Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
Editor,—The editors of theJournal of Epidemiology and Community Healthshould be congratulated on their far reaching audience. The comment entitled “Genes as causes: scientific fact or simplistic thinking?” is a good excuse for transdisciplinary discussion.1 The response to such comment could indeed be entitled “Genes as causes: simplistic thinking? No, scientific facts”. The “genes as causes” tenet is more than an opinion and there is ample evidence that, for certain severe diseases, “genes are sufficient causes for human suffering” as long as their effects are not so detrimental that they are incompatible with life, in which case proof of causation will be even more difficult to obtain. The issues raised by Porta and colleagues1 are legitimate but, in our opinion, they are extreme.
As epidemiologists well know, there are few human illnesses that can be attributed to a single factor. When so, this requires exposure—here used in a very broad term—to factors exerting very strong effects. You can refer to genes as “exposures”, as we are inevitably “exposed” to our genetic background and cannot in fact effectively prevent exposure to genes. Only very severe mutations, fortunately uncommon, are necessary and sufficient causes of disease: it is not worth providing a detailed list of gene mutations leading to the lack of a protein product or to protein devoid of function, such information can readily be found in classic genetics textbooks.2 For the majority of medical conditions, genes are only “one of the pool of exposures” leading to disease, therefore often being neither absolutely required nor sufficient for disease development. Even in the case of a severe illness, such as cystic fibrosis, mutations in the gene encoding the cystic fibrosis transmembrane regulator (CFTR) lead to a complex constellation of phenotypes ranging from classic cystic fibrosis to mild forms of disease such as absence of the vas deferens, among others.3 So called “severe mutations” in CFTR gene are inevitably associated with severe illness and “mild mutations” are likely to be associated with illness only in conjunction to other genetic or environmental exposures. Of course, the most common diseases in Western societies are multifactorial, the latter term including both genetic and environmental exposures.
The comment by the JECH editors is striking because the complex role of genes in disease is no different from the complex role of environmental or lifestyle exposures. Just an example. Despite the fact that there is no doubt that tobacco causes lung cancer, it is neither necessary nor sufficient cause for its development as not all smokers develop the disease and some non-smokers also develop lung cancer. We, at the bench, would never ask “Tobacco as a cause of lung cancer: scientific fact or simplistic thinking?”
Epidemiologists have laid the path to a stringent analysis of the role of genes as causes of disease and, when dealing with genes, there is no need to stay away from their paradigm. Bradford Hill proposed criteria of causality in epidemiological studies and various authors have dealt with their applicability and have made specific recommendations to establish causality.4 5
Time sequence. There is no question that the genes precede the development of disease.
Strength of association. This criterion does not always apply well to “genes as exposures”. In recessive conditions, both alleles need to be altered for disease to develop; in dominant conditions, only one is necessary. Therefore, genes often act in a dichotomous, rather than continuous, manner.
Experimental evidence/biological plausibility. Supporting data from in vitro work and animal models are generally stronger for severe than for mild mutations. In addition, the effect of mild mutations may not be reproduced experimentally unless the other factors contributing to disease development can be incorporated into the studies.
Other arguments, such as specificity, defined as a cause giving rise to a single effect,5require a very close scrutiny. For example, in some situations different mutations in a given gene yield very different diseases (that is, activating mutations in the ret oncogene are involved in familial forms of cancer whereas inactivating mutations are involved in Hirschprung's disease).6 Consistency, the reproducibility of an association in different populations under different circumstances,5 may be more easily demonstrated for severe genetic defects than for milder ones as the associated cofactors, genetic or environmental, may be variably represented in different populations.
These days, when a bulk analysis of the genome cannot be as refined as desirable, it is particularly risky to draw conclusions for each fragment of the genome or for specific subsets of patients. While many methodological aspects of genetic epidemiology deserve attention and call for solutions, there is no doubt that the study of the role of genes in disease will develop greatly in the future.
Geneticists, molecular biologists, epidemiologists, and clinicians need to work together to establish the role of genes as susceptibility factors for disease development. Misunderstandings—even lack of trust—are likely to occur, mostly because of the differences in the scientific approaches used by different disciplines, the power/arrogance of certain paradigms, and superficial communication. We very much hope that this response to the comment of the editors ofJECH will stimulate cross talk between epidemiologists and molecular biologists.
Authors' reply: How is causal interference practised in the biological sciences?
We appreciate the interest of the three distinguished biologists in our editorial.1-1 Indeed, we also believe there is a need to strengthen the dialogue between the biological and the public health sciences,1-2 1-3 and we are glad that the pages of theJournal of Epidemiology and Community Healthcan host views that would otherwise remain largely within the galaxy of biological journals.
Our commentary1-1 briefly addressed three interrelated but distinct issues. Firstly, the communication and socialisation processes in Western societies of the discoveries of the biological sciences, and the cultural constructs (including “risks”) that, as a result, sometimes are “down loaded in our collective imageries”. Secondly, the internal and external validity of causal inferences that biological sciences make both within and outside microbiological levels. And thirdly, the need to assess the clinical and epidemiological coherence of biological observations.1-3
The letter by Real and colleagues focuses on causal inferences. We agree with many of the points they make; for example, for most medical conditions, even for some severe illnesses, genes are only “one of the pool of exposures” leading to disease; the most common diseases in Western societies are multifactorial. We think that many other epidemiologists and public health professionals in general would also agree with Real et al. Non-deterministic models, for instance, have quite a tradition in public health,1-4-1-6 and have long been used to assess the causal effects of smoking. Hill's criteria themselves stem from one of such pioneering efforts, the 1964 Report of the Advisory Committee to the US Surgeon General, “Smoking and Health”.1-6
We note with some pleasure—and some surprise—their belief that “epidemiologists have laid the path to a stringent analysis of the role of genes as causes of disease”. If true, such a contribution to intellectual rigour would be awesome! It would certainly be something to add to any inventory of the accomplishments of epidemiology.1-7
Real et al also propose that, “when dealing with genes, there is no need to stay away from their [epidemiologists'] paradigm”. Is this true or just a brillianttour de force? Is it just for lack of a better alternative?. No matter what (though it does matter!), their application of “epidemiological” criteria of causality to the “genes as causes” tenet yields a remarkable sketch of a pragmatic framework for causal inferences in human genetics and “clinical biology”. Such framework, if adapted and developed, would perhaps enable us to better assess the potential causal roles of genes in human health—an issue that is obviously central to biology, but which in our view is also quite important for clinical medicine and for public health and social sciences, as the hypotheses and findings of biology continue to have such wide influence throughout the world.1-1-1-3 Certainly, space constraints did not allow Realet al to develop their causal exercise. None the less, their effort shows well the possibilities—and some limitations—of the causal criteria popularised by Hill and Susser.1-4-1-6
This brings us to one final set of questions. How are causal inferences achieved by the biological sciences nowadays?. How do biologists practise causal inference in daily (scientific) life? What properties do they seek most, in a cause? Are there many modern “biological” sets of causal criteria in use? Where and how are they taught and discussed? Having worked closely for over a decade with biologists—including Professor Real1-8—we do not ignore that the objects and processes of scientific inquiry, the nature and tools of studies, the timing and time span of experiments, and countless other circumstances differ in biology and epidemiology. Yet, it intrigues us whether biologists apply explicit rules that in some way resemble those applied in epidemiology.1-9
The letter by Real and colleagues shows that different “cultures of causation”1-1 could coexist, cross fertilise each other and find common ground—a reasonable endeavour for these times, when some cultures and processes converge while others suffer increased fragmentation? Perhaps epidemiological tools and theories on causality will further help integrate scientific cultures and evidence.1-2 The letter also suggests that molecular, genetic and cellular observations can be productively challenged by clinical and epidemiological evidence. And that biological facts can be assessed for coherence with facts and theories at the higher aggregate levels.1-1
We are indebted to Douglas Weed, Ana M García, Paolo Vineis, Ana Diez Roux and Alfredo Morabia for their generous sharing of ideas concerning some of the issues discussed here.