Each week, Garrison Keillor shares with National Public Radio listeners the latest news from Lake Wobegon where “all the women are strong, all the men are good looking, and all the children are above average.” The concept of “average” is deeply rooted in our scientific analysis of all health related traits such as height, weight and health indicators (such as blood sugar, cholesterol) and in assessing the likelihood of developing disease. There are several ways to measure average such as mean, median and mode that reflect different approaches towards evaluating central tendencies. For every characteristic, we graph the distribution values in the population (e.g. the Normal distribution, see figure) and quantify averages and variation in values in the “population” and its various subgroups.
When it comes to disease risk, nobody is average. In a recent tweet, Eric Topol stated: “your odds for cancer? Wrong! This is average; average is over…when do we get smart?” In this case, we already know that our odds of developing cancer vary by age, ethnicity, geographic location, occupation, education, cigarette smoking in addition to genetic differences among individuals.
Many scientists are frustrated by the “fixation about having some guidelines or recommendations for all people. It just doesn’t stop“. For example, the New York Times reported in 2013 why drinking 3 cups of coffee a day may be good for us. “Well, does that take into account that at least 20% of people carry an allele where the metabolism of caffeine is markedly reduced, and that risk allele has indeed been linked to a higher risk of heart attack? Why should there be a recommendation now that all of us should be drinking 3 cups of coffee a day?”
It is absolutely true that each of us is biologically unique. It makes sense that we should somehow use this uniqueness in disease prevention. A 2014 review paper expands on our emerging ability to measure our uniqueness, advocating comprehensive ‘‘omic’’ assessment of individuals, including DNA and RNA sequence, metabolome, microbiome, etc.. The resulting information could, in theory, individualize preventive strategies to promote health.
The issue here is not whether anybody is average but really what to do about such information for individualized prevention? It is true that prevention guidelines are typically designed to apply to “average” individuals in the population (e.g. breast cancer and colorectal cancer screening guidelines). Occasionally, evidence-based panels do venture into extremes of the risk distribution (e.g. BRCA counseling and testing in women with family history of breast and ovarian cancer). But it is not straightforward, without appropriate data, to develop evidence-based recommendations that apply to subsets of the population defined by all the traditional disease risk factors as well as new “omic” markers.
There are a few challenges to overcome. First, most complex chronic diseases are due to multiple genetic (as well as non-genetic) factors. With a few exceptions, we do not know the full complement of “heritability” explained by the common variants assessed in genomewide association studies. Second, specific genomic information may or may not be clinically actionable even when it is a risk factor. For example we have known for quite some time that the apoε4 allele increases our risk for Alzheimer’s disease, yet there is no proven strategy to prevent the disease. Third, the analysis of genetic-environmental interactions in population studies is still in its infancy. So far it has been difficult to extrapolate results from such analyses to apply on an individual level. One reason is that we need very large numbers of people in well designed epidemiologic studies to make sense of population “subgroups” stratified by many variables. Subgroup analyses in biomedical research can lead to high noise to signal ratio and failures to replicate.
Most genetic and non-genetic risk factors have a weak or moderate effect on the risk of common chronic disease. So stratification of disease risk based on known risk factors including genetics, leaves most people either “slightly above average or “slightly below average” risk (see figure). Even though no one is average, the evidence accumulated in relatively small studies makes it difficult to recommend different courses of action to preserve health for most people. Of course, there are relatively rare outliers at the end of the risk distribution for whom different medical or prevention interventions will be needed (e.g. BRCA) but for any given condition, we have not figured out yet how to use the whole risk distribution to individualize prevention strategies as opposed to population-based strategies that are good for every one (e.g. smoking cessation, physical activity, healthy diet, etc..).
An example of the challenge in using a targeted approach based on genetics is the prevention of type 2 diabetes (T2D). Although our understanding of the genetics of T2D is progressing quickly, the interactions between common genetic variants and lifestyle risk factors have not been systematically investigated in large studies. The EPIC InterAct study provided the ability to examine these effects in a large number (more than 10,000) of new T2D cases, from a cohort of more than 340 thousand participants. The authors studied the combined effects of a genetic T2D risk score and other risk factors. The effect of the genetic score was significantly greater in younger and leaner individuals. Obesity had a more profound effect on increasing T2D risk, however, and this effect was actually maintained across all levels of genetic risk. These results show the high risk of diabetes associated with obesity at any level of genetic score and highlight the importance of universal rather than targeted approaches based on genetics to lifestyle interventions.
Thus, in order to use genomics and other “omic” profiles to tailor prevention recommendations (such as screening, diet, and physical activity), beyond the “average”, more research is needed to assess the validity and utility of preventive interventions in subgroups of the population. Only then, we can actualize the concept of individualized disease prevention.