Thursday, January 04, 2018

When Diet Meets DNA: A Premature Love Story

For those of you who aren't familiar with doctoral student Kevin Klatt, he's one of the most thoughtful balanced voices on nutrition around. This fall, I noticed Kevin chatting on Twitter about the realities and limitations of nutrigenomics and invited him to write a guest post on same. Happily, he agreed, and as with all of Kevin's writing, his piece is deep, thoughtful, and fair. If you're considering spending money on genetic testing as related to diet, please read his terrific overview before shelling out your hard earned cash.
Tis the season for New Year’s resolutions and this year, buying a diet plan based on your unique genetics stands to be all the rage. As genetic/genomic technology has greatly advanced over the past 2 decades, we've seen a number of genetic testing services for you (and for your dog) pop up and they're more than willing to help you live your best life. Genetics test are increasingly marketed to health-conscious folks who want to maximize their nutritional health, promising to give you diet recommendations tailored to your specific genetics. Numerous articles have popped up over the years about 'eating for your genes' and, of course, those who want to eat for their own genes get links to websites that will happily take your cash and send you a diet report (some will even send you the food!) or supplements matching your specific genetics. So...how legit are these?

The science of 'nutrigenetics', or the study of the interaction between genetic variation and dietary needs, is a real, although very immature, science. Researchers have long known that individuals respond differently to different diets and that genetics probably plays a role in this variation. Despite the marketing trope that diet advice is 'one size fits all', inherent to the Dietary Reference Intakes (DRIs), estimated nutrient requirement and intake guidelines published by the Food and Nutrition Board (FNB) of the National Academies of Science, is the notion that nutrient needs vary. The DRIs establish estimated average needs for specific nutrients, and how variable these needs are, assuming the normal distribution pictured below; from there, policy makers and clinicians can assess whether a population or individual is at a high risk of inadequate nutrient intake. Why do individual nutrient needs vary? For a lot of reasons, such as differences in an individual’s growth and development, microbiomes, medications, physical activity levels, and, yes, genetics! The field of nutrigenetics focuses specifically on how certain genetic variants impact nutrient metabolism and affect nutrient requirements.

EAR: Estimated Average Nutrient Requirement; Intake at which 50% of the population has needs met
RDA: Recommended Dietary Allowances; Intake at which 97.5% of the population has needs met
AI: Adequate Intake; Unclear risk of adequacy at this intake but likely more than needs of the population.
Nutrigeneticists have generally assessed the interaction of diet and genes in a couple ways. In observational/epidemiological analyses, where we collect biomarkers of nutrient status and self-reported dietary intake, individuals in the cohort can be genotyped for a number of genetic variants. In the same way that we account for a other factors ('covariates') in these analyses, like age and sex, we can include genotypes in statistical models, and assess their independent effects and interactions with intake on nutrition-relevant outcomes. In addition to these observational analyses, there are many randomized controlled trials which have assessed the effect of some nutrition-related intervention on an outcome, and have included genotypes in their statistical analysis afterwards. You might see, for example, that a high protein diet intervention had an effect on weight loss, and when you include an obesity risk genotype into the mix, you might observe that one genotype loses more weight than the other. To date, these are the primary ways that nutrition scientists have addressed the topic of nutrigenetics.

While these are scientifically meaningful approaches, the interpretation of these gets a bit funky and can lead over enthusiastic entrepreneurs to selling more hype than science. A few things to consider about nutrigenetics research before you go buying a test:

1. Very little of the evidence establishes 'cause and effect' - The types of research that I described above include genotypes into the analysis of studies that weren't designed to primarily look at genotype-x-nutrient interactions. Ideally, we want studies that are adequately designed and powered to test the effect of a genotype-nutrient interaction from the get go. If you include genotypes into the analysis after the study was designed, it increases the likelihood that you observe a significant association only by chance. One can imagine that in a field with 20+ essential nutrients/dietary components that are of interest, an array of measured health outcomes, and now the addition of innumerable genotypes, we can make a lot of comparisons. While these comparisons might generate some nice hypotheses, they are at a high risk of being false positives.

What the field needs a lot more of are randomized controlled trials where individuals are genotyped before an intervention ('a priori') and then randomized to a dietary intervention. This allows for much stronger causal inference that a genotype does indeed have an effect. We also need repetition of these trials across diverse ancestries - because of the way genes are possibly inherited, genotypes that we think are causal may just be very close to the actual causal genetic variant on a chromosome! To date, there is one pretty good trial that looked-for individuals with the MTHFR who had 2 copies of the variant ('TT') and then randomized them to a riboflavin supplement. We need a lot more of these before we get too excited about genotype-based dietary recommendations.

2. Individualized isn't so individual - The philosophy that often goes along with genotype-based diet advice is that it moves nutrition beyond the ‘mean’ and really gets to the individual. While a fun marketing line that has led to ‘precision’ everything, the reality is a bit bleaker. Genotypes might sound super ‘individual’ but they really just represent subgroups of the population; similar to the way that we already give nutrient recommendations tailored to specific sub-groups of the population based on age, sex, and life stage, the science of nutrigenetics only refines our educated guess at your nutrient needs, it doesn’t define it. Furthermore, as mentioned above, a lot of other factors can affect nutrient needs, and its unlikely that a single genetic variant will be the only reason that individuals differ in their needs.

3. Effect Size - Popular discussions about 'eating for your genes' that have increased the hype surrounding nutrigenetics are not often concordant with the reality of the data. Effect estimates found in post-hoc, adjusted analyses are often modest (and causality still remains uncertain). Where the field has seen very large effect sizes of genetics variants in outright nutrient deficiency feeding studies, these studies are not available for the majority of nutrients nor do they apply to people consuming relatively nutrient sufficient diets.

In the case of the MTHFR C677T variant, where there is a substantial body of experimental and human data to strongly support the causal nature of the variant, its impact on nutrient requirements, in this case folate, doesn’t really change much about dietary recommendations. As the Academy of Nutrition and Dietetics noted in their position statement on nutrigenomics, there's not good evidence that having this variant increases folic acid needs higher than the current DRIs. To date, we're lacking evidence that genetic variants increase needs so dramatically that they suggest current recommendations are inadequate.

4. Outcomes - The field of nutrition has a pretty good track record of preventing outright nutrient deficiency; much of the population isn't walking around with scurvy. Preventing scurvy, however, is probably not the biggest concern of someone thinking about buying a nutrigenetics test. With nutrition, most folks are interested in the diet that will bring them closest to immortality, and at the very least, one that can help prevent chronic diseases. Unfortunately, it's quite hard to perform large randomized controlled trials (to assess causality) for long enough (to observe meaningful outcomes such as heart attacks) in the field of nutrition, which has led to a lot of confusion about how to best give dietary recommendations based on chronic disease outcomes. Do we put more faith in short term trials while measuring some intermediate marker of disease (like blood cholesterol), or do we rely on big epidemiological studies that can't assess causality and often rely on self-reported dietary intakes. These limitations still apply to nutrigenetics! There's some hope that incorporating genetic variants into current analyses might reduce some of the noise in the data, as seen in the relationship between caffeine and heart attack/hypertension. However, this is largely optimistic, and for most genetic variants advertised today, we lack the confidence to say that changing your diet accordingly will causally impact meaningful outcomes in the same way that we lack this inference from general nutrition data. Alas, immortality remains elusive!

5. Repetition, or lack thereof - I personally think the relationship between caffeine intake, genetic variants, and heart attacks is super interesting and one of the more 'low hanging fruit' hypotheses generated from the field of nutrigenetics. However, to date, we don't have additional studies to support these early interesting observations. Repetition of findings is a key component of an evidence-based recommendation. While companies selling you a genotyping kit might be more than happy to hype one promising early study from the field of nutrigenetics, keep in mind that the science in the field is generally super premature. We generally don't have numerous publications showing similar results across a number of studies and populations. It's quite easy to get super excited about genotypes that show promising relationships that sound biologically plausible, but without seeing that relationship often, across several populations, dietary recommendations are nothing short of rushed.

6. But isn't it motivating? #EatForYourGenes. For the clinicians out there, the science of whether genes affect metabolism and your nutrient requirements is cool but what we may really want to know is the likelihood that communicating that genetic information will lead people to eat a meaningfully better diet. Good science and good intentions coming from the nutrition community have tried to get folks to eat fewer calories, less fat, less sugar, more fiber, etc, but this hasn't always translated so well. With personalized nutrition, the field runs the risk of not only communicating recommendations in a way that might not lead to behavior change, but also having folks $pend a bunch of $ on genetic test$ that don't yield meaningful result$. The last thing the field of nutrition needs is another reason for the public not to trust it.

Where's the evidence at on this? The most recent Cochrane Review assessing the impact of conveying genetic information on behavior included 2 diet studies; the Cochrane review estimated a positive effect of conveying genetic information on changing self-reported diet intakes after pooling data from 2 studies. Before you get too excited, these data hardly support buying a genetic test to tailor your diet. The 2 included studies assessed whether conveying mutation information to individuals with familial hypercholesterolemia (FH) would lower fat intake, and the impact of self-reported changes in general diet after receiving information about ApoE genotype in adult offspring whose parents developed Alzheimer's. Neither of these populations are very relevant for the average person looking to get nutrigenetics testing, nor are their self-reported outcomes that impressive. Another systematic review and meta-analysis including 7 randomized and quasi-randomized controlled trials of nutrigenetics interventions on self-reported diet found no significant effect of the intervention.

Since the publishing of these meta-analyses, the Food4Me study was published, a large multi-centre trial where 3 levels of personalized nutrition (compared to 1 generalized diet) advice were provided, with one level including genotype information; the study reported no benefit from the genetic information arm (though the statistical analyses to determine this aren't shown) on total diet scores. A sub-analysis of this study also looked at whether the personalized nutrition advice would improve adherence specifically to a Mediterranean diet score, finding small added effects of genotype on overall Mediterranean diet adherence, an effect driven largely by an increase in tomato sauce consumption. Before you get too excited, we are talking incredibly small effect sizes here: increases of .25-.43 on a 14 point scale. Translation: letting participants know that they have/don't have 5 nutrient-related risk genotypes leads to a clinically meaningless, marginally significant effect on Mediterranean diet score in a secondary analysis of a large trial. Most folks advocating for nutrigenomics testing aren't going to convey the reality of the data to potential consumers. When the clinician digs a bit deeper, it becomes hard to recommend testing to induce motivation in even the most stubborn of clients with heavy wallets. This area also hasn’t thoroughly investigated the flip side in the real world of direct-to-consumer genetic testing – if I don’t have a genetic variant that increases risk, am I less likely to make a certain dietary change? Even if communicating genetic risk leads individuals to make dietary changes, most studies have not assessed whether communicating a lack of genetic risk runs counter to eating a diet in line with current guidelines.

Conclusion: My overall takeaway on nutrigenetics/personalized nutrition -- come for the science, but don't stay expecting to get a super prescriptive diet sent from the DNA gods. We still don't know most of the factors that go into telling what your specific individual nutrient needs are and the available evidence doesn't support the idea that you'll become super motivated and starting making substantial diet changes. It's always up to you if you're in the financial situation where you can blow $100 on a DNA test and want to see what they tell you that you should eat; i'll always recommend that you go over that with your primary care provider and a registered dietitian. Me? I'm gonna need a lot stronger data before anybody tells me to switch my coffee to decaf or go out of my way to eat more or less of some nutrient.

Kevin Klatt is a currently finishing his doctoral work in the Molecular Nutrition program at Cornell University's Division of Nutritional Sciences, where he conducts experimental studies to understand the interactions between 1-carbon and fatty acid metabolism using cell, animal and human trial approaches. You can find him on twitter, and blogging over here.