Mediation analysis with intermediate confounding: structural equation modeling viewed through the causal inference lens

The role of body image dissatisfaction in the relationship between body size and disordered eating and self-harm: complimentary Mendelian randomization and mediation analyses

Disordered eating and self-harm commonly co-occur in young people suggesting potential for shared underlying causes. Body image dissatisfaction (BID) has been recognised as a psychological correlate of body size, associated with both disordered eating and self-harm. However, the investigation into etiological pathways early in the lifecourse to provide detail on how body size and BID may foster disordered eating and self-harm remains largely unexplored. Employing data from two large population-based cohorts, the UK Biobank and the Avon Longitudinal Study of Parents And Children (ALSPAC), we conducted bidirectional Mendelian randomization (MR) to determine the causal direction of effect between genetically predicted prepubertal body size and two measures of BID indicating (i) desire to be smaller, and (ii) desire to be larger. We then used multivariable regression followed by counterfactual mediation analyses. Bidirectional MR indicated robust evidence that increased genetically predicted prepubertal body size increased desire to be smaller and decreased desire to be larger. Evidence for the reverse causal direction was negligible. These findings remained very similar across sensitivity analyses. In females and males, multivariable regression analyses demonstrated that being overweight increased the risk of disordered eating (risk ratio (RR), 95% confidence interval (CI): 1.19, 1.01 to 1.40 and 1.98, 1.28 to 3.05, respectively) and self-harm (RR, 95% CI: 1.35, 1.04 to 1.77 and 1.55, 0.86 to 2.81, respectively), while being underweight was protective against disordered eating (RR, 95% CI: 0.57, 0.40 to 0.81 and 0.81, 0.38 to 1.73, respectively). There was weak evidence of an increase in the risk of self-harm among underweight individuals. Mediation analyses indicated that the relationship between being overweight and subsequent disordered eating was largely mediated by the desire to be smaller. Our research carries important public health implications, suggesting distinct risk profiles for self-harm and disordered eating in relation to weight and body image. In addition, a better understanding of genetically predicted prepubertal BID may be valuable in the prevention and treatment of disordered eating and self-harm in adolescence.

Counterfactual Mediation Analysis with a Latent Class Exposure

Latent classes are a useful tool in developmental research, however there are challenges associated with embedding them within a counterfactual mediation model. We develop and test a new method "updated pseudo class draws (uPCD)" to examine the association between a latent class exposure and distal outcome that could easily be extended to allow the use of any counterfactual mediation method. UPCD extends an existing group of methods (based on pseudo class draws) that assume that the true values of the latent class variable are missing, and need to be multiply imputed using class membership probabilities. We simulate data based on the Avon Longitudinal Study of Parents and Children, examine performance for existing techniques to relate a latent class exposure to a distal outcome ("one-step," "bias-adjusted three-step," "modal class assignment," "non-inclusive pseudo class draws," and "inclusive pseudo class draws") and compare bias in parameter estimates and their precision to uPCD when estimating counterfactual mediation effects. We found that uPCD shows minimal bias when estimating counterfactual mediation effects across all levels of entropy. UPCD performs similarly to recommended methods (one-step and bias-adjusted three-step), but provides greater flexibility and scope for incorporating the latent grouping within any commonly-used counterfactual mediation approach.

Structural Relationships between Asthma and Dental Caries in Children: A Birth Cohort Study in Southern Brazil

INTRODUCTION

There is an inconclusive causal association between asthma symptoms and dental caries in the primary dentition. This study aimed to investigate, using SEM (structural equation modeling), a possible causal relation between asthma and dental caries in the primary dentition.

METHODS

Using data from the 2004 Pelotas Birth Cohort Study, a sub-sample of 1,303 individuals was selected. Dental caries was clinically evaluated at 5 years old based on decayed, missing, and filled tooth (dmft) index criteria. Asthma-related symptoms (wheezing and shortness of breath) at 1- and 4-year-olds composed a latent variable and were the main exposures to caries occurrence. SEM was used to identify possible direct, indirect, and mediated effects of asthma in primary dentition dental caries.

RESULTS

The general prevalence of caries at age 5 was 1.95 (SD: 3.56). When comparing the dmft values for children with asthma symptoms and those without, they presented similar values in both periods where asthma symptoms were evaluated (1- and 4-year-old). SEM analysis showed that asthma was neither directly nor indirectly related to dental caries.

CONCLUSION

Asthma, using a latent variable constructed based on asthma symptoms, showed no causal effect on dental caries occurrence in the primary dentition.

Can cognitive inflexibility reduce symptoms of anxiety and depression? Promoting the structural nested mean model in psychotherapy research

OBJECTIVE

To estimate the causal effect of executive functioning on the remission of depression and anxiety symptoms in an observational dataset from a vocational rehabilitation program. It is also an aim to promote a method from the causal inference literature and to illustrate its value in this setting.

METHOD

With longitudinal (four-time points over 13 months) data from four independent sites, we compiled a dataset with 390 participants. At each time point, participants were tested on executive function and self-reported symptoms of anxiety and depression. We used g-estimation to evaluate whether objectively tested cognitive flexibility affected depressive/anxious symptoms and tested for moderation. Multiple imputations were used to handle missing data.

RESULTS

The g-estimation showed a strong causal effect of cognitive inflexibility reducing depression and anxiety and modified by education level. In a counterfactual framework, a hypothetical intervention that could lower cognitive flexibility seemed to cause improvement in mental distress at the subsequent time-point (negative sign) for low education. The less flexibility, the larger improvement. For high education, the same but weaker effect was found, with a change in sign, negative during the intervention and positive during follow-up.

DISCUSSION

An unexpected and strong effect was found from cognitive inflexibility on symptom improvement. This study demonstrates how to estimate causal psychological effects with standard software in an observational dataset with substantial missing and shows the value of such methods.

Generational, sex, and socioeconomic inequalities in mental and social wellbeing during the COVID-19 pandemic: prospective longitudinal observational study of five UK cohorts

BACKGROUND

Research suggests that there have been inequalities in the impact of the coronavirus disease 2019 (COVID-19) pandemic and related non-pharmaceutical interventions on population mental health. We explored generational, sex, and socioeconomic inequalities during the first year of the pandemic using nationally representative cohorts from the UK.

METHODS

We analysed data from 26772 participants from five longitudinal cohorts representing generations born between 1946 and 2000, collected in May 2020, September-October 2020, and February-March 2021 across all five cohorts. We used a multilevel growth curve modelling approach to investigate generational, sex, and socioeconomic differences in levels of anxiety and depressive symptomatology, loneliness, and life satisfaction (LS) over time.

RESULTS

Younger generations had worse levels of mental and social wellbeing throughout the first year of the pandemic. Whereas these generational inequalities narrowed between the first and last observation periods for LS [-0.33 (95% CI -0.51 to -0.15)], they became larger for anxiety [0.22 (0.10, 0.33)]. Generational inequalities in depression and loneliness did not change between the first and last observation periods, but initial depression levels of the youngest cohort were worse than expected if the generational inequalities had not accelerated. Women and those experiencing financial difficulties had worse initial mental and social wellbeing levels than men and those financially living comfortably, respectively, and these gaps did not substantially differ between the first and last observation periods.

CONCLUSIONS

By March 2021, mental and social wellbeing inequalities persisted in the UK adult population. Pre-existing generational inequalities may have been exacerbated with the pandemic onset. Policies aimed at protecting vulnerable groups are needed.

Fertility treatment increases the risk of preterm birth independent of multiple gestations

Objective

To investigate the complex interplay between fertility treatment, multiple gestations, and prematurity.

Design

Retrospective cohort study linking the national Center for Disease Control and Prevention infant birth and death data from 2014 to 2018.

Setting

National database from Center of Disease Control and Prevention.

Patients

In total, 19,454,155 live-born infants with gestational ages 22-44 weeks, 114,645 infants born using non IVF fertility treatment (NIFT), and 179,960 via assisted reproductive technology (ART).

Intervention

Noninvasive fertility treatment or ART vs. spontaneously conceived pregnancies.

Main Outcome Measures

The main outcome assessed was prematurity. Formal mediation analysis was conducted to calculate the percentage mediated by multiple gestations.

Results

Newborns born using NIFT or ART compared with those with no fertility treatment had a higher incidence of multiple gestation (no fertility treatment = 3.0%; NIFT = 24.7%; ART = 32.7%; P<.001) and prematurity (no fertility treatment = 11.2%; NIFT = 23.4%; ART = 28.4%; P<.001). Mediation analysis demonstrates that 76.8% (95% confidence interval [CI], 75.2%-78.1%) of the effect of NIFT on prematurity was mediated through multiple gestations. Similarly, 71.2% (95% CI, 70.8%-72.7%) of the effect of ART on prematurity is mediated through multiple gestation. However, the direct effect of NIFT on prematurity is 20.4% (95% CI, 19.0%-22.0%). The direct effect of ART was 24.7% (95% CI, 23.7%-25.6%).

Conclusion

A significant proportion of prematurity associated with fertility treatment is mediated by the treatment itself, independent of multiple gestations.

The pathways from disadvantaged socioeconomic status in childhood to edentulism in mid-to-late adulthood over the life-course

BACKGROUND

This study aimed to examine the direct and indirect pathways from childhood socioeconomic status (SES) to the prevalence of edentulism in mid-to-late age Chinese individuals using structural equation modeling (SEM).

METHODS

This study analyzed data from 17,032 mid- to-late age Chinese individuals in the 2014 and 2015 China Health and Retirement Longitudinal Study (CHARLS). Childhood SES was determined based on the parents' education and occupation, financial situation of the family, primary residence, food availability, and medical convenience. Adulthood SES was established according to educational achievements of the individuals. Edentulism is defined as the loss of all natural teeth. SEM was used to examine the statistical significance of the association between childhood SES and edentulism, mediated by childhood health, adulthood SES, and adult health.

RESULTS

Childhood SES had significant indirect (β = -0.026, p < 0.01), and total (β = -0.040, p < 0.01) effects on edentulism. It was determined that 65% of the total effect of childhood SES on edentulism was indirect, and mainly mediated by adult SES. Also, the goodness-of-fit indices of the best-fitting model were acceptable.

CONCLUSION

This study revealed that childhood health, adult health and adult SES are mediators that explain the relationship between childhood SES and edentulism. The global attention to alleviate the inequality in edentulism should focus on exploring recommendations and intervention strategies from childhood to adulthood, by considering adult SES, childhood and adult health.

Association of Population Well-Being With Cardiovascular Outcomes

Importance

Mortality from cardiovascular disease (CVD) varies across communities and is associated with known structural and population health factors. Still, a population's well-being, including sense of purpose, social relationships, financial security, and relationship to community, may be an important target to improve cardiovascular health.

Objective

To examine the association of population level measures of well-being with rates of CVD mortality in the US.

Design, Setting, and Participants

This cross-sectional study linked data from the Gallup National Health and Well-Being Index (WBI) survey to county-level rates of CVD mortality from the Centers for Disease Control and Prevention Atlas of Heart Disease and Stroke. Participants were respondents of the WBI survey, which was conducted by Gallup with randomly selected adults aged 18 years or older from 2015 to 2017. Data were analyzed from August 2022 to May 2023.

Main Outcomes and Measures

The primary outcome was the county-level rate of total CVD mortality; secondary outcomes were mortality rates for stroke, heart failure, coronary heart disease, acute myocardial infarction, and total heart disease. The association of population well-being (measured using a modified version of the WBI) with CVD mortality was assessed, and an analysis of whether the association was modified by county structural factors (Area Deprivation Index [ADI], income inequality, and urbanicity) and population health factors (percentages of the adult population who had hypertension, diabetes, or obesity; were currently smoking; and were physically inactive) was conducted. Population WBI and its ability to mediate the association of structural factors associated with CVD using structural equation models was also assessed.

Results

Well-being surveys were completed by 514 971 individuals (mean [SD] age 54.0 [19.2] years; 251 691 [48.9%] women; 379 521 [76.0%] White respondents) living in 3228 counties. Mortality rates for CVD decreased from a mean of 499.7 (range, 174.2-974.7) deaths per 100 000 persons in counties with the lowest quintile of population well-being to 438.6 (range, 110.1-850.4) deaths per 100 000 persons in counties with the highest quintile of population well-being. Secondary outcomes showed similar patterns. In the unadjusted model, the effect size (SE) of WBI on CVD mortality was -15.5 (1.5; P < .001), or a decrease of 15 deaths per 100 000 persons for each 1-point increase of population well-being. After adjusting for structural factors and structural plus population health factors, the association was attenuated but still significant, with an effect size (SE) of -7.3 (1.6; P < .001); for each 1-point increase in well-being, the total cardiovascular death rate decreased by 7.3 deaths per 100 000 persons. Secondary outcomes showed similar patterns, with mortality due to coronary heart disease and heart failure being significant in fully adjusted models. In mediation analyses, associations of income inequality and ADI with CVD mortality were all partly mediated by the modified population WBI.

Conclusions and Relevance

In this cross-sectional study assessing the association of well-being and cardiovascular outcomes, higher well-being, a measurable, modifiable, and meaningful outcome, was associated with lower CVD mortality, even after controlling for structural and cardiovascular-related population health factors, indicating that well-being may be a focus for advancing cardiovascular health.

Disparities in 1-Year-Mortality in Infants With Cyanotic Congenital Heart Disease: Insights From Contemporary National Data

BACKGROUND

Racial inequities in congenital heart disease (CHD) outcomes are well documented, but contributing factors warrant further investigation. We examined the interplay between race, socioeconomic position, and neonatal variables (prematurity and small for gestational age) on 1-year death in infants with CHD. We hypothesize that socioeconomic position mediates a significant part of observed racial disparities in CHD outcomes.

METHODS

Linked birth/death files from the Natality database for all liveborn neonates in the United States were examined from 2014 to 2018. Infants with cyanotic CHD were identified. Non-Hispanic Black (NHB) and Hispanic infants were compared with non-Hispanic White (NHW) infants. The primary outcome was 1-year death. Socioeconomic position was defined as maternal education and insurance status. Variables included as mediators were prematurity, small for gestational age, and socioeconomic position. Structural equation modeling was used to calculate the contribution of each mediator to the disparity in 1-year death.

RESULTS

We identified 7167 NHW, 1393 NHB, and 1920 Hispanic infants with cyanotic CHD. NHB race and Hispanic ethnicity were associated with increased 1-year death compared to NHW (OR, 1.43 [95% CI, 1.25-1.64] and 1.17 [95% CI, 1.03-1.33], respectively). The effect of socioeconomic position explained 28.2% (CI, 15.1-54.8) of the death disparity between NHB and NHW race and 100% (CI, 42.0-368) of the disparity between Hispanic and NHW. This was mainly driven by maternal education (21.3% [CI, 12.1-43.3] and 82.8% [CI, 33.1-317.8], respectively) while insurance status alone did not explain a significant percentage. The direct effect of race or ethnicity became nonsignificant: NHB versus NHW 43.1% (CI, -0.3 to 63.6) and Hispanic versus NHW -19.0% (CI, -329.4 to 45.3).

CONCLUSIONS

Less privileged socioeconomic position, especially lower maternal education, explains a large portion of the 1-year death disparity in Black and Hispanic infants with CHD. These findings identify targets for social interventions to decrease racial disparities.

The role of pharmaceutical side-effects in depression among immigrants

OBJECTIVES

This study explores the role of pharmaceuticals with depression or suicidality as a side-effect in explaining the immigrant depression paradox. Immigrants generally report less depression than their native-born peers, despite the socio environments that are less conducive to well-being. This immigrant advantage in mental health tends to recede with time in the US and more acculturation. To date, an explanation for this pattern has remained elusive, partly because acculturation is also associated with many desirable outcomes, suggesting less depression with more acculturation.

DESIGN

Data came from seven two-year waves (2005-2006 to 2017-2018) of the National Health and Nutrition Examination Survey (NHANES). Depression was measured using the Patient Health Questionnaire (PHQ-9). Linear and logistic regressions were used to estimate the immigrant differences in depression, while controlling for sociodemographic characteristics, healthcare access, health conditions, and the use of medications with depression or suicidality as a side-effect.

RESULTS

30.3% and 22.7% of US-born adults used at least one medication with depression or suicidality as a side-effect, compared to 16.4% and 9.2% of foreign-born adults. Access to healthcare improved with time in the US and with acculturation, and both of these factors were also positively associated with the use of medications with depression or suicidality as a side-effect. The magnitude of the mediation associated with medication side-effects was significant, in many cases sufficient to eliminate the relationship between acculturation - whether expressed in terms of time in the US, English-language use, or nativity - and depression.

CONCLUSION

Exposure to medications with depression or suicidality as a side-effect helped explain part of the relative mental health advantage of foreign-born residents, as well as the diminishing advantage associated with time in the US and with acculturation.

Socioeconomic and sex inequalities in parent-reported adolescent mental ill-health: time trends in four British birth cohorts

BACKGROUND

Studies using symptom-based screeners have suggested that mental ill-health has increased in adolescents in recent decades, however, few studies have tested the equivalence of their instruments, which is critical for inferring changes in prevalence. In addition, little research has explored whether socioeconomic position (SEP) and sex inequalities in adolescent mental health have changed over time.

METHODS

Using structural equation modelling, we explored SEP and sex differences in harmonised parent reports of emotional and behavioural problems, using data from four UK birth cohorts: the 1958 National Child Development Study (NCDS'58; n = 10,868), the 1970 British Cohort Study (BCS'70; n = 8,242), the 1991-92 Avon Longitudinal Study of Parents and Children (ALSPAC'91; n = 5,389), and the 2000-01 Millennium Cohort Study (MCS'01; n = 9,338).

RESULTS

Compared with the two earliest cohorts, members of MCS'01 had higher latent mean scores on emotional problems (both sexes), and lower scores on behavioural problems (females only). The associations between four indicators of SEP and emotional problems were strongest in MCS'01, with housing tenure having the strongest association. All four SEP indicators were associated with behavioural problems in each cohort, with housing tenure again more strongly associated with problems in the MCS'01. Mediation analyses suggested that the increase in emotional problems occurred despite broadly improving socioeconomic conditions.

CONCLUSIONS

Our findings suggest that parent reports of adolescent emotional problems, but not behavioural problems, have risen in recent generations and this trend is not solely due to reporting styles. A failure to address widening social inequalities may result in further increases in mental ill-health amongst disadvantaged young people.

Area disadvantage and mental health over the life course: a 69-year prospective birth cohort study

PURPOSE

Existing evidence on the mental health consequences of disadvantaged areas uses cross-sectional or longitudinal studies with short observation periods. The objective of this research was to investigate this association over a 69-year period.

METHODS

Data were obtained from the MRC National Survey of Health and Development (NSHD; the British 1946 birth cohort), which consisted of 2125 participants at 69 years. We assessed longitudinal associations between area disadvantage and mental health symptoms at adolescence and adulthood with use of multilevel modelling framework.

RESULTS

After adjustment for father's social class, for each one percentage increase in area disadvantage at age 4, there was a 0.02 (95% CI 0.001, 0.04) mean increase in the total score of the neuroticism scale at age 13-15. After adjustment for father's social class, adult socio-economic position, cognitive ability and educational attainment, a one percentage increase in change score of area disadvantage between age 4 and 26 was associated with a mean increase in the total Psychiatric Symptom Frequency score (MD 0.06; 95% CI 0.007, 0.11). Similar associations were observed with change scores between ages 4, 53, 60 and total General Health Questionnaire-28 score at age 53 (MD 0.05; 95% CI 0.01, 0.11) and 60-64 (MD 0.06; 95% CI 0.009, 0.11).

CONCLUSIONS

Cohort members who experienced increasing area disadvantage from childhood were at increased risk of poor mental health over the life course. Population-wide interventions aiming at improving social and physical aspects of the early neighbourhood environment could reduce the socio-economic burden of poor mental health.

Maternal postnatal depressive symptoms and offspring emotional and behavioral development at age 7 years in a U.K. birth cohort: The role of paternal involvement

There is considerable variability in developmental outcomes of children whose mothers experience depression. Few longitudinal studies have examined contributions of paternal involvement in the association between maternal postnatal depression (PND) and offspring development. We examined pathways from maternal PND at 8 weeks (Edinburgh Postnatal Depression Scale; total score) to offspring emotional and behavioral development at 7 years (Strengths and Difficulties Questionnaire; total score) through behavioral, affective, and cognitive dimensions of paternal involvement in a U.K.-based birth cohort (Avon Longitudinal Study of Parents and Children; n = 3,434). Analyses were adjusted for baseline confounders and paternal PND (Edinburgh Postnatal Depression Scale; total score) as an intermediate confounder. Maternal PND was strongly associated with offspring development, but this association was not mediated by the combination of all indirect pathways through various dimensions of paternal involvement. Only father-child conflict emerged as a risk factor for adverse offspring development and as a mediator in the association between maternal PND and offspring development (albeit the effect size was small). If found causal, interventions that reduce father-child conflict may reduce the risk of adverse development in offspring of mothers with PND. (PsycInfo Database Record (c) 2023 APA, all rights reserved).

Could interventions on physical activity mitigate genomic liability for obesity? Applying the health disparity framework in genetically informed studies

Polygenic scores (PGS) are now commonly available in longitudinal cohort studies, leading to their integration into epidemiological research. In this work, our aim is to explore how polygenic scores can be used as exposures in causal inference-based methods, specifically mediation analyses. We propose to estimate the extent to which the association of a polygenic score indexing genetic liability to an outcome could be mitigated by a potential intervention on a mediator. To do this this, we use the interventional disparity measure approach, which allows us to compare the adjusted total effect of an exposure on an outcome, with the association that would remain had we intervened on a potentially modifiable mediator. As an example, we analyse data from two UK cohorts, the Millennium Cohort Study (MCS, N = 2575) and the Avon Longitudinal Study of Parents and Children (ALSPAC, N = 3347). In both, the exposure is genetic liability for obesity (indicated by a PGS for BMI), the outcome is late childhood/early adolescent BMI, and the mediator and potential intervention target is physical activity, measured between exposure and outcome. Our results suggest that a potential intervention on child physical activity can mitigate some of the genetic liability for childhood obesity. We propose that including PGSs in a health disparity measure approach, and causal inference-based methods more broadly, is a valuable addition to the study of gene-environment interplay in complex health outcomes.

New proposal to address mediation analysis interrogations by using genetic variants as instrumental variables

The application of causal mediation analysis (CMA) considering the mediation effect of a third variable is increasing in epidemiological studies; however, this requires fitting strong assumptions on confounding bias. To address this limitation, we propose an extension of CMA combining it with Mendelian randomization (MRinCMA). We applied the new approach to analyse the causal effect of obesity and diabetes on pancreatic cancer, considering each factor as potential mediator. To check the performance of MRinCMA under several conditions/scenarios, we used it in different simulated data sets and compared it with structural equation models. For continuous variables, MRinCMA and structural equation models performed similarly, suggesting that both approaches are valid to obtain unbiased estimates. When noncontinuous variables were considered, MRinCMA presented, overall, lower bias than structural equation models. By applying MRinCMA, we did not find any evidence of causality of obesity or diabetes on pancreatic cancer. With this new methodology, researchers would be able to address CMA hypotheses by appropriately accounting for the confounding bias assumption regardless of the conditions used in their studies in different settings.

Adverse Maternal Fetal Environment Partially Mediates Disparate Outcomes in Non-White Neonates with Major Congenital Heart Disease

OBJECTIVE

To determine whether differential exposure to an adverse maternal fetal environment partially explains disparate outcomes in infants with major congenital heart disease (CHD).

STUDY DESIGN

Retrospective cohort study utilizing a population-based administrative California database (2011-2017). Primary exposure: Race/ethnicity. Primary mediator: Adverse maternal fetal environment (evidence of maternal metabolic syndrome and/or maternal placental syndrome).

OUTCOMES

Composite of 1-year mortality or severe morbidity and days alive out of hospital in the first year of life (DAOOH). Mediation analyses determined the percent contributions of mediators on pathways between race/ethnicity and outcomes after adjusting for CHD severity.

RESULTS

Included were 2747 non-Hispanic White infants (reference group), 5244 Hispanic, and 625 non-Hispanic Black infants. Hispanic and non-Hispanic Black infants had a higher risk for composite outcome (crude OR: 1.18; crude OR: 1.25, respectively) and fewer DAOOH (-6 & -12 days, respectively). Compared with the reference group, Hispanic infants had higher maternal metabolic syndrome exposure (43% vs 28%, OR: 1.89), and non-Hispanic Black infants had higher maternal metabolic syndrome (44% vs 28%; OR: 1.97) and maternal placental syndrome exposure (18% vs 12%; OR, 1.66). Both maternal metabolic syndrome exposure (OR: 1.21) and maternal placental syndrome exposure (OR: 1.56) were related to composite outcome and fewer DAOOH (-25 & -16 days, respectively). Adverse maternal fetal environment explained 25% of the disparate relationship between non-Hispanic Black race and composite outcome and 18% of the disparate relationship between Hispanic ethnicity and composite outcome. Adverse maternal fetal environment explained 16% (non-Hispanic Black race) and 21% (Hispanic ethnicity) of the association with DAOOH.

CONCLUSIONS

Increased exposure to adverse maternal fetal environment contributes to racial and ethnic disparities in major CHD outcomes.

Sensitivity analysis for unobserved confounding in causal mediation analysis allowing for effect modification, censoring and truncation

Causal mediation analysis is used to decompose the total effect of an exposure on an outcome into an indirect effect, taking the path through an intermediate variable, and a direct effect. To estimate these effects, strong assumptions are made about unconfoundedness of the relationships between the exposure, mediator and outcome. These assumptions are difficult to verify in a given situation and therefore a mediation analysis should be complemented with a sensitivity analysis to assess the possible impact of violations. In this paper we present a method for sensitivity analysis to not only unobserved mediator-outcome confounding, which has largely been the focus of previous literature, but also unobserved confounding involving the exposure. The setting is estimation of natural direct and indirect effects based on parametric regression models. We present results for combinations of binary and continuous mediators and outcomes and extend the sensitivity analysis for mediator-outcome confounding to cases where the continuous outcome variable is censored or truncated. The proposed methods perform well also in the presence of interactions between the exposure, mediator and observed confounders, allowing for modeling flexibility as well as exploration of effect modification. The performance of the method is illustrated through simulations and an empirical example.

Mediating role of body composition and insulin resistance on the association of arterial stiffness with blood pressure among adolescents: The ALSPAC study

Background

Emerging evidence among adolescents suggests that arterial stiffness temporally precedes elevated blood pressure/hypertension in the casual pathway. It remains unknown whether insulin resistance and body composition mediate this relationship. Therefore, we examined the mediating role of total fat mass, lean mass, and insulin resistance in the association between arterial stiffness and blood pressure among adolescents.

Materials and methods

We studied 3,764 participants, aged 17 years from the Avon Longitudinal Study of Parents and Children (ALSPAC) United Kingdom birth cohort. Arterial stiffness accessed with Vicorder device measured carotid-femoral pulse wave velocity (cfPWV), body composition was measured by dual-energy Xray Absorptiometry, blood pressure by Omron device, and homeostatic model assessment of insulin resistance (HOMA-IR) was computed. Data were analysed with structural equation models mediation path analyses and adjusted for cardiometabolic and lifestyle factors.

Results

Among 1,678 [44.6%] male and 2,086 [55.4%] female participants, higher cfPWV was directly and independently associated with higher systolic and diastolic blood pressure, irrespective of the mediator [Standardized regression coefficient (β) = 0.248–0.370, p for all = 0.002]. Lean mass [β = 0.010; p = 0.026; 3.3% mediation] and HOMA-IR [β = 0.004; p = 0.033; 1.1% mediation] but not total fat mass [β &lt; 0.0001; p = 0.615; 0% mediation] partly mediated the association of cfPWV with systolic blood pressure after full adjustments. Similarly, lean mass [β = –0.004; p = 0.021; 1.4% mediation] and HOMA-IR [β = 0.007; p = 0.039; 2.8% mediation] but not total fat mass [β = –0.002; p = 0.665; 0.7% mediation] partly mediated the association of cfPWV with diastolic blood pressure.

Conclusion

Attenuating insulin resistance may be a potentially valuable strategy in lowering higher blood pressure precipitated by higher arterial stiffness.

The gendered impacts of delayed parenthood: A dynamic analysis of young adulthood

Young adulthood is a dynamic and demographically dense stage in the life course. This poses a challenge for research on the socioeconomic consequences of parenthood timing, which most often focuses on women. We chart the dynamics of delayed parenthood and its implications for educational and labor market trajectories for young adult women and men using a novel longitudinal analysis approach, the parametric g-formula. This method allows the estimation of both population-averaged effects (among all women and men) and average treatment effects (among mothers and fathers). Based on high-quality data from Finnish registers, we find that later parenthood exacerbates the educational advantage of women in comparison to men and attenuates the income advantage of men in comparison to women across young adult ages. Gender differences in the consequences of delayed parenthood on labor market trajectories are largely not explained by changes in educational trajectories. Moreover, at the time of entering parenthood, delayed parenthood improves the incomes of fathers more than those of mothers, thereby exacerbating existing gender differences. The results provide population-level evidence on how the delay of parenthood has contributed to the strengthening of women's educational position relative to that of men. Further, the findings on greater increases in fathers' than mothers' incomes at the time of entering parenthood, as followed by postponement, may help explain why progress in achieving gender equality in the division of paid and unpaid work in families has been slow.

Pharmaceutical Side Effects and the Sex Differences in Depression and Distress

INTRODUCTION

Women suffer from depression at higher rates than men. This difference is well established, although a consolidated explanation remains elusive. This study examines the role played by medications with depression or suicidality as a potential side effect in explaining the sex difference in depression.

METHODS

Data were analyzed for 224,810 U.S. adults aged ≥18 years from the 2008-2018 Medical Expenditure Panel Survey. Linear and logistic regressions were used to assess the sex differences in distress and depression while controlling for sociodemographic characteristics, healthcare access, health conditions, and the use of medications with depression or suicidality as a side effect.

RESULTS

41% and 28% of women used ≥1 medication with depression and suicidality as a side effect compared with 27% and 17% of men, respectively. When controlling for sociodemographic characteristics, healthcare access, and health conditions, women were more likely to report significant distress (OR=1.16, 95% CI=1.10, 1.24) and major depression (OR=1.12, 95% CI=1.07, 1.18) than men. In models that further adjusted for the use of medications with depression or suicidality as a side effect, the sex differences became statistically nonsignificant for both distress (OR=0.97, 95% CI=0.91, 1.03) and depression (OR=0.97, 95% CI=0.92, 1.02). Nonhormone medications (rather than hormone medications) with such side effects helped explain the sex differences in distress and depression.

CONCLUSIONS

Findings suggest a significant sex difference in pharmaceutical treatment and the potential consequences of pharmaceutical side effects on distress and depression. These results highlight the importance of pharmaceutical side effects in understanding health and health disparities.

Relationship of Sodium Intake With Granulocytes, Renal and Cardiovascular Outcomes in the Prospective EPIC-Norfolk Cohort

Background Experimental studies show that high-sodium intake affects the innate immune system, among others with increased circulating granulocytes. Whether this relationship exists on a population level and whether this relates to disease outcomes is unclear. We aimed to test the hypotheses that (1) sodium intake is associated with granulocytes on a population level; (2) granulocytes are associated with the presence of hypertension and both cardiovascular and renal outcomes; and (3) the relation between high-sodium intake and these outcomes is mediated by granulocytes. Methods and Results We performed an analysis in 13 804 participants from the prospective EPIC (European Prospective Investigation into Cancer)-Norfolk cohort, with a mean age of 58 years and median follow-up of 19.3 years. Analyses were carried out using calculated estimated sodium intake and sodium-to-potassium ratios from spot urines at baseline. The main outcomes were hypertension at baseline, and composite cardiovascular (mortality or cardiovascular events) and renal (mortality or renal events) outcomes during follow-up. Sodium intake and urine sodium-to-potassium ratio were positively associated with circulating granulocyte concentrations after adjustment for confounders (β=0.03; P=0.028 and β=0.06; P<0.001, respectively). Granulocytes significantly mediated the associations of, respectively, sodium intake and urine sodium-to-potassium ratio with hypertension at baseline, and cardiovascular and renal outcomes. Conclusions Sodium intake is positively associated with circulating granulocyte concentrations, and higher granulocyte concentrations associate with worse long-term cardiovascular and renal outcomes. Given the recently established immune-modulating effects of sodium and the role of immune cells in both cardiovascular and renal disease, causality for this pathway may need consideration in further studies.

How do psychologically based interventions for chronic musculoskeletal pain work? A systematic review and meta-analysis of specific moderators and mediators of treatment

Psychologically based interventions aim to improve pain-related functioning by targeting pain-related fears, cognitions and behaviors. Mediation and moderation analyses permit further examination of the effect of treatment on an outcome. This systematic review and meta-analysis aims to synthetize the evidence of specific mediators and moderators (i.e., treatment targets) of psychologically based treatment effects on pain and disability. A total of 28 mediation and 11 moderation analyses were included. Thirteen mediation studies were included in a meta-analysis, and the rest was narratively synthetized. Reductions in pain-related fear (indirect effect [IE]: -0.07; 95% confidence interval [CI]: -0.11, -0.04) and catastrophizing (IE: -0.07; 95%CI: -0.14, -0.00), as well as increases in self-efficacy (IE: -0.07; 95%CI: -0.11, -0.04), mediated effects of cognitive behavioral therapy on disability but not on pain intensity, when compared to control treatments. Enhancing pain acceptance (IE: -0.17; 95%CI: -0.31, -0.03) and psychological flexibility (IE: -0.30; 95%CI: -0.41, -0.18) mediated acceptance and commitment therapy effects on disability. The narrative synthesis showed conflicting evidence, which did not support a robust moderated effect for any of the examined constructs. Overall, the methodological quality regarding mediation was low, and some key pitfalls are highlighted alongside recommendations to provide a platform for future research.

Consumption of ultra-processed foods and IL-6 in two cohorts from high- and middle-income countries

This study evaluated the association between ultra-processed foods (UPF) on serum IL-6 and to investigate the mediation role of adiposity. Participants were 524 adults from the EPITeen Cohort (Porto, Portugal) and 2888 participants from the 1982 Pelotas Birth Cohort (Pelotas, Brazil). Dietary intake was collected using FFQ when participants were 21 years of age in the EPITeen and 23 years in the Pelotas Cohort. Serum IL-6 and body fat mass were evaluated when participants were 27 and 30 years old in the EPITeen and Pelotas, respectively. Generalised linear models were fitted to test main associations. Mediation of body fat mass was estimated using G-computation. After adjustment for socio-economic and behaviour variables, among females from the EPITeen, the concentration of IL-6 (pg/ml) increased with increasing intake of UPF from 1·31 (95 % CI 0·95, 1·82) in the first UPF quartile to 2·20 (95 % CI 1·60, 3·01) and 2·64 (95 % CI 1·89, 3·69) for the third and fourth UPF quartiles, respectively. A similar result was found among males in the Pelotas Cohort, IL-6 increased from 1·40 (95 % CI 1·32, 1·49) in the first UPF quartile to 1·50 (95 % CI 1·41, 1·59) and 1·59 (95 % CI 1·49, 1·70) in the two highest UPF quartiles. The P-value for the linear trend was < 0·01 in both findings. The indirect effect through fat mass was NS. Our findings suggest that the consumption of UPF was associated with an increase in IL-6 concentration; however, this association was not explained by adiposity.

Study protocol: randomised controlled hybrid type 2 trial evaluating the scale-up of two arts interventions for postnatal depression and Parkinson's disease

INTRODUCTION

Research on the benefits of 'arts' interventions to improve individuals' physical, social and psychological well-being is growing, but evidence on implementation and scale-up into health and social care systems is lacking. This protocol reports the SHAPER-Implement programme (Scale-up of Health-Arts Programmes Effectiveness-Implementation Research), aimed at studying the impact, implementation and scale-up of: Melodies for Mums (M4M), a singing intervention for postnatal depression; and Dance for Parkinson's (PD-Ballet) a dance intervention for Parkinson's disease. We examine how they could be embedded in clinical pathways to ensure their longer-term sustainability.

METHODS AND ANALYSIS

A randomised two-arm effectiveness-implementation hybrid type 2 trial design will be used across M4M/PD-Ballet. We will assess the implementation in both study arms (intervention vs control), and the cost-effectiveness of implementation. The design and measures, informed by literature and previous research by the study team, were refined through stakeholder engagement. Participants (400 in M4M; 160 in PD-Ballet) will be recruited to the intervention or control group (2:1 ratio). Further implementation data will be collected from stakeholders involved in referring to, delivering or supporting M4M/PD-Ballet (N=25-30 for each intervention).A mixed-methods approach (surveys and semi-structured interviews) will be employed. 'Acceptability' (measured by the 'Acceptability Intervention Measure') is the primary implementation endpoint for M4M/PD-Ballet. Relationships between clinical and implementation outcomes, implementation strategies (eg, training) and outcomes will be explored using generalised linear mixed models. Qualitative data will assess factors affecting the acceptability, feasibility and appropriateness of M4M/PD-Ballet, implementation strategies and longer-term sustainability. Costs associated with implementation and future scale-up will be estimated.

ETHICS AND DISSEMINATION

SHAPER-PND (the M4M trial) and SHAPER-PD (the PD trial) are approved by the West London and GTAC (20/PR/0813) and the HRA and Health and Care Research Wales (REC Reference: 20/WA/0261) Research Ethics Committees. Study findings will be disseminated through scientific peer-reviewed journals and scientific conferences.

TRIAL REGISTRATION NUMBERS

Both trials are registered with NIH US National Library of Medicine, ClinicalTrials.gov. The trial registration numbers, URLs of registry records, and dates of registration are: (1) PD-Ballet: URL: NCT04719468 (https://eur03.safelinks.protection.

OUTLOOK

com/?url=https%3A%2F%2Fwww.clinicaltrials.gov%2Fct2%2Fshow%2FNCT04719468%3Fterm%3DNCT04719468%26draw%3D2%26rank%3D1&data=04%7C01%7Crachel.davis%40kcl.ac.uk%7C11a7c5142782437919f808d903111449%7C8370cf1416f34c16b83c724071654356%7C0%7C0%7C6375441942616) (date of registration: 22 Jan 2021). (2) Melodies for Mums: NCT04834622 (https://clinicaltrials.gov/ct2/show/NCT04834622?term=shaper-pnd&draw=2&rank=1) (date of registration: 8 Apr 2021).

Educational note: causal decomposition of population health differences using Monte Carlo integration and the g-formula

One key objective of the population health sciences is to understand why one social group has different levels of health and well-being compared with another. Whereas several methods have been developed in economics, sociology, demography, and epidemiology to answer these types of questions, a recent method introduced by Jackson and VanderWeele (2018) provided an update to decompositions by anchoring them within causal inference theory. In this paper, we demonstrate how to implement the causal decomposition using Monte Carlo integration and the parametric g-formula. Causal decomposition can help to identify the sources of differences across populations and provide researchers with a way to move beyond estimating inequalities to explaining them and determining what can be done to reduce health disparities. Our implementation approach can easily and flexibly be applied for different types of outcome and explanatory variables without having to derive decomposition equations. We describe the concepts of the approach and the practical steps and considerations needed to implement it. We then walk through a worked example in which we investigate the contribution of smoking to sex differences in mortality in South Korea. For this example, we provide both pseudocode and R code using our package, cfdecomp. Ultimately, we outline how to implement a very general decomposition algorithm that is grounded in counterfactual theory but still easy to apply to a wide range of situations.

Identifying causal mechanisms in psychotherapy: What can we learn from causal mediation analysis?

Despite widespread interest in the development of process-based psychotherapies, little is still known about the underlying processes that underpin our most effective therapies. Statistical mediation analysis is a commonly used analytical method to evaluate how, or by which processes, a therapy causes change in an outcome. Causal mediation analysis (CMA) represents a new advancement in mediation analysis that employs causally defined direct and indirect effects based on potential outcomes. These novel ideas and analytical techniques have been characterized as revolutionary in epidemiology and biostatistics, although they are not (yet) widely known among researchers in clinical psychology. In this paper, I outline the fundamental concepts underlying CMA, clarify the differences between the CMA approach and the traditional approach to mediation, and identify two important data analytical aspects that have been emphasized as a result of these recent advancements. To illustrate the key ideas, assumptions, and mathematical definitions intuitively, an applied clinical example from a previously published randomized controlled trial is used. CMA's main contributions are discussed, as well as some of the key challenges. Finally, it is argued that the most significant contribution of CMA is the formalization of mediation in a unified causal framework with clear assumptions.

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.

Parametric-Regression-Based Causal Mediation Analysis of Binary Outcomes and Binary Mediators: Moving Beyond the Rareness or Commonness of the Outcome

In the causal mediation framework, several parametric-regression–based approaches have been introduced in the last decade for estimating natural direct and indirect effects. For a binary outcome, a number of proposed estimators use a logistic model and rely on specific assumptions or approximations that may be delicate or not easy to verify in practice. To circumvent the challenges prompted by the rare outcome assumption in this context, an exact closed-form natural-effects estimator on the odds ratio scale was recently introduced for a binary mediator. In this work, we further push this exact approach and extend it for the estimation of natural effects on the risk ratio and risk difference scales. Explicit formulas for the delta method standard errors are provided. The performance of our proposed exact estimators is demonstrated in simulation scenarios featuring various levels of outcome rareness/commonness. The total effect decomposition property on the multiplicative scales is also examined. Using a SAS macro (SAS Institute, Inc., Cary, North Carolina) we developed, our approach is illustrated to assess the separate effects of exposure to inhaled corticosteroids and placental abruption on low birth weight mediated by prematurity. Our exact natural-effects estimators are found to work properly in both simulations and the real data example.

DNA methylation as a mediator of associations between the environment and chronic diseases: A scoping review on application of mediation analysis

DNA methylation (DNAm) is one of the most studied epigenetic modifications. DNAm has emerged as a key biological mechanism and biomarkers to test associations between environmental exposure and outcomes in epidemiological studies. Although previous studies have focused on associations between DNAm and either exposure/outcomes, it is useful to test for mediation of the association between exposure and outcome by DNAm. The purpose of this scoping review is to introduce the methodological essence of statistical mediation analysis and to examine emerging epidemiological research applying mediation analyses. We conducted this scoping review for published peer-reviewed journals on this topic using online databases (PubMed, Scopus, Cochrane, and CINAHL) ending in December 2020. We extracted a total of 219 articles by initial screening. After reviewing titles, abstracts, and full texts, a total of 69 articles were eligible for this review. The breakdown of studies assigned to each category was 13 for smoking (18.8%), 8 for dietary intake and famine (11.6%), 6 for other lifestyle factors (8.7%), 8 for clinical endpoints (11.6%), 22 for environmental chemical exposures (31.9%), 2 for socioeconomic status (SES) (2.9%), and 10 for genetic factors and race (14.5%). In this review, we provide an exposure-wide summary for the mediation analysis using DNAm levels. However, we found heterogenous methods and interpretations in mediation analysis with typical issues such as different cell compositions and tissue-specificity. Further accumulation of evidence with diverse exposures, populations and with rigorous methodology will be expected to provide further insight in the role of DNAm in disease susceptibility.

Does postponing retirement affect cognitive function? A counterfactual experiment to disentangle life course risk factors

Evidence suggests that contemporaneous labor force participation affects cognitive function; however, it is unclear whether it is employment itself or endogenous factors related to individuals’ likelihood of employment that protects against cognitive decline. We exploit innovations in counterfactual causal inference to disentangle the effect of postponing retirement on later-life cognitive function from the effects of other life-course factors. With the U.S. Health and Retirement Study (1996–2014, n = 20,469), we use the parametric g-formula to estimate the effect of postponing retirement to age 67. We also study whether the benefit of postponing retirement is affected by gender, education, and/or occupation, and whether retirement affects cognitive function through depressive symptoms or comorbidities. We find that postponing retirement is protective against cognitive decline, accounting for other life-course factors (population: 0.34, 95% confidence interval (CI): 0.20,0.47; individual: 0.43, 95% CI: 0.26,0.60). The extent of the protective effect depends on subgroup, with the highest educated experiencing the greatest mitigation of cognitive decline (individual: 50%, 95% CI: 32%,71%). By using innovative models that better reflect the empirical reality of interconnected life-course processes, this work makes progress in understanding how retirement affects cognitive function.

•

Research is inconclusive as to employments' protection against cognitive decline.

•

Causal inference models can better reflect interconnected life-course processes.

•

The parametric g-formula shows a substantial protective effect of ongoing employment.

•

Protective effect holds for all subgroups, but is greatest for the highest educated.

Insights into the Cross-world Independence Assumption of Causal Mediation Analysis

Causal mediation analysis is a useful tool for epidemiologic research, but it has been criticized for relying on a "cross-world" independence assumption that counterfactual outcome and mediator values are independent even in causal worlds where the exposure assignments for the outcome and mediator differ. This assumption is empirically difficult to verify and problematic to justify based on background knowledge. In the present article, we aim to assist the applied researcher in understanding this assumption. Synthesizing what is known about the cross-world independence assumption, we discuss the relationship between assumptions for causal mediation analyses, causal models, and nonparametric identification of natural direct and indirect effects. In particular, we give a practical example of an applied setting where the cross-world independence assumption is violated even without any post-treatment confounding. Further, we review possible alternatives to the cross-world independence assumption, including the use of bounds that avoid the assumption altogether. Finally, we carry out a numeric study in which the cross-world independence assumption is violated to assess the ensuing bias in estimating natural direct and indirect effects. We conclude with recommendations for carrying out causal mediation analyses.

Using graphic modelling to identify modifiable mediators of the association between area-based deprivation at birth and offspring unemployment

BACKGROUND

Deprivation can perpetuate across generations; however, the causative pathways are not well understood. Directed acyclic graphs (DAG) with mediation analysis can help elucidate and quantify complex pathways in order to identify modifiable factors at which to target interventions.

METHODS AND FINDINGS

We linked ten Scotland-wide databases (six health and four education) to produce a cohort of 217,226 pupils who attended Scottish schools between 2009 and 2013. The DAG comprised 23 potential mediators of the association between area deprivation at birth and subsequent offspring 'not in education, employment or training' status, covering maternal, antenatal, perinatal and child health, school engagement, and educational factors. Analyses were performed using modified g-computation. Deprivation at birth was associated with a 7.3% increase in offspring 'not in education, employment or training'. The principal mediators of this association were smoking during pregnancy (natural indirect effect of 0·016, 95% CI 0·013, 0·019) and school absences (natural indirect effect of 0·021, 95% CI 0·018, 0·024), explaining 22% and 30% of the total effect respectively. The proportion of the association potentially eliminated by addressing these factors was 19% (controlled direct effect when set to non-smoker 0·058; 95% CI 0·053, 0·063) for smoking during pregnancy and 38% (controlled direct effect when set to no absences 0·043; 95% CI 0·037, 0·049) for school absences.

CONCLUSIONS

Combining a DAG with mediation analysis helped disentangle a complex public health problem and quantified the modifiable factors of maternal smoking and school absence that could be targeted for intervention. This study also demonstrates the general utility of DAGs in understanding complex public health problems.

Causal inference with observational data: the need for triangulation of evidence

The goal of much observational research is to identify risk factors that have a causal effect on health and social outcomes. However, observational data are subject to biases from confounding, selection and measurement, which can result in an underestimate or overestimate of the effect of interest. Various advanced statistical approaches exist that offer certain advantages in terms of addressing these potential biases. However, although these statistical approaches have different underlying statistical assumptions, in practice they cannot always completely remove key sources of bias; therefore, using design-based approaches to improve causal inference is also important. Here it is the design of the study that addresses the problem of potential bias - either by ensuring it is not present (under certain assumptions) or by comparing results across methods with different sources and direction of potential bias. The distinction between statistical and design-based approaches is not an absolute one, but it provides a framework for triangulation - the thoughtful application of multiple approaches (e.g. statistical and design based), each with their own strengths and weaknesses, and in particular sources and directions of bias. It is unlikely that any single method can provide a definite answer to a causal question, but the triangulation of evidence provided by different approaches can provide a stronger basis for causal inference. Triangulation can be considered part of wider efforts to improve the transparency and robustness of scientific research, and the wider scientific infrastructure and system of incentives.

Understanding disparities in cancer prognosis: An extension of mediation analysis to the relative survival framework

Mediation analysis can be applied to investigate the effect of a third variable on the pathway between an exposure and the outcome. Such applications include investigating the determinants that drive differences in cancer survival across subgroups. However, cancer disparities may be the result of complex mechanisms that involve both cancer-related and other-cause mortality differences making it difficult to identify the causing factors. Relative survival, a commonly used measure in cancer epidemiology, can be used to focus on cancer-related differences. We extended mediation analysis to the relative survival framework for exploring cancer inequalities. The marginal effects were obtained using regression standardization, after fitting a relative survival model. Contrasts of interests included both marginal relative survival and marginal all-cause survival differences between exposure groups. Such contrasts include the indirect effect due to a mediator that is identifiable under certain assumptions. A separate model was fitted for the mediator and uncertainty was estimated using parametric bootstrapping. The avoidable deaths under interventions can also be estimated to quantify the impact of eliminating differences. The methods are illustrated using data for individuals diagnosed with colon cancer. Mediation analysis within relative survival allows focus on factors that account for cancer-related differences instead of all-cause differences and helps improve our understanding on cancer inequalities.

Explaining better hearing in Norway: a comparison of two cohorts 20 years apart - the HUNT study

Background

The hearing function at a given age seems to have improved in more recent born cohorts in industrialized countries. But the reasons for the improvement have not yet been explained.

Methods

We investigated the extent to which better hearing in Norway is attributed to modifiable risk factors by using representative demographic and audiometric data from two cohorts of the Trøndelag Health Study, HUNT2 (1996–1998) and HUNT4 (2017–2019). We estimated natural indirect effects using causal inference methods in order to assess whether cohort improvement in hearing thresholds (HTs) was mediated by occupational noise exposure, recurrent ear infections, smoking and education.

Results

The improvement in HTs from HUNT2 to HUNT4 was 2.8 and 3.0 dB at low respectively high frequencies. Together all risk factors mediated this improvement by 0.8 dB (95% CI 0.7–0.9) and 0.8 dB (95% CI 0.7–0.9) respectively, corresponding to mediated proportions of 27 and 28%. Substantial mediation was specifically found for occupational noise in men and recurrent ear infections in women (mediated proportions of 11 and 17% at high frequencies, respectively).

Conclusions

Increased education, less occupational noise exposure, ear infections and smoking contributed considerably to better hearing in Norway the last two decades.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12889-021-10301-1.

Anthropogenic landscape decreases mosquito biodiversity and drives malaria vector proliferation in the Amazon rainforest

Inter-relationships among mosquito vectors, Plasmodium parasites, human ecology, and biotic and abiotic factors, drive malaria risk. Specifically, rural landscapes shaped by human activities have a great potential to increase the abundance of malaria vectors, putting many vulnerable people at risk. Understanding at which point the abundance of vectors increases in the landscape can help to design policies and interventions for effective and sustainable control. Using a dataset of adult female mosquitoes collected at 79 sites in malaria endemic areas in the Brazilian Amazon, this study aimed to (1) verify the association among forest cover percentage (PLAND), forest edge density (ED), and variation in mosquito diversity; and to (2) test the hypothesis of an association between landscape structure (i.e., PLAND and ED) and Nyssorhynchus darlingi (Root) dominance. Mosquito collections were performed employing human landing catch (HLC) (peridomestic habitat) and Shannon trap combined with HLC (forest fringe habitat). Nyssorhynchus darlingi abundance was used as the response variable in a generalized linear mixed model, and the Shannon diversity index (H') of the Culicidae community, PLAND, and the distance house-water drainage were used as predictors. Three ED categories were also used as random effects. A path analysis was used to understand comparative strengths of direct and indirect relationships among Amazon vegetation classes, Culicidae community, and Ny. darlingi abundance. Our results demonstrate that Ny. darlingi is negatively affected by H´ and PLAND of peridomestic habitat, and that increasing these variables (one-unit value at β0 = 768) leads to a decrease of 226 (P < 0.001) and 533 (P = 0.003) individuals, respectively. At the forest fringe, a similar result was found for H' (β1 = -218; P < 0.001) and PLAND (β1 = -337; P = 0.04). Anthropogenic changes in the Amazon vegetation classes decreased mosquito biodiversity, leading to increased Ny. darlingi abundance. Changes in landscape structure, specifically decreases in PLAND and increases in ED, led to Ny. darlingi becoming the dominant species, increasing malaria risk. Ecological mechanisms involving changes in landscape and mosquito species composition can help to understand changes in the epidemiology of malaria.

Formulating causal questions and principled statistical answers

Although review papers on causal inference methods are now available, there is a lack of introductory overviews on what they can render and on the guiding criteria for choosing one particular method. This tutorial gives an overview in situations where an exposure of interest is set at a chosen baseline ("point exposure") and the target outcome arises at a later time point. We first phrase relevant causal questions and make a case for being specific about the possible exposure levels involved and the populations for which the question is relevant. Using the potential outcomes framework, we describe principled definitions of causal effects and of estimation approaches classified according to whether they invoke the no unmeasured confounding assumption (including outcome regression and propensity score-based methods) or an instrumental variable with added assumptions. We mainly focus on continuous outcomes and causal average treatment effects. We discuss interpretation, challenges, and potential pitfalls and illustrate application using a "simulation learner," that mimics the effect of various breastfeeding interventions on a child's later development. This involves a typical simulation component with generated exposure, covariate, and outcome data inspired by a randomized intervention study. The simulation learner further generates various (linked) exposure types with a set of possible values per observation unit, from which observed as well as potential outcome data are generated. It thus provides true values of several causal effects. R code for data generation and analysis is available on www.ofcaus.org, where SAS and Stata code for analysis is also provided.

Invited Commentary: Frontiers of Power Assessment in Mediation Analysis

The development of tools for power and sample-size calculations for mediation analysis has lagged far behind the development of methods. The accompanying paper by Rudolph et al. (Am J Epidemiol. 2020;189(12):1559-1567) is a helpful contribution in using simulations as a tool for power calculations for more complex methods and settings. Much work remains to be done in the development of easy-to-use packages and simple online websites for carrying out power and sample-size calculations for mediation analysis. Much remains to be learned with respect to the relative power of different methods in different settings. There will likely be feedback between these 2 important frontiers of the tools we have available and of our understanding of power when conducting mediation analysis.

Socioeconomic Life Course Models and Oral Health: A Longitudinal Analysis

We compared socioeconomic life course models to decompose the direct and mediated effects of socioeconomic status (SES) in different periods of life on late-life oral health. We used data from 2 longitudinal Swedish studies: the Level of Living Survey and the Swedish Panel Study of Living Conditions of the Oldest Old. Two birth cohorts (older, 1925 to 1934; younger, 1944 to 1953) were followed between 1968 and 2011 with 6 waves. SES was measured with 4 indicators of SES and modeled as a latent variable. Self-reported oral health was based on a tooth conditions question. Variables in the younger and older cohorts were grouped into 4 periods: childhood, young/mid-adulthood, mid /late adulthood, late adulthood/life. We used structural equation modeling to fit the following into lagged-effects life course models: 1) chain of risk, 2) sensitive period with late-life effect, 3) sensitive period with early- and late-life effects, 4) accumulation of risks with cross-sectional effects, and 5) accumulation of risks. Chain of risk was incorporated into all models and combined with accumulation, with cross-sectional effects yielding the best fit (older cohort: comparative fit index = 0.98, Tucker-Lewis index = 0.98, root mean square error of approximation = 0.04, weighted root mean square residual = 1.51). For the older cohort, the chain of SES from childhood → mid-adulthood → late adulthood → late life showed the following respective standardized coefficients: 053, 0.92, and 0.97. The total effect of childhood SES on late-life tooth loss (standardized coefficient: -0.23 for older cohort, -0.17 for younger cohort) was mediated by previous tooth loss and SES. Cross-sectional effects of SES on tooth loss were observed throughout the life course, but the strongest coefficients were at young/mid-adulthood (standardized coefficient: -0.41 for older cohort, -0.45 for younger cohort). SES affects oral health cumulatively over the life course and through a chain of risks. Actions to improve socioeconomic conditions in early life might have long-lasting effects on health if they help prevent people from becoming trapped in a chain of risks.

The association between family structure and children's BMI over time-the mediating role of income

PURPOSE

Although both family structure and income have previously been indicated as being associated with body mass index (BMI), the extent to which the effect of family structure on BMI is mediated through income is incompletely understood. Taking the case of the United Kingdom, this study aims to investigate the association between family structure, defined in this study as whether children live in a one- or two-adult household, and childhood BMI, and whether this varies by child sex and with increased age. Second, the study aims to examine whether family equivalised income, as a proxy for socioeconomic status, mediates the association between family structure and childhood BMI.

METHODS

This study uses data from the Millennium Cohort Study. Data from 7478 children born between 2000 and 2001 in the UK at the ages of 3, 5, 7, 11, and 14 were used. Mediation analysis was used to consider, at each age, the extent to which the association between living in a one- or two-adult household and BMI was mediated through income overall and stratified by sex. To assess the robustness of the mediation analysis estimates, we used both E-values and multiple confounder adjustment.

RESULTS

At ages 3 and 5, there was no direct or indirect effect of family structure mediated by income on BMI. Between the ages of 7 and 11, the overall proportion of the association mediated vastly increased, from 19.70% at age 7 up to 42.70% at the age of 11. The E-values show that substantial unmeasured confounder associations would be needed to fully explain away the conclusions from the mediation analysis. Results remained significant when models were additionally adjusted for geographic region, the main respondent's (usually the mother's) highest educational attainment, and ethnicity.

CONCLUSIONS

An increasing proportion of the association between family structure and BMI is mediated by income as children grow older. The study focuses on the mediating role of income between family structure and BMI using the available data as an empirical application of the potential impact of income as mediator in the causal pathway.

Estimation of Controlled Direct Effects in Longitudinal Mediation Analyses with Latent Variables in Randomized Studies

In a randomized study with longitudinal data on a mediator and outcome, estimating the direct effect of treatment on the outcome at a particular time requires adjusting for confounding of the association between the outcome and all preceding instances of the mediator. When the confounders are themselves affected by treatment, standard regression adjustment is prone to severe bias. In contrast, G-estimation requires less stringent assumptions than path analysis using SEM to unbiasedly estimate the direct effect even in linear settings. In this article, we propose a G-estimation method to estimate the controlled direct effect of treatment on the outcome, by adapting existing G-estimation methods for time-varying treatments without mediators. The proposed method can accommodate continuous and noncontinuous mediators, and requires no models for the confounders. Unbiased estimation only requires correctly specifying a mean model for either the mediator or the outcome. The method is further extended to settings where the mediator or outcome, or both, are latent, and generalizes existing methods for single measurement occasions of the mediator and outcome to longitudinal data on the mediator and outcome. The methods are utilized to assess the effects of an intervention on physical activity that is possibly mediated by motivation to exercise in a randomized study.

The LifeCycle Project-EU Child Cohort Network: a federated analysis infrastructure and harmonized data of more than 250,000 children and parents

Early life is an important window of opportunity to improve health across the full lifecycle. An accumulating body of evidence suggests that exposure to adverse stressors during early life leads to developmental adaptations, which subsequently affect disease risk in later life. Also, geographical, socio-economic, and ethnic differences are related to health inequalities from early life onwards. To address these important public health challenges, many European pregnancy and childhood cohorts have been established over the last 30 years. The enormous wealth of data of these cohorts has led to important new biological insights and important impact for health from early life onwards. The impact of these cohorts and their data could be further increased by combining data from different cohorts. Combining data will lead to the possibility of identifying smaller effect estimates, and the opportunity to better identify risk groups and risk factors leading to disease across the lifecycle across countries. Also, it enables research on better causal understanding and modelling of life course health trajectories. The EU Child Cohort Network, established by the Horizon2020-funded LifeCycle Project, brings together nineteen pregnancy and childhood cohorts, together including more than 250,000 children and their parents. A large set of variables has been harmonised and standardized across these cohorts. The harmonized data are kept within each institution and can be accessed by external researchers through a shared federated data analysis platform using the R-based platform DataSHIELD, which takes relevant national and international data regulations into account. The EU Child Cohort Network has an open character. All protocols for data harmonization and setting up the data analysis platform are available online. The EU Child Cohort Network creates great opportunities for researchers to use data from different cohorts, during and beyond the LifeCycle Project duration. It also provides a novel model for collaborative research in large research infrastructures with individual-level data. The LifeCycle Project will translate results from research using the EU Child Cohort Network into recommendations for targeted prevention strategies to improve health trajectories for current and future generations by optimizing their earliest phases of life.

Evidence synthesis for constructing directed acyclic graphs (ESC-DAGs): a novel and systematic method for building directed acyclic graphs

Background

Directed acyclic graphs (DAGs) are popular tools for identifying appropriate adjustment strategies for epidemiological analysis. However, a lack of direction on how to build them is problematic. As a solution, we propose using a combination of evidence synthesis strategies and causal inference principles to integrate the DAG-building exercise within the review stages of research projects. We demonstrate this idea by introducing a novel protocol: ‘Evidence Synthesis for Constructing Directed Acyclic Graphs’ (ESC-DAGs)’.

Methods

ESC-DAGs operates on empirical studies identified by a literature search, ideally a novel systematic review or review of systematic reviews. It involves three key stages: (i) the conclusions of each study are ‘mapped’ into a DAG; (ii) the causal structures in these DAGs are systematically assessed using several causal inference principles and are corrected accordingly; (iii) the resulting DAGs are then synthesised into one or more ‘integrated DAGs’. This demonstration article didactically applies ESC-DAGs to the literature on parental influences on offspring alcohol use during adolescence.

Conclusions

ESC-DAGs is a practical, systematic and transparent approach for developing DAGs from background knowledge. These DAGs can then direct primary data analysis and DAG-based sensitivity analysis. ESC-DAGs has a modular design to allow researchers who are experienced DAG users to both use and improve upon the approach. It is also accessible to researchers with limited experience of DAGs or evidence synthesis.

A Viable Alternative When Propensity Scores Fail: Evaluation of Inverse Propensity Weighting and Sequential G-Estimation in a Two-Wave Mediation Model

Two methods from the potential outcomes framework - inverse propensity weighting (IPW) and sequential G-estimation - were evaluated and compared to linear regression for estimating the mediated effect in a two-wave design with a randomized intervention and continuous mediator and outcome. Baseline measures of the mediator and outcome can be considered confounders of the follow-up mediator - outcome relation for which adjustment is necessary to eliminate bias. To adjust for baseline measures of the mediator and outcome, IPW uses stabilized inverse propensity weights whereas sequential G-estimation uses regression adjustment. Theoretical differences between the models are described, and Monte Carlo simulations compared the performance of linear regression; IPW without weight truncation; IPW with weights truncated at the 1st/99th, 5th/95th, and 10th/90th percentiles; and sequential G-estimation. Sequential G-estimation performed similarly to linear regression, but IPW provided a biased estimate of the mediated effect, lower power, lower confidence interval coverage, and higher mean squared error. Simulation results show that IPW failed to fully adjust the follow-up mediator - outcome relation for confounding due to the baseline measures. We then compared the mediated effect estimates using data from a randomized experiment evaluating a steroid prevention program for high school athletes. Implications and future directions are discussed.

Bayesian Causal Mediation Analysis with Multiple Ordered Mediators

Causal mediation analysis provides investigators insight into how a treatment or exposure can affect an outcome of interest through one or more mediators on causal pathway. When multiple mediators on the pathway are causally ordered, identification of mediation effects on certain causal pathways requires a sensitivity parameter to be specified. A mixed model-based approach was proposed in the Bayesian framework to connect potential outcomes at different treatment levels, and identify mediation effects independent of a sensitivity parameter, for the natural direct and indirect effects on all causal pathways. The proposed method is illustrated in a linear setting for mediators and outcome, with mediator-treatment interactions. Sensitivity analysis was performed for the prior choices in the Bayesian models. The proposed Bayesian method was applied to an adolescent dental health study, to see how social economic status can affect dental caries through a sequence of causally ordered mediators in dental visit and oral hygiene index.

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.

Understanding cognitive impairment in mood disorders: mediation analyses in the UK Biobank cohort

BACKGROUND

Cognitive impairment is strongly linked with persistent disability in people with mood disorders, but the factors that explain cognitive impairment in this population are unclear.

AIMS

To estimate the total effect of (a) bipolar disorder and (b) major depression on cognitive function, and the magnitude of the effect that is explained by potentially modifiable intermediate factors.

METHOD

Cross-sectional study using baseline data from the UK Biobank cohort. Participants were categorised as having bipolar disorder (n = 2709), major depression (n = 50 975) or no mood disorder (n = 102 931 and n = 105 284). The outcomes were computerised tests of reasoning, reaction time and memory. The potential mediators were cardiometabolic disease and psychotropic medication. Analyses were informed by graphical methods and controlled for confounding using regression, propensity score-based methods and G-computation.

RESULTS

Group differences of small magnitude were found on a visuospatial memory test. Z-score differences for the bipolar disorder group were in the range -0.23 to -0.17 (95% CI -0.39 to -0.03) across different estimation methods, and for the major depression group they were approximately -0.07 (95% CI -0.10 to -0.03). One-quarter of the effect was mediated via psychotropic medication in the bipolar disorder group (-0.05; 95% CI -0.09 to -0.01). No evidence was found for mediation via cardiometabolic disease.

CONCLUSIONS

In a large community-based sample in middle to early old age, bipolar disorder and depression were associated with lower visuospatial memory performance, in part potentially due to psychotropic medication use. Mood disorders and their treatments will have increasing importance for population cognitive health as the proportion of older adults continues to grow.

DECLARATION OF INTEREST

I.J.D. is a UK Biobank participant. J.P.P. is a member of the UK Biobank Steering Committee.