Mediation Analysis with Survival Outcomes: Accelerated Failure Time vs. Proportional Hazards Models

Frailty and Metabolic Vulnerability in Heart Failure: A Community Cohort Study

BACKGROUND

Frailty is common in heart failure (HF) and is associated with death but not routinely captured clinically. Frailty is linked with inflammation and malnutrition, which can be assessed by a novel plasma multimarker score: the metabolic vulnerability index (MVX). We sought to evaluate the associations between frailty and MVX and their prognostic impact.

METHODS AND RESULTS

In an HF community cohort (2003-2012), we measured frailty as a proportion of deficits present out of 32 physical limitations and comorbidities, MVX by nuclear magnetic resonance spectroscopy, and collected extensive longitudinal clinical data. Patients were categorized by frailty score (≤0.14, >0.14 and ≤0.27, >0.27) and MVX score (≤50, >50 and ≤60, >60 and ≤70, >70). Cox models estimated associations of frailty and MVX with death, adjusted for Meta-Analysis Global Group in Chronic Heart Failure (MAGGIC) score and NT-proBNP (N-terminal pro-B-type natriuretic peptide). Uno's C-statistic measured the incremental value of MVX beyond frailty and clinical factors. Weibull's accelerated failure time regression assessed whether MVX mediated the association between frailty and death. We studied 985 patients (median age, 77; 48% women). Frailty and MVX were weakly correlated (Spearman's ρ=0.21). The highest frailty group experienced an increased rate of death, independent of MVX, MAGGIC score, and NT-proBNP (hazard ratio, 3.3 [95% CI, 2.5-4.2]). Frailty improved Uno's c-statistic beyond MAGGIC score and NT-proBNP (0.69-0.73). MVX only mediated 3.3% and 4.5% of the association between high and medium frailty groups and death, respectively.

CONCLUSIONS

In this HF cohort, frailty and MVX are weakly correlated. Both independently contribute to stratifying the risk of death, suggesting that they capture distinct domains of vulnerability in HF.

Secular Trends in Outcomes and Impact of Novel Oral Anticoagulants in Atrial Fibrillation-Related Acute Ischemic Stroke

BACKGROUND

Novel oral anticoagulants (NOACs) are currently recommended for the secondary prevention of stroke in patients with acute ischemic stroke (AIS) accompanied by atrial fibrillation (AF). However, the impact of NOACs on clinical outcomes in real-world practice remains ambiguous. This study analyzes the trend of clinical events in patients with AF-related AIS and determines how much the introduction of NOACs has mediated this trend.

METHODS

We identified patients with AIS and AF between January 2011 and December 2019 using a multicenter stroke registry. Annual rates of NOAC prescriptions and clinical events within 1 year were evaluated. The primary outcome was a composite of recurrent stroke, myocardial infarction, and all-cause mortality. To assess the mediation effect of NOACs on the relationship between the calendar year and these outcomes, we used natural effect models and conducted exposure-mediator, exposure-outcome, and mediator-outcome analyses using multivariable regression models or accelerated failure time models, adjusting for potential confounders.

RESULTS

Among the 12 977 patients with AF-related AIS, 12 500 (average age: 74.4 years; 51.3% male) were analyzed after excluding cases of valvular AF. Between 2011 and 2019, there was a significant decrease in the 1-year incidence of the primary composite outcome from 28.3% to 21.7%, while the NOAC prescription rate increased from 0% to 75.6%. A 1-year increase in the calendar year was independently associated with delayed occurrence of the primary outcome (adjusted time ratio, 1.10 [95% CI, 1.07-1.14]) and increased NOAC prescription (adjusted odds ratio, 2.20 [95% CI, 2.14-2.27]). Increased NOAC prescription was associated with delayed occurrence of the primary outcome (adjusted time ratio, 3.82 [95% CI, 3.17 to 4.61]). Upon controlling for NOAC prescription (mediator), the calendar year no longer influenced the primary outcome (adjusted time ratio, 0.97 [95% CI, 0.94-1.00]). This suggests that NOAC prescription mediates the association between the calendar year and the primary outcome.

CONCLUSIONS

Our study highlights a temporal reduction in major clinical events or death in Korean patients with AF-related AIS, mediated by increased NOAC prescription, emphasizing NOAC use in this population.

Impact of Finerenone-Induced Albuminuria Reduction on Chronic Kidney Disease Outcomes in Type 2 Diabetes : A Mediation Analysis

BACKGROUND

In patients with chronic kidney disease (CKD) and type 2 diabetes (T2D), finerenone, a nonsteroidal mineralocorticoid receptor antagonist, reduces cardiovascular and kidney failure outcomes. Finerenone also lowers the urine albumin-to-creatinine ratio (UACR). Whether finerenone-induced change in UACR mediates cardiovascular and kidney failure outcomes is unknown.

OBJECTIVE

To quantify the proportion of kidney and cardiovascular risk reductions seen over a 4-year period mediated by a change in kidney injury, as measured by the change in log UACR between baseline and month 4.

DESIGN

Post hoc mediation analysis using pooled data from 2 phase 3, double-blind trials of finerenone. (ClinicalTrials.gov: NCT02540993 and NCT02545049).

SETTING

Several clinical sites in 48 countries.

PATIENTS

12 512 patients with CKD and T2D.

INTERVENTION

Finerenone and placebo (1:1).

MEASUREMENTS

Separate mediation analyses were done for the composite kidney (kidney failure, sustained ≥57% decrease in estimated glomerular filtration rate from baseline [approximately a doubling of serum creatinine], or kidney disease death) and cardiovascular (cardiovascular death, nonfatal myocardial infarction, nonfatal stroke, or hospitalization for heart failure) outcomes.

RESULTS

At baseline, median UACR was 514 mg/g. A 30% or greater reduction in UACR was seen in 3338 (53.2%) patients in the finerenone group and 1684 (27.0%) patients in the placebo group. Reduction in UACR (analyzed as a continuous variable) mediated 84% and 37% of the treatment effect on the kidney and cardiovascular outcomes, respectively. When change in UACR was analyzed as a binary variable (that is, whether the guideline-recommended 30% reduction threshold was met), the proportions mediated for each outcome were 64% and 26%, respectively.

LIMITATION

The current findings are not readily extendable to other drugs.

CONCLUSION

In patients with CKD and T2D, early albuminuria reduction accounted for a large proportion of the treatment effect against CKD progression and a modest proportion of the effect against cardiovascular outcomes.

PRIMARY FUNDING SOURCE

Bayer AG.

The role of exposure to per- and polyfluoroalkyl substances in racial/ethnic disparities in hypertension: Results from the study of Women's health across the nation

BACKGROUND

Racial/ethnic disparities in hypertension are a pressing public health problem. The contribution of environmental pollutants including PFAS have not been explored, even though certain PFAS are higher in Black population and have been associated with hypertension.

OBJECTIVES

We examined the extent to which racial/ethnic disparities in incident hypertension are explained by racial/ethnic differences in serum PFAS concentrations.

METHODS

We included 1058 hypertension-free midlife women with serum PFAS concentrations in 1999-2000 from the multi-racial/ethnic Study of Women's Health Across the Nation with approximately annual follow-up visits through 2017. Causal mediation analysis was conducted using accelerated failure time models. Quantile-based g-computation was used to evaluate the joint effects of PFAS mixtures.

RESULTS

During 11,722 person-years of follow-up, 470 participants developed incident hypertension (40.1 cases per 1000 person-years). Black participants had higher risks of developing hypertension (relative survival: 0.58, 95% CI: 0.45-0.76) compared with White participants, which suggests racial/ethnic disparities in the timing of hypertension onset. The percent of this difference in timing that was mediated by PFAS was 8.2% (95% CI: 0.7-15.3) for PFOS, 6.9% (95% CI: 0.2-13.8) for EtFOSAA, 12.7% (95% CI: 1.4-22.6) for MeFOSAA, and 19.1% (95% CI: 4.2, 29.0) for PFAS mixtures. The percentage of the disparities in hypertension between Black versus White women that could have been eliminated if everyone's PFAS concentrations were dropped to the 10th percentiles observed in this population was 10.2% (95% CI: 0.9-18.6) for PFOS, 7.5% (95% CI: 0.2-14.9) for EtFOSAA, and 17.5% (95% CI: 2.1-29.8) for MeFOSAA.

CONCLUSIONS

These findings suggest differences in PFAS exposure may be an unrecognized modifiable risk factor that partially accounts for racial/ethnic disparities in timing of hypertension onset among midlife women. The study calls for public policies aimed at reducing PFAS exposures that could contribute to reductions in racial/ethnic disparities in hypertension.

Early life adversity and adult social relationships have independent effects on survival in a wild primate

Adverse conditions in early life can have negative consequences for adult health and survival in humans and other animals. What variables mediate the relationship between early adversity and adult survival? Adult social environments represent one candidate: Early life adversity is linked to social adversity in adulthood, and social adversity in adulthood predicts survival outcomes. However, no study has prospectively linked early life adversity, adult social behavior, and adult survival to measure the extent to which adult social behavior mediates this relationship. We do so in a wild baboon population in Amboseli, Kenya. We find weak mediation and largely independent effects of early adversity and adult sociality on survival. Furthermore, strong social bonds and high social status in adulthood can buffer some negative effects of early adversity. These results support the idea that affiliative social behavior is subject to natural selection through its positive relationship with survival, and they highlight possible targets for intervention to improve human health and well-being.

Explaining regional variations in colon cancer survival in Ontario, Canada: a population-based retrospective cohort study

OBJECTIVES

Regional variation in cancer survival is an important health system performance measurement. We evaluated if regional variation in colon cancer survival may be driven by differences in the patient population, their health and healthcare utilisation, and/or cancer care delivery.

DESIGN

Population-based retrospective cohort study using routinely collected linked health administrative data.

SETTING

Ontario, Canada.

PARTICIPANTS

Patients with colon cancer diagnosed between 1 January 2009 and 31 December 2012.

OUTCOME

Cancer-specific survival was compared across the province's 14 health regions. Using accelerated failure time models, we assessed whether regional survival variations were mediated through differences in case mix, including age, sex, comorbidities, stage at diagnosis and colon subsite, potential marginalisation and/or prediagnosis healthcare.

RESULTS

The study population included 16 895 patients with colon cancer. There was statistically significant regional variation in cancer-specific survival. Three regions had cancer-specific survival that was between 30% (95% CI 1.03 to 1.65) and 39% (95% CI 1.13 to 1.71) longer and one region had cancer-specific survival that was 26% shorter (95% CI 0.58 to 0.93) than the reference region. For three of these regions, case mix explained between 26% and 56% of the survival variation. Further adjustment for rurality explained 22% of the remaining survival variation in one region. Adjustment for continuity of primary care and the diagnostic interval length explained 10% and 11% of the remaining survival variation in two other regions. Socioeconomic marginalisation, recent immigration and colonoscopy history did not explain colon cancer survival variation.

CONCLUSIONS

Case mix accounted for much of the regional variation in colon cancer survival, indicating that efforts to monitor the quality of cancer care through survival metrics should consider case mix when reporting regional survival differences. Future work should repeat this approach in other settings and other cancer sites considering a broad range of potential mediators.

Perfluoroalkyl Substances and Incident Natural Menopause in Midlife Women: The Mediating Role of Sex Hormones

Perfluoroalkyl and polyfluoroalkyl substances (PFAS) have been associated with earlier natural menopause; however, the underlying mechanisms are not well understood, particularly the extent to which this relationship is mediated by sex hormones. We analyzed data (1999-2017) on 1,120 premenopausal women from the Study of Women's Health Across the Nation (SWAN). Causal mediation analysis was applied to quantify the degree to which follicle-stimulating hormone (FSH) and estradiol levels could mediate the associations between PFAS and incident natural menopause. Participants with higher PFAS concentrations had shorter times to natural menopause, with a relative survival of 0.82 (95% confidence interval (CI): 0.69, 0.96) for linear perfluorooctane sulfonate (n-PFOS), 0.84 (95% CI: 0.69, 1.00) for sum of branched-chain perfluorooctane sulfonate (Sm-PFOS), 0.79 (95% CI: 0.66, 0.93) for linear-chain perfluorooctanoate (n-PFOA), and 0.84 (95% CI: 0.71, 0.97) for perfluorononanoate (PFNA), comparing the highest tertile of PFAS concentrations with the lowest. The proportion of the effect mediated through FSH was 8.5% (95% CI: -11.7, 24.0) for n-PFOS, 13.2% (95% CI: 0.0, 24.5) for Sm-PFOS, 26.9% (95% CI: 15.6, 38.4) for n-PFOA, and 21.7% (6.8, 37.0) for PFNA. No significant mediation by estradiol was observed. The effect of PFAS on natural menopause may be partially explained by variations in FSH concentrations.

More than treatment refusal: a National Cancer Database analysis of adjuvant treatment refusal and racial survival disparities among women with endometrial cancer

BACKGROUND

Disparities in adjuvant treatment between Black and White women with endometrial cancer exist and contribute to worse outcomes among Black women. However, factors leading to disparate treatment receipt are understudied.

OBJECTIVE

We examined whether patient refusal of adjuvant treatment (chemotherapy or radiation) differed between Black and White women and whether treatment refusal mediated racial disparities in survival among women with endometrial cancer.

STUDY DESIGN

We used the National Cancer Database, a hospital-based cancer registry, to identify non-Hispanic Black and non-Hispanic White women diagnosed with endometrial cancer from 2004 to 2016 who either received or refused recommended radiation or chemotherapy. We used logistic regression to estimate multivariable-adjusted odds ratios and 95% confidence intervals for associations between race and treatment refusal. We also examined predictors of treatment refusal in race-specific models. Accelerated failure time models were used to estimate absolute differences in overall survival by race. We used causal mediation analysis to estimate the proportion of racial differences in overall survival attributable to racial differences in adjuvant treatment refusal. We considered the overall study population and strata defined by histology, and adjusted for sociodemographic, tumor, and facility characteristics.

RESULTS

Our analysis included 75,447 endometrial cancer patients recommended to receive radiation and 60,187 endometrial cancer patients recommended to receive chemotherapy, among which 6.4% and 11.4% refused treatment, respectively. Among Black women recommended for radiation or chemotherapy, 6.4% and 9.6% refused, respectively. Among White women recommended for radiation or chemotherapy, 6.4% and 11.8% refused, respectively. After adjusting for sociodemographic variables, facility characteristics, and tumor characteristics, Black women were more likely to refuse chemotherapy than White women (adjusted odds ratio, 1.26; 95% confidence interval, 1.15-1.37), but no difference in radiation refusal was observed (adjusted odds ratio, 1.00; 95% confidence interval, 0.91-1.11). Some predictors of radiation refusal varied by race, namely income, education, histology, stage, and chemotherapy receipt (P interactions<.05), whereas predictors of chemotherapy refusal were generally similar between Black and White women. Among women recommended for radiation, Black women survived an average of 4.3 years shorter than White women, which did not seem attributable to differences in radiation refusal. Among women recommended for chemotherapy, Black women survived an average of 3.2 years shorter than White women of which 1.9 months (4.9%) could potentially be attributed to differences in chemotherapy refusal.

CONCLUSION

We observed differences in chemotherapy refusal by race, and those differences may be responsible for up to about 2 months of the overall 3.2-year survival disparity between White and Black women. Radiation refusal did not explain any of the 4.3-year disparity among women recommended for radiation. Treatment refusal accounts for, at most, a small fraction of the total racial disparity in endometrial cancer survival. Although a better understanding of the reasons for patient treatment refusal and subsequent intervention may help improve outcomes for some women, other causes of disparate outcomes, particularly those reflecting the social determinants of health, must be investigated.

Child maltreatment and incident mental disorders in middle and older ages: a retrospective UK Biobank cohort study

BACKGROUND

Understanding the mental health consequences of child maltreatment at different life stages is important in accurately quantifying the burden of maltreatment. This study investigated the association between child maltreatment and incident mental disorders in middle and older age as well as the potential mediators and moderators.

METHODS

This is a retrospective cohort study of 56,082 participants from UK Biobank. Child maltreatment was recalled using the Childhood Trauma Screener. Incident mental disorders, including depressive, anxiety and affective disorders, behavioural syndromes, post-traumatic stress disorder (PTSD), schizophrenia, substance abuse, and dementia, after baseline assessment were ascertained through linkage to primary care records.

FINDINGS

There was a dose-response relationship between child maltreatment and mental disorder. Those who experienced three or more maltreatment types had the highest risk of all mental disorders (HR 1.85, 95% CI: 1.67-2.06) followed by those who experienced two (HR 1.48, 95% CI: 1.35-1.63) and then one (HR 1.26, 95% CI: 1.19-2.35). Child maltreatment was most strongly associated with PTSD (HR 1.59, 95% CI: 1.20-2.10 P=0.001). The excess risk was largely unexplained by the included mediators. The association between child maltreatment and all mental disorders were stronger among participants who binge drank (P_interaction=0.003) or had few social visits (P_interaction=0.003).

INTERPRETATION

The mental health consequence of child maltreatment could last decades, even among those who had no recorded mental disorders in early adulthood. In the absence of strong mediators, prevention of child maltreatment remains the priority.

FUNDING

Wellcome Trust Institutional Strategic Support Fund.

Loneliness, social network size, and mortality in older adults and the role of cortisol

BACKGROUND

Loneliness and social isolation have both been found to be associated with increased mortality in previous studies. One potential underlying mechanism is via the hypothalamic-pituitary-adrenal axis.

OBJECTIVE

This study aimed to examine the association between social network size and cortisol, to analyze the associations between both loneliness and social network size and mortality, and to examine to what extent the association between network size and/or loneliness and mortality is mediated by cortisol.

DESIGN

The study group consisted of 443 depressed and non-depressed participants of the Netherlands Study of Depression in the Elderly (NESDO). Cross-sectional analysis of the association between social network size and cortisol measures was followed by a survival analysis of the associations between both social network size and loneliness and mortality.

RESULTS

There were no significant associations between social network size and cortisol measures. Loneliness and small social network size were not associated with mortality. Age and partner status were more important predictors of mortality.

CONCLUSION

As people grow older the variety of factors that influence mortality risk increases, diminishing the effect of a single factor. Prevention of early morbidity and mortality in older adults should be tailored to specific needs and risks, instead of aiming at one specific factor.

Mediation analysis methods used in observational research: a scoping review and recommendations

Background

Mediation analysis methodology underwent many advancements throughout the years, with the most recent and important advancement being the development of causal mediation analysis based on the counterfactual framework. However, a previous review showed that for experimental studies the uptake of causal mediation analysis remains low. The aim of this paper is to review the methodological characteristics of mediation analyses performed in observational epidemiologic studies published between 2015 and 2019 and to provide recommendations for the application of mediation analysis in future studies.

Methods

We searched the MEDLINE and EMBASE databases for observational epidemiologic studies published between 2015 and 2019 in which mediation analysis was applied as one of the primary analysis methods. Information was extracted on the characteristics of the mediation model and the applied mediation analysis method.

Results

We included 174 studies, most of which applied traditional mediation analysis methods (n = 123, 70.7%). Causal mediation analysis was not often used to analyze more complicated mediation models, such as multiple mediator models. Most studies adjusted their analyses for measured confounders, but did not perform sensitivity analyses for unmeasured confounders and did not assess the presence of an exposure-mediator interaction.

Conclusions

To ensure a causal interpretation of the effect estimates in the mediation model, we recommend that researchers use causal mediation analysis and assess the plausibility of the causal assumptions. The uptake of causal mediation analysis can be enhanced through tutorial papers that demonstrate the application of causal mediation analysis, and through the development of software packages that facilitate the causal mediation analysis of relatively complicated mediation models.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12874-021-01426-3.

Association and pathways between shift work and cardiovascular disease: a prospective cohort study of 238 661 participants from UK Biobank

BACKGROUND

This study aimed to study the association between shift work and incident and fatal cardiovascular disease (CVD), and to explore modifying and mediating factors.

METHODS

This is a population-based, prospective cohort study with a median follow-up of 11 years; 238 661 UK Biobank participants who were in paid employment or self-employed at baseline assessment were included.

RESULTS

Shift workers had higher risk of incident [hazard ratio (HR) 1.11, 95% confidence interval (CI) 1.06-1.19] and fatal (HR 1.25, 95% CI 1.08-1.44) CVD compared with non-shift workers, after adjusting for socio-economic and work-related factors. The risk was higher with longer duration of shift work, in women and in jobs with little heavy manual labour. Current smoking, short sleep duration, poor sleep quality, adiposity, higher glycated haemoglobin and higher cystatin C were identified as the main potentially modifiable mediators. Mediators collectively explained 52.3% of the associations between shift work and incident CVDs.

CONCLUSIONS

Shift workers have higher risk of incident and fatal CVD, partly mediated through modifiable risk factors such as smoking, sleep duration and quality, adiposity and metabolic status. Workplace interventions targeting these mediators have the potential to alleviate shift workers' CVD risk.

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 (SIS) 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 (FDR) 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 TCGA lung cancer cohort. Two methylation sites (cg08108679 and cg26478297) are identified as potential mediating epigenetic markers.

AVAILABILITY

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.

Gender Differences in Mortality After Left Ventricular Assist Device Implant: A Causal Mediation Analysis Approach

We used the International Society for Heart and Lung Transplantation (ISHLT) Registry for Mechanically Assisted Circulatory Support (IMACS) database to examine 1) gender differences in post-left ventricular assist device (LVAD) mortality in the contemporary era and 2) preimplant clinical factors that might mediate any observed differences. Adults who received continuous-flow (CF)-LVAD from January 2013 to September 2017 (n = 9,565, age: 56.2 ± 13.2 years, 21.6% female, 31.1% centrifugal pumps) were analyzed. An inverse probability weighted Cox proportional hazards model was used to estimate association of female gender with all-cause mortality, adjusting for known covariates. Causal mediation analysis was performed to test plausible preimplant mediators mechanistically underlying any association between female gender and mortality. Females had higher mortality after LVAD (adjusted hazard ratio [HR]: 1.36; p < 0.0001), with significant gender × time interaction (p = 0.02). An early period of increased risk was identified, with females experiencing a higher risk of mortality during the first 4 months after implant (adjusted HR: 1.74; p < 0.0001), but not after (adjusted HR: 1.18; p = 0.16). More severe tricuspid regurgitation and smaller left ventricular end-diastolic diameter at baseline mediated ≈21.9% of the increased early hazard of death in females; however, most of the underlying mechanisms remain unexplained. Therefore, females have increased mortality only in the first 4 months after LVAD implantation, partially driven by worsening right ventricular dysfunction and LV-LVAD size mismatch.

Depression as a risk factor for dementia in older people with type 2 diabetes and the mediating effect of inflammation

AIMS/HYPOTHESIS

We aimed to determine the association of depression with dementia risk in people with type 2 diabetes, and to explore the possible mediating role of inflammation in this relationship.

METHODS

The Edinburgh Type 2 Diabetes Study is a prospective cohort of 1066 men and women with type 2 diabetes aged 60-75 years. Cox proportional hazards regression analysis was used to investigate the association between depression, assessed at baseline, and subsequent risk of dementia over 10 years. Depression was defined using the Hospital Anxiety and Depression Scale, while incident dementia was defined using medical records, prescription data and death certificates. The potential mediating effect of systemic inflammation was assessed by adjusting models for a generalised inflammation factor, derived from four inflammatory markers measured at baseline (C-reactive protein, IL-6, TNF-α and fibrinogen), and carrying out an exploratory mediation analysis.

RESULTS

Dementia developed in 105 participants over a median follow-up of 10.6 years. After adjusting for age and sex, depression was associated with over a 2.5-fold increase in risk of dementia (HR 2.59 [95% CI 1.62, 4.15]). Additional adjustment for the generalised inflammation factor and other covariates did not attenuate the size of association between depression and incident dementia and mediation analysis showed that it was not a mediator. Adjusted logistic regression models showed cross-sectional associations of C-reactive protein and IL-6 with depression.

CONCLUSIONS/INTERPRETATION

Depression is an important risk factor for dementia in people with type 2 diabetes. Some inflammatory markers were associated with depression, but systemic inflammation does not appear to mediate the relationship between depression and dementia. Graphical abstract.

The relative contributions of behavioral, biological, and psychological risk factors in the association between psychosocial stress and all-cause mortality among middle- and older-aged adults in the USA

Evidence of an association between psychosocial stress and mortality continues to accumulate. However, despite repeated calls in the literature for further examination into the physiological and behavioral pathways though which stress affects health and mortality, research on this topic remains limited. This study addresses this gap by employing a counterfactual-based mediation analysis of eight behavioral, biological, and psychological pathways often hypothesized to play a role in the association between stress and health. First, we calculated the survival rate of all-cause mortality associated with cumulative psychosocial stress (high vs. low/moderate) using random effects accelerated failure time models among a sample of 7108 adults from the Midlife in the United States panel study. Then, we conducted a multiple mediator mediation analysis utilizing a counterfactual regression framework to determine the relative contributions of each mediator and all mediators combined in the association between stress and mortality. Exposure to high psychosocial stress was associated with a 0.76 times reduced survival rate over the follow-up period 1995-2015, while adjusting for age, sex, race, income, education, baseline health, and study design effects. The mediators accounted for 49% of this association. In particular, smoking, sedentary behavior, obesity/BMI, and cardiovascular disease displayed significant indirect effects and accounted for the largest reductions in the total effect of stress on mortality, with natural indirect effects of 14%, 12%, 11%, and 4%, respectively. In conclusion, traditional behavioral and biological risk factors play a significant role in the association between psychosocial stress and mortality among middle and older adults in the US context. While eliminating stress and the socioeconomic disparities that so often deliver people into high-stress scenarios should be the ultimate goal, public health interventions addressing smoking cessation, physical activity promotion, and cardiovascular disease treatment may pay dividends for preventing premature mortality in the near-term.

Child maltreatment and cardiovascular disease: quantifying mediation pathways using UK Biobank

BACKGROUND

Child maltreatment is associated with cardiovascular disease (CVD), but mediation pathways have not been fully elucidated. The aim of the current study was to determine and quantify the underlying pathways linking child maltreatment and CVD.

METHODS

We conducted a retrospective cohort study using the UK Biobank. The number and types of child maltreatment, including abuse and neglect, were recalled by the participants. Lifestyle, biological, physical, and mental health factors measured at baseline were explored as potential mediators. Incident CVD was ascertained through record linkage after baseline measurement. Age, sex, ethnicity, area-based deprivation, and education level were adjusted for as confounders. Cox proportional hazard models were conducted to test for associations between child maltreatment and incident CVD.

RESULTS

A total of 152,040 participants who completed the child maltreatment assessment were included in the analyses, and one third reported at least one type of child maltreatment. There was a dose-response relationship between the number of maltreatment types and incident CVD. On average, each additional type of child maltreatment was associated with an 11% (95% CI 8-14%, P < 0.0001) increased risk of CVD. The majority (56.2%) of the association was mediated through depressive symptoms, followed by smoking (14.7%), high-density lipoprotein cholesterol (8.7%), and sleep duration (2.4%).

CONCLUSION

Child maltreatment is associated with incident CVD through a combination of mental health, lifestyle, and biological pathways. Therefore, in addition to interventions to reduce the occurrence of child maltreatment, attention should be targeted at promoting healthy lifestyles and preventing, identifying, and treating depression among children and adults who have previously been maltreated.

Smoking-induced risk of future cardiovascular disease is partly mediated by cadmium in tobacco: Malmö Diet and Cancer Cohort Study

BACKGROUND

Smoking is a strong risk factor for cardiovascular disease (CVD) and causes exposure to cadmium, which is a pro-atherosclerotic metal. Cadmium exposure has also been shown to increase the risk of CVD, even after adjustment for smoking. Our hypothesis was that part of the risk of CVD in smokers may be mediated by cadmium exposure from tobacco smoke. We examined this hypothesis in a mediation analysis, trying to assess how much of the smoking-induced CVD risk could be explained via cadmium.

METHODS

We used prospective data on CVD (incidence and mortality) in a Swedish population-based cohort of 4304 middle-aged men and women (the Malmö Diet and Cancer Study). Blood cadmium was analyzed in base-line samples from 1991, and clinical events were followed up for 16-19 years based on registry data. Mediation analysis was conducted to evaluate the indirect effect (via cadmium) of smoking on CVD. Survival was analyzed by the accelerated failure time (AFT) model and the Aalen additive hazard model.

RESULTS

The mean blood cadmium level in the study population was 0.43 μg/L (median 0.24 μg/L) and increased with recent and current smoking. As expected, shorter survival time (AFT model) and higher incidence rate (Aalen model) were found in current smokers for all CVD outcomes and this effect seemed to be partly mediated by cadmium. For the sum of acute myocardial infarction, bypass grafts and percutaneous coronary intervention, and death in ischemic heart disease, about half of the increased risk of such events in current smokers was mediated via cadmium, with similar results for the AFT and Aalen models.

CONCLUSIONS

Cadmium plays an important role in smoking-induced CVDs. This provides evidence for mechanisms and is of importance for both individuals and policy makers.

The Interplay Between Subjective Response to Alcohol, Craving, and Alcohol Self-Administration in the Human Laboratory

BACKGROUND

Despite a rich literature on human laboratory paradigms of subjective response (SR) to alcohol, craving for alcohol, and alcohol self-administration, few studies have examined the interplay across these 3 constructs. The present study addresses this gap in the literature by examining the interplay between SR, craving, and self-administration in the human laboratory.

METHODS

Data were culled from a medication study (NCT02026011) in which heavy drinking participants of East Asian ancestry completed 2 double-blinded and counterbalanced experimental sessions. In each experimental session, participants received a priming dose of intravenous (IV) alcohol to a target breath alcohol concentration (BrAC) of 0.06 g/dl and measures of SR (stimulation and sedation) and alcohol craving were collected across rising BrACs. The IV alcohol challenge was immediately followed by a 1-hour alcohol self-administration period.

RESULTS

Mixed model analyses found a positive and significant relationship between the slope of stimulation and the slope of craving during the alcohol challenge. The relationship between sedation and craving, however, was not significant. The slope of craving during the alcohol challenge significantly predicted a higher number of mini-drinks consumed and lower latency to first drink. Further, mediation analyses found that craving was a significant mediator of the relationship between stimulation and total number of mini-drinks consumed, but the same pattern was not found for sedation.

CONCLUSIONS

Insofar as alcohol self-administration represents the end point of interest for a host of experimental and clinical research questions, the present study suggests that alcohol craving represents a more proximal predictor of self-administration than measures of alcohol-induced stimulation. It is recommended that human laboratory models interpret measures of SR and craving in light of their relative predictive utility for drinking outcomes.

Evaluating a Method to Estimate Mediation Effects With Discrete-Time Survival Outcomes

The utility of evaluating mediation effects spans across research domains. The model facilitates investigation of underlying mechanisms of event timing and, as such, has the potential to help strengthen etiological research and inform intervention work that incorporates the evaluation of mediating variables. In order for the analyses to be maximally useful however, it is critical to employ methodology appropriate for the data under investigation. The purpose of this paper is to evaluate a regression-based approach to estimating mediation effects with discrete-time survival outcomes. We empirically evaluate the performance of the discrete-time survival mediation model in a statistical simulation study, and demonstrate that results are functionally equivalent to estimates garnered from a potential-outcomes framework. Simulation results indicate that parameter estimates of mediation in the model were statistically accurate and precise across the range of examined conditions. Type 1 error rates were also tolerable in the conditions studied. Adequate power to detect effects in the model, with binary X and continuous M variables, required effect sizes of the mediation paths to be medium or large. Possible extensions of the model are also considered.

Blood lead levels and timing of male sexual maturity: A longitudinal study of Russian boys

BACKGROUND

Higher blood lead levels (BLLs) have been linked to neurologic deficits and impaired growth, but few studies have evaluated their association with timing of sexual maturity or pubertal progression in boys.

METHODS

In a longitudinal cohort of Russian boys enrolled at age 8-9 and followed to adulthood, BLLs were measured at study entry, and pubertal staging (genitalia and pubic hair) and testicular volume (TV) measurements were obtained annually. We used interval-censored regression models to estimate differences between boys with higher BLL (≥5 μg/dL) and lower BLL in mean ages at sexual maturity (genitalia stage 5, pubic hair stage 5, or TV ≥ 20 mL) and duration of pubertal progression (onset to maturity), adjusting for potential confounders. Mediation analyses were conducted to quantify the percent of lead's effect attributable to its association with reduced somatic growth.

RESULTS

Among 481 evaluable boys, 28% had BLL ≥ 5 μg/dL. Adjusted mean ages at sexual maturity were 14.7 years for genitalia, 16.1 for pubic hair, and 13.9 for TV. In adjusted models, boys with BLLs ≥5 μg/dL had later maturity than those with lower levels by 4-5 months depending on pubertal indicator. In mediation analyses, height and body mass index at age 11 accounted for 40-71% of the shift in age at maturity for boys with higher compared to lower BLLs. Higher BLLs were not associated with pace of pubertal progression.

CONCLUSIONS

Higher lead levels were associated with later attainment of sexual maturity in males, but not with the duration of pubertal progression. A high proportion of the delay in sexual maturity for boys with higher as compared to lower BLL was shown to be attributable to mediating effects of BLL on reduced growth.

Education level and incident functional disability in elderly Japanese: The Ohsaki Cohort 2006 study

As the factors that link education level with incident functional disability in elderly Japanese have never been investigated, the present study investigated this issue in an elderly Japanese population. A 9-year prospective cohort study (2006–2015) was conducted among 8,680 Japanese individuals (≥65 years), Ohsaki city, Japan. In a baseline survey, we collected data on education level and potential mediators. Data on incident functional disability were retrieved from the Long-term Care Insurance database. The Cox proportional hazards model was used to estimate the hazard ratios (HRs) and 95% confidence intervals (CIs) for incident functional disability by education level (below upper-secondary education (reference), and upper secondary education and above). Mediating effects were estimated using accelerated failure time model and a logistic regression model. During 9-year follow-up period, 2,742 cases (31.6%) of incident functional disability were observed, and education level showed an inverse association with functional disability (P for trend <0.01). Participation in community activities had the largest mediating effect (34.7%) on the relationship between education level and incident functional disability. This effect remained among those aged 65–74 years (19.9%) but became negligible among those aged ≥75 years. Other potential mediators (such as smoking and drinking status) were also tested, but these showed only small mediating effects. The inverse association between education level and the incident risk of functional disability appears to be largely mediated by participation in community activities among elderly Japanese, especially those aged 65–74 years.

Mediation analysis with a time-to-event outcome: a review of use and reporting in healthcare research

BACKGROUND

Mediation analysis tests whether the relationship between two variables is explained by a third intermediate variable. We sought to describe the usage and reporting of mediation analysis with time-to-event outcomes in published healthcare research.

METHODS

A systematic search of Medline, Embase, and Web of Science was executed in December 2016 to identify applications of mediation analysis to healthcare research involving a clinically relevant time-to-event outcome. We summarized usage over time and reporting of important methodological characteristics.

RESULTS

We included 149 primary studies, published from 1997 to 2016. Most studies were published after 2011 (n = 110, 74%), and the annual number of studies nearly doubled in the last year (from n = 21 to n = 40). A traditional approach (causal steps or change in coefficient) was most commonly taken (n = 87, 58%), and the majority of studies (n = 114, 77%) used a Cox Proportional Hazards regression for the outcome. Few studies (n = 52, 35%) mentioned any of the assumptions or limitations fundamental to a causal interpretation of mediation analysis.

CONCLUSION

There is increasing use of mediation analysis with time-to-event outcomes. Current usage is limited by reliance on traditional methods and the Cox Proportional Hazards model, as well as low rates of reporting of underlying assumptions. There is a need for formal criteria to aid authors, reviewers, and readers reporting or appraising such studies.

The role of fetal growth restriction in the association between Down syndrome and perinatal mortality

Objective: Down syndrome (DS) is associated with significant risk of perinatal mortality. We hypothesize that this association is primarily mediated through the effects of fetal growth restriction (FGR).Methods: This was a retrospective cohort analysis using the US Natality Database from 2011 to 2013. Analysis was limited to singleton nonanomalous pregnancies or confirmed DS pregnancies without severe structural anomalies between 24 and 42 w in gestation. The risk of stillbirth (SB) associated with DS was estimated using both Cox proportional Hazard (HR) regression and accelerated failure time (AFT) methods. The risk of neonatal mortality was estimated using logistic regression analyses. Mediation analysis was then performed to estimate the effect of small for gestational age (SGA), defined as birthweight ≤10th percentile for gestational age, on perinatal mortality associated with DS. All regression models were selected using backward stepwise elimination method. The final regression models included adjustment for maternal age, hypertension, and diabetes.Results: The final cohort included 2446 DS cases among 9,804,718 births. The overall SB rate was 223.6/1000 births in DS group and 4.7/1000 births without DS (p < .001, adjusted hazard ratio (aHR): 58.25; 95% CI [53.44,63.49]). Based on the AFT model, DS survival-to-delivery rate is 4.3 times lower (TR: 0.23; 95% CI [0.22,0.24]). Thirty-five percentage of the effect of DS on stillbirth was mediated through SGA (% mediation:35.1%; 95% CI [33.7,36.4]). The rate of neonatal mortality among DS was 69.0/1000 births compared with 2.8/1000 births without DS (p < .001, adjusted odds ratio (aOR): 27.16; 95% CI: [22.63,32.60]). Only 11.6% of the effect of DS on neonatal deaths was mediated through SGA (%mediation:11.6%; 95% CI [8.4,10.6]).Conclusion: Over one-third of overall stillbirths were mediated through SGA. Routine surveillance of fetal growth and standard SGA surveillance protocols may reduce the risk of perinatal mortality in DS pregnancies. Conversely, it is important to point out that these surveillance strategies may not be effective two-third of the cases not affected by growth restriction.

Exploring causality mechanism in the joint analysis of longitudinal and survival data

In many biomedical studies, disease progress is monitored by a biomarker over time, eg, repeated measures of CD4 in AIDS and hemoglobin in end-stage renal disease patients. The endpoint of interest, eg, death or diagnosis of a specific disease, is correlated with the longitudinal biomarker. In this paper, we examine and compare different models of longitudinal and survival data to investigate causal mechanisms, specifically, those related to the role of random effects. We illustrate the methods by data from two clinical trials: an AIDS study and a liver cirrhosis study.

Advancing Analytic Approaches to Address Key Questions in Mechanisms of Behavior Change Research

OBJECTIVE

Interest in studying mechanisms of behavior change (MOBCs) in substance use disorder (SUD) treatments has grown considerably in the past two decades. Much of this work has focused on identifying which variables statistically mediate the effect of SUD treatments on clinical outcomes. However, a fuller conceptualization of MOBCs will require greater understanding of questions that extend beyond traditional mediation analysis, including better understanding of when MOBCs change during treatment, when they are most critical to aiding the initiation or maintenance of change, and how MOBCs themselves arise as a function of treatment processes.

METHOD

In the present study, we review why these MOBC-related questions are often minimally addressed in empirical research and provide examples of data analytic methods that may address these issues. We highlight several recent studies that have used such methods and discuss how these methods can provide unique theoretical insights and actionable clinical information.

RESULTS

Several statistical approaches can enhance the field's understanding of the timing and development of MOBCs, including growth-curve modeling, time-varying effect modeling, moderated mediation analysis, dynamic systems modeling, and simulation methods.

CONCLUSIONS

Adopting greater diversity in methods for modeling MOBCs will help researchers better understand the timing and development of key change variables and will expand the theoretical precision and clinical impact of MOBC research. Advances in research design, measurement, and technology are key to supporting these advances.

Mediation Analysis for Censored Survival Data Under an Accelerated Failure Time Model

Recent advances in causal mediation analysis have formalized conditions for estimating direct and indirect effects in various contexts. These approaches have been extended to a number of models for survival outcomes including accelerated failure time models, which are widely used in a broad range of health applications given their intuitive interpretation. In this setting, it has been suggested that under standard assumptions, the "difference" and "product" methods produce equivalent estimates of the indirect effect of exposure on the survival outcome. We formally show that these two methods may produce substantially different estimates in the presence of censoring or truncation, due to a form of model misspecification. Specifically, we establish that while the product method remains valid under standard assumptions in the presence of independent censoring, the difference method can be biased in the presence of such censoring whenever the error distribution of the accelerated failure time model fails to be collapsible upon marginalizing over the mediator. This will invariably be the case for most choices of mediator and outcome error distributions. A notable exception arises in case of normal mediator-normal outcome where we show consistency of both difference and product estimators in the presence of independent censoring. These results are confirmed in simulation studies and two data applications.

Delay in sexual maturation in perinatally HIV-infected youths is mediated by poor growth

OBJECTIVE

To evaluate the association between HIV infection and sexual maturation, and mediation of this association by HIV effects on growth.

DESIGN

Pooled data were analyzed from two longitudinal cohort studies, the International Maternal Pediatric Adolescent AIDS Clinical Trials P219/219C Study (1993-2007) and the Pediatric HIV/AIDS Cohort Study Adolescent Master Protocol (2007-2015), including perinatally HIV-infected (PHIV) and HIV-exposed uninfected (PHEU) youths.

METHODS

We evaluated age at sexual maturity among 2539 PHIV and PHEU adolescents based on annual physician-assessed pubertal staging measures. Interval-censored regression models were used to evaluate associations of HIV infection with age at maturity. Mediation analyses accounting for height and BMI Z-scores at specific ages were used to estimate direct and indirect effects of HIV infection on age at sexual maturity.

RESULTS

Mean ages at sexual maturity for PHIV girls (n = 1032) were 15.5 years for both female breast and pubic hair and 15.9 and 15.8 years for PHIV boys (n = 1054) for genitalia and pubic hair, respectively. PHIV youths matured approximately 6 months later on average than PHEU (n = 221 girls and 232 boys), and this difference persisted after adjustment for race/ethnicity and birth cohort. BMI and height Z-scores mediated the association between HIV infection and later maturation in girls, accounting for up to 74% of the total HIV effect. Only height Z-scores mediated the effect of HIV on male age at maturity, accounting for up to 98% of the HIV effect.

CONCLUSION

PHIV youths attain sexual maturity later on average than PHEU youths. Much of this difference may be attributable to deficient growth, suggesting directions for future interventions.