I've heard the phrase from other medical professionals that nutrition is a 'softer science'. Softer sciences are considered to be ones where it's difficult to establish strictly, objectively measurable criteria, and the term is often used in reference to psychology, sociology, anthropology, and other social sciences.
I've heard it said plenty but I've never really given it much thought, often chalking these statements up to individuals taking a cheap shot at the field or not fully understanding all of the lines of evidence - to a degree, I still feel this way. I thought a lot more deeply about the soft-ness of nutrition, however, when I saw the most recent reports making headlines - Dr Katz over at Yale published a public health review that concluded there is no best diet. Most of the summaries of this research making the social media rounds have been to 'eat real food' AKA don't eat processed food. A part of me rolls my eyes at this, especially as we seek to understand the complex interactions between nutrition and genotype, to promote optimal health of the individual, the notion that one can eat whatever they want as long as it is unprocessed is rooted a bit too far away from science. I also feel that sentiment undermines the decades of research determining optimal micronutrient consumption. But, as I've discussed before, other parts of me get the 'whole foods' approach - though I think it should be called, "well-planned, optimally processed, plant-centric whole foods" (I should've been a marketer!!). Overall, I think this report speaks more to the overwhelming limitations of nutritional sciences research AKA it's soft-science-ness, rather than supporting an 'eat any whole foods, in any macronutrient/micronutrient composition ratios' mantra.
What makes nutrition softer, or difficult to definitively measure?
1. Adherence - I would say that this is the major issue in nutritional sciences research, as well as a number of other fields- see here, here, and here. When you want to study the effects of a low-carb diet vs low-fat, you have your participants come in, you give them lectures on how to eat. Or maybe you've rounded up a group of self-ascribed low-carb/fat eaters. You then track their weight changes, blood lipids, etc over a certain course of time, and report your findings. This all sounds great in an ideal world - if only people stuck to eating the diet they were assigned/said. Take this study in the annals of internal medicine - it reports increases in all-cause mortality from low-carb diets. But if you look at the data, most of the low-carb group was eating 40% carbs, which is lower than the low end of the AMDR but much higher than any dogmatic low-carb diet would suggest. What can you tell about truly low-carb/Atkins style diets from this? Very little. So what evidence is there to suggest individuals on Atkins style low-carb diets are putting themselves at risk? Very little, mostly from inference. Be very vigilant with titles that state low fat/carb/Mediterranean/Paleo/etc, because the participants often fail to stick to these.
2. Time scale - But there is some evidence from Atkins-style low carb diets. Many fear high intakes of saturated fat in these diets will lead to aberrant lipid levels, but the data we have suggests the exact opposite. Does that mean Atkins-style low carb diets are safe? Length of the study becomes an issue. One could easily argue that CVD doesn't develop in the 6-24months that these individuals on both ketogenic and non-ketogenic diet were followed for. There aren't studies following these individuals for a decade or two. Could you even find a large population size willing to participate in a study that has been adhering to a low-carb diet for 2 decades? Most of the chronic diseases that affect industrialized citizens today take years and years to develop, if not decades. But we don't always follow people for decades - and when we do, it's rarely longitudinal or even cross sectional - with retrospective followups of self-reported intakes, there's always noise, and enough that anyone can find some flaw in a study to dismiss contrary data and support their dietary dogma. Until we start locking people up in metabolic wards for decades, we'll probably never definitely know (assuming we maintain the same varied food system). It simply comes down to - we don't know, do you want to take that risk?
3. Multifactorial - Let's assume I did find a population, and the ketosis diet, on average, had heightened CVD mortality risk. What was that from? Is it the ketones? Is it the high saturated/unsaturated fat intake? Is it the low-fiber intake? Is it from a low potassium:sodium ratio? Were they eating mostly animal sources of foods or plant? Were they deficient in some nutrient? Did they supplement? Let's say that population had no heart issues. Can we recommend ketogenic diets? We would also need to look at their effect on bones, the brain, skeletal muscle, etc, before we could truly recommend the efficacy and safety of them. Where are those well-controlled studies? I've said it many times before, and I'll say it again - nutrients don't work in isolation. When you study foods, you aren't just studying one chemical, as you may with a 'harder' science.
4. Recalls/FFQ's- How would I know you ate a low-carb diet? Likely, it'll come from a dietary recalls and food frequency questionnaire. But how accurate are these? There is an entire body of literature, from several fields, that shows these are often grossly inaccurate, leading to underreporting of energy intake and omission of specific foods, making it quite difficult to actually discern macro/micronutrient intakes in different populations (1-7).
5. Correlation and Causation - Vitamin D is the best example here. Google any disease and Vitamin D and you'll likely find search results - diabetes (8), CVD (9), breast cancer (10), autism (11). Is vitamin D deficiency causing all of these? It's truly difficult to discern, especially when many of these diseases are multi-factorial. One day vitamin D is causatively linked to autism via serotonin synthesis, and the next Harvard puts out a report about the 8 environmental chemicals disturb neurological development leading to autism - which is true?. Many people have a lot of diseases, and most people have nutritionally imperfect diets - I could draw you a lot of correlations, but which ones are causative? To answer this, you need very well designed/controlled studies and a great statistician - and you need the person reading the study and telling you about it to be well versed on these topics.
6. Animals, and cells, and people, oh my! - What do you do when cell line, animal, RCTs and observational data don't match up? Let's use phytoestrogens as an example here. Animal models and in vitro studies show very mixed effects as to the estrogenic effect exerted by them (12). Human RCT data gives mixed results as well, and, overall, doesn't seem to have much of an effect. However, population data shows that Japan, which has historically had significantly (magnitudes of 100) higher phytoestrogen intake than the US, suffer from significantly reduced incidence of breast cancer (13) - however, this is confounded by a multitude of lifestyle/genetic factors that can hardly be teased out. What do you tell patients about soy-foods and phytoestrogen supplements? In most biomedical sciences, RCTs are the gold standard -but I would argue this isn't true for nutritional sciences. For a good critique of RCTs in nutrition, check out CarbSanity's 2 part post. I often actually find observational studies to be some of the most informative about nutrition, especially in their ability to speak to the multifactorial interactions that go into disease processes. Medical literature today will report ApoE4 as being a major genetic risk factor for Alzheimer's and CHD, but observational data from Sub-Saharan Africans, with the highest prevalence of this genotype but an absence of this association, suggests there are many factors at play here (15). For a while, nutrition demonized all saturated fats, even though observational studies of Pacific Islanders, relying largely on whole coconuts for calories, showed little to no vascular disease (16). Quantitative dietary studies of hunter gatherers, with little to no vascular disease, often show high animal food consumption, which is often associated with increased CVD risk in Western population (17). Not to say that we should give evidence-based recommendations off of observational studies, but they certainly speak to the interactions between nutrition and lifestyle factors, and should be considered before making over-zealous statements about the effects of single nutrients. Studying nutrition requires the use of a number of different models, none of which are inherently flawless, and each provide unique pieces to the puzzle.
7. Funding - It's difficult to quantify exactly how much money Nutrition funding gets, but it is nowhere near other biomedical fields. One can look at NIH's estimated funding for specific diseases (14), and clearly see that most are not nutrition-related, and even the ones that are, are not solely looked at by nutritional scientists. Inadequate funding reduces the feasibility of doing long-term trials, or providing all of the food to be eaten over a 6month period, or having patients meet with RD's several times a week, or collecting all of the relevant biomarkers, etc etc. Funding can also be important because studies may be biased due to their funding sources - this isn't for all studies, but there have been some that have creeped into the literature. My favorite was this beef/egg at breakfast study funded by the beef/egg industry - telling us more of what we already knew..
8. Reductionist or Realist - There's often a debate one must undergo when wanting to know the effect of just one nutrient. Let's say I want to focus on the effects of omega 3 fatty acids in the diet. Assuming you're not wanting to supplement, you would either:
When I hear nutrition being called a soft science, I don't really take that as an insult, but more so interpret it to mean that it's just really f'ing hard to definitively study. It took me a while to really come to the realization that nutrition, while generally a biomedical science, isn't as hard a science as other biomedical sciences (not trying to play the whose science is harder game, they're all difficult). But I think being a soft science is what makes nutrition pretty interesting (even if it is often frustrating). It requires a broad knowledge base, founded in biochem, physiology, biostats, research methodology, and lab techniques.
While I probably just made the field sound fraught with issues, I think they should be seen as more of a caveat: no one study is perfect or complete, and not everyone can necessarily interpret the body of literature adequately. Making evidence-based nutrition recommendations requires a synthesis of a lot of different approaches and types of data, and there's room for flexibility. If anything, being a softer science stresses the need for having experts get together to interpret the available data (the IoM's DRI's/AMDRs), having trained practitioners (RDs/LDNs) help individuals construct a well-planned diet, and seeking out nutrition advice from credible sources, without biases. This isn't to say not following the recommendations will seriously harm the average person - but it's your choice if you want to be your own case-study.
1. http://www.ncbi.nlm.nih.gov/pubmed/10719403
2. http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2803049/
3. http://ajcn.nutrition.org/content/37/1/139.abstract
4. http://link.springer.com/article/10.1023%2FA%3A1019637632584
5. http://www.ncbi.nlm.nih.gov/pubmed/12612176
6. http://www.ncbi.nlm.nih.gov/pubmed/1454084
7. http://www.ncbi.nlm.nih.gov/pubmed/15054345
8.http://www.nih.gov/news/health/oct2013/niddk-21.htm
9.http://eurheartj.oxfordjournals.org/content/early/2013/06/08/eurheartj.eht166.short?rss=1
10.https://www.vitamindcouncil.org/health-conditions/breast-cancer/
11. http://www.sciencedaily.com/releases/2014/02/140226110836.htm
12. http://www.ncbi.nlm.nih.gov/pubmed/11694655
13. http://ajcn.nutrition.org/content/79/2/183.full
14. http://report.nih.gov/categorical_spending.aspx
15. http://onlinelibrary.wiley.com/store/10.1046/j.1469-1809.1999.6340301.x/asset/j.1469-1809.1999.6340301.x.pdf?v=1&t=htg2hgp0&s=4d69d75c9e727379a05787359f7e9edb18e69d5b
16.http://www.ncbi.nlm.nih.gov/pubmed/7270479/
17.http://www.nature.com/ejcn/journal/v56/n1s/abs/1601353a.html
I've heard it said plenty but I've never really given it much thought, often chalking these statements up to individuals taking a cheap shot at the field or not fully understanding all of the lines of evidence - to a degree, I still feel this way. I thought a lot more deeply about the soft-ness of nutrition, however, when I saw the most recent reports making headlines - Dr Katz over at Yale published a public health review that concluded there is no best diet. Most of the summaries of this research making the social media rounds have been to 'eat real food' AKA don't eat processed food. A part of me rolls my eyes at this, especially as we seek to understand the complex interactions between nutrition and genotype, to promote optimal health of the individual, the notion that one can eat whatever they want as long as it is unprocessed is rooted a bit too far away from science. I also feel that sentiment undermines the decades of research determining optimal micronutrient consumption. But, as I've discussed before, other parts of me get the 'whole foods' approach - though I think it should be called, "well-planned, optimally processed, plant-centric whole foods" (I should've been a marketer!!). Overall, I think this report speaks more to the overwhelming limitations of nutritional sciences research AKA it's soft-science-ness, rather than supporting an 'eat any whole foods, in any macronutrient/micronutrient composition ratios' mantra.
What makes nutrition softer, or difficult to definitively measure?
1. Adherence - I would say that this is the major issue in nutritional sciences research, as well as a number of other fields- see here, here, and here. When you want to study the effects of a low-carb diet vs low-fat, you have your participants come in, you give them lectures on how to eat. Or maybe you've rounded up a group of self-ascribed low-carb/fat eaters. You then track their weight changes, blood lipids, etc over a certain course of time, and report your findings. This all sounds great in an ideal world - if only people stuck to eating the diet they were assigned/said. Take this study in the annals of internal medicine - it reports increases in all-cause mortality from low-carb diets. But if you look at the data, most of the low-carb group was eating 40% carbs, which is lower than the low end of the AMDR but much higher than any dogmatic low-carb diet would suggest. What can you tell about truly low-carb/Atkins style diets from this? Very little. So what evidence is there to suggest individuals on Atkins style low-carb diets are putting themselves at risk? Very little, mostly from inference. Be very vigilant with titles that state low fat/carb/Mediterranean/Paleo/etc, because the participants often fail to stick to these.
2. Time scale - But there is some evidence from Atkins-style low carb diets. Many fear high intakes of saturated fat in these diets will lead to aberrant lipid levels, but the data we have suggests the exact opposite. Does that mean Atkins-style low carb diets are safe? Length of the study becomes an issue. One could easily argue that CVD doesn't develop in the 6-24months that these individuals on both ketogenic and non-ketogenic diet were followed for. There aren't studies following these individuals for a decade or two. Could you even find a large population size willing to participate in a study that has been adhering to a low-carb diet for 2 decades? Most of the chronic diseases that affect industrialized citizens today take years and years to develop, if not decades. But we don't always follow people for decades - and when we do, it's rarely longitudinal or even cross sectional - with retrospective followups of self-reported intakes, there's always noise, and enough that anyone can find some flaw in a study to dismiss contrary data and support their dietary dogma. Until we start locking people up in metabolic wards for decades, we'll probably never definitely know (assuming we maintain the same varied food system). It simply comes down to - we don't know, do you want to take that risk?
3. Multifactorial - Let's assume I did find a population, and the ketosis diet, on average, had heightened CVD mortality risk. What was that from? Is it the ketones? Is it the high saturated/unsaturated fat intake? Is it the low-fiber intake? Is it from a low potassium:sodium ratio? Were they eating mostly animal sources of foods or plant? Were they deficient in some nutrient? Did they supplement? Let's say that population had no heart issues. Can we recommend ketogenic diets? We would also need to look at their effect on bones, the brain, skeletal muscle, etc, before we could truly recommend the efficacy and safety of them. Where are those well-controlled studies? I've said it many times before, and I'll say it again - nutrients don't work in isolation. When you study foods, you aren't just studying one chemical, as you may with a 'harder' science.
4. Recalls/FFQ's- How would I know you ate a low-carb diet? Likely, it'll come from a dietary recalls and food frequency questionnaire. But how accurate are these? There is an entire body of literature, from several fields, that shows these are often grossly inaccurate, leading to underreporting of energy intake and omission of specific foods, making it quite difficult to actually discern macro/micronutrient intakes in different populations (1-7).
5. Correlation and Causation - Vitamin D is the best example here. Google any disease and Vitamin D and you'll likely find search results - diabetes (8), CVD (9), breast cancer (10), autism (11). Is vitamin D deficiency causing all of these? It's truly difficult to discern, especially when many of these diseases are multi-factorial. One day vitamin D is causatively linked to autism via serotonin synthesis, and the next Harvard puts out a report about the 8 environmental chemicals disturb neurological development leading to autism - which is true?. Many people have a lot of diseases, and most people have nutritionally imperfect diets - I could draw you a lot of correlations, but which ones are causative? To answer this, you need very well designed/controlled studies and a great statistician - and you need the person reading the study and telling you about it to be well versed on these topics.
6. Animals, and cells, and people, oh my! - What do you do when cell line, animal, RCTs and observational data don't match up? Let's use phytoestrogens as an example here. Animal models and in vitro studies show very mixed effects as to the estrogenic effect exerted by them (12). Human RCT data gives mixed results as well, and, overall, doesn't seem to have much of an effect. However, population data shows that Japan, which has historically had significantly (magnitudes of 100) higher phytoestrogen intake than the US, suffer from significantly reduced incidence of breast cancer (13) - however, this is confounded by a multitude of lifestyle/genetic factors that can hardly be teased out. What do you tell patients about soy-foods and phytoestrogen supplements? In most biomedical sciences, RCTs are the gold standard -but I would argue this isn't true for nutritional sciences. For a good critique of RCTs in nutrition, check out CarbSanity's 2 part post. I often actually find observational studies to be some of the most informative about nutrition, especially in their ability to speak to the multifactorial interactions that go into disease processes. Medical literature today will report ApoE4 as being a major genetic risk factor for Alzheimer's and CHD, but observational data from Sub-Saharan Africans, with the highest prevalence of this genotype but an absence of this association, suggests there are many factors at play here (15). For a while, nutrition demonized all saturated fats, even though observational studies of Pacific Islanders, relying largely on whole coconuts for calories, showed little to no vascular disease (16). Quantitative dietary studies of hunter gatherers, with little to no vascular disease, often show high animal food consumption, which is often associated with increased CVD risk in Western population (17). Not to say that we should give evidence-based recommendations off of observational studies, but they certainly speak to the interactions between nutrition and lifestyle factors, and should be considered before making over-zealous statements about the effects of single nutrients. Studying nutrition requires the use of a number of different models, none of which are inherently flawless, and each provide unique pieces to the puzzle.
7. Funding - It's difficult to quantify exactly how much money Nutrition funding gets, but it is nowhere near other biomedical fields. One can look at NIH's estimated funding for specific diseases (14), and clearly see that most are not nutrition-related, and even the ones that are, are not solely looked at by nutritional scientists. Inadequate funding reduces the feasibility of doing long-term trials, or providing all of the food to be eaten over a 6month period, or having patients meet with RD's several times a week, or collecting all of the relevant biomarkers, etc etc. Funding can also be important because studies may be biased due to their funding sources - this isn't for all studies, but there have been some that have creeped into the literature. My favorite was this beef/egg at breakfast study funded by the beef/egg industry - telling us more of what we already knew..
8. Reductionist or Realist - There's often a debate one must undergo when wanting to know the effect of just one nutrient. Let's say I want to focus on the effects of omega 3 fatty acids in the diet. Assuming you're not wanting to supplement, you would either:
- control all of the factors of all of the individuals diets and just vary omega 3's. This allows you, in your conclusion, to state: the effect seen is due to omega 3's. e.g. provide all of the food to participants and give different fish consumption recommendations.
- manipulate the levels of omega 3's while allowing the rest of the diet to vary, as it might in the real world. e.g. give different fish consumption recommendations and let the rest of the diet vary
While I probably just made the field sound fraught with issues, I think they should be seen as more of a caveat: no one study is perfect or complete, and not everyone can necessarily interpret the body of literature adequately. Making evidence-based nutrition recommendations requires a synthesis of a lot of different approaches and types of data, and there's room for flexibility. If anything, being a softer science stresses the need for having experts get together to interpret the available data (the IoM's DRI's/AMDRs), having trained practitioners (RDs/LDNs) help individuals construct a well-planned diet, and seeking out nutrition advice from credible sources, without biases. This isn't to say not following the recommendations will seriously harm the average person - but it's your choice if you want to be your own case-study.
1. http://www.ncbi.nlm.nih.gov/pubmed/10719403
2. http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2803049/
3. http://ajcn.nutrition.org/content/37/1/139.abstract
4. http://link.springer.com/article/10.1023%2FA%3A1019637632584
5. http://www.ncbi.nlm.nih.gov/pubmed/12612176
6. http://www.ncbi.nlm.nih.gov/pubmed/1454084
7. http://www.ncbi.nlm.nih.gov/pubmed/15054345
8.http://www.nih.gov/news/health/oct2013/niddk-21.htm
9.http://eurheartj.oxfordjournals.org/content/early/2013/06/08/eurheartj.eht166.short?rss=1
10.https://www.vitamindcouncil.org/health-conditions/breast-cancer/
11. http://www.sciencedaily.com/releases/2014/02/140226110836.htm
12. http://www.ncbi.nlm.nih.gov/pubmed/11694655
13. http://ajcn.nutrition.org/content/79/2/183.full
14. http://report.nih.gov/categorical_spending.aspx
15. http://onlinelibrary.wiley.com/store/10.1046/j.1469-1809.1999.6340301.x/asset/j.1469-1809.1999.6340301.x.pdf?v=1&t=htg2hgp0&s=4d69d75c9e727379a05787359f7e9edb18e69d5b
16.http://www.ncbi.nlm.nih.gov/pubmed/7270479/
17.http://www.nature.com/ejcn/journal/v56/n1s/abs/1601353a.html
Comments
Post a Comment