Limits...
Assessing causal relationships in genomics: From Bradford-Hill criteria to complex gene-environment interactions and directed acyclic graphs.

Geneletti S, Gallo V, Porta M, Khoury MJ, Vineis P - Emerg Themes Epidemiol (2011)

Bottom Line: However, such knowledge has seldom been applied to assess causal relationships in clinical genetics and genomics, even in studies aimed at making inferences relevant for human health.The method we develop in this paper provides a simple and rigorous first step towards this goal.The present paper is an example of integrative research, i.e., research that integrates knowledge, data, methods, techniques, and reasoning from multiple disciplines, approaches and levels of analysis to generate knowledge that no discipline alone may achieve.

View Article: PubMed Central - HTML - PubMed

Affiliation: Department of Statistics, London School of Economics, Houghton Street, London, UK. s.geneletti@lse.ac.uk.

ABSTRACT
Observational studies of human health and disease (basic, clinical and epidemiological) are vulnerable to methodological problems -such as selection bias and confounding- that make causal inferences problematic. Gene-disease associations are no exception, as they are commonly investigated using observational designs. A rich body of knowledge exists in medicine and epidemiology on the assessment of causal relationships involving personal and environmental causes of disease; it includes seminal causal criteria developed by Austin Bradford Hill and more recently applied directed acyclic graphs (DAGs). However, such knowledge has seldom been applied to assess causal relationships in clinical genetics and genomics, even in studies aimed at making inferences relevant for human health. Conversely, incorporating genetic causal knowledge into clinical and epidemiological causal reasoning is still a largely unexplored area.As the contribution of genetics to the understanding of disease aetiology becomes more important, causal assessment of genetic and genomic evidence becomes fundamental. The method we develop in this paper provides a simple and rigorous first step towards this goal. The present paper is an example of integrative research, i.e., research that integrates knowledge, data, methods, techniques, and reasoning from multiple disciplines, approaches and levels of analysis to generate knowledge that no discipline alone may achieve.

No MeSH data available.


DAG with a randomisation node R. R indicates whether X is randomised or allowed to arise naturally. A: U is a confounder. B: U is a mediator. Randomisation allows us to distinguish between these situations.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC3141807&req=5

Figure 3: DAG with a randomisation node R. R indicates whether X is randomised or allowed to arise naturally. A: U is a confounder. B: U is a mediator. Randomisation allows us to distinguish between these situations.

Mentions: In DTF we introduce randomisation as a variable R (Figure 3). To clarify, consider the following example. Assume that X is a binary variable that can be forced to take on a particular value or "set". It takes on two values: "active" (X = a), or "baseline" (X = b). The randomising variable R has the same settings as X as well as the observational setting R= Ф (the empty set). When R = a then X = a with no uncertainty (imagine forcing X to take on this value, say by administering the treatment to a compliant patient). Similarly, when R= b, X = b with no uncertainty. Finally when R = Ф, X is allowed to arise without intervention and can take on the values a and b as in an observational study. For causal inference in DTF we want to estimate (usually the expected value of) the outcome Y given that an intervention has happened. For example, if we want to know which treatment, active or baseline, is better for Y, we might look at the difference in the expected value of Y given these treatments: E(Y / R = a)- E(Y/R = b). This would then be a measure of the causal effect of a vs b. In observational studies, we do not have E(Y/R = a) the interventional expectation; rather, we have E(Y/ X = a, R = Ф) the observational expectation; similarly for b. The question is, therefore, how to make an inference about the former using the latter. One assumption that is often made is that all observed confounders U are observed. However, this is often not possible and other approaches that simulate randomisation, such as the instrumental variable approach known as Mendelian randomisation [37] can be used. See Dawid [41], Didelez [42], and Geneletti [43] for formal examples.


Assessing causal relationships in genomics: From Bradford-Hill criteria to complex gene-environment interactions and directed acyclic graphs.

Geneletti S, Gallo V, Porta M, Khoury MJ, Vineis P - Emerg Themes Epidemiol (2011)

DAG with a randomisation node R. R indicates whether X is randomised or allowed to arise naturally. A: U is a confounder. B: U is a mediator. Randomisation allows us to distinguish between these situations.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC3141807&req=5

Figure 3: DAG with a randomisation node R. R indicates whether X is randomised or allowed to arise naturally. A: U is a confounder. B: U is a mediator. Randomisation allows us to distinguish between these situations.
Mentions: In DTF we introduce randomisation as a variable R (Figure 3). To clarify, consider the following example. Assume that X is a binary variable that can be forced to take on a particular value or "set". It takes on two values: "active" (X = a), or "baseline" (X = b). The randomising variable R has the same settings as X as well as the observational setting R= Ф (the empty set). When R = a then X = a with no uncertainty (imagine forcing X to take on this value, say by administering the treatment to a compliant patient). Similarly, when R= b, X = b with no uncertainty. Finally when R = Ф, X is allowed to arise without intervention and can take on the values a and b as in an observational study. For causal inference in DTF we want to estimate (usually the expected value of) the outcome Y given that an intervention has happened. For example, if we want to know which treatment, active or baseline, is better for Y, we might look at the difference in the expected value of Y given these treatments: E(Y / R = a)- E(Y/R = b). This would then be a measure of the causal effect of a vs b. In observational studies, we do not have E(Y/R = a) the interventional expectation; rather, we have E(Y/ X = a, R = Ф) the observational expectation; similarly for b. The question is, therefore, how to make an inference about the former using the latter. One assumption that is often made is that all observed confounders U are observed. However, this is often not possible and other approaches that simulate randomisation, such as the instrumental variable approach known as Mendelian randomisation [37] can be used. See Dawid [41], Didelez [42], and Geneletti [43] for formal examples.

Bottom Line: However, such knowledge has seldom been applied to assess causal relationships in clinical genetics and genomics, even in studies aimed at making inferences relevant for human health.The method we develop in this paper provides a simple and rigorous first step towards this goal.The present paper is an example of integrative research, i.e., research that integrates knowledge, data, methods, techniques, and reasoning from multiple disciplines, approaches and levels of analysis to generate knowledge that no discipline alone may achieve.

View Article: PubMed Central - HTML - PubMed

Affiliation: Department of Statistics, London School of Economics, Houghton Street, London, UK. s.geneletti@lse.ac.uk.

ABSTRACT
Observational studies of human health and disease (basic, clinical and epidemiological) are vulnerable to methodological problems -such as selection bias and confounding- that make causal inferences problematic. Gene-disease associations are no exception, as they are commonly investigated using observational designs. A rich body of knowledge exists in medicine and epidemiology on the assessment of causal relationships involving personal and environmental causes of disease; it includes seminal causal criteria developed by Austin Bradford Hill and more recently applied directed acyclic graphs (DAGs). However, such knowledge has seldom been applied to assess causal relationships in clinical genetics and genomics, even in studies aimed at making inferences relevant for human health. Conversely, incorporating genetic causal knowledge into clinical and epidemiological causal reasoning is still a largely unexplored area.As the contribution of genetics to the understanding of disease aetiology becomes more important, causal assessment of genetic and genomic evidence becomes fundamental. The method we develop in this paper provides a simple and rigorous first step towards this goal. The present paper is an example of integrative research, i.e., research that integrates knowledge, data, methods, techniques, and reasoning from multiple disciplines, approaches and levels of analysis to generate knowledge that no discipline alone may achieve.

No MeSH data available.