The impact of communicating genetic risks of disease on risk-reducing health behaviour: systematic review with meta-analysis
BMJ 2016; 352 doi: https://doi.org/10.1136/bmj.i1102 (Published 15 March 2016) Cite this as: BMJ 2016;352:i1102
All rapid responses
Rapid responses are electronic comments to the editor. They enable our users to debate issues raised in articles published on bmj.com. A rapid response is first posted online. If you need the URL (web address) of an individual response, simply click on the response headline and copy the URL from the browser window. A proportion of responses will, after editing, be published online and in the print journal as letters, which are indexed in PubMed. Rapid responses are not indexed in PubMed and they are not journal articles. The BMJ reserves the right to remove responses which are being wilfully misrepresented as published articles or when it is brought to our attention that a response spreads misinformation.
From March 2022, the word limit for rapid responses will be 600 words not including references and author details. We will no longer post responses that exceed this limit.
The word limit for letters selected from posted responses remains 300 words.
Hollands et al provide an important contribution to evidence on the potential impact of DNA-based disease risk information on health behaviours. Based on the findings of their systematic review and meta-analysis, the authors conclude that: “existing evidence does not support expectations that such interventions could play a major role in motivating behaviour change to improve population health (1).” However, we contend that this conclusion is premature. In particular, there has been limited population-based research using risk estimates based on multiple genomic variants, for a broad range of health behaviours, and lack of evaluation of whether the impact may be influenced by the presence of other risk factors. We also agree with other rapid responses (Hay and McBride, Janssens, Burton) that highlight other limitations of the previous studies and the need for more research to more fully assess the potential role of genomics in facilitating behaviour change.
With regards to “population health”, few studies in the review were representative of the general population. Most of the studies were based on ‘high-risk’ groups and some studies were restricted to small population subgroups such as factory workers or university students. A number of differences in characteristics between ‘high-risk’ groups and the broader average-risk population have been observed, such as base rates of screening and other health habits, and awareness of genetic testing and genetic literacy (2). Such differences could mean that genomic risk information might impact the broader population differently, but to date, few studies examining the impact of DNA-based disease risk information have been undertaken in populations not defined by a particular characteristic, such as being at higher risk of a particular condition.
Many of the studies in this systematic review communicated disease risk estimates based on single or few genomic variants. Single-gene common variants underestimate the complexity of genomic contributions to common diseases, have little effect on personal risk and are likely to produce risk messages of low motivational potency (2,3). Furthermore, expectations for the role of DNA-based disease risk information in improving population health are arguably based on testing for multiple (as opposed to single or few) genomic variants.3 As technology is advancing, we are finding more genomic variants that contribute to common, complex diseases. In our field of melanoma prevention, common genomic variants have been found in at least 20 genes (4) and they make a strong contribution to melanoma risk prediction (5,6). We recently conducted a pilot randomised controlled trial in which we provided the public with personalised genomic risk as a combined risk estimate based on variants in 21 melanoma genes. Our results showed strong interest, feasibility and acceptability of giving such information to the public, and potential ‘clinically important’ improvements to prevention behaviours (manuscript submitted). However, a larger, adequately-powered study is required to fully assess the impact of this intervention.
Only one study (Glanz et al, 2013)(7) in this review measured sun-related behaviours as a primary outcome. Despite only 73 people in the study, they found a standardised mean difference of 0.43 (95% CI -0.03 to 0.90, P=0.07), indicating a borderline-significant increase in sun protection for the intervention group compared to the control group. Another finding by Glanz et al, not reported in the review by Hollands et al, was a significant increase in the frequency of skin self-examinations (P=0.002), which is important for secondary prevention of melanoma. Hollands et al argue that “clear justification” is required to conduct additional large scale trials in this field. We submit that these moderate effect sizes represent justification “that efficacy of a clinically important degree is possible” and deserve further investigation in a larger study, and with a broader population since the people in this study had a family history of melanoma. Equally important is a wide-ranging and interdisciplinary discussion as to what a “clear justification” can and should comprise.
In summary, we contend that in order to make firm conclusions about a role for DNA-based disease risk information to improve population health, we need more population-based research using risk estimates based on multiple genomic variants, and evaluation of whether the impact may be influenced by the presence of other risk factors. More research is also needed to assess whether the effect of genomic risk information might differ when used as a strategy for primary prevention (e.g. targeting health behaviours to reduce disease incidence) versus secondary prevention (e.g. targeting screening behaviours to improve early detection) of disease. These investigations should be accompanied by psycho-social and ethical evaluations of the impact of this information, and deliberation over what criteria are appropriate to use as justifications for this kind of data.
References
1. Hollands GJ, French DP, Griffin SJ, et al. The impact of communicating genetic risks of disease on risk-reducing health behaviour: systematic review with meta-analysis. BMJ 2016;352:i1102.
2. McBride CM, Koehly LM, Sanderson SC, et al. The behavioral response to personalized genetic information: will genetic risk profiles motivate individuals and families to choose more healthful behaviors? Annu Rev Public Health 2010;31:89-103.
3. Bloss CS, Madlensky L, Schork NJ, et al. Genomic information as a behavioral health intervention: can it work? Per Med 2011;8(6):659-67.
4. Law MH, Bishop DT, Lee JE, et al. Genome-wide meta-analysis identifies five new susceptibility loci for cutaneous malignant melanoma. Nat Genet 2015;47(9):987-95.
5. Cust AE, Goumas C, Vuong K, et al. MC1R genotype as a predictor of early-onset melanoma, compared with self-reported and physician-measured traditional risk factors: an Australian case-control-family study. BMC Cancer 2013;13(1):406.
6. Fang S, Han J, Zhang M, et al. Joint effect of multiple common SNPs predicts melanoma susceptibility. PLoS One 2013;8(12):e85642.
7. Glanz K, Volpicelli K, Kanetsky PA, et al. Melanoma genetic testing, counseling, and adherence to skin cancer prevention and detection behaviors. Cancer Epidemiol Biomarkers Prev 2013;22(4):607-14.
Competing interests: No competing interests
We read with interest the recently published Cochrane review examining the behavioral impact of genomic-based assessments of health risks. The authors came to similar conclusions as their prior 2010 review that there is a lack of evidence that genomic risk-based feedback relating to common chronic diseases has utility for motivating behavior change, such as smoking cessation, improvements in physical activity, and diet. While the authors acknowledge that the evidence was of generally low quality, the specific requirements of Cochrane review methodology limit their ability to characterize study weaknesses. Considering the relative paucity of research, the established conceptual and empirical importance of effective risk communication as a precursor of behavior change, and several other caveats noted below, we urge the research community to consider this review as a spring board for a new generation of genomic risk communication and behavior change research and not a death knell.
To this end, we suggest several aspects of the reviewed studies that limit their external validity and point investigators to much-needed improvements in research rigor. First, the majority have confounded study participation with preexisting motivation to change; if study participants are highly motivated to change behaviors at baseline, as they have been when choosing to participate in novel behavioral-genomic interventions, then the motivational impact of the intervention will undoubtedly be attenuated. Second, the study samples have been relatively small and demographically homogeneous; some communications have been paired with other interventions, others were not. In our view, this range of variability in a relatively small number of studies during a time when genetic risk information has been evolving raises more questions than answers. Third, decades of research has shown that risk perception may be necessary but not sufficient to induce behavior change. Those of us in the field know all too well that risk information must be paired with skills-building and social support to effectively influence behavior change. Thus, we should not be surprised that genetic feedback alone does not lead people to change entrenched, complex habits such as tobacco use and diet.
The rigor of the Cochrane review process and the terse punch lines they afford run the risk of losing sight of these numerous and important caveats. This is particularly concerning given that the field of genomic risk communication is at such an early stage (18 studies with most published in the past 6 years) and a time when genomic risk information is in flux and evolving quickly. Further, the review could conclude relatively little about differential behavioral effects of positive versus average risk genetic risk information, due to the paucity of published data. Most of the studies reviewed evaluated one-time only risk communication. New technological capacity such as whole genome sequencing will not only increase the literacy demands on participants who receive genomic information, but likely will lead to serialized communications rather than one-time risk results. Moreover, in the realm of chronic disease risk, the challenge going forward will be to effectively convey how genetic variation interacts with environmental exposures. This has largely been overlooked by current approaches. These types of communication are at the heart of envisioned precision medicine approaches and will require rigorous evaluation regarding effects on prevention and treatment adherence. Thus, the challenges of risk communication will continue to be of paramount importance for the field of behavior change.
In closing, momentum is rapidly building for stratified, precision medicine approaches to disease treatment and prevention. We contend that there are many evolving challenges regarding how to use genomic risk communication in public health. Indeed, research to consider optimal applications of genomic information for reducing risks of common, preventable illnesses remains largely unexplored. Well-designed studies are needed, agnostic to outcome. Such rigorous research efforts could position us to shape improvements in behavior change approaches that will be critical for achieving the vision of precision medicine in the coming years.
Competing interests: No competing interests
According to Hollands et al., their review “suggests that communicating DNA-based disease risk estimates has little or no effect on risk-reducing health behavior.” This conclusion needs to be nuanced.
First, communicating risk estimates based on one or a few single nucleotide polymorphisms was never expected to impact health behavior for diseases that are caused by complex interactions between numerous genetic and nongenetic risk factors. The randomized trials lacked equipoise; they were premature and naive, and their results just showed what was already known.
Second, the authors conducted separate meta-analyses for smoking cessation, diet, physical activity and screening attendance, rather than separate meta-analyses for each genetic test. For example, among the seven studies that investigated the role of genetic testing on diet were studies that tested: the APOE gene among people with a family history of Alzheimer’s disease; the FTO gene for obesity risk in students; and multiple type 2 diabetes genes in people who were overweight or obese. Altogether, the 18 trials investigated 15 different genetic tests predicting 13 different diseases in 16 different groups of people. Pooling such different studies in a meta-analysis assumes that genetic tests are expected to have a similar impact on health behavior, irrespective of their predictive ability and of what outcome is predicted in whom, but this is unlikely a valid assumption. Genetic tests may have an impact on preventive behavior when the predictive ability of the test is sufficiently high and when behavior change is known to reduce the risk of disease.
Finally, most genetic tests in the review are already outdated. Eleven out of 18 studies tested a single genetic variant and also the studies that predicted outcomes using multiple polymorphisms could now consider a much larger number of variants, thereby improving the predictive ability of the test. The review shows that testing one or a few genetic variants has no impact on health behavior, but this conclusion cannot be extended to DNA-based disease risk estimates in general. While it may be unlikely that genetic tests for common diseases will ever become predictive enough to motivate behavioral change, this lack of impact cannot be concluded and anticipated from this review of old, premature tests.
Competing interests: No competing interests
About the commentary by Hilary Burton.
Until now, the argument of the "individual interest" of drugs and tests that could not be identified properly by randomized controlled trials was used only by homeopathy supporters. However, any drug has “individual” effects, and drugs, just as tests, are designed to be used in groups of individuals--that is, a population--which can thus be considered in statistical terms.
Moreover, what about the adverse effects, such as anxiety caused by the announcement of a genetic risk?
In the minds of the population genetic risk means fatality.
This could also be an explanation for the absence of behavior modification after the announcement of a genetic risk.
Furthermore, what would be the interest of tests that are one among many factors and that modify only “slightly” the risks for individuals in most common diseases?
Isn’t the debate already “polarized” by lobbyists who are attempting to dislocate the whole drug’s regulation system on behalf of the hopes raised by so called “precision medicine”? (cf 21st century Cure’s act and adaptive pathway).
Competing interests: No competing interests
The results of this update of the 2011 Cochrane review on the impact of communicating genetic risks of disease on risk-reducing behaviour comes as no surprise to our group, the PHG Foundation, who have been working in the area of genetic and genomic technologies for almost two decades. We know that there is a paucity of evidence linking risk communication, genomic or otherwise, and the motivation of sustained behaviour change and that the quality of existing evidence is poor. In other contexts, we know that motivating behaviour change is extraordinarily difficult, and there appears to be no reason to suppose that genomic information is endowed with greater influencing power for any given level of risk than other types of information.
Furthermore, as the authors point out, when it comes to assessing the impact of communicating predictive genetic information, the quality of the evidence presented is poor, the studies selected for review typically have poor or flawed study design or utilise outcomes or timings of outcome assessment that are insufficiently robust. Additionally, the criteria for inclusion of research within the review required that each study adopted a randomised controlled trial design, meaning that the results represent population level data rather than individual level data, which in the context of attempting to assess an intrinsically individualised approach is problematic.
As an organisation committed to making the best use of biomedical science to improve health, we consider that genomics has an important role to play in this endeavour as one of a number of relevant biomarkers of common, complex disease susceptibility. Collective use of these biomarkers will enable the development of a molecular taxonomy of disease that can in turn drive the development of more accurately tailored and targeted interventions offered to population subgroups and individuals. We strongly agree with the authors that DNA tests ‘may have a role in stratifying populations by risk, to enable clinical and behavioural interventions to be targeted at those at increased risk’. Preventive interventions might include screening tests and surveillance, drug treatments or even occasionally screening.
However, it is now widely accepted in the field of personalised medicine that for common complex diseases, genomic information alone will be insufficient to either motivate behaviour change, or (more realistically) to stratify populations to receive more rational and targeted interventions. Apart from anything else, inclusion of genetic data alongside other personal, physiological or biomarker data adds little to the accuracy of the risk prediction for these diseases and the size of effect is small. Thus for an individual the eventual risk estimate will only be slightly increased or slightly decreased. Why then would these small differences intrinsically motivate significant changes in behaviour? The public has more sense than to react differently just because of the ‘genomic’ origins of the information when the absolute risk estimates are very similar.
In this sense, this paper tackles a question that is no longer considered of central relevance within the field. We, and others working in this area anticipate that any potential future utility of genomic information in preventing or treating disease will arise from the appropriate matching of individuals or groups of individuals (classified on the basis of genomic and a host of other biomarkers and environmental factors) with interventions known to be most effective in these groups.
Rather than dismissing the role of genomics and personalised medicine more widely on the basis of sparse and poor quality evidence relating to a question of only marginal relevance to the endeavour at hand, we suggest that health policy makers should focus on generating robust evidence on how genomic and other personal physiological and environmental information could be optimised to guide clinical and behavioural interventions; increase transparency around what is known about the scientific validity and clinical utility of genomics (especially in direct-to-consumer contexts) and develop more nuanced approaches that acknowledge the potential impact of both genomic and other risk information in empowering individuals to improve their health.
Progress is dogged by a determination amongst both personalised medicine evangelists and sceptics to present genomic analysis in polarised ‘all or nothing’ terms. As we learn more about the widely varying contribution of genomics to disease it becomes ever clearer to us at the PHG Foundation that a more balanced and nuanced approach to its role in prevention is essential. The construction and demolition of ‘straw man’ arguments of the type described in this article are not helpful in encouraging such a balanced approach to health policy making. Instead, as health policy researchers and advisers we should be seeking to critically evaluate and, where the evidence warrants it, support the work of those researchers and clinicians aiming to develop personalised medicine approaches in the areas in which its potential benefits are most urgently sought and required, such as rare diseases and cancer.
As better quality evidence accrues in the future that addresses the potential role for genomics in ‘personalising prevention’ for common chronic diseases it will of course be important to evaluate this evidence and consider its implications for health policy. It is hard, however, to see the value in undertaking such evaluation prematurely, as it risks undermining the current efforts within the field of biomedical and clinical research to develop this evidence base more fully.
Competing interests: No competing interests
Re: The impact of communicating genetic risks of disease on risk-reducing health behaviour: systematic review with meta-analysis
Hay and McBride highlight a number of caveats about our review (1) some of which we share.
First, we agree that the results of our review should not be read as a “death knell” for research on this topic but rather as the basis for subsequent studies with more methodologically robust designs than many of those in the 18 included RCTs.
Second, while we agree that in theory the null findings might reflect ceiling effects if all participants in the RCTs were very highly motivated to change their behaviour, we think that this is very unlikely. The studies varied in the populations studied (including general populations as well as clinical populations), the settings in which the studies were conducted and the behaviours observed.
Third, while we acknowledge that the included studies varied in sample size and the interventions surrounding the communication of test results, the studies share many similarities with each other, and with commercially available interventions, in terms of their key intervention components and statistically are not heterogeneous.
Fourth, we agree that risk information is rarely sufficient for behavior change. We would add that neither is it always necessary given that many of our routine or habitual behaviours are elicited by cues in our environments, often without our awareness (2). Indeed, we would suggest that focusing on environmental cues is likely to influence such behaviours more than repeatedly providing information about genetic risk, gene environment interactions or the results of genome sequencing. Additionally we note that the studies in our review assessing the impact of communicating disease risks on non-routine behaviours i.e. attendance at screening or attending for behavioural support, also indicated an absence of effect. While these comprised only two studies, both were at low risk of bias.
Finally, we endorse Hay and McBride’s call for more methodologically rigorous research, but don't share their optimism for the promised benefits of “precision medicine” which, as Caulfield (3) notes, “distracts people from more evidence-based approaches to improving population health…and helps to legitimize the marketing of unproven genetic-testing services”.
Theresa Marteau, Gareth Hollands, David French, Simon Griffin, Toby Prevost, Stephen Sutton and Sarah King
1. Hollands GJ, French DP, Griffin SJ, Prevost AT, Sutton S, King S, & Marteau TM. (2016) The impact of communicating genetic risks of disease on risk-reducing health behaviour: systematic review with meta-analysis. BMJ; 35210.1136/bmj.i1102
2. Hollands GJ, Marteau TM, Fletcher PC. (2016) Non-conscious processes in changing health-related behaviour: a conceptual analysis and framework. Health Psychol Rev:1-14; 10.1080/17437199.2015.1138093
3. Caulfield T. The limits of personalized medicine. The Atlantic 16 March 2016. http://www.theatlantic.com/health/archive/2016/03/does-knowing-personal-...
Competing interests: No competing interests