examined the association of exposure to green and blue spaces with lung function in children, and assessed the mediation effect of air pollution and physical activity.

METHODS

The study used data from the Generation XXI, a population-based birth cohort from the Porto Metropolitan Area (Portugal). Residential Normalised Difference Vegetation Index (NDVI) at different buffers (100, 250 and 500 m), the accessibility to urban green spaces (UGS) within 400 and 800 m and the minimum distance to the nearest UGS and to the nearest blue spaces were assessed at birth, 4, 7 and 10 years of age. Three life-course measures were calculated: averaged exposure, early-life exposure (birth) and exposure trend over time (change in exposure). Forced vital capacity (FVC), forced expiratory volume in 1 s (FEV1) and forced expiratory flow between 25% and 75% of FVC (FEF25-75%) at 10 years were used as outcomes. To assess associations, linear regression models and path analysis were used.

RESULTS

This study included 3278 children. The adjusted models showed that increasing the NDVI exposure over time within 100 m of the child's residence was associated with higher values of FEV1 (L) and FEF25-75% (L·s-1) (β 0.01, 95% CI 0.0002-0.03 and β 0.02, 95% CI 0.001-0.05, respectively). No significant associations were observed for the remaining measures of exposure, and no mediation effect was found for pollution or physical activity.

CONCLUSION

Increasing exposure to greenness at close proximity from residences was associated with improved lung function. While the mechanism remains unknown, this study brings evidence that city greening may improve children's respiratory health.

Association of Adverse Childhood Experiences and Social Isolation With Later-Life Cognitive Function Among Adults in China

IMPORTANCE

Studies investigating the association of threat-related and deprivation-related adverse childhood experiences (ACEs) with later-life cognitive decline are lacking.

OBJECTIVES

To evaluate the independent association of threat-related and deprivation-related ACEs with cognitive decline over time among middle-aged and older Chinese adults and to examine the modifying role of social isolation in such associations.

DESIGN, SETTING, AND PARTICIPANTS

This prospective cohort study used cognitive data from the China Health and Retirement Longitudinal Study (CHARLS) baseline survey that was administered between June 1, 2011, and March 31, 2012, and the CHARLS follow-up survey administered between July 1 and September 30, 2015. The life history survey with information of ACEs was additionally administered between June 1 and December 31, 2014. Statistical analysis was performed from March 1 to July 31, 2022. The study population consisted of middle-aged and older adults (age range, 45-97 years) with complete data on ACEs and 2 cognitive assessments and without cognitive impairment at baseline.

EXPOSURES

Five threat-related ACEs (ie, physical abuse, household substance abuse, domestic violence, unsafe neighborhood, and bullying) and 5 deprivation-related ACEs (ie, emotional neglect, household mental illness, incarcerated household member, parental separation or divorce, and parental death) before 17 years of age were queried by questionnaires. The cumulative scores of the 2 ACE dimensions were calculated and grouped into 3 categories as 0, 1, and 2 or more in main analyses.

MAIN OUTCOMES AND MEASURES

Cognitive function was measured by episodic memory and executive function. Global cognition was further calculated as the total score of these 2 dimensions. The raw scores of each cognitive test were standardized to z scores using baseline means and SDs. Linear mixed-effects models were constructed to examine the association between 2 dimensions of ACEs and the rate of annual cognitive decline. The modifying role of baseline social isolation in such associations was assessed with 3-way interaction tests.

RESULTS

Of the 6466 participants included in main analyses, 3301 (51.1%) were men and the mean (SD) age was 57.2 (8.3) years. Compared with no exposures, experience of 1 deprivation-related ACE was associated with faster cognitive decline in global cognition (β = -0.012 [95% CI, -0.022 to -0.002] SD/y) and executive function (β = -0.010 [95% CI, -0.020 to -0.00002] SD/y), whereas individuals with at least 2 childhood deprivations had faster cognitive declines in all cognitive tests (β = -0.035 [95% CI, -0.050 to -0.019] SD/y for global cognition; β = -0.047 [95% CI, -0.068 to -0.025] SD/y for episodic memory; β = -0.019 [95% CI, -0.034 to -0.004] SD/y for executive function). However, such an association was not observed for threat-related ACEs. In addition, baseline social isolation was a significant modifier in the associations between deprivation-related ACEs and cognitive declines in global cognition (β = -0.033 [95% CI, -0.061 to -0.005] SD/y; P = .02 for 3-way interaction) and executive function (β = -0.032 [95% CI, -0.059 to -0.005] SD/y; P = .02 for 3-way interaction).

CONCLUSIONS AND RELEVANCE

Deprivation-related ACEs, but not threat-related ACEs, were associated with faster decline in later-life cognitive function, whereas social isolation could modify such detrimental impact. These findings highlight the potential benefits of promoting social integration in maintaining later-life cognitive function among individuals who have experienced childhood deprivation.

BMI and HbA1c are metabolic markers for pancreatic cancer: Matched case-control study using a UK primary care database

Background

Weight loss, hyperglycaemia and diabetes are known features of pancreatic cancer. We quantified the timing and the amount of changes in body mass index (BMI) and glycated haemoglobin (HbA1c), and their association with pancreatic cancer from five years before diagnosis.

Methods

A matched case-control study was undertaken within 590 primary care practices in England, United Kingdom. 8,777 patients diagnosed with pancreatic cancer (cases) between 1^st January 2007 and 31^st August 2020 were matched to 34,979 controls by age, gender and diabetes. Longitudinal trends in BMI and HbA1c were visualised. Odds ratios adjusted for demographic and lifestyle factors (aOR) and 95% confidence intervals (CI) were calculated with conditional logistic regression. Subgroup analyses were undertaken according to the diabetes status.

Results

Changes in BMI and HbA1c observed for cases on longitudinal plots started one and two years (respectively) before diagnosis. In the year before diagnosis, a 1 kg/m² decrease in BMI between cases and controls was associated with aOR for pancreatic cancer of 1.05 (95% CI 1.05 to 1.06), and a 1 mmol/mol increase in HbA1c was associated with aOR of 1.06 (1.06 to 1.07). ORs remained statistically significant (p < 0.001) for 2 years before pancreatic cancer diagnosis for BMI and 3 years for HbA1c. Subgroup analysis revealed that the decrease in BMI was associated with a higher pancreatic cancer risk for people with diabetes than for people without (aORs 1.08, 1.06 to 1.09 versus 1.04, 1.03 to 1.05), but the increase in HbA1c was associated with a higher risk for people without diabetes than for people with diabetes (aORs 1.09, 1.07 to 1.11 versus 1.04, 1.03 to 1.04).

Conclusions

The statistically significant changes in weight and glycaemic control started three years before pancreatic cancer diagnosis but varied according to the diabetes status. The information from this study could be used to detect pancreatic cancer earlier than is currently achieved. However, regular BMI and HbA1c measurements are required to facilitate future research and implementation in clinical practice.

Longitudinal Analysis of Amplitude-Integrated Electroencephalography for Outcome Prediction in Hypoxic-Ischemic Encephalopathy

OBJECTIVE

To investigate the prognostic accuracy of longitudinal analysis of amplitude-integrated electroencephalography (aEEG) background activity to predict long-term neurodevelopmental outcome in neonates with hypoxic-ischemic encephalopathy (HIE) receiving therapeutic hypothermia.

STUDY DESIGN

This single-center observational study included 149 neonates for derivation and 55 neonates for validation with moderate-severe HIE and of gestational age ≥35 weeks at a tertiary neonatal intensive care unit. Single-channel aEEG background pattern, sleep-wake cycling, and seizure activity were monitored over 84 hours during therapeutic hypothermia and rewarming, then scored for each 6-hour interval. Neurodevelopmental outcome was assessed using the Bayley Scales of Infant Development, Second Edition. Favorable outcome was defined as having both a Mental Development Index (MDI) score and Psychomotor Development Index (PDI) score ≥70, and adverse outcome was defined as either an MDI or a PDI <70 or death. Regression modeling for longitudinal analysis of repeatedly measured data was applied, and area under the receiver operating characteristic curve (AUC) was calculated.

RESULTS

Longitudinal aEEG background analysis combined with sleep-wake cycling score had excellent predictive value (AUC, 0.90; 95% CI, 0.85-0.95), better than single aEEG scores at any individual time point. The model performed well in the independent validation cohort (AUC, 0.87; 95% CI, 0.62-1.00). The reclassification rate of this model compared with the conventional analysis of aEEG background at 48 hours was 18% (24 patients); 14% (18 patients) were reclassified correctly. Our results were used to develop a user-friendly online outcome prediction tool.

CONCLUSIONS

Longitudinal analysis of aEEG background activity and sleep-wake cycling is a valuable and accurate prognostic tool.

Physical Activity and Sedentary Time in Pregnancy: An Exploratory Study on Oxidative Stress Markers in the Placenta of Women with Obesity

Regular moderate-to-vigorous physical activity (MVPA) and reduced sedentary time (ST) improve maternal glucose metabolism in pregnancy. More MVPA and less ST outside pregnancy increase antioxidant capacity, hence, are beneficial in preventing oxidative stress. The placenta is the first line of defense for the fetus from an adverse maternal environment, including oxidative stress. However, effects of MVPA and ST on oxidative stress markers in the placenta are unknown. The purpose of this study was to assess the association of MVPA and ST in pregnancy with oxidative stress markers in placentas of overweight/obese women (BMI ≥ 29 kg/m2). MVPA and ST were objectively measured with accelerometers at <20 weeks, 24−27 and 35−37 weeks of gestation. Using linear Bayesian multilevel models, the associations of MVPA and ST (mean and changes) with mRNA expression of a panel of 11 oxidative stress related markers were assessed in 96 women. MVPA was negatively correlated with HSP70 mRNA expression in a sex-independent manner and with GCLM expression only in placentas of female fetuses. ST was positively associated with HO-1 mRNA expression in placentas of male neonates. None of the other markers were associated with MVPA or ST. We speculate that increasing MVPA and reducing ST attenuates the oxidative stress state in placentas of obese pregnant women.

The SMART Framework: Selection of Machine learning Algorithms with ReplicaTions - a Case Study on the Microvascular Complications of Diabetes

Over 34 million people in the US have diabetes, a major cause of blindness, renal failure, and amputations. Machine learning (ML) models can predict high-risk patients to help prevent adverse outcomes. Selecting the 'best' prediction model for a given disease, population, and clinical application is challenging due to the hundreds of health-related ML models in the literature and the increasing availability of ML methodologies. To support this decision process, we developed the Selection of Machine-learning Algorithms with ReplicaTions (SMART) Framework that integrates building and selecting ML models with decision theory. We build ML models and estimate performance for multiple plausible future populations with a replicated nested cross-validation technique. We rank ML models by simulating decision-maker priorities, using a range of accuracy measures (e.g., AUC) and robustness metrics from decision theory (e.g., minimax Regret). We present the SMART Framework through a case study on the microvascular complications of diabetes using data from the ACCORD clinical trial. We compare selections made by risk-averse, -neutral, and -seeking decision-makers, finding agreement in 80% of the risk-averse and risk-neutral selections, with the risk-averse selections showing consistency for a given complication. We also found that the models that best predicted outcomes in the validation set were those with low performance variance on the testing set, indicating a risk-averse approach in model selection is ideal when there is a potential for high population feature variability. The SMART Framework is a powerful, interactive tool that incorporates various ML algorithms and stakeholder preferences, generalizable to new data and technological advancements.

Accounting for Repeat Pregnancies in Risk Prediction Models

BACKGROUND

In perinatal epidemiology, the development of risk prediction models is complicated by parity; how repeat pregnancies influence the predictive accuracy of models that include obstetrical history is unclear.

METHODS

To assess the influence of repeat pregnancies on the association between predictors and the outcomes, as well as the influence of ignoring the nonindependence between pregnancies, we created four analytical cohorts using the Clinical Practice Research Datalink. The cohorts included (1) first deliveries, (2) a random sample of one delivery per woman, (3) all eligible deliveries per woman, and (4) all eligible deliveries and censoring of follow-up at subsequent pregnancies. Using Plasmode simulations, we varied the predictor-outcome association across cohorts.

RESULTS

We found minimal differences in the relative contribution of predictors to the overall predictions and the discriminative accuracy of models in the cohort of randomly sampled deliveries versus the all deliveries cohort (C-statistic: 0.62 vs. 0.63; Nagelkerke's R2: 0.03 for both). Accounting for clustering and censoring upon subsequent pregnancies also had negligible influence on model performance. We found important differences in model performance between the models developed in the cohort of first deliveries and the random sample of deliveries.

CONCLUSIONS

In our study, a model including first deliveries had the best predictive accuracy but was not generalizable to women of varying parities. Moreover, including repeat pregnancies did not improve the predictive accuracy of the models. Multiple models may be needed to improve the transportability and accuracy of prediction models when the outcome of interest is influenced by parity.

Infant pacifier sanitization and risk of challenge-proven food allergy: A cohort study

BACKGROUND

Environmental microbial exposure plays a role in immune system development and susceptibility to food allergy.

OBJECTIVE

We sought to investigate whether infant pacifier use during the first postnatal year, with further consideration of sanitization, alters the risk of food allergy by age 1 year.

METHODS

The birth cohort recruited pregnant mothers at under 28 weeks' gestation in southeast Australia, with 894 families followed up when infants turned 1 year. Infants were excluded if born under 32 weeks, with a serious illness, major congenital malformation, or genetic disease. Questionnaire data, collected at recruitment and infant ages 1, 6, and 12 months, included pacifier use and pacifier sanitization (defined as the joint exposure of a pacifier and cleaning methods). Challenge-proven food allergy was assessed at 12 months.

RESULTS

Any pacifier use at 6 months was associated with food allergy (adjusted odds ratio, 1.94; 95% CI, 1.04-3.61), but not pacifier use at other ages. This overall association was driven by the joint exposure of pacifier-antiseptic use (adjusted odds ratio, 4.83; 95% CI, 1.10-21.18) compared with no pacifier use. Using pacifiers without antiseptic at 6 months was not associated with food allergy. Among pacifier users, antiseptic cleaning was still associated with food allergy (adjusted odds ratio, 3.56; 95% CI, 1.18-10.77) compared with no antiseptic use. Furthermore, persistent and repeated antiseptic use over the first 6 months was associated with higher food allergy risk (P = .029).

CONCLUSIONS

This is the first report of a pacifier-antiseptic combination being associated with a higher risk of subsequent food allergy. Future work should investigate underlying biological pathways.

Age and life expectancy clocks based on machine learning analysis of mouse frailty

The identification of genes and interventions that slow or reverse aging is hampered by the lack of non-invasive metrics that can predict the life expectancy of pre-clinical models. Frailty Indices (FIs) in mice are composite measures of health that are cost-effective and non-invasive, but whether they can accurately predict health and lifespan is not known. Here, mouse FIs are scored longitudinally until death and machine learning is employed to develop two clocks. A random forest regression is trained on FI components for chronological age to generate the FRIGHT (Frailty Inferred Geriatric Health Timeline) clock, a strong predictor of chronological age. A second model is trained on remaining lifespan to generate the AFRAID (Analysis of Frailty and Death) clock, which accurately predicts life expectancy and the efficacy of a lifespan-extending intervention up to a year in advance. Adoption of these clocks should accelerate the identification of longevity genes and aging interventions.

Temporal trends in behavioral risk and protective factors and their association with mortality rates: results from Brazil and Argentina

Background

Despite available information on trends in behavioral factors for Brazil and Argentina, little is known about the association of these trends with mortality. Understanding this association is important to avoid early deaths. Therefore, we aimed to evaluate temporal trends in behavioral risk and protective factors in Brazil and Argentina, and to assess their association with overall and cause-specific mortality rates.

Methods

Ecological study with data from two population surveys from Brazil and Argentina. Weighted prevalence of tobacco smoking, excessive alcohol consumption, fruit and vegetable consumption and physical activity for the 27 Brazilian state capitals and for the 23 Argentinean provinces were used as behavioral factors. Information on overall mortality as well as cardiovascular diseases and cancer mortality for the year 2015 was collected from national mortality banks of both countries. Estimated prevalence rates were used to describe trends in behavioral factors from 2006 to 2014 in Brazil, and from 2005 to 2013 in Argentina, while Pearson’s correlation and linear regression models were used to assess their association with overall and cause-specific mortality rates.

Results

Brazil presented improvements in behavioral risk and protective factors: sharp decrease in tobacco smoking prevalence (from 15 to 9%), increase in regular fruit and vegetable consumption (from 28 to 36%), and increase in physical activity (45 to 51%). In Argentina, results were more disappointing: small reduction in tobacco smoking (from 55 to 50%) and decrease in physical activity (from 55 to 45%). In both countries, excessive alcohol consumption remained stable, with increase only among women. The association between behavioral factors and mortality showed that in those Brazilian capitals with higher prevalence of regular consumption of fruits and vegetables, there were lower overall mortality rates. Stratification by gender revealed that significant results were only found among women.

Conclusion

Prevalence of regular consumption of fruits and vegetables increased in Brazilian capitals and was associated with lower overall mortality rate, suggesting a positive impact of Brazilian policies to improve dietary intake patterns on its population’s mortality. Approaches focusing on behavioral factors are especially needed in Argentina to reach similar results of those seen in Brazil.

Dynamic prediction model to identify young children at high risk of future overweight: Development and internal validation in a cohort study

BACKGROUND

Primary prevention of overweight is to be preferred above secondary prevention, which has shown moderate effectiveness.

OBJECTIVE

To develop and internally validate a dynamic prediction model to identify young children in the general population, applicable at every age between birth and age 6, at high risk of future overweight (age 8).

METHODS

Data were used from the Prevention and Incidence of Asthma and Mite Allergy birth cohort, born in 1996 to 1997, in the Netherlands. Participants for whom data on the outcome overweight at age 8 and at least three body mass index SD scores (BMI SDS) at the age of ≥3 months and ≤6 years were available, were included (N = 2265). The outcome of the prediction model is overweight (yes/no) at age 8 (range 7.4-10.5 years), defined according to the sex- and age-specific BMI cut-offs of the International Obesity Task Force.

RESULTS

After backward selection in a Generalized Estimating Equations analysis, the prediction model included the baseline predictors maternal BMI, paternal BMI, paternal education, birthweight, sex, ethnicity and indoor smoke exposure; and the longitudinal predictors BMI SDS, and the linear and quadratic terms of the growth curve describing a child's BMI SDS development over time, as well as the longitudinal predictors' interactions with age. The area under the curve of the model after internal validation was 0.845 and Nagelkerke R² was 0.351.

CONCLUSIONS

A dynamic prediction model for overweight was developed with a good predictive ability using easily obtainable predictor information. External validation is needed to confirm that the model has potential for use in practice.

Harnessing repeated measurements of predictor variables for clinical risk prediction: a review of existing methods

BACKGROUND

Clinical prediction models (CPMs) predict the risk of health outcomes for individual patients. The majority of existing CPMs only harness cross-sectional patient information. Incorporating repeated measurements, such as those stored in electronic health records, into CPMs may provide an opportunity to enhance their performance. However, the number and complexity of methodological approaches available could make it difficult for researchers to explore this opportunity. Our objective was to review the literature and summarise existing approaches for harnessing repeated measurements of predictor variables in CPMs, primarily to make this field more accessible for applied researchers.

METHODS

MEDLINE, Embase and Web of Science were searched for articles reporting the development of a multivariable CPM for individual-level prediction of future binary or time-to-event outcomes and modelling repeated measurements of at least one predictor. Information was extracted on the following: the methodology used, its specific aim, reported advantages and limitations, and software available to apply the method.

RESULTS

The search revealed 217 relevant articles. Seven methodological frameworks were identified: time-dependent covariate modelling, generalised estimating equations, landmark analysis, two-stage modelling, joint-modelling, trajectory classification and machine learning. Each of these frameworks satisfies at least one of three aims: to better represent the predictor-outcome relationship over time, to infer a covariate value at a pre-specified time and to account for the effect of covariate change.

CONCLUSIONS

The applicability of identified methods depends on the motivation for including longitudinal information and the method's compatibility with the clinical context and available patient data, for both model development and risk estimation in practice.

A tutorial on dynamic risk prediction of a binary outcome based on a longitudinal biomarker

Dynamic risk predictions based on all available information are useful in timely identification of high‐risk patients. However, in contrast with time to event outcomes, there is still a lack of studies that clearly demonstrate how to obtain and update predictions for a future binary outcome using a repeatedly measured biomarker. The aim of this study is to give an illustrative overview of four approaches to obtain such predictions: likelihood based two‐stage method (2SMLE), likelihood based joint model (JMMLE), Bayesian two‐stage method (2SB), and Bayesian joint model (JMB). We applied the approaches to provide weekly updated predictions of post–molar gestational trophoblastic neoplasia (GTN) based on age and repeated measurements of human chorionic gonadotropin (hCG). Discrimination and calibration measures were used to compare the accuracy of the weekly predictions. Internal validation of the models was conducted using bootstrapping. The four approaches resulted in the same predictive and discriminative performance in predicting GTN. A simulation study showed that the joint models outperform the two‐stage methods when we increase the within‐ and the between‐patients variability of the biomarker. The applicability of these models to produce dynamic predictions has been illustrated through a comprehensive explanation and accompanying syntax (R and SAS^®).

Incorporating repeated measurements into prediction models in the critical care setting: a framework, systematic review and meta-analysis

BACKGROUND

The incorporation of repeated measurements into multivariable prediction research may greatly enhance predictive performance. However, the methodological possibilities vary widely and a structured overview of the possible and utilized approaches lacks. Therefore, we [1] propose a structured framework for these approaches, [2] determine what methods are currently used to incorporate repeated measurements in prediction research in the critical care setting and, where possible, [3] assess the added discriminative value of incorporating repeated measurements.

METHODS

The proposed framework consists of three domains: the observation window (static or dynamic), the processing of the raw data (raw data modelling, feature extraction and reduction) and the type of modelling. A systematic review was performed to identify studies which incorporate repeated measurements to predict (e.g. mortality) in the critical care setting. The within-study difference in c-statistics between models with versus without repeated measurements were obtained and pooled in a meta-analysis.

RESULTS

From the 2618 studies found, 29 studies incorporated multiple repeated measurements. The annual number of studies with repeated measurements increased from 2.8/year (2000-2005) to 16.0/year (2016-2018). The majority of studies that incorporated repeated measurements for prediction research used a dynamic observation window, and extracted features directly from the data. Differences in c statistics ranged from - 0.048 to 0.217 in favour of models that utilize repeated measurements.

CONCLUSIONS

Repeated measurements are increasingly common to predict events in the critical care domain, but their incorporation is lagging. A framework of possible approaches could aid researchers to optimize future prediction models.

Medical students' empathy positively predicts charitable donation behavior

Empathy is known to motivate prosocial behavior. This relationship, however is complex and influenced by the social context and the type of prosocial behavior. Additionally, empathy is a complex psychological capacity, making it important to examine how different components of empathy influence different prosocial behaviors. The current study uses a unique longitudinal sample to assess how changes in cognitive and affective components of empathy relate to charitable giving. Measures of empathy were collected from medical students in the fall and spring of students' first three years of medical school. After this time, students had the opportunity to donate to charity. Positive changes in students' cognitive empathy predicted their charitable giving, with students who demonstrated greater increases in cognitive empathy giving more money. This study points to an important role for cognitive empathy in certain prosocial behaviors, and suggests that long term changes in empathy influence individual differences in prosocial behavior.