Natural Direct Effects Research Articles

Examining the potential direct cardiovascular benefit of tumor-necrosis factor inhibitor in rheumatoid arthritis: Natural and controlled direct effect analyses.

Tumor necrosis factor inhibitors (TNFi) may have a direct benefit on cardiovascular (CV) disease beyond reducing rheumatoid arthritis (RA) disease activity measured by the Clinical Disease Activity Index (CDAI). We compared TNFi initiators and methotrexate (MTX) monotherapy initiators from the CorEvitas RA registry. Two approaches to the "direct effect" of TNFi beyond CDAI were used. In the natural direct effect (NDE) analysis, the potential CV benefit of TNFi was partitioned into NDE and the natural indirect effect (NIE) mediated by CDAI during the first 6 months. We also estimated the controlled direct effects (CDE), corresponding to the direct benefit of TNFi when CDAI trajectories were hypothetically equalized between the initiators of TNFi and MTX monotherapy at a constant value. Estimates were given on the hazard ratio scale. We identified 5764 initiators of TNFi and 3588 initiators of MTX monotherapy. TNFi initiators were younger (58 vs. 64 years) with a shorter disease duration. Our total effect estimates (TNFi vs. MTX [reference]) were protective in direction (0.76-0.91). The NDE estimate was 0.76 [95% confidence interval (CI) 0.59, 0.98], whereas the NIE estimate was 1.00 [95%CI 1.00, 1.00]. In the CDE analyses accounting for longitudinal CDAI, the CDE estimates was 1.27 [95%CI 0.60, 2.69]. We could not convincingly demonstrate a direct benefit of TNFi outside its impact on CDAI. At present, the emphasis should be on the stringent control of RA disease activity, a known important CV risk factor, regardless of medication choice.

The role of housing characteristics in racial and ethnic disparities in SARS-CoV-2 antibody seropositivity among New York City adults: A population representative study

Black and Latino populations have been disproportionately burdened by COVID-19 morbidity and mortality. Subsidized housing, crowding, and neighborhood poverty might be associated with increased COVID-19 transmission and play a role in observed racial and ethnic disparities, yet research is limited. Our study investigated whether these housing variables mediate the relationship between race and ethnicity and SARS-CoV-2 antibody seropositivity among New York City (NYC) adults. We analyzed data from a SARS-CoV-2 serosurvey (n = 1074), nested within the 2020 cross-sectional NYC Community Health Survey (June–October 2020). We defined SARS-CoV-2 seropositivity as either a positive blood test for SARS-CoV-2 antibodies or a self-reported positive test result. We used causal mediation analyses to test whether subsidized housing, crowding, and neighborhood poverty mediate a relationship between race and ethnicity and seropositivity. After controlling for potential confounding, we found elevated prevalence ratios of SARS-CoV-2 seropositivity among Black (APR = 1.74, 95% CI = 1.10–2.73) and Latino (APR = 1.58, 95% CI = 1.05–2.37) residents compared with White residents and for those living in crowded housing (APR = 1.48, 95% CI = 1.03–2.12) and high-poverty neighborhoods (APR = 1.54, 95% CI = 1.12–2.11) but not for subsidized housing. We observed statistically significant natural direct effects for all three mediators. While living in crowded housing and high-poverty neighborhoods contributed to racial and ethnic disparities in seropositivity the estimated contribution from living in subsidized housing was −9% (Black) and − 14% (Latino). Our findings revealed racial and ethnic disparities in seropositivity of SARS-CoV-2 antibodies among NYC adults. Unlike crowding and neighborhood poverty, living in subsidized housing did not explain racial and ethnic disparities in COVID-19.

Intracranial carotid arteriosclerosis mediates the association between blood pressure and cerebral small vessel disease: the Rotterdam Study

Abstract Background Cerebral arteriosclerosis could explain the physiopathological mechanisms linking high blood pressure (BP) and cerebral small vessel disease (CSVD). Objectives To test the hypothesis that ICAC mediates the association between BP and CSVD. Methods 1458 stroke-free participants from the Rotterdam Study underwent nonenhanced computed tomography to quantify ICAC and brain magnetic resonance imaging scans to assess CSVD. ICAC subtypes included atherosclerotic (intimal) and non-atherosclerotic internal elastic lamina (IEL) calcifications. We analyzed systolic BP (SBP), diastolic BP (DBP), pulse pressure (PP), and mean arterial pressure (MAP). Mediation analysis included a two way decomposition to compute the natural direct effect (NDE), natural indirect effect (NIE) and percentage of mediation (%) of ICAC on the association between BP and CSVD. Results The study population had a mean age of 68.0 years old, and 52% (n=758) of the participants were women. In analyses including participants with predominantly IEL calcification, we observed that larger log-ICAC volume was positively related to a higher pulse pressure (β=0.020; P<0.001), and lower diastolic BP (β=0.024; P=0.001). None of the BP components were associated with log-ICAC volume among participants with predominantly intimal calcifications (β≤0.008; P≥0.060). Among all participants, log-ICAC volume mediated the association of DBP (NIE, 0.003; −14.5%) and PP (NIE,0.003; 16.5%) with log-white matter hyperintensities (log WMH). In participants with IEL calcification, log-ICAC volume mediated the association of DBP with log-WMH (NIE, 0.004, −19.5%); no mediations were observed for intimal ICAC. For Lacunes, in all participants, log-ICAC volume mediated the association of DBP (NIE, −0.015, −40%) and PP (NIE,0.015; 26.9%). In participants with IEL calcification; the NIE was 0.020 (45.8%) for DBP and 0.017 (18.2%) for PP. No interactions were detected. Conclusions ICAC mediated the association between BP and CSVD. Non-atherosclerotic IEL calcification, considered a proxy of arterial stiffness, was the main physiopathological mechanism explaining how BP links to CSVD due to cerebral arteriosclerosis. Funding Acknowledgement Type of funding sources: None.

High cardiovascular disease risk-associated with the incidence of Type 2 diabetes among prediabetics

BackgroundPrediabetes is a risk factor for developing Type 2 diabetes mellitus (T2D). We report on the first cohort study of the association between high cardiovascular diseases (CVD) risk with the incidence of T2D in prediabetics. First, estimate the direct effect of developing T2D on patients with prediabetes who have high CVDs risk; and 2) assess the potential increased risk of developing T2D mediated by statins. MethodsWe conducted a population-based cohort study using a subset of data from the Canadian Primary Care Sentinel Surveillance Network (CPCSSN) from 2000 to 2015. Cox proportional hazards (PH) regressions were conducted to estimate our primary outcome, which is the time to T2D among patients with prediabetes. ResultsFrom the 4995 filtered prediabetic participants identified between 2000 and 2015, 2800 participants were diagnosed with high CVDs risk scores as measured by the Framingham risk score. 2195 participants were non-high CVDs risk controls. The covariate-adjusted hazard ratio (HR) of 1.24 [95% confidence interval (CI), 1.10-1.31] for T2D by CVDs risk among prediabetics was observed. The total effect of CVDs risk on developing T2D was decomposed to a natural direct effect of high CVDs risk HR= 1.18 [95% CI, 1.01-1.48] and an indirect effect through statin therapy of HR= 1.06 [95% CI, 0.97-1.30]. ConclusionPatients with prediabetes and high CVDs risk had a 24% higher chance of developing T2D. The high CVDs risk effect was mediated by statin therapy. Regular monitoring and counselling of prediabetics using statins is likely warranted to prevent the incidence of T2D.

In utero particulate matter exposure in association with newborn mitochondrial ND4L10550A>G heteroplasmy and its role in overweight during early childhood

BackgroundMitochondria play an important role in the energy metabolism and are susceptible to environmental pollution. Prenatal air pollution exposure has been linked with childhood obesity. Placental mtDNA mutations have been associated with prenatal particulate matter exposure and MT-ND4L10550A>G heteroplasmy has been associated with BMI in adults. Therefore, we hypothesized that in utero PM2.5 exposure is associated with cord blood MT-ND4L10550A>G heteroplasmy and early life growth. In addition, the role of cord blood MT-ND4L10550A>G heteroplasmy in overweight during early childhood is investigated.MethodsThis study included 386 mother-newborn pairs. Outdoor PM2.5 concentrations were determined at the maternal residential address. Cord blood MT-ND4L10550A>G heteroplasmy was determined using Droplet Digital PCR. Associations were explored using logistic regression models and distributed lag linear models. Mediation analysis was performed to quantify the effects of prenatal PM2.5 exposure on childhood overweight mediated by cord blood MT-ND4L10550A>G heteroplasmy.ResultsPrenatal PM2.5 exposure was positively associated with childhood overweight during the whole pregnancy (OR = 2.33; 95% CI: 1.20 to 4.51; p = 0.01), which was mainly driven by the second trimester. In addition, prenatal PM2.5 exposure was associated with cord blood MT-ND4L10550A>G heteroplasmy from gestational week 9 – 13. The largest effect was observed in week 10, where a 5 µg/m3 increment in PM2.5 was linked with cord blood MT-ND4L10550A>G heteroplasmy (OR = 0.93; 95% CI: 0.87 to 0.99). Cord blood MT-ND4L10550A>G heteroplasmy was also linked with childhood overweight (OR = 3.04; 95% CI: 1.15 to 7.50; p = 0.02). The effect of prenatal PM2.5 exposure on childhood overweight was mainly direct (total effect OR = 1.18; 95% CI: 0.99 to 1.36; natural direct effect OR = 1.20; 95% CI: 1.01 to 1.36)) and was not mediated by cord blood MT-ND4L10550A>G heteroplasmy.ConclusionsCord blood MT-ND4L10550A>G heteroplasmy was linked with childhood overweight. In addition, in utero exposure to PM2.5 during the first trimester of pregnancy was associated with cord blood MT-ND4L10550A>G heteroplasmy in newborns. Our analysis did not reveal any mediation of cord blood MT-ND4L10550A>G heteroplasmy in the association between PM2.5 exposure and childhood overweight.

Metal mixture mediates associations between socioeconomic indicators and blood pressure outcomes in rural Bangladeshi adults: Evidence from a prospective population study

BACKGROUND AND AIM: There is limited evidence on the causal mechanism in the socioeconomic differentials in blood pressure (BP) in low and middle-income countries (LMICs). We, therefore, examined the associations of socioeconomic indicators (SEIs) and BP, socioeconomic disparities in metal mixture exposures, and the potential mediation effect of metal mixtures in the associations between SEIs and BP outcomes in rural Bangladeshi adults. METHODS: This study included 900 rural Bangladeshi adults prospectively followed for six years in the Bangladesh vitamin E and selenium Trial (BEST). Blood pressure was measured at baseline and during three biennial follow-up examinations. Blood levels of arsenic, lead, selenium, and urinary arsenic, including socioeconomic indicators-years of education, and agricultural land ownership (ALO), were measured at baseline. Weighted quantile sum (WQS) regression was used to estimate the joint effects of metal mixture on systolic blood pressure (SBP), diastolic blood pressure (DBP), pulse pressure (PP), and mean arterial pressure (MAP) taken at the third follow-up. Causal mediation effect of the metal mixture was estimated using the parametric g-computation method. RESULTS: Education was positively, but ALO was inversely associated with BP. Participants with higher SEIs were less likely to be exposed to metal mixtures than their counterparts. Metal mixture mediated the associations of SEIs with SBP, DBP, MAP, and PP. For example, the randomized analogue of total natural direct effect (rRTNDE) was 5.31 (95% CI: 1.91, 8.77) and the pure natural indirect effect (rRPNIE) was -0.67 (95% CI: -1.34, -0.22) for education and SBP. Similarly, the rRTNDE was -2.31 (95% CI: -4.37, -0.33) and the rRPNIE was -0.29 (95% CI: -0.70, -0.03) for ALO and SBP. There was no evidence of the exposure-mediator interaction in the mediation models (Pint>0.20). CONCLUSIONS: Our study suggests that metal mixture mediates socioeconomic differentials in BP. KEYWORDS: Blood Pressure; Metals; SES; Mediation; Mixture

The association between prenatal greenspace exposure and Autism spectrum disorder, and the potentially mediating role of air pollution reduction: A population-based birth cohort study.

Autism spectrum disorder (ASD) incidence has increased in past decades. ASD etiology remains inconclusive, but research suggests genetic, epigenetic, and environmental contributing factors and likely prenatal origins. Few studies have examined modifiable environmental risk factors for ASD, and far fewer have examined protective exposures. Greenspace has been associated with positive child development, but very limited greenspace research has examined ASD risk or prenatal exposures. Only one ecological study in 2017 has evaluated the association between greenspace and ASD, observing protective benefits. Greenspace may have direct effects on ASD risk and indirect effects by reducing air pollution exposure, a growing suspected ASD risk factor. To measure the association between prenatal greenspace exposure and ASD risk and examine if reduced air pollution levels in areas of higher greenspace mediate this association. We linked a population-based birth cohort of all deliveries in Metro Vancouver, Canada, from 2004 to 2009, with follow-up to 2014. Diagnoses were based on Autism Diagnostic Observation Schedule and Autism Diagnostic Interview-Revised instruments. Greenspace was quantified as the average of the annual mean Normalized Difference Vegetation Index (NDVI) within a 250m buffer of a residential postal code. Air pollutant exposures-particulate matter with a diameter less than 2.5µm (PM2.5), nitric oxide (NO), and nitrogen dioxide (NO2)-were derived from previously developed and temporally adjusted land use regression models. We estimated air pollutant exposures as the mean concentration per month during pregnancy. We calculated odds ratios (ORs) using logistic regression per NDVI interquartile range (IQR) increase, adjusting for child sex, birth month and year, maternal age and birthplace, and neighborhood-level urbanicity and income. To estimate the health impact of greenspace on ASD at the population level, we used the logistic regression model and marginal standardization to derive risk differences (RDs). Lastly, to quantify the mediating effect of greenspace on ASD risk through air pollution reduction, we used marginal structural models and a potential outcomes framework to calculate marginal risk differences (RDs) to decompose the total effect of greenspace on ASD into natural direct and indirect effects. Of 129,222 births, 1,921 (1.5%) children were diagnosed with ASD. The adjusted OR for ASD per NDVI IQR (0.12) increase was 0.96 (95% CI: 0.90, 1.02) in 250m buffer zones and 0.94 (95% CI: 0.89, 1.00) in 100m buffer zones. On the additive scale, the adjusted RDs were null. Natural direct, natural indirect, and total effect RDs were null for PM2.5, NO, and NO2 mediation models. Prenatal greenspace exposure was associated with reduced odds of ASD, but in the additive scale, this effect was null at the population level. No mediating effect was observed through reduced air pollution, suggesting that air pollution may act as a confounder rather than as a mediator.

33028 Direct and indirect treatment effect of guselkumab on anxiety and depression in patients with psoriasis by adjusting for Psoriasis Area and Severity Index (PASI) – A mediation analysis

Introduction: Guselkumab effectively treats psoriasis and is associated with improvements in anxiety and depression symptoms. Whether improvements in anxiety and depression are due to resolution of skin symptoms or the direct result of guselkumab treatment is unclear. Methods: Patients received guselkumab at Week(W)0, W4, then Q8W; placebo (W16→guselkumab); or adalimumab for moderate-to-severe psoriasis. Mediation analysis evaluated direct effects of guselkumab on change from baseline in Hospital Anxiety and Depression Scale (HADS) score at W24, after adjusting for indirect effects mediated by clinical assessment of skin clearance using Psoriasis Area and Severity Index (PASI) improvement and achievement of ≥90% improvement from baseline (PASI90). Regression analyses were used to determine 95% confidence interval (CI). Results: Improvements in HADS were partially mediated indirectly through skin response among guselkumab-treated patients at W24. Compared with adalimumab, HADS anxiety score was mediated through PASI improvement by 25.5%, with independent treatment effect of -0.74 (95% CI -1.22,-0.27; P < .002), or 74.5% of total effect. Similarly, treatment effect was mediated by 24% through PASI90 response, with independent treatment effect of -0.76 (95% CI -1.24, -0.28; P < .002), or 76% of total effect. Treatment effect of guselkumab (versus adalimumab) on change in HADS depression total score was mediated through PASI improvement by 49.8%; through PASI90, 40.5%. Guselkumab and adalimumab had similar natural direct effects on changes in depression score. Conclusions: Guselkumab demonstrated an independent treatment response in reducing anxiety and depression after adjusting for clinical response of skin clearance. The mechanism of the direct effect of guselkumab beyond skin clearance warrants additional research.

O7-8 Mediation effect of physical activity and sedentary behavior on the association of gender with quality of life in adolescents

BackgroundIf boys and girls are known to have different levels of quality of life (HRQoL), less is known about behavioural factors such as physical activity (PA), sedentary behaviour (SB) that can explain it. We aimed to analyse the mediation effect of PA and SB on the association of sex and HRQoL among adolescents.Methods2448 adolescents (1378, 56.3% of Girls) from the PRALIMAP trial with 2-year follow-up were included. HRQoL (physical, mental, social and general dimension scores; range 0-100), PA (walking, moderate vigorous and total) and SB (sitting time) were assessed using the Duke Health profile and the IPAQ questionnaires, respectively. The 2-year mean PA and SB were calculated and categorized into high or low PA (? 1hour/day) or SB (2 hours/day out of school). We conducted a mediation analysis to investigate the causal mechanism of gender (reference=Girls) on HRQoL through PA and SB by estimating total effect (TE), natural direct effect (NDE), natural indirect effects (NIE) as well as proportion mediated (PM: proportion of sex-HRQoL relationship mediated by PA or SB).ResultsThe direct relationship of gender and HRQoL (NDE) was significant whatever the HRQoL dimension. When using vigorous PA as mediator, PM was estimated at 13.7% (TE:11.0 [9.5 to 12.5]; NIE:1.5[0.9 to 2.1]), 19.4% (TE: 12.3 [10.6 to 14.1]; NIE: 2.4 [1.6 to 3.1]), 70.0% (TE:4.6 [3.1 to 6.2] ; NIE: 3.3 [2.5 to 4.0]) and 25.6% (TE: 9.3 [8.1 to 10.5]; NIE:2.4 [1.8 to 2.9]) for physical, mental, social and general HRQoL, respectively. Similar but less important mediation effect was observed for moderate and walking PA. For Total PA, PM was estimated 7.0%, 8.7%, 47.4% and 14.7% for physical, mental, social and general HRQoL while for SB, the mediation was inverse and less important (PM: -1.5%, -1.4%, -2.0% and -1.5% for physical, mental, social and general HRQoL respectively).ConclusionHRQoL is significantly higher in boys compared to girls among adolescents and this difference is shown to be partially mediated by differences in PA practice, especially vigorous PA. Additionally to the impact of adolescent PA in preventing non-communicable diseases in adulthood, it can also enhance their well-being.

Bayesian joint modeling for causal mediation analysis with a binary outcome and a binary mediator: Exploring the role of obesity in the association between cranial radiation therapy for childhood acute lymphoblastic leukemia treatment and the long-term risk of insulin resistance

Mediation analysis with a binary outcome is notoriously more challenging than with a continuous outcome. A new Bayesian approach for performing causal mediation with a binary outcome and a binary mediator, named the t-link approach, is introduced. This approach relies on the Bayesian multivariate logistic regression model introduced by O'Brien and Dunson (2004) and its Student-t approximation. By re-expressing the Mediation Formula, it is shown how to use this multivariate latent model for estimating the natural direct and indirect effects of an exposure on an outcome in any measure scale of interest (e.g., odds or risk ratio, risk difference). The t-link mediation approach has several valuable features which, to our knowledge, are not found together in existing binary-binary mediation analysis approaches. In particular, it allows for sensitivity analyses regarding the impact of unmeasured mediator-outcome confounders on the natural effects estimates. The proposed mediation approach was evaluated and compared with two other benchmark approaches using simulated data. Results revealed the usefulness of the t-link mediation approach when the sample size is small or moderate. Lastly, the t-link approach was applied for assessing the impact of cranial radiation therapy given to treat childhood acute lymphoblastic leukemia on the long-term risk of insulin resistance, where this effect is possibly mediated by obesity.

Health problems, turnover intention, and actual turnover among shift work female nurses: Analyzing data from a prospective longitudinal study.

This study investigated health problems, turnover intention, and actual turnover among shift work nurses. While turnover intention is often used as a proxy variable for turnover, the relationship between these variables requires clarification. This study tested for relevant associations using prospective longitudinal data with a time lag of 12 months. We also tested for associations between health problems (sleep disturbance, fatigue, and depression) and turnover intentions/turnover, with a focus on the mediating role of turnover intention. This study conducted a secondary analysis of data from the Shift Work Nurses' Health and Turnover project, which is a prospective longitudinal cohort study. We analyzed health problems, turnover intention, and actual turnover. The data were analyzed via descriptive statistics, the Pearson's chi-squared test, independent t-test, univariable logistic regression, multiple logistic regression, and causal mediation. Participants included 491 shift work female nurses. Of these, 112 (22.8%) had turnover intention, while 38 (7.7%) left their jobs within the 12-month period of investigation. Of the 112 with turnover intention, 22 left their jobs (OR 5.68. 95% CI 2.84-11.36). The logistic regression analysis showed that sleep disturbance and fatigue were associated with turnover intentions and actual turnover, while depression was only associated with turnover intention. The causal mediation analysis showed that turnover intention mediated the relationship between health problems (sleep disturbance and fatigue) and actual turnover (sleep disturbance OR 1.31, 95%CI = 1.02-1.60; fatigue OR 2.11, 95%CI = 1.50-2.68); sleep disturbance had a natural direct effect on actual turnover (OR 2.66, 95%CI,2.07-3.21). Turnover intention strongly predicted actual turnover. Sleep disturbances may result in turnover, even in cases without existing turnover intention. These findings highlight the need for early interventions aimed at preventing and alleviating sleep disturbances for shift work female nurses.

Synovitis mediates the association between bone marrow lesions and knee pain in osteoarthritis: data from the Foundation for the National Institute of Health (FNIH) Osteoarthritis Biomarkers Consortium

Although subchondral bone marrow lesions (BMLs) and synovitis have been well acknowledged as important sources of pain in knee osteoarthritis (KOA), it is unclear if synovitis plays the mediating role in the relationship between BMLs and knee pain. We analyzed 600 subjects with magnetic resonance imaging (MRI) in the Foundation for National Institutes of Health Osteoarthritis Biomarkers Consortium (FNIH) cohort at baseline and 24-month. BMLs and synovitis were measured according to the MRI Osteoarthritis Knee Score (MOAKS) scoring system. BMLs were scored in five subregions. A summary synovitis score of effusion and Hoffa-synovitis was calculated. Knee pain was evaluated using the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC). Linear regression models were applied to analyze the natural direct effect (NDE) of BMLs and synovitis with knee pain, respectively, and natural indirect effect (NIE) mediated by synovitis. 590 participants (58.8% females, with a mean age of 61.5) were included in the present analyses. For NDE, knee pain was cross-sectionally associated with medial femorotibial BMLs (β=0.23, 95% CI: 0.09, 0.38) and synovitis (β=0.40, 95% CI: 0.20, 0.60). Longitudinal associations retained significant [medial femorotibial BMLs (β=0.37, 95% CI: 0.21, 0.53); synovitis (β= 0.72, 95% CI: 0.45, 0.99)]. In the NIE analyses, synovitis mediated the association between medial femorotibial BML and knee pain at baseline (β=0.051, 95% CI: 0.01, 0.09) and over 24 months (β=0.079, 95% CI: 0.023, 0.15), with the mediating proportion of 17.8% and 22.4%, respectively. Synovitis partially mediates the association between medial femorotibial BMLs and knee pain.

Identification and robust estimation of swapped direct and indirect effects: Mediation analysis with unmeasured mediator-outcome confounding and intermediate confounding.

Counterfactual-model-based mediation analysis can yield substantial insight into the causal mechanism through the assessment of natural direct effects (NDEs) and natural indirect effects (NIEs). However, the assumptions regarding unmeasured mediator-outcome confounding and intermediate mediator-outcome confounding that are required for the determination of NDEs and NIEs present practical challenges. To address this problem, we introduce an instrumental blocker, a novel quasi-instrumental variable, to relax both of these assumptions, and we define a swapped direct effect (SDE) and a swapped indirect effect (SIE) to assess the mediation. We show that the SDE and SIE are identical to the NDE and NIE, respectively, based on a causal interpretation. Moreover, the empirical expressions of the SDE and SIE are derived with and without an intermediate mediator-outcome confounder. Then, a multiply robust estimation method is derived to mitigate the model misspecification problem. We prove that the proposed estimator is consistent, asymptotically normal, and achieves the semiparametric efficiency bound. As an illustration, we apply the proposed method to genomic datasets of lung cancer to investigate the potential role of the epidermal growth factor receptor in the treatment of lung cancer.

Sensitivity analysis for direct and indirect effects in the presence of exposure-induced mediator-outcome confounders.

Questions of mediation are often of interest in reasoning about mechanisms, and methods have been developed to address these questions. However, these methods make strong assumptions about the absence of confounding. Even if exposure is randomized, there may be mediator-outcome confounding variables. Inference about direct and indirect effects is particularly challenging if these mediator-outcome confounders are affected by the exposure because in this case these effects are not identified irrespective of whether data is available on these exposure-induced mediator-outcome confounders. In this paper, we provide a sensitivity analysis technique for natural direct and indirect effects that is applicable even if there are mediator-outcome confounders affected by the exposure. We give techniques for both the difference and risk ratio scales and compare the technique to other possible approaches.

Methods of analysis for survival outcomes with time-updated mediators, with application to longitudinal disease registry data.

Mediation analysis is a useful tool to illuminate the mechanisms through which an exposure affects an outcome but statistical challenges exist with time-to-event outcomes and longitudinal observational data. Natural direct and indirect effects cannot be identified when there are exposure-induced confounders of the mediator-outcome relationship. Previous measurements of a repeatedly-measured mediator may themselves confound the relationship between the mediator and the outcome. To overcome these obstacles, two recent methods have been proposed, one based on path-specific effects and one based on an additive hazards model and the concept of exposure splitting. We investigate these techniques, focusing on their application to observational datasets. We apply both methods to an analysis of the UK Cystic Fibrosis Registry dataset to identify how much of the relationship between onset of cystic fibrosis-related diabetes and subsequent survival acts through pulmonary function. Statistical properties of the methods are investigated using simulation. Both methods produce unbiased estimates of indirect and direct effects in scenarios consistent with their stated assumptions but, if the data are measured infrequently, estimates may be biased. Findings are used to highlight considerations in the interpretation of the observational data analysis.

Posttreatment confounding in causal mediation studies: A cutting-edge problem and a novel solution via sensitivity analysis.

In causal mediation studies that decompose an average treatment effect into indirect and direct effects, examples of posttreatment confounding are abundant. In the presence of treatment-by-mediator interactions, past research has generally considered it infeasible to adjust for a posttreatment confounder of the mediator-outcome relationship due to incomplete information: for any given individual, a posttreatment confounder is observed under the actual treatment condition while missing under the counterfactual treatment condition. This paper proposes a new sensitivity analysis strategy for handling posttreatment confounding and incorporates it into weighting-based causal mediation analysis. The key is to obtain the conditional distribution of the posttreatment confounder under the counterfactual treatment as a function of not only pretreatment covariates but also its counterpart under the actual treatment. The sensitivity analysis then generates a bound for the natural indirect effect and that for the natural direct effect over a plausible range of the conditional correlation between the posttreatment confounder under the actual and that under the counterfactual conditions. Implemented through either imputation or integration, the strategy is suitable for binary as well as continuous measures of posttreatment confounders. Simulation results demonstrate major strengths and potential limitations of this new solution. A reanalysis of the National Evaluation of Welfare-to-Work Strategies (NEWWS) Riverside data reveals that the initial analytic results are sensitive to omitted posttreatment confounding.

The Mediating Effect of Sleep Duration on the Association between Food Insecurity and Childhood Obesity.

Background: Childhood obesity has increased globally during the past four decades. Food insecurity could heighten the risk of obesity. However, little is known about the underlying mechanism. This study aims to investigate the mediating role of sleep duration in the association between food insecurity and childhood obesity and whether there are differences by age, sex, and race/ethnicity. Method: Data from the National Survey of Children's Health for the years 2016-2020 were used, including children 10-17 years of age. We employed causal mediation analysis within a counterfactual framework to decompose the total effect of food insecurity into natural direct and indirect effects and estimate the proportion mediated. Result: The prevalence of obesity was 15.8% in our study sample. Children with food insecurity had 78% higher odds [odds ratio (OR) = 1.78, 95% confidence interval (CI): 1.70-1.86] of having obesity compared with children who were food secure. Overall, only 6.13% of this association was mediated by sleep duration. The association between food insecurity and obesity was stronger in females (OR = 1.96, 95% CI: 1.84-2.10) than males (OR = 1.66, 95% CI: 1.56-1.75), but the proportion mediated by sleep duration was larger in males (7.13%) than females (5.22%). Evidence of mediation was more pronounced in children 10-11 years of age and non-Hispanic Asian children (proportion mediated = 14.85% and 11.21%, respectively). Conclusion: Food insecurity is associated with an increased prevalence of obesity among children. Although a small proportion of this association is mediated by sleep duration, these results suggest that sleep should be considered when assessing the link between food insecurity and childhood obesity.

Intakes of Vegetables and Potatoes and Associations with Incident Diabetes

ObjectivesHigher vegetable intakes are hypothesized to reduce the risk of developing diabetes but the findings are weak. Using Diet, Cancer and Health cohort, we aimed to 1) analyze the relationship between intakes of total vegetable, vegetable subgroups and incident diabetes, 2) analyze separately the association between intakes of total potatoes (excluding fries/chips) and incident diabetes, and 3) understand whether the relationship between vegetable intake and incident diabetes is mediated by baseline BMI.MethodsCross-sectional associations between exposures (baseline intakes of total vegetables, vegetable subgroups, and potatoes, estimated from a food frequency questionnaire), and baseline BMI were assessed by multivariable-adjusted linear regression models. Non-linear associations between exposures and incident diabetes were examined by multivariable-adjusted Cox proportional hazards models with restricted cubic splines. Mediation by BMI was quantified through natural direct and indirect effects.ResultsAmong 54,984 participants (median age: 56 years), 7836 diabetes cases were recorded during median follow-up of 16.3 years. Participants in the highest total vegetable intake quintile (median: 319 g/d) had a 0.29 kg/m2 (95% CI: −0.39, −0.18) lower BMI and a 20% (95% CI: 14, 25%) lower risk of incident diabetes after multivariable adjustments, compared to those in the lowest quintile (median: 67 g/d). Baseline BMI mediated 20% (95% CI: 13, 31%) of the association between vegetable intakes and diabetes. Conversely, those in the highest potato intake quintile (median: 256 g/d) had a 9% (95% CI: 2, 16%) higher risk of diabetes after multivariable adjustments and compared to those in the lowest quintile (median: 52 g/d). Of the vegetable subclasses, higher intakes of green leafy, cruciferous, yellow/orange/red, other vegetables and legumes were significantly associated with a lower risk of diabetes. Stratified analyses on the absolute scale suggest that those at a higher risk of diabetes (BMI ≥30) may benefit the most from higher vegetable intakes.ConclusionsThe findings provide evidence that a higher vegetable, but not potato, intake might help to mitigate diabetes risk, partly by reducing BMI.Funding SourcesPratik Pokharel is supported by an Edith Cowan University Higher Degree by Research Scholarship Australia.

1204-P: Evaluation of Mediation via the Physical Activity Environment of Neighborhood, Socioeconomic Environment, and Type 2 Diabetes Associations

Inequitable access to neighborhood leisure-time physical activity (LTPA) resources may explain geographic disparities in type 2 diabetes (T2D) . We evaluated the LTPA environment as a potential mediating pathway through which the neighborhood socioeconomic environment (NSEE) affects T2D. We used harmonized measures and analytic approaches in three independent study samples in the Diabetes Location, Environmental Attributes, and Disparities Network: the national Veterans Administration Diabetes Risk (VADR) cohort, comprising electronic health records (EHR) of 4.1 million T2D-free veterans; Reasons for Geographic and Racial Differences in Stroke (REGARDS) , a national prospective cohort (n=11,2T2D-free) ; and a case-control study in the Geisinger EHR in Pennsylvania (15,888 T2D cases) . We measured NSEE with an index of census tract-level Census metrics (higher indicates disadvantage) . The LTPA environment was evaluated by PA facility density within street network buffers and population-weighted distance to parks. In models stratified by levels of community urbanicity, we estimated natural direct effects and indirect effects for each mediator. PA facility density mediated NSEE-T2D effects, but the magnitude and direction of the indirect effects differed by urbanicity and study sample. In rural communities, we found positive indirect effects via PA facility density (incidence rate difference [95% confidence interval] of NSEE quartile 4 vs. 1: 0.0[0.0038, 0.0099] in VADR; 1.53 [0.25, 3.05] in REGARDS) , representing a proportion mediated of 3% (VADR) and 51% (REGARDS) . We found small indirect effects for all other community types in VADR but none in the Geisinger sample. Distance to parks mediated NSEE-T2D effects only in VADR rural communities. Results suggest regional or population-specific differences in how and whether the LTPA environment mediates NSEE-T2D relations, which could support local T2D prevention efforts. Disclosure K.A. Moon: None. C.M. Nordberg: None. A. Zhu: Research Support; Amgen Inc. J. Uddin: None. P.M. Lopez: None. S.L. Orstad: None. M.D. Schwartz: None. V. Ryan: None. B.S. Schwartz: None. A.P. Carson: Research Support; Amgen Inc. D. Long: Research Support; Amgen Inc. J.R. Brown: None. G.S. Lovasi: None. S. Adhikari: Research Support; Johnson &amp; Johnson. R. Kanchi: None. S. Avramovic: None. K.R. Siegel: None. M.N. Poulsen: None. Funding Centers for Disease Control (U01DP006293, Drexel University; U01DP006296, Geisinger-Johns Hopkins University; U01DP006299, New York University School of Medicine; U01DP006302, University of Alabama at Birmingham) .

Association between Parental Feeding Styles and Excess Weight, and Its Mediation by Diet, in Costa Rican Adolescents.

Background. Whereas parental feeding styles (PFS) influence children’s diet, less is known about this relationship in adolescents. Methods. A cross-sectional study in 686 Costa Rican adolescents (13–18 years) evaluated 4 validated PFS scores: healthy eating verbal encouragement; scolding; directly controlling diet; instrumental/emotional. Diet was evaluated through 3-day food records, deriving a Traditional Costa Rica Adolescents Diet Score (TCRAD). Excess weight (EW) measured by BMI was dichotomized following standards. Regression-based mediation analysis estimated the overall and sex-stratified odds ratios of EW for natural direct (NDE), natural indirect (NIE), and total effects (TE) of the pathway PFS→TCRAD→EW. Results. A one-unit increase in the direct control PFS score was associated with higher EW odds overall [(TE: 1.55; 95% CI: 1.04–2.31; p-value = 0.033), (NDE: 1.52; 95% CI: 1.02–2.27; p-value = 0.039)], and in boys [(TE: 2.13; 95% CI: 1.04–4.38; p-value = 0.039), (NDE: 2.10; 95% CI: 1.03–4.31; p-value = 0.042)]. Non-significant mediation by TCRAD was observed for the healthy eating verbal encouragement PFS overall (p-value = 0.06). Associations for the instrumental/emotional and scolding PFS were not significant. Conclusions. Direct diet control from parents may contribute to adolescents’ excess weight, particularly among boys. Parents encouraging healthy eating might support adolescents’ healthy weight through a healthy diet. Longitudinal research should clarify the association between PFS and diet-related outcomes among diverse adolescents.