Causal Inference In Epidemiology Research Articles

Scientific guidelines for evaluating the validity of forensic feature-comparison methods

When it comes to questions of fact in a legal context—particularly questions about measurement, association, and causality—courts should employ ordinary standards of applied science. Applied sciences generally develop along a path that proceeds from a basic scientific discovery about some natural process to the formation of a theory of how the process works and what causes it to fail, to the development of an invention intended to assess, repair, or improve the process, to the specification of predictions of the instrument’s actions and, finally, empirical validation to determine that the instrument achieves the intended effect. These elements are salient and deeply embedded in the cultures of the applied sciences of medicine and engineering, both of which primarily grew from basic sciences. However, the inventions that underlie most forensic science disciplines have few roots in basic science, and they do not have sound theories to justify their predicted actions or results of empirical tests to prove that they work as advertised. Inspired by the “Bradford Hill Guidelines”—the dominant framework for causal inference in epidemiology—we set forth four guidelines that can be used to establish the validity of forensic comparison methods generally. This framework is not intended as a checklist establishing a threshold of minimum validity, as no magic formula determines when particular disciplines or hypotheses have passed a necessary threshold. We illustrate how these guidelines can be applied by considering the discipline of firearm and tool mark examination.

Specificity of association in epidemiology

The epidemiologist Bradford Hill famously argued that in epidemiology, specificity of association (roughly, the fact that an environmental or behavioral risk factor is associated with just one or at most a few medical outcomes) is strong evidence of causation. Prominent epidemiologists have dismissed Hill’s claim on the ground that it relies on a dubious `one-cause one effect’ model of disease causation. The paper examines this methodological controversy, and argues that specificity considerations do have a useful role to play in causal inference in epidemiology. More precisely, I argue that specificity considerations help solve a pervasive inferential problem in contemporary epidemiology: the problem of determining whether an exposure-outcome correlation might be due to confounding by a social factor. This examination of specificity has interesting consequences for our understanding of the methodology of epidemiology. It highlights how the methodology of epidemiology relies on local tools designed to address specific inference problems peculiar to the discipline, and shows that observational causal inference in epidemiology can proceed with little prior knowledge of the causal structure of the phenomenon investigated. I also argue that specificity of association cannot (despite claims to the contrary) be entirely explained in terms of Woodward’s well-known concept of “one-to-one” causal specificity. This is because specificity as understood by epidemiologists depends on whether an exposure (or outcome) is associated with a `heterogeneous’ set of variables. This dimension of heterogeneity is not captured in Woodward’s notion, but is crucial for understanding the evidential import of specificity of association.

Using Mendelian randomization as the cornerstone for causal inference in epidemiology.

Mendelian randomization (MR) is attracting considerable critical attention. This paper aimed to explore the characteristics of the publications of MR, to reach an insight in this field and prospect the future trend. A bibliometric analysis was performed to identify published MR-related research. The articles were selected from the Web of Science Core Collection database. Excel 2019, VOSviewer 1.6.9, and CiteSpace 5.7.R3 were used to analyze the information. A total of 1783 papers of MR were identified, and the first included literature appeared in 2003. A total of 2829 institutions from 72 countries participated in the relevant research, while the UK contributed to 852 articles and were in a leading position. The most productive institution was the University of Bristol, and Smith GD who has posted the most articles (n=202) was also from there. The Int J Epidemiol (100 publications, 6861 citations) was the most prolific and high citation journal. Related topics of frontiers will still focus on coronary heart disease, diabetes, cancer, psychiatric disorder, body mass index, and lifestyle factors. We summarized the publication information of MR-related literature from 2003 to 2020, including country and institution of origin, authors, and publication journal. We analyzed former research hotspots in the field of MR and predicted future areas of interest. Exposures and outcomes detected in this paper will be the hotspots and frontiers of research in the next few years.

Invited Perspective: Oil and Gas Development and Adverse Birth Outcomes: What More Do We Need to Know?

Invited Perspective: Oil and Gas Development and Adverse Birth Outcomes: What More Do We Need to Know?

Qualitative research in epidemiology: theoretical and methodological perspectives

Increasingly, modern epidemiology has adopted complex causal frameworks incorporating individual- and population-level determinants of health. Despite the growing use of qualitative methodologies in public health research generally, discussion of causal reasoning in epidemiology rarely considers evidence derived from qualitative research. This article argues for a coherent role of qualitative research within epidemiology through analysis of the principles of causal reasoning that underlie current debates about causal inference in epidemiology. It introduces two approaches to causal inference by Russo and Williamson (2009) and Reiss (2012) that emphasize the relevance of both the nature of causation and how knowledge is gained about causation in assessing evidence for a causal relation. Both theories have scope for incorporating multiple types of evidence to assess causal claims. We argue that these theories align with the empirical focus of epidemiology and allow for different types of evidence to evaluate causal claims, including evidence originating from qualitative research; such evidence can contribute to a mechanistic understanding of causal relations and to understanding the effects of context on health-related outcomes. Finally, we discuss this approach in light of previous literature on the role of qualitative research in epidemiology and implications for future epidemiologic research.

機械学習・深層学習・計量経済学に関する先行研究の整理 (Previous Studies on Machine Learning, Deep Learning, and Econometrics)

Japanese Abstract: 本研究は、機械学習・深層学習・強化学習の分野と計量経済学の分野の近接性を探す点に特徴をもつ。計量経済学において時系列分析やパネル分析を学んでいれば、機械学習は分析手法とともに計量経済学の考え方に多くの共通点をもつことから今後は関係性をより体系的に考察することが重要となると考える。また計量経済学も因果推論の分野では機械学習が活用されており、その関係性を説明する。さらに機械学習と計量経済学の近年の研究の進展を概観するため、画像認識や音声認識における機械学習のビジネスへの応用やCATEやLASSO、 政策配分問題としてのELMや疫学における因果推論としてCMA（因果媒介分析）・Causal Forest・SDD・構造推定などを幅広く扱う。 English Abstract: This research is characterized in that it seeks closeness between the fields of machine learning, deep learning and reinforcement learning and the field of econometrics. If you are studying time series analysis or panel analysis in econometrics, machine learning has much in common with econometrics as well as analytical methods, so it is important to consider relationships more systematically in the future. I think it will be. In econometrics, machine learning is used in the field of causal inference, and the relationship is explained. In order to give an overview of recent advances in research on machine learning and econometrics, we have applied machine learning to business in image recognition and speech recognition, CATE and LASSO, ELM as a policy distribution problem, and CMA (causal inference in epidemiology). Mediation analysis), Causal Forest, SDD, structure estimation, etc.

FIFA World Cup and climate change: correlation is not causation. [Copa del Mundo FIFA y cambio climático: correlación no implica causalidad

Sports analytics is defined as the process that identifies and acquires knowledge and insight about players and teams’ performances. To do so, analysts use a wide variety of data sources coming from matches and individual players’ performances (O'Donoghue & Holmes 2014; Jayal, McRobert, Oatley & O’Donoghue, 2018). Nowadays, detailed data from different nature including technical skills, individual physiological performances, team formations, or injuries are analysed on a daily basis by the analytics departments belonging to sports clubs and professional franchises. Even private companies like STATS or OPTA generate important revenues offering their movement tracking values and advanced metrics to media and fans. In the emerging field of Sports Analytics, as in many others, analysts must be aware of spurious correlations. These can come up due to the size (not nature) of data, a common-causal variable or just due to serendipity. For this reason, we always must keep in mind the lessons of the statistician Stephen John Senn and his sharp quote on the matter: “Statistics is not just about merely warning that correlation is not causation. Sometimes correlation isn’t even correlation”. Thus, we will explain an example of how climate change can be affecting, or not, on the FIFA World Cup performance statistics. https://doi.org/10.5232/ricyde2019.057ed References/referencias Allmers, S., & Maennig, W. (2009). Economic impacts of the FIFA Soccer World Cups in France 1998, Germany 2006, and outlook for South Africa 2010. Eastern Economic Journal, 35 (4), 500-519. https://doi.org/10.1057/eej.2009.30 Castellano, J.; Casamichana, D., & Lago, C. (2012). The use of match statistics that discriminate between successful and unsuccessful soccer teams. Journal of Human Kinetics, 31 (1), 137-147. https://doi.org/10.2478/v10078-012-0015-7 Jayal, A., McRobert, A., Oatley, G., & O’Donoghue, P. (2018). Sports Analytics: Analysis, Visualisation and Decision Making in Sports Performance . Routledge. https://doi.org/10.4324/9781315222783 Kakamu, T.; Wada, K.; Smith, D. R.; Endo, S., & Fukushima T. (2017). Preventing heat illness in the anticipated hot climate of the Tokyo 2020 Summer Olympic Games. Environmental Health and Preventive Medicine 22 (68), 1-6. https://doi.org/10.1186/s12199-017-0675-y Matzarakis, A., & Frohlich, D. (2015). Sport events and climate for visitors-the case of FIFA World Cup in Qatar 2022. International Journal of Biometeorology, 59 (4), 481-586. https://doi.org/10.1007/s00484-014-0886-5 Mohr, M.; Nybo, L.; Grantham, J., & Racinais, S. (2012). Physiological Responses and Physical Performance during Football in the Heat. PLoS ONE, 7 (6), e39202. https://doi.org/10.1371/journal.pone.0039202 NASA (2018). Global Climate Change, Vital Signs of the Planet . [Internet] http://climate.nasa.gov/ (last accessed, Jul. 18) Nassis, G. P.; Brito, J.; Dvorak, J.; Chalabi, H.; & Racinais, S. (2015). The association of environmental heat stress with performance: analysis of the 2014 FIFA World Cup Brazil. British Journal of Sports Medicine, 49 (9), 609-613. https://doi.org/10.1136/bjsports-2014-094449 Nicolau, J. L., & Sharma, A. (2018). A generalization of the FIFA World Cup effect. Tourism Management, 66 (June), 315-317. https://doi.org/10.1016/j.tourman.2017.12.014 Liu, H.; Gomez, M. A.; Lago-Penas, C., & Sampaio, J. (2015). Match statistics related to winning in the group stage of 2014 Brazil FIFA World Cup. Journal of Sports Sciences, 33 (12), 1205-1213. https://doi.org/10.1080/02640414.2015.1022578 O’Donoghue, P., & Holmes, L. (2014). Data analysis in sport . Routledge. https://doi.org/10.4324/9781315816357 Pachuari R. K.; Allen, M. R.; Barros, V. R.; Broome, J.; Cramer, W.; Christ, R., et al. (2014). Climate change 2014: synthesis report. Contribution of working groups I, II and III to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. Intergovernmental Panel on Climate Change, Geneva. Retrieved from: https://epic.awi.de/id/eprint/37530/1/IPCC_AR5_SYR_Final.pdf Ramdas, B.; van Gaalen, R., & Bolton, J. (2015). The Announcement Impact of Hosting the FIFA World Cup on Host Country Stock Markets. Procedia Economics Finance, 30 , 226-238. https://doi.org/10.1016/S2212-5671(15)01290-3 Robinson, W. S. (1950). Ecological correlations and the behaviour of individuals. American Sociological Review, 15 (3), 351-357. https://doi.org/10.2307/2087176 Rumpf, M. C.; Silva, J. R.; Hertzog, M.; Farooq, A., & Nassis, G. (2017). Technical and physical analysis of the 2014 FIFA World Cup Brazil: winners vs. losers. The Journal of sports medicine and physical fitness, 57 (10), 1338-1343. Vandenbroucke, J. P.; Broadbent, A., & Pearce, N. (2016). Causality and causal inference in epidemiology: the need for a pluralistic approach. International Journal of Epidemiology, 45 (6), 1776-1786. https://doi.org/10.2307/2087176 Vigen, T. (2015). Spurious Correlations . Hachette Books. Wikipedia (2018). FIFA World Cup-related lists . [Internet] (last accessed, Jul. 18). Retrieved from: https://en.wikipedia.org/wiki/Category:FIFA_World_Cup-related_lists

The C-word, the P-word, and realism in epidemiology

This paper considers an important recent (May 2018) contribution by Miguel Hernán to the ongoing debate about causal inference in epidemiology. Hernán rejects the idea that there is an in-principle epistemic distinction between the results of randomized controlled trials and observational studies: both produce associations which we may be more or less confident interpreting as causal. However, Hernán maintains that trials have a semantic advantage. Observational studies that seek to estimate causal effect risk issuing meaningless statements instead. The POA proposes a solution to this problem: improved restrictions on the meaningful use of causal language, in particular “causal effect”. This paper argues that new restrictions in fact fail their own standards of meaningfulness. The paper portrays the desire for a restrictive definition of causal language as positivistic, and argues that contemporary epidemiology should be more realistic in its approach to causation. In a realist context, restrictions on meaningfulness based on precision of definition are neither helpful nor necessary. Hernán’s favoured approach to causal language is saved from meaninglessness, along with the approaches he rejects.

DAG-informed regression modelling, agent-based modelling and microsimulation modelling: a critical comparison of methods for causal inference.

The current paradigm for causal inference in epidemiology relies primarily on the evaluation of counterfactual contrasts via statistical regression models informed by graphical causal models (often in the form of directed acyclic graphs, or DAGs) and their underlying mathematical theory. However, there have been growing calls for supplementary methods, and one such method that has been proposed is agent-based modelling due to its potential for simulating counterfactuals. However, within the epidemiological literature, there currently exists a general lack of clarity regarding what exactly agent-based modelling is (and is not) and, importantly, how it differs from microsimulation modelling—perhaps its closest methodological comparator. We clarify this distinction by briefly reviewing the history of each method, which provides a context for their similarities and differences, and casts light on the types of research questions that they have evolved (and thus are well suited) to answering; we do the same for DAG-informed regression methods. The distinct historical evolutions of DAG-informed regression modelling, microsimulation modelling and agent-based modelling have given rise to distinct features of the methods themselves, and provide a foundation for critical comparison. Not only are the three methods well suited to addressing different types of causal questions, but, in doing so, they place differing levels of emphasis on fixed and random effects, and also tend to operate on different timescales and in different timeframes.

Web of Causation between Dietary Patterns and Childhood Obesity: Applying Hill's Criteria

Abstract Since their publication in 1965, the Bradford Hill criteria for causality have been largely used as a framework for causal inference in epidemiology. We aim at employing this classical approach to shed new light onto the web of causation of childhood obesity. Although the fundamental cause of obesity is the long-term imbalance between energetic need and intake, this medical condition is multifactorial in its origin, influenced by genetic, behavioral, socioeconomic, and environmental factors. This imbalance leads to accumulation of excessive adipose tissue. Observational studies tend to mostly quantify association between dietary factors and accumulation of adipose tissue. On the other hand, multivariate analysis proved some of these associations to be spurious, therefore prospective trials are needed to demonstrate causality. Short term experimental studies have been conducted to identify unique dietary pattern changes on specific outcomes, but long term, community-based studies would offer more comprehensive answers on dietary pattern effects. We conducted a literature review on PubMed, Scopus, Web of Science, and Google Scholar. From a total of 323 papers identified at first stage, we further discuss the applicability of Bradford Hill criteria for 31 articles, by examples of dietary patterns and accumulation of excess body fat as exposure-response associations. We also put forward and analyzed the evidence prospective studies would bring, as foundation for future interventions.

Directed Acyclic Graphs: Alternative tool for causal inference in epidemiology and biostatistics research and teaching

The issue of causation is one of the major challenges for epidemiologists who aim to understand the association between an exposure and an outcome to explain disease patterns and potentially provide a basis for intervention. Suitably designed experimental studies can offer robust evidence of the causal relationships. The experimental studies, however, are not popular, difficult or even unethical and impossible to conduct; it would be desirable if there is a methodology for reducing bias or strengthening the causal inferences drawn from observational studies. The traditional approach of estimating causal effects in such studies is to adjust for a set of variables judged to be confounders by including them in a multiple regression. However, which variables should be adjusted for as confounders in a regression model has long been a controversial issue in epidemiology. From my observation, the adjustments using only “statistical artifacts” methods such as the p-value<0.2 in univariate analysis, stepwise (forward/backward) are widely used in research and teaching in Epidemiology and Statistics but without appropriated notice on the biological or clinical relationships between exposure and outcome which may induce the bias in estimating causal effects. In this mini-review, we introduce an interesting method, namely Directed Acyclic Graphs (DAGs), which can be used to reduce the bias in estimating causal effects; it is also a good application for Epidemiology and Biostatistics teaching.

Invited Commentary: Detecting Individual and Global Horizontal Pleiotropy in Mendelian Randomization-A Job for the Humble Heterogeneity Statistic?

Mendelian randomization (MR) is gaining in recognition and popularity as a method for strengthening causal inference in epidemiology by utilizing genetic variants as instrumental variables. Concurrently with the explosion in empirical MR studies, there has been the steady production of new approaches for MR analysis. The recently proposed “global and individual tests for direct effects” (GLIDE) approach fits into a family of methods that aim to detect horizontal pleiotropy—at the individual single nucleotide polymorphism level and at the global level—and to adjust the analysis by removing outlying single nucleotide polymorphisms. In this commentary, we explain how existing methods can (and indeed are) being used to detect pleiotropy at the individual and global levels, although not explicitly using this terminology. By doing so, we show that the true comparator for GLIDE is not MR-Egger regression (as Dai et al., the authors of the accompanying article (Am J Epidemiol. 2018;187(12):2672–2680), claim) but rather the humble heterogeneity statistic.

Causal inference in epidemiology: a fruitful debate

Causal inference in epidemiology: a fruitful debate

Response: Formalism or pluralism? A reply to commentaries on 'Causality and causal inference in epidemiology'.

Response: Formalism or pluralism? A reply to commentaries on 'Causality and causal inference in epidemiology'.

Commentary: Causal inference in epidemiology: potential outcomes, pluralism and peer review.

Commentary: Causal inference in epidemiology: potential outcomes, pluralism and peer review.

Robust causal inference using directed acyclic graphs: the R package 'dagitty'.

Directed acyclic graphs (DAGs), which offer systematic representations of causal relationships, have become an established framework for the analysis of causal inference in epidemiology, often being used to determine covariate adjustment sets for minimizing confounding bias. DAGitty is a popular web application for drawing and analysing DAGs. Here we introduce the R package 'dagitty', which provides access to all of the capabilities of the DAGitty web application within the R platform for statistical computing, and also offers several new functions. We describe how the R package 'dagitty' can be used to: evaluate whether a DAG is consistent with the dataset it is intended to represent; enumerate 'statistically equivalent' but causally different DAGs; and identify exposure-outcome adjustment sets that are valid for causally different but statistically equivalent DAGs. This functionality enables epidemiologists to detect causal misspecifications in DAGs and make robust inferences that remain valid for a range of different DAGs. The R package 'dagitty' is available through the comprehensive R archive network (CRAN) at [https://cran.r-project.org/web/packages/dagitty/]. The source code is available on github at [https://github.com/jtextor/dagitty]. The web application 'DAGitty' is free software, licensed under the GNU general public licence (GPL) version 2 and is available at [http://dagitty.net/].

Commentary: The formal approach to quantitative causal inference in epidemiology: misguided or misrepresented?

Commentary: The formal approach to quantitative causal inference in epidemiology: misguided or misrepresented?

The role of counterfactual theory in causal reasoning

In this commentary I review the fundamentals of counterfactual theory and its role in causal reasoning in epidemiology. I consider if counterfactual theory dictates that causal questions must be framed in terms of well-defined interventions. I conclude that it does not. I hypothesize that the interventionist approach to causal inference in epidemiology stems from elevating the randomized trial design to the gold standard for thinking about causal inference. I suggest that instead the gold standard we should use for thinking about causal inference in epidemiology is the thought experiment that, for example, compares an actual disease frequency under one exposure level with a counterfactual disease frequency under a different exposure level (as discussed in Greenland and Robins (1986) and Maldonado and Greenland (2002)). I also remind us that no method should be termed “causal” unless it addresses the effect of other biases in addition to the problem of confounding.

Re: Causality and causal inference in epidemiology: the need for a pluralistic approach

Re: Causality and causal inference in epidemiology: the need for a pluralistic approach

Alcohol consumption as a cause of cancer.

There is increasing research evidence about the causal role of alcohol in cancer, accompanied by unclear and conflicting messages in the media. This paper aimed to clarify the strength of the evidence for alcohol as a cause of cancer, and the meaning of cause in this context. Recent epidemiological and biological research on alcohol and cancer was reviewed and summarized, drawing upon published meta-analyses identified from the Medline database and the archives of the International Agency for Research on Cancer. More recent epidemiological studies not included in these publications were also reviewed. A brief description of the nature of causal inference in epidemiology was used to frame discussion of the strength of the evidence that alcohol causes cancer, and contrast this with the case for a protective association of alcohol with cardiovascular disease. The usual epidemiological understanding of a cause is a factor that increases the incidence of a condition in the population. In the context of a body of epidemiological evidence of an association of alcohol consumption with a disease, the inference that it is a causal association requires alternative explanations of the observed finding to be judged unlikely. Even without complete knowledge of biological mechanisms, the epidemiological evidence can support the judgement that alcohol causes cancer of the oropharynx, larynx, oesophagus, liver, colon, rectum and breast. The measured associations exhibit gradients of effect that are biologically plausible, and there is some evidence of reversibility of risk in laryngeal, pharyngeal and liver cancers when consumption ceases. The limitations of cohort studies mean that the true effects may be somewhat weaker or stronger than estimated currently, but are unlikely to be qualitatively different. The same, or similar, epidemiological studies also commonly report protection from cardiovascular disease associated with drinking but a high level of scepticism regarding these findings is now warranted. There is strong evidence that alcohol causes cancer at seven sites in the body and probably others. Current estimates suggest that alcohol-attributable cancers at these sites make up 5.8% of all cancer deaths world-wide. Confirmation of specific biological mechanisms by which alcohol increases the incidence of each type of cancer is not required to infer that alcohol is a cause.