Causal criteria and counterfactuals; nothing more (or less) than scientific common sense

Directed acyclic graph for epidemiological studies in childhood food allergy: Construction, user's guide, and application

Understanding modifiable prenatal and early life causal determinants of food allergy is important for the prevention of the disease. Randomized clinical trials studying environmental and dietary determinants of food allergy may not always be feasible. Identifying risk/protective factors for early-life food allergy often relies on observational studies, which may be affected by confounding bias. The directed acyclic graph (DAG) is a causal diagram useful to guide causal inference from observational epidemiological research. To date, research on food allergy has made little use of this promising method. We performed a literature review of existing evidence with a systematic search, synthesized 32 known risk/protective factors, and constructed a comprehensive DAG for early-life food allergy development. We present an easy-to-use online tool for researchers to re-construct, amend, and modify the DAG along with a user's guide to minimize confounding bias. We estimated that adjustment strategies in 57% of previous observational studies on modifiable factors of childhood food allergy could be improved if the researchers determined their adjustment sets by DAG. Future researchers who are interested in the causal inference of food allergy development in early life can apply the DAG to identify covariates that should and should not be controlled in observational studies.

Commentary: On the reliability of causal claims

Causal assessments in epidemiology are a complex process due to the many methods involved. The general scientific method lords over the process joined by study designs and statistical methods. Other methods include those that evaluate quality and bias along with the research synthesis methods such as the systematic narrative review, meta-analysis, and the criteria-based methods. When different investigators apply these methods to the same evidence and come up with different causal assessments, as described in the review by Goodman et al. in this issue, a key question becomes, how can the differences be explained? A prime candidate involves different methodologic choices. A deeper question emerges from this same situation: are the methods used for causal assessments reliable? Reliability is a hallmark of scientific practice. The methods used to make claims about causality should be reliable. Given the complexity of the causal assessment process, an objective evaluation of reliability is challenging but clearly worth the effort. Fortunately, Hill's criterion of analogy, much maligned in epidemiology, provides a clue. This commentary explores the issue of the reliability of causal claims using the Goodman et al. systematic review as its foil along with the claims by EPA, IARC, and ATSDR about the relationship between perchloroethylene and non-Hodgkin lymphoma, the claims Goodman et al. believe are wrong.

Assessing causality in epidemiology: revisiting Bradford Hill to incorporate developments in causal thinking

The nine Bradford Hill (BH) viewpoints (sometimes referred to as criteria) are commonly used to assess causality within epidemiology. However, causal thinking has since developed, with three of the most prominent approaches implicitly or explicitly building on the potential outcomes framework: directed acyclic graphs (DAGs), sufficient-component cause models (SCC models, also referred to as 'causal pies') and the grading of recommendations, assessment, development and evaluation (GRADE) methodology. This paper explores how these approaches relate to BH's viewpoints and considers implications for improving causal assessment. We mapped the three approaches above against each BH viewpoint. We found overlap across the approaches and BH viewpoints, underscoring BH viewpoints' enduring importance. Mapping the approaches helped elucidate the theoretical underpinning of each viewpoint and articulate the conditions when the viewpoint would be relevant. Our comparisons identified commonality on four viewpoints: strength of association (including analysis of plausible confounding); temporality; plausibility (encoded by DAGs or SCC models to articulate mediation and interaction, respectively); and experiments (including implications of study design on exchangeability). Consistency may be more usefully operationalised by considering an effect size's transportability to a different population or unexplained inconsistency in effect sizes (statistical heterogeneity). Because specificity rarely occurs, falsification exposures or outcomes (i.e., negative controls) may be more useful. The presence of a dose-response relationship may be less than widely perceived as it can easily arise from confounding. We found limited utility for coherence and analogy. This study highlights a need for greater clarity on BH viewpoints to improve causal assessment.

Assessing causality in epidemiology: revisiting Bradford Hill to incorporate developments in causal thinking

The nine Bradford Hill (BH) viewpoints (sometimes referred to as criteria) are commonly used to assess causality within epidemiology. However, causal thinking has since developed, with three of the most prominent approaches implicitly or explicitly building on the potential outcomes framework: directed acyclic graphs (DAGs), sufficient-component cause models (SCC models, also referred to as 'causal pies') and the grading of recommendations, assessment, development and evaluation (GRADE) methodology. This paper explores how these approaches relate to BH's viewpoints and considers implications for improving causal assessment. We mapped the three approaches above against each BH viewpoint. We found overlap across the approaches and BH viewpoints, underscoring BH viewpoints' enduring importance. Mapping the approaches helped elucidate the theoretical underpinning of each viewpoint and articulate the conditions when the viewpoint would be relevant. Our comparisons identified commonality on four viewpoints: strength of association (including analysis of plausible confounding); temporality; plausibility (encoded by DAGs or SCC models to articulate mediation and interaction, respectively); and experiments (including implications of study design on exchangeability). Consistency may be more usefully operationalised by considering an effect size's transportability to a different population or unexplained inconsistency in effect sizes (statistical heterogeneity). Because specificity rarely occurs, falsification exposures or outcomes (i.e., negative controls) may be more useful. The presence of a dose-response relationship may be less than widely perceived as it can easily arise from confounding. We found limited utility for coherence and analogy. This study highlights a need for greater clarity on BH viewpoints to improve causal assessment.

INFERENCE: An Evidence-Based Approach for Medicolegal Causal Analyses

A fundamental purpose of forensic medical, or medicolegal, analysis is to provide legal factfinders with an opinion regarding the causal relationship between an alleged unlawful or negligent action and a medically observed adverse outcome, which is needed to establish legal liability. At present, there are no universally established standards for medicolegal causal analysis, although several different approaches to causation exist, with varying strengths and weaknesses and degrees of practical utility. These approaches can be categorized as intuitive or probabilistic, which are distributed along a spectrum of increasing case complexity. This paper proposes a systematic approach to evidence-based assessment of causation in forensic medicine, called the INtegration of Forensic Epidemiology and the Rigorous EvaluatioN of Causation Elements (INFERENCE) approach. The INFERENCE approach is an evolution of existing causal analysis methods and consists of a stepwise method of increasing complexity. We aimed to develop a probabilistic causal analysis approach that (1) fits the needs of legal factfinders who require an estimate of the probability of causation, and (2) is still sufficiently straightforward to be applied in real-world forensic medical practice. As the INFERENCE approach is most relevant in complex cases, we also propose a process for selecting the most appropriate causal analysis method for any given case. The goal of this approach is to improve the reproducibility and transparency of causal analyses, which will promote evidence-based practice and quality assurance in forensic medicine, resulting in expert opinions that are reliable and objective in legal proceedings.

Difficulties in establishing a causal link between chemical exposures and cancer cannot be overcome by court assessments

Scientific data are often used in lawsuits to prove, or dismiss, causation by a claimed factor of a claimed disease. Recent media reports of million-dollar compensations awarded to some cancer patients who had been exposed to certain chemical substances motivated me to examine how solid the causal links really were. Here, I discuss the limitations of epidemiological research on cancer causation and highlight how new knowledge of cancer genetics makes it unrealistic to expect that cancer causation can be clearly demonstrated. I then present two exposure-cancer cases, namely talcum powder-ovarian cancer and glyphosate-non-Hodgkin lymphoma, that led to civil lawsuits decided, in the United States, in favor of the claimants. Both these cancers have several risk factors, among which the claimed exposure presents only a minor, if any, increased risk. Through these cases, I explain why the use of epidemiological data is inappropriate to define causal associations in complex diseases like cancer. I close by suggesting a fairer approach, called proportional liability, to resolving future cancer litigation cases.

Causal criteria: time has come for a revision

Epidemiologists study associations but they are usually interested in causation that could lead to disease prevention. Experience show, however, that many of the associations we identify are not the causes we take an interest in (correlation is not causation). In order to proper translate association into causes, a set of causal criteria was developed 50-60 years ago and they became important tools guiding this translational process (sometimes correlation is causation). Best known of these are the Bradford Hill 'criteria'. In these last 50 years, epidemiologic theory and infrastructure have advanced rapidly without changes in these causal criteria. We think time has come to revisit the 'old' criteria to see which ones we should keep and which ones should be taken out or be replaced by new measures of association. Robustness of these criteria in attempts to make the association go away should have high priority. A group of renowned internationally recognized researchers should have this task. Since classifying associations as causes is often done in order to reduce or eliminate the exposures of concern results from conditional outcome research should also be used. We therefore suggest to add a 'consequence' criterion. We argue that a consequence criterion that provides a framework for assessing or prescribing action worthy or right in social contexts is needed. A consequence criterion will also influence how strict our causal criteria need to be before leading to action and will help in separating the 'causal discussion' and the discussion on what to do about it. A consequence criterion will be a tool in handling dilemmas over values (as social solidarity, fairness, autonomy). It will have implications for the interpretation and use of the procedural criteria of causality. Establishing interconnected procedural and consequence criteria should be a task for institutions representing and being recognized by experts, civil society and the state.

Hill's Heuristics and Explanatory Coherentism in Epidemiology

In this essay, I argue that Ted Poston's theory of explanatory coherentism is well-suited as a tool for causal explanation in the health sciences, particularly in epidemiology. Coherence has not only played a role in epidemiology for more than half a century as one of Hill's viewpoints, it can also provide background theory for the development of explanatory systems by integrating epidemiologic evidence with a diversity of other error-independent data. I propose that computational formalization of Hill's viewpoints in an explanatory coherentist framework would provide an excellent starting point for a formal epistemological (knowledge-theoretical) project designed to improve causal explanation in the health sciences. As an example, I briefly introduce Paul Thagard's ECHO system and offer my responses to possible objections to my proposal.

For and Against Methodologies: Some Perspectives on Recent Causal and Statistical Inference Debates

I present an overview of two methods controversies that are central to analysis and inference: That surrounding causal modeling as reflected in the "causal inference" movement, and that surrounding null bias in statistical methods as applied to causal questions. Human factors have expanded what might otherwise have been narrow technical discussions into broad philosophical debates. There seem to be misconceptions about the requirements and capabilities of formal methods, especially in notions that certain assumptions or models (such as potential-outcome models) are necessary or sufficient for valid inference. I argue that, once these misconceptions are removed, most elements of the opposing views can be reconciled. The chief problem of causal inference then becomes one of how to teach sound use of formal methods (such as causal modeling, statistical inference, and sensitivity analysis), and how to apply them without generating the overconfidence and misinterpretations that have ruined so many statistical practices.

Adolescent idiopathic scoliosis and back pain

This broad narrative review addresses the relationship between adolescent idiopathic scoliosis (AIS) and back pain. AIS can be responsible for low back pain, particularly major cases. However, a linear relationship between back pain and the magnitude of the deformity cannot be expected for any individual patient. A large number of juvenile patients can remain pain-free. The long-term prognosis is rather benign for many cases and thus a tailored approach to the individual patient seems mandatory. The level of evidence available does not allow stringent recommendations for any of the disorders included in this review.

Causation or only correlation? Application of causal inference graphs for evaluating causality in nano-QSAR models

In this paper, we suggest that causal inference methods could be efficiently used in Quantitative Structure-Activity Relationships (QSAR) modeling as additional validation criteria within quality evaluation of the model. Verification of the relationships between descriptors and toxicity or other activity in the QSAR model has a vital role in understanding the mechanisms of action. The well-known phrase "correlation does not imply causation" reflects insight statistically correlated with the endpoint descriptor may not cause the emergence of this endpoint. Hence, paradigmatic shifts must be undertaken when moving from traditional statistical correlation analysis to causal analysis of multivariate data. Methods of causal discovery have been applied for broader physical insight into mechanisms of action and interpretation of the developed nano-QSAR models. Previously developed nano-QSAR models for toxicity of 17 nano-sized metal oxides towards E. coli bacteria have been validated by means of the causality criteria. Using the descriptors confirmed by the causal technique, we have developed new models consistent with the straightforward causal-reasoning account. It was proven that causal inference methods are able to provide a more robust mechanistic interpretation of the developed nano-QSAR models.

From monocausality to systems thinking: a complementary and alternative conceptual approach for better understanding the development and prevention of sports injury

The science of sports injury control, including both its cause and prevention, has largely been informed by a biomedical and mechanistic model of health. Traditional scientific practice in sports injury research has routinely involved collapsing the broader socioecological landscape down in order to analyse individual-level determinants of injury - whether biomechanical and/or behavioural. This approach has made key gains for sports injury prevention research and should be further encouraged and allowed to evolve naturally. However, the public health, Applied Human Factors and Ergonomics, and injury epidemiological literature more broadly, has accepted the value of a socioecological paradigm for better understanding disease and injury processes, and sports injury research will fall further behind unless it does the same. A complementary and alternative conceptual approach towards injury control known as systems thinking that builds on socioecological science, both methodologically and analytically, is readily available and fast developing in other research areas. This review outlines the historical progression of causal concepts in the field of epidemiology over the course of the modern scientific era. From here, causal concepts in injury epidemiology, and models of aetiology as found in the context of sports injury research are presented. The paper finishes by proposing a new research agenda that considers the potential for a systems thinking approach to further enhance sports injury aetiological understanding. A complementary systems paradigm, however, will require that sports injury epidemiologists bring their knowledge and skillsets forwards in an attempt to use, adapt, and even refine existing systems-based approaches. Alongside the natural development of conventional scientific methodologies and analyses in sports injury research, progressing forwards to a systems paradigm is now required.

Research gaps in evaluating the relationship of meat and health

Humans evolved as omnivores and it has been proposed that cooking meat allowed for evolution of larger brains that has led to our success as a species. Meat is one of the most nutrient dense foods, providing high-quality protein, heme iron, zinc, and vitamins B6 and B12. Despite these advantages, epidemiologic studies have linked consumption of red or processed meat with obesity, type 2 diabetes, cardiovascular diseases, and cancers of multiple organs. Most observational studies report small, increased relative risks. However, there are many limitations of such studies including inability to accurately estimate intake, lack of prespecified hypotheses, multiple comparisons, and confounding from many factors - including body weight, fruit/vegetable intake, physical activity, smoking, and alcohol - that correlate significantly either positively or negatively with meat intake and limit the reliability of conclusions from these studies. The observational studies are heterogeneous and do not fulfill many of the points proposed by AB Hill in 1965 for inferring causality; his most important factor was strength of the association which in dietary studies is usually <1.5 but is not considered adequate in virtually all other areas of epidemiology outside nutrition. Accepting small, statistically significant risks as "real" from observational associations, the field of nutrition has a long list of failures including beta-carotene and lung cancer, low-fat diets and breast cancer or heart disease that have not been confirmed in randomized trials. Moderate intake of a variety of foods that are enjoyed by people remains the best dietary advice.

Using qualitative comparative analysis (QCA) in systematic reviews of complex interventions: a worked example

BACKGROUND

Systematic reviews that address policy and practice questions in relation to complex interventions frequently need not only to assess the efficacy of a given intervention but to identify which intervention - and which intervention components - might be most effective in particular situations. Here, intervention replication is rare, and commonly used synthesis methods are less useful when the focus of analysis is the identification of those components of an intervention that are critical to its success.

METHODS

Having identified initial theories of change in a previous analysis, we explore the potential of qualitative comparative analysis (QCA) to assist with complex syntheses through a worked example. Developed originally in the area of political science and historical sociology, a QCA aims to identify those configurations of participant, intervention and contextual characteristics that may be associated with a given outcome. Analysing studies in these terms facilitates the identification of necessary and sufficient conditions for the outcome to be obtained. Since QCA is predicated on the assumption that multiple pathways might lead to the same outcome and does not assume a linear additive model in terms of changes to a particular condition (that is, it can cope with 'tipping points' in complex interventions), it appears not to suffer from some of the limitations of the statistical methods often used in meta-analysis.

RESULTS

The worked example shows how the QCA reveals that our initial theories of change were unable to distinguish between 'effective' and 'highly effective' interventions. Through the iterative QCA process, other intervention characteristics are identified that better explain the observed results.

CONCLUSIONS

QCA is a promising alternative (or adjunct), particularly to the standard fall-back of a 'narrative synthesis' when a quantitative synthesis is impossible, and should be considered when reviews are broad and heterogeneity is significant. There are very few examples of its use with systematic review data at present, and further methodological work is needed to establish optimal conditions for its use and to document process, practice, and reporting standards.

The conundrum of causality in tumor virology: the cases of KSHV and MCV

Controversy has plagued tumor virology since the first tumor viruses were described over 100 years ago. Methods to establish cancer causation, such as Koch's postulates, work poorly or not at all for these viruses. Kaposi's sarcoma herpesvirus (KSHV/HHV8) and Merkel cell polyomavirus (MCV) were both found using nucleic acid identification methods but they represent opposite poles in the patterns for tumor virus epidemiology. KSHV is uncommon and has specific risk factors that contribute to infection and subsequent cancers. MCV and Merkel cell carcinoma (MCC), in contrast, is an example in which mutations to our normal viral flora contribute to cancer. Given the near-ubiquity of human MCV infection, establishing cancer causality relies on molecular evidence that does not fit comfortably within traditional infectious disease epidemiological models. These two viruses reveal some of the challenges and opportunities for inferring viral cancer causation in the age of molecular biology.

Hume, Mill, Hill, and the sui generis epidemiologic approach to causal inference

The epidemiologic approach to causal inference (i.e., Hill's viewpoints) consists of evaluating potential causes from the following 2, noncumulative angles: 1) established results from comparative, observational, or experimental epidemiologic studies; and 2) reviews of nonepidemiologic evidence. It does not involve statements of statistical significance. The philosophical roots of Hill's viewpoints are unknown. Superficially, they seem to descend from the ideas of Hume and Mill. Hill's viewpoints, however, use a different kind of evidence and have different purposes than do Hume's rules or Mill's system of logic. In a nutshell, Hume ignores comparative evidence central to Hill's viewpoints. Mill's logic disqualifies as invalid nonexperimental evidence, which forms the bulk of epidemiologic findings reviewed from Hill's viewpoints. The approaches by Hume and Mill cannot corroborate successful implementations of Hill's viewpoints. Besides Hume and Mill, the epidemiologic literature is clueless about a plausible, pre-1965 philosophical origin of Hill's viewpoints. Thus, Hill's viewpoints may be philosophically novel, sui generis, still waiting to be validated and justified.

Revisiting the concepts of risk and protective factors for understanding the etiology and development of substance use and substance use disorders: implications for prevention

Over the past 20 years we have accumulated a greater knowledge and understanding of the genetic, neurobiological, and behavioral factors that may be associated with young people initiating the use of drugs and other substances and to progressing from use to abuse and dependence. This knowledge suggests that individuals may be "predisposed" to substance use disorders (SUD) and that the actual engagement in these behaviors depends on their environmental experiences from micro to macro levels. This paper summarizes this knowledge base and supports a developmental framework that examines the interaction of posited genetic, psychological, and neurobiological "predispositions" to SUD and those environmental influences that exacerbate this vulnerability.

Inference in Forensic Pathology

Forensic pathologists make inferences about cause and manner of death. Those inferences have come under increasing scrutiny by the courts, by social critics of our findings, and by society at large. Much of this criticism is due to our inability to explain our inferential process. Forensic pathologists should be able to cogently explain the reasoning behind their findings, and express it in terms useful to stakeholders. This requires that we have a basic understanding of different kinds of inference and the scientific method, how they are used and their limitations. Medical diagnosis is not a simple matter of application of cookbook-style inferential laws, but involves a combination of deduction, induction, abduction, dialectic, and informal inference. There are significant differences between the way physicians make inferences compared to how they justify them. A discussion of different kinds of inference, inferential fallacies, evaluation of evidence, causation, and the scientific method is provided, with illustrations from the practice of forensic pathology.

Causal models for investigating complex disease: I. A primer

BACKGROUND/AIMS

To illustrate the utility of causal models for research in genetic epidemiology and statistical genetics. Causal models are increasingly applied in risk factor epidemiology, economics, and health policy, but seldom used in statistical genetics or genetic epidemiology. Unlike the statistical models usually used in genetic epidemiology, causal models are explicitly formulated in terms of cause and effect relationships occurring at the individual level.

METHODS

We describe two causal models, the sufficient component cause model and the potential outcomes model, and show how key concepts in genetic epidemiology, including penetrance, phenocopies, genetic heterogeneity, etiologic heterogeneity, gene-gene interaction, and gene-environment interaction, can be framed in terms of these causal models. We also illustrate how potential outcomes models can provide insight into the potential for confounding and bias in the measurement of causal effects in genetic studies.

RESULTS

Our analysis illustrates how causal models can elucidate the relationships among underlying causal mechanisms and measures obtained from statistical analysis of observed data.

CONCLUSION

Causal models can enhance research aimed at identifying causal genes.

The role of causal criteria in causal inferences: Bradford Hill's "aspects of association"

As noted by Wesley Salmon and many others, causal concepts are ubiquitous in every branch of theoretical science, in the practical disciplines and in everyday life. In the theoretical and practical sciences especially, people often base claims about causal relations on applications of statistical methods to data. However, the source and type of data place important constraints on the choice of statistical methods as well as on the warrant attributed to the causal claims based on the use of such methods. For example, much of the data used by people interested in making causal claims come from non-experimental, observational studies in which random allocations to treatment and control groups are not present. Thus, one of the most important problems in the social and health sciences concerns making justified causal inferences using non-experimental, observational data. In this paper, I examine one method of justifying such inferences that is especially widespread in epidemiology and the health sciences generally - the use of causal criteria. I argue that while the use of causal criteria is not appropriate for either deductive or inductive inferences, they do have an important role to play in inferences to the best explanation. As such, causal criteria, exemplified by what Bradford Hill referred to as "aspects of [statistical] associations", have an indispensible part to play in the goal of making justified causal claims.

Dealing with uncertainties in environmental burden of disease assessment

Disability Adjusted Life Years (DALYs) combine the number of people affected by disease or mortality in a population and the duration and severity of their condition into one number. The environmental burden of disease is the number of DALYs that can be attributed to environmental factors. Environmental burden of disease estimates enable policy makers to evaluate, compare and prioritize dissimilar environmental health problems or interventions. These estimates often have various uncertainties and assumptions which are not always made explicit. Besides statistical uncertainty in input data and parameters - which is commonly addressed - a variety of other types of uncertainties may substantially influence the results of the assessment. We have reviewed how different types of uncertainties affect environmental burden of disease assessments, and we give suggestions as to how researchers could address these uncertainties. We propose the use of an uncertainty typology to identify and characterize uncertainties. Finally, we argue that uncertainties need to be identified, assessed, reported and interpreted in order for assessment results to adequately support decision making.

Questions on causality and responsibility arising from an outbreak of Pseudomonas aeruginosa infections in Norway

In 2002, Norway experienced a large outbreak of Pseudomonas aeruginosa infections in hospitals with 231 confirmed cases. This fuelled intense public and professional debates on what were the causes and who were responsible. In epidemiology, other sciences, in philosophy and in law there is a long tradition of discussing the concept of causality. We use this outbreak as a case; apply various theories of causality from different disciplines to discuss the roles and responsibilities of some of the parties involved. Mackie's concept of INUS conditions, Hill's nine viewpoints to study association for claiming causation, deterministic and probabilistic ways of reasoning, all shed light on the issues of causality in this outbreak. Moreover, applying legal theories of causation (counterfactual reasoning and the "but-for" test and the NESS test) proved especially useful, but the case also illustrated the weaknesses of the various theories of causation.

We conclude that many factors contributed to causing the outbreak, but that contamination of a medical device in the production facility was the major necessary condition. The reuse of the medical device in hospitals contributed primarily to the size of the outbreak. The unintended error by its producer – and to a minor extent by the hospital practice – was mainly due to non-application of relevant knowledge and skills, and appears to constitute professional negligence. Due to criminal procedure laws and other factors outside the discourse of causality, no one was criminally charged for the outbreak which caused much suffering and shortening the life of at least 34 people.

Linking causal concepts, study design, analysis and inference in support of one epidemiology for population health

In this paper, which is dedicated to Dr. Calvin Schwabe, I review the concepts of causation and how they impact on study design, analysis and interpretation of results. Notwithstanding the fact that no observational study can prove causation, there are a number of issues that if addressed sufficiently, can improve the validity and usefulness of our studies. These are elaborated with specific recommendations for the conduct of future observational research. Approaches that have been useful in my teaching and research experience are also described.

Getting causal considerations back on the right track

In their commentary on my paper Phillips and Goodman suggested that counterfactual causality and considerations on causality like those by Bradford Hill are only "guideposts on the road to common sense". I argue that if common sense is understood to mean views that the vast majority of researchers share, Hill's considerations did not lead to common sense in the past--precisely because they are so controversial. If common sense is taken to mean beliefs that are true, then Hill's considerations can only lead to common sense in the simple and well-understood causal systems they apply to. Counterfactuals, however, are largely common sense in the latter meaning.I suggest that the road of scientific endeavour should lead epidemiologic research toward sound strategies that equip researchers with skills to separate causal from non-causal associations with minimal error probabilities. This is undeniably the right direction and the one counterfactual causality leads to. Hill's considerations are merely heuristics with which epidemiologists may or may not find this direction, and they are likely to fail in complex landscapes (causal systems). In such environments, one might easily lose orientation without further aids (e.g., defendable assumptions on biases). Counterfactual causality tells us when and how to apply these heuristics.