A recent study looking at saturated fat intake in the context of a low-carbohydrate diet came out in PLoSOne the other day (1), and the headlines are covering the topic in their typical misleading fashion - see medicalnewstoday's article here titled "Rise of Saturated Fat in Diet does not Raise Fats in Blood". Even Greg Miller, PhD tweeted the study to me saying:
At face value, seeing that headline and Greg's comment, I would think...hmm okay, researchers got together, set up a tracer study using some kind of labeled fatty acids in individuals consuming typical diets and showed that lipid fractions in the blood weren't enriched in these labeled fats, in any individual, obese or not, and that there was some functional outcome. Having read the study, I can tell you that's not even close to what was done - so let's investigate further:
The study was entitled "Effects of Step-Wise Increases in Dietary Carbohydrate on Circulating Saturated Fatty Acids and Palmitoleic Acid in Adults with Metabolic Syndrome":
UPDATE: I realized that PloS One posts who the editor of the publication is. It lists the following: Editor: Pankaj K. Singh, University of Nebraska Medical Center, United States of America . This is a bit problematic and might explain the fact that this study really doesn't show anything novel - Dr. Singh is a cancer researcher, not a nutritional scientist - http://www.unmc.edu/news.cfm?match=12615
UPDATE: I realized that PloS One posts who the editor of the publication is. It lists the following: Editor: Pankaj K. Singh, University of Nebraska Medical Center, United States of America . This is a bit problematic and might explain the fact that this study really doesn't show anything novel - Dr. Singh is a cancer researcher, not a nutritional scientist - http://www.unmc.edu/news.cfm?match=12615
Study Population: 16 men and women ranging from 30-66 with BMIs ranging from 27-50 (overweight through class III obesity). The subjects had metabolic syndrome, but were excluded if they had type 2 diabetes. Fasting insulin exhibited a large range that did not exceed the reference value of less than 174 pmol.
Diets: One can see that the diets were indeed relatively similar in calories and protein (kudos to the study design on the protein part). A couple points to note here: fiber intake is pretty low in the low-carb phases, and PUFA intake is variable throughout (particularly if you look at the range and not just the average). We got no micronutrient data. Note that the baseline diet is determined by 3 day diet records, and we are not given dietary information for the run-in period (weird...)
.png)
Results: The results are nothing too surprising. The participants lose more weight in the beginning and tapers off as more carbs are placed in - could be water weight, could be compliance to kcal intake. The lipid responses aren't too abnormal either - slightly higher TGs as carbs increase (semi interesting blip there at mid-level CHO), LDL gets lowest on the higher carb, HDL increases slightly. These are difficult to interpret as we know that individuals are losing weight (a major confounding factor throughout the study) and PUFAs were variable. Insulin resistance as measured by the HOMA-IR method (not the gold standard) improves - again, this is expected with weight loss. Systolic blood pressure decreased across the diet and takes a bigger dip with the higher carbs at the end. Diastolic drops initially and increases slightly with the higher carb phase.
.png)
.png)
Conclusion: The authors conclude that high intakes of saturated fat do not contribute to an accumulation of plasma SFA in the context of a low-carb diet, and that decreases in SFA/increases in CHO increase the proportion of plasma palmitoleic acid. "These findings contradict the perspective that dietary saturated fat per se is harmful, and underscore the importance of considering the level of dietary carbohydrate that accompanies saturated fat consumption".
My Comments:
1. Weight Loss - it's rather unfortunate that this wasn't an eucaloric trial. Weight loss confounds a number of issues here: everything from lipid changes to individual fatty acids. We truly can't tell whether these changes in fatty acids are due to liberation from the adipocyte, to fats from the diet that are spared from oxidation as CHO are increased, or are de-novo synthesized. The authors suggest in the discussion that de-novo lipogenesis (DNL) is the source of 16:1 under higher carbohydrate consumption - I was shocked (but not really given the authors and their history of being low-carb advocates) that they didn't discuss this further and Hellerstein's literature (2) on the topic of DNL(3,4). Given the effect of insulin on de novo lipogenesis, and the wide range of insulin in this group, if 16:1 is due to DNL, the variation in palmitoleic acid as CHO increases is not surprising at all. Seeing as these results are being generalized to the media as universally applicable, i'm disappointed at this lack of an unbiased discussion regarding the interpretation of these results.
2. Palmitoleic Acid -Palmitoleic acid is a more controversial fatty acid, because it can be a marker of increased De Novo Lipogenesis, or it can come from the diet (where it appears to act more like a saturated fat) (5). Some animal models suggest that Palmitoleic Acid acts as a positive correlate of insulin sensitivity but studies in humans have been conflicting (6,7,8). A more recent RCT showed that palmitoleic acid supplementation improved serum lipids and decreased inflammation in an obese cohort (9).
I found palmitoleic acid (16:1) increases as a main outcome, and one that is suggested to be inherently detrimental, to be quite strange. The authors state that they focus in on this because in different cohort/cross sectional studies it is a predictor of metabolic disease (note: these studies are in individuals consuming a higher carbohydrate diet). How is it a predictor? These cohort studies are looking at 16:1 levels in cases (diabetics) versus controls (non-diabetics), or within an individual at time points before and after diabetes diagnosis. They see increases in average levels between diabetics/non, and increases within individuals as they progress from non-diabetic to diabetic. This isn't all that surprising - as insulin rises with worsening glucose tolerance, it will promote DNL/activity of desaturases, leading to more 16:1. Does that make 16:1 inherently bad? No. It just makes it a marker.
This PloS study is much different than these cohort studies, because we are comparing palmitoleic acids within the context of a macronutrient distribution change, not before and after the onset of disease. To state that these palmitoleic acid changes are a marker of metabolic decline (in HC) or improvement (in LC) would be over-extrapolating from the results. For me to believe the higher-carb diet led to worse outcomes based on palmitoleic acid, I'd have to see a worsening from their baseline levels. But we don't get those measures, only those taken after the free-living period (remember that was 3 weeks of a low-carb diet). So to say the high carb-diet worsened their metabolic status is not possible given the data presented to us.
Assuming that the 16:1 does come from DNL, I don't see how it would serve as any marker of metabolic health in low-carbohydrate diets (AKA we can't interpret the decreased levels in the low-carb phase as some sort of beneficial outcome or improvement). There is no literature suggesting in low-carbohydrate diet consuming individuals that it would be. Increases in 16:1 are a marker of metabolic dysfunction because of dysregulated DNL, which is an issue of carbohydrate metabolism. If you remove the carbohydrate,you remove the substrate for DNL, making 16:1 no longer a biomarker. Literature looking at 16:1 within adipocytes supports this idea that there is less 16:1 in adipocytes under low-carbohydrate diets conditions (10) . It would be one thing if we had data saying that palmitoleic acid in the plasma is inherently harmful- but we don't. It simply serves as a marker of metabolic decline over time in cohort studies, as we know that obese individuals undergo higher rates of DNL than lean individuals in non-overfeeding contexts (11). I'm really struggling to see why 16:1 is a positive/negative outcome for the authors of this study. 16:1 as an outcome feels like an afterthought of the data....
Also, just to note, the cited studies looking at 16:1 levels as a marker of metabolic decline also look at 16:1 to 16:0 ratios to attempt to show that 16:1 results from DNL (it's hard to measure the desaturase enzyme activity in humans). While it's not a perfect index, I'm curious as to why the PloS paper didn't include this...
3. Foods used in the Diet - I would've enjoyed a more detailed description of the diets beyond just the nutrient profiles and the snippet that notes they were rich in whole grains and Low glycemic index foods.
4. Micronutrients - if the food was prepared, why aren't we getting a micronutrient profile? Some of my issues with very low carbohydrate diets are not only their being lower on the fiber end (which is arguably relevant in the context of a LC diet), but also that they tend to require much more planning to hit recommended intakes of micronutrients. A reason I've always been okay with the 'reduce your SFA intake' (regardless of our ability to independently link them as a risk factor to CVD outcomes) is that they don't add extra essential nutrients beyond calories (and the questionable relevance of minimal associated n-3's/CLA and vitK2). Nothing about this current study says to me "eat whole milk greek yogurt instead of skim greek yogurt and some almonds" - i'll keep doing the latter.
5. Recommendations: This reciprocal relationship between saturated fat and carbohydrates, while being an interesting model, makes me question whether the authors understand the recommendations. The highest carbohydrate phase of the diet has twice as much saturated fat as it does PUFA, on average. The recommendations aren't to replace saturated fats with carbs, but replace them with PUFAs. Individuals were getting 11% of their kcals from saturated fats, while only getting 5% of their kcals from PUFAs - pretty much exactly the opposite of the recommendations. The best data we have suggests a reduction in CVD outcomes for those that replace saturated fat with mixed PUFAs, in the context of a higher carbohydrate diet (23,24). If these authors want to change the recommendations, there's gonna have to be some harder endpoints.
6. Not addressing concerns: Studies like these do not address concerns outside the context of obesity and weight loss. A concern of low-carb, high SFA compared to higher MUFA and PUFA diets is the issue of flow mediated dilation, a marker of endothelial function (12, 13). I've seen literature by Jeff Volek and others responding to this concern (14,15) but his studies failed to address the issue of weight maintenance and were not in lean adults. Under isocaloric conditions in healthy adults, FMD appears to be impaired by high SFA relative to higher MUFA/PUFA diets (15). See below for more discussion of concerns.
7. Take-Aways: It's an interesting concept for a study and I get that they were trying to show that consuming a lot of saturated fat doesn't necessarily cause a large rise in blood levels (something I kinda thought we already knew - see here). I think it's interesting to see that palmitate was reduced in cholesteryl esters from C1 to C5. Unfortunately, there's not really a big take away from this study. There were no functional health outcomes. It occurred in the context of weight loss. It was only done in obese people, which doesn't tell us much about what happens for lean individuals consuming this diet. And it was only 16 men and women with greatly varying BMIs across 18 weeks.
To be completely honest, I actually find this study really weird, because I'm not actually sure what they're trying to say or what this does for evidence-based nutrition. The conclusion seems to point out that eating a lot of saturated fat on a low carb diet is different than eating a lot of saturated fat on a high carb diet. Great. But that doesn't address people's concerns about low carb, very high saturated fat diets. Besides what I mentioned above related to micronutrients and FMD, there is also the concern that they keep LDL cholesterol elevated. As we saw here, the higher carb end of the diet led to a greater drop in LDL - it would've been great if we could've seen a more robust measure like LDL particle number- but we don't. This short term trial doesn't address those concerns - we quite literally have no functional outcome related to health. Further concerns about low-carbohydrate diets can be found here. The authors also seem to try to imply that higher-carbohydrate diets are bad for obese individuals based off of palmitoleic acid levels, which as I mentioned above, is a pretty questionable outcome and not a hard clinical one by any means. If these were their end points, I still can't fathom why they did not make it a eucaloric trial. I find myself thinking back to Shai et al (17) and still considering that higher levels of carbohydrate are great options for weight loss for the obese individuals, as long individuals can adhere to them.
I would also note that the 2013 AHA/ACC/TOS guidelines for overweight and obese individuals mention that low-carbohydrate diets are an option for weight loss (18), so the 'recommendations' have already accepted the efficacy of low-carbohydrate diets in the short term. If the low-carb community wants to show that low-carb diets are 'better', they're going to have to do much better than this current publication. And if they ever want to change recommendations, they're going to have to start doing a lot better job at looking at functional outcomes, and demonstrating efficacy in normal weight individuals. I'm not against low-carb diets, but I am all for evidence. And if you ask me what low-carb diets do in the long term, the only evidence that I can show you is low(ish)-quality data that suggests higher mortality from them (19, 20, 21, 22).
As a general aside, some people have pointed out that the funding sources are pretty questionable, particularly given the clear bias in the discussion.
Design: This 18 week study's experimental approach is depicted below. There were 6 diets, that started off low in carbohydrate(CHO) and high in saturated fat (SFAs), and slowly increased carbs, focusing on low glycemic index carbs, while reducing SFAs. Diets were designed to produce a 300kcal/d deficit and protein intake was held relatively constant. This was not a metabolic ward trial, though all food was provided. The run-in period was 3weeks long.
Diets: One can see that the diets were indeed relatively similar in calories and protein (kudos to the study design on the protein part). A couple points to note here: fiber intake is pretty low in the low-carb phases, and PUFA intake is variable throughout (particularly if you look at the range and not just the average). We got no micronutrient data. Note that the baseline diet is determined by 3 day diet records, and we are not given dietary information for the run-in period (weird...)
Results: The results are nothing too surprising. The participants lose more weight in the beginning and tapers off as more carbs are placed in - could be water weight, could be compliance to kcal intake. The lipid responses aren't too abnormal either - slightly higher TGs as carbs increase (semi interesting blip there at mid-level CHO), LDL gets lowest on the higher carb, HDL increases slightly. These are difficult to interpret as we know that individuals are losing weight (a major confounding factor throughout the study) and PUFAs were variable. Insulin resistance as measured by the HOMA-IR method (not the gold standard) improves - again, this is expected with weight loss. Systolic blood pressure decreased across the diet and takes a bigger dip with the higher carbs at the end. Diastolic drops initially and increases slightly with the higher carb phase.
The authors measure individual fatty acid changes in triglycerides, cholesteryl esters and phospholipids, focusing in on Palmitoleic acid(16:1), a monounsaturated fatty acid. They note that, although much individual variation is seen in 16:1, there is a trend that going from lower to higher carbohydrate increases the amount. I would strongly stress the variation here.
Conclusion: The authors conclude that high intakes of saturated fat do not contribute to an accumulation of plasma SFA in the context of a low-carb diet, and that decreases in SFA/increases in CHO increase the proportion of plasma palmitoleic acid. "These findings contradict the perspective that dietary saturated fat per se is harmful, and underscore the importance of considering the level of dietary carbohydrate that accompanies saturated fat consumption".
My Comments:
1. Weight Loss - it's rather unfortunate that this wasn't an eucaloric trial. Weight loss confounds a number of issues here: everything from lipid changes to individual fatty acids. We truly can't tell whether these changes in fatty acids are due to liberation from the adipocyte, to fats from the diet that are spared from oxidation as CHO are increased, or are de-novo synthesized. The authors suggest in the discussion that de-novo lipogenesis (DNL) is the source of 16:1 under higher carbohydrate consumption - I was shocked (but not really given the authors and their history of being low-carb advocates) that they didn't discuss this further and Hellerstein's literature (2) on the topic of DNL(3,4). Given the effect of insulin on de novo lipogenesis, and the wide range of insulin in this group, if 16:1 is due to DNL, the variation in palmitoleic acid as CHO increases is not surprising at all. Seeing as these results are being generalized to the media as universally applicable, i'm disappointed at this lack of an unbiased discussion regarding the interpretation of these results.
2. Palmitoleic Acid -Palmitoleic acid is a more controversial fatty acid, because it can be a marker of increased De Novo Lipogenesis, or it can come from the diet (where it appears to act more like a saturated fat) (5). Some animal models suggest that Palmitoleic Acid acts as a positive correlate of insulin sensitivity but studies in humans have been conflicting (6,7,8). A more recent RCT showed that palmitoleic acid supplementation improved serum lipids and decreased inflammation in an obese cohort (9).
I found palmitoleic acid (16:1) increases as a main outcome, and one that is suggested to be inherently detrimental, to be quite strange. The authors state that they focus in on this because in different cohort/cross sectional studies it is a predictor of metabolic disease (note: these studies are in individuals consuming a higher carbohydrate diet). How is it a predictor? These cohort studies are looking at 16:1 levels in cases (diabetics) versus controls (non-diabetics), or within an individual at time points before and after diabetes diagnosis. They see increases in average levels between diabetics/non, and increases within individuals as they progress from non-diabetic to diabetic. This isn't all that surprising - as insulin rises with worsening glucose tolerance, it will promote DNL/activity of desaturases, leading to more 16:1. Does that make 16:1 inherently bad? No. It just makes it a marker.
This PloS study is much different than these cohort studies, because we are comparing palmitoleic acids within the context of a macronutrient distribution change, not before and after the onset of disease. To state that these palmitoleic acid changes are a marker of metabolic decline (in HC) or improvement (in LC) would be over-extrapolating from the results. For me to believe the higher-carb diet led to worse outcomes based on palmitoleic acid, I'd have to see a worsening from their baseline levels. But we don't get those measures, only those taken after the free-living period (remember that was 3 weeks of a low-carb diet). So to say the high carb-diet worsened their metabolic status is not possible given the data presented to us.
Assuming that the 16:1 does come from DNL, I don't see how it would serve as any marker of metabolic health in low-carbohydrate diets (AKA we can't interpret the decreased levels in the low-carb phase as some sort of beneficial outcome or improvement). There is no literature suggesting in low-carbohydrate diet consuming individuals that it would be. Increases in 16:1 are a marker of metabolic dysfunction because of dysregulated DNL, which is an issue of carbohydrate metabolism. If you remove the carbohydrate,you remove the substrate for DNL, making 16:1 no longer a biomarker. Literature looking at 16:1 within adipocytes supports this idea that there is less 16:1 in adipocytes under low-carbohydrate diets conditions (10) . It would be one thing if we had data saying that palmitoleic acid in the plasma is inherently harmful- but we don't. It simply serves as a marker of metabolic decline over time in cohort studies, as we know that obese individuals undergo higher rates of DNL than lean individuals in non-overfeeding contexts (11). I'm really struggling to see why 16:1 is a positive/negative outcome for the authors of this study. 16:1 as an outcome feels like an afterthought of the data....
Also, just to note, the cited studies looking at 16:1 levels as a marker of metabolic decline also look at 16:1 to 16:0 ratios to attempt to show that 16:1 results from DNL (it's hard to measure the desaturase enzyme activity in humans). While it's not a perfect index, I'm curious as to why the PloS paper didn't include this...
3. Foods used in the Diet - I would've enjoyed a more detailed description of the diets beyond just the nutrient profiles and the snippet that notes they were rich in whole grains and Low glycemic index foods.
4. Micronutrients - if the food was prepared, why aren't we getting a micronutrient profile? Some of my issues with very low carbohydrate diets are not only their being lower on the fiber end (which is arguably relevant in the context of a LC diet), but also that they tend to require much more planning to hit recommended intakes of micronutrients. A reason I've always been okay with the 'reduce your SFA intake' (regardless of our ability to independently link them as a risk factor to CVD outcomes) is that they don't add extra essential nutrients beyond calories (and the questionable relevance of minimal associated n-3's/CLA and vitK2). Nothing about this current study says to me "eat whole milk greek yogurt instead of skim greek yogurt and some almonds" - i'll keep doing the latter.
5. Recommendations: This reciprocal relationship between saturated fat and carbohydrates, while being an interesting model, makes me question whether the authors understand the recommendations. The highest carbohydrate phase of the diet has twice as much saturated fat as it does PUFA, on average. The recommendations aren't to replace saturated fats with carbs, but replace them with PUFAs. Individuals were getting 11% of their kcals from saturated fats, while only getting 5% of their kcals from PUFAs - pretty much exactly the opposite of the recommendations. The best data we have suggests a reduction in CVD outcomes for those that replace saturated fat with mixed PUFAs, in the context of a higher carbohydrate diet (23,24). If these authors want to change the recommendations, there's gonna have to be some harder endpoints.
6. Not addressing concerns: Studies like these do not address concerns outside the context of obesity and weight loss. A concern of low-carb, high SFA compared to higher MUFA and PUFA diets is the issue of flow mediated dilation, a marker of endothelial function (12, 13). I've seen literature by Jeff Volek and others responding to this concern (14,15) but his studies failed to address the issue of weight maintenance and were not in lean adults. Under isocaloric conditions in healthy adults, FMD appears to be impaired by high SFA relative to higher MUFA/PUFA diets (15). See below for more discussion of concerns.
7. Take-Aways: It's an interesting concept for a study and I get that they were trying to show that consuming a lot of saturated fat doesn't necessarily cause a large rise in blood levels (something I kinda thought we already knew - see here). I think it's interesting to see that palmitate was reduced in cholesteryl esters from C1 to C5. Unfortunately, there's not really a big take away from this study. There were no functional health outcomes. It occurred in the context of weight loss. It was only done in obese people, which doesn't tell us much about what happens for lean individuals consuming this diet. And it was only 16 men and women with greatly varying BMIs across 18 weeks.
To be completely honest, I actually find this study really weird, because I'm not actually sure what they're trying to say or what this does for evidence-based nutrition. The conclusion seems to point out that eating a lot of saturated fat on a low carb diet is different than eating a lot of saturated fat on a high carb diet. Great. But that doesn't address people's concerns about low carb, very high saturated fat diets. Besides what I mentioned above related to micronutrients and FMD, there is also the concern that they keep LDL cholesterol elevated. As we saw here, the higher carb end of the diet led to a greater drop in LDL - it would've been great if we could've seen a more robust measure like LDL particle number- but we don't. This short term trial doesn't address those concerns - we quite literally have no functional outcome related to health. Further concerns about low-carbohydrate diets can be found here. The authors also seem to try to imply that higher-carbohydrate diets are bad for obese individuals based off of palmitoleic acid levels, which as I mentioned above, is a pretty questionable outcome and not a hard clinical one by any means. If these were their end points, I still can't fathom why they did not make it a eucaloric trial. I find myself thinking back to Shai et al (17) and still considering that higher levels of carbohydrate are great options for weight loss for the obese individuals, as long individuals can adhere to them.
I would also note that the 2013 AHA/ACC/TOS guidelines for overweight and obese individuals mention that low-carbohydrate diets are an option for weight loss (18), so the 'recommendations' have already accepted the efficacy of low-carbohydrate diets in the short term. If the low-carb community wants to show that low-carb diets are 'better', they're going to have to do much better than this current publication. And if they ever want to change recommendations, they're going to have to start doing a lot better job at looking at functional outcomes, and demonstrating efficacy in normal weight individuals. I'm not against low-carb diets, but I am all for evidence. And if you ask me what low-carb diets do in the long term, the only evidence that I can show you is low(ish)-quality data that suggests higher mortality from them (19, 20, 21, 22).
As a general aside, some people have pointed out that the funding sources are pretty questionable, particularly given the clear bias in the discussion.
But as I'd like to point out, for all the attention we give to funding as a competing interest, the field of nutrition needs to start noting personal biases. For anyone whose followed the low-carb vs low-fat debates, a lot of these author names will sound quite familiar. The two last authors even have their own website, http://www.artandscienceoflowcarb.com/ . I went into this study not expecting a really broad, unbiased discussion of the results, and I wasn't surprised by what I found - that expectation had nothing to do with the Dairy Institute Funding it.
Since I'm calling out personal biases, for anyone who cares, I go on and off eating a high and lower carb diets, depending on whether i'm doing aerobic exercises that day or am just sitting in my office. I don't particularly stick to one paradigm for myself, and don't 'believe' in any one macronutrient distribution over another - my personal scientific opinion is that it all comes down to genotype anyway. I generally can't stand the low-fat vs low-carb wars because it tends to just be an ego match of "my diet is better than yours", and in reality, working with an individual to find a healthy lifestyle that they can maintain for years to come is the real answer.
1. http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0113605
2. http://ajcn.nutrition.org/content/74/6/707.full
3. http://ajcn.nutrition.org/content/73/2/253.full
4. http://www.ncbi.nlm.nih.gov/pmc/articles/PMC185982/
5. http://www.jlr.org/content/35/4/656.full.pdf
2. http://ajcn.nutrition.org/content/74/6/707.full
3. http://ajcn.nutrition.org/content/73/2/253.full
4. http://www.ncbi.nlm.nih.gov/pmc/articles/PMC185982/
5. http://www.jlr.org/content/35/4/656.full.pdf
6. http://onlinelibrary.wiley.com/doi/10.1002/ejlt.201100187/pdf
7.http://www.ncbi.nlm.nih.gov/pubmed/24022867
8.http://www.ncbi.nlm.nih.gov/pubmed/20943795
9. http://www.sciencedirect.com/science/article/pii/S1933287414002815
10. http://ajcn.nutrition.org/content/93/1/186.full
11. http://ajcn.nutrition.org/content/74/6/737.long
12. http://journals.cambridge.org/action/displayAbstract?fromPage=online&aid=8976486&fileId=S000711451300216X
13. http://hyper.ahajournals.org/content/51/2/376.long
14. http://www.ncbi.nlm.nih.gov/pubmed/19632695
15. http://ajcn.nutrition.org/content/87/3/567.full.pdf
16. http://atvb.ahajournals.org/content/25/6/1274.full
17. http://www.nejm.org/doi/full/10.1056/NEJMoa0708681
18. http://content.onlinejacc.org/article.aspx?articleid=1770219
19. http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0055030
20. http://jaha.ahajournals.org/content/3/5/e001169.abstract
21. http://www.bmj.com/content/344/bmj.e4026
22. http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2989112/
23. http://www.ncbi.nlm.nih.gov/pubmed/21118617
24. http://www.ncbi.nlm.nih.gov/pubmed/21679479
Really excellent critique.
ReplyDelete