On Well-defined Hypothetical Interventions in the Potential Outcomes Framework

Interpretations of Studies on SARS-CoV-2 Vaccination and Post-acute COVID-19 Sequelae

This article discusses causal interpretations of epidemiologic studies of the effects of vaccination on sequelae after acute severe acute respiratory syndrome coronavirus 2 infection. To date, researchers have tried to answer several different research questions on this topic. While some studies assessed the impact of postinfection vaccination on the presence of or recovery from post-acute coronavirus disease 2019 syndrome, others quantified the association between preinfection vaccination and postacute sequelae conditional on becoming infected. However, the latter analysis does not have a causal interpretation, except under the principal stratification framework-that is, this comparison can only be interpreted as causal for a nondiscernible stratum of the population. As the epidemiology of coronavirus disease 2019 is now nearly entirely dominated by reinfections, including in vaccinated individuals, and possibly caused by different Omicron subvariants, it has become even more important to design studies on the effects of vaccination on postacute sequelae that address precise causal questions and quantify effects corresponding to implementable interventions.

Mediterranean Diet Adherence and Risk of All-Cause Mortality in Women

Importance

Higher adherence to the Mediterranean diet has been associated with reduced risk of all-cause mortality, but data on underlying molecular mechanisms over long follow-up are limited.

Objectives

To investigate Mediterranean diet adherence and risk of all-cause mortality and to examine the relative contribution of cardiometabolic factors to this risk reduction.

Design, Setting, and Participants

This cohort study included initially healthy women from the Women's Health Study, who had provided blood samples, biomarker measurements, and dietary information. Baseline data included self-reported demographics and a validated food-frequency questionnaire. The data collection period was from April 1993 to January 1996, and data analysis took place from June 2018 to November 2023.

Exposures

Mediterranean diet score (range, 0-9) was computed based on 9 dietary components.

Main Outcome and Measures

Thirty-three blood biomarkers, including traditional and novel lipid, lipoprotein, apolipoprotein, inflammation, insulin resistance, and metabolism measurements, were evaluated at baseline using standard assays and nuclear magnetic resonance spectroscopy. Mortality and cause of death were determined from medical and death records. Cox proportional hazards regression was used to calculate hazard ratios (HRs) for Mediterranean diet adherence and mortality risk, and mediation analyses were used to calculate the mediated effect of different biomarkers in understanding this association.

Results

Among 25 315 participants, the mean (SD) baseline age was 54.6 (7.1) years, with 329 (1.3%) Asian women, 406 (1.6%) Black women, 240 (0.9%) Hispanic women, 24 036 (94.9%) White women, and 95 (0.4%) women with other race and ethnicity; the median (IQR) Mediterranean diet adherence score was 4.0 (3.0-5.0). Over a mean (SD) of 24.7 (4.8) years of follow-up, 3879 deaths occurred. Compared with low Mediterranean diet adherence (score 0-3), adjusted risk reductions were observed for middle (score 4-5) and upper (score 6-9) groups, with HRs of 0.84 (95% CI, 0.78-0.90) and 0.77 (95% CI, 0.70-0.84), respectively (P for trend < .001). Further adjusting for lifestyle factors attenuated the risk reductions, but they remained statistically significant (middle adherence group: HR, 0.92 [95% CI, 0.85-0.99]; upper adherence group: HR, 0.89 [95% CI, 0.82-0.98]; P for trend = .001). Of the biomarkers examined, small molecule metabolites and inflammatory biomarkers contributed most to the lower mortality risk (explaining 14.8% and 13.0%, respectively, of the association), followed by triglyceride-rich lipoproteins (10.2%), body mass index (10.2%), and insulin resistance (7.4%). Other pathways, including branched-chain amino acids, high-density lipoproteins, low-density lipoproteins, glycemic measures, and hypertension, had smaller contributions (<3%).

Conclusions and Relevance

In this cohort study, higher adherence to the Mediterranean diet was associated with 23% lower risk of all-cause mortality. This inverse association was partially explained by multiple cardiometabolic factors.

The Mediterranean Diet, Cardiometabolic Biomarkers, and Risk of All-Cause Mortality: A 25-Year Follow-Up Study of the Women's Health Study

Background

Higher consumption of Mediterranean diet (MED) intake has been associated with reduced risk of all-cause mortality but limited data are available examining long-term outcomes in women or the underlying molecular mechanisms of this inverse association in human populations. We aimed to investigate the association of MED intake with long-term risk of all-cause mortality in women and to better characterize the relative contribution of traditional and novel cardiometabolic factors to the MED-related risk reduction in morality.

Methods

In a prospective cohort study of 25,315 initially healthy women from the Women's Health Study, we assessed dietary MED intake using a validated semiquantitative food frequency questionnaire according to the usual 9-category measure of MED adherence. Baseline levels of more than thirty cardiometabolic biomarkers were measured using standard assays and targeted nuclear magnetic resonance spectroscopy, including lipids, lipoproteins, apolipoproteins, inflammation, glucose metabolism and insulin resistance, branched-chain amino acids, small metabolites, and clinical factors. Mortality and cause of death was ascertained prospectively through medical and death records.

Results

During a mean follow-up of 25 years, 3,879 deaths were ascertained. Compared to the reference group of low MED intake (0-3, approximately the bottom tertile), and adjusting for age, treatment, and energy intake, risk reductions were observed for the middle and upper MED groups with respective HRs of 0.84 (95% CI 0.78-0.90) and 0.77 (95% CI 0.70-0.84), p for trend <0.0001. Further adjusting for smoking, physical activity, alcohol intake and menopausal factors attenuated the risk reductions which remained significant with respective HRs of 0.92 (95% CI 0.85-0.99) and 0.89 (95% CI 0.82-0.98), p for trend 0.0011. Risk reductions were generally similar for CVD and non-CVD mortality. Small molecule metabolites (e.g., alanine and homocysteine) and inflammation made the largest contributions to lower mortality risk (accounting for 14.8% and 13.0% of the benefit of the MED-mortality association, respectively), followed by triglyceride-rich lipoproteins (10.2%), adiposity (10.2%) and insulin resistance (7.4%), with lesser contributions (<3%) from other pathways including branched-chain amino acids, high-density lipoproteins, low-density lipoproteins, glycemic measures, and hypertension.

Conclusions

In the large-scale prospective Women's Health Study of 25,315 initially healthy US women followed for 25 years, higher MED intake was associated with approximately one fifth relative risk reduction in mortality. The inverse association was only partially explained by known novel and traditional cardiometabolic factors.

Visualizing the (Causal) Effect of a Continuous Variable on a Time-To-Event Outcome

Visualization is a key aspect of communicating the results of any study aiming to estimate causal effects. In studies with time-to-event outcomes, the most popular visualization approach is depicting survival curves stratified by the variable of interest. This approach cannot be used when the variable of interest is continuous. Simple workarounds, such as categorizing the continuous covariate and plotting survival curves for each category, can result in misleading depictions of the main effects. Instead, we propose a new graphic, the survival area plot, to directly depict the survival probability over time and as a function of a continuous covariate simultaneously. This plot utilizes g-computation based on a suitable time-to-event model to obtain the relevant estimates. Through the use of g-computation, those estimates can be adjusted for confounding without additional effort, allowing a causal interpretation under the standard causal identifiability assumptions. If those assumptions are not met, the proposed plot may still be used to depict noncausal associations. We illustrate and compare the proposed graphics to simpler alternatives using data from a large German observational study investigating the effect of the Ankle-Brachial Index on survival. To facilitate the usage of these plots, we additionally developed the contsurvplot R-package, which includes all methods discussed in this paper.

Tell me what you want, what you really really want: Estimands in observational pharmacoepidemiologic comparative effectiveness and safety studies

PURPOSE

Ideally, the objectives of a pharmacoepidemiologic comparative effectiveness or safety study should dictate its design and data analysis. This paper discusses how defining an estimand is instrumental to this process.

METHODS

We applied the ICH-E9 (Statistical Principles for Clinical Trials) R1 addendum on estimands - which originally focused on randomized trials - to three examples of observational pharmacoepidemiologic comparative effectiveness and safety studies. Five key elements specify the estimand: the population, contrasted treatments, endpoint, intercurrent events, and population-level summary measure.

RESULTS

Different estimands were defined for case studies representing three types of pharmacological treatments: (1) single-dose treatments using a case study about the effect of influenza vaccination versus no vaccination on mortality risk in an adult population of ≥60 years of age; (2) sustained-treatments using a case study about the effect of dipeptidyl peptidase 4 inhibitor versus glucagon-like peptide-1 agonist on hypoglycemia risk in treatment of uncontrolled diabetes; and (3) as needed treatments using a case study on the effect of nitroglycerin spray as-needed versus no nitroglycerin on syncope risk in treatment of stabile angina pectoris.

CONCLUSIONS

The case studies illustrated that a seemingly clear research question can still be open to multiple interpretations. Defining an estimand ensures that the study targets a treatment effect that aligns with the treatment decision the study aims to inform. Estimand definitions further help to inform choices regarding study design and data-analysis and clarify how to interpret study findings.

Adverse childhood experiences and adult outcomes using a causal framework perspective: Challenges and opportunities

BACKGROUND

Research on the effect of adverse childhood experiences (ACEs) on adult outcomes has typically relied on retrospective assessment of ACEs and cumulative scores. However, this approach raises methodological challenges that can limit the validity of findings.

OBJECTIVE

The aims of this paper are 1) to present the value of directed acyclic graphs (DAGs) to identify and mitigate potential problems related to confounding and selection bias, and 2) to question the meaning of a cumulative ACE score.

RESULTS

Adjusting for variables that post-date childhood could block mediated pathways that are part of the total causal effect while conditioning on adult variables, which often serve as proxies for childhood variables, can create collider stratification bias. Because exposure to ACEs can affect the likelihood of reaching adulthood or study entry, selection bias could be introduced via restricting selection on a variable affected by ACEs in the presence of unmeasured confounding. In addition to challenges regarding causal structure, using a cumulative score of ACEs assumes that each type of adversity will have the same effect on a given outcome, which is unlikely considering differing risk across adverse experiences.

CONCLUSIONS

DAGs provide a transparent approach of the researchers' assumed causal relationships and can be used to overcome issues related to confounding and selection bias. Researchers should be explicit about their operationalization of ACEs and how it is to be interpreted in the context of the research question they are trying to answer.

Tying research question and analytical strategy when variables are affected by medication use

Ill-defined research questions could be particularly problematic in an epidemiological setting where measurements fluctuate over time due to intercurrent events, such as medication use. When a research question fails to specify how medication use should be handled methodologically, arbitrary decisions may be made during the analysis phase, which likely leads to a mismatch between the intended question and the performed analysis. The mismatch can result in vastly different or meaningless interpretations of estimated effects. Thus, a research question such as "what is the effect of X on Y?" requires further elaboration, and it should consider whether and how medication use has affected the measurements of interest. In our study, we will discuss how well-defined questions can be formulated when medication use is involved in observational studies. We will distinguish between a situation where an exposure is affected by medication use and where the outcome of interest is affected by medication use. For each setting, we will give examples of different research questions that could be asked depending on how medication use is considered in the estimand and discuss methodological considerations under each question.

COVID-19 and the eye: alternative facts The 2022 Bowman Club, David L. Easty lecture

In addition to catastrophic loss of life, and dramatic and unwanted alterations to the daily lives of those left behind, the COVID-19 pandemic has fostered the publication and dissemination of an unprecedented quantity of peer-reviewed medical and scientific publications on a single subject. In particular, the ophthalmic literature is now replete with clinical and laboratory studies on putative eye involvement by SARS-CoV-2, the aetiologic agent of COVID-19. In this review, we critically appraise the published literature on COVID-19, and suggest that the quality of scientific peer review and editorial decision-making also suffered during the COVID-19 pandemic.

Adolescent neighborhood disadvantage and memory performance in young adulthood

OBJECTIVE

Using data from the National Longitudinal Study of Adolescent to Adult Health, we estimated the average causal effect of neighborhood disadvantage in adolescence on memory performance in young adulthood. We contrasted several different ways of operationalizing a continuous measure of neighborhood disadvantage including a continuous neighborhood disadvantage score and ordinal measures.

RESULTS

Neighborhood disadvantage was measured in Wave I when participants were a mean age of 15.41 years (SE: 0.12) and memory performance was measured in Wave IV when participants were a mean age of 28.24 years (SE: 0.12). We found that adolescent neighborhood disadvantage was associated with decreased memory performance in young adulthood. Notably, we observed a linear decline in word recall score among those in the less disadvantaged tail of the distribution (neighborhood disadvantage <1), a finding not observed using traditional ordinal variable classifications of disadvantage.

CONCLUSION

Experiencing neighborhood disadvantage in adolescence may have lasting impacts on cognitive health throughout the life course.

Constructed Measures and Causal Inference: Towards a New Model of Measurement for Psychosocial Constructs

Psychosocial constructs can only be assessed indirectly, and measures are typically formed by a combination of indicators that are thought to relate to the construct. Reflective and formative measurement models offer different conceptualizations of the relation between the indicators and what is sometimes conceived of as a univariate latent variable supposed to correspond in some way to the construct. I argue that the empirical implications of reflective and formative models will often be violated by data since the causally relevant constituents will generally be multivariate, not univariate. These empirical implications can be formally tested but factor analysis is not adequate to do so. I argue that formative models misconstrue the relation between the constructed measures and the underlying reality by which causal processes operate, but that reflective models misconstrue the nature of the underlying reality itself by typically presuming that the constituents of it that are causally efficacious are unidimensional. I discuss the ensuing problems arising from these misconstruals. I propose a causal interpretation of associations between constructed measures and various outcomes that is applicable to both reflective and formative models and is applicable even if the usual assumptions of these models are violated. I put forward an outline for a new model of the process of measure construction and discuss the practical implications of these observations and proposals for the provision of definitions, the selection of items, item-by-item analyses, the construction of measures, and the interpretation of the associations of these measures with subsequent outcomes.

Emulating Target Trials to Improve Causal Inference From Agent-Based Models

Agent-based models are a key tool for investigating the emergent properties of population health settings, such as infectious disease transmission, where the exposure often violates the key "no interference" assumption of traditional causal inference under the potential outcomes framework. Agent-based models and other simulation-based modeling approaches have generally been viewed as a separate knowledge-generating paradigm from the potential outcomes framework, but this can lead to confusion about how to interpret the results of these models in real-world settings. By explicitly incorporating the target trial framework into the development of an agent-based or other simulation model, we can clarify the causal parameters of interest, as well as make explicit the assumptions required for valid causal effect estimation within or between populations. In this paper, we describe the use of the target trial framework for designing agent-based models when the goal is estimation of causal effects in the presence of interference, or spillover.

Urinary sodium excretion and the risk of CVD: a community-based cohort study in Taiwan

Urinary Na excretion is a potential risk factor for CVD. However, the underlying biological mechanisms and effects of salt sensitivity are unclear. The purpose of this study was to characterise the relative contribution of biological factors to the Na–CVD association. A total of 2112 participants were enrolled in this study. Structured questionnaires and blood and urine samples were obtained. Twenty-four-hour Na excretion was estimated using a single overnight urine sample. Hypertension, the metabolic syndrome and overweight status were considered to indicate salt sensitivity. Cox proportional hazard models were used to investigate the effects of salt sensitivity on urinary Na excretion and CVD risk. The traditional mediation approach was used to calculate the proportion of mediation. The mean age (sd) of the 2112 participants was 54·5 (sd 12·2) years, and they were followed up for a mean of 14·1 (sd 8·1) years. Compared with those in the lowest quartile, the highest baseline urinary Na excretion (>4·2 g/24 h) was associated with a 43 % higher CVD risk (hazard ratio, 1·43; 95 % CI 1·02, 1·99). Participants with high urinary Na excretion, hypertension or the metabolic syndrome had a significantly high risk of CVD. The carotid intima-media thickness had the largest mediating effect (accounting for 35 % of the Na–CVD association), followed by systolic blood pressure (BP) (33 %), left ventricular mass (28 %) and diastolic BP (14 %). Higher urinary Na excretion increased the risk of CVD, which was explained largely by carotid media-thickness and systolic BP.

Mortality Effects of Hypothetical Interventions on Physical Activity and TV Viewing

INTRODUCTION

Long-term effects of physical activity and television (TV) viewing on mortality have been inferred from observational studies. The associations observed do not allow for inferences about the effects of population interventions and could be subject to bias due to time-varying confounding.

METHODS

Using data from the Australian Diabetes, Obesity and Lifestyle Study, collected in 1999-2000 (T0), 2004-2005 (T1), and 2011-2012 (T2), we applied the parametric g-formula to estimate cumulative risks of death under hypothetical interventions on physical activity and/or TV viewing determined from self-report while adjusting for time-varying confounding.

RESULTS

In the 6377 participants followed up for 13 yr from 2004 to 2005 to death or censoring in 2017, 781 participants died. The observed cumulative risk of death was 12.2%. The most effective hypothetical intervention was to increase weekly physical activity to >300 min (risk ratio (RR), 0.66 (0.46-0.86) compared with a "worst-case" scenario; RR, 0.83 (0.73-0.94) compared with no intervention). Reducing daily TV viewing to <2 h in addition to physical activity interventions did not show added survival benefits. Reducing TV viewing alone was least effective in reducing mortality (RR, 0.85 (0.60-1.10) compared with the worst-case scenario; RR, 1.06 (0.93-1.20) compared with no intervention).

CONCLUSIONS

Our findings suggested that sustained interventions to increase physical activity could lower all-cause mortality over a 13-yr period, and there might be limited gain from intervening to reduce TV viewing time in a relatively healthy population.

Estimating the Effect of Healthcare-Associated Infections on Excess Length of Hospital Stay Using Inverse Probability-Weighted Survival Curves

BACKGROUND

Studies estimating excess length of stay (LOS) attributable to nosocomial infections have failed to address time-varying confounding, likely leading to overestimation of their impact. We present a methodology based on inverse probability-weighted survival curves to address this limitation.

METHODS

A case study focusing on intensive care unit-acquired bacteremia using data from 2 general intensive care units (ICUs) from 2 London teaching hospitals were used to illustrate the methodology. The area under the curve of a conventional Kaplan-Meier curve applied to the observed data was compared with that of an inverse probability-weighted Kaplan-Meier curve applied after treating bacteremia as censoring events. Weights were based on the daily probability of acquiring bacteremia. The difference between the observed average LOS and the average LOS that would be observed if all bacteremia cases could be prevented was multiplied by the number of admitted patients to obtain the total excess LOS.

RESULTS

The estimated total number of extra ICU days caused by 666 bacteremia cases was estimated at 2453 (95% confidence interval [CI], 1803-3103) days. The excess number of days was overestimated when ignoring time-varying confounding (2845 [95% CI, 2276-3415]) or when completely ignoring confounding (2838 [95% CI, 2101-3575]).

CONCLUSIONS

ICU-acquired bacteremia was associated with a substantial excess LOS. Wider adoption of inverse probability-weighted survival curves or alternative techniques that address time-varying confounding could lead to better informed decision making around nosocomial infections and other time-dependent exposures.

Association of the Mediterranean Diet With Onset of Diabetes in the Women's Health Study

IMPORTANCE

Higher Mediterranean diet (MED) intake has been associated with reduced risk of type 2 diabetes, but underlying biological mechanisms are unclear.

OBJECTIVE

To characterize the relative contribution of conventional and novel biomarkers in MED-associated type 2 diabetes risk reduction in a US population.

DESIGN, SETTING, AND PARTICIPANTS

This cohort study was conducted among 25 317 apparently healthy women. The participants with missing information regarding all traditional and novel metabolic biomarkers or those with baseline diabetes were excluded. Participants were invited for baseline assessment between September 1992 and May 1995. Data were collected from November 1992 to December 2017 and analyzed from December 2018 to December 2019.

EXPOSURES

MED intake score (range, 0 to 9) was computed from self-reported dietary intake, representing adherence to Mediterranean diet intake.

MAIN OUTCOMES AND MEASURES

Incident cases of type 2 diabetes, identified through annual questionnaires; reported cases were confirmed by either telephone interview or supplemental questionnaire. Proportion of reduced risk of type 2 diabetes explained by clinical risk factors and a panel of 40 biomarkers that represent different physiological pathways was estimated.

RESULTS

The mean (SD) age of the 25 317 female participants was 52.9 (9.9) years, and they were followed up for a mean (SD) of 19.8 (5.8) years. Higher baseline MED intake (score ≥6 vs ≤3) was associated with as much as a 30% lower type 2 diabetes risk (age-adjusted and energy-adjusted hazard ratio, 0.70; 95% CI, 0.62-0.79; when regression models were additionally adjusted with body mass index [BMI]: hazard ratio, 0.85; 95% CI, 0.76-0.96). Biomarkers of insulin resistance made the largest contribution to lower risk (accounting for 65.5% of the MED-type 2 diabetes association), followed by BMI (55.5%), high-density lipoprotein measures (53.0%), and inflammation (52.5%), with lesser contributions from branched-chain amino acids (34.5%), very low-density lipoprotein measures (32.0%), low-density lipoprotein measures (31.0%), blood pressure (29.0%), and apolipoproteins (23.5%), and minimal contribution (≤2%) from hemoglobin A1c. In post hoc subgroup analyses, the inverse association of MED diet with type 2 diabetes was seen only among women who had BMI of at least 25 at baseline but not those who had BMI of less than 25 (eg, women with BMI <25, age- and energy-adjusted HR for MED score ≥6 vs ≤3, 1.01; 95% CI, 0.77-1.33; P for trend = .92; women with BMI ≥25: HR, 0.76; 95% CI, 0.67-0.87; P for trend < .001).

CONCLUSIONS AND RELEVANCE

In this cohort study, higher MED intake scores were associated with a 30% relative risk reduction in type 2 diabetes during a 20-year period, which could be explained in large part by biomarkers of insulin resistance, BMI, lipoprotein metabolism, and inflammation.

Educational note: types of causes

We explore the different types of causes that are commonly investigated by epidemiologists. We first distinguish between causes which are events (including actions) and causes which are states. Second, we distinguish between modifiable and non-modifiable states. This yields three types of causes: fixed states (non-modifiable), dynamic states (modifiable) and events (including actions). Different causes may have different characteristics: the methods available to study them, the types of possible biases, and therefore the types of evidence needed to infer causality, may differ according to the specific cause-effect relationship under study. Nevertheless, there are also substantial commonalities. This paper is intended to improve understanding of the different types of causes, and the different types of causality, that underpin epidemiological practice.

Invited Commentary: Counterfactuals in Social Epidemiology-Thinking Outside of "the Box"

There are tensions inherent between many of the social exposures examined within social epidemiology and the assumptions embedded in quantitative potential-outcomes-based causal inference framework. The potential-outcomes framework characteristically requires a well-defined hypothetical intervention. As noted by Galea and Hernán (Am J Epidemiol. 2020;189(3):167-170), for many social exposures, such well-defined hypothetical exposures do not exist or there is no consensus on what they might be. Nevertheless, the quantitative potential-outcomes framework can still be useful for the study of some of these social exposures by creative adaptations that 1) redefine the exposure, 2) separate the exposure from the hypothetical intervention, or 3) allow for a distribution of hypothetical interventions. These various approaches and adaptations are reviewed and discussed. However, even these approaches have their limits. For certain important historical and social determinants of health such as social movements or wars, the quantitative potential-outcomes framework with well-defined hypothetical interventions is the wrong tool. Other modes of inquiry are needed.

Causal Diagrams: Pitfalls and Tips

Graphical models are useful tools in causal inference, and causal directed acyclic graphs (DAGs) are used extensively to determine the variables for which it is sufficient to control for confounding to estimate causal effects. We discuss the following ten pitfalls and tips that are easily overlooked when using DAGs: 1) Each node on DAGs corresponds to a random variable and not its realized values; 2) The presence or absence of arrows in DAGs corresponds to the presence or absence of individual causal effect in the population; 3) “Non-manipulable” variables and their arrows should be drawn with care; 4) It is preferable to draw DAGs for the total population, rather than for the exposed or unexposed groups; 5) DAGs are primarily useful to examine the presence of confounding in distribution in the notion of confounding in expectation; 6) Although DAGs provide qualitative differences of causal structures, they cannot describe details of how to adjust for confounding; 7) DAGs can be used to illustrate the consequences of matching and the appropriate handling of matched variables in cohort and case-control studies; 8) When explicitly accounting for temporal order in DAGs, it is necessary to use separate nodes for each timing; 9) In certain cases, DAGs with signed edges can be used in drawing conclusions about the direction of bias; and 10) DAGs can be (and should be) used to describe not only confounding bias but also other forms of bias. We also discuss recent developments of graphical models and their future directions.

The Effects of Reverse Causality and Selective Attrition on the Relationship Between Body Mass Index and Mortality in Postmenopausal Women

Concerns about reverse causality and selection bias complicate the interpretation of studies of body mass index (BMI, calculated as weight (kg)/height (m)2) and mortality in older adults. The objective of this study was to investigate methodological explanations for the apparent attenuation of obesity-related risks in older adults. We used data from 68,132 participants in the Women's Health Initiative (WHI) clinical trial for this analysis. All of the participants were postmenopausal women aged 50-79 years at baseline (1993-1998). To examine reverse causality and selective attrition, we compared rate ratios from inverse probability of treatment- and censoring-weighted Poisson marginal structural models with results from an unweighted adjusted Poisson regression model. The estimated mortality rate ratios and 95% confidence intervals for BMIs of 30.0-34.9, 35.0-39.9 and ≥40.0 were 0.86 (95% confidence interval (CI): 0.77, 0.96), 0.85 (95% CI: 0.72, 0.99), and 0.88 (95% CI: 0.72, 1.07), respectively, in the unweighted model. The corresponding mortality rate ratios were 0.96 (95% CI: 0.86, 1.07), 1.12 (95% CI: 0.97, 1.29), and 1.31 95% CI: (1.08, 1.57), respectively, in the marginal structural model. Results from the inverse probability of treatment- and censoring-weighted marginal structural model were attenuated in low BMI categories and increased in high BMI categories. The results demonstrate the importance of accounting for reverse causality and selective attrition in studies of older adults.

Commentary: Causal Inference for Social Exposures

Social epidemiology seeks to describe and quantify the causal effects of social institutions, interactions, and structures on human health. To accomplish this task, we define exposures as treatments and posit populations exposed or unexposed to these well-defined regimens. This inferential structure allows us to unambiguously estimate and interpret quantitative causal parameters and to investigate how these may be affected by biases such as confounding. This paradigm has been challenged recently by some critics who favor broadening the exposures that may be studied beyond treatments to also consider states. Defining the exposure protocol of an observational study is a continuum of specificity, and one may choose to loosen this definition, incurring the cost of causal parameters that become commensurately more vague. The advantages and disadvantages of broader versus narrower definitions of exposure are matters of continuing debate in social epidemiology as in other branches of epidemiology.