Wim van Drongelen, in, 2018 14.1 IntroductionIt is common practice to decompose multichannel data into its components (e.g., Chapters 7 and 28) to detect structure in complex data sets. Another question that is often posed concerns the causal structure between channels or components: that is, does one channel generate another? In neuroscience the underlying question is: does one area in the brain activate others? A typical example is when an epileptologist examines multichannel recordings of brain electrical activity and attempts to find the source (focus) from where epileptic seizures originate.

What does causality mean? Causality is defined by the lexicographers at Oxford Dictionaries as The relationship between cause and effect., The principle that everything has a cause. ‘To deny causality is to deny divine wisdom, for causality is a necessary relation.’ ‘I also point out the difference between causality and correlation.’ ‘Note, however, that such a correlation does not indicate which is the relation of causality between the two variables.’.

Often this task is accomplished by finding the signals that lead or lag; the leading signals are then considered as causing the lagging ones. Cross-correlation or nonlinear equivalents (such as mutual information) can be used to formalize and quantify timing differences between signal pairs in multichannel data sets ( Chapter 13). We have to start pessimistically by pointing out that translation from lead-lag to causality is not strictly possible—the example in Fig. 14.1 demonstrates this. If we record from areas A and B in Fig. 14.1A, our method of interpreting lead-lag as a causal relationship A → B is correct. However, if we measure signals from A, B, and C ( Fig. 14.1B), we conclude that A → B, A → C, and B → C. We are only partly correct: the two former relationships are correctly inferred but the latter is not. It would get even worse if we hadn't recorded from A in this example: then we only find B → C and we would be 100% incorrect.

So equating lead-lag with causality/connectivity can be incorrect. Having said this, in many studies in neuroscience this is (conveniently) ignored and timing in signals is frequently used as an argument for connectivity. Often, authors use terms such as “functional connectivity” or “synaptic flow” to (implicitly) indicate the caveats above. Because we often know the typical conduction velocity and delays caused by synaptic transmission, we can (at least) recognize unrealistic delays.

For example, if we know that areas B and C in Fig. 14.1B are 10 cm apart and that conduction velocities of the fibers between B and C are ∼1 m/s, we can expect delays ∼100 ms plus a few milliseconds for each synapse involved. Now suppose that in this example the delay between B and C ( Δ t 2 − Δ t 1 ) is ∼5 ms, such a value is far below the expected delay of over 100 ms, which is an indication that direct connectivity doesn't play a role in the observed lead-lag between B and C.

Causality assessment is the method by which the extent of relationship between a drug and a suspected reaction is established, i.e., to attribute clinical events to drugs in individual patients or in case reports. 25 Many systems can be used to make a causality assessment of ADR reports among which the causality categories described by the Uppsala Monitoring Centre ( Table 16.3) and the Naranjo algorithm ( Tables 16.4 and 16.5) are the most commonly used. 26,27 Both of them have not been validated, and studies evaluating agreement between the two did not give consistent results. 28,29 The adoption of a particular system will be based on the consensus of the causality assessment committee involved. However, in case of analysis of serious ADRs, considerations of both systems should be used to ascertain causality. TermAssessment CriteriaCertainA clinical event, including laboratory test abnormality, occurring in a plausible time relationship to drug administration, and which cannot be explained by concurrent disease or other drugs or chemicals.

The response to withdrawal of the drug (dechallenge) should be clinically plausible. The event must be definitive pharmacologically or phenomenologically, i.e., an objective and specific medical disorder or a recognized pharmacological phenomenon, using a satisfactory rechallenge procedure if necessary.Probable/likelyA clinical event, including laboratory test abnormality, with a reasonable time sequence to the administration of the drug. Unlikely to be attributed to concurrent disease or other drugs or chemicals, and which follows a clinically reasonable response to withdrawal (dechallenge).

Rechallenge information is not required to fulfill this definition.PossibleA clinical event, including laboratory test abnormality, with a reasonable time sequence to administration of the drug, but which could also be explained by concurrent disease or other drugs or chemicals. Franklin, in, 2016 AbstractCausality is not often disputed in criminal prosecutions because the nature of the exposure (eg, gunshot, knife, blunt trauma) is strongly associated with the injury outcome (penetrating wound, skull fracture, etc.) such that causality is determined as a matter of common sense. In some cases that are prosecuted criminally, however, there are questions that require a reliable analysis of comparative risk in order to assess pivotal probabilistic issues pertaining to guilt or innocence. In this chapter a variety of examples of how forensic epidemiology methods are used in both the prosecution and defense of criminal matters are presented, illustrating the flexibility and duplicability of the methods. Mutual causality recognizes that there might be a relationship between two things, yet the outcome is contextual and variable, based on multidirectional influences, feedback from other parts, and rule-governed processes that control the system. This variability leads to increased adaptability as the more variables that there are in a system, the more mechanisms that can potentially provide redundancy. Mutual causality provides for an integrated understanding of life and is better able to explain how the body handles and assimilates the many different influences that it encounters than can linear causality (see Fig.

‘Reality appears as a dynamically interdependent process. All factors, mental and physical, subsist in a web of mutual causal interaction, with no element or essence held to be immutable or autonomous’ ( Macy 1991). Causality (a) Linear causality: there is a direct correlation between cause and effect; (b) mutual causality: a single outcome can be due to the interaction of various factors; (c) mutual causality: a single factor can result in various outcomes.Mutual causality explains the uniqueness of individuals. The impact of accidents or stressful situations is not the same for every person, individuals with the same disease often have different symptoms, diseases progress differently for each person, the impact of the same symptom is experienced differently, etc.

The same food for one person can be medicine; for another poison. The same treatment can have opposite effects on different individuals. Life is complex, human beings are complex. Mutual causality recognizes this complexity, the interrelationship of individuals to their environment, and the uniqueness of individuals. Jytte Brender, in, 2006 Perils and PitfallsThe causality analysis is the weakest point of the method because the interpretation and therefore the final conclusion is completely dependent on this analysis.

It can be difficult to uncover the skeletons in an organization and to discuss the problems without hurting somebody's feelings. Alternatively, someone may try to cover up and hide the problems.On the other hand, the method is less dependent on whether all divergences are included or whether you get the right divergences defined as key divergences because the actual causality will often manifest itself in more than just one symptom. Rachel Dankner, in, 2019 Reverse CausalityReverse causality is a possible explanation for associations between diabetes and certain types of cancer. In a study of new users of diabetes medications, patients with upper gastrointestinal cancers (esophageal, stomach, pancreatic, liver cancers) were more likely than patients with other cancer diagnoses to have initiated insulin use during the 6 months prior to the cancer diagnosis. This supports the possibility of reverse causation, that is, that upper gastrointestinal cancers promoted diabetes, rather than the converse. To account for reverse causality, studies of the relationship between the two diseases have often excluded from analysis the early period, ranging from 3 months to 2 years, following the diabetes diagnosis.

Avigan, in, 2013 Bayesian MethodsCausality assessment of individual liver injury cases using a Bayesian approach is probabilistic and requires a large databank of epidemiological and clinical information about all causes of liver injury gathered from a background population that is demographically similar to the suspect case 52. Starting with a slate of all possible etiologies of acute liver injury and their individual probabilities quantified according to incidence in a similar background population, as potential nondrug causes are individually excluded and clinical information and laboratory and histopathological data are incorporated into a logistic regression analysis, probabilities of the remaining potential causes, including DILI, are quantitatively recalculated and strengthened. Although such an iterative analysis is theoretically quantitative, and could be plausibly supported with computer-based Bayesian computational aids, an essential characteristic of causality assessment in individual liver injury cases is the inevitability of residual uncertainty. This uncertainty is the result of current gaps in knowledge surrounding all possible etiologies and phenotypes of acute liver disease. In the Acute Liver Failure Study Group’s recently reported prospective study despite information gathering and a comprehensive diagnostic workup of each enrolled study subject, approximately 15% of all patients with ALF were found to have an indeterminate cause for their liver injury 53–55. It is conceivable that causes of liver failure in an indeterminate group include such entities as undiagnosed acetaminophen toxicity, acute type E viral hepatitis for which routine commercial serological testing must be developed 56, known hepatotoxic pathogens that were not detected, viruses or organisms not currently recognized by biomedical scientists as possible pathogens, and exposure to hepatotoxic drugs not reported to the investigators.

In addition to such undetermined causes, in some cases the possibility that drug-drug or drug-disease interactions have synergized to cause liver injury can also pose a significant challenge for assessing the causal role of a suspect drug. Hatton, in, 2014 3.1.2 Research DesignCausality is typically inferred from high-quality group experimental, group quasi-experimental, and single-subject designs ( Cook & Cook, 2013; Council for Exceptional Children, 2014; Gersten et al., 2005; Horner et al., 2005; Kratochwill et al., 2010; Odom et al., 2005). Regardless of the design, research must be of high quality to be trustworthy. Therefore, quality indicators for varying types of research must be used to determine if studies are of sufficient quality to be included as documentation for EBPs. For group designs, quality indicators based on Gersten et al. (2005), such as those developed by the National Professional Development Center on Autism Spectrum Disorders (NPDC-ASD; Wong et al., 2014), may be employed. For single-subject designs, quality indicators based on Horner et al.

(2005) and/or Kratochwill et al. Mercenary kings ign. (2010) may be employed. Council for Exceptional Children (2014) recently published a comprehensive guide for identifying EBPs that includes quality indicators for group and single-subject designs.

William Lee, Matthew Hotopf, in, 2012 Reverse causalityReverse causality occurs when the exposure is caused by the disorder rather than the disorder being caused by the exposure. In the life event literature, one possibility would be that patients with depression are more likely to suffer life events. Their depression may lead to an under-performance at work that may cause them to lose their job. Much of the research on life events has sought to address these concerns.Sometimes, reverse causality may be quite subtle. For example, obstetric complications are associated with schizophrenia.

The usual assumption is that this is because obstetric complications lead to subtle brain damage which predisposes to schizophrenia. An alternative view, however, is that individuals who later develop schizophrenia already have developmental brain abnormalities in utero that make them more prone to difficult deliveries, since childbirth requires the active participation of the fetus as well as the mother.Having satisfied yourself that the reported association was not due to chance, bias, confounding or reverse causality, you are then in a position to take the association at face value – to give credence to the idea of the exposure in the study really being the cause of the outcome.