Interventional probability of causation (IPoC) with epidemiological and partial mechanistic evidence: benzene vs. formaldehyde and acute myeloid leukemia (AML).

INTRODUCTION: Causal epidemiology for regulatory risk analysis seeks to evaluate how removing or reducing exposures would change disease occurrence rates. We define interventional probability of causation (IPoC) as the change in probability of a disease (or other harm) occurring over a lifetime or other specified time interval that would be caused by a specified change in exposure, as predicted by a fully specified causal model. We define the closely related concept of causal assigned share (CAS) as the predicted fraction of disease risk that would be removed or prevented by a specified reduction in exposure, holding other variables fixed. Traditional approaches used to evaluate the preventable risk implications of epidemiological associations, including population attributable fraction (PAF) and the Bradford Hill considerations, cannot reveal whether removing a risk factor would reduce disease incidence. We argue that modern formal causal models coupled with causal artificial intelligence (CAI) and realistically partial and imperfect knowledge of underlying disease mechanisms, show great promise for determining and quantifying IPoC and CAS for exposures and diseases of practical interest. METHODS: We briefly review key CAI concepts and terms and then apply them to define IPoC and CAS. We present steps to quantify IPoC using a fully specified causal Bayesian network (BN) model. Useful bounds for quantitative IPoC and CAS calculations are derived for a two-stage clonal expansion (TSCE) model for carcinogenesis and illustrated by applying them to benzene and formaldehyde based on available epidemiological and partial mechanistic evidence. RESULTS: Causal BN models for benzene and risk of acute myeloid leukemia (AML) incorporating mechanistic, toxicological and epidemiological findings show that prolonged high-intensity exposure to benzene can increase risk of AML (IPoC of up to 7e-5, CAS of up to 54%). By contrast, no causal pathway leading from formaldehyde exposure to increased risk of AML was identified, consistent with much previous mechanistic, toxicological and epidemiological evidence; therefore, the IPoC and CAS for formaldehyde-induced AML are likely to be zero. CONCLUSION: We conclude that the IPoC approach can differentiate between likely and unlikely causal factors and can provide useful upper bounds for IPoC and CAS for some exposures and diseases of practical importance. For causal factors, IPoC can help to estimate the quantitative impacts on health risks of reducing exposures, even in situations where mechanistic evidence is realistically incomplete and individual-level exposure-response parameters are uncertain. This illustrates the strength that can be gained for causal inference by using causal models to generate testable hypotheses and then obtaining toxicological data to test the hypotheses implied by the models-and, where necessary, refine the models. This virtuous cycle provides additional insight into causal determinations that may not be available from weight-of-evidence considerations alone.

Physicians' Explanatory Models of Pediatric Inflammatory Bowel Disease: A Qualitative Interview Study.

Explanatory models are culturally informed representations of illness that convey understandings of the etiology and expected course of disease. Substantial research has explored lay explanatory models, but examining physicians' clinical explanatory models can also provide insight into patients' understandings of illness because physicians are a foundational source of authoritative knowledge that shapes lay concepts of illness and disease. This study characterized the explanatory models used by pediatric gastroenterologists when explaining inflammatory bowel disease (IBD) to children. We conducted semi-structured qualitative interviews with 20 pediatric gastroenterologists across the United States about their clinical communication and explanatory models. We identified two primary explanatory models used to describe immune dysregulation in pediatric IBD: the defense and protection model, which characterizes the immune system as an army that erroneously sees the body as "non-self" and attacks it; and the switch model, which conceptualizes treatment as activating a switch that turns off a faulty immune response. We also identified two models used by some physicians to describe inflammation: the scratch and scrape model, which compares IBD inflammation to scratches or scrapes on the skin; and the bonfire model, which compares inflammation to a fire in need of extinguishing. While the use of militaristic metaphors is pervasive in medicine, describing autoimmunity as a battle against the self may lead children to perceive their body as the enemy. This may be compounded by describing the immune system as "confused" while noting its ongoing protective function. Use of these explanatory models may nevertheless improve patient disease-related knowledge.

Should Cognitive Screening Tests Be Corrected for Age and Education? Insights From a Causal Perspective.

Cognitive screening tests such as the Mini-Mental State Examination are widely used in clinical routine to predict cognitive impairment. The raw test scores are often corrected for age and education, although documented poorer discrimination performance of corrected scores has challenged this practice. Nonetheless, test correction persists, perhaps due to the seemingly counterintuitive nature of the underlying problem. We used a causal framework to inform the long-standing debate from a more intuitive angle. We illustrate and quantify the consequences of applying the age-education correction of cognitive tests on discrimination performance. In an effort to bridge theory and practical implementation, we computed differences in discrimination performance under plausible causal scenarios using Open Access Series of Imaging Studies (OASIS)-1 data. We show that when age and education are causal risk factors for cognitive impairment and independently also affect the test score, correcting test scores for age and education removes meaningful information, thereby diminishing discrimination performance.

Constructing Causal Life-Course Models: Comparative Study of Data-Driven and Theory-Driven Approaches.

Life-course epidemiology relies on specifying complex (causal) models that describe how variables interplay over time. Traditionally, such models have been constructed by perusing existing theory and previous studies. By comparing data-driven and theory-driven models, we investigated whether data-driven causal discovery algorithms can help in this process. We focused on a longitudinal data set on a cohort of Danish men (the Metropolit Study, 1953-2017). The theory-driven models were constructed by 2 subject-field experts. The data-driven models were constructed by use of the temporal Peter-Clark (TPC) algorithm. The TPC algorithm utilizes the temporal information embedded in life-course data. We found that the data-driven models recovered some, but not all, causal relationships included in the theory-driven expert models. The data-driven method was especially good at identifying direct causal relationships that the experts had high confidence in. Moreover, in a post hoc assessment, we found that most of the direct causal relationships proposed by the data-driven model but not included in the theory-driven model were plausible. Thus, the data-driven model may propose additional meaningful causal hypotheses that are new or have been overlooked by the experts. In conclusion, data-driven methods can aid causal model construction in life-course epidemiology, and combining both data-driven and theory-driven methods can lead to even stronger models.

Identifying the causal effects of long-term exposure to PM2.5 and ground surface ozone on individual medical costs in China-evidence from a representative longitudinal nationwide cohort.

BACKGROUND: There is little evidence on whether PM2.5 and ground surface ozone have consistent effects on increased individual medical costs, and there is a lack of evidence on causality in developing countries. METHODS: This study utilized balanced panel data from 2014, 2016, and 2018 waves of the Chinese Family Panel Study. The Tobit model was developed within a counterfactual causal inference framework, combined with a correlated random effects and control function approach (Tobit-CRE-CF), to explore the causal relationship between long-term exposure to air pollution and medical costs. We also explored whether different air pollutants exhibit comparable effects. RESULTS: This study encompassed 8928 participants and assessed various benchmark models, highlighting the potential biases from failing to account for air pollution endogeneity or overlooking respondents without medical costs. Using the Tobit-CRE-CF model, significant effects of air pollutants on increased individual medical costs were identified. Specifically, margin effects for PM2.5 and ground-level ozone signifying that a unit increase in PM2.5 and ground-level ozone results in increased total medical costs of 199.144 and 75.145 RMB for individuals who incurred fees in the previous year, respectively. CONCLUSIONS: The results imply that long-term exposure to air pollutants contributes to increased medical costs for individuals, offering valuable insights for policymakers aiming to mitigate air pollution's consequences.

Invited Commentary: Machine Learning in Causal Inference-How Do I Love Thee? Let Me Count the Ways.

In this issue of the Journal, Mooney et al. (Am J Epidemiol. 2021;190(8):1476-1482) discuss machine learning as a tool for causal research in the style of Internet headlines. Here we comment by adapting famous literary quotations, including the one in our title (from "Sonnet 43" by Elizabeth Barrett Browning (Sonnets From the Portuguese, Adelaide Hanscom Leeson, 1850)). We emphasize that any use of machine learning to answer causal questions must be founded on a formal framework for both causal and statistical inference. We illustrate the pitfalls that can occur without such a foundation. We conclude with some practical recommendations for integrating machine learning into causal analyses in a principled way and highlight important areas of ongoing work.

Causal pathways to health-related quality of life in children with juvenile idiopathic arthritis: results from the ReACCh-Out cohort.

OBJECTIVE: Structural equation modelling was applied to data from the Research in Arthritis in Canadian Children emphasizing Outcomes (ReACCh-Out) cohort to help elucidate causal pathways to decreased health-related quality of life (HRQoL) in children with JIA. METHODS: Based on published literature and clinical plausibility, a priori models were constructed with explicit root causes (disease activity, treatment intensity) and mediators (pain, disease symptoms, functional impairments) leading to HRQoL [measured by the Quality of my Life (QoML) scale and the Juvenile Arthritis Quality of Life Questionnaire (JAQQ)] at five disease stages: (i) diagnosis, (ii) 3-9 months after diagnosis, (iii) flare, (iv) remission on medications, (v) remission off medications. Following structural equation modelling, a posteriori models were selected based on data fit and clinical plausibility. RESULTS: We included 561, 887, 137, 186 and 182 patients at each stage, respectively. In a posteriori models for active disease stages, paths from disease activity led through pain, functional impairments, and disease symptoms, directly or through restrictions in participation, to decreased QoML scores. Treatment intensity had detrimental effects through psychosocial domains; while treatment side effects had a lesser role. Pathways were similar for QoML and JAQQ, but JAQQ models provided greater specificity. Models for remission stages were not supported by the data. CONCLUSION: Our findings support disease activity and treatment intensity as being root causes of decreased HRQoL in children with JIA, with pain, functional impairments, and participation restrictions being mediators for disease activity; they support psychosocial effects and side effects as being mediators for treatment intensity.

Modeling the Impact of Social Determinants of Health on HIV.

There is growing evidence for the key role of social determinants of health (SDOH) in understanding morbidity and mortality outcomes globally. Factors such as stigma, racism, poverty or access to health and social services represent complex constructs that affect population health via intricate relationships to individual characteristics, behaviors and disease prevention and treatment outcomes. Modeling the role of SDOH is both critically important and inherently complex. Here we describe different modeling approaches and their use in assessing the impact of SDOH on HIV/AIDS. The discussion is thematically divided into mechanistic models and statistical models, while recognizing the overlap between them. To illustrate mechanistic approaches, we use examples of compartmental models and agent-based models; to illustrate statistical approaches, we use regression and statistical causal models. We describe model structure, data sources required, and the scope of possible inferences, highlighting similarities and differences in formulation, implementation, and interpretation of different modeling approaches. We also indicate further needed research on representing and quantifying the effect of SDOH in the context of models for HIV and other health outcomes in recognition of the critical role of SDOH in achieving the goal of ending the HIV epidemic and improving overall population health.

Are complex causal models less likely to be true than simple ones? A critical comment on Trafimow (2017).

Trafimow (2017) used probabilistic reasoning to argue that more complex causal models are less likely to be true than simpler ones, and that researchers should be skeptical of causal models involving more than a handful of variables (or even a single correlation coefficient) [Trafimow, D. (2017). The probability of simple versus complex causal models in causal analyses. Behavior Research Methods, 49, 739-746]. In this comment, I point out that Trafimow's argument is misleading, and reduces to the observation that more informative models (that make definite statements about certain causal relations) are less likely to be true than less informative models (that remain silent about those relations, by omitting some variables from consideration). This correct but trivial statement does not deliver the epistemological leverage promised in the paper. When complexity is evaluated with reasonable criteria (such as the number of nonzero effects in alternative models involving the same variables), more complex models can be more, less, or equally likely to be true compared with simpler ones. I also discuss Trafimow's claim that, if a model is unlikely to be true a priori, researchers will seldom be able to gather evidence of sufficient quality to support it; in practice, even low-probability models can receive strong support without the need for extraordinary evidence. Researchers should evaluate the plausibility of causal models on a case-by-case basis, and be skeptical of overblown claims about the dangers of complex theories.

Implications of Local Friendliness Violation for Quantum Causality.

We provide a new formulation of the Local Friendliness no-go theorem of Bong et al. [Nat. Phys. 16, 1199 (2020)] from fundamental causal principles, providing another perspective on how it puts strictly stronger bounds on quantum reality than Bell's theorem. In particular, quantum causal models have been proposed as a way to maintain a peaceful coexistence between quantum mechanics and relativistic causality while respecting Leibniz's methodological principle. This works for Bell's theorem but does not work for the Local Friendliness no-go theorem, which considers an extended Wigner's Friend scenario. More radical conceptual renewal is required; we suggest that cleaving to Leibniz's principle requires extending relativity to events themselves.

The causal structure of natural kinds.

One primary goal for metaphysical theories of natural kinds is to account for their epistemic fruitfulness. According to cluster theories of natural kinds, this epistemic fruitfulness is grounded in the regular and stable co-occurrence of a broad set of properties. In this paper, I defend the view that such a cluster theory is insufficient to adequately account for the epistemic fruitfulness of kinds. I argue that cluster theories can indeed account for the projectibility of natural kinds, but not for several other epistemic operations that natural kinds support. Natural kinds also play a role in scientific explanations and categorizations. A theory of natural kinds can only account for these additional kind-based epistemic practices if it also analyzes their causal structure.

Relating causal and probabilistic approaches to contextuality.

A primary goal in recent research on contextuality has been to extend this concept to cases of inconsistent connectedness, where observables have different distributions in different contexts. This article proposes a solution within the framework of probabi- listic causal models, which extend hidden-variables theories, and then demonstrates an equivalence to the contextuality-by-default (CbD) framework. CbD distinguishes contextuality from direct influences of context on observables, defining the latter purely in terms of probability distributions. Here, we take a causal view of direct influences, defining direct influence within any causal model as the probability of all latent states of the system in which a change of context changes the outcome of a measurement. Model-based contextuality (M-contextuality) is then defined as the necessity of stronger direct influences to model a full system than when considered individually. For consistently connected systems, M-contextuality agrees with standard contextuality. For general systems, it is proved that M-contextuality is equivalent to the property that any model of a system must contain 'hidden influences', meaning direct influences that go in opposite directions for different latent states, or equivalently signalling between observers that carries no information. This criterion can be taken as formalizing the 'no-conspiracy' principle that has been proposed in connection with CbD. M-contextuality is then proved to be equivalent to CbD-contextuality, thus providing a new interpretation of CbD-contextuality as the non-existence of a model for a system without hidden direct influences. This article is part of the theme issue 'Contextuality and probability in quantum mechanics and beyond'.

Non-Gaussian Methods for Causal Structure Learning.

Causal structure learning is one of the most exciting new topics in the fields of machine learning and statistics. In many empirical sciences including prevention science, the causal mechanisms underlying various phenomena need to be studied. Nevertheless, in many cases, classical methods for causal structure learning are not capable of estimating the causal structure of variables. This is because it explicitly or implicitly assumes Gaussianity of data and typically utilizes only the covariance structure. In many applications, however, non-Gaussian data are often obtained, which means that more information may be contained in the data distribution than the covariance matrix is capable of containing. Thus, many new methods have recently been proposed for using the non-Gaussian structure of data and inferring the causal structure of variables. This paper introduces prevention scientists to such causal structure learning methods, particularly those based on the linear, non-Gaussian, acyclic model known as LiNGAM. These non-Gaussian data analysis tools can fully estimate the underlying causal structures of variables under assumptions even in the presence of unobserved common causes. This feature is in contrast to other approaches. A simulated example is also provided.

Causally Modeling Adaptation to the Environment.

Brandon claims that to explain adaptation one must specify fitnesses in each selective environment and specify the distribution of individuals across selective environments. Glymour claims, using an example of the adaptive evolution of costly plasticity in a symmetric environment, that there are some predictive or explanatory tasks for which Brandon's claim is limited. In this paper, I provide necessary conditions for carrying out Brandon's task, produce a new version of the argument for his claim, and show that Glymour's reasons for making his claim are problematic. I provide a few interpretations of Glymour's argument but ultimately raise worries for what I take to be the key premises.

Neurodevelopmental outcomes in children in relation to placental abruption.

OBJECTIVE: Placental abruption has a profound impact on perinatal mortality, but implications for neurodevelopment during childhood remain unknown. We examined the association between abruption and neurodevelopment at 8 months and 4 and 7 years and evaluated the extent to which these associations were mediated through preterm delivery. DESIGN: Secondary analysis of a multicenter prospective cohort study. SETTING: Multicenter US National Collaborative Perinatal Project (1959-76). POPULATION: Women that delivered singleton live births. METHODS: Analyses of IQ scores were based on marginal structural models (MSM) to account for losses to follow-up. We also carried out a causal mediation analysis to evaluate if the association between abruption and cognitive deficits was mediated through preterm delivery, and performed a sensitivity analysis for unobserved confounding. MAIN OUTCOME MEASURES: We evaluated cognitive development based on the Bayley scale at 8 months (Mental and Motor Scores), and intelligent quotient (IQ) based on the Stanford-Binet scale at 4 years and the Wechsler Intelligence Scale for Children at 7 years. RESULTS: The confounder and selection-bias adjusted risk ratio (RR) of abnormal 8-month Motor and Mental assessments were 2.35 (95%CI 1.39, 3.98) and 2.03 (95%CI 1.13, 3.64), respectively, in relation to abruption. The associations at 4 years were attenuated and resolved at 7 years. The proportion of children with abruption-associated neurological deficits mediated through preterm delivery ranged from 27 to 75%. Following adjustment for unobserved confounding the proportion mediated through preterm delivery was attenuated. CONCLUSION: The effect of abruption on neurodevelopmental outcomes appears restricted to an effect that is largely mediated through preterm delivery. TWEETABLE ABSTRACT: Increased risk of cognitive deficits in relation to abruption appears to be mediated through preterm delivery.

Discovering phenotypic causal structure from nonexperimental data.

The evolutionary potential of organisms depends on how their parts are structured into a cohesive whole. A major obstacle for empirical studies of phenotypic organization is that observed associations among characters usually confound different causal pathways such as pleiotropic modules, interphenotypic causal relationships and environmental effects. The present article proposes causal search algorithms as a new tool to distinguish these different modes of phenotypic integration. Without assuming an a priori structure, the algorithms seek a class of causal hypotheses consistent with independence relationships holding in observational data. The technique can be applied to discover causal relationships among a set of measured traits and to distinguish genuine selection from spurious correlations. The former application is illustrated with a biological data set of rat morphological measurements previously analysed by Cheverud et al. (Evolution 1983, 37, 895).

Religious Service Attendance and Lower Depression Among Women-a Prospective Cohort Study.

OBJECTIVE: Previous studies on the association between religious service attendance and depression have been mostly cross-sectional, subject to reverse causation, and did not account for the potential feedback between religious service attendance and depression. We prospectively evaluated evidence whether religious service attendance decreased risk of subsequent risk of depression and whether depression increased subsequent cessation of service attendance, while explicitly accounting for feedback with potential effects in both directions. METHOD: We included a total of 48,984 US nurses who were participants of the Nurses' Health Study with mean age 58 years and who were followed up from 1996 to 2008. Religious service attendance was self-reported in 1992, 1996, 2000, and 2004. Depression was defined as self-reported physician-diagnosed clinical depression, regular anti-depressant use, or severe depressive symptoms. Multivariate logistic regression and marginal structural models were used to estimate the odds ratio of developing incident depression, adjusted for baseline religious service attendance, baseline depression, and time-varying covariates. RESULTS: Compared with women who never attended services, women who had most frequent and recent religious service attendance had the lowest risk of developing depression (odds ratio [OR] = 0.71, 95 % confidence interval [CI] 0.62-0.82). Compared with women who were not depressed, women with depression were less likely to subsequently attend religious services once or more per week (OR = 0.74, 95 % CI 0.68-0.80). CONCLUSIONS: In this study of US women, there is evidence that higher frequency of religious service attendance decreased the risk of incident depression and women with depression were less likely to subsequently attend services.

Using meta-analytic path analysis to test theoretical predictions in health behavior: An illustration based on meta-analyses of the theory of planned behavior.

OBJECTIVE: Synthesizing research on social cognitive theories applied to health behavior is an important step in the development of an evidence base of psychological factors as targets for effective behavioral interventions. However, few meta-analyses of research on social cognitive theories in health contexts have conducted simultaneous tests of theoretically-stipulated pattern effects using path analysis. We argue that conducting path analyses of meta-analytic effects among constructs from social cognitive theories is important to test nomological validity, account for mediation effects, and evaluate unique effects of theory constructs independent of past behavior. We illustrate our points by conducting new analyses of two meta-analyses of a popular theory applied to health behaviors, the theory of planned behavior. METHOD: We conducted meta-analytic path analyses of the theory in two behavioral contexts (alcohol and dietary behaviors) using data from the primary studies included in the original meta-analyses augmented to include intercorrelations among constructs and relations with past behavior missing from the original analysis. RESULTS: Findings supported the nomological validity of the theory and its hypotheses for both behaviors, confirmed important model processes through mediation analysis, demonstrated the attenuating effect of past behavior on theory relations, and provided estimates of the unique effects of theory constructs independent of past behavior. CONCLUSIONS: Our analysis illustrates the importance of conducting a simultaneous test of theory-stipulated effects in meta-analyses of social cognitive theories applied to health behavior. We recommend researchers adopt this analytic procedure when synthesizing evidence across primary tests of social cognitive theories in health.

Cause and context: place-based approaches to investigate how environments affect mental health.

OBJECTIVES: Our surroundings affect our mood, our recovery from stress, our behavior, and, ultimately, our mental health. Understanding how our surroundings influence mental health is central to creating healthy cities. However, the traditional observational methods now dominant in the psychiatric epidemiology literature are not sufficient to advance such an understanding. In this essay we consider potential alternative strategies, such as randomizing people to places, randomizing places to change, or harnessing natural experiments that mimic randomized experiments. METHODS: We discuss the strengths and weaknesses of these methodological approaches with respect to (1) defining the most relevant scale and characteristics of context, (2) disentangling the effects of context from the effects of individuals' preferences and prior health, and (3) generalizing causal effects beyond the study setting. RESULTS: Promising alternative strategies include creating many small-scale randomized place-based trials, using the deployment of place-based changes over time as natural experiments, and using fluctuations in the changes in our surroundings in combination with emerging data collection technologies to better understand how surroundings influence mood, behavior, and mental health. CONCLUSIONS: Improving existing research strategies will require interdisciplinary partnerships between those specialized in mental health, those advancing new methods for place effects on health, and those who seek to optimize the design of local environments.

Dynamic gamma frequency feedback coupling between higher and lower order visual cortices underlies perceptual completion in humans.

To perceive a coherent environment, incomplete or overlapping visual forms must be integrated into meaningful coherent percepts, a process referred to as "Gestalt" formation or perceptual completion. Increasing evidence suggests that this process engages oscillatory neuronal activity in a distributed neuronal assembly. A separate line of evidence suggests that Gestalt formation requires top-down feedback from higher order brain regions to early visual cortex. Here we combine magnetoencephalography (MEG) and effective connectivity analysis in the frequency domain to specifically address the effective coupling between sources of oscillatory brain activity during Gestalt formation. We demonstrate that perceptual completion of two-tone "Mooney" faces induces increased gamma frequency band power (55-71Hz) in human early visual, fusiform and parietal cortices. Within this distributed neuronal assembly fusiform and parietal gamma oscillators are coupled by forward and backward connectivity during Mooney face perception, indicating reciprocal influences of gamma activity between these higher order visual brain regions. Critically, gamma band oscillations in early visual cortex are modulated by top-down feedback connectivity from both fusiform and parietal cortices. Thus, we provide a mechanistic account of Gestalt perception in which gamma oscillations in feature sensitive and spatial attention-relevant brain regions reciprocally drive one another and convey global stimulus aspects to local processing units at low levels of the sensory hierarchy by top-down feedback. Our data therefore support the notion of inverse hierarchical processing within the visual system underlying awareness of coherent percepts.