Mediation analysis with causally ordered mediators using Cox proportional hazards model

Insulin resistance potentiates the effect of remnant cholesterol on cardiovascular mortality in individuals without diabetes

BACKGROUND AND AIMS

Remnant cholesterol (RC) and insulin resistance (IR) have been independently associated with cardiovascular risk. Here, we evaluated the role of IR and RC on cardiovascular disease (CVD) mortality.

METHODS

We conducted an analysis of 16,113 individuals ≥20 years without diabetes from the National Health and Nutrition Examination Survey (NHANES-III/IV). RC levels were calculated using total cholesterol, non-HDL-c, and LDL-c; IR was defined as HOMA2-IR≥2.5 and CVD mortality as a composite of cardiovascular and cerebrovascular mortality. Multiple linear regression was used to assess the relationship between HOMA2-IR and RC and Cox regression models to assess their joint role in CVD mortality. Causally ordered mediation models were used to explore the mediating role of IR in RC-associated CVD mortality.

RESULTS

We identified an association between higher HOMA2-IR and higher RC levels. The effect of IR on CVD mortality was predominant (HR 1.32, 95%CI 1.18-1.48) and decreased at older ages (HR 0.934, 95%CI 0.918-0.959) compared to RC (HR 0.983, 95%CI 0.952-1.014). Higher risk of CVD mortality was observed in individuals with IR but normal RC (HR 1.37, 95%CI 1.25-1.50) and subjects with IR and high RC (HR 1.24, 95%CI 1.13-1.37), but not in subjects without IR but high RC. In mediation models, HOMA2-IR accounted for 78.2% (95%CI 28.11-98.89) of the effect of RC levels on CVD mortality.

CONCLUSIONS

Our findings suggest that RC potentiates the risk of CVD mortality through its effect on whole-body insulin sensitivity, particularly among younger individuals.

Exposure to greenness during pregnancy and the first three years after birth and autism spectrum disorder: A matched case-control study in shanghai, China

Causes of autism spectrum disorder (ASD) have not been fully understood. Previous studies have linked environmental factors with ASD. However, evidence for the greenness-ASD association is limited, especially in China. To fill this gap, we conducted a matched case-control study to examine the association between greenness and ASD in China. Participants in this study were 84,934 children aged 3-12 years in Shanghai, China, selected using a multi-stage cluster sampling method. ASD cases were firstly screened by questionnaires completed by both children's parents and teachers, and were then confirmed by clinical examinations. Further, 10 healthy controls were randomly selected to match each ASD case by age and sex. The final analyses included 146 ASD cases and 1460 healthy controls. Participants' exposure to greenness before and after birth was assessed by normalized difference vegetation index (NDVI) and enhanced vegetation index (EVI) from NASA's Earth Observing System according to their residential locations. We used conditional logistic regression to examine the ASD-greenness association. Per interquartile range (IQR) increase in EVI500m and NDVI500m during the year before birth were associated with lower risks of ASD with adjusted odds ratios (ORs) and 95% confidence intervals (CIs) of 0.96 (95%CI: 0.946, 0.975, IQR = 0.074) and 0.937 (95%CI: 0.915, 0.959, IQR = 0.101). Exposure to greenness during the first 3 years after birth was also significantly associated with lower risk of ASD [IQR ORs for EVI500m and NDVI500m were 0.935 (95%CI: 0.91, 0.962, IQR = 0.06) and 0.897 (95%CI: 0.861, 0.935, IQR = 0.09), respectively]. Air pollution showed mediation effects on thegreenness-ASD association. Greenness was observed to have stronger beneficial effects on children without historical diseases and term birth. More greenness exposure before and after birth were significantly associated with lower risks of ASD in children. Our results highlight the importance of greenness in urban planning.

On the Conventional Definition of Path-specific Effects: Fully Mediated Interaction With Multiple Ordered Mediators

Path-specific effects are a critical measure for assessing mediation in the presence of multiple mediators. However, the conventional definition of path-specific effects has generated controversy because it often causes misinterpretation of the results of multiple mediator analysis. For in-depth analysis of this issue, we propose the concept of decomposing fully mediated interaction from the average causal effect. We show that misclassification of fully mediated interaction is the main cause of misinterpretation of path-specific effects. We propose two strategies for specifying fully mediated interaction: isolating and reclassifying fully mediated interaction. The choice of strategy depends on the objective. Isolating fully mediated interaction is the superior strategy when the main objective is elucidating the mediation mechanism, whereas reclassifying it is superior when the main objective is precisely interpreting the mediation analysis results. To compare performance, this study used the two proposed strategies and the conventional decomposition strategy to analyze the mediating roles of dyspnea and anxiety in the effect of impaired lung function on poor health status in a population of patients with chronic obstructive pulmonary disease. The estimation result showed that the conventional decomposition strategy underestimates the importance of dyspnea as a mechanism of this disease. Specifically, the strategy of reclassifying fully mediated interaction revealed that 50% of the average causal effect is attributable to mediating effects, particularly the mediating effect of dyspnea.

Prediction of Gestational Diabetes Mellitus under Cascade and Ensemble Learning Algorithm

Gestational diabetes mellitus (GDM) is one of the risk factors for fetal dysplasia and maternal pregnancy difficulties. Therefore, the prediction of the risk of GDM in advance has become a big demand for millions of families. Therefore, machine learning technology is adopted to study GDM prediction. Firstly, the data is preprocessed, and the mean value is used for outlier processing. After preprocessing of the data, the IV value method is used to screen the features. Of the 83 features in the original sample data, 40 important features are screened out through feature engineering. On this basis, Logistics regression model, Lasso-Logistics, Gradient Boosting Decision Tree (GBDT), Extreme Gradient Boosting (Xgboost), Light Gradient Boosting Machine (Lightgbm), and Gradient Boosting Categorical Features (Catboost) are established, and multiple learners are integrated. Finally, the constructed model is tested on data sets. The accuracy of the proposed model is 80.3%, the accuracy is 74.6%, the recall rate is 79.3%, and the running time is only 2.53 seconds. This means that the proposed model is superior to the previous models in terms of accuracy, precision, recall rate, and F1 value, and the time consumption is also in line with the actual engineering requirements. The proposed scheme provides some ideas for the research of machine learning technology in disease prediction.

Path-specific effects in the presence of a survival outcome and causally ordered multiple mediators with application to genomic data

Causal multimediation analysis (i.e. the causal mediation analysis with multiple mediators) is critical for understanding the effectiveness of interventions, especially in medical research. Deriving the path-specific effects of exposure on the outcome through a set of mediators can provide detail about the causal mechanism of interest However, existing models are usually restricted to partial decomposition, which can only be used to evaluate the cumulative effect of several paths. In genetics studies, partial decomposition fails to reflect the real causal effects mediated by genes, especially in complex gene regulatory networks. Moreover, because of the lack of a generalized identification procedure, the current multimediation analysis cannot be applied to the estimation of path-specific effects for any number of mediators. In this study, we derive the interventional analogs of path-specific effect for complete decomposition to address the difficulty of nonidentifiability. On the basis of two survival models of the outcome, we derive the generalized analytic forms for interventional analogs of path-specific effects by assuming the normal distributions of mediators. We apply the new methodology to investigate the causal mechanism of signature genes in lung cancer based on the cell cycle pathway, and the results clarify the gene pathway in cancer.

Mediation analysis for survival data with high-dimensional mediators

MOTIVATION

Mediation analysis has become a prevalent method to identify causal pathway(s) between an independent variable and a dependent variable through intermediate variable(s). However, little work has been done when the intermediate variables (mediators) are high-dimensional and the outcome is a survival endpoint. In this paper, we introduce a novel method to identify potential mediators in a causal framework of high-dimensional Cox regression.

RESULTS

We first reduce the data dimension through a mediation-based sure independence screening method. A de-biased Lasso inference procedure is used for Cox's regression parameters. We adopt a multiple-testing procedure to accurately control the false discovery rate when testing high-dimensional mediation hypotheses. Simulation studies are conducted to demonstrate the performance of our method. We apply this approach to explore the mediation mechanisms of 379 330 DNA methylation markers between smoking and overall survival among lung cancer patients in The Cancer Genome Atlas lung cancer cohort. Two methylation sites (cg08108679 and cg26478297) are identified as potential mediating epigenetic markers.

AVAILABILITY AND IMPLEMENTATION

Our proposed method is available with the R package HIMA at https://cran.r-project.org/web/packages/HIMA/.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

Survival mediation analysis with the death-truncated mediator: The completeness of the survival mediation parameter

In medical research, the development of mediation analysis with a survival outcome has facilitated investigation into causal mechanisms. However, studies have not discussed the death-truncation problem for mediators, the problem being that conventional mediation parameters cannot be well defined in the presence of a truncated mediator. In the present study, we systematically defined the completeness of causal effects to uncover the gap, in conventional causal definitions, between the survival and nonsurvival settings. We propose a novel approach to redefining natural direct and indirect effects, which are generalized forms of conventional causal effects for survival outcomes. Furthermore, we developed three statistical methods for the binary outcome of survival status and formulated a Cox model for survival time. We performed simulations to demonstrate that the proposed methods are unbiased and robust. We also applied the proposed method to explore the effect of hepatitis C virus infection on mortality, as mediated through hepatitis B viral load.

Mediation analysis for mixture Cox proportional hazards cure models

Mediation analysis aims to decompose a total effect into specific pathways and investigate the underlying causal mechanism. Although existing methods have been developed to conduct mediation analysis in the context of survival models, none of these methods accommodates the existence of a substantial proportion of subjects who never experience the event of interest, even if the follow-up is sufficiently long. In this study, we consider mediation analysis for the mixture of Cox proportional hazards cure models that cope with the cure fraction problem. Path-specific effects on restricted mean survival time and survival probability are assessed by introducing a partially latent group indicator and applying the mediation formula approach in a three-stage mediation framework. A Bayesian approach with P-splines for approximating the baseline hazard function is developed to conduct analysis. The satisfactory performance of the proposed method is verified through simulation studies. An application of the Alzheimer's disease (AD) neuroimaging initiative dataset investigates the causal effects of APOE-ϵ4 allele on AD progression.

Impact of undiagnosed type 2 diabetes and pre-diabetes on severity and mortality for SARS-CoV-2 infection

INTRODUCTION

Diabetes and hyperglycemia are risk factors for critical COVID-19 outcomes; however, the impact of pre-diabetes and previously unidentified cases of diabetes remains undefined. Here, we profiled hospitalized patients with undiagnosed type 2 diabetes and pre-diabetes to evaluate its impact on adverse COVID-19 outcomes. We also explored the role of de novo and intrahospital hyperglycemia in mediating critical COVID-19 outcomes.

RESEARCH DESIGN AND METHODS

Prospective cohort of 317 hospitalized COVID-19 cases from a Mexico City reference center. Type 2 diabetes was defined as previous diagnosis or treatment with diabetes medication, undiagnosed diabetes and pre-diabetes using glycosylated hemoglobin (HbA1c) American Diabetes Association (ADA) criteria and de novo or intrahospital hyperglycemia as fasting plasma glucose (FPG) ≥140 mg/dL. Logistic and Cox proportional regression models were used to model risk for COVID-19 outcomes.

RESULTS

Overall, 159 cases (50.2%) had type 2 diabetes and 125 had pre-diabetes (39.4%), while 31.4% of patients with type 2 diabetes were previously undiagnosed. Among 20.0% of pre-diabetes cases and 6.1% of normal-range HbA1c had de novo hyperglycemia. FPG was the better predictor for critical COVID-19 compared with HbA1c. Undiagnosed type 2 diabetes (OR: 5.76, 95% CI 1.46 to 27.11) and pre-diabetes (OR: 4.15, 95% CI 1.29 to 16.75) conferred increased risk of severe COVID-19. De novo/intrahospital hyperglycemia predicted critical COVID-19 outcomes independent of diabetes status.

CONCLUSIONS

Undiagnosed type 2 diabetes, pre-diabetes and de novo hyperglycemia are risk factors for critical COVID-19. HbA1c must be measured early to adequately assess individual risk considering the large rates of undiagnosed type 2 diabetes in Mexico.

Association of air pollution and greenness with carotid plaque: A prospective cohort study in China

Previous studies indicated that exposure to air pollution was associated with the progress of atherosclerosis, but evidence is very limited in China and even in the world. This study aims to assess the associations of long-term exposures to air pollution and greenness with the occurrence of carotid plaque. Participants of this cohort study were urban residents and office workers who visited Hebei General Hospital for routine physical examination annually from September 2016 through to December 2018. Eligible participants were people diagnosed the absence of carotid plaque clinically at their first hospital visit and were followed up at their second or third hospital visit. Exposure to particulate matter with aerodynamic diameter less than 2.5 μm (PM2.5), nitrogen dioxide (NO2) and ozone (O3) were estimated using an inverse distance weighted (IDW) method. The level of greenness was assessed using the Normalized Difference Vegetation Index (NDVI) and Enhanced Vegetation Index (EVI). The associations were evaluated using Cox proportional hazards regression models. Among 4,137 participants, 575 showed the occurrence of carotid plaque during the follow-up period. After controlling for potential confounders, the hazard ratios (HRs) and 95% confidence intervals (95%CIs) of carotid plaque associated with per interquartile range (IQR) increase in PM2.5, NO2, and O3 were 1.78 (1.55, 2.03), 1.32 (1.14, 1.53) and 1.99 (1.71, 2.31), respectively. Increased EVI and NDVI were significantly associated with lower risk of carotid plaque [HR (and 95%CI): 0.84 (0.77, 0.93) and 0.87 (0.80, 0.94)]. PM2.5 significantly mediated 80.47% or 93.00% of the estimated association between EVI or NDVI and carotid plaque. In light of the significant associations between air pollution, greenness and carotid plaque in this study, continued efforts are needed to curb air pollution and plan more green space considering their effects on vascular disease.

Cardiovascular risk factor mediation of the effects of education and Genetic Risk Score on cardiovascular disease: a prospective observational cohort study of the Framingham Heart Study

OBJECTIVES

Level of education and genetic risk are key predictors of cardiovascular disease (CVD). While several studies have explored the causal mechanisms of education effects, it remains uncertain to what extent genetic risk is mediated by established CVD risk factors. This study sought to investigate this and explored the mediation of education and genetic effects on CVD by established cardiovascular risk factors in the Framingham Heart Study (FHS).

DESIGN

Prospective observational cohort study.

PARTICIPANTS

7017 participants from the FHS.

SETTING

Community-based cohort of adults in Framingham, Massachusetts, USA.

PRIMARY OUTCOME MEASURE

Incident CVD. The total effects of education and genetic predisposition using a 63-variant genetic risk score (GRS) on CVD, as well as those mediated by established CVD risk factors, were assessed via mediation analysis based on the counterfactual framework using Cox proportional hazards regression models.

RESULTS

Over a median follow-up time of 12.0 years, 1091 participants experienced a CVD event. Education and GRS displayed significant associations with CVD after adjustment for age and sex and the established risk factors smoking, total cholesterol (TC), high-density lipoprotein cholesterol (HDL-C), body mass index, systolic blood pressure (SBP) and diabetes. For education effects, smoking, HDL-C and SBP were estimated to mediate 18.8% (95% CI 9.5% to 43%), 11.5% (95% CI 5.7% to 29.0%) and 4.5% (95% CI 1.6% to 13.3%) of the total effect of graduate degree, respectively, with the collective of all risk factors combined mediating 38.5% (95% 24.1% to 64.9%). A much smaller proportion of the effects of GRS were mediated by established risk factors combined (17.6%, 95% CI 2.4% to 35.7%), with HDL-C and TC mediating 11.5% (95% CI 6.2% to 21.5%) and 3.1% (95% CI 0.2% to 8.3%), respectively.

CONCLUSIONS

Unlike education inequalities, established risk factors mediated only a fraction of GRS effects on CVD. Further research is required to elucidate the underlying causal mechanisms of genetic contributions to CVD.

Predicting Mortality Due to SARS-CoV-2: A Mechanistic Score Relating Obesity and Diabetes to COVID-19 Outcomes in Mexico

BACKGROUND

The SARS-CoV-2 outbreak poses a challenge to health care systems due to its high complication rates in patients with cardiometabolic diseases. Here, we identify risk factors and propose a clinical score to predict COVID-19 lethality, including specific factors for diabetes and obesity, and its role in improving risk prediction.

METHODS

We obtained data of confirmed and negative COVID-19 cases and their demographic and health characteristics from the General Directorate of Epidemiology of the Mexican Ministry of Health. We investigated specific risk factors associated to COVID-19 positivity and mortality and explored the impact of diabetes and obesity on modifying COVID-19-related lethality. Finally, we built a clinical score to predict COVID-19 lethality.

RESULTS

Among the 177 133 subjects at the time of writing this report (May 18, 2020), we observed 51 633 subjects with SARS-CoV-2 and 5,332 deaths. Risk factors for lethality in COVID-19 include early-onset diabetes, obesity, chronic obstructive pulmonary disease, advanced age, hypertension, immunosuppression, and chronic kidney disease (CKD); we observed that obesity mediates 49.5% of the effect of diabetes on COVID-19 lethality. Early-onset diabetes conferred an increased risk of hospitalization and obesity conferred an increased risk for intensive care unit admission and intubation. Our predictive score for COVID-19 lethality included age ≥ 65 years, diabetes, early-onset diabetes, obesity, age < 40 years, CKD, hypertension, and immunosuppression and significantly discriminates lethal from non-lethal COVID-19 cases (C-statistic = 0.823).

CONCLUSIONS

Here, we propose a mechanistic approach to evaluate the risk for complications and lethality attributable to COVID-19, considering the effect of obesity and diabetes in Mexico. Our score offers a clinical tool for quick determination of high-risk susceptibility patients in a first-contact scenario.

Mendelian randomization using semiparametric linear transformation models

Mendelian randomization (MR) uses genetic information as an instrumental variable (IV) to estimate the causal effect of an exposure of interest on an outcome in the presence of unknown confounding. We are interested in the causal effect of cigarette smoking on lung cancer survival, which is subject to confounding by underlying pulmonary functions. Despite the well-developed IV analyses for the continuous and binary outcomes, the scarcity of methodology for the survival outcome limits its utility for the time-to-event data collected in many observational studies. We propose an IV analysis method in the survival context, estimating causal effects on a transformed survival time and survival probabilities using semiparametric linear transformation models. We study the conditions under which hazard ratio and the effect on survival probability can be approximated. For statistical inference, we construct estimating equations to circumvent the difficulty in deriving joint likelihood of the exposure and the outcome, due to the unknown confounding. Asymptotic properties of the proposed estimators are established without parametric assumptions about confounders. We study the finite sample performance in extensive simulation studies. The MR analysis of a lung cancer study suggests a harmful prognostic effect of smoking pack-years that would have been missed by the crude association.

Effect decomposition through multiple causally nonordered mediators in the presence of exposure-induced mediator-outcome confounding

Avin et al (2005) showed that, in the presence of exposure-induced mediator-outcome confounding, decomposing the total causal effect (TCE) using standard conditional exchangeability assumptions is not possible even under a nonparametric structural equation model with all confounders observed. Subsequent research has investigated the assumptions required for such a decomposition to be identifiable and estimable from observed data. One approach was proposed by VanderWeele et al (2014). They decomposed the TCE under three different scenarios: (1) treating the mediator and the exposure-induced confounder as joint mediators; (2) generating path-specific effects albeit without distinguishing between multiple distinct paths through the exposure-induced confounder; and (3) using so-called randomised interventional analogues where sampling values from the distribution of the mediator within the levels of the exposure effectively marginalises over the exposure-induced confounder. In this paper, we extend their approach to the case where there are multiple mediators that do not influence each other directly but which are all influenced by an exposure-induced mediator-outcome confounder. We provide a motivating example and results from a simulation study based on from our work in dental epidemiology featuring the 1982 Pelotas Birth Cohort in Brazil.